Accelerated human liver progenitor generation from pluripotent stem cells by inhibiting formation of unwanted lineages

Despite decisive progress in differentiating pluripotent stem cells (PSCs) into diverse cell-types, the often-lengthy differentiation and functional immaturity of such cell-types remain pertinent issues. Here we address the first challenge of prolonged differentiation in the generation of hepatocyte-like cells from PSCs. We delineate a roadmap describing the extracellular signals controlling six sequential branching lineage choices leading from pluripotency to endoderm, foregut, and finally, liver progenitors. By blocking formation of unwanted cell-types at each lineage juncture and manipulating temporally-dynamic signals, we accelerated generation of 89.0±3.1% AFP+ human liver bud progenitors and 87.3±9.4% ALBUMIN+ hepatocyte-like cells by days 6 and 18 of PSC differentiation, respectively. 81.5±3.2% of hepatocyte-like cells expressed metabolic enzyme FAH (as assayed by a new knock-in reporter line) and improved short-term survival in the Fah-/-Rag2-/-Il2rg-/- mouse model of liver failure. Collectively the timed signaling interventions indicated by this developmental roadmap enable accelerated production of human liver progenitors from PSCs.


Introduction
A quintessential goal of regenerative medicine has been to manufacture various human celltypes in a dish from pluripotent stem cells (PSCs). Despite decisive progress, PSC differentiation into many lineages often takes weeks or months of in vitro differentiation; yields 20 cell-types with limited functional maturity; and may produce heterogeneous populations containing other unwanted cell-types (Cohen and Melton, 2011). With regard to the first challenge (prolonged differentiation), one notion was that PSC differentiation might be an innately lengthy process. By contrast, PSC differentiation towards neuronal and mesodermal lineages can be considerably accelerated by precisely manipulating developmental signals in a 25 temporally-dynamic fashion (Chambers et al., 2012;Qi et al., 2017). Here we focus on likewise accelerating the production of liver cells (hepatocytes) from PSCs by first understanding the temporally-dynamic signals controlling liver differentiation. There has been major headway in generating enriched populations of hepatocyte-like cells from hPSCs (Agarwal et al., 2008;Basma et al., 2009;Cai et al., 2007;Carpentier et al., 30 2014;Espejel et al., 2010;Han, 2012;Ogawa et al., 2013;Rashid et al., 2010;Si Tayeb et al., 2010;Song et al., 2009;Touboul et al., 2010;Zhao et al., 2012). The considerable interest in generating hepatocytes has stemmed from the fact that they are vital for bodily metabolism and neutralize harmful waste products within the body, amongst other tasks (Stanger, 2015).
Attesting to the liver's importance, acute liver failure rapidly results in coma or death as toxins 35 accumulate in the body (Karl et al., 1953). However the sequence of lineage intermediates through which early hepatocytes develop from pluripotent cells remains to be fully defined, as are the extracellular signals that specify liver fate at each lineage juncture. Importantly, it might be possible to exclusively generate liver cells by deciphering the signals that induce non-liver cells at each branching lineage decision and then inhibiting the signals that otherwise specify 40 non-liver fates.
Here we decompose early liver development into a sequence of six consecutive lineage 80 choices and detail the signals at each juncture that specify each cell-type (either liver or nonliver precursors). We show that multiple developmental signals (e.g., retinoid and other signals) can have opposing effects within 24 hours, initially specifying one fate and then subsequently repressing its formation. This map of liver development allowed us to providently manipulate signals in a temporally-dynamic fashion-and to repress signals at each juncture that specified 85 unwanted cell-types-enabling accelerated production of liver bud progenitors from hPSCs within 6 days. Importantly, the hPSC-derived liver bud progenitors produced in this accelerated fashion are still capable of differentiating into hepatocyte-like cells, the latter of which can engraft the Fah -/-Rag2 -/-Ilr2g -/mouse model of liver injury. This suggests that accelerated differentiation by precisely manipulating developmental signals does not impede the 90 functionality of downstream cell-types. Finally, to quantitatively track this accelerated liver differentiation process, we furnish new tools; namely cell-surface markers identifying liver progenitors and a knock-in reporter hESC line to track the expression of metabolic enzyme FAH during hepatocyte differentiation. 95 RESULTS RA BMP and FGF activation and TGFβ inhibition rapidly differentiate endoderm into posterior foregut with the competence to later generate liver bud progenitors Pluripotent cells first differentiate into primitive streak and subsequently definitive endoderm before turning into liver (Fig. 1a). We previously identified signals to differentiate hPSCs into 100 >99% pure MIXL1-GFP + primitive streak cells in 24 hours (Fig. 1b) and subsequently into >98% pure SOX17-mCherry + definitive endoderm by day 2 of differentiation ( Fig. 1c) . These nearly-pure endoderm populations provided an ideal starting point to examine signals that could further drive day 2 endoderm into day 3 posterior foregut, and later, day 6 liver bud progenitors (Fig. 1b,c). 105 We found that transient activation of the retinoic acid (RA), BMP and FGF pathways together with inhibition of TGFβ signaling for 24 hours was necessary to differentiate day 2 definitive endoderm into day 3 posterior foregut (Fig. 1d).
RA pathway agonists initially promoted foregut specification on day 3 of differentiation. Treatment of day 2 endoderm with the RA agonist all-trans RA (ATRA, 2 μM) or TTNPB (75 nM) 110 for 24 hours enhanced the formation of day 3 posterior foregut that was later competent to differentiate into liver bud (Fig. 1e,f,g) and downstream hepatocytes (Fig. S1a,b) on later days of differentiation. Thus RA is crucial for human posterior foregut specification, consistent with how inhibiting RA synthesis abrogates both liver and pancreas formation in zebrafish embryos (Stafford and Prince, 2002). 115 Foregut specification by day 3 was also promoted by TGFβ inhibition (Fig. 1h) and activation of the BMP and FGF pathways (Fig. S1d,e,h,i). Each of these manipulations enhanced the generation of posterior foregut that was capable of subsequently differentiating into liver bud progenitors and/or hepatocytes ( Fig. 1h; Fig. S1c). Taken together, day 2 definitive endoderm could be converted into day 3 posterior foregut by the simultaneous 120 activation of RA, BMP and FGF pathways together with inhibition of TGFβ signaling for 24 hours; such day 3 posterior foregut populations expressed HHEX (Fig. S1j), a marker of ~E8.5 mouse ventral posterior foregut (Thomas et al., 1998).

