the metabolome regulates the epigenetic landscape during naive … · 2020. 8. 16. · pluripotent...

ART ICLES

The metabolome regulates the epigenetic landscapeduring naive-to-primed human embryonic stemcell transitionHenrik Sperber1,2,3,17, Julie Mathieu1,2,17, Yuliang Wang4,5,17, Amy Ferreccio1,2, Jennifer Hesson2,6, Zhuojin Xu2,Karin A. Fischer1,2, Arikketh Devi1,2,7, Damien Detraux1,2, Haiwei Gu2,8, Stephanie L. Battle2,9, Megan Showalter10,Cristina Valensisi2,9, Jason H. Bielas11,12, Nolan G. Ericson11, Lilyana Margaretha2, Aaron M. Robitaille2,Daciana Margineantu11, Oliver Fiehn10,13, David Hockenbery11, C. Anthony Blau2,14, Daniel Raftery2,8,11,AdamA.Margolin4,5, R. David Hawkins2,9, Randall T. Moon2,15,16, Carol B. Ware2,6 and Hannele Ruohola-Baker1,2,18

For nearly a century developmental biologists have recognized that cells from embryos can differ in their potential to differentiateinto distinct cell types. Recently, it has been recognized that embryonic stem cells derived from both mice and humans exhibittwo stable yet epigenetically distinct states of pluripotency: naive and primed. We now show that nicotinamideN-methyltransferase (NNMT) and the metabolic state regulate pluripotency in human embryonic stem cells (hESCs). Specifically,in naive hESCs, NNMT and its enzymatic product 1-methylnicotinamide are highly upregulated, and NNMT is required for lowS-adenosyl methionine (SAM) levels and the H3K27me3 repressive state. NNMT consumes SAM in naive cells, making itunavailable for histone methylation that represses Wnt and activates the HIF pathway in primed hESCs. These data support thehypothesis that the metabolome regulates the epigenetic landscape of the earliest steps in human development.

Pluripotent stem cells are able to self-renew and have the capacityto regenerate all tissues in the body. These cells hold promisefor understanding early human development as well as developingtherapies in regenerative medicine. Recent findings have revealed thatpluripotency does not represent a single defined state; diverse statesof pluripotency, with differences inmeasurable characteristics relatingto gene expression, epigenetics and cellular phenotype, providean experimental system for studying potential key regulators thatconstrain or expand the developmental capacity of pluripotent cells1–4.Two stable pluripotent states have been derived in the mouse, andnow in humans; preimplantation naive and postimplantation primedembryonic stem cell (ESC) states5–12. As naive, preimplantationhuman ESCs (hESCs) show higher developmental potential thanpostimplantation, primed hESCs (refs 8,12), it is critical to understandthe key molecular differences between these pluripotent cell types.

Metabolic signatures are highly characteristic for a cell andmay actas a leading cause for cell fate changes13–20. Recent data have shownthat pluripotent stem cells have a uniquemetabolic pattern. The naive-to-primed mouse ESC (mESC) transition accompanies a pronouncedmetabolic switch fromabivalent to a highly glycolytic state20. However,the primed state of inert mitochondria rapidly changes to highlyrespiring mitochondria during further differentiation. It is not yetunderstood how and why the pluripotent cells enter the highlyglycolytic, metabolically cancer-like (Warburg effect) state and how adifferentiating cell leaves this state.

In mESCs, the metabolism of threonine and that of S-adenosylmethionine (SAM) are coupled, resulting in regulation of histonemethylation marks21. Methionine and SAM are also required for theself-renewal of hESCs, because depletion of SAM leads to reducedH3K4me3marks and defects inmaintenance of the hESC state22. SAM

1Department of Biochemistry, University of Washington, Seattle, Washington 98195, USA. 2Institute for Stem Cell and Regenerative Medicine, University ofWashington, Seattle, Washington 98109, USA. 3Department of Chemistry, University of Washington, Seattle, Washington 98195, USA. 4Department of BiomedicalEngineering, Portland, Oregon 97239, USA. 5Computational Biology Program, School of Medicine, Oregon Health & Science University, Portland, Oregon 97239, USA.6Department of Comparative Medicine, University of Washington, Seattle, Washington 98195, USA. 7Department of Genetic Engineering, SRM University,Kattankulathur 603203, India. 8Northwest Metabolomics Research Center, Department of Anesthesiology and Pain Medicine, University of Washington, Seattle,Washington 98195, USA. 9Department of Medicine, Division of Medical Genetics and Department of Genome Sciences, University of Washington, Seattle, Washington98195, USA. 10University of California Davis Genome Center, California 95616, USA. 11Fred Hutchinson Cancer Research Center, Seattle, Washington 98109, USA.12Department of Pathology, University of Washington School of Medicine, Seattle, Washington 98195, USA. 13King Abdulaziz University, Faculty of Science,Biochemistry Department, Jeddah 21589, Saudi Arabia. 14Department of Medicine, Division of Hematology, University of Washington, Seattle, Washington 98195,USA. 15Howard Hughes Medical Institute, Chevy Chase, Maryland 20815, USA. 16Department of Pharmacology, University of Washington, Seattle, Washington 98109,USA. 17These authors contributed equally to this work.18Correspondence should be addressed to H.R.-B. (e-mail: [email protected])

Received 30 June 2015; accepted 1 October 2015; published online 16 November 2015; DOI: 10.1038/ncb3264

NATURE CELL BIOLOGY VOLUME 17 | NUMBER 12 | DECEMBER 2015 1523

© 2015 Macmillan Publishers Limited. All rights reserved

mailto:[email protected]

ART ICLES

0200400600800

1,0001,2001,4001,600 HIF1α

Tubulin

H7Elf1

Elf1 Elf1 AF

HIF1α

Tubulin

100130Mr (K)

Mr (K)

55

55100130

a b

∗∗

∗∗∗

NS

Naive Primed

0

0.2

0.4

0.6

0.8

1.0

1.2R

elat

ive

OC

Rch

ange

s af

ter

FCC

P

05

101520253035

OC

R c

hang

esaf

ter

FCC

P

D0 D1 D2 D3

Naive-to-primedtransition (Elf1)

∗∗

∗∗∗∗∗∗

Rel

ativ

e O

CR

cha

nges

afte

r FC

CP

∗∗∗

c d

h i j

e f

g

0

0.2

0.4

0.6

COX7B2

COX7BCOX5A

COX5B

2

1

0

–1

–2

–3

–4

COX3COX6B1COX6CCOX7A2COX4I1COX6A2

COX11COX7A1

COX10COX7A2LCOX18COX15COX19COX6B2COX6A1COX7CCOX8ACOX17COX16COX8CCOX412

Theunissen et al.10 naiveCell lines

Theunissen et al.10 primed

0.8

1.0

1.2

mtD

NA

cop

ies

(per

dip

loid

gen

ome)

Naive-to-primedtransition (WIN1)

Elf1AF – + + +

D0 D2WIN1AF – +

PrimedSALL2

SMCA4

JARID2LDHA

NaiveMOES

VIME

log2 fold change protein expression0 2

–log

10 P

val

ue

–2–4 4

0

2

4

6

8

LDHA

Tubulin

H7 Elf1 Elf1AF

JARID2

0

2

4

6

8

ND1/ND2 CommonmtDNA site

Elf1Elf1 AF

Del

etio

n fr

eque

ncy

(×10

–7 p

er b

p)

NS

NS

NS

130

55

2535

Elf1 A

FElf1

Label-freequantification

Digestion andfractionation

Inte

nsity

Inte

nsity

m/z m/z

Quantification andidentification

Nano LC-MS/MS(Fusion Orbitrap)and MaxQuant

Trypsin andonline HPLC

Elf1 naiveElf1 primed

Elf1

H1 4iL

IF H1 H7

Chan et al.6 3iL hESCChan et al.6 hESC

0

PC1 (55.3%)

PC

2 (1

8.9%

)

0

0.2

0.20.3

–0.2

–0.2–0.1

0.1

0.1

–0.1

Elf1Finkbeiner H9Grow46 Elf1Grow46 Elf1 AF

H1 4iLIFH1

H1 ENCODE47

Lis1Takashima et al.9 H9Takashima et al.9 H9 reset

Primed Naive–8

–4

0

4

0 5–5–10 10 15

PC1

PC

2

PTPRZ1USP44GRPR

NLGN4X LECT1FZD7

DUSP6BEX1

STC1KIF1A

SALL2

IDO1NELL2

CRABP1

LCK

MFGE8SFRP1 MAP7

SLCA73HAPLN3

CLDN3 ZDHHC22

GPR176

DNMT3B

LDHAJARID2 WNT3A

HIGDA1

COX5BCOX7B

NNMT

WNT5AWNT10B

WNT8A T

WNT5B

WNT9AKLF4

DPPA3

DPPA5

NLRP7

ALPPL2KHDC3L

FGF4ZYG11AGDE3CHST2

MT1ENODAL

HPGDCDX1HAND1MXL1APLNR WLS

SP5

COL3A1

DNMT3L

k lm

Time (min)

H1 4iLIFH1

Elf1

235

210

186

161

137

112

87

63

38

14

–110 19 38 57 76 94 113 132

Nor

mal

ized

OC

R (p

mol

min

–1) Oligo FCCP Anti/Rot

Figure 1 Naive and primed ESCs are metabolically different. (a) Principalcomponent analysis (PCA) of RNA-seq data from this study (Elf1, H1 4iLIF,Lis1, H1) and other studies6,9,46,47 (Finkbeiner data can be accessed atArrayExpress, accession number E-MTAB-3158). ComBat was applied on thecombined RNA-seq data set. (b) Genes contributing to principal componentsseparating primed versus naive hESCs. The size of the dots is proportional tothe square of the first principal component (PC1) value. Top contributinggenes are darker. (c) Metabolic profile of naive (Elf1 and H1 4iLIF) andprimed (H1) hESCs. A trace of oxygen consumption rate (OCR) changesin response to oligomycin, FCCP and antimycin/rotenone is shown under aMitoStress protocol (s.e.m., n=6 biological replicates). (d) Primed hESCs(H7 and H1) have reduced OCR changes in response to FCCP comparedwith naive hESCs (Elf1 and H1 4iLIF), n= 18 (H1, H1 4iLIF) or 24(Elf1 and H7) biological replicates; s.e.m.; P=0.122 for H1 4iLIF versusElf1, P=0.0001 for H1 versus Elf1, P=0.0014 for H7 versus Elf1; two-tailed t-test. NS, not significant. (e,f) Transition of naive hESCs Elf1 (e)and WIN1 (f) towards a more primed state by culture in activin-A–FGF(AF) media reduced OCR changes in response to FCCP after 1 to 3 days(D0 to D3) (n= 29 for Elf1 AF D1, n= 20 for Elf1 AF D2, n= 28 for

Elf1 AF D3, n= 33 for Elf1, n= 18 for WIN1 and WIN1 AF; s.e.m.;P =0.0013 for Elf1 AF D1 versus Elf1, P <0.001 for Elf1 AF D2 versusElf1, Elf1 AF D3 versus Elf1 and WIN1 AF versus WIN1; two-tailed t-test).(g) Heatmap log2 fold expression change of mitochondrial complex genesbetween primed and naive stages10. (h,i) Naive hESCs (Elf1) and primedhESCs (Elf1 AF) have similar mitochondrial DNA copy number (h, n=3)and mitochondrial mutation frequencies (i, n=3). s.e.m.; P=0.7802 (h),P =0.37 and 0.6 (i); two-tailed t-test. (j) HIF1α protein is stabilized inprimed hESCs (H7 and Elf1 AF). (k) Proteomic workflow used to identifydifferentially regulated protein expression in primed versus naive hESCs.(l) Volcano plot of differentially expressed proteins in primed hESCs (green,Elf1 AF) versus naive cells (blue, Elf1). Significant hits are shown (falsediscovery rate (FDR) < 0.05). Proteins were quantified by nano-LC–MS/MSon a Fusion Orbitrap. (m) JARID2 and LDHA proteins are upregulatedin primed hESCs (Elf1 AF and H7) compared with naive hESCs (Elf1),as revealed by western blot analysis. Unprocessed original scans of blotsare shown in Supplementary Fig. 9. For raw data, see SupplementaryTable 4. n represents the number of biological replicates. ∗∗P < 0.01;∗∗∗P<0.001.

1524 NATURE CELL BIOLOGY VOLUME 17 | NUMBER 12 | DECEMBER 2015


https://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-3158/

ART ICLES

therefore is shown to be a key regulator for maintaining the ESCundifferentiated state and regulating their differentiation. However,little is known about SAM levels or its regulation during the transitionbetween naive and primed human embryonic states. Recent derivationof naive hESCs allows a deeper analysis of the human naive-to-primed transition6–12. These studies have already revealed that theepigenetic landscape changes from the naive to primed state throughincreased H3K27me3 repressive methylation marks. However, theregulation of this process or the metabolomics of this transition havenot been dissected.

We now show that the upregulation of H3K27me3 repres-sive epigenetic marks during the naive-to-primed hESC tran-sition is controlled by the metabolic enzyme nicotinamideN -methyltransferase (NNMT). Knockdown of NNMT in naivehESCs increased H3K27me3 repressive marks in developmentalas well as key metabolic genes that regulate the metabolic switchin naive-to-primed transition. CRISPR-Cas9-based NNMT KOnaive hESC lines show upregulation of SAM, H3K27me3 marks,hypoxia-inducible factor (HIF) activation, Wnt repression and ageneral gene expression shift towards the primed stage. These datashow that NNMT consumes SAM in naive cells, making it unavailablefor histone methylation. Histone methylation further regulates thekey signalling pathways important for the metabolic changes that arenecessary for early human development.

RESULTSA pronounced metabolic switch occurs in mESCs between the preim-plantation (naive) and postimplantation (primed) states20. The humannaive counterpart has been recently toggled or derived from embryos.Principal component analysis (PCA) of the expression signatures ofthese new cell types confirmed that all derived naive hESCs are in asignificantly earlier stage than primed hESCs (Fig. 1a,b and Supple-mentary Fig. 1A–C and Supplementary Table 1A)6,8–10,23. To assess themetabolic profiles of the naive and primed hESCs, we analysed thecells’ oxygen consumption rates (OCR) using a Seahorse extracellularflux analyser. As seen previously in mESCs (ref. 20), we detected asignificant increase in OCR after FCCP injection in the newly derivednaive hESCs (Elf1; ref. 12; WIN1; ref. 10) whereas a much smallerincrease was observed in primed hESCs (H1, H7) or cells transitioningto the primed state (Elf1 AF, WIN1 AF; (Fig. 1c–f and SupplementaryFig. 1E–I). Likewise, cells ‘toggled’ back to a more naive state (H1 2iF(ref. 12), H1 4iLIF) showed increased OCR in response to FCCP to alevel similar to mESCs (Fig. 1a,c,d and Supplementary Fig. 1G,J–K).These results indicate that the primed hESCs have a lower mitochon-drial respiration capacity than naive hESCs.

The higher mitochondrial capacity of naive hESCs (Elf1) reflectsneithermorematuremitochondria12, nor an increase inmitochondrialDNA (mtDNA) copy number compared to primed hESCs (Elf1 AF,H7, H1; Fig. 1h and Supplementary Fig. 1L,M). Further, no obviousincrease in mtDNA mutation frequency was detected in primedcompared to naive hESCs (Fig. 1i and Supplementary Fig. 1N,O),suggesting that reduction of oxidative respiration in primed hESCsis not caused by a deteriorating mitochondrial genome. However,consistent with the mouse data, RNA-seq data from our studyand microarray or RNA-seq data from other studies3,6,8–10 showedthat expression of most mitochondrial electron transport chain

complex IV (cytochrome c oxidase (COX)) genes is significantlydownregulated in the primed state compared with the naive state(Fig. 1g and Supplementary Table 1B–D). Also, consistent withthe mouse data20, HIF1α is stabilized in primed but not in naivehESCs (Fig. 1j), correlating with a significant change in expressionof prolyl hydroxylase domain-containing protein 2, PHD2 (EGLN1),the primary regulator of HIF1α steady-state levels24,25 (SupplementaryTable 1B). Further support for HIF1α stabilization and activity at theprimed state comes fromour proteomic analysis revealing a significantincrease in the protein expression of the HIF targets Ldha and Jarid2at the primed hESC state (Elf1 AF compared with Elf1; Fig. 1k,land Supplementary Table 1E and Supplementary Fig. 1P, validationin Fig. 1m).

Differential metabolites between naive and primed embryonicstem cellsTo search for critical metabolites that control the metabolictransitions between naive and primed mESCs and hESCs, weperformed metabolic profiling using gas chromatography timeof flight (GC–TOF), liquid chromatography quadrupole time offlight (LC–QTOF) and liquid chromatography triple quadrupole(LC–QQQ) mass spectrometry (MS) analysis (Fig. 2a and Supple-mentary Table 1F–L).

PCA of the metabolite data reveals a difference in metaboliteprofiles betweennaive andprimed cells, regardless of species (Fig. 2b–fand Supplementary Fig. 2). Multiple naive and primed cell lines fromhuman and mouse separated clearly by naive versus primed stateon the basis of the PCA plot of GC metabolomics data (Fig. 2b,d–fand Supplementary Fig. 2A–C). Stearic acid and cholesterol are themetabolites that contribute the most to the separation within the firstPC, indicating that when ESCs transition from the naive to the primedstate, a major switch occurs in the lipid metabolism. A similar trendof naive and primed ESC separation is observed in PCA plots of theLCmetabolomics data (Fig. 2c and Supplementary Fig. 2D). However,H1 2iF, which is a primed cell line, ‘toggled’ towards the naive state, andElf1, the naive hESC line, clustered midway between the mouse naivecell line (R1) and primed cell lines (R1 AF, EpiSC, Elf1 AF, H7 andH1) in LC analysis, suggesting that H1 2iF and Elf1 have not reachedthe same naive state as observed in mouse cells, with respect to lipidsignature (Fig. 2c and Supplementary Fig. 2D).

