The Current State of Naïve Human Pluripotency

Abstract Naïve or ground state pluripotency is a cellular state in vitro which resembles cells of the preimplantation epiblast in vivo. This state was first observed in mouse embryonic stem cells and is characterized by high rates of proliferation, the ability to differentiate widely, and global hypomethylation. Human pluripotent stem cells (hPSCs) correspond to a later or “primed” stage of embryonic development. The conversion of hPSCs to a naïve state is desirable as their features should facilitate techniques such as gene editing and more efficient differentiation. Here we review protocols which now allow derivation of naïve human pluripotent stem cells by transgene expression or the use of media formulations containing inhibitors and growth factors and correlate this with pathways involved. Maintenance of these ground state cells is possible using a combination of basic fibroblast growth factor and human leukemia inhibitory factor together with dual inhibition of glycogen synthase kinase 3 beta, and mitogen‐activated protein kinase kinase (MEK). Close similarity between the ground state hPSC and the in vivo preimplantation epiblast have been shown both by demonstrating similar upregulation of endogenous retroviruses and correlation of global RNA‐seq data. This suggests that the human naïve state is not an in vitro artifact. Stem Cells 2015;33:3181–3186


INTRODUCTION
In mice, two pluripotent states have been captured in vitro. Mouse embryonic stem cells (mESCs) are sourced from the inner cell mass (ICM) of the preimplantation blastocyst [1,2]. When derived and maintained using a combination of leukemia inhibitory factor (LIF) and 2i (dual inhibition of extracellular signal-regulated protein kinases 1/2 [ERK1/2] pathway and glycogen synthase kinase 3 beta [GSK3b]) they are described as being in a na€ ıve or ground state [3]. When injected back into an early embryo, these cells can contribute to all lineages without tumorigenesis [4]. A more recent discovery has been mouse epiblast stem cells (mEpiSCs- Fig. 1). These are sourced from postimplantation epiblast cells [5,6] and are termed primed, due to their inability to integrate into a preimplantation blastocyst. They can, however, be differentiated into all three germ layers in vitro.
The most striking difference is the very high expression of de novo methyltransferases, which leads to condensing of chromatin [7]. Additionally, these cells require basic fibroblast growth factor (bFGF also known as FGF2) and transforming growth factor beta (TGFb) for selfrenewal, instead of 2i and LIF [3,5]. mEpiSCs can be converted back to the na€ ıve state by transfection with Klf4 or other reprogramming factors or using small molecules [8,9].
Na€ ıve pluripotent stem cells have been successfully captured in vitro from primed rhesus monkey induced pluripotent stem cell (iPSC) lines using specialized media containing 2i and LIF [10]. Since na€ ıve pluripotent stem cells can be generated from primates, this suggests that the state of naivety might be conserved across species. Using primate cells also allows dissection of genetic background and species to species differences. Primate na€ ıve iPSCs require bFGF, whereas bFGF causes differentiation in mESCs. Additionally, TGFb is not required for maintenance of primate na€ ıve iPSCs, indicating that TGFb might not be essential in the human system [10].
Embryogenesis is inherently different between species, which is reflected by the difficulties in generating truly na€ ıve human pluripotent stem cells (hPSCs) in vitro. For ethical reasons, information on human embryogenesis is lacking and many assumptions are made based on the mouse model [11]. Despite being sourced from the same point in development as mESCs, hESCs resemble mEpiSCs. Both form large, flat, 2D colonies and require bFGF for self-renewal. The ability to convert mEpiSCs to mESCs has led to the prediction that na€ ıve hPSCs might also be accessible by reverting primed hESCs. This has prompted several recent publications of strategies to capture the human na€ ıve state, either relying on transgene overexpression [12][13][14] or different combinations of small molecule inhibitors [15][16][17][18][19][20]. Here we review and compare all these published protocols, including a protocol devised by Duggal et al. [16] published in this issue.
KEY CHARACTERISTICS OF THE NA ..