TGFβ, BMP, PKA activation and WNT inhibition accelerate differentiation of foregut into 125
liver bud progenitors by day 6 of differentiation Though day 3 posterior foregut was initially specified by RA activation and TGFβ inhibition, the subsequent differentiation of posterior foregut into day 6 liver bud progenitors was suppressed by continued RA activation and TGFβ inhibition. Instead, liver bud specification on days 4-6 required activation of the TGFβ, BMP and PKA pathways, together with WNT inhibition, for 72 130 hours (Fig. 2a).
Hence our findings reconcile conflicting findings that RA is overall required for zebrafish liver induction (Stafford and Prince, 2002) but that RA-coupled beads inhibit liver bud marker 135 formation in HH10 stage zebrafish embryos (Bayha et al., 2009)-there is a temporally dynamic requirement for RA in liver bud specification. Akin to RA signaling, the role for TGFβ in liver specification was also temporally dynamic: TGFβ inhibition initially promoted foregut formation by day 3 (Fig. 1e,h) but 24 hours later, TGFβ activation promoted liver bud specification on days 4-6, leading to enhanced liver 140 bud marker expression by day 6 (Fig. 2b,d). Emphasizing the importance of TGFβ activation, we found that TGFβ inhibition at this stage instead abrogated the differentiation of foregut into liver bud (Fig. 2b,d), contrasting with earlier use of TGFβ inhibitors to differentiate hPSCderived endoderm into liver Sampaziotis et al., 2015;Touboul et al., 2010). BMP and PKA activation, together with WNT blockade, also cooperated with TGFβ 145 activation to differentiate day 3 foregut into day 6 liver bud progenitors. Consistent with earlier findings (Chung et al., 2008;Rossi et al., 2001;Shin et al., 2007;Si Tayeb et al., 2010;Wandzioch and Zaret, 2009;Zhao et al., 2012), first we found that BMP activation differentiated foregut into liver bud while blocking pancreas formation; by contrast BMP inhibition suppressed liver formation (Fig. S2a,g). This mirrors how bmp2b overexpression promotes liver 150 specification at the expense of pancreatic progenitors in zebrafish embryos (Chung et al., 2008) and how BMP induces liver from mouse embryonic endoderm explants (Rossi et al., 2001). Second, we found PKA agonists (e.g., 8-bromo-cAMP) also potently specified liver bud from foregut ( Fig. S2b), supporting the notion that prostaglandin E2 specifies zebrafish liver progenitors by activating the PKA cascade (Nissim et al., 2014). Finally, WNT inhibition (using 155 C59) from days 4 to 6 suppressed mid/hindgut (MHG; posterior endoderm) formation ( Fig. S1fg) and instead promoted the formation of liver bud progenitors that had the ability to subsequently form hepatocytes (Fig. S2f) and cholangiocytes (Fig. S4g). This is consistent with how WNT signaling blocks foregut formation and instead specifies MHG in Xenopus (McLin et al., 2007). Taken together, simultaneous activation of TGFβ, BMP and PKA but suppression of 160 WNT drove day 3 posterior foregut into day 6 liver bud progenitors while simultaneously blocking pancreatic and MHG differentiation. Combining the above signals enabled the generation of a rather-uniform liver bud progenitor population by day 6 of hPSC differentiation, which was accelerated by comparison to extant methods. Quantification of differentiation efficiencies by intracellular flow cytometry 165 demonstrated the consistent generation of an 89.0±3.1% pure AFP + liver bud progenitor population from the H1, H7 and H9 hPSC lines (Fig. 2e). This new combined approach ("SR2") for liver bud specification was more rapid than extant liver induction methods (Si Tayeb et al., 2010;Zhao et al., 2012), yielding significantly higher expression of liver bud markers and purer populations of liver bud progenitors (94.1±7.35% HNF4A + ) by day 6 of hPSC differentiation ( Fig.  170 2e,f,g; Fig. S2i), whereas HNF4A + liver bud progenitors were instead efficiently formed on later days (by day 14) in other differentiation protocols (Si Tayeb et al., 2010;Zhao et al., 2012). Finally day 6 hPSC-derived liver bud progenitors expressed high levels of liver bud transcription factors including AFP, HNF4A, TBX3, HNF6 and CEBPA, which were low or undetectable in MHG (Fig. S2c-e). Reciprocally hPSC-derived liver bud progenitors did not 175 express CDX or HOX genes ( Fig. S2c-e), which are markers of the MHG, a developmentally-related lineage that emanates from adjacent posterior endoderm and is also specified by BMP and FGF signals (Sherwood et al., 2011;Spence et al., 2010). Taken together, though liver bud and MHG (intestinal) progenitors are spatially adjacent in vivo, they are transcriptionally distinct lineages and can be produced in mutually-exclusive 180 signaling conditions in vitro from hPSCs. Importantly, while there is a common requirement for BMP and FGF in liver and MHG specification, we reveal signals that uniquely specify liver (Fig.  2a), thus clarifying how these lineages become segregated from one another. This knowledge enabled the efficient generation of human liver bud progenitors by day 6 of PSC differentiation, which is >2 times faster than extant methods. 185 A surface marker signature for hPSC-derived liver bud progenitors To quantitatively track this accelerated time course of human liver bud progenitor specification at the single-cell level, we next sought to identify cell-surface markers that would demarcate day 6 liver bud progenitors from developmentally-earlier fates (day 0 hPSCs and day 2 definitive 190 endoderm). Systematically screening the expression of 242 cell-surface antigens on these 3 lineages revealed stage-specific surface markers ( Fig. 3a,b). First, CD10 was expressed in 92.6±5.6% of undifferentiated hPSCs but was abruptly downregulated upon differentiation, being expressed in <2% of cells in day 2 definitive endoderm or day 6 liver bud populations ( Fig.  3c; Fig. S3b; Table S1). Second, CD184/CXCR4 (a known definitive endoderm marker 195 (D'amour et al., 2005)) was enriched in day 2 definitive endoderm by comparison to day 0 hPSCs and day 6 liver bud progenitors (Fig. 3c,d; Table S1). Finally, CD99 was highly expressed on day 6 liver bud progenitors by comparison to preceding hPSCs or DE (Fig. 3c,d; Table S1). This pattern was consistent across a panel of 4 hPSC lines (H7, HES2, H1 and BJC1; Table S1; Fig. S3a,b). Thus, positive selection for CD99 hi cells may enrich for liver bud 200 progenitors, especially when combined with negative selection to eliminate CD10 + hPSCs and CD184 + DE. These cell-surface markers can be used to track early liver progenitor specification and to distinguish them from developmentally-earlier cell-types.

