Derivation of Naïve Human Embryonic Stem Cells

Derivation of naïve human embryonic stem cells

Carol B. Warea,b,1, Angelique M. Nelsona,b, Brigham Mechamc, Jennifer Hessona,b, Wenyu Zhoua,d, Erica C. Jonlina,e, Antonio J. Jimenez-Caliania,f, Xinxian Dengg, Christopher Cavanaugha,b, Savannah Cooka,b, Paul J. Tesarh, Jeffrey Okadaa,e, Lilyana Margarethaa,e, Henrik Sperbera,i, Michael Choia,i, C. Anthony Blaua,e, Piper M. Treutingb, R. David Hawkinsa,j, Vincenzo Cirullia,f, and Hannele Ruohola-Bakera,d,i

aInstitute for Stem Cell and Regenerative Medicine, Departments of bComparative Medicine, dBiology, and iBiochemistry, and Divisions of eHematology, fMetabolism, Endocrinology and Nutrition, and jMedical Genetics, Department of Medicine, University of Washington, Seattle, WA 98195; cSage Bionetworks, Seattle, WA 98109; gDepartment of Pathology, University of Washington, Seattle, WA 98195; and hDepartment of Genetics, Case Western Reserve University, Cleveland, OH 44106

Edited* by Stanley M. Gartler, University of Washington, Seattle, WA, and approved February 12, 2014 (received for review October 24, 2013) The naïve pluripotent state has been shown in mice to lead to of a naïve hESC line, Elf1. We report on the naïve state of hu- broad and more robust developmental potential relative to primed man ESCs capable of unlimited culture in 2i. mouse epiblast cells. The human naïve ES cell state has eluded derivation without the use of transgenes, and forced expression Results of OCT4, KLF4, and KLF2 allows maintenance of human cells in Primed Mouse and Human Pluripotent Cells Acclimate to Naïve 2i a naïve state [Hanna J, et al. (2010) Proc Natl Acad Sci USA 107 Culture. In an earlier report, we described toggling mouse and (20):9222–9227]. We describe two routes to generate nontrans- human ESCs to an earlier state using a histone deacetylase in- genic naïve human ES cells (hESCs). The first is by reverse toggling hibitor (HDACi) (7). Thus, we used this approach as a stepping of preexisting primed hESC lines by preculture in the histone stone to reverse primed cells to a naïve state. mEpiSCs and deacetylase inhibitors butyrate and suberoylanilide hydroxamic hESCs were converted to 2i culture following one to three pas- acid, followed by culture in MEK/ERK and GSK3 inhibitors (2i) with sages in a mixture of HDACi consisting of sodium butyrate and FGF2. The second route is by direct derivation from a human em- suberoylanilide hydroxamic acid (SAHA, vorinostat) (B/S); no- bryo in 2i with FGF2. We show that human naïve cells meet mouse Methods

menclature is explained in ). Without preexposure to BIOLOGY criteria for the naïve state by growth characteristics, antibody la- B/S, the majority of the primed stage mouse and human cells beling profile, gene expression, X-inactivation profile, mitochon- differentiated following exposure to 2i (Fig. 1 A and B and Fig. DEVELOPMENTAL drial morphology, microRNA profile and development in the S1 B–G), as expected (3). Once cells were passaged through B/S, context of teratomas. hESCs can exist in a naïve state without both mouse and human responded to 2i through formation of the need for transgenes. Direct derivation is an elusive, but attain- tight colony edges reflective of mESCs (Fig. 1B and Fig. S1B). able, process, leading to cells at the earliest stage of in vitro plu- These human cultures remained dependent on basic fibroblast ripotency described for humans. Reverse toggling of primed cells growth factor (FGF2; F). All stages of pluripotency reflected to naïve is efficient and reproducible. under these culture conditions express POU class 5 homeobox 1 (OCT4), Nanog homeobox (NANOG), and the cell surface t has become clear with the derivation of mouse epiblast stem proteins TRA-1-60 and TRA-1-81. Unlike naïve mESCs, both Icells (mEpiSCs) that pluripotency encompasses more than one naïve and primed hESCs are stage-specific embryonic antigen-4 stage of development (1, 2). The earlier “naïve” stage represents (SSEA4)-positive while remaining SSEA1-negative (Fig. S2A). the preimplantation inner cell mass, typified by mouse ES cells Primed H1, intermediate H1 (in B/S), and naïve H1 (in 2iF) (mESCs), and the “primed,” the postimplantation epiblast, typified by mEpiSCs and human ES cells (hESCs). The challenge in Significance naïve cell maintenance has been protecting cells from differentiation stimuli. This has been achieved in mESCs through ex- We report on generation of nontransgenic, naïve human plu- posure to leukemia inhibitory factor (LIF), whereas addition of ripotent cells that represent the developmentally earliest state extracellular signal-regulated kinase (MEK) and glycogen syn- described for human established cells. Existing human ES cell thase kinase 3 (GSK3) inhibitors (2i) in defined medium allows lines in the later primed state can be toggled in reverse to naïve the cells to attain a homogeneous ground state (3). Defining by exposure to histone deacetylase inhibitors prior to naïve characteristics of the naïve/ground vs. primed states are shown in A culture. A new line was established directly from an eight-cell Fig. S1 . In humans, the naïve stage has been difficult to capture embryo under naïve culture conditions. We describe the naïve as a stable in vitro state. state in humans and show that naïve human ES cells have There are practical advantages that come with a human naïve expanded endoderm developmental capacity. state. Among them is ease of trypsin passage and developmental capacity. Whole animals can be generated from good naïve mESCs Author contributions: C.B.W., W.Z., C.A.B., R.D.H., and H.R.-B. designed research; C.B.W., through tetraploid complementation (4), and mEpiSCs cannot A.M.N., J.H., W.Z., A.J.J.-C., X.D., C.C., S.C., J.O., H.S., M.C., R.D.H., and V.C. performed contribute to chimerism. Being more comparable to mESCs, naïve research; E.C.J. and P.J.T. contributed new reagents/analytic tools; C.B.W. and A.M.N. ’ hESCs will likely allow increased developmental potential and obtained consent for embryo donation; E.C.J. obtained human subjects approval/consent; C.B.W., B.M., W.Z., J.O., L.M., H.S., P.M.T., R.D.H., and V.C. analyzed data; and a more accurate correlation to mESC data. C.B.W. wrote the paper. It has been reported that human induced pluripotent cells The authors declare no conflict of interest. (h-iPSCs) can be maintained in the naïve state if the pluri- *This Direct Submission article had a prearranged editor. potency-inducing transgenes are not silenced (5). Only recently have hESCs been maintained in a naïve state without transgenes Freely available online through the PNAS open access option. (6). Our primary aim was to generate naïve hESCs not de- Data deposition: The Agilent array data reported in this paper has been deposited in the Synapse Commons Repository database, https://synapse.sagebase.org/ [accession nos. pendent upon transgenes for stable culture. We toggled existing syn1447097 (H1-2iF vs. H1), syn1447098 (Elf1-3iL vs. Elf1-AF), and syn1447101 (mESC vs. human ESC and mouse mEpiSC lines back from the primed mEpiSC vs. mEpiSC-2iL)]. state to grow under the influence of 2i without the need for 1To whom correspondence should be addressed. E-mail: [email protected]. Activin A. This helped us to define appropriate culture con- This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. ditions for human naïve cells and allowed the de novo derivation 1073/pnas.1319738111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1319738111 PNAS Early Edition | 1of6 Downloaded by guest on September 26, 2021 would represent altered states as determined by expression. The difference between right and left quadrants would be more profound than the difference between top and bottom. Although the toggled naïve mESCs [mEpiSC-(B/S)2iL] have the morpho- logical appearance of mESCs (Fig. S1B), these cells are more similar to delayed-implantation icm (Fig. 2B). Thus, there is likely an advantage to direct naïve cell derivation. Therefore, we generated a new hESC line from a donated embryo under the naïve culture conditions determined by H1-2iF cells.

