Author contributions: B.B. and C.M.V.: wrote the review together.
Disclosure of potential conflicts of interest is found at the end of this article.
First published online in STEM CELLSEXPRESS November 21, 2012.
This article discusses the implications of the recent discovery that rat bone marrow-derived multipotent adult progenitor cells (rMAPCs), a cell type with broad somatic differentiation potential but of uncertain lineage identity, are similar to rat blastocyst-derived extraembryonic endoderm precursor (rXENP) cells, which appear to represent the committed extraembryonic endoderm precursor of the blastocyst. It was found that under rMAPC culture conditions, rXENP cells can be homogeneously cultured and similar cells, named rat hypoblast stem cells (rHypoSCs), can be derived from rat blastocysts more rapidly and directly. The detailed comparison of rHypoSCs, rXENP cells, and rMAPCs revealed highly similar gene expression profiles and developmental potentials. The significance of these findings for embryology, stem cell biology, and medicine is discussed. Specifically, the results assign a lineage identity to rMAPCs, indicate that rMAPCs originated by environmental reprogramming, and imply that HypoSCs, XENP cells, and MAPCs possess lineage plasticity. The MAPC-HypoSC relation also strengthens the consistency of rat and mouse embryology and consequently the idea that HypoSCs represent the committed extraembryonic endoderm precursor of the blastocyst. On this basis, it is argued that the direct comparison of HypoSCs (now available in pure form) with embryonic stem cells will be highly useful for the understanding of pluripotency and plasticity. Finally, the new findings suggest an explanation for an obscure observation on stem cell-induced transplantation tolerance. Thus, the HypoSC/XENP/MAPC phenotype provides a unique but broadly instructive model with which to study stem cell plasticity, reprogramming, and transplantation tolerance, all central themes in regenerative medicine. STEM CELLS2013;31:620–626
RODENT MULTIPOTENT ADULT PROGENITOR CELLS AND EXTRAEMBRYONIC ENDODERM PRECURSOR CELLS
Rodent (mouse and rat) multipotent adult progenitor cells (MAPCs) were first described in 2002 as cell lines that arise from explanted bone marrow, can be expanded without obvious signs of senescence, and have multipotent somatic differentiation capacity . While the lineage identity of MAPCs remained unclear until recently, their products were found to include known cell types from at least two germ layers, that is, mesoderm (endothelial, smooth muscle, blood cells, and osteoblasts) and definitive endoderm (hepatocytes and pancreatic beta cells); neuroectodermal differentiation seems also possible but has not yet been sufficiently substantiated (for an overview of the known differentiation potential of MAPCs, see ). Gene expression profiling of rodent MAPCs revealed in 2007 that these cells coexpress the pluripotency factor Oct4 with a group of markers deemed characteristic of the extraembryonic endoderm . Just 1 year earlier, the committed extraembryonic endoderm precursor (ExenP; Fig. 1) had been identified in the inner cell mass (ICM) of murine preimplantation embryos (blastocysts) and found to distinguish itself from the subsequent developmental stages by the presence of Oct4 . Thus, the profiling of MAPCs raised the possibility of an early extraembryonic endodermal lineage identity. In 2009, rat blastocyst-derived cell lines with ExenP-like features were published, which appeared to have gene expression features very similar to MAPCs. These cell lines, named extraembryonic endoderm precursor cells (XENP cells), express Oct4 along with extraembryonic endoderm markers, differentiate into parietal and visceral endoderm in vitro, and contribute to the same lineages in vivo . It was therefore logical to perform a direct comparison of rodent MAPCs and XENP cells, and this article will discuss the results and implications of this comparison.
