Genetic approach to track neural cell fate decisions using human embryonic stem cells

With their capability to undergo unlimited self-renewal and to differentiate into all cell types in the body, human embryonic stem cells (hESCs) hold great promise in human cell therapy. However, there are limited tools for easily identifying and isolating live hESC-derived cells. To track hESC-derived neural progenitor cells (NPCs), we applied homologous recombination to knock-in the mCherry gene into the Nestin locus of hESCs. This facilitated the genetic labeling of Nestin positive neural progenitor cells with mCherry. Our reporter system enables the visualization of neural induction from hESCs both in vitro (embryoid bodies) and in vivo (teratomas). This system also permits the identification of different neural subpopulations based on the intensity of our fluorescent reporter. In this context, a high level of mCherry expression showed enrichment for neural progenitors, while lower mCherry corresponded with more committed neural states. Combination of mCherry high expression with cell surface antigen staining enabled further enrichment of hESC-derived NPCs. These mCherry+ NPCs could be expanded in culture and their differentiation resulted in a down-regulation of mCherry consistent with the loss of Nestin expression. Therefore, we have developed a fluorescent reporter system that can be used to trace neural differentiation events of hESCs.


INTRODUCTION
Human embryonic stem cells (hESCs) undergo unlimited self-renewal and retain pluripotency to differentiate into all cell types of the body, making them invaluable tools for development of cellular therapeutics (Fu and Xu, 2011). Modeling human disease has further become possible with advances in genetic modification in hESCs either through homologous recombination (Song et al., 2010;Hockemeyer et al., 2011) or gene knockdown (Ordonez et al., 2012), as well as reprogramming of cells having disease-causing mutations into an ESC-like state (human induced pluripotent stem cells or hiPSCs) (Unternaehrer and Daley, 2011). Given the limitation of using mouse systems to develop human therapeutics, the use of these approaches can provide more physiologically relevant insight into disease pathology and help to identify more efficacious drug targets. These approaches additionally permit disease modeling of tissues that are previously unattainable or that show pathologies too advanced in donor tissues. This includes studies on the mechanisms underlying neuronal degeneration associated with central nervous system (CNS) diseases (e.g. Parkinson's, Alzheimer's or Amyotrophic lateral sclerosis).
Studies using human pluripotent stem cell derived neurons to model these diseases or generate neural tissue replacements rely upon the effectiveness of protocols that recapitulate normal embryonic development to direct the differentiation of hESCs into the neural cell types afflicted by the disease (Aas et al., 1996). A key step in this process is to differentiate hESCs into neural stem cells (NSCs) that can be further differentiated into neurons, astrocytes and glia. While significant progress has been made in improving the differentiation of hESCs into NSCs (Reubinoff et al., 2001;Itsykson et al., 2005;Wu et al., 2007;Chambers et al., 2009), this step remains an imperfect process with resulting cultures comprising a mixture of undifferentiated ESCs, NSCs, neurons, glia and non-CNS cell types including neural crest. Subsequent differentiation of these mixed cultures would fail to generate single pure lineages needed for transplantation, modeling or population based gene expression studies (e.g. RNA-seq).
To circumvent these issues, several groups have had success in using combinatorial cell surface marker staining to enrich for NSCs (Yuan et al., 2011;Israel et al., 2012). However this method requires further sorting strategies for continued monitoring of resultant cell differentiation states. Promoter-based viral reporter systems for the neural progenitor marker Nestin has also been used to specifically label NSCs within the human fetal brain and could be used to purify these cells for expansion or for further differentiation both in culture and following transplantation (Keyoung et al., 2001). Several groups have used Nestin reporter systems to identify and track neural progenitors both in vivo and in vitro (Sawamoto et al., 2001;Lenka et al., 2002;Mignone et al., 2004;Noisa et al., 2010). These transgenic or viral based strategies, however, involve integration of the exogenous DNA into random genomic loci that can modulate the promoter activity, depending on the epigenetic state around the integration site. To address this issue, we decided to generate knock-in cell lines using a Nestin-based fluorescent reporter system that would permit not only a simpler sorting strategy for isolating NPCs/NSCs but also would enable their live tracking during expansion or differentiation. Unlike transgenic approaches (Placantonakis et al., 2009), this system would ensure reporter expression specifically driven by the endogenous Nestin promoter and/or enhancer, and would not introduce any other genetic alterations that arise from random integration strategies. This would permit specific mCherry expression during neural differentiation of hESCs and the ability to generate enriched NSC cultures where multi-potentiality could be easily monitored. These cells will provide a unique opportunity for studies requiring purer neural cultures and live cell tracking, including transplantation, disease modeling or whole transcriptome analyses (Peljto and Wichterle, 2011).

