Using insights from mammalian liver development, we developed a differentiation protocol to generate hepatocyte-like cells from hESC by sequential induction of primitive streak/mesendoderm/definitive endoderm, followed by gradual hepatic maturation. The same protocol, with omission of FCS, can also be used to induce differentiation of rMAPC.
Two distinct populations of endoderm are sequentially induced during mouse development. Concomitant with implantation (E4.5), the PrE delaminates from the surface of the blastocoele of the inner cell mass (ICM). In contrast to PrE, cells fated to become DE, are derived from epiblast cells which ingress the PS during the process of gastrulation. DE is then specified to hepatic endoderm, hepatoblasts and finally mature hepatocytes in response to a series of factors, some secreted by surrounding cells 
. As both types of endoderm express many transcripts in common (Foxa2
, Sox17, Hnf4α, Afp
, …), one should take into account the minimal differences in transcript and protein expression between these two types of endoderm when investigating in vitro
differentiation of pluripotent stem cells towards hepatic endoderm. Thus, monitoring the sequential and transient expression of a complement of genes/proteins consistent with PS/ME, during the course of the in vitro
differentiation process can help to distinguish between DE and PrE. Although some of the PS/ME/DE genes assessed here, are also expressed in VE (i.e. Mixl1
) or trophectoderm (i.e. Eomes
) or later during development (i.e. Cxcr4
), the transient up-regulation of all of these genes together in response to Activin-A and Wnt3a and their subsequent down-regulation upon withdrawal of these cytokines in both hESC and rMAPC, strongly indicates transition through a PS-like intermediate prior to acquisition of hepatic characteristics. These findings at the transcriptome level were also substantiated at the protein level. It should be noted that even though we demonstrate that at least 5% of rMAPC differentiate to Sox7−
DE cells, another fraction of rMAPC may differentiate towards VE, given the very early rise of VE genes such as Tmprss2
, Afp, Ttr
, and persistent expression of Sox7 protein on d6. In contrast, no substantial up-regulation of SOX7, AFP, TTR
was found before d10 in hESC. In addition, the highest levels of PS/ME/DE gene expression was obtained at an earlier time point in hESC compared to rMAPC, consistent with the notion that hESC are developmentally comparable with epiblast/embryonic ectoderm cells, the stage just prior to gastrulation, while we hypothesise that rMAPC first have to switch fate from a PrE-like phenotype to epiblast-like cell, thereby gaining the ability to undergo “gastrulation”.
It is at first sight surprising that cells with a phenotype consistent with PrE can be fated to intra-embryonic cells (PS/ME/DE) and hepatocyte-like cells. It is commonly believed that DE cells derived from the epiblast are the sole cells that, following gastrulation, give rise to epithelium of the digestive tract, pancreas, liver and lungs, whereas VE is derived from the subpopulation of Nanog-
cells in the ICM. However, a recent lineage tracing study in early embryos in vivo
has demonstrated that this dogma may not be true 
. Kwon et al
demonstrated that VE cells can, at the time of gastrulation, intersperse with epiblast derived cells, enter the embryo proper, proliferate and contribute to 10–40% of the epithelium of the foregut, midgut and hindgut of 12–18 somite embryos 
. Even though the study did not address whether these VE cells themselves are patterned to cells of PS, ME and then DE, they found down-regulation of the VE marker, HNF4α
, in the distally positioned VE-derived cells that are eventually incorporated in the embryo proper. There is also mounting evidence that fate decisions early during development between PrE and epiblast are metastable. For instance, expression levels of Nanog
in ESC and the ICM can fluctuate, where Nanoglow
cells may start to re-express Nanog
, allowing differentiation not to VE and PE, but also intra-embryonic cell types 
. In contrast to the findings that cells initially expressing Oct4, Sox17, Foxa2
can be fated to PS/ME/DE and further to hepatocyte-like phenotype, Séguin et al, recently demonstrated that forced expression of SOX7
in hESC results in the commitment of hESC to cells with a phenotype similar to extra-embryonic endoderm (XEN) cells, which still co-express NANOG
. In contrast to rMAPC, SOX7-hESC could not be induced to differentiate to cells with hepatic or pancreatic characteristics 
. As SOX7
decreased Wnt/β-catenin-stimulated transcription 
, it is possible that the inability to suppress the constitutively expressed SOX7
, which may prevent activation of β-catenin via Wnt3a signaling, is responsible for the absence of PS/ME/DE induction. How rMAPC exactly can commit to cells expressing PS/ME/DE transcripts, why some rMAPC differentiate to VE and others apparently to PS/DE and finally, whether PrE cells derived from rat blastocysts, like XEN-P cells 
can be fated to PS/ME/DE and hepatic endoderm using this protocol, will require further evaluation.
A prerequisite for the further activation of the hepatic gene program is the significant induction of LETFs. As a combination of these LETFs became expressed at substantial levels, hESC and rMAPC progeny should be poised to differentiate further towards mature hepatocytes, if supplied with additional proper cues. Indeed, following exposure to 3 additional steps, cells with different levels of maturation emerged by the end of the differentiation protocol, as evidenced by the significant expression of post-natal hepatic specific markers (G6p, Cx32, Cyp’s, Pepck), combined with a maintained expression of genes expressed in more primitive hepatic cells, such as Afp, which normally rapidly decreases at birth and is no longer expressed in mature hepatocytes. Indeed, most hESC and rMAPC-progeny stained for Afp at the protein level, some in co-expression with Alb, while only a minor fraction was Alb+/Afp−.
Although the hepatocyte-like cells derived from both hESC and rMAPC display several hepatic functions, such as albumin secretion, urea production, glycogen storage, Cyp450 and GST activity, levels were in general still approximately 5 to 10-fold lower than in mature hepatocytes. Thus, despite the use of cytokine cocktails known to play a role during liver development, the in vitro culture system still does not recreate all signals present in vivo that govern a coordinated maturation from pluripotent cells to terminally differentiated mature hepatocytes. As organogenesis does not solely depend on soluble factors but also cell-cell interactions, development of 3-dimensional culture systems wherein the anatomical features of developing liver lobules are recreated and the use of bio-reactors making it possible to more closely control physiological parameters such as pH and glycemia, may be needed to allow the creation of hepatocytes with fully mature characteristics and functions.