Our data have revealed several important insights about differentiation from hPSCs. First, our data showed that hESCs and hiPSCs make specified derivatives that are nearly equivalent transcriptionally. This was surprising considering the vastly different circumstances by which hESCs and hiPSCs are derived, and in light of well-documented differences between them at the epigenetic and transcriptional levels in the undifferentiated state 2, 3, 4, 5, 6, 8, 9, 43, 44
. This similarity might be attributed to the fact that only high-passage hiPSCs (> 40) were used. However, even at high passage, a small number of genes still appeared to distinguish the undifferentiated hiPSC and hESC lines used. The fact that these differences were largely undetectable in the differentiated state could suggest that the progeny of these cell types are more similar than their parental cells or that the genes differentially expressed between them are not expressed in the specified progeny.
Second, upon differentiation, we did not detect appreciable re-expression of the exogenous reprogramming factors. Because the expression of OCT4 was undetectable in any PSC derivative, it is unlikely that the loci representing the retroviral reprogramming factors were re-activated upon differentiation or even continued culture. This finding could suggest that, at least in the contexts analyzed here, concerns over re-expression of oncogenic factors from hiPSC lines generated by viral integration could be mitigated by specification of hiPSCs to even a progenitor state, such as NPC, hepatoblast, or FB.
Third, both hESCs and hiPSCs made progeny that continued to express a group of genes known to play roles in very early embryonic development. While the progeny of ESCs have been proposed to represent embryonic cell types because of the primitive nature of the starting cell types 23, 25, 45
, to our knowledge human pluripotent cell-derived progeny have not yet been placed into such a narrow developmental context. The most logical interpretation of our data is that the PSC derivatives generated here represent cells similar to those found earlier than 6 weeks of development.
It remains possible, however, that current protocols to make hPSC derivatives generate cells that do not completely represent cells found in tissue at any developmental stage. With limited access to tissues representing the earliest stages of development, this possibility will remain unexplored for now. It is also possible that new culture conditions could be defined that improve differentiation in vitro to better recapitulate that which occurs in vivo. In addition, it is possible that experimentally manipulating the expression of early embryonic genes described here could be used to accelerate development in vitro. Regardless of the fact that PSC derivatives produced in other labs also express some of the same early embryonic genes suggest that many protocols lead to a consistent result, so perhaps a shift in differentiation strategy or expectations is required.
Our data also suggest that simple approaches, such as continued culture, can further the development of PSC-derived cells, though this method also has its limitations as described above. Nevertheless, it is tempting to speculate that, upon specification, the differentiation process is more or less pre-determined by mechanisms that lock cells into a process that takes a specific amount of time, or number of cell divisions complete, regardless of the culture conditions employed. This idea has significant support from studies with in vitro
murine development 45
, where differentiation of mESCs under minimal conditions allowed appropriate temporal and regional specification of neural tissue.
Recent work suggested that hPSC derivatives share hallmarks of gene expression with oncogenic cells 46
. Our data suggest that the presence of oncogenic hallmarks could be explained by persistent expression of genes in hPSC derivatives that are typical of the early embryo and known to be re-expressed in cancers. For example, while LIN28
is not normally expressed in post-natal mammalian tissue, it was recently shown to be re-expressed in 15% of human cancers; and cancers with poor prognosis that are high in LIN28
expression are low in let-7 family expression 39
. This study also showed that overexpression of LIN28
drove transformation of FBs, consistent with its proposed role in reprogramming 35, 38
. Therefore, if the derivatives of PSCs are to be used clinically, it could be important to take into account the residual expression of the early embryonic genes, particularly, LIN28
. Finally, one of the great benefits of iPS technology is the ability to model human diseases in vitro
using patient-derived cells. Our data would suggest that it could be difficult to model human diseases in this context, unless a phenotype manifests very early in development.