COMPARISON OF MAPCS WITH XENP CELLS LED TO ISOLATION OF HYPOBLAST STEM CELLS
Derivation of rat MAPCs (rMAPCs) has been extensively described [1, 6-8]. Since 2007, several cell lines were isolated from the bone marrow of E18-3-week-old Fisher and Sprague-Dawley rats using the rMAPC isolation methods described in . From the 12 independent isolations, starting with 2–12 × 107 bone marrow cells, four isolations (33%) yielded one to four homogenous populations of cells. Appearance of the typical refractile small cells occurred after 6–10 weeks of culture of the CD45− cell fraction obtained 1 month after initial plating of the bone marrow cells by CD45-column depletion. All new cell lines expressed transcripts for Oct4, Sall4, Gata4, Gata6, Sox7, Sox17, and Foxa2; Oct4, Sox7, Sox17, Gata4, and Gata6 protein; as well as CD31 and SSEA1, both markers of the ICM [9, 10]. Thus, although derivation of rodent MAPCs is still not easy, for nearly a decade, it has been possible to maintain the above-mentioned established undifferentiated rodent MAPCs in culture in homogeneous form [3, 6, 7]. In contrast, rat XENP (rXENP) cells isolated on fibroblast feeder layers with Leukemia Inhibitory Factor showed extensive extraembryonic endodermal differentiation under their original culture conditions . Application of reciprocal culture conditions to rMAPCs and rXENP cell lines led to an immediate reciprocal change in cell line morphology and differentiation marker expression . Low-density-plated rXENP cells now formed colonies with comparable efficiency as under original rXENP conditions, but without evidence of differentiation. Growth factor requirements, lineage markers, and somatic in vitro differentiation products of the rMAPC medium-shifted rXENP cells were highly similar with those of rMAPCs.
Consequently, rMAPC culture conditions were tested for direct cell line isolation from rat blastocysts. Cell lines with rMAPC morphology emerged rapidly from the explanted blastocysts and were dubbed hypoblast stem cells (HypoSCs). The rat HypoSCs (rHypoSCs) were highly similar to rMAPCs and MAPC medium-shifted rXENP cells, not only by morphology but also by growth factor requirements and in vitro differentiation potential . Microarray analysis confirmed the near-identity of gene expression patterns of rMAPCs and rHypoSCs, and both rMAPCs and rHypoSCs exhibited a similar in vivo potential, that is, contributions to both visceral and parietal endoderm, plus some contribution to trophoblast, as previously described for rXENP cells. Thus, the comparison of rXENP cells and rMAPCs revealed a significant similarity between the two phenotypes and led to the isolation of rHypoSCs. Overall, these similarities justify the claim that rXENP cells, rHypoSCs, and rMAPCs represent the same cell type. Note that this claim does not rule out the possibility of differences that might be meaningful in certain contexts. For example, (female) HypoSCs, with their likely ExenP origin (section “HypoSCs as a cell line model of the committed extraembryonic endoderm precursor”), are expected to show paternal X chromosome inactivation , while (female) MAPCs, with their likely somatic origin (section “Identity, plasticity, and origin of rat MAPCs”), may show random X chromosome inactivation, even if this still needs to be demonstrated. Also, it would not be surprising if MAPCs maintained an epigenetic memory of their somatic tissue of origin, similar to findings with induced pluripotent stem cells (iPSCs) . In the following, we will draw some conclusions, develop some hypotheses, and raise some questions that emerge from these findings.
CONSEQUENCES OF THE HYPOSC-XENP-MAPC RELATIONSHIP
HypoSCs as a Cell Line Model of the Committed Extraembryonic Endoderm Precursor
In the blastocyst, the committed extraembryonic endoderm precursor (ExenP) and the committed epiblast precursor (EpiP) are two small, intermingled groups of cells that constitute the ICM (Fig. 1) and give rise to the yolk sac endoderm and fetus/amnion/extraembryonic mesoderm, respectively. Right after their specification, the ExenP cells sort to the surface of the ICM and form the morphologically and molecularly distinct hypoblast layer that is also known as the primitive endoderm, or PrE [4, 13]. Like ExenP and PrE, transplanted rXENP cells contributed to all extraembryonic endoderm lineages, but unlike PrE and like ExenP cells, rXENP cells express Oct4. Furthermore, like transplanted ExenP and unlike transplanted PrE, rXENP cells can contribute to the trophectoderm (TE) (compare  with ). These observations argue that rXENP cells are more like ExenP than PrE.