ÏVE STATE
A key difference between na€ ıve and primed cells lie in their differentiation potential. For assessing human cells, Gafni et al. [15] used chimera assays, where human na€ ıve or primed cells are injected into mouse morulas. Unlike the primed cells, the progeny of the na€ ıve cells were subsequently detected in all tissues [15]. However, Theunissen et al. [19] found this method unreproducible, since no human cells derived from na€ ıve stem cells were detected when performing the assay in their laboratory, despite using cells generated by Gafni et al. as a control [19]. A less rigorous but widely used assay measures teratoma formation following injection of PSCs in immunocompromised mice and assessment of presence of mesoderm, endoderm, and ectoderm lineages. Na€ ıve and primed human pluripotent cells form mature, high grade teratomas [14,15,[18][19][20], with one study suggesting that na€ ıve cells form teratomas of increased volume in a shorter time in comparison to primed cells [17]. Na€ ıve PSCs, like primed PSCs, can readily form embryoid bodies containing cells of all three germline lineages [14][15][16][17][18]. Directed differentiation protocols have also been performed [14,16,18,19]. Most notably, Duggal et al. [16] show improved efficiency and homogeneity of directed differentiation toward neuronal, mesodermal, and endodermal lineages in comparison to primed cells.
Respiration is different between the two cell types: primed cells are almost entirely glycolytic, whereas metabolism in na€ ıve cells uses greater mitochondrial respiration [14,20]. This shift is also observed in vivo. Before implantation of the blastocyst in mouse, cells rely on oxidative phosphorylation [21,22], whereas after implantation a shift toward glycolytic metabolism occurs [23]. Increasing evidence (reviewed in [24]) is emerging that the regulation of energy metabolism is connected with epigenetic modifying machinery, which is also involved in progression from the na€ ıve state.
Na€ ıve cells show higher survival of single cell passaging in comparison to their primed counterparts [15]. They also differ in their doubling time of approximately 16 hours instead of 36 in hESC [5]. Differences in morphology are also widely reported [14-17, 19, 20]. In their na€ ıve stage, hESCs and mESCs form rounded 3D colonies, whereas primed cells grow in flat monolayers. This may play a role in diffusion of signaling molecules and cell-cell adhesion pathways.
A key observation is the difference in enhancer landscape between na€ ıve and primed cells. Globally, more enhancers are active during the na€ ıve state, whereas inactive enhancer complexes in similar positions are observed in primed cells. These are termed "seed enhancers" and seem to prepare for larger enhancer complexes and reorganization [25,26]. Gafni et al. [15] and Theunissen et al. [19] took advantage of the differential use of enhancers of the OCT4 gene as a way to assay for optimal na€ ıve maintenance conditions. Although both enhancers activated the gene to the same extent, the proximal enhancer is mainly active in primed cells, whereas the distal enhancer is used in the na€ ıve state [15,19].
Gene expression is different between na€ ıve and primed cells. These changes have been reported on a global scale [14-17, 19, 20]. Transcript levels have been shown to correlate between na€ ıve hPSC and mESC [14,19], and this similarity has been used to assess naivety, albeit being cross-species comparison [19]. For a within-species comparison, Wang et al. analyzed available RNA-seq data from cells taken from the ICM of early human embryos and compared these expression patterns to na€ ıve cells generated in vitro [27]. The authors argue that this comparison to human in vivo data is more relevant than comparisons to mouse, especially since they discovered a primate-specific transcript, human endogenous retrovirus subfamily H (HERVH), as a key component of naivety. They disrupted either HERVH or its binding partner LBP9 which showed that HERVH is essential for self-renewal in na€ ıve PSCs [27]. Expression of endogenous retroviruses were confirmed by Grow et al. [28], who report high expression of HERVK in preimplantation epiblast cells and in the na€ ıve cell line Elf1 generated by Ware et al. [20] and also in cells converted to the na€ ıve state using the protocol by Chan et al. [17].
RNA methylation has been shown to play a role in the ability to maintain and exit ground state pluripotency. Two recent publications by Batista et al. and Geula et al. [29,30] came to the same conclusion: Knockout of the N 6 -methyladenosine (m 6 A) transferase METTL3 causes reduced m 6 A RNA methylation and failure to resolve the na€ ıve state. These cells have been referred to as being hyperpluripotent due to their inability to differentiate. Transcripts marked with m 6 A decay faster and therefore allow the cells to make changes and differentiate [29,30]. However, these results contradict an earlier publication by Wang et al. [31] who reported that m 6 A might be required for maintenance of the ground state. Contrary to the two more recent publications, their cells with knockdown of METTL3 and METTL14 were unable to maintain pluripotency and differentiated [31] [32]. Na€ ıve cells in both species have been shown to be hypomethylated. This is particularly evident in female cells, as both X chromosomes are still active in the early embryo, which is also reported in na€ ıve pluripotent stem cells. X-inactivation by heterochromatin formation is observed in primed cells and thus can be a marker to distinguish between both states [14,15,18,20].