Generation of a New Naïve hESC Line. One eight-cell embryo formed what seemed to be a naïve hESC line, with mounded colony morphology having bright edges (Fig. S1I). The efficiency

Fig. 1. Effect of 2i culture on mouse and human pluripotent cells. (A)Al- kaline phosphatase stain of mouse pluripotent colonies. Left two plates: 2i added to mEpiSC colonies causes differentiation and loss of the alkaline phosphatase positive cells. Right four plates: mEpiSCs grown in butyrate plus SAHA for a minimum of 1 passage before addition of 2i allows pluripotent colonies to flourish. (B) hESC (H1) toggled backward on Matrigel to 2i through butyrate or forward to differentiation when not exposed to B/S before 2i culture, indicating that 2i must follow B/S exposure to prevent differentiation. (Scale bars, 100 μM.)

cultures grown on Matrigel were karyotyped by G-banding at high passage number (passage 93). H1 grown in B/S (H1p93B/ S45) and grown in B/S followed by 2iF [H1p93(B/S17)2iF27] displayed a normal 46XY karyotype (5/5 spreads analyzed each). The companion cells [H1p93(M12)T19], grown in the primed state throughout, gained genetic material with all cells carrying trisomy 12 and trisomy 17 (5/5 spreads) and some of those carrying an additional marker chromosome of unknown origin (3/5). To explore the biological effect of LIF, naïve H1s were exposed to human LIF or FGF2. Cells that were supported on feeders with LIF exhibited a naïve mESC-like morphology by mounding, whereas colonies were flatter in FGF2 (Fig. S1 C–F). Despite the naïve mESC morphology following one passage, H1-2iL began showing signs of differentiation by the third passage and were lost to differentiation by the third to fifth passage (Fig. S1G). Along with five hESC lines (H1, H7, H9, HUES1, and HUES2), two mEpiSC lines (mEpiSC5 and mEpiSC7, 2) and two human induced pluripotent lines were converted to 2i culture. In addition, one nonhuman primate line (MF-1; Macaca fascicularis;agiftof Eliza Curnow, Washington National Primate Research Center, Seattle) was converted to 2i culture (Fig. S1H). Unlike hESCs, the nonhuman primate line could survive over several passages in LIF Fig. 2. Genomic analysis of naïve hESCs. (A) RNA expression heat map of HIF2α (EPAS1) target genes in H1-2iF cells relative to parent H1 [H1p58 with no FGF2 and did not require HDACi to reverse toggle, in- M. fascicularis (B/ST2)2iF14 vs. H1p58 in TeSR2; run in quadruplicate]. (B) Principal compo- dicating a species difference between humans and . nent analysis comparison of mouse whole genome Agilent array data comparing Hunter et al. (8) embryo data (Left) to mESC equivalents: R1p22 Gene Expression of H1 Cultured As Naïve Indicates an Earlier Pattern. (mESC-2iL, naïve), mEpiSC7p24AF (mEpiSC-AF, primed), and mEpiSC7p55 We analyzed the differences between the stages of human plu- (AF7,B/S1)2iL20 (mEpiSC-2iL, toggled to naïve) (Right). Elf1 naïve (3iL, green ripotency, primed and toggled naïve through Agilent whole hu- squares) and primed (AF, blue squares) expression data are compared with man genome array comparisons. Hypoxia inducible factor 2α the in vivo mouse embryo data in the plot on the left. (C) Comparison of in- (HIF2α) (EPAS1) seems to be diagnostic of the 2iF stage in that house Elf1 expression array data as the standard against which to measure × × a significant number of HIF2α targets show a unique expression in-house (UW) data (H1-2iF four repetitions; primed: H1 four repetitions) and that generated by Hanna et al. (5). Naïve: C1.1, C1.2, WIBR3.1, WIBR3.3, pattern in naïve H1 cultures relative to primed (Fig. 2A; hyper- P < WIBR3.5; primed: first grouping-BG01, BG01-mTeSR; BG01-NANOGtgk; sec- geometric distribution, 0.01). ond grouping-WIBR1, WIBR2, WIBR3rep1, WIBR3rep2, and WIBR3–5%O2. To determine whether cells that were toggled through B/S to Note that the lines tested on left side that are compared with naïve Elf1 are 2i could be defined as “naïve” we used principal component grouped identically on the Elf1 primed side of the graph. Presumed naïve analysis. Microarray expression data derived from the mEpiSCs, cell lines are represented by dark blue dots and presumed primed are or- mEpiSCs converted to 2iL, and mESCs were compared with ange. (D) DNase I hypersensitivity analysis of the POU5F1 enhancer regions a mouse developmental yardstick (8) (GSE8881), in which inner for Elf1 (lower line black) and H1 (line above in blue). The first exon of POU5F1 is shown above the H1 data along with a 2-kb size bar to map the cell mass (icm) was collected from mouse embryos at pre- proximal enhancer (PE) and distal enhancer (DE). (E) ChIP-seq H3K27me3 implantation (E3.7; thought to be equivalent to naïve mESC), comparison of primed hESCs [orange line, data taken from Gafni et al. (6)] to delayed implantation icm of diapause (E7.5, i.e., 3-d delay), and naïve Elf1-2iL (blue line) using the genes in C that intersect with Gene On- postimplantation epiblast (E6.5). Clustering within a quadrant togeny “developmental genes” (n = 648).