However, some observations weakened the conclusion that rXENP cells are similar to the ExenP. First, during the derivation of rXENP cells from blastocysts, there appeared to not be lineage continuity: Oct4 mRNA disappeared within approximately 4 days after blastocyst explantation but re-emerged approximately 1 week later, coinciding with the emergence of rXENP cells. Second, the rXENP cells exhibited uncontrollable differentiation and could therefore not be maintained undifferentiated; therefore it was difficult to phenotype them in detail, and it could not be proven that the observed lineage contributions came directly from the subpopulation of rXENP cells in the cultures. These two limitations were overcome by recognizing the rMAPC connection, as rMAPC culture conditions allowed the homogenous culture of rXENP cells and the rapid isolation of similar or identical cells (i.e., rHypoSCs) without discernable discontinuity. Even though phenotypically, there appear not to be differences between rXENP cells and rHypoSCs, in view of their continuous, more rapid outgrowth, rHypoSCs would appear as the more legitimate ExenP representative. Therefore, from here on, we will discuss the blastocyst-derived ExenP-like cell line entity in terms of HypoSCs and mention rXENP cells only in selected cases.
Still, a subtle problem regarding the identity of HypoSCs remains. When aggregated with morulae, ExenP cells frequently change their lineage identity toward epiblast fates , but contribution of transplanted rHypoSCs/rXENP cells to the epiblast lineage was not observed [5, 8]. However, this difference between HypoSCs and ExenP (if confirmed by direct comparison within the same species) would appear analogous to the embryonic stem cell (ESC)-EpiP relationship. ESCs, albeit considered to be a model of the naïve epiblast , do not efficiently contribute to extraembryonic lineages in vivo , but transplanted EpiP cells do . Thus, ESCs are clearly not fully identical with their presumed in vivo correlates; in fact, they are known to exhibit in vitro adaptations such as the upregulation of self-renewal transcripts and repressive epigenetic markers , and the same may be expected for HypoSCs. Although the disparate in vivo behaviors of the ICM cell types and their closest known cell line correlates are not yet fully understood, these differences do not appear to be fundamental given the ease with which HypoSCs and ESCs show in vitro a reciprocal lineage plasticity that is comparable to the reciprocal lineage plasticity shown in vivo by ExenP and EpiP cells (this point is discussed in the separate section on reciprocal plasticity below).
Regardless of the subtleties in assigning an exact identity to HypoSCs, these cell lines offer exciting prospects for studying yolk sac and developmental biology, especially in conjunction with genetic approaches that are applicable to cell lines. In this context, it should be an immediate priority to further improve the degrees of chimerism achieved with rHypoSCs.
The MAPC-HypoSC Connection as a Bridge Between Rat and Mouse Embryology
The foregoing discussion tacitly mingled rat and mouse embryology, since HypoSCs have been isolated from rat only, but most of what we know about the ExenP cells was discovered in the mouse. Is such mingling justified?
A general argument comes from the fact that stem cell isolates thought to represent EpiP and TE are highly similar between mouse and rat: ESCs are very similar in morphology, molecular signature, and developmental potential between mouse and rat [19, 20]; even rat-mouse chimeras have been produced using rat and mouse ESCs or the ESC-like iPSCs [21, 22]. Likewise, the blastocyst-derived trophoblast stem cells (TSCs) appear highly similar between mouse and rat [23, 24]. Thus, two out of the three early blastocyst cell types are represented in vitro by cell lines that are highly similar between mouse and rat, suggesting that cell line correlates of the third cell type (ExenP) would also be highly similar between mouse and rat.
Indeed, several observations argue directly that mouse HypoSCs may be isolated. First, as outlined already, the lineage markers and developmental potential of the rHypoSCs closely match the known specifics of the mouse ExenP. Second, from mouse ESC cultures, a small fraction with characteristics similar to those of mouse ExenP cells and rHypoSCs has been enriched: these Nanog-negative cells coexpressed Oct4 and SSEA1 together with extraembryonic endoderm markers and contributed to visceral and parietal endoderm in vivo . These “spontaneously” arising ExenP-like cells that were not grown as a line, may be responsible for the fact that ESCs can, at low frequency, contribute to the yolk sac .
Third, rHypoSCs and rXENP cells can chimerize the parietal and visceral endoderm not just of the rat but also of the mouse [5, 8, 14]. Fourth, mouse MAPCs show extraembryonic endoderm lineage markers, growth factor requirements, and in vitro differentiation potentials similar to rMAPCs, and notably also express Oct4, albeit at lower levels than rMAPCs . Interestingly, one mouse MAPC line in  was also capable of contributing to the embryo proper, albeit at low frequency. Whether contribution to extraembryonic endoderm is also possible with mouse MAPCs/HypoSCs is currently being tested.