In addition, targeted analysis of metabolites was performed usingLC–QQQ MS with naive Elf1 and primed H1 hESCs (Fig. 2g andSupplementary Table 1H,I) and using GC–TOF with naive Elf1 andprimed Elf1 AF (Supplementary Fig. 2B and Supplementary Table 1J).Metabolites upregulated in the primed state include fructose(1,6/2,6)-bisphosphate (F16BP or F26BP), lactate, methionine, nicotinamideand kynurenine (Fig. 2f,g and Supplementary Fig. 2B). Upregulationof F16BP is in concord with highly active glycolysis; however,phosphoenolpyruvate (PEP), a downstreammetabolite of F16BP, doesnot increase in primed hESCs (Fig. 2h,i). Intermediates before PEP canbe conserved for biosynthetic purposes: 3-phosphoglycerate can bediverted to serine and glycine synthesis, which can supply one-carbonunits to multiple methylation reactions; dihydroxyacetone phosphate(DHAP) can be converted to glycerol, which serves as the backbone ofglycerolipids. Therefore, we tested potential changes in lipid and fattyacid metabolism and amino acid pathways (Fig. 2h).



ART ICLES

0.3

0.5

1.0

2.0

4.0

8.0

Glucos

e

G1P/G

6P/F

6P/F

1P

F16P

/F26

BP

D-GA3P

/DHAP

PEP

Pyruv

ate

Fold

cha

nge

ofex

pre

ssio

n H

1 ve

rsus

Elf1

Relative expression of glycolysismetabolites in H1 versus Elf1

h

i

a b c

d e

f g

∗∗

∗∗

Glucose

G1P/G6P/F6P

F16BP/F26BP

D-GA3P/DHAP

PEP

Pyruvate

Amino acidsynthesis

Lipidsynthesis

mEpiSC

hESC

mESC

Elf1

Post-implantation

Pre-implantation

Culturing embryonicstem cells

R1H1 2iF

Elf1

R1 AFmEpiSC

H1

Mass spec. GC metabolomics LC metabolomics

R1 AF versus R1 EpiSC versus R1

H1 versus H1 2iF H1 versus Elf1

5

–10

–5

5

0

–10

–5

5

10

15

0

7.5

5.0

0

2.5

0

4

6

8

2

2

1

00

4

6

2

0 –5 –105 10 150PC1 PC1

PC

2

PC

2

–5–10

20–2–4 1 20–1–2

2.50log2 fold change metabolite level log2 fold change metabolite level

log2 fold change metabolite level log2 fold change metabolite level

–log

10 P

val

ue–l

og10

P v

alue

–log

10 P

val

ue–l

og10

P v

alue

–2.5–5.0 2.50–2.5–5.0

GC-TOFNon-targeted,aq. metabolites

LC-QTOFLipophilicmetabolites

LC-QQQTargeted, aq. metabolites

ZX YAnnotation of peaks

Data analysis

Mouse

Human

EpiSCH1H1 2iFR1R1 AF

No sig. changeHigher in R1 AFHigher in R1

Methionine

Succinic acidGlucose

Lactic acidNicotinamide

Hypoxanthine

Methionine

Aspartic acidKynurenate

F16BP/F26BPXanthosine

Taurine

TrimethylamineCholine ATP

L-Kynurenine

GlucoseLactic acid

Nicotinamide

Hypoxanthine

Methionine

Succinic acid

Glucose

Lactic acid

Nicotinamide

Hypoxanthine

No sig. changeHigher in H1Higher in H1 2iF

No sig. changeHigher in H1Higher in Elf1

No sig. changeHigher in EpiSCHigher in R1

EpiSCH1H1 2iFR1R1 AF

Figure 2 Metabolomic analysis of naive and primed ESCs. (a) Schematicshowing the mass spectrometry (MS) experiments performed for metaboliteson naive (pre-implantation) and primed (post-implantation) mESCs andhESCs. GC, gas chromatography; LC, liquid chromatography; see maintext for full definitions. (b,c) Naive and primed stem cells can be clearlyseparated on the basis of their metabolic profiles. (b) PCA plot of water-soluble untargeted GC–MS metabolomics data. The first principal component(PC), which separates the primed cell types (left) from the naive cell types(right), explained 50.5% of total variance. (c) PCA plot of untargeted LCmetabolomics data. Three clusters are along the first PC: primed cells (left),primed cells toggled back to naive cells (middle) and naive cells (right). The

first PC explained 68.2% of total variance. (d–g) Volcano plots of differentiallyabundant metabolites between primed and naive cells in mESCs detectedby GC–TOF (d,e) and hESCs (f, GC–TOF; g, LC–QQQ MS). Metabolites ofbiological interest for further analysis are labelled. (h) Visualization of theglycolysis pathway and its connections to lipid and amino acid synthesis.(i) Fold change of glycolysis metabolites (n=3, s.e.m.; glucose (P=0.6630),G1P–G6P–F6P–F1P (P=0.3713), F16BP–F26BP (P=0.0070), D-GA3P–DHAP (P=0.0058), PEP (P =0.1925), pyruvate (P =0.1416); two-tailedt-test) after log2 transformation and mean centring in H1 versus Elf1,detected by targeted LC–QQQ MS. For raw data, see Supplementary Tables 1and 4. n represents the number of biological replicates. ∗∗P<0.01.

Differential fatty acid metabolism in naive and primed ESCsFurther lipid analysis was performed using a LC–QTOF instrumenton naive Elf1 and primed H1 cells (Supplementary Table 1K) aswell as non-targeted LC–QTOF analysis on Elf1, Elf1 AF, R1 andmEpi ESCs (Supplementary Table 1L). For 119 lipidomic features withidentified molecular formulae and structures, lipids more abundant inH1 have higher numbers of carbons (Wilcoxon rank sum test P value4.30× 10−4, Fig. 3a). For 320 features with just identified mass, lipidsmore abundant in H1 are heavier (P value = 1.66 × 10−10, Fig. 3b).Non-targeted LC–QTOF also showed that lipids more abundant in

mEpi than in R1 have a significantly higher number of carbons(P value = 0.012, Supplementary Fig. 3G). When sorting on thebasis of unsaturation, lipids more abundant in primed R1 AF havea higher number of double bonds than lipids more abundant innaive R1 (P value = 0.044, Fig. 3c), which can also be observed inprimed Elf1 AF compared with naive Elf1 (P value = 7.35 × 10−5,Supplementary Fig. 3F).

In concordance with the significant increase in long-carbon-chain lipids observed in primed mouse and human ESCs, we alsodetected a significant increase in accumulation of lipid droplets in



ART ICLES

P value: 1.66 × 10–10 P value: 0.044P value: 4.3 × 10–4a b c d

NS

g

PALM/BSA

PALM/BSA ETO ETO

Elf1 PALMElf1 BSAH7 PALMH7 BSA

i

PALM/BSA

PALM/BSA ETO ETO

012345

miR

10a

miR

-9

6

miR

NA

exp

ress

ion

rela

tive

to R

NU

66

Elf1H1

Lipid abundance groupHigher in Elf1 Higher in H1

Lipid abundance groupHigher in Elf1 Higher in H1

Lipid abundance group

7.5

Num

ber

of d

oub

le b

ond

s

5.0

2.5

1,250

1,000

Mas

s 750

250

500

60

50

Num

ber

of c

arb

ons

40

20

30

Higher in R1 Higher in R1 AF

j k l

H9 primedH1 primed

WIRB3 primedC1 primed

WIRB3 naiveBGO1 naive

C1 naiveElf1 naive

C1 primedWIRB3 naiveBGO1 naive

C1 naive

WIRB3 primed

C1 primedWIRB3 naiveBGO1 naive

WIRB3 primed

C1 naive

H3K27me3

H3K4me3

H3K27Ac

∗

∗∗

log 2

fold

cha

nge

log2 fold change miRNA expression

0

–1

–2

H1 ve

rsus

Elf1

Grow46

Theu

nisse

n10

Taka

shim

a9

Chan6

Gafni8

Mou

se

in viv

o

–3

–4

–5

h

e

f CPT1A expression(primed versus naive)

Elf1 AF

BO

DIP

YB

right

fiel

d

Elf1

EpiSCmESC

BO

DIP

YB

right

fiel

d

∗∗∗

020406080

100120140160180200

NaiveBSA

NaivePALM

PrimedBSA

PrimedPALM

OC

R c

hang

es a

fter

ETO

ad

diti

on

HumanMouse

–log

10 P

val

ue

30miR-10a-5p

miR-372-3p

miR-193b-3p

miR-520f-3p

miR-9-5p

No sig. diff. Higher in H1 Higher in Elf1

miR-33a-3p

20

10

–10 10

0

50100150200250300350400

050 100

Time (min)

Nor

mal

ized

OC

R(p

mol

min

–1)

Nor

mal

ized

OC

R(p

mol

min

–1)

150 2000 50 100Time (min)

150 200

0

0

3.07

× 1

0–5

3.19

× 1

0–26

8.06

× 1

0–5

1.98

× 1

0–30

1.31

× 1

0–5

1.26

× 1

0–2

4.04

× 1

0–10

50

100

150

200

250

300

0

R1 PALMR1 BSAEpi PALMEpi BSA

68.595 mb

68.6 mb

68.605 mb

68.61 mb

CP

T1A

01234

ELF

1H

3K27

me3

01234

C1

naiv

eH

3K27

me3

01234

BG

O1

naiv

eH

3K27

me3

01234

WIB

R3

naiv

eH

3K27

me3

01234

C1

prim

edH

3K27

me3

01234

WIB

R3

prim

edH

3K27

me3

01234

H1

prim

edH

3K27

me3

01234

H9

prim

edH

3K27

me3

00.5

11.5

22.5

3

C1

naiv

eH

3K4m

e3

00.5

11.5

22.5

3

BG

O1

naiv

eH

3K4m

e3

00.5

11.5

22.5

3

WIB

R3

naiv

eH

3K4m

e3

00.5

11.5

22.5

3

C1

prim

edH

3K4m

e3

00.5

11.5

22.5

3

WIB

R3

prim

edH

3K4m

e3

00.10.20.30.40.5

C1

naiv

eH

3K27

ac

00.10.20.30.40.5

BG

O1

naiv

eH

3K27

ac

00.10.20.30.40.5

WIB

R3

naiv

eH

3K27

ac

00.10.20.30.40.5

C1

prim

edH

3K27

ac

00.10.20.30.40.5

WIB

R3

prim

edH

3K27

ac

CPT1A

04

30

0.50

Figure 3 Primed ESCs accumulate lipids whereas naive ESCs use fatty acidsas a source of energy. (a,b) More abundant lipids in primed hESCs (H1) havemore carbon atoms (a) and larger mass (b) than more abundant lipids innaive hESCs (Elf1). (c) More abundant lipids in primed mESCs (R1 AF) aremore unsaturated than more abundant lipids in naive mESCs (R1). n=6,P values Wilcoxon rank–sum test. Boxes represent median, 25th and 75thquantiles. Whiskers extend 1.5 IQR above the 75th quantile and below the25th quantile. Dots represent values beyond whiskers. (d,e) BODIPY 493/503staining shows an increase of lipid droplet accumulation in primed human(Elf1, d) and mouse (EpiSCs, e) ESCs compared with naive human (Elf1 AF,d) and mouse (R1, e) ESCs. Scale bars, 50 µm. (f) CPT1A is downregulated inhuman and mouse primed ESCs compared with naive ESCs in our study andothers. n from left to right (primed,naive): 2,1; 3,3; 2,5; 3,3; 3,3; 3,9; 3,2.Negative binomial test P values are shown. (g) ChIP-seq analysis of the CPT1Agene shows more repressive H3K27me3 marks and fewer active H3K4me3and H3K27ac marks in primed hESCs (C1, WIBR3 (ref. 8), H1, H9 (ref. 46))than in naive hESCs (Elf1 (ref. 12); naive C1, naive BGO1, naive WIBR3

(ref. 8)). (h) Volcano plot of microRNA expression in naive hESCs (Elf1)and primed hESCs (H1, ENCODE47, Supplementary Table 1M). (i) qPCRexpression of hsa-miR-9 and hsa-miR-10a (predicted to target CPT1A andFASN, respectively). hsa-miR-10a is 34-fold higher, and hsa-miR-9 is 4-fold lower in Elf1 versus H1 (n=3, s.e.m.; miR-10a: P =0.004, miR-9:P=0.022; two-tailed t-test). (j–l) Seahorse palmitate assay shows that naivehuman and mouse ESCs use fatty acids as a source of energy. A trace of OCRchanges after palmitate (PALM) or BSA vehicle addition, followed by ETOin hESCs (naive Elf1 and primed H7, j) and mESCs (naive R1 and primedEpiSCs, k). n=4 for Elf1 BSA, R1 BSA, Epi BSA, n=5 for Elf1 PALM,R1 PALM, Epi PALM, n=6 for H7 BSA, H7 PALM; s.e.m. Changes afterETO injections were quantified in l (Elf1 BSA (n=12), R1 BSA (n=12),Epi BSA (n=12), Elf1 PALM (n=15), R1 PALM (n=15), Epi PALM (n=15),H7 BSA (n=18), H7 PALM (n=18), s.e.m.; naive hESCs: P =0, 0096,primed hESCS: P=0.354, naive mESCs: P=0.03, primedmESCs: P=0.88,two-tailed t-test). For raw data, see Supplementary Table 4. n represents thenumber of biological replicates. ∗P<0.05; ∗∗P<0.01.

the primed state, as observed by Oil red O and BODIPY staining(Fig. 3d,e and Supplementary Fig. 3A,B). These data indicatean increased synthesis and/or decreased beta-oxidation in primedcells. Interestingly, several of the enzymes involved in fatty acidtransport into the mitochondria and fatty acid beta-oxidation are

significantly downregulated in primed hESCs, as well as in themouse in vivo postimplantation state (Supplementary Fig. 3H,I).Carnitine acyltransferase 1 (CPT1) transfers acyl groups from long-chain fatty acid CoA to carnitine, facilitating the initial step in long-chain fatty acid transfer to the mitochondrial matrix. Interestingly,



ART ICLES

0

0.002

0.004

0.006

0.008

0.010

0.012

0.014

0.016

0.018

0.020

H7

5iLI

FW

IN1

5iLA

FH

7 5i

LAF

WIN

1 5i

LA

H1

2iF

Elf1

H1

4iLT

F

Lis1

WIN

1 F

H1

H7

Elf1

AF

mR

NA

exp

ress

ion

rela

tive

to A

CTB

0

Elf1 2

iLIF

GC H1/

H1 2iF

GC Elf1

AF/

Elf1

GC H7/

Elf1

GC H1/

Elf1

QQQ H1/

Elf1

HILIC H

1/Elf1

HILIC H

1/W

IN1

H1 2iF

H1 4iL

IF H1 H7

Elf1 A

F

0.005

0.010

0.015

0.020

IDO

mR

NA

exp

ress

ion,

norm

ailz

ed t

o A

CTB

0

10

20

30

40

50

60

Elf1 Elf1 AF1Y20AOE

Elf1 AF1NNMT

OE

OC

R c

hang

es a

fter

FC

CP

No sig. diff. Higher in H1Higher in Elf1

log2 fold change mRNA expression

–log

10 P

val

ue

0 5 10–5–10

0

5

10

15

NNMT IDO1NNMT expression e f

h

∗∗∗

i

259 242

160

81 79

31 30

∗

k

∗

∗∗∗

∗∗∗ ∗∗∗

∗∗∗

∗∗∗ ∗∗∗

∗∗∗ ∗∗∗

Nor

mal

ized

OC

R(p

mol

min

–1)

Time (min)

Oligo FCCP Anti/Rot

WIN1 + SAM 5 hWIN1

IDO1 expression

a

b c

Prim

ed (K

YN

/Trp

) ver

sus

naiv

e (K

YN

/Trp

)

∗

∗∗∗∗∗∗

∗∗∗∗

∗∗

∗

∗∗∗

∗∗∗

∗∗∗

Methionine

d

NNMT

1-MNA

Nicotinamide

SAM

Histone,DNA

Methylatedhistone,DNA

Primed

0

0.2

0.4

0.6

0.8

1.0

1.2

WIN1 WIN1+SAM

Rel

ativ

e O

CR

cha

nges

afte

r FC

CP

Non-targeted MSTargeted MS

j

285

21∗∗∗

g

0

0.05

0.10

0.15

0.20

0.25

Elf1 Elf1 AF H1 0–31533516987

105123141159177

13 26 39 52 66 79 92 105 118 131Sum

-nor

mal

ized

SA

M le

vels SAM

H1

4iLI

F

0

2

4

6

8

10

12

14

SAH

Tryptophan

NNH2

NH2NH2

NH2

NH2

NH2

NH2

S O

OH OH

HO2C

+NH3

NH2

CH3

CH3

NH2OHOH

OH

OHHO

OO O

O

O

O

O

N N

N

N

N

N N

N

N

N

–OO S S

+OH

IDO1

Kynurenine

AADATDegradation

WIN

1El

f1H

1 2i

FH

1 4i

LIF H1

WIN

1 Te

SREl

f1 A

F

Elf1

gN

NM

T 7.

2.1

Elf1

gN

NM

T 6.

