LETTERdoi:10.1038/nature13441

Mutant IDH inhibits HNF-4a to block hepatocytedifferentiation and promote biliary cancerSupriya K. Saha1*, Christine A. Parachoniak1*, Krishna S. Ghanta1, Julien Fitamant1, Kenneth N. Ross1, Mortada S. Najem1,Sushma Gurumurthy1, Esra A. Akbay2, Daniela Sia3,4,5, Helena Cornella3, Oriana Miltiadous4, Chad Walesky6, Vikram Deshpande1,Andrew X. Zhu1, Aram F. Hezel7, Katharine E. Yen8, Kimberly S. Straley8, Jeremy Travins8, Janeta Popovici-Muller8,Camelia Gliser8, Cristina R. Ferrone1, Udayan Apte6, Josep M. Llovet3,4,9,10, Kwok-Kin Wong2, Sridhar Ramaswamy1,11

& Nabeel Bardeesy1

Mutations in isocitrate dehydrogenase 1 (IDH1) and IDH2 are amongthe most common genetic alterations in intrahepatic cholangiocar-cinoma (IHCC), a deadly liver cancer1–5. Mutant IDH proteins inIHCC and other malignancies acquire an abnormal enzymatic activ-ity allowing them to convert a-ketoglutarate (aKG) to 2-hydroxyglu-tarate (2HG), which inhibits the activity of multiple aKG-dependentdioxygenases, and results in alterations in cell differentiation, sur-vival, and extracellular matrix maturation6–10. However, the molecularpathways by which IDH mutations lead to tumour formation re-main unclear. Here we show that mutant IDH blocks liver progenitorcells from undergoing hepatocyte differentiation through the pro-duction of 2HG and suppression of HNF-4a, a master regulator ofhepatocyte identity and quiescence. Correspondingly, genetically engi-neered mouse models expressing mutant IDH in the adult liver showan aberrant response to hepatic injury, characterized by HNF-4a si-lencing, impaired hepatocyte differentiation, and markedly elevatedlevels of cell proliferation. Moreover, IDH and Kras mutations, gen-etic alterations that co-exist in a subset of human IHCCs4,5, coop-erate to drive the expansion of liver progenitor cells, developmentof premalignant biliary lesions, and progression to metastatic IHCC.These studies provide a functional link between IDH mutations, he-patic cell fate, and IHCC pathogenesis, and present a novel genetic-ally engineered mouse model of IDH-driven malignancy.

Gain-of-function IDH1/IDH2 mutations occur in ,25% of IHCCs1,3–5,a liver malignancy that exhibits bile duct differentiation, but have notbeen identified in hepatocellular carcinomas (HCCs), which exhibit hepa-tocyte differentiation (http://www.sanger.ac.uk/cosmic). To examine therole of IDH mutations in liver tumorigenesis we isolated mouse hepa-toblasts, which are embryonic progenitors that give rise to hepatocytesand bile duct cells and correspond with adult liver progenitors11,12. Hepa-toblasts expressing mutant IDH1 (R132C, R132H) or IDH2 (R140Q,R172K) produced increased 2HG, but exhibited morphology and pro-liferation rates indistinguishable from vector and IDH wild-type con-trols (Extended Data Fig. 1a–d). However, unlike control hepatoblasts,which underwent hepatocyte differentiation when transferred fromcollagen-coated plates to uncoated plates13, forming hepatocyte clusters,decreasing proliferation, and activating a large program of hepatocyte-specific genes including Adh1 and Aldob, IDH-mutant cells were refrac-tory to differentiation (Fig. 1a–d and Extended Data Fig. 1e–g). IDH1(R132C) and IDH2(R172K) caused the most pronounced effects, cor-relating with relative 2HG levels. Treatment of R132C-expressing hepa-toblasts with AGI-5027 (also known as ML309), a specific inhibitor of

mutant IDH1 (ref. 14), attenuated 2HG production (Extended DataFig. 1h) and restored hepatocyte differentiation (Fig. 1e, f). Conversely,differentiation of wild-type hepatoblasts was counteracted by (R)- or(S)-2HG octyl esters (Extended Data Fig. 1i, j). In contrast to the com-plete inhibition of hepatocyte differentiation, mutant IDH did not impairbiliary differentiation of hepatoblasts in matrigel (Fig. 1g and ExtendedData Fig. 1k, l). Thus, mutant IDH specifically blocks hepatocyte lin-eage progression through 2HG production.