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1319738111 Ware et al. Downloaded by guest on September 26, 2021 of establishing a human pluripotent line in 2i was quite low at one line from 128 donated embryos, primarily frozen as blastocysts. Elf1 is positive for OCT4, NANOG, TRA-1-60, TRA-1-81, and SSEA4 while remaining negative for SSEA1 (Fig. S2A), consistent with recent reports (5, 6). The line was karyotyped at passage 9 and found to be a normal female (Fig. S2B). Although Elf1 was started in 2iF, we attempted to convert it to 3i (includes an FGF receptor inhibitor) + human LIF, to determine whether Elf1 could grow independently of FGF2. Elf1- 3iL resulted in a more mounded morphology characteristic of mESC (Fig. S1I). Unlike H1-2iF, Elf1-3iL could be stabilized without FGF2. Elf1 could be maintained as naïve indefinitely in 3iL and 2iL and as such has been grown stably for more than 60 passages. However, when FGF is not present there is an increased background of differentiated cells. There was negligible culture drift when maintained in 2iF. Two questions concerning culture arose. What is the operative signaling mechanism of FGF2 in the presence of an MEK inhibitor? Mammalian target of rapamycin (mTOR) survival signaling is a possible alternative pathway used by FGF2. Both naïve and primed Elf1 were sensitive to mTOR inhibition through exposure to rapamycin (Fig. S3A). The second question concerns the role of the Janus kinase/signal transducers and activators of transcription (JAK/STAT) pathway in naïve Elf1 cells. A STAT3 inhibitor (NSC74859) was added to the 2iL medium, which caused naïve cells to slow growth, whereas Elf1s

induced to the primed stage through Activin and FGF (AF) were Fig. 3. Analysis of hESC stages of pluripotency. (A) MicroRNA analysis as- BIOLOGY B