Segregation of human liver bud progenitors into hepatocyte vs. biliary fates by 205
competing NOTCH, TGFβ, PKA and other signals Next we tested whether these hPSC-derived day 6 liver bud progenitors-obtained via accelerated differentiation-were still competent to subsequently give rise to ALBUMIN + hepatocyte-like cells or SOX9 + biliary cells (cholangiocytes) on later days of differentiation. Suppression of both NOTCH and TGFβ signaling on days 7 to 8 was necessary to block biliary 210 formation and consolidate hepatocyte fate. First, we found that NOTCH blockade (by DAPT) prevented hPSC-derived liver bud progenitors from adopting a biliary fate (as assessed by reduced SOX9 expression) and instead diverted cells into ALBUMIN + hepatocytes ( Fig.  S4a,b,e). Second, brief TGFβ blockade after the liver bud stage also enhanced expression of various liver genes including ALBUMIN (Fig. S4c,d). Reciprocally, TGFβ activation (using 215 ACTIVIN) reduced ALBUMIN expression but induced SOX9 expression in a dose-dependent fashion (Fig. S4f). Collectively, NOTCH and TGFβ signaling drive biliary differentiation and hence must be suppressed to promote hepatocyte fate from liver bud progenitors. Conversely, consistent with earlier studies (Ogawa et al., 2015;Sampaziotis et al., 2015), activation of TGFβ and NOTCH together with insulin differentiated day 6 liver bud progenitors 220 into CK7 + /CK19 + biliary progenitors by day 13 of hPSC differentiation (Fig. S4g). This suggests the bipotent capacity of liver bud progenitors at a population level to give rise to either hepatic or biliary fates (which remains to be formally demonstrated in vivo by single-cell lineage tracing) and identifies the signals that control the mutually-exclusive allocation of hepatocyte-like cells vs. cholangiocytes. 225 Beyond expression of pan-hepatocyte marker ALBUMIN, we sought to assess what other key liver enzymes and markers were expressed by hPSC-derived hepatocyte-like cells and we screened for signals that could augment the acquisition of certain physiologic characteristics. In particular we sought to promote expression of tyrosine metabolism pathway components in hPSC-derived hepatocytes, as genetic deficiency of tyrosine metabolic enzymes 230 (e.g., fumarylacetoacetate hydrolase [FAH]) results in hereditary tyrosinemia in human patients, a metabolic disorder (St-Louis and Tanguay, 1997). High insulin levels together with a stabilized ascorbic acid derivative (ascorbic acid-2-phosphate [AAP]) greatly promoted the expression of tyrosine metabolic pathway genes PAH, HGD, HPD, TAT, MAI and FAH and other liver markers when applied on days 7-12 of hPSC differentiation (Fig. 4b, c). PKA agonists (e.g., 8-bromo-235 cAMP and forskolin) also had a similar effect (Fig. S4h). Microarray profiling showed treatment with NOTCH inhibitors, AAP, forskolin or insulin upregulated genes associated with gene ontology classifications pertaining to metabolic processes ( Fig. 4a; Fig. S4i; Table S5), asserting that these individual manipulations each promote the metabolic competence of hPSCderived hepatocytes. 240 Coordinated application of these hepatocyte-specifying signals-together with inhibition of cholangiocyte-specifying signals TGFβ and NOTCH-enabled the generation of an enriched population of ALB + hepatocyte-like cells by day 18 of hPSC differentiation ( Fig. 4d; Fig. S4j). Intracellular flow cytometric analysis revealed the consistent generation of, on average, an 87.3±9.4% pure ALBUMIN + hepatocyte population from the H1, H7 and H9 hPSC lines (Fig. 4e). 245 This new combined approach yielded significantly higher expression of hepatocyte markers such as ALBUMIN and CPS1 within a span of 18 days when compared with other methods (Si Tayeb et al., 2010;Zhao et al., 2012) at the same timepoint, as revealed by widefield fluorescent imaging of entire wells of differentiated cells (Fig. 4d).
We further tested if the coordinated application of these signals enabled the generation 250 of hepatocyte like-cells that secreted ALBUMIN and possessed P450 cytochrome activities, which other differentiation methods have successfully attained (Basma et al., 2009;Carpentier et al., 2014;Si Tayeb et al., 2010;Zhao et al., 2012). First, day 18 hPSC-derived hepatocytes stained positive for periodic acid Schiff (Fig. S4k). Second, they secreted 5.89±0.71 µg/mL of ALBUMIN protein (Fig. 4f) and 1.38±0.26 µg/mL of FIBRINOGEN protein into the medium (Fig.  255 4g)-levels in the same order of magnitude as the amount secreted in vivo by hepatocytes in human beings (Bernardi et al., 2012;Tennent et al., 2007). Third, they also possessed CYP3A4 enzymatic activity at levels higher than that in HepG2 cells, which was attenuated by the CYP3A4 inhibitor ketoconazole (Fig. 4h). Given that these three criteria have been previously applied to characterize hPSC-derived hepatocyte-like cells (Espejel et al., 2010;Ogawa et al., 260 2013; Rashid et al., 2010;Si Tayeb et al., 2010;Takebe et al., 2013;Zhao et al., 2012), we sought to further focus on expression of tyrosine metabolic pathway enzymes (Fig. 4b, To track the expression of FAH at a single-cell level in hPSC-derived hepatocytes, we 265 used CRISPR/Cas9-mediated gene editing to generate a FAH-2A-Clover knock-in H1 hESC reporter line in which the FAH coding sequence was left intact (Fig. 5a,b). Optimal guide RNAs that efficiently directed Cas9 cleavage of the FAH locus were first evaluated in HEK293T cells ( Fig. S5a-c) and then used to direct homologous recombination in hESCs to knock-in a Clover fluorescent reporter gene downstream of FAH. Allele-specific integration of the reporter cassette 270 was confirmed by PCR ( Fig. S5d-f) and sequencing (Fig. S5g). Applying the above differentiation protocol to the resultant FAH-2A-Clover hESCs yielded 81.5±3.2% FAH-Clover + population in the day 18 hepatocyte-like population (Fig. 5c,d). Attesting to the faithfulness of the reporter, FACS-sorted Clover + day 18 cells were significantly enriched for FAH, ALBUMIN and HGD mRNAs by comparison to Clovercells (Fig. 5e,f). The expression of FAH in the 275 majority of hPSC-derived hepatocyte-like cells motivated us to test whether these populations could be used to treat the Fah -/-Rag2 -/-Ilr2g -/mouse model of liver injury.