Generation of a knock-in Nestin reporter cell line
We have recently developed a bacterial artificial chromosome (BAC) based targeting strategy that enables a high efficiency of homologous recombination in hESCs (Song et al., 2010). To knock-in the mCherry coding region into the Nestin locus, we constructed a BAC targeting vector in which the ATG of one endogenous Nestin allele was replaced with that of mCherry ( Fig. 1A and 1B). In order to screen for homologous recombinants, one homologous arm of the BAC vector was shortened to 6.5 kb. Recombinants were then screened by Southern blotting to identify both the 8.8 kb germline band and the 11.2 kb mutant band (Fig. 1B). The LoxP-flanked selection marker cassette (CAG-Neo) was subsequently excised from the genome (Fig. 1C) through transient expression of the Cre enzyme (Song et al., 2007). Knock-in hESCs showed normal karyotypes ( Fig. 2A) and retained the potential to differentiate into all three germ layers including neuroectoderm ( Fig. 2B and 2C). Furthermore, mCherry-positive neural-like cells were found to migrate from teratoma tissues when they were cultured in neural progenitor medium (Fig. 2F-H). Therefore, our knock-in cell line may provide a means for monitoring the formation of neural lineages during early hESC differentiation.

mCherry fluorescence follows Nestin expression dynamics
To further validate whether our genetic tagging of the Nestin locus could allow tracing of neural differentiation from hESCs, Nestin:mCherry hESCs were differentiated with an established embryoid body (EB) neural induction protocol (Reubinoff et al., 2001;Zhang et al., 2001). mCherry positive cells could be observed soon after EB treatment with neural induction medium coincident with the expected emergence of neural progenitors (Fig. 3A). Further, mCherry positive cells were found to migrate out from attached EBs ( Fig. 3B and 3C) and rapidly expanded to confluence similarly to that observed for teratoma tissues (Fig. 2F-H). This confirms that our Nestin:mCherry reporter system can monitor neural specification from both in vitro and in vivo differentiated hESCs.
Following out-migration from EB or teratoma tissues, mCherry positive cells displayed properties consistent with neural progenitor cells. They could be further expanded in culture for several passages (data not shown); showed significant proliferation as indicated by Ki-67 staining (Fig. 3D); and co-expressed endogenous Nestin (Fig. 3E). Furthermore, high mCherry expression was localized to neural rosette-like structures, as was previously reported for a transgenic Nestin reporter (Noisa et al., 2010), and colocalized with the neural stem cell marker Sox2 (Fig. 3F). Therefore, mCherry expression can faithfully label Nestin positive cycling progenitors during differentiation of hESCs into the neural lineage both in vitro and in vivo.

mCherry levels dropped during NPC differentiation
While our results indicate that the knock-in of mCherry into the Nestin locus recapitulates its endogenous expression, we next sought to define whether mCherry specifically labels neural progenitors. To do address this, mCherry positive cells were characterized using flow cytometry (Figs. 4A and 5). Interestingly, two different fractions showing low and high mCherry expression were revealed with global passaging of neural cultures coincident with an emerging population of CD44 positive cells (Fig. 4A) representing more differentiated neural cell types (Liu et al., 2004). Continued culture in this manner resulted in the progressive loss of mCherry expression over several passages (data not shown). Our results are consistent with mCherry Lo cells potentially representing more committed neural cell types as has been previously reported for a transgenic hESC system (Noisa et al., 2010). In fact, mouse Nestin reporters exhibited gradient levels of reporter activity during differentiation, with high levels associated with neural stem cells, lower levels observed in more differentiated yet dividing neuronal progenitors and finally little to no activity in fully differentiated neurons (Sawamoto et al., 2001;Lenka et al., 2002;Mignone et al., 2004). Consistently, mCherry expression was lost in terminally differentiated neurons that were labeled with a synapsin-GFP lentiviral reporter ( Fig. 4B   addition, mCherry Hi NPCs could be differentiated into both GFAP + astrocytes and Tuj1 + neurons ( Fig. 4D and 4E), confirming their multipotency. Therefore, our results demonstrate that mCherry levels correspond with endogenous Nestin expression, specifically labeling neural progenitors and lost upon their further differentiation. One significant advantage to the labeling of NPCs with mCherry would be the potential for sorting and enriching this population. Furthermore, given the potential heterogeneity of neural progenitors in terms of stage of commitment and potentiality, it would be advantageous to specifically isolate or enrich bone fide neural stem cells. To determine the specificity of mCherry to NSCs or NPCs compared to more committed cell lineages, we used cell surface markers that highlight these different neural subpopulations. Previous reports have shown the use of multiple surface markers to identify NSCs based on a CD184 + CD271 -CD44 -CD24 + staining signature (Yuan et al., 2011). Both CD271 and CD44 identify undifferentiated hESCs, as well as neural contaminants including neural crest stem cells, peripheral neural cells, more differentiated neuronal cell types, astrocyte progenitors and glia (Morrison et al., 1999;Luo et al., 2002;Liu et al., 2004;Lee et al., 2007). Alternatively, both CD184 and CD24 have been shown to be expressed on the surface of NSCs. Consistent with mCherry high expressers representing a more neural stem cell state, the CD184 + CD44 -CD24 + population localized more within the mCherry Hi expressing cells in contrast to the CD184 -CD24population (Fig. 5). Furthermore, the CD271 -CD44 -CD24 + cells also mapped more to the mCherry Hi expressers unlike a similar CD24population (Fig. 5). These results indicate that mCherry high expression is in part predicative of a stem cell state, but is not specific to this population (Fig. 5). This is likely the result of continued high Nestin expression in more committed neural progenitors as well as a potential lag in the mCherry expression compared to endogenous Nestin that is a result of their differing protein stabilities.