To uncover the molecular program underlying these defects, we ex-amined the effect of mutant IDH1/2 on the transcriptome of hepato-blasts grown on collagen. Transcriptional profiles of mutant IDH1 and2 clustered together and gene set enrichment analysis (GSEA) demon-strated reduced expression of targets of HNF-4a—a master transcrip-tional regulator of hepatocyte differentiation12—and of HNF-1a, whichacts downstream of HNF-4a15. Moreover, canonical HNF-4a- and HNF-1a-binding sites were strongly enriched at the promoters of differentiallyexpressed genes (Fig. 2a and Extended Data Fig. 2a–c).

HNF-4a has multiple isoforms expressed from separate promoters.The P2 promoter (encoding HNF-4a(7–9)) is active in hepatoblasts andgradually extinguished in adult hepatocytes, whereas the P1 promoter(encoding HNF-4a(1–6)) is hepatocyte-specific16. Hnf4a(7–9) messen-ger RNA and protein were reduced in IDH-mutant hepatoblasts, as wasexpression of HNF-4a targets (Extended Data Fig. 2d–g). Moreover,under hepatocyte differentiation conditions, mutant IDH completelyinhibited the pronounced induction of HNF-4a(1–6) and its target—OCLN—that is observed in control cells (Fig. 2b and Extended DataFig. 2h). Mutant IDH or octyl-2HG treatment blocked Hnf4a(1–6)mRNA induction, whereas AGI-5027 restored Hnfa(1–6) levels in R132C-expressing cells (Fig. 2c and Extended Data Fig. 2i–k). Histone H3 lysine4 trimethylation (H3K4me3) is associated with active transcription andwas specifically reduced at the P1 promoter in R132C hepatoblasts, con-sistent with the observed silencing of Hnf4a(1–6), whereas the repress-ive marks, H3K27me3 and H3K9me3, were unaffected (Extended DataFig. 2l, m and data not shown). Importantly, HNF-4a knockdown im-paired hepatocyte differentiation of wild-type hepatoblasts without inhib-iting biliary differentiation, whereas ectopic HNF-4a expression rescueddifferentiation of IDH-mutant cells (Fig. 2d–h and Extended Data Figs2n, o, 3a, b). Thus, mutant IDH alters the epigenetic state of the P1promoter—through targeting either direct regulators of the locus or moreupstream factors—and prevents induction Hnf4a(1–6), thereby block-ing hepatocyte lineage progression. Notably, HNF-4a and HNF-1ahavecritical antiproliferative and tumour suppressor functions in the adult

*These authors contributed equally to this work.

1Massachusetts General Hospital Cancer Center, Harvard Medical School, Boston, Massachusetts 02114, USA. 2Department of Medical Oncology, Dana-Farber Cancer Institute, Department of Medicine,Harvard Medical School, Boston, Massachusetts 02115, USA. 3HCC Translational Research Laboratory, Barcelona-Clınic Liver Cancer Group, Liver Unit, Institut d’Investigacions Biomediques August Pi iSunyer (IDIBAPS), Hospital Clınic, University of Barcelona, Catalonia 08036, Spain. 4Mount Sinai Liver Cancer Program, Division of Liver Diseases, Dept of Medicine. Icahn School of Medicine at Mount Sinai,New York 10029, USA. 5Gastrointestinal Surgery and Liver Transplantation Unit, National Cancer Institute, and Department of Experimental Oncology, Milan 20133, Italy. 6Department of Pharmacology,Toxicology and Therapeutics, University of Kansas Medical Center, Kansas City, Kansas 66160, USA. 7University of Rochester Medical Center, Rochester, New York 14642, USA. 8Agios Pharmaceuticals,Cambridge, Massachusetts 02139, USA. 9Institucio Catalana de Recerca i Estudis Avançats, Barcelona, Catalonia 08010, Spain. 10University of Barcelona, Catalonia 08036, Spain. 11Broad Institute ofHarvard and MIT, Cambridge, Massachusetts 02142, USA.