hPSC-derived hepatocytes engraft the injured mouse liver and improve survival during liver injury 280
Finally we sought to determine whether bulk populations of hPSC-derived day 18 hepatocytelike cells generated using the above signaling strategy could engraft the injured mouse liver and improve short-term survival. Specifically, we tested whether FAH + hPSC-derived hepatocytes derived using our differentiation schema could engraft a genetic mouse model of liver injurynamely Fah -/-Rag2 -/-Il2rg -/-(FRG) mice (Azuma et al., 2007;Overturf et al., 1996). FRG mice are 285 an immunodeficient mouse model of Tyrosinemia Type I, an inherited liver failure syndrome caused by FAH mutations in patients (Labelle et al., 1993;St-Louis and Tanguay, 1997). Like human patients, FRG mice suffer severe and eventually fatal liver injury in the absence of a hepatoprotective drug, NTBC (Azuma et al., 2007;Overturf et al., 1996). To this end, day 18 hPSC-derived hepatocytes were doubly labeled with constitutively-290 expressed genetic reporters (EF1A-BCL2-2A-GFP and UBC-tdTomato-2A-Luciferase)  and then were intrahepatically injected into neonatal FRG mice (less than 2 days old). Injected mice were allowed to grow until 6 weeks of age before liver injury was induced by cyclical withdrawal of hepatoprotective drug NTBC (Azuma et al., 2007;Overturf et al., 1996;Zhu et al., 2014) (Fig. 6a). Strikingly, hPSC-derived hepatocytes engrafted in 7 out of 15 mice 295 ( Fig. 6b) and continued to expand in vivo as shown by an increase in bioluminescence intensity over time (Fig. 6b, S6a). After ~10 weeks of liver injury, all surviving mice had detectable bioluminescence (indicating the presence of transplanted cells); mice that initially had little/no bioluminescence had died by this timepoint (Fig. 6b). This suggests that engrafted hPSCderived hepatocytes promoted the survival of the FRG mice. Bioluminescent imaging of 300 dissected livers (Fig. S6b,c) and sectioning revealed the presence of human ALBUMIN + tdTomato + liver cells near the vasculature within the livers of mice that survived 5 months posttransplantation (Fig. S6b,c). We further tested the ability of our hPSC-derived hepatocytes to ameliorate liver failure in a second model by intrasplenically transplanting them into 4-to 6-week-old adult FRG mice 305 (Fig. S6d) and inducing chronic liver injury by intermittent NTBC withdrawal (Fig. 6a). One month post-transplantation, human ALBUMIN + liver cells were detected in the mouse liver, some of which were localized near the vasculature ( Fig. 6d; Fig. S6e). On average 120±36 ng/mL of human ALBUMIN was detected in the mouse bloodstream (N = 7 mice) indicating modest ALBUMIN secretion by the transplanted hPSC-derived hepatocytes (Fig. 6g). Bilirubin 310 levels (reflecting the extent of liver injury) were reduced in the serum of these mice (Fig. 6f). 72.7% of the FRG mice transplanted with hPSC-derived hepatocytes survived longer compared to the negative control mice that were not injected with any cells (Fig. 6e). Together, these results indicated that transplanted hPSC-derived hepatocytes could engraft the injured adult mice liver, secrete human serum ALBUMIN and ameliorate liver injury. This demonstrates that 315 hPSC-derived liver bud progenitors obtained via accelerated differentiation are still competent to generate hepatocyte-like cells capable of some degree of in vivo engraftment and that they can improve short-term survival in the FRG mouse model of hereditary tyrosinemia.

320
Contrary to the view that hPSC differentiation into various cell-types is innately a protracted process, multiple groups including our own have shown that precisely timed manipulation of extracellular signals can accelerate differentiation (Chambers et al., 2012;Qi et al., 2017). Indeed during development, differentiating progenitors interpret signals in a temporally-dynamic way, with drastically different responses in even a 24-hour interval 325 (Wandzioch and Zaret, 2009). Thus manipulating these signals in vitro with equal dynamism is necessary to achieve efficient differentiation. Here we show that starting from highly-pure definitive endoderm populations, it is possible to generate enriched liver bud progenitors by day 6 of hPSC differentiation while suppressing differentiation into alternate endodermal fates (pancreas and MHG). These liver bud progenitors are produced >2 times faster than possible 330 by extant methods. Though derived in accelerated fashion, these hPSC-derived liver bud progenitors are still competent to differentiate into hepatocyte-like cells and cholangiocytes at later days. Through the use of FAH-2A-Clover reporter hESCs, we demonstrate that the resultant hepatocyte-like cells express tyrosine metabolic enzyme FAH and that accordingly, they can engraft and improve short-term survival in the Fah -/-Rag2 -/-Il2rg -/mouse model of 335 hereditary tyrosinemia. It is hoped that this accelerated system to generate hPSC-derived liver bud progenitors will avail the final goal of generating mature hepatocytes, which has yet to be realized. Our work highlighted two principles to more specifically generate liver at the expense of other endodermal lineages. First, we found that signals that specified liver fate were temporally 340 dynamic, initially promoting but then inhibiting liver fate. Indeed over a brief 24-hour interval, RA activation and TGFβ blockade were necessary to drive definitive endoderm into posterior foregut progenitors, but the next day, these same signals blocked the further progression of posterior foregut towards a liver bud fate. Hence understanding the temporal dynamics with which these signals act is key, as these signals must be manipulated with equal dynamism to 345 guide efficient differentiation as cells pass through transient windows of inductive competence. Second, at each lineage segregation, we determined the signals that promoted the formation of both our desired fate as well as the unwanted lineage(s). With this knowledge, these lineage segregations could be efficiently negotiated by providing the relevant inductive signal(s) to drive differentiation towards the desired lineage while repressing the signal(s) that 350 otherwise promoted the alternate fate . For example, NOTCH and TGFβ signaling differentiated liver bud progenitors into cholangiocytes while suppressing hepatocyte formation. Conversely, blockade of NOTCH and TGFβ pathways induced differentiation into hepatocytes, in part by excluding formation of SOX9 + cholangiocytes. Taken in collective our results suggest a signaling code for accelerated liver 355 specification ( Fig. 2a), extending beyond the use of BMP and FGF to induce liver progenitors from endoderm (Gouon-Evans et al., 2006;Shiraki et al., 2008;Si Tayeb et al., 2010;Spence et al., 2010;Touboul et al., 2010;Zhao et al., 2012). Indeed, liver and MHG both share a common requirement for BMP and FGF, raising the question of how these two lineages are distinguished from one another. We show that transient exposure of endoderm to RA activation and TGFβ 360 blockade specifies the competence of posterior foregut, such that upon subsequent exposure to BMP and FGF, it specifically differentiates into liver (instead of MHG). Moreover, while we confirm that BMP and FGF are necessary to differentiate posterior foregut into liver bud, this is only fully realized in the context of other signals (TGFβ and PKA activation together with WNT inhibition). We also introduce new tools to track this accelerated differentiation process: (i) the 365 surface marker CD99 which, together with absence of CD10 and CD184/CXCR4, identifies day 6 liver bud progenitors and (ii) an FAH-Clover reporter hESC line to track the subsequent generation of FAH + hepatocyte-like cells.
Finally, we show that hPSC-derived hepatocyte-like cells, though derived in accelerated fashion, are competent to engraft in vivo and improve the survival of Fah -/-Rag2 -/-Il2rg -/mice. 370 Substantial progress has been made in generating human hepatocytes from direct lineage reprogramming that can engraft sublethal models of mouse liver injury (Huang et al., 2014;Sekiya and Suzuki, 2011;Song et al., 2016;Yu et al., 2013;Zhu et al., 2014). However, generating functionally mature human hepatocytes from hPSCs remains a significant challenge. It is hoped that human liver differentiation roadmap described here and this accelerated system 375 to generate liver bud progenitors will provide insight into human liver development and avail the as-of-yet-unrealized goal of generating mature hepatocytes.