mCherry sorting strategy for NSC enrichment
While high mCherry expression occurs in neural stem cells (Fig. 3F), it in itself is not sufficient to fully purify this fraction from the global population. In order to achieve this, mCherry Hi expression would need to be combined with cell  surface marker staining. Interestingly, neural cultures that were gated on the mCherry Hi expression and that were additionally either CD271 -CD44or CD44 -CD184 + (Fig. 6A) were almost entirely CD24 + when neural cells were collected directly from induced EBs. This indicates a redundancy for this marker at this stage in neural differentiation. However, this trend was not necessarily observed after passaging of the total population, likely due to the spontaneous differentiation of NPCs (data not shown). Therefore, these results indicate that mCherry Hi can substitute for the CD24 + staining when isolating NSCs directly from induced EBs, enabling a simpler sorting strategy for these cells. Consistently, the sorted CD184 + CD44 -CD271population, which was found to map to a mCherry Hi population (Fig. 6B), showed significant further enrichment of the NSC markers Sox1, Sox2 and Pax6 over the mCherry Hi population alone (Fig. 6C). Use of dual SMAD inhibitors Noggin and SB431542 during neural induction (Chambers et al., 2009) was found to significantly increase the number of NSCs sorted and showed a further increase in Sox1 and Pax6 expression levels in the enriched mCherry Hi population (data not shown). Therefore, these results indicate that the combination of mCherry Hi with a more limited surface marker sorting strategy that excludes CD24 can be sufficient to significantly enrich for NSCs. Furthermore, substituting CD24 for mCherry expression provides the distinct advantage of being able to further monitor the differentiation status of the NPC culture and permits tracking of terminal differentiation events.

DISCUSSION
Our studies demonstrate a knock-in approach that can be used together with existing strategies of purifying NPCs/ NSCs derived from hESCs while additionally providing a means of tracking these cells live, both in vitro and in vivo. Transgenic or viral Nestin reporters have been widely used to label neural progenitors (Keyoung et al., 2001;Sawamoto et al., 2001;Lenka et al., 2002;Mignone et al., 2004;Noisa et al., 2010)   Pax6). These results indicate that our reporter system can be combined with additional strategies for specifically isolating or enriching NSCs including and not limited to: surface marker staining; additional knock-in strategies conferring NSC or lineage-specific drug resistance; or knock-in fluorescent reporters of additional neural lineage markers. Furthermore, our knock-in line was able to achieve significant enrichment of NSCs with a smaller subset of surface markers (CD184 + CD44 -CD271 -) as has been previously used (Yuan et al., 2011), providing a clear advantage over wild-type hESCs. The potential to replace the CD24 surface marker with intrinsic fluorescent reporter activity further provides a means to monitor the NSC cultures and track terminal differentiation. Our lineage-specific knock-in approach will enable further studies tracing the commitment of neural progenitors derived from hESCs; to perform population based studies; and to improve current protocols for directed differentiation of cell types needed for cell replacement therapy.

Construction of BAC-based targeting vector
The Nestin BAC clone was purchased from Life Technologies Inc. and the targeting vector was constructed by recombineering as we previously described (Song et al., 2010).
(MEFs) feeder layer in knockout medium supplemented with 10% knockout serum replacement, 10% plasmanate, 1 mmol/L glutamine, 0.1 mmol/L nonessential amino acids, 10 ng/mL bFGF, and 0.1 μmol/L β-mercaptoethanol. To passage hESCs, confluent culture was washed with phosphate buffered saline (PBS), dissociated for 5 min with TrypLE and resuspended into single cells with culture medium. All tissue culture reagents were purchased from Life Technologies Inc. except where indicated otherwise.

Teratoma formation of hESCs in SCID mice
Two million hESCs were suspended in the knockout medium supplemented with 30% matrigel (BD Biosciences) and injected subcutaneously into severe combined immunodeficiency (SCID) mice. After six to eight weeks, teratomas were recovered. For histological assessment, teratomas were fixed in 10% buffered formalin, embedded in paraffin, sectioned and stained with hematoxylin and eosin. To derive NPCs from teratomas, teratomas were chopped into small pieces and cultured in NPC medium. After several passages, the adherent culture became homogenous with mCherry positive cells.