not obviously affected (Fig. S3 ). sociated with pluripotency. (B) XIST labeling shows that Elf1-3iLs (Left)do DEVELOPMENTAL not display a XIST cloud by FISH, whereas Elf1s primed have two XIST signals Elf1 Gene Expression Indicates the Line Is Naïve. Using the Sage and cells differentiated for 10 d (Right) gain a single XIST signal (red dot) resources, Agilent expression data were compared with archived within the nucleus. When the nucleus is highlighted using DAPI and the field array data. A set of 628 naïve Elf1 specific enriched (fourfold or magnified XIST remains undetectable in naïve Elf1 (Lower Left), whereas the greater) genes and of 639 primed Elf1 specific enriched genes XIST signal can be detected upon differentiation on one or both X chro- mosomes (red dots, white arrows, Lower Right two panels). (C) Bisulfite were used as a yardstick against which to compare our data with sequencing of the XIST promoter using the 11L-11R primers shows that XIST that in the literature. Naïve Elf1 expression, but not primed Elf1, remains methylated throughout the naïve and primed stages. Using the was found to correlate with previously reported naïve h-iPSC 7L-7R primers, methylation seems to diminish in naïve relative to primed expression using persistent expression of transgenes (5) (Fig. cells, 10-d in vitro-differentiated cells and in the 98-d teratoma. Note that 2C). All of the primed hESC lines showed an expression signa- the 1, 9, and 11 primer sets did not vary from data shown for set 11 upon ture that was clearly different from the naïve signature. Fur- differentiation. Open circles indicate unmethylated and filled circles indicate thermore, cell lines treated as primed showed the most variability methylated CpGs. (D) Graphs representing cloning efficiency (percent) and in expression pattern. This could partially explain some of the doubling times (hours) of Elf1 naive (green), Elf1 primed (yellow), H1 naïve (dark biological variability seen among primed hESCs. blue), and H1 primed (light blue). (E) Electron microscopy of mitochondria. The left panels show the mitochondrial shape difference between Elf1-3iL and Elf1- The Elf1 array data were compared with the same previously B AF. This is quantified in the graph on the right, where increased ratio indicates generated mouse embryo data used in Fig. 2 . The naïve Elf1 a rounder mitochondrial population (± SEM). ***P < 0.001, **P < 0.01, and *P < cells did fall into the same quadrant with mouse peri-implanta- 0.05 as determined by t test. tion icm as seen in delayed implantation embryo analysis whereas primed Elf1s reflected mEpiSCs (Fig. 2B). This is re- markable given the diversity of expression between mice and did remain reasonably stable in naïve and primed hESCs (Elf1, humans and the differences in experimental process, indicating H1, and H7), as expected. Expression of miRs 518b and 520b that Elf1s are within the mouse naïve definition by expression. and 520c increased in the naïve state relative to primed while Curiously, primed Elf1 cells could be switched back to naïve miR-372 expression increased in Elf1 relative to H1 and H7 conditions (2iL) without the need for a B/S intermediate step, (Fig. 3A). indicating a fundamental difference between hESCs derived To compare our ELF1 data to a recent report (6) on naïve under naïve conditions, but cultured as primed, and hESCs hESCs we looked at two parameters, POU5F1 (OCT4) enhancer generated initially as primed cells. use and presence of H3K27me3. Naïve cells depend upon the MicroRNA (miR) pattern change is implicated during initia- distal OCT4 enhancer, whereas primed cells use the proximal tion of development. For this reason, we decided to look at select enhancer (5, 6). We detected a similar phenomenon using miRs, miR-302b, miR-372, miR-518b, miR-520b, and miR520c. DNase I hypersensitivity analysis, which is useful for pinpointing These miRs are expressed in the primed state of human pluripotency (9, 10). miR-302b is thought to remain relatively stable poised and active regulatory regions. Increased hypersensitivity throughout human pluripotency and was used as a positive mapped to the distal enhancer in Elf1 and to the proximal en- D control for the pluripotent state. miR-372 may increase in hESC hancer in H1 (Fig. 2 ). Recently, naïve hESCs were shown to developmentally earlier than primed (10, 11). As part of a pri- exhibit reduced H3K27me3 at developmental gene promoters mate-specific chromosome 19 miR cluster (CM19C), miR-518b, (6). We generated H3K27me3 ChIP-seq data in Elf1 naïve con- miR-520b, and mir-520c are known to be expressed in the ditions and compared this to published H3K27me3 hESC-primed primed state and to decline upon differentiation (9). Thus, an levels (6) at 648 developmental genes used for comparison in Fig. increase in these miR levels relative to miR 302b would indicate 2C. We see an overall reduction of H3K27me3 signal in naïve cells − that the cells are earlier in pluripotent development. miR-302b (P < 2.2 × 10 16), in agreement with previous findings.

Ware et al. PNAS Early Edition | 3of6 Downloaded by guest on September 26, 2021 Elf1 Cells Have Two Active Xs with Appropriate X Inactivation on Differentiation. Elf1 is a female line, which allowed us to explore X inactivation. FISH analysis for X-inactive specific tran- script (XIST) indicated the absence of a signal in naïve Elf1. Two XIST clouds arose in primed cells, and two clouds were present in some of the cells upon random differentiation for 10 d in FBS, but there were also cells with only one cloud (Fig. 3B). Identi- fication of both Xs via XIST-FISH before X inactivation is consistent with a previous observation on human embryos (12). Thus, we looked at XIST promoter methylation as an indicator of an inactive X, using four primer sets (13). Sets 1L-1R, 9F-9R, and 11F-11R indicated similar extensive promoter methylation in the naïve, primed, and differentiated states. The promoter region detected by the 7L-7R set indicated demethylation of XIST in naïve cells relative to primed and differentiated states. Absence of X inactivation may, in part, be a feature of low oxygen culture, as previously described (13).

Naïve hESCs Double More Rapidly and Clone More Efficiently Than Primed hESCs. We measured both cloning efficiency and population doubling time in Elf1 naïve (2iL), Elf1 primed (AF), H1 naïve (2iF), and H1 primed (Fig. 3D). Elf1 naïve cloning efficiency was twice that of the other lines, and doubling time was halved relative to H1. Elf1 primed and H1 naïve characteristics are intermediate. The trend seen is similar to that reported (5, 6), although absolute numbers are not identical, likely owing to technical differences in the assays.