GENE EDITING EFFICIENCY
The efficiency of homologous recombination is significantly higher in mESCs in comparison to hESCs [33]. This has led to the hypothesis that the na€ ıve state might be more amenable to gene editing. Buecker et al. [34] generated human na€ ıvelike cells by transgene expression and measured random insertion of 10-20 kb cassettes containing a fluorescent marker and drug resistance gene. They showed a 200-fold increase of insertion frequency in their na€ ıve-like cells. The authors also targeted hypoxanthine-guanine phosphoribosyltransferase with a puromycin selection cassette with 4-4.5 kb homology arms and reported correctly targeted insertion rates as high as 1% but did not compare this to primed cells [34].
A different approach was taken by Gafni et al. [15], who measured rates of correct insertion in the two endogenous loci COL1A and OCT4 by using a puromycin selection cassette with homology arms of 2.1-2.5 kb and 4-4.5 kb in length, respectively. They showed relatively high correctly targeted integration rates of 11%-14.5% in na€ ıve cells, whereas integration in primed cells was low (0%-0.3%) [15]. Both groups show high rates of homologous recombination in na€ ıve cells using standard electroporation techniques of dsDNA plasmid template and without the need for nucleases-this is an advantage as site-specific nucleases including CRISPR-cas9 have been shown to exhibit off-target effects, which can only be ruled out after whole genome sequencing [35]. A reason for the difference between primed and na€ ıve editing efficiencies may be due to chromatin accessibility which has been shown to affect gene editing [36]-the more open chromatin state in na€ ıve cells might facilitate targeting. However, conclusive evidence for this is lacking. Moreover, gene editing in primed cells is technically challenging due to the requirement of clonal steps. Increased single cell survival of na€ ıve cells, together with higher rates of proliferation, facilitates genetic manipulations that require cloning steps [34].
Takashima et al. [14] overexpressed KLF2 and NANOG which are key transcription regulators for the acquisition of the na€ ıve state. This rewiring of the pluripotency circuitry, together with their media formulation, leads to stable selfrenewing na€ ıve pluripotent cells even after silencing of transgene expression [14]. Five other recent publications report achieving na€ ıve pluripotency without using any transgenes, by using different media compositions containing small molecule inhibitors and growth factors (Supporting Information Table  S1) [15,[17][18][19][20]. All available protocols rely on inhibition of MEK and GSK3b and on addition of bFGF, which represses differentiation. Most protocols also continuously add hLIF, with the exception of Ware et al. [20]. Other components either cause demethylation, repress differentiation or are inhibitors targeting MAPK pathways (summarized in Fig. 2). Some teams have reported that low oxygen level aids conversion [14,19,20], whereas others have reported no benefit of lowered oxygen [15,17,18] (Supporting Information Table S1).
Different strategies (Supporting Information Fig. S3) have been used for na€ ıve derivation, so the resulting cells have different properties. Evidence supporting naivety of these cells is summarized in Supporting Information Table S2. Chan et al. [17] were able to generate cells without transgene expression but did not bring forward as much evidence as most other protocols. Gafni et al. were able to demonstrate differentiation ability of their na€ ıve cells by generating cross-species chimeric mouse embryos containing differentiated cells derived from the human na€ ıve cells in several different tissues. However, they did not perform in vitro differentiation [15]. Ground state cells generated by Valamehr et al. [18] did not exhibit na€ ıve morphology and their protocol requires single cell cloning, however this workflow has been designed for high throughput generation of homogenous cells which share properties of naivety, and therefore has its own applications. The protocol by Ware et al. requires reverse toggling with HDAC inhibitors and is not particularly efficient. However, Ware et al. [20] have generated a stable na€ ıve cell line Elf1, which is banked and available. Takashima et al. bring forward comprehensive evidence for the naivety of their cells, including evidence for a switch to mitochondrial respiration. However, this protocol requires transgene delivery and therefore is less practical [14]. Theunissen et al. [19] also show that their cells are na€ ıve, however their protocol can induce karyotypic abnormalities and their na€ ıve female cells undergo X inactivation, indicating a later stage in development. The most recent publication (Duggal et al.) includes a demonstration of enhanced directed differentiation in comparison to their primed parental cells [16]. Reproduction of these protocols by other laboratories will establish how robust they are.