In addition to the cell lines just mentioned, extraembryonic endoderm cell lines (XEN cells) have previously been isolated from mouse blastocysts . Mouse XEN cells show important similarities to rHypoSCs [5, 8], but lack the ICM markers that are typical of HypoSCs (such as Oct4, SSEA1, CD31, and Rex1 [5, 26, 27]) and show appreciable contribution to the visceral endoderm only after treatment with Nodal or Cripto . We therefore believe that mouse XEN cells, rather than being the mouse equivalent of rHypoSCs, represent a slightly more advanced developmental stage, the PrE. This hypothesis predicts not only that distinct rat XEN cells and mouse HypoSCs may eventually be isolated but also that XEN cells are less plastic than HypoSCs (see the next two sections). Figure 2 summarizes these ideas by proposing an amalgamated rat-mouse schematic that includes the published rodent stem cell lines with a known lineage identity.
HypoSCs Versus ESCs: Reciprocal Lineage Plasticity and the Study of Pluripotency
rHypoSCs promise to expand our options for studying the ExenP in its normal developmental context. However, rHypoSCs are also of interest beyond yolk sac biology, mainly because of their lineage plasticity and “special relationship” with pluripotent ESCs.
Although untreated rHypoSCs apparently do not easily undergo a somatic lineage switch in the morula aggregation assay, they easily differentiate somatically in vitro ; this was also discovered as a consequence of their similarity with rMAPCs that had known somatic multipotentiality, as summarized in . Interestingly, the differentiation patterns of HypoSCs appear reciprocal to those of ESCs. Indeed, although ESCs are essentially restricted to the fetal lineages in vivo [17, 33], they can easily differentiate into extraembryonic endoderm in vitro [34, 35]. The reciprocal behavior of HypoSCs and ESCs is likely to reflect the developmental relationship of their presumed in vivo equivalents, ExenP and EpiP, which can easily be converted into each other in situ by manipulating fibroblast growth factor signaling [15, 16, 36].
The apparently reciprocal relationship between HypoSCs on one hand and ESCs on the other hand allows us to take advantage of the historically more detailed knowledge about ESCs and make predictions about the behavior of HypoSCs. That is, as extraembryonic endodermal differentiation of ESCs seems to pass through a HypoSC-like step , we hypothesize that somatic differentiation of HypoSCs passes through an ESC-like step (Fig. 1); this could be tested by monitoring Nanog expression or the female X chromosome inactivation patterns . Moreover, as culture conditions for rat ESCs are known [19, 20], it should be possible to capture that ESC-like step. Then, rHypoSCs should be able to differentiate into all the same cell types as ESCs. This would then suggest that rHypoSCs are probably fully pluripotent, albeit in a latent form; “latent” meaning that the pluripotency unfolds under specific culture conditions. Regardless of whether that hypothesis is true, it will be interesting to delineate the exact trans-specification path of somatically differentiating HypoSCs. The early steps of the transition of the ESC phenotype toward the extraembryonic endoderm lineage are under much scrutiny (e.g., [25, 35, 37]), while the reciprocal pathway has not yet been studied at all, which now should be possible.
In view of the close and reversible relationship between ExenP and EpiP [15, 16, 36], a detailed comparison between HypoSCs (plastic, latently pluripotent) and ESCs (plastic, overtly pluripotent) should be instructive for a sharper molecular definition of pluripotency and for separating that phenomenon from the plasticity phenomenon. At least in the rat, such a comparison is now directly possible as we have homogenous ESC and HypoSC lines.