Mitochondrial Metabolism As a Functional Marker of Pluripotency Stage. mESCs are bivalent and can convert between mitochondrial and glycolytic metabolism depending on culture conditions, whereas mEpiSCs and hESCs are dependent primarily on aer- Fig. 4. Teratomas generated from naïve, naïve toggled to primed, and primed toggled to naïve cells reflect extensive endoderm developmental obic glycolysis (14). Consistent with the mouse data, expression c capacity. (A) H&E-labeled sections of Elf1p17-2iL10 (naïve; 42 d) and of cytochrome oxidase assembly protein and estrogen receptor- Elf1p15T8 (primed, 67 d) teratomas. (B) Endoderm-specific labeling of sec- related receptor B were significantly higher in the naïve Elf1 tions from the Elf1 teratomas shown in A. The upper two panels of both −4 cultures relative to primed Elf1 (P < 1 × 10 and P < 0.01, tumors are sequential sections, the upper labeled to highlight liver de- respectively; Agilent array data), signifying effective mitochon- velopment (red, albumin; green, α-fetoprotein; and blue, E-cadherin) and drial oxidative metabolism. Primed Elf1 increased RNA expres- the second set to highlight pancreatic development (red, PDX1; green, SOX9; sion of HIF1α targets lactate dehydrogenase A, phosphoinositide- and blue, E-cadherin). The next three panels (descending) for both tumors dependent kinase-1 (PDK1), and phosphorylase, glycogen, liver also represent different sequential sections with the first set representing P < −5 P < P < α liver development, the second set pancreatic development, and the third set (PYGL) ( 10 , 0.01, and 0.01, respectively). HIF1 is liver development by alternative markers (labeled as above and red, CYP3A known to play an instrumental role in support of aerobic glycolysis. and green, HNF4A). The lower Elf1p17-2iL10 naïve panel are in place to The dependence of the primed state on glycolysis likely confers reinforce the impression of the level of organization of endodermal de- a growth advantage that may be correlated to that seen in cancer velopment within these tumors (red, FOXA2; green, SOX9; and blue, cells through the Warburg effect. PDK1, 2, and 3 are all up-regu- E-cadherin), and the bottom right panel (Elf1p15T8) is a negative control. − latedinprimedElf1cells(P < 0.01, P < 0.01, and P < 10 5, re- (C) H&E sections of an H1 naïve, 44-d teratoma. The graph below the H&E spectively), indicating that some of the mechanism of primed cell sections represents quantitation of the areas stained for either E-cadherin mitochondrial shutdown may be similar to that seen in the (epithelial cells) or PDX1 (pancreatic progenitors) in primed (H1p44-AF9), Warburg effect (15, 16). naïve [H1p49(B/S3)2iF10], and naïve revertedtoprimed[H1p49(B/S3,2iF4) AF5] H1 generated teratomas, indicating that total epithelial developmental Mitochondrial morphology was assessed in naïve Elf1 and potential and the pancreatic subset are both enhanced in the naïve state primed Elf1 by electron microscopy. Reflective of the mouse relative to primed. (D) Top three panels are H&E sections of an mESC tera- naïve vs. primed cells (14), naïve Elf1 mitochondria appeared toma (naïve, 13 d). The lower panels show immunofluorescent labeling of less mature, in that they were generally round with few cristae, sections from this mESC teratoma. (Scale bars in A and D, 100 μM; they de- whereas mitochondria in primed cells were elongated with fine the scale for all H&E-stained sections.) somewhat more developed cristae (Fig. 3E).

Teratoma Formation. Teratoma formation has conventionally hESC-generated teratomas have noted the disproportionate been used in pluripotent cell assessment to determine the amount of mesoderm in the tumor mass (amounting to between ability of a line to generate the three germ lineages. Repre- two-thirds and three-fourths of the tumor) (17, 18). As shown in sentatives of the three germ lineages are present in teratomas Fig. 4B, teratomas derived from both naïve and primed Elf1s from toggled and de novo-derived naïve hESCs, proving their contained significant compartments occupied by E-cadherin– pluripotency (Fig. 4). This assay has the potential to highlight positive areas whose phenotype could be clearly ascribed to the lineage differentiation potential of any individual plurip- definitive endoderm cell lineages owing to the proximity to the otent line. It is in this regard that we pursued teratoma for- developing liver. Specific examples are shown for the liver cell mation to determine the endoderm lineage contribution. lineage, marked by the expression of albumin, α-fetoprotein, and Tumors were harvested from animals injected with naïve Elf1 cytochrome P450, family 3, subfamily A (CYP3A), whereas in cells within 42–50 d and primed Elf1 by 67 d, when they were adjacent sections pancreatic cell lineage marked by the expression first detectable by palpation. Studies by others with primed of pancreatic and duodenal homeobox 1 (PDX1) and SOX9 in

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1319738111 Ware et al. Downloaded by guest on September 26, 2021 E-cadherin–positive cells seems associated with, yet separate from, improve naïve hESC derivation rate. We note that the amount of liver development. Immunoreactivity for hepatocyte nuclear factor GSK3i used influenced culture stability such that cells self- 4, alpha (HNF4A), CYP3A, forkhead box A2 (FOXA2), and/or renewed in 0–2 μM GSK3i, whereas 5 μM drove differentiation. sex determining region Y, box 9 (SOX9) also identifies regions This varies from mESCs, where 3 μM GSK3i is optimal (3), but is occupied by more immature common progenitors for liver and similar to observations made with naïve rat ESCs, for which 1 μM pancreas lineages. This phenotype contrasts with that observed in GSK3i is optimal (22, 23). Excess GSK3i may enable Wnt-induced teratomas derived from primed H1 cells, in which clusters of im- nuclear events that are counter to the naïve state. mature endoderm were reduced relative to naïve H1. E-cadherin A key advantage of naïve hESCs is the predicted ability to immunoreactivity, indicating epithelial cells and PDX1 immuno- better reflect development. This expectation has been borne out reactivity, indicating pancreas progenitors, were both increased in in the teratoma assays. Generation of the endoderm lineage is the naïve H1 (Fig. 4C, graph). Reversion of naïve cells to primed via particularly robust from naïve cells. We interpret this to mean culture reduces both E-cadherin– and PDX1-positive areas. Lineage that some tissue restriction for teratoma formation occurs when marker data were combined for H1 and Elf1 naïve (n = 6) vs. primed hESCs are established in the primed state. A caveat to claiming (n = 4) teratomas (Fig. S4). Endoderm was increased in the naïve cell- that teratoma developmental potential reflects fetal development generated tumors explored relative to the two primed cell-generated arises out of a recent paper (24) showing that mouse cells derived tumors, whereas neural and mesenchymal lineages were reduced. as primed mEpiSCs can integrate correctly into fetal tissue if in- Because mESCs are the naïve gold standard, we assessed troduced into epiblast stage embryos. However, given the nature mESC teratomas. These were harvested 13 d following injection. of the postimplantation epiblast chimera assay that disallows mESC-generated tumors looked similar to hESC-generated reinitiation of pregnancy to term, it is difficult to assess the dif- tumors. Endoderm was well developed (Fig. 4D), although ference between the lineage contribution. Changes to teratoma immature, relative to day-42 to day-67 naïve human teratomas. tissue contribution do not necessarily reflect true embryonic de- Overall, the teratoma data indicate that robust endoderm developmental potential, but it is the best assay we currently have for velopment is attenuated when hESCs are derived under primed human pluripotent cells. conditions. Lineage potential seems to be protected throughout LIF-related receptor function is active in the naïve state, as in vitro pluripotent culture if hESCs are derived as naïve. Ro- determined by array analysis and STAT3 inhibition. Thus, up- bust lineage potential is regained when H1s are toggled back- regulation of LIFR and gp130 in the context of stem cell mor-