0 0 M O N T H 2 0 1 4 | V O L 0 0 0 | N A T U R E | 1

Macmillan Publishers Limited. All rights reserved©2014

liver17–20, suggesting the relevance of this pathway to mutant-IDH-mediated tumorigenesis and prompting us to extend our studies in vivo.

We generated transgenic mice with doxycycline-inducible express-ion of IDH2(R140Q) or IDH2(R172K) (Tet-R140Q, Tet-R172K strains)specifically in adult hepatocytes—R140Q was detected in virtually allhepatocytes and R172K showed more scattered expression, and liver2HG levels were elevated (Extended Data Figs 4a–d, 5a). Since mutantIDH blocks liver progenitors from undergoing hepatocyte differenti-ation in vitro, we sought to address whether it acts analogously in vivoto specifically override differentiation from a progenitor cell state, orconversely, whether it broadly alters homeostasis of mature hepato-cytes. Although normally quiescent, the liver has extensive regenerative

capacity after injury involving replication of mature hepatocyte and bil-iary cells, or activation of bipotential progenitors (oval cells) that mayarise from either lineage11,21. In the absence of injury, Tet-R140Q micewere healthy for up to 48 weeks, and had normal liver histology, markerexpression, proliferation, and liver function (Fig. 3d, Extended DataFig. 5b and data not shown). By contrast, pronounced defects in res-toration of hepatocyte differentiation were observed in mice fed a dietcontaining 3,5-diethoxycarbonyl-1,4-dihydrocollidin (DDC) for 5 daysand then switched to a normal diet for 3 weeks (Fig. 3a), a protocol caus-ing hepatocyte cell death and transient oval cell activation21,22. Hepa-tocyte markers including HNF-4awere downregulated 3–10-fold, whilebiliary markers were unchanged, and proliferation was increased .40-fold relative to wild-type controls (Fig. 3b–d). Despite this depletion ofmature hepatocytes, no changes were seen in parameters of liver func-tion (Extended Data Fig. 5c, d and data not shown), consistent with thepersistence of hepatocytes surviving short-term DDC treatment and theestablished capacity of reduced hepatocyte numbers to maintain nor-mal physiology.

EV

1

EV

2

R132C

1

R132C

2

EV

1

EV

2

R132C

1

R132C

2

Collagen Uncoated

Collagen Uncoated

EV

IDH

1 W

T

R132C

R132H

EV

IDH

1 W

T

R132H

R132C

Collagen Uncoated

a

b

f

d

c

e IDH1 WT R132C

+ A

GI-

5027

+ D

MS

O

0

4

12

Time (h)

Cell

no

. (fo

ld c

hang

e) R132C + DMSO

EV + DMSOR132C + AGI-5027EV + AGI-5027

x

x

xx

24 72 120

8

*

min max

Expressionlevel

Hep

ato

cyte

Bili

ary

Krt19

Krt7Ttr

Rnase4

Mgst1

Hal

Habp2

Gjb1

Cyp27a1

C9

C3

C2

Aldob

EV

R132C

g 104 + 41

243 + 5

R132HR132C

R172KR140QIDH2 WT

IDH1 WTEV

Aldob

Adh1

103

102

101

100

**

Rela

tive e

xp

ressio

n

AGI-5027

*

Rela

tive

exp

ressio

n

WTEV R132C WTEV R132C– – – – – – ++++++

Aldob

Adh1

*

Cell

no

. (fo

ld c

hang

e)