Design and construction of Cas9 plasmids
Primers used for constructing the plasmids are listed in  (Cong et al. 2013), was digested with XbaI & AarI (Thermo, ER1582) and ligated with EF1α promoter. Next, the modified plasmid was digested with SspI and XbaI and then ligated with the Amp pUC fragment. The BmsBI chimaeric gRNA cassette was amplified from a gBlock based on px335 (Addgene plasmid # 42335), a gift from Feng Zhang (Cong et al., 2013). The cassette was subsequently ligated into XbaI cut site of our modified Cas9 plasmid. To add a 2A-linked mRuby2, the plasmid was first digested with PmlI and EcoRI, and the Cas9 3' fragment was amplified from px330 and 2AmRuby2 (Lam et al., 2012) fragment was amplified from a gBlock. Next, these two PCR fragments were ligated with the digested plasmid. Enhanced specificity of Cas9 was attained by specific mutations of the Cas9-protein sequence (Kleinstiver et al., 2016;Slaymaker et al., 2015). Such mutations (K848A/K1003A/R1060A) were introduced to our Cas9 plasmid by cloning in a gBlock. 5' fragment of Cas9 and 3' fragment plus 2AmRuby2 were amplified from the plasmid, and mutated sites from gBlock. The plasmid was digested with FseI and EcoRI and fragments ligated in the digest. hPSCs are very sensitive to DNA damage (Momcilovic et al., 2009;Momcilovic et al., 2010) and we found that Cas9 targeted hPSCs had low survival and low number of correctly targeted clones. Inhibiting the TP53 checkpoint could increase survival of targeted hPSCs (Ihry et al., 2017). mtp53 dominant negative fragment was amplified from gBlock based on pCE-mp53DD (Addgene plasmid # 41856), a gift from Shinya Yamanaka (Okita et al. 2013) and ligated it into our Cas9 plasmid digested with EcoRI to generate our final construct pMIA3 1sg-eSpCas9-2AmRuby2-2Amp53DD".

Design of gRNA
The genomic sequence of the end of human FAH CDS (chr15:80186000-80187000) was uploaded to Benchling (https://benchling.com/) and single gRNAs were designed using the online search algorithm. One gRNA (GAGCAGAGAAAATCTCATGA, negative strand) that overlaps the FAH-stop codon was selected. Oligos with the gRNA sequence and the complementary sequence (TCATGAGATTTTCTCTGCTC), with CACC-and AAAC-added to the 5' end of each oligo, were purchased from IDT. Finally, the oligos were annealed and ligated into BsmBI-digested pMIA3 1sg-eSpCas9-2AmRuby2-2Amp53DD.

GFP reconstitution assay
The gRNA cutting efficiency was confirmed by GFP reconstitution assay using pCAG-EGxxFP plasmid (Addgene plasmid # 50716), a gift from Masahito Ikawa (Mashiko et al., 2013), as described previously. Briefly, the target sequence was amplified from H1 hPSCs genomic DNA and then cloned into the SalI cut site on pCAG-eGxxFP (for primers see Table S2). The plasmid was then transfected into HEK293T cells with or without FAH-pMIA3. 48h later, strong GFP signal was observed when both plasmids were transfected indicating Cas9 cleavage activity (Fig. S5a,b,c).
To swap the resistance gene from puromycin to blasticidin, the mPgk1 promoter & polyA fragments were first amplified from OCT4-2A-eGFP-PGK-Puro and Blasticidin (Bsd) resistance gene from pCMV-Bsd plasmid. These fragments were ligated into BsrGI & AscI double-digested donor plasmid to generate a construct containing OCT4-2A-eGFP-PGK-Bsd-FAH. The modified plasmid was then digested with NheI & PacI to allow ligation of P2A-linked Clover fluorophore (Lam et al., 2012), that had been amplified from a gBlock. The PGK promoter was amplified from OCT4-2A-eGFP-PGK-Puro, to make OCT4-2A-Clover-PGK-Bsd-FAH. A Gly-Ser-Gly sequence was added to the start of the P2A sequence as this has been shown to increase the 2A-peptide cleavage efficiency (Kim et al., 2011;Szymczak et al., 2004). The Bsd CDS was amplified from pCMV-Bsd and the negative selection agent thymidine kinase (TK) CDS from pLOX-TERT-iresTK (Addgene plasmid # 12245), a gift from Didier Trono (Salmon et al. 2000).
The two PRC fragments were then ligated into PacI & AflIII digested donor plasmid, to generate the construct OCT4-2AClover-PGK-Bsd2ATK-FAH. NotI was then used to digest this plasmid and a PGK-DTA negative selection cassette was ligated outside the homology arms, amplified from OCT4-2A-eGFP-PGK-Puro and a gBlock. Finally the 5' homology arm of human FAH was amplified from H1 hPSC gDNA and then ligated into the SbfI-& NheI-digested plasmid to complete the FAH-2AClover-PGK-Bsd2ATK-FAH-PGK-DTA donor plasmid.
Generation of FAH-2AClover H1 hES reporter cell line 1.5x10 6 H1 hPSCs were used in one reaction of an Amaxa nucleofection (Lonza) targeting. 5 g of both FAH-pMIA3 and FAH-2AClover-PGK-Bsd2ATK-FAH-PGK-DTA donor plasmid were nucleofected with the P3 Primary Cell kit (V4XP-3024) using CM-113 program. Prior to nucleofection the cells were pretreated with ROCK-inhibitor (Y-27632, Tocris) for at least 1 hour and then dissociated into a single-cell suspension with StemPro Accutase (A1110501, ThermoFisher). After nucleofection, the cells were seeded in mTeSR supplemented by CloneR (05888, Stem Cell Technologies). After 24h, the media was changed to mTeSR and cells were left to recover. At approximately 48-72h, Bsd (A1113903, ThermoFisher) was added to the media at 10μg/ml to select for targeted cells. Individual colonies were manually picked from wells after 10-14 days of selection and further expanded for screening.
12 clones were picked from the targeted wells. 11 clones survived the expansion and gDNA was isolated for screening. Primers for amplifying the FAH target region in the eGxxFP plasmid (also annotated "eGxxFP primers" were used for screening the targeted clones alongside FAHclover forward and reverse primers (Table S2). Expected PCR product size for wild-type cells is 422 bp with eGxxFP primers and for targeted cells 2.3 kbp with "FAH-clo" primers. 10 out of 11 clones picked amplified PCR bands at both WT and targeted sizes, suggesting they were heterozygous clones (Fig. S5d,e). 3 out of the 10 clones had no mutations on the WT allele as confirmed by Sanger sequencing (Fig. S5g).
Heterozygous clones that were confirmed were further expanded and targeted with 10μg of pCAG-Cre:GFP (Matsuda & Cepko 2007) (Addgene plasmid # 13776), a gift from Connie Cepko. These cells were allowed to recover for 48h after which they were sorted for GFP. GFP + cells were plated in CloneR supplemented mTeSR at a density of 100 to 1000 cells per well in a 6well tissue culture plate. After 48h, the media was changed to mTeSR and clones were allowed to expand for 10-14 days. Clones were then manually picked for expansion and screening. PCR screening and Sanger sequencing was done with eGxxFP primers as above. Expected PCR amplicon size for correctly targeted allele is 1.3 kbp (Fig. S5f). Next, confirmed clones were expanded and used in downstream experiments.