2.4

∗∗∗

∗∗∗∗∗∗∗∗∗

1-MNA

Primed

Sum

-nor

mal

ized

1-M

NA

leve

ls

0

0.1

0.2

0.3

0.4

0.5

0.6

Figure 4 Amino acids methionine and tryptophan are differentially regulatedin naive and primed hESCs. (a) Model of the tryptophan–kynurenine pathway.(b) IDO1 is highly expressed in primed hESCs as compared with naive hESCs(qPCR, n=3 for H1 4iLIF, Elf1 AF, n=4 for H1 2iF, H1, H7, n=5 forElf1 2iLIF; s.e.m.; ∗∗∗P<0.001; two-tailed t-test). (c) The kynurenine versustryptophan ratio is higher in primed than naive hESCs, as detected by targeted(n= 6) and non-targeted (HILIC: n= 4, QQQ n= 3) mass spectrometry.(s.e.m.; ∗P <0.05, ∗∗P <0.01, ∗∗∗P <0.001; one-tailed t-test.) (d) Modelof SAM pathway and NNMT. Metabolites in red are upregulated in primedhESCs. Metabolites and enzymes in blue are upregulated in naive hESCs.(e) Volcano plot of RNA-seq data from naive hESCs (Elf1) and primed hESCs(H1). Genes with greater than twofold change and FDR < 0.05 are coloured.NNMT and IDO1 are among the most differentially expressed genes. (f) NNMTis highly upregulated in naive hESCs compared with primed hESCs (qPCR).Numbers indicate fold changes of naive hESCs compared with H1 and H7primed hESCs. (n=3 for WIN1, H7 5iLAF, Elf1, H1 4iLIF, H1 4iLTF, Lis1,

WIN1F, Elf1 AF, n=4 for WIN1 5iLA, H1 2iF, H1, H7, n=5 for H7 5iLIF;s.e.m.; ∗∗∗P<0.001; two-tailed t-test.) (g) Naive hESCs (n=4 each) havehigher amounts of 1-MNA than primed hESCs (n=4 for H1, n=6 for WIN1TeSR, Elf1 AF) (s.e.m., ∗∗∗P < 0.001, two-tailed t-test). 1-MNA was notdetected in the Elf1 CRISPR-Cas9 KO mutant of NNMT (gNNMT 7.2.1,n=6; gNNMT 6.2.4, n=6). (h) SAM levels are higher in primed hESCs (H1n=4, Elf1 AF n=6) than in naive hESCs (Elf1 n=4) (s.e.m.; P=0.0089for Elf1 AF versus Elf1, P=0.0376 for H1 versus Elf1; two-tailed t-test).(i,j) SAM induces a ‘primed-like’ metabolic profile in naive hESCs. Additionof SAM (500 µM) for 5 h in media without methionine reduces OCR changesin response to FCCP in naive hESCs (WIN1). A Seahorse trace is shown in i(n=6; s.e.m). OCR changes after FCCP are quantified in j (n=23; s.e.m.;P=0.017). (k) Overexpression of NNMT delays the metabolic transition fromnaive to primed (n=4; s.e.m.; P =0.028, two-tailed t-test). For raw dataand exact P values, see Supplementary Table 4. n represents the number ofbiological replicates. ∗P<0.05; ∗∗P<0.01; ∗∗∗P<0.001.



ART ICLES

the rate-limiting fatty acid transporter CPT1A is downregulated inboth mouse in vivo postimplantation and human primed ESC statecompared with all analysed naive states (Fig. 3f). The decrease inCPT1A expression in the primed state could be due to increasedH3K27me3 and decreased H3K4me3 and H3K27ac marks observedin the CPT1A promoter in primed hESCs (Fig. 3g and SupplementaryFig. 3C)6,8,12. Several of the microRNAs (miRNAs) predicted totarget CPT1A and other enzymes involved in beta-oxidation areupregulated in primed hESCs (for example, miR-9, miR-33a-3p,Fig. 3h and Supplementary Table 1M). Moreover, miRNAs predictedby Targetscan and miRTarBase to target enzymes involved in fattyacid synthesis were downregulated in primed cells (for example, miR-10a and miR-193, Fig. 3h). Concomitantly, key fatty acid synthesisgenes were up in primed H1 hESCs compared with the naive Elf1state (SLC25A1,ACLY,ACACA, FASN and SREBP-1c; SupplementaryFig. 3D). We further validated some of the miRNAs by qPCR analysisand showed that miR-9, predicted to target CPT1A, was upregulated,whereas miR-10a, predicted to target SREBP-1c (a regulator of fattyacid and cholesterol synthesis), was downregulated in the humanprimed state (Fig. 3i).

To test the level of fatty acid beta-oxidation in naive and primedhuman and mouse ESCs, we performed a palmitate-oxidation assayin the Seahorse metabolic flux analyser26. Importantly, both mouseand human naive ESCs were capable of using palmitate as an energysource, whereas primed mouse or human ESCs were not (Fig. 3j–land Supplementary Fig. 3E). This result suggests that primed humanand mouse ESCs are not capable of significant beta-oxidation and,in combination with increased fatty acid synthesis, may explainthe accumulation of lipids observed in this state (Fig. 3d,e andSupplementary Fig. 3a,b).

Differential amino acid metabolism in naive and primed ESCsIn addition to glycolysis and fatty acid metabolism, primed cells showchanges in amino acid metabolism pathways. In primed versus naivehESCs we observed a large enrichment of the tryptophan degradationproduct kynurenine, which can act as a ligand for the nuclear receptorAHR (Fig. 4a,c)27. Interestingly, tryptophan is shown to be criticalfor primed hESC growth22. Accordingly, RNA-seq and qPCR datashow a large increase of the tryptophan-metabolizing enzyme IDO1 inprimed hESCs compared with naive hESCs and in vivo 8-cell humanembryos23 (Figs 1b and 4b,e and Supplementary Fig. 4A). IDO1 levelsquickly drop during differentiation, indicating a specific functionfor IDO1 in the primed state12,28,29 (Supplementary Fig. 4B,C andSupplementary Table 1N).

Interestingly, a strong upregulation of 1-methylnicotinamide(1-MNA), a product of NNMT activity, was observed in all naivecompared with primed hESCs (Fig. 4g).Methionine and nicotinamidedownregulation along with 1-MNA upregulation in the naive statecorrelates with upregulation of NNMT, shown previously to create ametabolic methyl sink, thereby promoting epigenetic remodelling incancer28 (Fig. 4d–g and Supplementary Fig. 4G). Primed hESCs showan increase in SAM levels compared with the naive state (Fig. 4h andSupplementary Fig. 4H and Supplementary Table 1O). The increasein SAM correlates with the sharp decrease in NNMT enzyme levelsobserved in primedhESCs (Fig. 4h,e,f)6,8,11, suggesting that SAM levelsmay be reduced in the naive state by high NNMT activity. Further,

significant expression changes of NNMT among various tissues revealthat NNMT is dynamically regulated during development and suggestthat NNMT might act as regulator of SAM levels also in a develop-mental context, not only in cancer30 (Supplementary Fig. 4D–F andSupplementary Table 1N). Accordingly, we show that high levels ofSAM induce a primed-like metabolic profile in naive hESCs whereasoverexpression of NNMT, but not the mutated form of NNMT,delays the naive-to-primed hESC metabolic switch (Fig. 4h–k andSupplementary Fig. 4I).

NNMT regulates repressive histone modificationsReduction of NNMT levels during the naive-to-primed transitioncorrelates with a significant increase in SAM levels and in H3K27me3histone methylation marks in 648 developmentally regulatedgenes6,8,10,12 (Fig. 5a). Moreover, western blot analysis revealed anoverall increase of H3K27me3 and H3K9me3 marks in primedhESCs compared with naive hESCs, whereas the H3K9 and H3K14acetylation marks remained unchanged (Fig. 5b and SupplementaryFig. 5H). ChIP-seq analysis of other marks (H3K4me1, H3K4me3and H3K27ac) did not show a significant change between primedand naive hESCs (Supplementary Fig. 5A–G)6,8. RNA-seq analysisof histone methyltransferases and histone demethylases involved inH3K27 and H3K9 methyl marks did not show changes in expressionlevels that could explain the significant increase in repressivemethylation marks observed at the primed state (SupplementaryFig. 5I–J). Furthermore, we showed by western blot analysis that theprotein levels of the polycomb repressive complex 2 (PRC2) regulator,EED, are not increased in the primed state compared to the naivestate (Fig. 5b).

Overexpression of NNMT in primed cells reduced H3K27me3marks (Supplementary Fig. 4I). The direct reduction of NNMT levelsin naive hESCs significantly reduced the enzymatic product, 1-MNA,reduced naive hESC enriched microRNAs and increased H3K27me3andH3K9me3marks, as analysed bywestern blots, whereasH3K9 andH3K14 acetylationmarks did not change (Fig. 5c,d and SupplementaryFig. 5K,L). RNA-seq analysis of NNMT knockdown identified asignificant overlap between genes changed in NNMT KD and thenaive-to-primed transition (Fig. 5e). In particular, themost significantoverlap is observed between genes upregulated by NNMT knockdownand genes expressed higher in Elf1 AF versus Elf1. Therefore, on thebasis of transcriptome signatures, NNMT knockdown samples do notonly show increased repressive H3K27me3 marks but are also movingtowards the primed state.

To analyse the early NNMT-responsive genes for H3K27me3marks, we altered the NNMT regulator, LIF–STAT pathway. In naivehESCs LIF activates STAT3 (ref. 31). Activated STAT was previouslyshown to bind the NNMT promoter and activate its transcription32.H1 toggled to a more naive state using 2iF (ref. 12) without externalLIF addition also has a high level of NNMT (Fig. 4f) and activatedLIF–STAT pathway, suggesting that the LIF pathway is endogenouslyactivated in the naive hESC state (Supplementary Fig. 5M). Weshow that treating naive hESCs with a STAT3 inhibitor affectsNNMT expression and the repressive histone methylation pattern.qPCR analysis showed a reduction of NNMT expression in Elf1cells as early as 6 h after STAT3 inhibitor addition (SupplementaryFig. 5N). Importantly, reduction of NNMT in naive hESCs by STAT3



ART ICLES

a

d

H3K9/K14Ac

Tubulin

H3K9me3

H3K27me3

LucsiRNA

NNMTsiRNA

Elf1 h H3K27me3overlap

Elf1 STAT3iversus Elf1

Primed versusnaive

313

c

0

0.5

1.0

1.5

∗∗∗

NNMT

∗

Elf1 luciferase siRNA 72 hElf1 NNMT siRNA 72 h

Elf1

WNTEGLN1

i

b

H3K27me3

e

f

g

j

H3K9/K14Ac

Tubulin

H3K9me3

H3K27me3

Elf1H7

1525Mr (K)

Mr (K)

Mr (K)

1525

1525

55

mR

NA

exp

ress

ion

rela

tive

to A

CTB

Sum

-nor

mal

ized

1-M

NA

leve

ls

55

55

Tubulin

EED

1525

1525

1525

55

1525

1525

55

TSS 2,500–2,500–5,000 5,000

Elf1 STAT3i 6 hElf1

H3K27me3

TSS

–2,5

00–5

,000

5,00

02,

500

H9primed

WIBR2primed

WIBR3 primedC1 primed

WIBR3 naiveC1 naive

BGO1 naiveElf1 naive

STAT3i 6 h

Elf1

AF

vers

us E

lf1(G

row

46)

H1

vers

us E

lf1

Taka

shim

a9

Cha

n6

Theu

niss

en10

Gaf

ni8

In primedIn NNMT siRNA

In primedIn NNMT siRNA

0

0.05

0.10

0.15

0.20

0.25

Rea

ds

per

mill

ion

Elf1H7

Elf1 Elf1AF

Tubulin

H3K27me3 1525

55

9.99× 10–21

3.05× 10–14

0.00516 4.03× 10–9

0.969 0.047

1.84× 10–5

DMSOSTAT3i

–0

–0

+6 h

–24 h

Elf1

H3K27me3

H3K9me3

Tubulin

–5,000 –5,000

0 0

2

4

6

8

10

0.5

1.0

1.5

2.0

2.5

Elf1 naiveC1 naiveBGO1 naive

WIBR3 naiveWIBR2 naiveC1 primed

WIBR3 primedH1 primedH9 primed

WIBR2 primedChan et al.6 naiveChan et al.6 primed

12

–2,500 TSSTSSGenomic position

Genomic region (5′ → 3′)

TSS 2,500–2,500–5,000 5,000

Genomic region (5′ → 3′)

Genomic position2,500 5,000 5,000

1-MNA

0.594 0.261 5.35× 10–7

0.981 0.237

0.10

0.15

0.20

Rea

d c

ount

per

mill

ion

map

ped

rea

ds 0.25

0.1

0.2

0.3

Rea

d c

ount

per

mill

ion

map

ped

rea

ds

0.4

0.5

0 0.2 0.5 0.7 1.0

Elf1naive

Elf1+STAT3i

C1naive

BGO1naive

WIBR2naive

C1primed

H1primed

Figure 5 High NNMT expression in naive hESCs regulates histone methylationstatus. (a) H3K27me3 reads (mapped 5 kb around transcription start sites(TSS) of 648 developmental genes) were plotted for Ware et al.12, Gafniet al.8, Theunissen et al.10, Bernstein et al.48 (left panel) and Chan et al.6

(right panel) ChIP-seq data sets. (b) Western blot analyses show higherH3K27me3 and H3K9me3 levels in primed hESCs (H7 and Elf1 AF)than naive hESCs (Elf1). (c) qPCR analysis shows a knockdown regulationof NNMT using siRNA (50nM, 72h) in naive hESCs (Elf1), inducinga decrease of 1-MNA levels (qPCR n= 3; s.e.m., P = 0.001, two-tailedt-test; HILIC n=4, s.e.m., P =0.039 one-tailed t-test). (d) Western blotanalysis of histone marks in Elf1 cells treated with siRNA against NNMTor siRNA against luciferase as a control. (e) Hypergeometric test P valuesfor the overlap between genes expressed higher (lower) in NNMT siRNAcompared with LUC siRNA and genes expressed higher (lower) in primed linescompared with naive lines from multiple studies. Colour shade is proportionalto negative log10 of P values. LUC siRNA transcriptomic signature has

significant overlap with the Elf1 AF versus Elf1 data set. (f) Western blotanalysis of histone modifications after treatment of Elf1 cells with 100 µMof STAT3 inhibitor (STAT3i). (g) 6 h treatment with STAT3i (100 µM) inElf1 cells increases H3K27me3 marks, as shown by ChIP-seq analysis onall genes. (h) Wnt ligands and EGLN1 are among the 313 overlappinggenes with increased H3K27me3 mark in primed versus naive hESCs(refs 8,10,48), and Elf1 cells treated for 6 h with 100 µM STAT3i versusElf1 cells. (i) Windowed chromatin heatmaps of H3K27me3 profile ±5 kbof promoters of the 313 overlapping genes with increased H3K27me3.(j) H3K27me3 reads from ChIP-seq data mapped 5 kb around TSS wereplotted for naive hESCs (C1, WIBR3, BGO1 (ref. 8) and Elf1 (ref. 12)),primed hESCs (C1, WIBR3 (ref. 8) and H1 (ref. 12)) and naive hESCsElf1 treated for 6 h with 100 µM of STAT3 inhibitor. Unprocessed originalscans of blots are shown in Supplementary Fig. 9. For raw data, seeSupplementary Table 4. n represents the number of biological replicates.∗P<0.05; ∗∗∗P<0.001.

inhibitor also increased H3K27me3 and H3K9me3 marks, as shownby western blot analysis (Fig. 5f). We characterized H3K27me3in naive hESCs by ChIP-seq analysis and observed a significantincrease in H3K27me3 marks at promoters after 6 h STAT3 inhibitortreatment (Fig. 5g and Supplementary Table 1P,Q). Interestingly, morethan 25% (313 genes) of genes with primed-enriched H3K27me3marks already showed increased H3K27me3 marks after 6 h STAT3

inhibitor treatment (Fig. 5h and Supplementary Table 1R).Windowedheatmap and average profile of ChIP-seq signal revealed a markedincrease in H3K27me3 marks close to the transcription start site ofthese 313 genes in the naive-to-primed transition as well as after6 h STAT3 inhibitor treatment5,9 (Fig. 5i,j). Among these genes arecomponents of the Wnt pathway and regulators of HIF (Fig. 5h andSupplementary Table 1R).



ART ICLES

0

50

100

150

200

250

DMSO IWP2 DMSO IWP2

Gen

e ex

pre

ssio

n re

lativ

eto

AC

TB

∗∗∗

∗

Gen

e ex

pre

ssio

n re

lativ

eto

RN

U66

miR-372 inElf1

ed

NNMT

1-MNA

SAM

H3K27me3WNT

Primed hESC

f

Bright field Venus DsRed

Elf1

2iL

IFE

lf1 A

F

c

Elf1

2iL

IF

DMSO IWP2

Brig

ht fi

eld

Venu

s

XAV939

0

0.2

0.4

0.6

0.8

1.0

1.2

Elf1 WIN1 WIN1AF (3D)

Rel

ativ

e O

CR

cha

nges

afte

r FC

CP

No IWP2IWP2

∗∗∗∗∗

∗∗∗

Time (min)53

37332824201612

840

–466 79 92 105 118 131

FCCP Anti/Rot

Nor

mal

ized

OC

R(p

mol

min

–1)

Elf1 + IWP2Elf1

0

0.0005

0.0010

0.0015

0.0020

NNMT inElf1

ba NaivehESCs

PrimedhESCs

Cell linesGrow46 Elf1

Grow46 Elf1 AF

Sperber Elf1

Sperber H1

BAR: β-catenin-activated reporterWnt

β-catenin

12xTCF/LEF-binding sites

Ubiquitinpromoter

Venus DsRedWNT10BWISP1LEF1WNT5BWNT8ADKK1CCND1AXIN2WNT11WNT10AWNT9BWNT5AWNT2TCF7WNT1WNT3AWNT6WNT3MYCWNT9AWNT7AWNT2BWNT4WNT16WISP2WNT8BWNT7B

2

0

–2

–4

–6

Figure 6 The Wnt pathway is active in naive hESCs. (a) Heatmap of geneexpression of Wnt ligands and Wnt targets in primed hESCs (H1 and Elf1 AF)and naive hESCs (Elf1). (b) Wnt is activated in naive hESCs. EndogenousWnt signalling in naive (Elf1) and primed (Elf1 AF) BAR reporter cells. Scalebars, 200 µm. (c) The Wnt inhibitor IWP2 (2 µM) and the Wnt antagonistXAV939 (5 µM) inhibit the reporter activity in naive Elf1 cells after 72h.Scale bars, 200 µm. (d) Wnt inhibition by IWP2 (2 µM, 72h) decreases OCRchanges after FCCP in naive hESCs (Elf1 and WIN1) and in naive hESCstransitioning to primed (WIN1 AF). A trace of OCR changes is presented inElf1 (n=8 for Elf1, n=6 for Elf1+IWP2; s.e.m.). OCR changes after FCCP

were quantified (n=8 for Elf1, n=6 for Elf1+IWP2, WIN1, WIN1+IWP2,WIN1 AF, n=7 for WIN1 AF+IWP2; s.e.m.; P =0.0009 for Elf1+IWP2versus Elf1, P = 0.0084 for WIN1+IWP2 versus WIN1, P = 0.0006 forWIN1 AF+IWP2 versus WIN1 AF; two-tailed t-test). (e) Wnt inhibition byIWP2 (2 µM, 72h) downregulates NNMT (normalized to ACTB) and miR-372(normalized to RNU66) expression in naive hESCs (Elf1) as shown by qPCRanalysis. (n=3; s.e.m.; P=0.04 for miR-372, P=6.44×10−6 for NNMT;one-tailed t-test.) (f) Model of self-reinforcing loop between WNT and NNMTin primed hESCs. For raw data, see Supplementary Table 4. n represents thenumber of biological replicates. ∗P<0.05; ∗∗P<0.01; ∗∗∗P<0.001.