ward through HDACi to a naïve state. phology is likely a good indicator of the naïve state in the ab- BIOLOGY

sence of other markers. The presence of miR-302b indicates DEVELOPMENTAL Discussion pluripotency, whereas an increase in miR-518b, miR-520b, and We show that naïve human pluripotent cells can be derived directly miR-520c indicates a move to an earlier pluripotency stage. Mi- from the preimplantation human embryo and that hESCs derived tochondrial morphology is also a good indicator of a naïve line. beyond a point in the transition from naïve to primed maintain the Once an appropriate morphology is detected, a few relatively ability to become stably naïve in response to culture conditions. inexpensive tests of mitochondrial morphology and gene expres- This is supported by a recent report (6) that, in addition to 2iLF, sion can detect the naïve state. Elf1 cells can be used to develop an used increased MAP kinase pathway inhibition through p38 and in vitro differentiation assay to detect the extent of endoderm JNK inhibition along with transforming growth factor beta development as a naïve hallmark, and the naïve parameters above (TGF-β) addition to allow reverse toggling and maintenance of serve as indirect evidence of developmental potential. de novo-derived naïve hESCs. In our hands, reverse toggling of In summary, a stable, naïve hESC state does exist without the primed hESCs is facilitated by HDACi exposure in humans, need for transgenes and can be maintained indefinitely through whereas in mice naïve cells can arise directly from mEpiSC cultures trypsin passage. The cells seem karyotypically stable with hall- under naïve conditions (19). However, we see that preculture in marks consistent with a mouse naïve state, including gene ex- HDACi assists mEpiSCs in surviving initial exposure to 2i. pression, teratoma formation, and immature mitochondrial and The ability to maintain reverse-toggled hESCs in a naïve state overall morphology. Naïve hESCs are capable of robust differ- seems to hinge around the continued presence of FGF. Human entiation to all three germ lineages, with a heightened capability response to FGF2 does not completely phenocopy the mouse of endoderm formation not previously seen in hESCs. during the naïve-to-primed transition. Unlike in mice, human hypoblast formation does not require FGF2 (20), whereas FGF2 Methods response in the presence of MEK/ERK inhibitor indicates that Oversight. Animals. All procedures with mice followed the recommendations FGF is likely to have a positive effect on self-renewal in both the and approval of the Institutional Animal Care and Use Committee of the naïve and primed states, although perhaps through different sig- University of Washington and were performed within an American Associ- naling pathways. Active MEK signaling is important in the primed ation for the Advancement of Laboratory Animal Care-accredited specific state, whereas phosphoinositide 3-kinase/mTOR signaling or an- pathogen-free facility at the University of Washington. other FGF-induced cell signal suppresses differentiation in both Human subjects. Approval was received from the University of Washington Institutional Review Board to obtain consent for donation of excess cry- the naïve and primed states. In support of this, mouse primordial opreserved human embryos from fertility clinic patients to generate new germ cells require a 1-d exposure to FGF2 to toggle to pluri- human ESC lines. All embryo donations followed informed consent. potency, and forskolin can substitute for FGF2 in this context (21). Embryonic stem cell research oversight. All embryonic stem cell work was pre- Because human iPSCs can survive as naïve if cultured continuously approved by the University of Washington Embryonic Stem Cell Research in forskolin (5), this suggests that forskolin can, at least partially, Oversight Committee. substitute for continuous presence of FGF2 to stabilize hESC in the naïve state. We have been unable to generate a new hESC line Cell Line Nomenclature. directly in 2iL, and the one line generated in 2iF was thawed as an Cell line name “p” + number: total passage number eight-cell embryo. The presence of FGF2 and/or the stage of development at freeze/thaw may have been crucial to halting de- L: leukemia inhibitory factor velopment in the naïve state. The extremely poor direct derivation F: FGF2 rate of naïve hESCs from embryos may be due to culture con- A: Activin A ditions marginal with regard to halting human icm development. + Increased inhibition of MAP kinase signaling, in addition to MEK B/S: butyrate SAHA HDACi mixture inhibition through the use of p38 and JNK inhibitors (6), may 2i: PD0325901 (MEK inhibitor) + CHIR99021 (GSK3 inhibitor)