Intracellular FACS
AFP antibodies (DakoCytomation, A000829) were conjugated with R-phycoerythin (rPE) using rPE labeling kit (abcam, ab102918). Cells (either undifferentiated hPSCs or hPSC-derived liver cells) were dissociated into single cells using TrypLE Express (Gibco, 1260413) and centrifuged at 2000 rpm for 3 minutes. Each cell pellet was washed with 1x PBS (Gibco, 14190235), strained using a 100µm strainer (Falcon,) and counted using a hemocytometer before fixation in BD Cytofix/Cytoperm buffer (BD Bioscience, 554714) on ice for 20 minutes. Next, fixed cells were washed 2 times with 2 mL 1x BD Perm/Wash buffer (pre-warmed to room temperature) (BD Bioscience, 554723) at room temperature and pelleted. The cell pellet was then resuspended in 1x BD Perm/Wash buffer and 100μLof cell suspension was aliquoted into individual tubes for separate stains. Stained or unstained controls were included. Either anti-ALB-APC (R&D, IC1455A, 0.4μL per 150,000 cells) or anti-AFP-PE (0.33μL per 150,000 cells) was used. Anti-AFP-PE antibody/cell mixture was incubated at room temperature for 30 minutes in the dark while anti-ALB-APC antibody/cell mixture was incubated at room temperature for 20 minutes in the dark. Subsequently, the unstained or stained cells were washed 2 times with 2 mL 1x BD perm/wash buffer at room temperature and pelleted. The pellet was then resuspended in 300μL 1x BD perm/wash. FACS run was performed using BD LSR Fortessa X20 and FACS data were analyzed using FlowJo.

Sorting of Clover + and Clovercells by FACS
Cells were dissociated as single cells and stained with DAPI prior to FACS. Separate samples of FAH-Clover + and Cloverpopulations were gated and collected and then harvested for gene expression analyses.

High-throughput antibody screens
BD Lyoplate Screening panel (BD Biosciences, 560747) antibodies (Table S1) were reconstituted with deionized water before their use to stain cells following manufacturer's instructions. Cell suspension was filtered through 100 µm strainer to remove clumps. 75 µL cell suspension was added to each antibody-containing well, pipette-mixed 3 times and incubated in the dark at 4°C for 20-30mins. The cells were pelleted at 500g for 6 minutes and supernatant was discarded by plate inversion. The cells were then washed twice with 200 µL cell staining buffer by pipette mixing, resuspended in DAPI-containing cell staining buffer and analysed with BD LSR Fortessa X20.

Immunostaining and image analyses
Cells were washed once gently with 1x PBS, fixed with 4% formaldehyde in PBS for 15-20 minutes at room temperature, washed 3 times with 1x PBS and blocked with blocking buffer (10% Donkey Serum + 0.1% Triton X in 1x PBS) for 1 hour at room temperature. Next, the cells were stained with antibodies diluted in 1% Donkey Serum + 0.1% Triton X in 1x PBS at 4°C overnight (see Table S3 for all antibodies used). The next day, the cells were washed 3 times with washing buffer (0.1% Triton X in 1x PBS) and then stained with fluorophore-conjugated secondary antibodies diluted in at 1:1000 (1% Donkey Serum + 0.1% Triton X in 1x PBS) in the dark at room temperature for 1 hour. Finally, the cells were washed once with 100 ng/mL DAPI/PBS once and twice with 1x PBS before visualization using Zeiss Axio Vision. "No primary antibody staining" or undifferentiated hPSC negative controls were used to adjust exposure times for minimal background fluorescence detection. The stained cells were counted using Image J (Abràmoff and Magalhães, 2004;Collins, 2007). Image preprocessing including gray scale conversion, threshold setting, image segmentation and noise reduction were performed.

Western blotting (WB)
Cell samples were lysed using radioimmunoprecipitation assay (RIPA) buffer supplemented with protease inhibitor cocktail (Nacalai Tesque, 25955-11). Protein concentration was determined using Pierce BCA Protein Assay (ThermoScientific, 23225). 40 µg of protein lysates were separated by SDS-PAGE and transferred onto a PVDF membrane (100V at 4°C, for approximately 1 hour). Membranes were blocked with 5% milk in TBS for 1 hour at room temperature. Membranes were briefly washed with Tris-buffered saline (TBS) and after this incubated with the primary antibody (see Table S3 for all antibodies used) at 4°C overnight. The following day, membranes were washed 3 times with TBS and incubated with the appropriate HRP-conjugated IgG secondary antibody for 2 hours at room temperature. All antibodies, primary and secondary, were used with 5% milk in TBS. Following incubation with secondary antibody, the membranes were washed 3 times in TBS. Washed membranes were developed using SuperSignal West Pico ECL (Thermo Scientific, 34077) and imaged with BioRad ChemiDoc MP Imager.

Immunohistochemistry (paraffin)
Immunohistochemistry was performed on paraffinized murine tissue that was embedded onto glass microscope slides and sectioned using a microtome. Sections were deparaffinized through a series of 5-minute washes with xylene, 100% ethanol, 95% ethanol, 70% ethanol and lastly milli-Q water. Endogenous peroxidase activity in sections were blocked using blocking buffer (65 mL of 100% methanol, 3.5 mL of 30% hydrogen peroxidase (Sigma) and 31.5 mL of milli-Q water) for 30-minute at room temperature. Finally, antigen retrieval was performed at boiling temperature for 30 minutes in a 10% pH 6-citrate buffer.
The tissues were subsequently blocked with donkey serum for 1 hour at room temperature in a hydration chamber. Sections were stained with primary antibodies (Vector Laboratories, Vectastain ABC kit; PK-6101) at 4°C overnight, later washed 3 times with 0.1% Triton in 1x PBS done in 5-minute incubation intervals (with fresh buffer applied during each wash) and stained with secondary biotinylated antibody at room temperature for 30 minutes. Thereafter, the samples were lightly rinsed with 1x PBS for 5 minutes on an orbital shaker. 250 μl of ABC reagent was added to each sample and left to incubate at room temperature for 1 hour. The slides were then washed 3 times with 1x PBS for 5 minutes.
For peroxidase detection, DAB kit (Vector Laboratory, SK-4100) was used for substrate binding. The substrate concoction (5 mL of milli-Q water, 2 drops of buffer stock, 4 drops of DAB and 2 drops of hydrogen peroxide) was applied onto the samples and left to incubate for 5 minutes at room temperature. After the 5 minutes incubation, the samples were rinsed 3 times with milli-Q water, changing fresh water each time. The samples were then placed into a vertical plastic chamber containing hematoxylin and incubated for 20 minutes.
After the hematoxylin counterstaining step, the samples were rinsed 3 times with milli-Q water before proceeding to the dehydration steps. Dehydration steps comprises: 70% ethanol, 95% ethanol, 100% ethanol and xylene wash steps, all with 5 minutes incubation. The counterstained samples were permanently mounted using Vectamount reagent (H-5000) from Vector Laboratory.