Most Wnt ligands and target genes are downregulated in primedcompared with naive hESCs, suggesting that the Wnt pathway mightbe inactivated during the naive-to-primed transition (Fig. 6a andSupplementary Fig. 6A–DandFig. 1b,l, Vime; ref. 33). Previous studieshave revealed that in human andmouse primedESCs theWnt pathwayis not significantly activated and forced activation of the pathway leadsto differentiation34,35. We now show using a Wnt-pathway activityreporter34 that although the reporter is not activated in primedhESCs, strong activation is observed in naive hESCs (Fig. 6b andSupplementary Fig. 6F). We furthermore show that the Wnt activityin naive hESCs is dependent on β-catenin because short interfering

RNA (siRNA; β-cat) or XAV939 treatment markedly downregulatedthe reporter activity (Fig. 6c and Supplementary Fig. 6E). In addition,the Wnt ligand is produced by the naive hESCs because IWP2, aninhibitor that represses Wnt palmitylation, also represses the reporteractivity in naive cells (Fig. 6c). Inhibition of Wnt in naive hESCsreduces expression of the naive hESC enriched markers, NNMT andmicroRNA miR-372 (ref. 12), and accelerates the transition towardsthe primed metabolic state (Fig. 6d–f and Supplementary Fig. 6G).These data reveal that the robustWnt activity in naive hESCs is amongthe earliest responders to the repressive H3K27me3 marks duringnaive-to-primed hESC transition.



ART ICLES

∗∗∗

∗

∗∗ ∗∗∗

Tubulin

HIF1

Elf1 iCas9gHIF1 6.3.1 gHIF1 6.2.1

55

100130

a

b g h

f

EGLN1

Elf1

H1

H9

H1

WIB

R3

prim

edW

IBR

3na

ive

BG

O1

naiv

eE

lf1E

lf1 +

STA

T3i

H3K

27m

e3(C

hIP

-seq

)R

NA

exp

.(R

NA

-seq

)

0.4

00.4

00.4

00.4

00.4

04

04

0

40

0

40

0

HIF1 CRISPR KO

WT

6.2.1

6.3.1

e

0.06250.1250.25

DN

MT3

L

NN

MT

PD

K1

VE

GFID

O1

0.51248

1632

Fold

cha

nge

(Elf1

gH

IF1

6.3.

1 ve

rsus

Elf1

)

0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Elf1iCas9

gHIF16.2.1

gHIF16.3.1

Rel

ativ

e O

CR

cha

nge

afte

r FC

CP

ad

diti

on

2iLIFAF (D3)

∗∗∗

VHL

OH OH

OH Ub

Ub

Ub

Proteasomaldegradation

OH

HIF1α

HIF1αPro Pro

EGLN1(PHD2)

bHLH

bHLH

PAS

PAS

PAS

P564P402

NTAD CTAD

bHLH

HIF1α

Mr (K)

231.5 mb

231.51 mb

231.52 mb

231.53 mb

231.54 mb

231.55 mb

231.56 mb

d

c CRISPR KO: gHIF1 6.2.1 (DNA seq)

CRISPR KO: gHIF1 6.3.1 (DNA seq)

Trac

e fil

e

DN

A s

eq.

alig

n

+1 bp

WT

WT

6-2-1

WT6-2-1

WT

6-2-1

∗

Trac

e fil

e

DN

A s

eq.

alig

n

Pro

t. s

eq.

alig

nP

rot.

seq

.al

ign

–76 bp

WTWT

6-3-1

WT

6-3-1

Figure 7HIF1α is required for the naive-to-primed hESC transition. (a) Screenshot of RNA expression and H3K27me3 marks of EGLN1 (PHD2) in naivehESCs (Elf1 (ref. 12), WIBR3 naive and BGO1 naive8), primed hESCs(WIBR3 primed8, H1 and H9; ref. 48) and Elf1 treated with STAT3 inhibitor(100 µM) for 6 h. (b) HIFα is hydroxylated on prolyl residues by EGLN1(PHD2), leading to VHL-mediated proteolysis. (c,d) Sequencing trace files,DNA sequences and protein sequences of HIF1α CRISPR-Cas9 knockout (KO)mutant clones (gHIF1 6.2.1, c; gHIF1 6.3.1, d). (e) Schematic representationof wild-type HIF1α protein and proteins resulting from the CRISPR-Cas9KO mutants gHIF1 6.2.1 and gHIF1 6.3.1. bHLH, basic helix–loop–helixdomain; PAS, Per–Arnt–Sim domain; NTAD, amino-terminal transcriptionalactivation domain; CTAD, carboxy-terminal transcriptional activation domain.(f) HIF1α is not expressed in CRISPR-Cas9 KO mutants. Western blotanalysis of HIF1α expression in cells pushed towards the primed stage byculture in TeSR1 for 5 days (5D) in wild-type Elf1 cells (iCas9 Elf1), and

two CRISPR-Cas9 KO mutants of HIF1α (gHIF1 6.2.1 and gHIF1 6.3.1).(g) qPCR analysis of hESCs transitioning to primed reveals that naive markers(DNMT3L and NNMT) are still expressed higher in Elf1 HIF1α CRISPR-Cas9 KO cells compared with wild-type Elf1, whereas primed marker IDO1and HIF target genes (PDK1 and VEGFA) are downregulated (n=3; s.e.m.;P=0.024 for DNMT3L, P=0.0005 for NNMT, P=0.001 for IDO1, P=0.12for PDK1, P=0.004 for VEGFA; two-tailed t-test). (h) KO of HIF1α preventsthe metabolic switch occurring during the transition of hESCs from thenaive to the primed state as shown by measuring OCR after FCCP additionusing Seahorse. n=3 for gHIF1 6.3.1 2iLIF and AF and n=4 for Elf1iCas9 and gHIF1 6.2.1 2iLIF and AF; s.e.m.; P=0.0117 for gHIF1 6.2.1versus Elf1 iCas9, P = 0.0032 for gHIF1 6.3.1 versus Elf1 iCas9; two-tailed t-test. Unprocessed original scans of blots are shown in SupplementaryFig. 9. For raw data, see Supplementary Table 4. n represents the number ofbiological replicates. ∗P<0.05; ∗∗P<0.01; ∗∗∗P<0.001.



ART ICLES

Elf1 gNNMT 7.2.1 (DNA seq) Tr

ace

file

DN

A s

eq.

alig

n

Pro

t. s

eq. a

lign

Trac

e fil

e

DN

A s

eq.

alig

n

Pro

t. s

eq. a

lign

0.0 0.2

SAM

HIF1

WNT

H3K27me3 JARID2/PRC2?

Metabolic switch

hESCs hESCs

EGLN1

NNMT

NaivehESCs

Primed hESCs

versus Elf1 versus Elf1

In primed In gNNMT 7.2.1

In primed In gNNMT 7.2.1

Gen

e ex

pre

ssio

n le

vel

norm

aliz

ed t

o A

CTB

∗∗∗

∗

Elf1 iCas9

gNNMT6.2.4

NNMT CRISPR KO

WT

7.2.1

6.2.4

55

100130Mr (K)

Mr (K)

1525

55

100

1525

4 ×

10–8

0.00

3 1

× 10

–5

0.2

6.8

× 10

–5

0.00

1 4.

6 ×

10–1

0

8.2

× 10

–39

4.1

× 10

–27

2.7

× 10

–5

–2.5–2.0–1.5–1.0–0.5

00.51.01.52.02.5

log 2

fold

cha

nge

(gN

NM

T 7.

2.1

vers

us E

lf1)

NN

MT

WN

T5A

W

NT5

B

WN

T9B

P

DK

1 P

DK

2 H

K2

PK

M

LDH

A

VE

GFA

0.40.2

0–0.2–0.4

PC

2 (1

2%)

–0.6

Methyltransferase

Methyltransferase

Methyltr

Elf1 gNNMTiCas9

HIF1α

H3K27me3

Tubulin

HIF1α

H3K27me3

Tubulin

Elf1Elf1 gNNMT 6.2.2 Grow46 Elf1

Grow46 Elf1 AFElf1 gNNMT 7.2.1

Elf1 gNNMT 7.3.5 H1H1 4iLIFLis1

7.2.1 0.0070.0060.0050.0040.0030.0020.001

Elf1 ct

rl

gNNM

T 6.

2.4

gNNM

T 7.

2.1

0

2.04 × 10–175 4.34 × 10–211 1.7 × 10–135 3.07 × 10–105 7.59 × 10–22 5.87 × 10–5

1.39 × 10–37 4.34 × 10–211 1.06 × 10–58 5.6 × 10–73 0.00887 5.1 × 10–9

Elf1 AF Takashima9 Chan6 Theunissen10 Gafni8H1

–0.4 –0.2 0.0 0.2PC1 (61%)

Naive Primed

C C CC CCC C C CCC C C CC C C C C CC C CC CCA A A A A A A A A A A A AA AT T C T TG G G T T T TG G GG T TT G G G G G G G G G AGG GT T TGG G GGG N G N N N NN

0 240

5′ 3′ WT

7-2-1

WT

7-2-1

250

a

c e

b

f

i

g h j

260 270 350 360 370 380

WT

+1 bp

–119 bp

Elf1 gNNMT 6.2.4 (DNA seq) 5′ 3′ WT

–13 bp

–14 bp

WT

6-2-4

WT

6-2-4

C C CT T C T C T T T T TG C C C C CG G G G GA A A A C CA AN N N G G G GN N N N N N N N

d

∗∗∗∗∗∗

Sum

-nor

mal

ized

SA

M le

vels

Elf1

Elf1 g

NNMT

7.2.

1

Elf1 g

NNMT

6.2.

4

SAM

0

0.05

0.10

0.15

0.20

0.25

Figure 8 NNMT affects the naive-to-primed hESC transition by repressingthe Wnt pathway and activating the HIF pathway. (a,b) Sequencing tracefiles, DNA sequences, protein sequences and three-dimensional proteinstructures predicted from the sequence (Pymol) of various NNMT CRISPR-Cas9 KOmutant clones (gNNMT 7.2.1, a; gNNMT 6.2.4, b). Green representsthe truncated NNMT protein in the CRISPR-Cas9 mutant. (c) Schematicrepresentation of the wild-type NNMT protein and proteins resulting fromthe CRISPR-Cas9 KO mutants gNNMT 7.2.1 and gNNMT 6.2.4. (d) Elf1NNMT CRISPR-Cas9 KO cells have higher amounts of SAM than wild-typeElf1 cells (n=6; s.e.m.; P=1.23×10−05 for gNNMT 7.2.1, P=5.47×10−06

for gNNMT 6.2.4; two-tailed t-test). (e) Western blot analysis reveals higherHIF1α expression and H3K27me3 marks in the Elf1 CRISPR-Cas9 KOmutants gNNMT 7.2.1 and gNNMT 6.2.4 compared with control Elf1 (iCas9)cells. (f) qPCR analysis of the naive marker DNMT3L in wild-type Elf1cells (n=6) and the Elf1 CRISPR-Cas9 KO mutants gNNMT 7.2.1 (n=5)and gNNMT 6.2.4 (n= 3). s.e.m.; P=0.0009 for gNNMT 6.2.4 versusElf1, P=0.027 for gNNMT 7.2.1 versus Elf1; two-tailed t-test. (g) log2

fold expression change of NNMT, Wnt ligands and HIF target genes in Elf1CRISPR-Cas9 KO gNNMT 7.2.1 compared with wild-type Elf1 cells (RNA-seq). (h) PCA plot of CRISPR NNMT knockout line and different naive andprimed lines sequenced in this study. gNNMT 6.2.2 and gNNMT 7.3.5 areheterozygous controls. PC1 explains most of the variation in the data (61%),and the gNNMT 7.2.1 knockdown line moved along the x axis substantiallyaway from other naive lines and towards the primed state. (i) Hypergeometrictest P values for the overlap between genes expressed higher (lower) ingNNMT 7.2.1 compared with Elf1 and genes expressed higher (lower) inprimed lines compared with naive lines from multiple studies. Colour shade isproportional to negative log10 of P values. The gNNMT 7.2.1 transcriptomicsignature has significant overlap with all published primed transcriptomicdata sets, supporting its transition towards a primed stage. (j) Model ofthe intricate relationship between metabolism and epigenetics in hESCs.Unprocessed original scans of blots are shown in Supplementary Fig. 9.For raw data, see Supplementary Table 4. n represents the number ofbiological replicates. ∗P<0.05; ∗∗∗P<0.001.



ART ICLES

A significant increase of H3K27me3 marks was observed in themetabolic gene, prolyl hydroxylase 2 (EGLN1), ECHS1, HIGD1and miR-193 (Figs 5h and 7a) promoters in the primed state aswell as in STAT3-inhibitor-treated Elf1 cells, versus the naive state(Supplementary Table 1R). The increased repressive H3K27me3 markin the EGLN1 promoter correlated with the observed reduced geneexpression in the primed state (Fig. 7a). As EGLN1 induces VHL andubiquitylation-dependent HIF1α degradation, its repression in theprimed state (Fig. 7a) could cause HIF1α stabilization in the primedhESC stage (Figs 7b and 1j). Furthermore, ectopic HIF1α stabilizationaccelerates primed hESC markers (Supplementary Fig. 7A–C). To testwhether HIF1α KO affects the naive-to-primed transition in hESCs,we generated CRISPR-Cas9-based HIF1α KO lines and analysed theircapacity to develop from the preimplantation to postimplantationstage in human development (Fig. 7c–f and Supplementary Fig. 7D).Importantly, the gene expression analysis revealed that naive markerswere upregulated and primedmarkers downregulated, suggesting thatHIF1α KO naive hESCs are not capable of transitioning to the primedstage (Fig. 7g). Furthermore, Seahorse flux analysis showed thatHIF1αKO hESCs do not exhibit an OCR reduction, suggesting that they aredefective in the metabolic switch normally observed in the naive-to-primed hESC transition (Fig. 7h and Supplementary Fig. 7E). Thesedata show thatHIF1α is required for naive-to-primed hESC transition.

To validate the NNMT-based regulation of the Wnt and HIFpathways in the naive-to-primed transition (Fig. 5h) we generatedCRISPR-Cas9-based NNMT KO mutant lines that lack NNMT enzy-matic activity as shownby a lack of 1-MNAproduction and an increasein SAM levels (Fig. 8a–d and Supplementary Fig. 8A–D,G and Fig. 4g).As observed with NNMT KD experiments (Fig. 5d), NNMT KO linesshowed a significant increase in H3K27me3 marks (Fig. 8e). This epi-genetic modification was accompanied by pronounced stabilizationof the HIF1α protein and downregulation of the naive hESC markerDNMT3L (Fig. 8e,f). Furthermore, genome-wide RNA sequencingresults revealed that the NNMT KO line 7-2-1 shows upregulationof HIF pathway genes, downregulation of the Wnt pathway (Fig. 8g)and a genome-wide gene expression signature indicative of a generalshift towards the primed stage, even in the presence of primed stageinhibitors (2iLIF; Fig. 8h,i and Supplementary Fig. 8E-F,H). Downreg-ulationwas also observed among a significant fraction of the 313 geneswith increased repressiveH3K27me3marks in the primed state and af-ter STAT3 inhibitor treatment (hypergeometric test P value< 0.0036).In summary, these data show that NNMT affects the naive-to-primedhESC transition through epigenetic alterations that repress the Wntpathway and activate the HIF pathway.

DISCUSSIONWe show that human naive and primed ESCs exhibit distinctmetabolic profiles and the switch between these metabolic states isregulated by NNMT, which controls the amount of SAM available forPRC2-dependent H3K27me3 histone methylation. Repressive histonemethylation then controls the primed hESC specific metabolismthrough the Wnt and HIF pathways (Fig. 8j). The naive-to-primedhESC transition shows a reduction in Wnt signalling, electrontransport chain activity, and fatty acid beta-oxidation and an increasein mechanisms involved in lipid biosynthesis and HIF1α stabilization.In naive hESCs NNMT and its enzymatic product 1-MNA are

highly upregulated, whereas the substrates nicotinamide and SAMare downregulated, correlating with reduced H3K27me3 marks.Inhibition of the NNMT regulator, STAT3, in naive hESCs increasesH3K27me3 repressive marks in developmental and metabolic genes,including Wnt signalling and the HIF1 repressor, prolyl hydroxylaseEGLN1. Further validations using Wnt pathway inhibitors reveal thatWnt activity is critical for the naive state. The HIF1α KO mutant linegenerated by the CRISPR-Cas9 system is incapable of the naive-to-primed metabolic and fate switch, showing that HIF1α is requiredfor the naive-to-primed hESC transition. NNMT KO naive hESClines show increased H3K27me3 marks, HIF1α stabilization and Wntligand reduction, all indications of a transition towards the primedstate, even in the presence of naive state stabilizers (MEK and GSK3inhibitors and LIF). These data show that NNMT consumes SAMin naive cells, making it unavailable for histone methylation thatrepresses the Wnt pathway and electron transport chain activityand activates the HIF pathway and lipid synthesis, facilitating themetabolic switch in the naive-to-primed hESC transition (Fig. 8j).Therefore, differential metabolites between pluripotent states controlepigenetic dynamics and signalling.