Time (h)

R132C

IDH1 WT

EV

24 72 120

10

15

20

25

0

5

104

103

102

101

100

Figure 1 | Mutant IDH blocks hepatocyte differentiation. a–d, Hepatoblastsexpressing empty vector (EV) or the indicated IDH alleles at physiological levels(see Extended Data Fig. 1a), were transferred to uncoated plates to inducehepatocyte differentiation. a, Hepatocyte sphere formation. WT, wild type.b, c, Proliferation (b) and hepatocyte marker expression (c) (quantitativepolymerase chain reaction with reverse transcription (qRT–PCR)). d, Heatmap of hepatocyte and biliary gene expression. e, f, Hepatoblasts treated with2.5mM AGI-5027 or dimethylsulphoxide (DMSO) vehicle. e, Hepatocytesphere formation (top), proliferation (bottom). f, Hepatocyte markerexpression. g, Hepatoblasts in matrigel assessed for biliary differentiation.Tubular structures per 6 cm dish 6 standard deviation (s.d.) are quantified.*P , 0.05. Scale bars, 100 mm.

a

LI_WILMS_TUMOR_VS_FET_KID_1_DN

KARLSSON_TGFB1_TARGETS_UP

SMID_BREAST_CANCER_BASAL_UP

SERVITJA_ISLET_HNF1A_TARGETS_DN

DER_IFN_ALPHA_RESPONSE_UP

LUCAS_HNF4A_TARGETS_UP

–4 0 4 NS

NES

eEV 1 R132C

+ H

NF

-4α(

1)

+ E

V 2

20

10

0

25

15

5

Time (h)

24 72 120

Cell

no

. (fo

ld c

hang

e) R132C + EV 2

R132C + HNF-4α(1)

EV 1 + HNF-4α(1)

EV 1 + EV 2

x

x

x

x

*

Adh1

Aldob

Alb

f

g h

b

HNF-4α(1–6)

Ocln

IDH1

Actin

EV

IDH

1 W

T

R132C

EV

IDH

1 W

T

R132C

0

10

20

30

40Hnf4a(1–6)

Rela

tive e

xp

ressio

n

Collagen

EV

R132C

Uncoated

EV

R132C

Cell

no

. (fo

ld c

hang

e)

Time (h)24 72 120

shHnf4a + EV

shHnf4a + HNF-4αshCTL + HNF-4α

shCTL + EV

x

x x

x

*

0

5

10

20

15

25

Collagen Uncoated

shCTL

shHnf4a#1

shHnf4a#2

Mutant IDHvs EV

Mutant IDHvs WT

103

102

101

100

EV 2

HNF-4α(1)

+–

+–

–+

*

Rela

tive e

xp

ressio

n

R132CEV1

c d

Figure 2 | Mutant IDH blocks hepatocyte differentiation by silencing HNF-4a. a, Heat map of GSEA showing top-ranked gene sets distinguishingIDH1(R132C) or IDH2(R172K) from wild-type (WT) or empty vector (EV)control hepatoblasts (pairwise analysis; replicates for each condition; seeMethods). NES, normalized enrichment score; NS, not significant.b, c, Hepatoblasts analysed by immunoblot (b) and qRT–PCR (c). d, e, Analysisof wild-type hepatoblasts expressing the indicated short hairpin (sh)RNAs(uncoated plates). CTL, control. d, Hepatocyte sphere formation.e, Proliferation of shRNA-expressing hepatoblast cells co-expressing EV orshRNA-resistant Hnf4a(1) complementary DNA. f–h, Control and R132C-expressing hepatoblasts co-expressing vector control (EV2) or HNF-4a, grownon uncoated plates. f, Hepatocyte sphere formation. g, Hepatocyte geneexpression. h, Proliferation. Scale bars, 100mm (d), 250mm (f). *P , 0.05.