Immunohistochemistry (cryoblock)
Mouse liver tissue was fixed in 4% formaldehyde diluted in 1x PBS at 4°C overnight and then immersed in 30% sucrose solution for cryoprotection. The liver tissue was then embedded in Optimal Cutting Temperature (OCT), sectioned and deposited onto microscope slides (A*STAR IMCB histology lab). The microscope slides were thawed in room temperature and stained following the immunostaining method described above.

Gene expression profiling by RT-qPCR
RNA was extracted using Zymo RNA extraction kit. Reverse transcription was performed with random primers (Applied Biosystems RT kit) to generate cDNA. Gene expression was quantified using gene-specific primers and Quantitect SYBR Green master mix (Qiagen) or KAPA SYBR® Fast qPCR kit (KAPA Biosystems, KK4602). NCBI primer designing tool was used to design gene-specific primer sequences. Primers used were validated for their specificity and those with efficiency between 90-110% were used (Table S4). Gene expression profile was analysed using Microsoft excel and heatmaps were generated using GenePattern version 2.0 (Reich et al., 2006).

Microarray
Total RNA was extracted from hPSCs and hepatic derivatives using the RNeasy Micro kit (Qiagen) and profiled for RNA integrity using the 2100 Bioanalyzer (Agilent). 500ng of high purity and high integrity samples with 260/280 and 260/230 absorbance ratios >1.9 and RNA integrity numbers (RIN) >8.0 were reverse transcribed into cDNA and in vitro transcribed into biotin-labelled cRNA using the TargetAmp-Nano labeling kit for Illumina Expression BeadChip (Epicentre). The cRNA was hybridised on HumanHT-12 v4 Expression BeadChips and scanned on the HiScan system (Illumina) according to the manufacturer's specifications. The raw microarray data was background subtracted using the BeadStudio Data Analysis Software v3.1.3.0 (Illumina) and normalised using the cross-correlation method (Chua et al., 2006). Differential gene expression was defined based on a fold-change cutoff of >1.5 compared to the average of the undifferentiated hESC baseline controls ( Table S5). Heatmaps of the foldchange in gene expression on a Log 2 scale were generated using Cluster and TreeView. Gene ontology analyses on the genes downregulated upon removal of DAPT, Forskolin, Ascorbic acid and Insulin were conducted using DAVID/EASE (Huang et al., 2008;Huang et al., 2009). Background subtraction using Illumina HT-12 v3 database was chosen in the settings.

Animal husbandry and blood sampling
The IACUC and IRB committee had approved all procedures performed in the study. FRG mice on NOD or C57BL6 background (Yecuris, 10-0008 or 10-0001) were handled and housed under aseptic conditions. Each mouse had its ears notched for long-term tracking. The mice were fed ad libitum with irradiated LabDiet 5LJ5 chow formulated with high fat and low protein content to avoid excessive tyrosine levels, which can lead to liver damage. NTBC (Yecuris, 20-0026) was dissolved in sodium carbonate to generate 1 mg/mL stock solution. A final dose of 16 mg/L was given to breeders or non-experimental mice, whereas 0 to 8 mg/L NTBC was provided to experimental mice on selected days. All experimental mice were treated with the same NTBC cycling condition. 3% Dextrose was added to the drinking water to offset the bitter taste of NTBC. Antibiotics such as Sulfamethoxazole (SMX) and Trimethoprin (TMP) (Yecuris, 20-0037) were added to the water once every other week for 2-4 days to prevent infection in the immunecompromised mice. To aid in caloric intake, each cage was supplemented with a dish of liquid diet, prepared by adding 1 volume of STAT high caloric liquid (PRN Pharmaceutical, G8270) to 1 volume of 3% Dextrose (Sigma) drinking water. Masses of FRG mice were measured weekly. Blood was collected using lancets for submandibular bleeding and allowed to coagulate in 4°C overnight. The next day, serum was harvested by centrifugation and removal of red blood cells.

Intrahepatic injections into neonate and adult livers
For neonatal intrahepatic transplantations, 0-48h old FRG pups were injected with approximately 200,000 cells directly into the liver using 31G needles. The pups were then rubbed with bedding and returned to the cages. Pertaining to intrahepatic transplantation into adult livers, 50 µL of cells were injected into multiple sites within the liver using 26G-31G needles. Bleeding was stopped by gently applying pressure on the puncture site upon withdrawal of needle.

Bioluminescence imaging
Mice were pre-transplanted with hPSC-derived hepatocyte-like cells, which overexpress luciferase gene. Prior to bioluminescence imaging, the transplanted mice were anesthetized with 1 to 3% isoflurane-mixed oxygen. To minimize imaging signal interference, mice were depilated at abdominal region to fully expose skin surface as hair will absorb and scatter light which may result in lower signal output. Mice were then weighed to determine their weight. Dluciferin solution was reconstituted by adding 30 mL of saline/1 x PBS to 1g of D-luciferin stock (Promega, InvivoGlo) and then filtered. 167 μg/g firefly D-luciferin was injected intraperitoneally into each mouse. After 10min of incubation time, the mice were positioned in the Perkin Elmer IVIS Spectrum facing upright (0% relative to supine position) in which the whole liver region was exposed to the camera's field of view. The parameters of bioluminescence imaging were as follow: Exposure -40s; binning -medium; excitation -block; emission -open; structure -no; FOV -D; Fstop -1; and height -1.5.

Enzyme linked immunosorbent assay (ELISA)
ELISA Accessory kit (Bethyl Laboratories, Inc, E101) was used to quantify human serum albumin levels. Assay was performed as per manufacturer' instructions. Absorbance was measured using an ELISA plate reader at 450 nm.

Bilirubin quantification
Bilirubin levels in mouse serum were measured using a colorimetric Bilirubin assay kit (Sigma, MAK126). Total, direct or blank working reagents were prepared as per manufacturer's instructions. 20 µL serum was added per well and total, direct or blank working reagents were added to each sample. Colorimetric product was measured at 530 nm using the Sunrise TM microplate reader. Bilirubin concentration was then calculated using the following formula [(A530sample -A530blank) x 5 mg/dL]/ [A530 calibrator -A530 water].

Survival curve analyses
Mice were checked for survival daily and Kaplan-Meier survival analysis was conducted using GraphPad Prism v7.00 for Mac (GraphPad Software, La Jolla, CA, USA). Statistical analyses were performed using the one-sided Mantel-Cox log-rank test. Data from 3 independent transplantation experiments were analyzed.