Primed ESCs are dependent on glycolysis14,17,36–39. We nowshow that although early glycolysis metabolites are upregulated, thedownstream metabolites are downregulated in primed-state hESCs,suggesting that metabolites are being channelled off to increase theamount of glycerol backbone available for biosynthesis of lipids inprimed cells, or for the one-carbon cycle for methylation reactionsby SAM. SAM can also be regulated by NNMT, whose enzymaticproduct, 1-MNA is markedly increased in naive versus primed hESCs.As NNMT is considered to create amethyl sink28, reduction of NNMTin primed hESCs can make SAM available as a substrate for DNA andhistonemethylation.We show that a difference in SAM levels betweennaive and primed hESCs correlates with pronounced changes inH3K27me3 marks6,8,10–12 and reveal NNMT as a key regulator of thesechanges. Although previous studies have shown SAM-dependentregulation of histone methylation in stable primed hESCs, the effectwas mainly observed in H3K4me3, not in H3K27me3 marks22. It isplausible that H3K27me3marks, once established, are less dynamic inprimed hESCs than H3K4me3 marks40.

Although H3K27me3 marks are reduced in naive compared withprimed hESCs, the enzymes required for this methylation (EZH2 andEED) are not downregulated. We now show that high NNMT activityin naive hESCs sequesters the methylation substrate, SAM, therebyrepressing H3K27me3 marks. Furthermore, the PRC2-recruitingprotein Jarid2 is upregulated in primed hESCs compared with naive,which may give further specificity to PRC2 action in the naive-to-primed hESC transition41,42 (Figs 1b,l,m and 8j).

In this study we show a direct impact of SAM levels and NNMTfunction on histone marks in naive hESCs, revealing that changesin the metabolic profile of hESCs shape the epigenetic landscapeduring development. Although previous studies have revealed theimportance of Wnt and HIF pathways in naive or primed pluripotentstem cells35,39,43–45, we now show that these pathways are regulated bymetabolite levels. We propose that the availability of SAM triggersthe cascade by activating PRC2 and thereby increasing repressiveH3K27me3 epigenetic marks in the promoters of key regulatorsof the naive-to-primed transition, HIF repressor and Wnt ligands



ART ICLES

(Fig. 8j). These studies pave the way for further understanding anduse of metabolite-specific molecular mechanisms for cell fate changesin general. �

METHODSMethods and any associated references are available in the onlineversion of the paper.

Note: Supplementary Information is available in the online version of the paper

ACKNOWLEDGEMENTSWe thank members of the Ruohola-Baker laboratory for helpful discussionsthroughout this work. We thank D. Djukovic and J. K. Meissen for helpwith mass spectrometry, D. Jones for help with RNA-seq analysis, W. Zhou,A. Nelson, S. Shannon, S. Sidhu, C. Cavanaugh, Y. Zhang, W. Heath, K. Sankary,E. Engelhart and K. Toor for technical help, K. Au for performing karyotypeanalysis, A. Madan for performing RNA-seq, P. Treuting for teratoma analysis,K. Bomsztyk (University of Washington, USA) for providing EED antibody,B. Cravatt (Scripps Research Institute, USA) for providing NNMT overexpressionconstructs and R. Jaenisch (Whitehead Institute for Biomedical Research, USA)and J. Hanna (Weizmann Institute of Science, Israel) for providing WIN1 cellsand LIS1 cells, respectively. This work is supported in part by the Universityof Washington’s Proteomics Resource (UWPR95794). R.T.M. is an Investigator,and A.M.R. and Z.X. are Associates, of the HHMI. This work was supported byfellowship from the American Heart Association to J.M., AG-NS-0577-09 fromthe Ellison Medical Foundation for J.H.B., Schultz Fellowship for H.S., grantsfrom the National Institutes of Health 1U24DK097154 for O.F., RO1ES019319for J.H.B., R01GM097372, R01GM97372-03S1 and R01GM083867 for H.R.-B.,1P01GM081619 for C.A.B., R.T.M., C.B.W., A.M. and H.R.-B., and the NHLBIProgenitor Cell Biology Consortium (U01HL099997; UO1HL099993) for H.R.-B.

AUTHOR CONTRIBUTIONSH.S., J.M., Y.W. and H.R.-B. designed the study. H.S., J.M., Y.W., A.F., J.H., Z.X.,K.A.F., A.D., D.D., H.G., S.L.B.,M.S., C.V., N.G.E., L.M., A.M.R. andD.M. performedall experiments in the study under the guidance of J.H.B., O.F., D.H., C.A.B.,D.R., A.M., R.D.H., R.T.M., C.B.W. and H.R.-B. H.S., J.M., Y.W. and H.R.-B. wrotethe manuscript.

COMPETING FINANCIAL INTERESTSThe authors declare no competing financial interests.

Published online at http://dx.doi.org/10.1038/ncb3264Reprints and permissions information is available online at www.nature.com/reprints

1. Buecker, C. et al. Reorganization of enhancer patterns in transition from naive toprimed pluripotency. Cell Stem Cell 14, 838–853 (2014).

2. Factor, D. C. et al. Epigenomic comparison reveals activation of “Seed”enhancers during transition from naive to primed pluripotency. Cell Stem Cell 14,854–863 (2014).

3. Tesar, P. J. et al. New cell lines from mouse epiblast share defining features withhuman embryonic stem cells. Nature 448, 196–199 (2007).

4. Wu, J. et al. An alternative pluripotent state confers interspecies chimaericcompetency. Nature 521, 316–321 (2015).

5. Brons, I. G. et al. Derivation of pluripotent epiblast stem cells from mammalianembryos. Nature 448, 191–195 (2007).

6. Chan, Y. S. et al. Induction of a human pluripotent state with distinctregulatory circuitry that resembles preimplantation epiblast. Cell Stem Cell 13,663–675 (2013).

7. Duggal, G. et al. Alternative routes to induce naive pluripotency in human embryonicstem cells. Stem Cells 33, 2686–2698 (2015).

8. Gafni, O. et al. Derivation of novel human ground state naive pluripotent stem cells.Nature 504, 282–286 (2013).

9. Takashima, Y. et al. Resetting transcription factor control circuitry toward ground-state pluripotency in human. Cell 158, 1254–1269 (2014).

10. Theunissen, T. W. et al. Systematic identification of culture conditions for inductionand maintenance of naive human pluripotency. Cell Stem Cell 15, 524–526 (2014).

11. Valamehr, B. et al. Platform for induction and maintenance of transgene-free hiPSCsresembling ground state pluripotent stem cells. Stem Cell Rep. 2, 366–381 (2014).

12. Ware, C. B. et al. Derivation of naive human embryonic stem cells. Proc. Natl Acad.Sci. USA 111, 4484–4489 (2014).

13. Bracha, A. L., Ramanathan, A., Huang, S., Ingber, D. E. & Schreiber, S. L.Carbon metabolism-mediated myogenic differentiation. Nat. Chem. Biol. 6,202–204 (2010).

14. Folmes, C. D. et al. Somatic oxidative bioenergetics transitions into pluripotency-dependent glycolysis to facilitate nuclear reprogramming. Cell Metab. 14,264–271 (2011).

15. Greer, S. N., Metcalf, J. L., Wang, Y. & Ohh, M. The updated biology of hypoxia-inducible factor. EMBO J. 31, 2448–2460 (2012).

16. Mathieu, J. et al. Hypoxia-inducible factors have distinct and stage-specificroles during reprogramming of human cells to pluripotency. Cell Stem Cell 14,592–605 (2014).

17. Panopoulos, A. D. et al. The metabolome of induced pluripotent stem cellsreveals metabolic changes occurring in somatic cell reprogramming. Cell Res. 22,168–177 (2012).

18. Rafalski, V. A., Mancini, E. & Brunet, A. Energy metabolism and energy-sensingpathways in mammalian embryonic and adult stem cell fate. J. Cell Sci. 125,5597–5608 (2012).

19. Yanes, O. et al. Metabolic oxidation regulates embryonic stem cell differentiation.Nat. Chem. Biol. 6, 411–417 (2010).

20. Zhou, W. et al. HIF1α induced switch from bivalent to exclusivelyglycolytic metabolism during ESC-to-EpiSC/hESC transition. EMBO J. 31,2103–2116 (2012).

21. Shyh-Chang, N. et al. Influence of threonine metabolism on S-adenosylmethionineand histone methylation. Science 339, 222–226 (2013).

22. Shiraki, N. et al. Methionine metabolism regulates maintenance and differentiationof human pluripotent stem cells. Cell Metab. 19, 780–794 (2014).

23. Yan, L. et al. Single-cell RNA-Seq profiling of human preimplantation embryos andembryonic stem cells. Nat. Struct. Mol. Biol. 20, 1131–1139 (2013).

24. Berra, E. et al. HIF prolyl-hydroxylase 2 is the key oxygen sensor setting low steady-state levels of HIF-1α in normoxia. EMBO J. 22, 4082–4090 (2003).

25. Simonson, T. S. et al. Genetic evidence for high-altitude adaptation in Tibet. Science329, 72–75 (2010).

26. Nguyen-Tran, D. H. et al. Molecular mechanism of sphingosine-1-phosphate actionin Duchenne muscular dystrophy. Dis. Model. Mech. 7, 41–54 (2014).

27. Opitz, C. A. et al. An endogenous tumour-promoting ligand of the human arylhydrocarbon receptor. Nature 478, 197–203 (2011).

28. Ulanovskaya, O. A., Zuhl, A. M. & Cravatt, B. F. NNMT promotes epigeneticremodeling in cancer by creating a metabolic methylation sink. Nat. Chem. Biol.9, 300–306 (2013).

29. Mathieu, J. et al. Hypoxia induces re-entry of committed cells into pluripotency. StemCells 31, 1737–1748 (2013).

30. Kraus, D. et al. Nicotinamide N-methyltransferase knockdown protects against diet-induced obesity. Nature 508, 258–262 (2014).

31. Graf, U., Casanova, E. A. & Cinelli, P. The role of the Leukemia Inhibitory Factor(LIF)—pathway in derivation and maintenance of murine pluripotent stem cells.Genes (Basel) 2, 280–297 (2011).

32. Tomida, M., Ohtake, H., Yokota, T., Kobayashi, Y. & Kurosumi, M. Stat3 up-regulatesexpression of nicotinamide N-methyltransferase in human cancer cells. J. CancerRes. Clin. Oncol. 134, 551–559 (2008).

33. Gilles, C. et al. Transactivation of vimentin by β-catenin in human breast cancer cells.Cancer Res. 63, 2658–2664 (2003).

34. Davidson, K. C. et al. Wnt/β-catenin signaling promotes differentiation, not self-renewal, of human embryonic stem cells and is repressed by Oct4. Proc. Natl Acad.Sci. USA 109, 4485–4490 (2012).

35. ten Berge, D. et al. Embryonic stem cells require Wnt proteins to preventdifferentiation to epiblast stem cells. Nat. Cell Biol. 13, 1070–1075 (2011).

36. Prigione, A. & Adjaye, J. Modulation of mitochondrial biogenesis and bioenergeticmetabolism upon in vitro and in vivo differentiation of human ES and iPS cells. Int.J. Dev. Biol. 54, 1729–1741 (2010).

37. Varum, S. et al. Energy metabolism in human pluripotent stem cells and theirdifferentiated counterparts. PLoS ONE 6, e20914 (2011).

38. Zhang, J. et al. UCP2 regulates energy metabolism and differentiation potential ofhuman pluripotent stem cells. EMBO J. 30, 4860–4873 (2011).

39. Zhou, W. et al. Assessment of hypoxia inducible factor levels in cancer celllines upon hypoxic induction using a novel reporter construct. PLoS ONE 6,e27460 (2011).

40. Trojer, P. & Reinberg, D. Histone lysine demethylases and their impact on epigenetics.Cell 125, 213–217 (2006).

41. Escobar, T. M. et al. miR-155 activates cytokine gene expression in Th17 cells byregulating the DNA-binding protein Jarid2 to relieve polycomb-mediated repression.Immunity 40, 865–879 (2014).

42. Landeira, D. & Fisher, A. G. Inactive yet indispensable: the tale of Jarid2. Trends CellBiol. 21, 74–80 (2011).

43. Blauwkamp, T. A., Nigam, S., Ardehali, R., Weissman, I. L. & Nusse, R. EndogenousWnt signalling in human embryonic stem cells generates an equilibrium of distinctlineage-specified progenitors. Nat. Commun. 3, 1070 (2012).

44. Clevers, H., Loh, K. M. & Nusse, R. Stem cell signaling. An integral program fortissue renewal and regeneration: Wnt signaling and stem cell control. Science 346,1248012 (2014).

45. Mazumdar, J. et al. O2 regulates stem cells through Wnt/β-catenin signalling. Nat.Cell Biol. 12, 1007–1013 (2010).

46. Grow, E. J. et al. Intrinsic retroviral reactivation in human preimplantation embryosand pluripotent cells. Nature 522, 221–225 (2015).

47. ENCODE Project Consortium. An integrated encyclopedia of DNA elements in thehuman genome. Nature 489, 57–74 (2012).

48. Bernstein, B. E. et al. The NIH roadmap epigenomics mapping consortium. Nat.Biotechnol. 28, 1045–1048 (2010).



http://dx.doi.org/10.1038/ncb3264





METHODS DOI: 10.1038/ncb3264

METHODSCulture of primed and naive embryonic stem cells. Primed hESCs (H1 (WA-01, WiCell) and H7 (WA-07, WiCell)) and naive hESCs (Elf1 (NIHhESC-12-0156,University of Washington), WIN1 (NIHhESC-14-0299, MIT)) and Lis1 (ref. 8)(Weizmann Institute of Science) were cultured as previously described10,12. Briefly,the cells were cultured on a feeder layer of irradiated primary mouse embryonicfibroblasts (MEFs) in hESC media (Supplementary Table 3). One passage before theexperiments, the cells were transferred to growth-factor-reduced Matrigel (BectonDickinson) in MEF-conditioned media (CM). Reverse toggling of H1 and H7cells was performed using supplemented media (Supplementary Table 3). H1 cellswere pushed towards a more naive state by culture in 4iLIF (modified from Gafniet al.: 1 µM GSK3 inhibitor (CHIR99021), 1 µM of MEK inhibitor (PD0325901),5 µM JNK inhibitor (SP600125), 2 µM p38 inhibitor (BIRB796), 10 ngml−1 humanLIF, 5 ngml−1 IGF and 10 ngml−1 bFGF) for 3 passages (Supplementary Fig. 1D).In addition, human naive cells (Elf1 and WIN1) were pushed towards a moreprimed state by culturing them in either TeSR1 (STEMCELLTechnologies), or bFGF(10 ngml−1) with orwithout activinA (10 ngml−1) for 3 passages (Elf1AF,WIN1AF,WIN1 F, WIN1 TeSR). Mouse ESCs (R1, EpiSC, R1 AF) were cultured as describedin Supplementary Table 3. All cells were grown at 37 ◦C, 5% CO2 and 5% O2. Cellshave been karyotyped by Diagnostic Cytogenetics, and were tested for mycoplasmadetection using MycoAlert detection kit (Lonza).

OCR and ECAR measurement using Seahorse cellular flux assays. Naiveand primed ESCs were seeded onto 96-well Seahorse plates pre-coated withMatrigel at 25× 104 or 40× 104 cells per well. Culture media were exchangedfor base media (unbuffered DMEM, Sigma D5030) supplemented with sodiumpyruvate (Gibco, 1mM) and with 25mM glucose (for the MitoStress assay),25mM glucose and 50 µM carnitine (for palmitate assay), or 2mM glutamine(for glucose stress assay) 1 h before the assay. Substrates and selective inhibitorswere injected during the measurements to achieve final concentrations ofglucose (2.5mM), 4-(trifluoromethoxy)phenylhydrazone (FCCP, 300 nM–500 nM),oligomycin (2.5 µM), antimycin (2.5 µM), rotenone (2.5 µM), palmitate (50 µM inBSA), BSA and ETO (50 µM). The OCR and ECAR values were normalized to thenumber of cells present in each well, quantified by the Hoechst staining (HO33342;Sigma-Aldrich). Changes inOCRandECAR in response to the addition of substratesand inhibitors were defined as the maximal change after the chemical injectioncompared with the last OCR value before the injection.

MitochondrialDNAmutation frequency and copynumber analysis.TheDNAof Elf1 and H7 cells was isolated using DNAzol (Invitrogen). TaqMan primers wereused to quantify mitochondrial and genomic DNA (Supplementary Table 2). Theratio of mtDNA to genomic DNA was determined using a standard curve for eachprimer. Each reaction contained 2 ng of DNA extract, 1× TaqMan Universal PCRMasterMix NoAmpErase UNG, 500 nM of each primer, and 200 nM of the TaqManprobe. Using the 7300 real-time PCR system (Applied Biosystems), the reactionswere amplified by incubation at 50 ◦C for 2min, 95 ◦C for 10min, and then 40cycles of 15 s at 95 ◦C followed by 1min at 60 ◦C where the intensity of fluorescencewas measured.

Naive hESCs (Elf1) and primed hESCs (H1 and Elf1 AF) were grown in triplicatefor mutation analysis. Elf1 were analysed between passage 19 and 23, Elf1 AF wereanalysed at passage 25 and H1 at passage 65. All lines were grown on Matrigel forthe last passage before analysis.

DNAwas isolated fromhESCswith theDNeasy Blood andTissueKit (QIAGEN).Raremutation-bearingmolecules were selectively enriched through endonucleolyticdestruction of wild-type target sites by sequential additions of TaqI (New EnglandBiolabs). Complete cleavage of wild-type TaqI sites was verified by PCR followedby restriction digest and gel electrophoresis. Reaction droplets were prepared asdescribed previously49, with primers and probe sets specific to each target region(Supplementary Table 2). Fragments for point mutation detection and mtDNAcopy number measurement were amplified as follows: 95 ◦C for 10min, followedby 40 cycles of 94 ◦C for 30 s, and 60 ◦C for 1min. For digital deletion detection,thermal cycling was as follows: 95 ◦C for 10min, followed by 50 cycles of 94 ◦Cfor 30 s, and 63.5 ◦C for 2min. The thermally cycled droplets were analysed byflow cytometry in a QX100 Droplet Digital Reader (Bio-Rad) for fluorescenceanalysis and quantification of mutation frequencies and mtDNA copy number, asdescribed elsewhere49.