Ware et al. PNAS Early Edition | 5of6 Downloaded by guest on September 26, 2021 3i: 2i + SU5402 (FGFR inhibitor) 5% O2 atmosphere throughout. Four days after thaw, the zona pellucida was removed using Acidified Tyrode’s Solution and the zona-free embryo T: TeSR2 medium plated into one well of a 96-well plate seeded with feeders. The first passage M: MEF conditioned medium was mechanical 3 d following the removal of the zona. Seven days later, passage was through mechanical plucking of a growth above the plated Parentheses: previous factor exposure colony. Subsequent passages used Trypsin/EDTA. The culture seemed to the right of parentheses: current factor exposure established by passage 5. Henceforth, the cells were passaged two times per week using trypsin/EDTA. Elf1 was bulk-frozen at passage 8 in hESC culture associated numbers: number of passages under those conditions medium plus 10–12 ng/mL FGF2, 3 μM CHIR 99021, and 0.4 μM PD0329501 For example, H1p52(B/S3)2iF10 indicates H1 cultured for a total of 52 (2iF) using the cryoprotectant dimethylsufloxide (10%; Sigma) and given the passages. Cells were exposed to butyrate + SAHA for three passages from name Elf1, “El” to honor the Ellisons, who donated the funds for the work, p40–p42 and then to 2i + FGF for 10 passages from p43 to p52. and “f1” for female, line one. Elf1 is on the NIH Stem Cell Registry (0156; approved May 15, 2012) and is being banked at WiCell for distribution. Culture of Pluripotent Cells. Initial cultures of hESC [H1 (NIH code WA01), H7 (NIH code WA07), H9 (NIH code WA09), BG02, and HUES2], nonhuman pri- Microarray Analysis. Whole mouse and whole human genome arrays (Agilent) mate M. fascicularis (MF-1), mEpiSC5, and mEpiSC7 ES cells were grown on were hybridized with total RNA from ESCs cultured with additives as in- a feeder layer of gamma-irradiated (3,000 Rad) primary mouse embryonic dicated in the text. These were run in independent quadruplicate through the fibroblasts (MEFs) (7, 25). For cultures without feeders, cells were plated on array facility at Seattle Biomedical Research Institute. Matrigel (BD Biosciences) diluted according to the manufacturer’s instruc-

tions. Cultures were incubated at 37 °C in 5% C02,5%O2, and 90% N2. hESC Microarray Data Preprocessing. The data for each experiment were stored as culture medium consisted of high-glucose DMEM/F12 containing GlutaMax raw, preprocessed, and processed in Synapse Commons Repository (https:// supplemented with 20% KnockOut serum replacer (SR), 1 mM sodium py- synapse.sagebase.org/). With the exception of the Hunter et al. study (8) ruvate, 0.1 mM nonessential amino acids, 50 U/mL penicillin, 50 mg/mL (GSE8881), we implemented the default one-color Affymetrix and default streptomycin (all from Invitrogen), and 0.1 mM β-mercaptoethanol (Sigma– two-color Agilent workflows for each dataset. The former removes intensity- Aldrich). To reverse toggle, hESCs were first cultured in the presence of an dependent array effects, and the latter removes both intensity-dependent HDACi mixture containing 0.1 mM sodium butyrate (Sigma–Aldrich) plus array and dye effects. A single study, the Elf1 dataset, required a more so- 50 nM SAHA (B/S; Cayman). Human LIF was used when LIF is specified for phisticated model, which in this case meant correcting for the effects of an human and nonhuman primate ESCs. Factors were used at the following outlier sample. Further methods are defined in Supporting Information. concentrations: 10 ng/mL hLIF (Chemicon), 10 ng/mL Activin A (Humanzyme), 10 ng/mL FGF2 (Invitrogen), 0.4–0.8 μM PD0325901 (Selleck), 1.0–3.0 μM Immunohistochemistry. Triple immunofluorescent labeling to detect endo- CHIR99021 (Selleck), 2 μM SU5402 (Calbiochem), 0.1 μM NSC74859 (Selleck), derm in teratoma sections was performed as previously published with minor and 10 μM Y-27632 (Selleck). Primed cells and human cells in B/S were exposed modifications (26). Methods are further defined in Supporting Information. to 1.2 U/mL dispase (Invitrogen) dissolved in PBS supplemented with SR. Naïve Further methods for array analysis, immunohistochemistry, flow cytometry, cells and mouse cells in B/S were passaged using trypsin/EDTA (Invitrogen). quantitative PCR, XIST FISH, bisulfite sequencing, teratoma formation, karyo- mESCs were cultured in either DMEM or DMEM/F12 with the additives type analysis, and electron microscopy can be found in Supporting Information. described above, except the serum replacer was substituted with 20% FBS (ES qualified; Invitrogen), unless otherwise stated. MEF-free cultures were grown ACKNOWLEDGMENTS. Eric Hayes and Eliza Curnow kindly provided the on Matrigel- coated dishes using medium supplemented as indicated in the M. fascicularis embryonic stem cells (MF1) for this project. We thank David text. Mouse LIF was used at 1,000 units/mL for mESCs (ESGRO; Chemicon). Hockenbery, Daciana Margineantu, and Scott Hansen for helpful discussions Cells for undirected differentiation were seeded at ∼2 × 105 cells per during the progress of this work. John Stamatoyannopoulos, Richard Sand- uncoated, gelatinized, or Matrigel-treated 10-cm plate in DMEM (Invi- strom, Theresa Canfield, Sandra Stehling-Sun, and Scott Hansen were helpful trogen) containing 20% FBS (ES cell tested; Invitrogen) with penicillin and in all aspects of the DNase I hypersensitivity analysis. Faria Ahmed and Abhinav streptomycin (Invitrogen). Cells were allowed to attach and fluid was Nagulapally performed, processed, and analyzed the ChIP-seq analysis of Elf1. David-Changsoo Kim facilitated the teratoma analysis. We are greatly indebted changed every other day. to The Ellison Foundation and the State of Washington Life Science Discovery Fund (4553677) for gift and nonfederal funds in support of this study. This Establishment of Elf1. Elf1 was thawed as a six- to eight-cell embryo, cultured work was supported by National Institutes of Health, National Institute of to expanded blastocyst in blastocyst medium (Sage), and kept in a 5% CO2, General Medical Sciences Grant P01 GM081619-01 and R01 GM097372.