Statistics
No statistical method was used to pre-determine sample size for in vitro or in vivo experiments. Experiments were not randomized. The investigators were not blinded to the allocation during experiments or outcome assessment.

Fig. S1
Figure S1 | Regulation of early foregut competence. a) Experimental strategy to treat DE with RA or TGFβ modulators on the day 2-3 interval and evaluating its subsequent impact on day 18 hepatocyte gene expression as shown in subpanels b, c. b) qPCR gene expression of day 18 hepatocyte markers after inhibition (BMS) or activation of retinoid signaling (using ATRA, 2 µM or various doses of TTNPB, 10-100 nM) in the presence of base condition A83BF (A83BF: A8301, 1 µM; BMP4, 30 ng/mL; FGF2, 10 ng/mL) on the day 2-3 interval. c) qPCR gene expression of day 18 hepatocyte markers after inhibition (A8301, 1 µM or SB505124, 1 µM) or activation (ACTIVIN, 10 ng/mL) of TGFβ signaling in the presence of base condition BF (BF: BMP4, 30 ng/mL; FGF2 10, ng/mL) on the day 2-3 interval. d) Experimental strategy to treat definitive endoderm (DE) with FGF2 at 10 ng/mL) on the day 2-3 interval to produce day 3 posterior foregut (PFG) and assaying subsequent effects on hepatic gene expression by day 8, as shown in subpanel e. e) ALBUMIN qPCR of day 8 hepatic progenitors cells generated from endoderm treated on the day 2-3 interval with FGF2. f) Experimental strategy to treat posterior foregut (PFG) or liver bud progenitors (LB) with a WNT inhibitor (C59, 1 µM) or R-SPONDIN3 (R100, 100 ng/mL) and WNT3A at varying doses of 50 or 100 ng/mL on the day 3-4 or day 4-5 interval to produce day 4 or day 5 liver bud progenitors, as shown in subpanel g. g) CDX2 qPCR of day 5 liver bud progenitors cells generated from PFG treated on the day 3-4 interval with WNT modulators. h) Experimental strategy to treat DE with BMP4 or DM (DM3189) on the day 2-3 interval and evaluating its impact on day 6 liver bud differentiation, as shown in subpanel i. i) qPCR gene expression of day 5 liver bud progenitors after inhibition (DM) or activation of BMP signaling on day 2-3 interval. j) qPCR gene expression of hPSC, day 3 hPSC-derived PFG and day 5 hPSC-derived liver bud (LB) and midgut/hindgut (MHG) progenitors.    Figure S2 | Lineage bifurcation between liver bud and pancreatic fates. a) qPCR gene expression of day 5 liver bud cells generated from endoderm treated on the day 4-5 interval with a BMP inhibitor (DM: DM3189, 250 nM) or varying doses of BMP4 (B, 10-100 ng/mL) in the presence of base condition A10 (A10: ACTIVIN at 10 ng/mL). MHG denotes hPSC-derived midgut/hindgut cells. b) Gene expression of day 6 liver bud cells after 2-day treatment of PKA inhibitor (rRpCAMP) or PKA agonist (BrCAMP) in the presence of base condition A83B10 (A83B10: A8301, 1 µM; BMP4, 10 ng/mL) during a day 4-5 interval and day 6 hPSC-derived MHG, as shown by qPCR. c) Gene expression of day 6 hPSC-derived liver bud and MHG cells derived as shown by qPCR. d) Cartoon of known organ domain markers expressed in ~E8.5 mouse endoderm. e) Gene expression of day 5 liver bud progenitors, MHG or pancreatic endoderm cells (Panc End) derived from hPSCs as shown by qPCR. f) qPCR gene expression of day 18 hepatocytes generated from PFG cells treated on the day 4-6 interval with WNT inhibitors (C59, 1 µM or XAV939, 1 µM) or CHIR99201 (CHIR) of varying doses (1 µM, 2 µM or 3 µM). g) qPCR gene expression of day 6 cells after treatment of PFG cells on the day 4-6 interval a BMP inhibitor (DM: DM3189, 250 nM) or varying doses of BMP4 (3-25 ng/mL) in the presence of pancreatic inducing base condition (ACPRS = ACTIVIN + C59 + PD0325901 + ATRA + SANT1). A10B10: ACTIVIN, 10 ng/mL; BMP4, 10 ng/mL. h) qPCR gene expression of day 6 cells after treatment of PFG cells with a FGF/ERK inhibitor (PD032: PD0325901, 500 nM) or varying doses of FGF2 (10-20 ng/mL) in the presence of pancreatic inducing base condition (ACPRS = ACTIVIN + C59 + PD0325901 + ATRA + SANT1) on the day 4-6 interval. i) qPCR gene expression of H9 hPSC-derived liver bud progenitors generated from SR2 described in the present study and methods previously described in the literature (Si Tayeb et al., 2010;Zhao et al., 2012).  16 hepatocytes generated from liver bud progenitors treated on the day 7-12 interval with a NOTCH inhibitor (DAPT, 10 µM). Each qPCR heatmap is representative of 4 independent experiments with technical duplicates. c) Experimental strategy to treat day 6 liver bud progenitors (LB) with a TGFβ inhibitor (SB505124, 1 µM) on the day 7-8, 9-10, 11-12 or 7-12 intervals and assaying downstream effects on hepatic gene expression by day 16, as shown in subpanel d. d) qPCR gene expression of day 16 hepatocytes generated from liver bud progenitors treated on the day 7-12, 12-16 or 7-16 intervals with a TGFβ inhibitor (SB505124, 1µM). Each qPCR heatmap is representative of 3 independent experiments with duplicates. e) qPCR gene expression of day 6 hPSC-derived liver bud (LB) progenitors before and after 2 day treatment of 10 µM DAPT in the absence or presence of 10 µM DEX. f) qPCR gene expression of day 10 hPSC-derived hepatic progenitors after 4-day treatment with TGFβ inhibitors A83 (A8301, 1 µM) or SB (SB505124, 1 µM) or TGFβ activator ACTIVIN at varying doses (10, 30, and 50 ng/mL) in the presence of base media. g) Immunostaining of CK7 and CK19 expression in day 13 hPSCderived biliary progenitors, with DAPI nuclear counterstain DAPI, scale = 100µm. h) qPCR gene expression of day 12 liver cells after treatment with 8-BromocAMP (cAMP) and Forskolin (Fsk) in the presence of Dexamethasone (Dex). i) Gene ontology terms enriched amongst the genes regulated by DAPT, AAP or INS during the liver bud à hepatic progenitor differentiation step. j) Protein expression of ALBUMIN, HGD and GAPDH in hPSCs (undifferentiated H1, H7, H9), day 18 H1-, H7-and H9-derived hepatocytes and adult human hepatocytes (AHH) as shown by western blot. k) Day 18 hPSC-derived hepatocytes stained with Periodic Acid-Schiffs (PAS) stain; scale = 400µm. Each western blot is a representative of 2 independent experiments.