Oct4 May Be a Key Factor that Maintains Lineage Plasticity
As just pointed out, the study of HypoSCs (especially in comparison with ESCs) is likely to be useful for understanding the molecular basis of pluripotency, plasticity, and early lineage determination. It deserves interest in this context that rHypoSCs, and their presumed in vivo correlate, the ExenP, express Oct4. Oct4 is considered a key determinant of pluripotency and a regulator of early lineage choice [38–40]. However, the function of Oct4 in HypoSCs is apparently not to maintain directly pluripotency (see previous section), and Oct4 is clearly not a pluripotency factor in ExenP cells, which normally produce the same differentiated cells (the yolk sac endoderm) as their immediate progeny, the Oct4-negative PrE. We suggest that the function of Oct4 in the ExenP and HypoSCs is related to their lineage plasticity. Indeed, a recent study exploring the plasticity of the mouse ICM observed that plasticity is lost in the extraembryonic endoderm lineage at the same time that Oct4 expression is lost, that is, the PrE is no longer plastic . Likewise, TE in bovine blastocysts, which implant later than mouse blastocysts, expresses unlike in mouse, Oct4, and maintains its plasticity . Furthermore, exogenous Oct4 appears to facilitate reprogramming in other cellular contexts and independently of pluripotency [42–44]. Thus, HypoSCs may be ideal to study the plasticity function of naturally expressed Oct4. Our hypothesis predicts that knockout/knockdown of Oct4 in HypoSCs will not affect their extraembryonic endoderm lineage potential, but will reduce their ability for somatic differentiation. Regardless of whether the hypothesis is true, the study of Oct4 in this new cellular context is likely to reveal interesting new facets in the biology of this important factor.
Plasticity and the Origin of XENP Cells
As already mentioned, rXENP cells appeared to originate from Oct4-negative and morphologically distinct blastocyst outgrowths . With appropriate feeder cells, these same outgrowths also gave rise to cell lines that, unlike the rXENP cells, maintained the original outgrowth morphology  and later turned out to be trophoblastic in nature . Similar or identical cell lines (termed “extraembryonic stem cells”) from rat blastocysts were also isolated in another laboratory and found to give rise to both giant trophoblast and extraembryonic endoderm cells in vitro . The identity of these cells is unknown, but they might represent the first committed stage of the TE lineage. In any case, these observations indicate that the plasticity of the early blastocyst cells is not limited to the epiblast and yolk sac endoderm lineages, and that stem cell lines representing the ExenP phenotype may originate directly (HypoSCs) or indirectly (XENP cells) (Fig. 1). Recent findings suggest that the same applies to the stem cells representing the EpiP phenotype: mouse ESCs also may originate directly, that is, by continuously preserving the naïve epiblast state, or indirectly, by passing through a germ cell precursor state [48, 49].
Identity, Plasticity, and Origin of Rat MAPCs
For years, the lineage identity of rat and mouse MAPC had been a riddle. Initially, they were deemed to be a subpopulation of mesenchymal stem cells or related to mesenchymal stem cells , as MAPCs arise from the fraction of the bone marrow that adheres to culture plates and lacks the CD45 surface antigen. With time, improvements of the culture medium have led to a better-defined and better-reproducible (but probably also slightly different) phenotype that eventually appeared more similar to endodermal or extraembryonic endodermal cells . But it was the discovery that rMAPCs are similar to rXENP cells that enabled us to assign an exact lineage identity to rMAPCs. The immediate implication of this finding is, of course, that the somatic differentiation of rMAPCs is based on lineage plasticity, just as already discussed for rHypoSCs.
A second implication concerns the origin of rMAPCs. The assignment of an extraembryonic lineage identity to rMAPCs immediately suggests strongly that they originate through reprogramming (Fig. 1). A further argument is provided by the analysis of Oct4 gene expression. Although rodent MAPCs, as well as many other multipotent/progenitor cells, isolated from bone marrow and other tissues, express Oct4 (see , for a list), Oct4 could not be detected in bone marrow or other adult somatic tissues (excluding gonads) using Oct4-Enhanced Green Fluorescent Protein reporter mice [50, 51]. Moreover, the ablation of Oct4, using a conditional knockout strategy, did not show impairment in tissue homeostasis and regeneration capacity in the adult animals . Consistent with this observation, we have no evidence that rMAPCs can be found in bone marrow . In line with the idea that rMAPCs are not present in the bone marrow are the facts that their derivation is extremely slow and inefficient and that bone marrow explants from which rMAPCs emerge do not express the rMAPC molecular signature . It is also hard to imagine that an early, preimplantation-stage extraembryonic cell type persists in the bone marrow, especially when considering that rXENP cells and rHypoSCs are highly tumorigenic in ectopic locations [8, 14]. Nevertheless, given the recent finding that some gut cells can be traced back to the extraembryonic endoderm precursor , a theoretical possibility remains that a closely related, but Oct4-negative, extraembryonic endoderm descendent occurs rarely in the bone marrow.