CONCLUSIONS
The concept of na€ ıve hPSCs has been contentious. Pera [41] argues that since this state was actively searched for in humans, it is highly likely that it is purely an artifact generated in the lab. However, Wang et al. used RNA-seq data which was available from cells taken directly from the ICM of early embryos and showed a tight correlation to na€ ıve cells generated in vitro [27]. This was confirmed when Huang, Maruyama, and Fan took a systems biology approach and compared datasets from many previous publications [42]. Their analysis revealed poor conservation of gene networks between mPSCs and hPSCs but a high resemblance to the ICM of their respective blastocysts. They also found variations in transcriptomes from different na€ ıve conversion protocols, but all established na€ ıve cells  [17], whereas all other protocols use CHIR99021. KLF2 and NANOG are overexpressed in the protocol developed by Takashima et al. [14], which is depicted by a halo around these transcription factors. TGFb is used as a supplement by Gafni et al. [15] and is present in the basal media of the protocol devised by Chan et al. [17]. The PKC inhibitor G€ o6983 is optional in the protocol by Gafni et al. [15]. Abbreviations used: BIRB, BIRB796; CHIR, CHIR99021; FGF4, fibroblast growth factor; Go, G€ o6983; GSK3, glycogen synthase kinase 3; HDAC, histone deacetylase; hLIF, human leukemia inhibitory factor; PD17, PD173074; PD03, PD0325901; SB59, SB590885; SP6, SP600125; TGFb, transforming growth factor beta [14, 16-20, 32, 37-39]. showed clear resemblance to human late preimplantation embryos. According to this study, na€ ıve cells generated by Takashima et al. [14] and Theunissen et al. [19] most closely resembled the human preimplantation blastocyst. The protocols by Valamehr et al. [18] and Duggal et al. [16] were not included in the study. In conclusion, the authors propose comparing the combination of transcriptome analysis and epigenetic characterization to in vivo data from embryogenesis as a gold standard for naivety [42].
The description of just two states, na€ ıve and primed, is an oversimplification [11,27,43,44]. Two studies [27,43] used single-cell RNA-seq and reported a polyclonal spectrum of cell states ranging between these extremes and that na€ ıve PSCs are present as a subpopulation in cultures previously considered entirely primed. Wang et al. [27] used a reporter system based on the endogenous retrovirus HERVH's LTR7 promoter which is only active in na€ ıve cells. This approach showed a consistent 4% of cells with na€ ıve reporter expression which can be selected for using 2i and LIF and do not need prior conversion. Recently, Wu et al. were able to capture another alternative state designated "region-selective primed" pluripotency in vitro in both mouse and human which are distinct from both na€ ıve and primed states [44].
There remain many challenges in the field of na€ ıve pluripotency. All protocols for generating human na€ ıve PSCs yield slightly different cellular states. It is still unclear which of these is closest to its in vivo counterpart. The in vivo na€ ıve state is inherently transient, so continuous in vitro culture may be detrimental. For example, female cells maintained in the na€ ıve state that do not exhibit X-inactivation might suffer from double dosage effects. With protocols now readily available which allow the generation and maintenance of na€ ıve cells, these questions can be addressed. Meanwhile, their faster rate of growth, single cell survival, and enhanced gene editing efficiency will be used. In the near future, na€ ıve hPSCs may be useful for accessing paths of differentiation which have been previously unreachable.

ACKNOWLEDGMENTS
We thank Dr. Paul Fairchild for proof-reading, helpful discussions, and feedback. B.T.D. is funded by BBSRC industrial case DPhil training grant BB/L015447/1. S.C. receives financial support from the Wellcome Trust (WTISSF121302) and the Oxford Martin School (LC0910-004) and the Monument Trust Discovery Award from Parkinson's UK. R.F. is funded by EU IMI (StemBANCC), who provide the following statement: The research leading to these results has received support from the Innovative Medicines Initiative Joint Undertaking under grant agreement no 115439, resources of which are composed of financial contribution from the European Union's Seventh Framework Programme (FP7/2007-2013) and EFPIA companies' in kind contribution. This publication reflects only the author's views and neither the IMI JU www.imi.europa.eu nor EFPIA nor the European Commission are liable for any use that may be made of the information contained therein.

AUTHOR CONTRIBUTIONS
B.T.D.: conception and design, collection and/or assembly of data, and manuscript writing; R.F.: conception and design, manuscript writing, and financial support; S.A.C.: conception and design, manuscript writing, financial support, administrative support, and final approval of manuscript.

DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
Research funding includes grants from BBSRC industrial case DPhil training grant BB/L015447/1 with industrial partner F. Hoffmann-La Roche AG (B.T.D.) and EU IMI STEMBANCC grant number 115439 (R.F., S.A.C.).