1. Brons IG, et al. (2007) Derivation of pluripotent epiblast stem cells from mammalian 15. Lu CW, Lin SC, Chen KF, Lai YY, Tsai SJ (2008) Induction of pyruvate dehydrogenase embryos. Nature 448(7150):191–195. kinase-3 by hypoxia-inducible factor-1 promotes metabolic switch and drug re- 2. Tesar PJ, et al. (2007) New cell lines from mouse epiblast share defining features with sistance. J Biol Chem 283(42):28106–28114. human embryonic stem cells. Nature 448(7150):196–199. 16. Hitosugi T, et al. (2011) Tyrosine phosphorylation of mitochondrial pyruvate de- 3. Ying QL, et al. (2008) The ground state of embryonic stem cell self-renewal. Nature hydrogenase kinase 1 is important for cancer metabolism. Mol Cell 44(6):864–877. 453(7194):519–523. 17. Hentze H, et al. (2009) Teratoma formation by human embryonic stem cells: Evaluation 4. Nagy A, Rossant J, Nagy R, Abramow-Newerly W, Roder JC (1993) Derivation of of essential parameters for future safety studies. Stem Cell Res (Amst) 2(3):198–210. completely cell culture-derived mice from early-passage embryonic stem cells. Proc 18. Ozolek JA, Castro CA (2011) Teratomas derived from embryonic stem cells as models – Natl Acad Sci USA 90(18):8424 8428. for embryonic development, disease, and tumorigenesis. Embryonic Stem Cells: Basic 5. Hanna J, et al. (2010) Human embryonic stem cells with biological and epigenetic char- Biology to Bioengineering, ed Kallos MS (InTech, New York), pp 231–262. – acteristics similar to those of mouse ESCs. Proc Natl Acad Sci USA 107(20):9222 9227. 19. Bao S, et al. (2009) Epigenetic reversion of post-implantation epiblast to pluripotent 6. Gafni O, et al. (2013) Derivation of novel human ground state naive pluripotent stem embryonic stem cells. Nature 461(7268):1292–1295. cells. Nature 504(7479):282–286. 20. Roode M, et al. (2012) Human hypoblast formation is not dependent on FGF signal- 7. Ware CB, et al. (2009) Histone deacetylase inhibition elicits an evolutionarily con- ling. Dev Biol 361(2):358–363. served self-renewal program in embryonic stem cells. Cell Stem Cell 4(4):359–369. 21. Leitch HG, et al. (2013) Rebuilding pluripotency from primordial germ cells. Stem Cell 8. Hunter SM, Mansergh FC, Evans MJ (2008) Optimization of minuscule samples for use Rev 1(1):66–78. with cDNA microarrays. J Biochem Biophys Methods 70(6):1048–1058. 22. Chen Y, Blair K, Smith A (2013) Robust self-renewal of rat embryonic stem cells requires 9. Ren J, Jin P, Wang E, Marincola FM, Stroncek DF (2009) MicroRNA and gene expres- – sion patterns in the differentiation of human embryonic stem cells. J Transl Med 7:20. fine-tuning of glycogen synthase kinase-3 inhibition. Stem Cell Rev 1(3):209 217. 23. Meek S, et al. (2013) Tuning of β-catenin activity is required to stabilize self-renewal 10. Stadler B, et al. (2010) Characterization of microRNAs involved in embryonic stem cell – states. Stem Cells Dev 19(7):935–950. of rat embryonic stem cells. Stem Cells 31(10):2104 2115. 11. Greve TS, Judson RL, Blelloch R (2013) microRNA control of mouse and human plu- 24. Huang Y, Osorno R, Tsakiridis A, Wilson V (2012) In Vivo differentiation potential ripotent stem cell behavior. Annu Rev Cell Dev Biol 29:213–239. of epiblast stem cells revealed by chimeric embryo formation. Cell Rep 2(6): 12. Okamoto I, et al. (2011) Eutherian mammals use diverse strategies to initiate 1571–1578. X-chromosome inactivation during development. Nature 472(7343):370–374. 25. Abbondanzo SJ, Gadi I, Stewart CL (1993) Derivation of embryonic stem cell lines. 13. Lengner CJ, et al. (2010) Derivation of pre-X inactivation human embryonic stem cells Methods Enzymol 225:803–823. under physiological oxygen concentrations. Cell 141(5):872–883. 26. Yebra M, et al. (2003) Recognition of the neural chemoattractant Netrin-1 by 14. Zhou W, et al. (2012) HIF1α induced switch from bivalent to exclusively glycolytic integrins alpha6beta4 and alpha3beta1 regulates epithelial cell adhesion and metabolism during ESC-to-EpiSC/hESC transition. EMBO J 31(9):2103–2116. migration. Dev Cell 5(5):695–707.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1319738111 Ware et al. Downloaded by guest on September 26, 2021