Proteomics. Naive hESCs (Elf1 2iLIF) and primed hESCs (Elf1 AF) were washedin 1×PBS and flash frozen. Cell pellets were lysed in 1M urea, 50mM ammoniumbicarbonate, pH 7.8, and heated to 50 ◦C for 20min.Normalized quantities of proteinwere reduced, alkylated, and digested overnight with trypsin. The resulting peptideswere desalted on Waters Sep-Pak C18 cartridges. Peptides were measured by nano-LC–MS/MS on a Thermo Scientific Fusion Orbitrap. Peptides were separated byreverse-phase chromatography in a 180min gradient (1–45% acetonitrile). The

Fusion was operated in data-dependent mode with the following settings: 60,000resolution, 400–1,600m/z full scan, Top Speed 3 s, and an 1.8m/z isolation window.Identification and label-free quantification of peptides was done with MaxQuant1.5 (ref. 50) using a 1% false discovery rate (FDR) against the human Swiss-Prot/TrEMB database downloaded from Uniprot on 11 October 2013. We analysedtwo biological and three technical replicates per condition. Proteins that weresignificantly regulated between conditions were identified using a permutation-based t-test (S1, FDR 5%) in Perseus 1.4.1.3.

Non-targeted GC–TOF and LC–QTOF analysis for metabolites in mouseand human ESCs. For the first set of experiments (Fig. 2b–f) 5–20 million cellsper replicate (grown on ‘ghost’: irradiated MEFs lysed using a detergent solution0.5% Triton and 0.034% (v/v) NH4OH (Sigma-Aldrich) to retain their extracellularmatrix) were scraped in PBS, pelleted in 6 replicates per condition, and frozen at−80 ◦C. For the first set of experiments cells were thawed on ice and mixed with2ml of ice-cold degassed acetonitrile, and then vortexed for 20 s and sonicated for5min. One millilitre of material was taken and centrifuged for 5min at 14,000 rcf(Eppendorf 5415D Centrifuge). Supernatant was divided into a 500 µl (GC–TOF)and a 250 µl (LC–QTOF) aliquot. Lyophilized aliquots were resuspended in 500 µl of1:1 acetonitrile/H2O and centrifuged. Supernatants were lyophilized until analysis.For the second set of experiments cells were thawed on ice and extracted aspreviously reported51. The upper layer was used for LC–MS analysis and the bottomlayer was used for GC–MS analysis; both layers were evaporated to dryness.

For mass spectrometry-based untargeted metabolomics, primary metaboliteswere derivatized in 40 µgml−1 methoxyamine/pyridine and subsequentlytrimethylsilylated. GC–TOF MS analysis was performed as previously described52,53

using a Leco Pegasus IV time-of-flight MS coupled to an Agilent 6890 GC equippedwith a 30-m-long 0.25mm id Rtx5Sil-MS column and a Gerstel MPS2 automaticliner exchange system. Data were processed by ChromaTOF and BinBase filteringfor metabolite identifications52.

LC–QTOF MS was used for analysis of complex lipids. For the first set ofexperiments, lyophilized material was redissolved in 100 µl initial LC gradientsolvent and analysed within 24 h. HILIC and reversed-phase LC–QTOF analysisand data processing were performed as previously described53 using an Agilent 1200series HPLC equipped with either an Agilent Zorbax Eclipse Plus C18 2.1×150mmcolumn for reversed phase or a Waters 1.7 µm Acquity BEH HILIC 2.1× 150mmcolumn. LC eluents were analysed with an Agilent 6530 accurate mass Q-TOFmass spectrometer. For the second set of experiments the lipid extracted phasewas analysed as previously described51. Method blanks and human pooled plasmasamples were used as QC controls. MZmine 2.10 was used to process the raw dataand metabolites were reported when present in 50% of the samples in each group.Annotations were made based on an in-house accurate mass and retention timelibrary created using LipidBlast, described previously52,54.

Multivariate analysis of primed versus naive metabolomes. Metabolitemeasurements of known and unknownGCmetabolites and known lipidmetabolites(from ESI(+) and ESI(−) modes) were submitted using R to DeviumWeb (v 0.3.2;ref. 51) and normalized using unit norm normalization. O-PLS-DA, a multivariateclassification model, was used to identify differences between primed and naivehuman andmouse cells. Robust model performance statistics were generated by 100rounds of Monte Carlo cross-validation using training and testing compared withthe permuted model (random chance).

LC–QTOF for Elf1 and H1 hESCs. For lipid extraction Elf1 and H1 cells weregrown onMatrigel for one passage. Cells were washed with PBS and 37 ◦C deionizedwater followed by the addition of 0.5ml of a−75 ◦C solution of internal standards52and incubation on dry ice for 15min. Cells were scraped into Eppendorf tubes and1ml of chloroformwas added, followed by 15min incubation on dry ice and spun for5min at 4 ◦C at 18,000 rcf (Beckman Coulter Microfuge 18), after which the lowerphase was collected and stored at−80 ◦C.

LC–QTOFMS experiments were performed using an Agilent 1200 SL LC systemcoupled online with an Agilent 6520 Q-TOF mass spectrometer. Each sample (4 µlfor positive ESI ionization, 8 µl for negative ESI ionization) was injected onto anAgilent Zorbax 300 SB-C8 column, which was heated to 50 ◦C.

Targeted LC–QQQ MS analysis for water-soluble metabolites. TargetedLC–QQQ MS analysis was performed to detect a different set of water-solublemetabolites to complement the other non-targeted mass spectrometry experiments.Elf1 and H1 cells were grown on Matrigel for one passage, washed with PBSand ice-cold deionized water followed by the addition of a −75 ◦C 0.75ml 9:1methanol/chloroform solution. The plates were incubated on dry ice for 15minbefore scraping into Eppendorf tubes, whichwere spun at 18,000g (BeckmanCoulterMicrofuge 18) for 5min at 4 ◦C. All soluble extract was transferred into a newEppendorf tube and vacuum dried. Samples were stored at−80 ◦C.

NATURE CELL BIOLOGY


DOI: 10.1038/ncb3264 METHODS

Dried samples were reconstituted in 200 µl 5mM ammonium acetate in40% water/60% acetonitrile + 0.2% acetic acid, and filtered through 0.45 µmPVDF filters (Phenomenex) before LC–MS analysis. LC–MS/MS was performedusing an Agilent 1260 LC AB Sciex 5500 QQQ MS. Both chromatographicseparations were performed in HILIC mode on two SeQuant ZIC-cHILIC columns.The mobile phase was composed of Solvents A (5mM ammonium acetate in90%H2O/10% acetonitrile + 0.2% acetic acid) and B (5mM ammonium acetate in90%acetonitrile/10%H2O+ 0.2% acetic acid). The chromatographic separation,MSionization and data acquisition were performed using an AB Sciex QTrap 5500massspectrometer equipped with an electrospray ionization (ESI) source. The instrumentwas controlled by Analyst 1.5 software. Targeted data acquisition was performedin multiple-reaction-monitoring (MRM) mode. The extracted MRM peaks wereintegrated using MultiQuant 2.1 software.

TargetedHILIC–QTOFmass spectrometrymetabolite quantifications ofme-thionine metabolites. Cells were grown on Matrigel for one passage, scraped andwashed with PBS at room temperature, pelleted and flash frozen in liquid nitrogen.Samples were extracted by adding 1ml cold 3:1 cold methanol/water to the cellpellet, vortexed, placed at−20 ◦C for 30min, and centrifuged for 10min at 14,000 rcf(Eppendorf 5415D Centrifuge). The supernatant was transferred and then cen-trifuged again, and then the supernatant was evaporated to dryness. Samples wereresuspended in 80:20 acetonitrile/water containing Val-Try-Val. Standard curvedilutions for quantifications were prepared using amixture of 1-methylnicotinamideHCl (1-MNA), S-methyl-5′-thioadenosine (MTA), S-adenosyl methionine (SAM),S-adenosyl homocysteine (SAH), methionine, kynurenine and tryptophan (Sigma).

Hydrophilic interaction chromatography (HILIC) analysis of standard curvesand samples was performed using an Agilent 1290 Infinity Ultrahigh PressureLiquid Chromatography stack equipped with an auto-sampler (4 ◦C) using 5 µlinjections into an Acquity UPLC BEHAmide column (Waters Corporation). Mobilephases were prepared with 10mM ammonium formate and 0.125% formic acidin either 100% LC–MS-grade water for mobile phase A or 95:5 acetonitrile/waterfor mobile phase B. Metabolites were detected and quantified by an Agilent 6530accurate mass quadrupole time-of-flight (QTOF) mass spectrometer with a jetstream ESI source in positive ion mode. Mass calibration was maintained byconstant reference ion infusion, with MS data collected at 4 spectra s−1. Data fileswere analysed using Agilent Mass Hunter TOF Quantitative Analysis software.Peak filtering was performed manually to eliminate peaks with a signal-to-noiseratio of less than 3. Retention times and major adducts for each compound areas follows: 1-MNA (m/z 137.0715) 6.345min, MTA (m/z 297.0896) 2.583minM + H, tryptophan (m/z 204.0899) 6.904min M + H, kynurenine (m/z 208.0848)6.971min M + H & M + Na, methionine (m/z 149.0511) 7.493min M + H &M + 2Na + H, SAH (m/z 384.1216) 8.810min M + H, SAM (m/z 399.1451)9.768min. Metabolites at undetectable levels and metabolites whose levels saturatedthe system were given the lowest and highest detectable values respectively insubsequent analysis.

Metabolite levels were sum-normalized for each sample using the methioninemetabolite values (methionine, nicotinamide, MTA, 1-MNA, SAM and SAH).P values were calculated using a one-tailed t-test.

Transcriptomic data analysis. RNA-seq data processing was performed accordingto Takashima and colleagues9. Raw RNA-seq reads from this study and 3 otherstudies (Chan et al.6, Takashima et al.9 and Yan et al.23) were aligned to hg19/GRCh37with STAR aligner55. Transcript quantificationwas performedwith htseq-count fromthe HTSeq package56 using GENCODE v15. Differential expression analysis wasperformedwithDESeq after filtering out geneswhose total read count across samplesis below the 40th quantile of all genes.

Size factors used to normalize by library size were computed using the DESeqpackage57. Reads were further normalized by gene length.

Affymetrix Human Gene Array 1.0 ST arrays from Gafni et al.8 were processedwith the oligo package58 and normalized using Robust Multi-array Average59.Multiple probes mapping into the same gene were summarized into a singleexpression value by taking the maximum. Affymetrix PrimeView arrays fromTheunissen et al.10 were processed with the Affy package60 and normalized withRMA. Microarray differential expression analysis was performed using the limmapackage. RNA-seq and microarray data were combined as previously described9,and expression levels were converted to log2 fold change relative to the mean ofhuman embryo-derived PSC samples within each study. One-to-one orthologousgenes between mouse–human were mapped as previously described9. PCA plots ofall samples from all studies were generated using the princomp function from the Rstats package.

An alternative PCA analysis was performed where the ComBat tool61 wasapplied to correct for batch effects and naive samples were not normalized toprimed samples. PCA was applied on batch effect-corrected, gene-wise mean-centred expression values.

Global metabolomic data analysis. All global metabolomic data were mean-centred within each sample. The prcomp function in R is used for principlecomponent analysis of metabolomics data. Differentially abundant metabolites weredefined as metabolites with a twofold change in abundance and a Benjamini–Hochberg adjusted false discovery rate < 0.2.

For the lipidomics data, features missing in more than half of all samples (4 ormore out of 6) were removed from further analysis. Missing values were replacedwith minimum detected values within each sample before mean centring.

ChIP-seq data analysis. ChIP-seq data of H3K27me3 H3K4me3, H3K9me3 andH3K27ac modifications from Chan et al.6, Gafni et al.8, Theunissen et al.10 andBernstein et al.48 were downloaded from Array Express, GEO and the ENCODEproject website. Reads were aligned to hg19 using Bowtie version 1.0.0. allowing1 mismatch (-N 1). ngsplot was used to generate plots of reads around 5 kb oftranscription start sites of a priori defined developmental genes. Reads of replicatesamples for the same cell type were merged for ngsplot. Reads with mappingquality above 20 were used by ngsplot. Differentially marked genomic regions wereidentified with diffReps version 1.55.4 (ref. 62) and annotated to the closest genes.Genes associatedwith at least one significant genomic region (FDR less than 0.05 andfold change > 1) were classified as differentially marked. When a gene is annotatedwith multiple significant genomic regions, the most significant one is assigned tothat gene.

Lipid droplet visualization using Oil red O and BODIPY staining. Naive andprimed ESCs were fixed with 4% PFA at room temperature for 10min, washed twicewith PBS and stained with Oil red O dye (Sigma) for 10min at room temperature.Alternatively, lipid droplets were stained using BODIPY 493/503 (Molecular Probes)for 15min on a rocking platform at room temperature. Pictures were taken using afluorescentmicroscope (Leica). Lipid droplet analysis at the in vivo postimplantationstage has proved to be difficult.

Protein extraction and western blot analysis. Cellular extracts were preparedusing a lysis buffer containing 20mMTrisHCl (pH 7.5), 150mMNaCl, 15%glycerol,1% Triton, 25mM β-glycerolphosphate, 50mM NaF, 10mM Na pyrophosphate,orthovanadate, phenylmethylsulphonyl fluoride (all chemicals are from Sigma-Aldrich), Protease inhibitor cocktail (Roche) and 2% SDS. Twenty-five units ofbenzonase nuclease (EMD Chemicals) and 20mM of dithiothreitol (Sigma) wereadded to the lysis buffer right before use. Fifteenmicrograms of protein (determinedby Bradford) was loaded, separated by 4–20% SDS–PAGE, and transferred topolyvinylidene difluoride membranes, blocked with 5% non-fat dry milk for 60minat room temperature, and incubated overnight at 4 ◦C with primary antibody.After incubation for one hour with horseradish peroxidase-conjugated secondaryantibodies, they were visualized by enhanced chemiluminescence (Millipore Corp).Antibodies used in this study are: H3K27me3 (1:1,000, Abcam, ab6002), H3K9me3(1:1,000, Abcam, ab8898), H3K9/14Ac (1:1,000, Cell Signaling, 9677s), EED(1:1,000, gift from K. Bomsztyk63), HIF1α (1:2,000, BD Biosciences, 610958), LDHA(1:1,000, Cell Signaling, 3582), JARID2 (1:1,000, Cell Signaling, 13594), pSTAT3(1:1,000, Abcam, Ab76315) and γ-tubulin (1:10,000, Promega, G712A).

RNA extraction and qPCR analysis. RNA was extracted using Trizol andanalysed by SYBRgreen qPCR with the 7300 real-time PCR system (AppliedBiosystems)64 and TaqMan qPCR (Applied Biosystems). Primers used are listed inSupplementary Table 2.

qPCR of miRNAs was conducted using TaqMan miRNA assays (AppliedBiosystems). Raw Ct values for miRNAs were normalized to RNU (ref. 63). Linearexpression values for all qPCR experiments were calculated using the 2(−1Ct) method.P values were calculated using Student’s t-test.

ChIP-seq experiment. Naive hESCs Elf1 2iLIF grown on Matrigel were treatedwith 100 µM of STAT3 inhibitor (Selleckchem) for 6 h or 24 h and analysed formethylation marks by western blot and ChIP Seq. For ChIP-seq analysis, cells werecrosslinked and chromatin processed as previously described64 with minor modifi-cations. Briefly, cells were collected with accutase and crosslinked in suspensionwith1% formaldehyde solution for 10min at room temperature. Reaction was quenchedwith glycine and crosslinked cells were rinsedwith ice-cold PBS. Nuclei were isolatedand chromatin sonicated using a Covaris E210 to approximately 200–500 bp sizerange. ChIP-seq was conducted as previously described64 with minor modifications.Briefly, magnetic Dynabeads were incubated overnight rotating at 4◦Cwith antibodyagainstH3K27me3 (ActiveMotif, cat. no. 39155). Sonicated chromatin fromapprox-imately 200,000 cells was added to the bead-bound-antibodies and incubated at 4◦Crotating overnight. Beads were washed and bound chromatin was eluted from beadsand reverse crosslinked overnight. Purified DNA was prepared for next-generationsequencing through end repair, A-tailing, ligation of custom Y-adapters and PCRamplification to generate the final DNA library following gel size selection.

NATURE CELL BIOLOGY


METHODS DOI: 10.1038/ncb3264

Generation of BAR-Elf1 reporter cell line. Elf1 cells grown in naive media (2iFor 2iLIF) or primed media (AF) were infected with BAR reporter lentivirus34,65 andseeded onto Matrigel-coated plates in MEF-CM with 10 µM Y-27632, and 1 µMthiazovivin (ROCK inhibitors, Tocris). Transduced Elf1 cells were cultured for aweek on Matrigel, and then passaged onto MEF plates for further selection andexpansion. Single Elf1 naive reporter cells were collected using TrypLE Expressand FACS sorted for the population with both Venus- and DsRed-positive signals.DsRed-positive colonies of Elf1 primed reporter cells were transferred onto MEFplates, and the same positive selection was repeated 1–2 more rounds. Negativecolonies were removed as a negative selection.

Manipulation of Wnt pathway. Wnt secretion and signalling were inhibited innaive hESCs (Elf1, WIN1) by treatment with IWP2 (2 µM, Tocris) or XAV939(5 µM, Sigma). The Wnt pathway was activated in primed Elf1 AF reporter cellsusing a GSK3 inhibitor, CHIR99021 (72 h, 10 µM, AxonMedChem). Both IWP2 andCHIR99021 were reconstituted in dimethylsulphoxide (DMSO).

Production of conditioned medium (LCM and Wnt3A-CM). L and L-Wnt3Acells (ATCC) were cultured in 15 cm plates in 10% FBS/DMEM media until ∼90%confluent. Medium was collected every 48 h for three batches. Biological activity ofsecretedWnt3A in the individual batches of the conditional medium was confirmedin 293T-BAR reporter cells34, and then batches were pooled and filtered. Primed(Elf1 AF) reporter cells were grown on Matrigel with 50% LCM or 50%Wnt3A-CMfor 3 days before taking bright-field and fluorescent pictures (Leica microscope).