In any case, the mechanism of origin of rMAPCs is a worthwhile research topic. If rMAPCs originate by reprogramming in vitro, which seems almost certain, then the reproducibility of their isolation, despite the low efficiency, together with the fact that cytogenetically normal rMAPCs have been isolated, would argue for a nonmutagenic mechanism, that is, rMAPCs probably represent a case of environmentally driven reprogramming. Environmentally driven reprogramming has been observed before [35, 53-56], but the origin of MAPCs is particularly spectacular, as it appears to bridge a much larger epigenetic distance, in fact a distance that is comparable with the creation of ESC-like cells (iPSCs) from fibroblasts. Therefore, it should not surprise that (based on microarray analysis) rMAPCs are not 100% identical with rHypoSCs , just like iPSCs and ESCs maintain subtle differences . The exciting difference from iPSCs is that the derivation of rMAPCs was achieved without the direct manipulation of an intracellular process. Clearly, clarification of the origin of (at least rodent) MAPCs is of great interest for understanding the interaction of environment and epigenome, and hence for the reprogramming field and regenerative medicine.
Noteworthy, human (h)MAPCs  exhibit significant differences with rodent MAPCs. hMAPCs do not express the same complement of hypoblast-related transcripts. Furthermore, human cells have less robust capability to commit to endodermal cell types and do not form yolk sac tumors as described here for the rat equivalent.
A Lesson for Transplantation Tolerance?
The lineage plasticity of rMAPCs, rHypoSCs, and rXENP cells and the phenotypic similarity of these three cell line isolates suggest a reassessment of older findings regarding stem cell-induced transplantation tolerance. Ten years ago, it was shown that the above-mentioned trophoblastic rat blastocyst-derived cell lines , when injected into the portal veins of adult, immunocompetent, fully Major Histocompatibility Complex-mismatched rats, were able to induce the acceptance (without further host conditioning) of transplanted hearts of the same strain origin as the injected cells, while third-party hearts were rejected . The phenomenon was linked to the development of partial-lineage mixed hematopoietic chimerism and could be prevented by thymectomy. Initially, the injected cells were deemed ESC-like (based on morphology and the markers SSEA1 and Alkaline Phosphatase), but they lack Oct4 and Nanog while expressing the TE marker Cdx2 and easily form giant trophoblast cells in vitro . As mentioned before, similar if not identical Oct4-negative rat extraembryonic stem cells spontaneously formed extraembryonic endoderm cells  that, with present knowledge, almost certainly were rXENP cells, as the culture conditions were almost identical. It is plausible that the trophoblast-related immunological properties of the Cdx2-positive cells enabled their initial survival in the allogeneic liver, similar to what was recently demonstrated for mouse TSCs . The injected TSCs also provided immune protection for coinjected cells ; therefore, it is conceivable that any XENP-like cell that could have formed from the extraembryonic Cdx2-positive cells (see above) may survive and enter somatic differentiation, eventually leading to hematopoietic chimerism. In support of this scenario, tail vein-injected mouse MAPCs, which are very similar to rMAPCs , have been shown to give rise to blood chimerism in vivo .
The unlikely encounter of a bone marrow-derived cell type (MAPCs) with a blastocyst-derived cell type (HypoSCs/XENP cells) has led to important conclusions and new opportunities that are of broader interest: we now have a homogeneous cell line model that is at least very similar to the ICM-stage rat extraembryonic endoderm precursor; we have realistic hope to isolate such cell lines also from mouse; we have a new in vitro model to study the role of Oct4 in early embryonic plasticity and to better define pluripotency via comparisons with ESCs and the committed epiblast precursor. In addition, we identified a dramatic case of adult cell reprogramming caused by merely altering the environment yielding rMAPCs, which have a similar lineage identity as the ICM-stage extraembryonic endoderm precursor. Finally, we can now postulate an explanation for an obscure finding on stem cell-induced transplantation tolerance.
Research in the lab of B.B. has been supported by KOSEF project #0000179 and NRF project #2012007774. C.M.V. was supported by an Odysseus grant from FWO, Flanders, and by an OT, GOA, CoE, and PF K.U. Leuven award.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
B.B. declares no conflict of interest. C.M.V. is a consultant for ReGenesys.