RNA interference experiments.Naive Elf1 2iLIF cells were transfected onMatrigelin MEF-CM supplemented with ROCK inhibitors (Tocris) using LipofectamineRNAiMAX (Life Technologies). siRNA targeting NNMT (Hs-NNMT-8) waspurchased from Qiagen as Flexitube siRNA premix, and siRNA targeting luciferasewas used as control. siRNAs against NNMT and luciferase were used at 50 nMfinal concentration. Protein and RNA were extracted 72 h after transfection. siRNAtargeting beta-catenin (Invitrogen, CTNNB1, Silencer Select ID s437) and SilencerSelect Negative Control 1 (Invitrogen) were transfected in naive Elf1 2iLIF cellsat a 10 nM final concentration following a reverse transfection protocol. Bright-field and fluorescence images were taken after 3 days. Efficacy was confirmed byqPCR analysis.

Overexpression of NNMT. Naive hESCs were transfected with NNMToverexpression construct or inactive NNMT mutant overexpression construct(Y20A; ref. 28). Cells were plated the following day into Matrigel-coated Seahorseplates with primed hESCmedia (conditionedmedia+AF). TheMitoStress protocolin the Seahorse flux analyser was performed 2 days later. Alternatively, primedhESCs Elf1 AF were transfected with NNMT overexpression construct28 andproteins were extracted 3 days later for detection of H3K27me3 marks by westernblot analysis.

Treatment of naive hESCs with themetabolite SAM.WIN1 cells were seeded inSeahorse plates 2 days before change of media with media without L-methionine(Sigma-Aldrich 0422 supplemented with 0.584 gm l−1 L-glutamine (Invitrogen))and addition of SAM (500 µM), and 5 h later the Seahorse MitoStress protocolwas performed.

HIF1α overexpression in naive hESCs. Naive hESCs (Elf1) were infected with anon-degradable form of HIF1α overexpressing construct (Addgene plasmid 19005,Yan et al.23) or a pBABE empty vector construct in the presence of 4 ngml−1Polybrene (Invitrogen). Twenty-four days later RNA and proteins were collectedor medium was changed into naive hESC media with LIF but without GSK3 andMEK inhibitors (2i). Pictures were taken 3 days after infection and primed-likemorphology of colonies quantified.

CRISPR-Cas9-based NNMT and HIF1 KO lines. Six gRNAs were designedfor each of the genes NNMT and HIF1α using the GECKO library and Zhangwebsite (http://mit.edu) and annealed and ligated into the LentiCRISPRv2 (http://Addgene.com), which was previously digested with BsmB1 and dephosphorylated.The validated constructs were transfected using GeneJuice to naive hESCs iCas9Elf1 (ref. 66) treated with doxycycline (2 µgml−1) for 1 or 2 days before andduring transfection. On day 3 the cells were trypsinized and replated on MEF-coated plates. On day 6 single colonies were picked and amplified as describedin Supplementary Fig. 8. Genomic DNA was collected using DNAzol reagent(Invitrogen). NNMT KO 6.2.4 mutant was generated by transfecting in vitrotranscribed gRNA to naive hESC Elf1 iCas9 cells66. T7-gRNA in vitro transcription(IVT) template (120 bp) was generated using T7 promoter-gRNA forward primerwith a reverse primer against the scaffold. T7-gRNA PCR products were used astemplates for IVT (MAXIscript T7 kit, Applied Biosystems). iCas9 Elf1 cells were

treated with doxycycline (2 µgml−1) for 1 or 2 days before and during transfection.For transfection, cells were dissociated with trypsin, replated onto MEF-coatedplates, and transfected in suspension with gRNAs using Lipofectamine RNAiMAX(Life Technologies). gRNA was added at a 10 nM final concentration. A secondtransfection was performed after 24 h. Two days after the last gRNA transfection,iCas9 Elf1 cells were dissociated into single cells and replated onto MEF-coatedplates. Single colonies were randomly selected and amplified. Genomic DNA wascollected using DNAzol. Genomic regions flanking the CRISPR target sites werePCR amplified, purified and sent to Genewiz for sequencing. Alternatively, Samtoolsmpileup and BCF tools were used to identify variants on the basis of aligned RNA-seq BAM files.

Teratoma formation. Naive hESCs H1 4iLIF were cultured on Matrigel-coatedplates. Cells (2× 106) were resuspended in Matrigel supplemented with a cocktailof prosurvival factors29 and injected into the femoral muscle of female 43–49-day-old SCID-Beige mice (Charles River). Palpable tumour masses developed inapproximately 5 weeks. The tumour-bearing mice were euthanized, and tumourtissue was fixed in 10% formalin for 24 h and stored in 70% ethanol until paraffinembedding. Five-micrometre sections of the tumourwere stainedwith haematoxylinand eosin (Supplementary Fig. 1D). The experiment was performed in compliancewith ethical regulations, IACUC protocol no. 4152-01.

Embryonic body formation. H1 4iLIF naive-toggled cells were trypsinized andtransferred into low-attachment plates in differentiation medium (DMEM, high-glucose, 20% FBS, 0.1mM NEAA, 2mM L-glutamine, 1mM sodium pyruvate, 1%Pen/Strep, 0.1mM β-mercaptoethanol). The medium was changed every other dayandwhole genome RNA-seq was performed after 2 weeks (Supplementary Table 1S).

Accession numbers. RNA-seq and ChIP-seq data sets generated for this studyare available from the NCBI GEO database under accession number GSE60955.Gene expression and ChIP-seq data from Grow et al.46 (GSE63570), Gafni et al.8(GSE52824), Chan et al.6 (E-MTAB-2031 and E-MTAB-2041), Theunissen et al.10(GSE59435), and Takashima et al.9 (GSE60945) were also analysed. mRNA-seq, microRNA-seq and ChIP-seq of H1 and ChIP-seq of H9 hESC cell linesfrom the ENCODE project were downloaded from the ENCODE project websitehosted at the UCSC genome browser (https://genome.ucsc.edu/ENCODE). RNA-seq data of multiple lineages derived from H1 hESCs were downloaded fromrelease 9 of the Roadmap Epigenome project (http://www.genboree.org/EdaccData/Current-Release).

Reproducibility of experiments. The number of independent experiments foreach figure panel is described in the corresponding figure legend and raw data areavailable in Supplementary Table 4. For HIF1α, LDHA and JARID2 western blotsin Elf1, Elf1 AF and H7, there were 3 independent experiments (Fig. 1j,m). ForBODIPY 493/503 staining and Oil red O of lipid droplets in primed and naivehuman (Fig. 3d and Supplementary Fig. 3A,B) and mouse (Fig. 3e) ESCs, therewere 4 independent experiments. For western blots of histone marks and EED inElf1, Elf1 AF and H7 (Fig. 5b) there were 3 to 8 independent experiments. Forwestern blots of histonemarks in Elf1 and Elf1 treated with siRNA against NNMTorluciferase control (Fig. 5d) and Elf1 treated with STAT3 inhibitor or DMSO control(Fig. 5f) there were 3 independent experiments. ForWnt sensor analysis in naive andprimed cells, there were 3 independent experiments (Fig. 6b,c and SupplementaryFig. 6E,F). For western blot analysis of HIF1α expression in cells pushed towardsthe primed stage (Fig. 7f), there were 3 independent experiments. For western blotof HIF1α expression and H3K27me3 marks in Elf1 CRISPR-Cas9 KO mutants andcontrol cells (Fig. 8e), therewere 4 independent experiments. For teratoma generatedfromH1 4iLIF (Supplementary Fig. 1D), there were 3 independent experiments. Forwestern blot of H3K27me3 in Elf1 AF cells overexpressing NNMT (SupplementaryFig. 4I), there were 3 independent experiments. For western blot of H3K27me3marks in naive WIN1 and WIN1 TeSR, there were 3 independent experiments(Supplementary Fig. 5H). For western blot of HIF1α and H3K27me3 in WIN1 cellsafter transfection with siRNA against NNMT or luciferase (Supplementary Fig. 5L)there were 3 independent experiments. For western blot of phosphorylated STAT3in H1 and H1 2iF (Supplementary Fig. 5M), there were 2 independent experiments.For morphology of Elf1 transfected with EV or HIF1αOE, there were 3 independentexperiments (Supplementary Fig. 7B).

There is no estimate of variation in each group of data and the variance is similarbetween the groups.No statisticalmethodwas used to predetermine sample size. Theexperiments were not randomized. The investigators were not blinded to allocationduring experiments and outcome assessment. RNA samples with 260 nm/280 nm<1.80 were discarded.

49. Taylor, S. D. et al. Targeted enrichment and high-resolution digital profiling ofmitochondrial DNA deletions in human brain. Aging Cell 13, 29–38 (2014).

NATURE CELL BIOLOGY


http://mit.edu

http://Addgene.com

http://Addgene.com

http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE60955







https://genome.ucsc.edu/ENCODE

http://www.genboree.org/EdaccData/Current-Release

http://www.genboree.org/EdaccData/Current-Release

DOI: 10.1038/ncb3264 METHODS

50. Cox, J. & Mann, M. MaxQuant enables high peptide identification rates, individualizedp.p.b.-range mass accuracies and proteome-wide protein quantification. Nat.Biotechnol. 26, 1367–1372 (2008).

51. Liesenfeld, D. B. et al. Metabolomics and transcriptomics identify pathwaydifferences between visceral and subcutaneous adipose tissue in colorectal cancerpatients: the ColoCare study. Am. J. Clin. Nutr. 102, 433–443 (2015).

52. Kind, T. et al. FiehnLib: mass spectral and retention index libraries for metabolomicsbased on quadrupole and time-of-flight gas chromatography/mass spectrometry. Anal.Chem. 81, 10038–10048 (2009).

53. Meissen, J. K. et al. Induced pluripotent stem cells show metabolomic differencesto embryonic stem cells in polyunsaturated phosphatidylcholines and primarymetabolism. PLoS ONE 7, e46770 (2012).

54. Kind, T. et al. LipidBlast in silico tandem mass spectrometry database for lipididentification. Nat. Methods 10, 755–758 (2013).

55. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29,15–21 (2013).

56. Anders, S., Pyl, P. T. & Huber, W. HTSeq-a Python framework to work with high-throughput sequencing data. Bioinformatics 31, 166–169 (2014).

57. Anders, S. & Huber, W. Differential expression analysis for sequence count data.Genome Biol. 11, R106 (2010).

58. Carvalho, B. S. & Irizarry, R. A. A framework for oligonucleotide microarraypreprocessing. Bioinformatics 26, 2363–2367 (2010).

59. Irizarry, R. A. et al. Exploration, normalization, and summaries of high densityoligonucleotide array probe level data. Biostatistics 4, 249–264 (2003).

60. Gautier, L., Cope, L., Bolstad, B. M. & Irizarry, R. A. affy–analysis of AffymetrixGeneChip data at the probe level. Bioinformatics 20, 307–315 (2004).

61. Johnson, W. E., Li, C. & Rabinovic, A. Adjusting batch effects in microarray expressiondata using empirical Bayes methods. Biostatistics 8, 118–127 (2007).

62. Shen, L. et al. diffReps: detecting differential chromatin modification sites fromChIP-seq data with biological replicates. PLoS ONE 8, e65598 (2013).

63. Denisenko, O. N. & Bomsztyk, K. The product of the murine homolog of the Drosophilaextra sex combs gene displays transcriptional repressor activity. Mol. Cell Biol. 17,4707–4717 (1997).

64. Sperber, H. et al. miRNA sensitivity to Drosha levels correlates with pre-miRNAsecondary structure. RNA 20, 621–631 (2014).

65. Biechele, T. L., Adams, A. M. & Moon, R. T. Transcription-based reporters of Wnt/β-catenin signaling. Cold Spring Harb. Protoc. 2009, pdb prot5223 (2009).

66. Gonzalez, F. et al. An iCRISPR platform for rapid, multiplexable, andinducible genome editing in human pluripotent stem cells. Cell Stem Cell 15,215–226 (2014).

NATURE CELL BIOLOGY


Supplementary Table 3 was replaced with an updated version on 26 November 2015


S U P P L E M E N TA RY I N F O R M AT I O N

WWW.NATURE.COM/NATURECELLBIOLOGY 1

DOI: 10.1038/ncb3264

An updated version of this Supplementary Information file was uploaded on 4 February 2016, with new figure headings and two new references.


http://www.nature.com/doifinder/10.1038/ncb3264


2 WWW.NATURE.COM/NATURECELLBIOLOGY

DOI: 10.1038/ncb3264Supplemental figure 1

Teratoma from H1 4iLIF

A B

C D

E F

0

2

4

6

8

10

12

14

16

18

Cha

nge

in E

CA

R D

ue to

O

ligom

ycin

Elf Elf + AF O/N Elf + AF 2d Elf + AF 3d

H

I J K

L

0.00E+00 1.00E-07 2.00E-07 3.00E-07 4.00E-07 5.00E-07 6.00E-07 7.00E-07 8.00E-07 9.00E-07

Del

etio

n fr

eque

ncy

(per

bp)

mtDNA site

Elf1

H1

M P

DNMT3L gene expression

TIME (min)

Nor

mal

ized

OC

R (p

Mol

s/M

in)

TIME (min)

Oligo FCCP Anti/Rot WIN1 5iLAF WIN1 AF

G ***

*** ***

NS

NS

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

norm

aliz

ed E

CA

R c

hang

e af

ter

Olig

omyc

in in

ject

ion

Del

etio

n fre

quen

cy (p

er b

p)

8E-7

O

6E-7

4E-7

2E-7

0 0.00E+00

2.00E-07

4.00E-07

6.00E-07

8.00E-07

1.00E-06

1.20E-06

1.40E-06

1.60E-06

1.80E-06

2.00E-06

12S rRNA

COXII

Mut

atio

n fr

eque

ncy

(per

bp)

mtDNA site

Elf1

H1

NS

NS

Mut

atio

n fre

quen

cy (p

er b

p)

2E-6

1.6E-6

1.2E-6

0.8E-6

0.4E-6

N

0

200

400

600

800

1000

1200

1400

Elf1 H1

NS

Nor

mal

ized

OC

R (p

Mol

s/M

in) Oligo FCCP Anti/Rot

TIME (min)

ELF1 H7

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Elf1 H7

mtD

NA

/gD

NA

NS

mtD

NA

cop

ies

(p

er d

ilpoi

d ge

nom

e

0

***

NS NS

0

20

40

60

80

100

120

mouse naïve

H1 2iF H1

OC

R c

hang

es a

fter F

CC

P

***

Supplementary Figure 1 Differences between naive and primed human and mouse ESCs. A: PCA of RNA-seq and microarray data from this study, Margaretha et al (in preparation) and other studies6,8-10,23, normalized to primed hESC in each study. Analysis was performed as described previously9. B: PCA of naïve and primed microarray data8-10. ComBat was applied on the combined dataset. PC1 is associated with naïve vs. primed difference and explained majority of variation (65.9%) and recapitulated the trend in Fig.1A. C:DNMT3L expression (qPCR) across naïve (n=6) and primed hESCs (n=3). SEM ***p<0.001, 2-tailed t-test. D: Histopathology of a mature cystic teratoma generated from H1 4iLIF, composed of primitive to well-differentiated tissues from ecto- (Ec, neuroectodermal rossettes), meso- (Me, bone), and endodermal (En, hepatoid exocrine glands) embryonic layers. Bar in top panel= 100µm (100X). Boxed regions (200X). Teratoma were obtained from 3/3 mice. E-F: A trace of OCR changes during naïve (Elf1, E and WIN1, F) to primed (cultured in AF 1-3 days) hESC transition. Elf1: n= 33, Elf1AF 1D: n=29, Elf1AF 2D: n=20, Elf1AF 3D: n=28; WIN1: n=4, WIN1AF: n=6; s.e.m. G: ECAR changes after oligomycin injection in naïve (Elf1: n=39, H1 4iLIF: n=6) and primed (H1: n=12, H7: n=27) hESC transition. S.e.m; ***p<0.001,

2-tailed t-test. H: ECAR changes after oligomycin injection during naïve (Elf1: n=33) to primed (Elf1AF overnight: n=29, 2 days: n=20, 3 days: n=28) hESC transition. S.e.m; *** p<0.0001, p=0.59 (Elf1AF ON vs. Elf1), p=0.11 (Elf1AF 2D vs Elf1), 2-tailed t-test. I: A trace of OCR changes in H7 and Elf1 hESCs, n=6, s.e.m. J-K: Metabolic profile of H1 (primed, n=10), H1 2iF (naïve toggled, n=11) hESCs, and R1 (naïve, n=10) mouse ESCs. A trace of OCR changes (J, s.d.). OCR changes after FCCP treatment for primed (H1) hESCs, and naïve human (H1 2iF) and mouse (R1) ESCs (K, s.e.m.; p=0.225 for R1 vs. H12iF and p<0.0001 for H1 vs. H12iF). L-M: Mitochondrial DNA copy numbers in Elf1, H1 and H7 hESCS. n=3; s.e.m.; p=0.3327(L), p=0.13(M), 2-tailed t-test. N: Mitochondrial DNA mutation frequency in Elf1 and H1 hESCs. n=3; s.e.m.; NS.:non significant p=0.34 for 12S rRNA and p=0.062 for COXII, 2-tailed t-test. O: Mitochondrial DNA deletion frequency in Elf1 and H1 hESCs. n=3; s.e.m.; p=0.44 for ND1/ND2 and p=0.46 for Common; 2-tailed t-test. P: Label-free quantification of protein expression is reproducible. Tryptic digestions of Elf1 2iLIF and ElfAF hESCs were analyzed in triplicate by nano-LC MS/MS (average Pearson correlation: 0.86). For raw data and p-values, see Supplementary Table 4. n= number of biological replicates.


http://www.nature.com/doifinder/10.1038/ncb3264



H1 2iF R1 H1 R1AF epi

Heatmap of metabolite levels (GC-TOF)

Supplemental figure 2

0 2.7 3.3

A

C D

log2 fold change 0 2 -2 -4

-log1

0 p-

valu

e

4

2

0

6

4 6

B

LC metabolomics GC metabolomics

8

nicotinic acid

kynurenine methionine

Elf1 AF v.s. Elf1

Supplementary Figure 2 Mass spectrometry profiling of naive and primed human and mouse ESCs. A: Heatmap generated through hierarchical Pearson clustering of metabolite expression from GC-TOF shows that naïve ESCs cluster away from primed ESCs, regardless of origin (mouse or human). B: volcano plot of differentially abundant metabolites between

primed hESCs (Elf1 AF) and naïve hESCs (Elf1) detected by GC-TOF. C-D: PCA plot of water-soluble untargeted GC-MS (C) and LC-MS (D) metabolomics data showing a separation of naïve ESCs (mouse:R1 and human: Elf1) and primed ESCs (mouse: Epi, and human: Elf1 AF, H1, H7).





CPT1A H3K27me3 (Chan et al)

Primed Naive

H1 4iLIF PALM H1 4iLIF BSA H1 PALM H1 BSA

E

Elf1 H7

Oil Red O staining H1 vs. Elf1

F G P-value: 0.012

Num

ber o

f car

bon

Higher in R1 Higher in Epi

20

30

40

50

60


Higher in Elf1 Higher in Elf1AF


Num

ber o

f dou

ble

bond

s

0.0

2.5

7.5

10.0

5.0

P-value: 7.354e-5

Genes involved in fatty acid transport and beta oxidation (human: H1 vs. Elf1)

Genes involved in fatty acid transport and beta oxidation (mouse: Epi vs. ICM)

H I

A

D C B

Elf1 AF H1 H7 H1 2iF H1 4iLIF ELf1

Bod

ipy

Brig

ht fi

eld

Naïve Primed

Supplementary Figure 3 Lipid metabolism in naive and primed human and mouse ESCs. A: Lipid droplets are known to be increased in human pluripotent and embryonic stem cells67,68. BODIPY 493/503 staining shows an increase of lipid droplet accumulation in primed hESCs (H7, H1, Elf1 AF) compared to naïve hESCs (Elf1, H1 2iF, H1 4iLIF). Images were taken at 5X magnification. Scale bar represents 100µm. B: Oil Red O staining shows an increase of lipid droplet accumulation in primed hESCs (H7) compared to naïve hESCs (Elf1). Scale bar represents 20µm. C: H3K27me3 reads mapped 5kb around transcription start site (TSS) of CPT1A were plotted for Chan et al ChIP-seq data sets. Primed cells have more H3K27me3 repressive marks around TSS of CPT1A. D: Expression of key genes involved in fatty acid synthesis from RNA-seq analysis in naïve (Elf1, n=2) and primed (H1, n=1) hESCs. Negative binomial test p-values are shown. E: A trace of OCR changes under Seahorse palmitate assay with addition of 2 doses of palmitate or BSA vehicle followed by 2 doses of ETO

in hESCs H1 pushed toward a more naïve state using 4iLIF (H1 4iLIF) and primed hESCs H1. n=6 for H1BSA, n=5 for H1PALM, H1 4iLIF BSA and H1 4iLIF PALM. F: More abundant lipids in primed human cells (Elf1AF) are more unsaturated than more abundant lipids in naïve human cells (Elf1), n=6. G: More abundant lipids in primed mouse cells (EpiSC) have more carbon atoms than more abundant lipids in naïve mouse cells (R1), n=6. Boxes represent median, 25th and 75th quantiles. Whiskers extend 1.5 IQR above 75th quantile and below 25th quantile. Dots represent values beyond whiskers. H-I: Relative fold change of expression of genes involved in transport of fatty acids into mitochondria and genes involved in 4 steps of fatty acid beta-oxidation from RNASeq analysis in human ESCs (H1 vs. Elf1, H) and mouse ESCs (Epiblasts vs.ICM, I). Three isoforms have been described: CPT1A, CPT1B and CPT1C; however, CPT1C is not involved in fatty acid beta-oxidation48. For raw data, see Supplementary Table 4. n= number of biological replicates.





A C IDO1 expression in various lineages

0

5

10

15

20

25

30

35

8 cell hESC p0

hESC p10

Expr

essi

on (R

PKM

)

IDO B Microarray data

D D E F

G

0

0.05

0.1

0.15

0.2

0.25

naive primed

Sum

nor

mal

ized

SAM

leve

ls

SAM

0

0.05

0.1

0.15

0.2

0.25

naive primed Sum

nor

mal

ized

nic

otin

amid

e le

vels

Nicotinamide H I

** **

55kDa

25kDa 15kDa

55kDa

Gen

e ex

pres

sion

Supplementary Figure 4 Tryptophan and SAM metabolic pathways in naive and primed hESCs. A: RNA expression of IDO1 in human 8-cell embryo and primed hESCs at passage 0 (hESCp0) and 10 (hESCp10) analyzed by single cell RNA Seq data23. B: RNA expression of IDO1, IDO2, TDO2 and AADAT in naïve hESCs Elf1, primed hESCs H1 and H1 differentiated during 4 days12,49 detected by microarray (n=2). C: IDO expression in H1 cells and H1 differentiated toward various lineages47, n=2, error bar represents 2 SEM above and below the mean. D: NNMT expression in various tissues analyzed by RNASeq50. E: NNMT expression in various organs of rats over time (from 2 to 104 weeks) analyzed by

RNASeq50. F: NNMT expression in H1 cells and H1 differentiated toward various lineages47. G-H: Primed hESCs (H1 n=4, ELF AF n=6, WIN1 TESR n=6) have higher amounts of nicotinamide (G) and SAM (H), than naïve hESCs (WIN1 n=4, Elf1 n=4, H1 2iF n=4, H1 4iLIF n=4). S.e.m.; p=0.0019 for nicotinamide, p=0.0026 for SAM; 2-tailed t-test. I: Overexpression of NNMT decreases H3K27me3 marks in primed Elf AF cells 3 days after transient transfection. Antibody against NNMT detected a single band around 50kDA. Unprocessed original scans of blots are shown in Supplementary Fig.9. For raw data, see Supplementary Table 4. n= number of biological replicates.




H3K4me3 (Theunissen et al) WIBR2 primed WIRB2 6i naïve

H3K9me3 (Gafni et al) H3K27ac (Gafni et al)


H3K4me3 (Chan et al) H3K27ac (Chan et al)

G

B

Primed Naïve

Primed Naïve

Primed Naïve Primed

Naïve

A H3K4me1 (Gafni et al)

H3K4me3 (Gafni et al) D E

Primed Naïve

Primed Naïve

H

I K L

M N

F

C

55kDa

25kDa 15kDa

55kDa

100kDa

15kDa 25kDa

130kDa

55kDa 70kDa 100kDa

J

NS NS

NS

NS NS

Supplementary Figure 5 Histone methylation, NNMT and miRNA expression changes. A-B: Reads for H3K4me3 (A) and H3K27ac (B) mapped 5kb around TSS were plotted for Chan et al ChIP-seq data set. C-F: Reads for H3K4me1 (C), H3K4me3 (D), H3K9me3 (E) and H3K27ac (F) mapped 5kb around transcription start sites (TSS) were plotted for Gafni et al. ChIP-seq data set. G: Reads for H3K4me3 mapped 5kb around TSS were plotted for Theunissen et al. ChIP-seq data set. H: Western blot analysis of H3K27me3 mark in naïve hESCs (WIN1) and cells pushed toward the primed stage (WIN1 TeSR). I-J: RNA-seq data of histone methyltransferases (I) and histone demethylases (J) in naïve (Elf1) and primed (H1) hESCs. K: qPCR analysis of miR-10a, miR-518b and miR-520f in Elf1 cells after

transfection with siRNA against NNMT or luciferase (50nM, 72h). n=3, S.e.m., N.S.=non significant, p=0.16(miR-518b) p=0.36(miR-520f) and p=0.058(miR-10a), 2-tailed t-test L: Western blot analysis of HIF1α and H3K27me3 in WIN1 cells after transfection with siRNA against NNMT or luciferase (50nM, 72h). M: STAT3 is phosphorylated in H1 cells pushed toward a more naïve stage (H1 2iF), even without LIF addition to the media. N: qPCR analysis of NNMT expression after treatment of Elf1 cells with 100µM of STAT3 inhibitor. n=3, S.e.m., N.S.=non significant, p=0.139 (Stat3i 6h vs. Elf1) and p=0.267 (Stat3i 24h), 2-tailed t-test. Unprocessed original scans of blots are shown in Supplementary Fig.9. For raw data, see Supplementary Table 4. n= number of biological replicates.





E

F G

A B C

D

Supplementary Figure 6 Wnt pathway gene expression and effects by Wnt inhibitors. A-D: Wnt ligands and Wnt targets are up-regulated in naïve hESCs compared to primed hESCs, as detected by RNA seq in Takashima et al (A), Chan et al (B) and microarray in Theunissen et al (C) and Gafni et al. (D). E: siRNA against β-catenin inhibits the reporter activity in naïve Elf1 cells after 72h. Scale bars represent 200µm. F: treatment of primed

hESCs (Elf1 AF) with Wnt3a CM or GSK3 inhibitor (CHIR99021, 10µM) for 3 days induces differentiation and re-activation of the BAR reporter. G: A trace of OCR following mitostress protocol in naïve hESCs (WIN1) with or without treatment with Wnt inhibitor IWP2 (2 µM, 48h). n=5 for Win1 and n=3 for WIN1+IWP2. For raw data, see Supplementary Table 4. n= number of biological replicates.





A

EV

H

IF1

OE

Cells grown without 2i

* B C

D Elf1 gHIF1 6.2.1 (RNA seq )

E CRISPR KO gHIF1

Nor

mal

ized

OC

R (p

Mol

s/M

in)

FCCP Anti/Rot

TIME (min)

gHIF1 6.2.1 AF 3D

gHIF1 6.2.1

G

0

0.00005

0.0001

0.00015

0.0002

0.00025

0.0003

0.00035

EV rep1

EV rep2

HIF1 OE rep1

HIF1 OE rep2

Gen

e ex

pres

sion

rela

tive

to

ACTB

IDO1

0 10 20 30 40 50 60 70 80

EV HIF1 OE

% o

f prim

ed-li

ke c

olon

ies

Supplementary Figure 7 Responses to HIF1 overexpression and knockout experiments. A:Expression of IDO1 in naïve hESCs (Elf1) after infection with an empty vector (EV) or HIF1 overexpressing (HIF1 OE) virus, analyzed by qPCR (n=2 independent replicates). B-C: HIF1α OE pushes in Elf1 cells toward primed-like cells. When cultured in a “permissive media” (with LIF but without GSK3 and MEK inhibitors) naïve hESCs Elf1α overexpressing HIF1α for 3 days show higher percentage of primed hESC-like colonies than cells infected with an empty vector control (EV). Morphology is shown in B and quantification of the percentage of

primed hESC-like colonies in C. Scale bar represents 200µm. n=261 (EV) and n=327 (HIF1 OE) colonies were counted from 3 independent experiments, s.e.m, * p=0.0183, 2-tailed t-test. D: Analysis of mutation in the CRISPR-Cas9 KO of HIF1α (Elf1 gHIF1 6.2.1) by RNA seq: the variant site carries the AG -> AGG insertion, as noted by the purple insertion symbol. E: A trace of OCR following mitostress protocol in CRISPR HIF1α KO cells (gHIF1 6.2.1) in cultured in naïve conditions or pushed to primed for 3 days (AF D3). N=6 ; s.e.m. For raw data, see Supplementary Table 4. n= number of biological replicates.




Supplemental Figure 8

A

Elf1 gNNMT 6.2.2 (DNA seq) Elf1 gNNMT 7.3.5 (DNA seq) C D

B

E

G

Fatty acid synthesis (NNMT CRISPR KO 7.2.1 vs. Elf1)

H

F

in primed in gNNMT 7.2.1

in primed in gNNMT 7.2.1

Elf1 gNNMT 7.2.1 (RNA seq)

Fatty acid transport and beta oxidation (NNMT CRISPR KO 7.2.1 vs. Elf1)

C

Supplementary Figure 8 NNMT CRISPR knockout experiments. A-B: schematic representation of the CRISPR-Cas9 KO experimental procedure using plasmid transfection (A) or gRNA transfection (B). In vitro transcribed gRNAs targeting NNMT were transfected in doxycycline-induced iCas9 hESCs Elf1 or electroporated as plasmids. Single clones were picked and amplified prior analysis by DNA sequencing. C-D: Sequencing trace files of heterozygous CRISPR-Cas9 mutant clones (NNMT 6.2.2,C; gNNMT 7.3.5, D). The guide RNA is represented in green. E: Relative fold change of expression of genes involved in transport of fatty acids into mitochondria and genes involved in 4 steps of fatty acid beta-oxidation from RNASeq analysis in human ESCs (Elf1 CRISPR-Cas9 KO gNNMT 7.2.1 vs. Elf1)

as detected by RNA seq. F: Expression of key genes involved in fatty acid synthesis from RNA-seq analysis in naïve (Elf1) and naïve NNMT CRISPR-Cas9 KO (Elf1 gNNMT 7.2.1) hESCs. In E and F, n=1 for gNNMT 7.2.1 and n=2 for Elf1. G: Detection of mutation in the CRISPR-Cas9 KO of NNMT (Elf1 gNNMT 7.2.1) by RNA seq. The variant site carries the TC -> TCC insertion, as noted by the purple insertion symbol. H: Number of genes for the overlap between genes expressed higher (lower) in gNNMT 7.2.1 compared to Elf1 and genes expressed higher (lower) in primed lines compared to naïve lines from multiple studies. Color shade is proportional to the number of overlapping genes. Significance is shown in Fig. 8I. n= number of biological replicates.




Figure 1J

100kDa 130kDa

Elf1 H7

HIF1α

55kDa Tubulin

Elf1 Elf1 AF

Figure 1M

25kDa 35kDa

130kDa

LDHA

JARID2

H7 Elf1 AF Elf1

55kDa Tubulin

H3K27me3 15kDa 25kDa

Elf1 H7

H3K9me3 15kDa 25kDa

H3K9/K14Ac

Tubulin 55kDa

15kDa 25kDa

Figure 5B

EED 55kDa 70kDa

Elf1 H7

55kDa Tubulin

15kDa 25kDa

55kDa

Elf1 Elf1 AF

Tubulin

H3K27me3

Figure 5D

15kDa 25kDa H3K27me3

H3K9me3

siLuc

15kDa 25kDa

siNNMT siLuc siNNMT

15kDa 25kDa

55kDa

H3K9/K14Ac

Tubulin

Tubulin 55kDa

15kDa 25kDa H3K27me3

H3K9me3 15kDa 25kDa

Figure 5F

Supplementary Figure 9 Unprocessed original scans of Western blots




100kDa 130kDa

55kDa

55kDa

gHIF1 6.3.1 TeSR 5D

gHIF1 6.2.1 TeSR 5D

Elf1 iCas9 TeSR 5D

Figure 7F

100kDa 130kDa

HIF1

15kDa H3K27me3

55kDa Tubulin

HIF1

Tubulin

Figure 8E

HIF1 100kDa 130kDa

gNNMT 6.2.4

Elf iCas9

15kDa

H3K27me3

Tubulin

55kDa

Tubulin

Tubulin

H3K27me3

gNNMT 7.2.1 Elf iCas9

Supplementary Figure 9 continued




Suppl Fig4I

55kDa NNMT

Elf AF NNMT OE - +

H3K27me3

Tubulin 55kDa

25kDa 15kDa

Suppl Fig5H Suppl Fig5L

H3K27me3 15kDa 25kDa

55kDa

WIN1 5iLAF

WIN1 TeSR

Tubulin

Suppl Fig5M

H3K27me3

Tubulin

HIF1 130kDa 100kDa

25kDa 15kDa

55kDa

WIN1

siLuc siNNMT

H1 H1 2iF

pSTAT3 70kDa 100kDa

55kDa Tubulin

Supplementary Figure 9 continued




Supplementary Table Legends

Supplementary Table 1 A: Gene contribution to PCA in Figure 1A. B: mRNA-seq data of H1 and Elf1 human embryonic stem cells. C: COX genes expression, mean-centered log2 expression values used for Figure 1G. D: Differential expression analysis of mitochondrial complexes genes. E: Proteomics mass spec data of H1 and Elf1 human embryonic stem cells. F: Mass spec raw data from non-targeted GC-TOF analysis of mouse R1, epi and R1 AF ESCs, and human H1 and H1 2i ESCs. G: Mass spec raw data from non-targeted LC-QTOF analysis of mouse R1, epiSC and R1 AF ESCs, and human H1 and H1 2i ESCs. H-I: Normalized mass spec data from targeted LC-QQQ-MS of H1 and Elf1 hESCs. J: Mass spec raw data from non-targeted GC-TOF analysis of mouse R1 and epi ESCs, and human Elf1, Elf AF, H1 and H7 ESCs. K: Mass spec data from lipidomic analysis (LC-QTOF) of H1 and Elf1 human embryonic stem cells. L: Mass spec raw data from non-targeted LC-QTOF analysis of mouse R1 and epi ESCs, and human Elf1, Elf AF, H1 and H7 ESCs. M: miRNA-seq data of H1 and Elf1 human embryonic stem cells. N: Normalized mRNA read counts O: Raw data from targeted HILIC mass spec analysis of metabolites in methionine and tryptophan metabolism. P: Top GO-terms for genes that are upregulated in Elf1 human embryonic stem cells after 6h STAT3 inhibitor treatment. Q: Top GO-terms for genes that are downregulated in Elf1 human embryonic stem cells after 6h STAT3 inhibitor treatment. R: List of 313 genes that make up the intersection between genes that go down in Elf1 human embryonic stem cells after 6h STAT3 inhibitor treatment, and genes that have increased H3K27me3 marks in primed H1 compared to naïve Elf1 human embryonic stem cells. S: List of genes upregulated in 14 day embryonic bodies differentiated from H1 4iLIF compared to H1 4iLIF.

Supplementary Table 2 List of primers and Taqman probes used for amplification and quantification of genomic DNA, mitochondrial DNA and cDNA. List of sequences of CRISPR guide RNA used to KO HIF1 and NNMT and primers used to analyzed mutations.

Supplementary Table 3 Composition of media used for culture of naïve and primed human and mouse ESCs.

Supplementary Table 4 Statistics source data.

References

67. Thangaselvam Muthusamy, T., Mukherjee, O., Menon, R., Megha, P. B. & Panicker, M. M. Stem Cell Rep. 3,169–184 (2014).68. Stringari, C., Sierra, R., Donovan, P. J. & Gratton, E. J. Biomed Opt. 17, 046012 (2012).


the metabolome regulates the epigenetic landscape during naive … · 2020. 8. 16. · pluripotent...

Documents