RESEARCH LETTER

2 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 4

Macmillan Publishers Limited. All rights reserved©2014

Serial analyses of wild-type and Tet-R140Q livers revealed compar-able numbers of proliferating periductal HNF-4a2/CK192 oval cells at1 week, and resolution of this population after 3 weeks (Extended DataFig. 6a, b and Fig. 3e, f). However, Tet-R140Q livers exhibited prom-inent induction of proliferating non-periductal cells with hepatocytemorphology but reduced or absent HNF-4a expression (HNF-4a1 andHNF-4a2/CK192), which persisted after 3 weeks (Fig. 3e, f and ExtendedData Fig. 6a–d). Similar but more tempered phenotypes were seen inTet-R172K mice, consistent with focal transgene expression (ExtendedData Fig. 6e, f). Thus, mutant IDH specifically blocks restoration ofhepatocyte differentiation after acute liver injury, leading to aberrantproliferation in the hepatic parenchyma. As HNF-4a levels increase inliver progenitors undergoing hepatocyte commitment23,24, these prolif-erating HNF-4a-low/absent cells appear to be committed progenitors(derived from dedifferentiated hepatocytes or oval cells) whose differ-entiation to hepatocytes is subverted by mutant IDH2. The transgenewas not expressed in SOX91 oval cells, precluding assessment of effectson these earlier progenitors.

We developed an additional transgenic strain (IDH2R172K) expres-sing human IDH2(R172K) in SOX91 biliary cells but not in hepato-cytes (Fig. 3g and Extended Data Fig. 7a, b). These animals displayednormal liver histology, gene expression, and proliferation at 3 months(Extended Data Fig. 7c and data not shown). However, by 20 monthsof age there was pronounced accumulation of HNF-4a2/SOX91 ovalcells expressing IDH2(R172K) .25mm away from any bile duct or portalstructure (zones 2/3) (Fig. 3g). Collectively, our results suggest that mu-tant IDH2 abrogates differentiation of adult progenitors (activated spon-taneously during ageing or by injury), specifically blocking hepatocytelineage progression.

These findings are concordant with our observations in hepatoblastsand suggest that failed HNF-4a induction could contribute to the liverphenotypes caused by mutant IDH in vivo. In this regard, acute Hnf4adeletion in hepatocytes is reported to provoke hepatocyte differentiation

defects and, upon diethylnitrosamine (DEN) treatment, oval cellaccumulation and formation of tumours showing HCC and IHCCmorphology20. To expand upon potential parallels with IDH-mutantmice, we characterized these phenotypes in further depth. Importantly,the HCC lesions in DEN-treated Hnf4a conditional knockout mice wereuniformly HNF-4a1, indicating that they arose from cells that escapeddeletion of the locus (Extended Data Fig. 8a). By contrast, SOX91 ovalcells and CK191 IHCCs were HNF-4a2 and were never observed inDEN-treated controls (Extended Data Fig. 8b, c). Thus, HNF-4a abla-tion in DEN-treated livers drives progenitor expansion and progressionspecifically to IHCC. These findings establish Hnf4a as an IHCC tumoursuppressor and are consistent with HNF-4a acting downstream of mu-tant IDH in liver growth control.

The capacity of mutant IDH to silence Hnf4a and impair differenti-ation would be expected to confer sensitivity to transformation by addi-tional oncogenic lesions. In this regard, IDH and KRAS mutations existconcurrently in human IHCC4,5 and acute myeloid leukaemia25. Previ-ously, we showed that Kras(G12D) expression in mouse liver causesmixed IHCC/HCC with long latency, a phenotype accelerated byTP53deletion26. Intercrossing LSL-IDH2R172K, LSL-KrasG12D and Alb-Cre micerevealed dramatic oncogenic cooperation, with six out of six Alb-Cre;LSL-IDH2R172K;LSL-KrasG12D (IDH2R172K;KrasG12D) animals devel-oping poor body condition and palpable liver tumours between 33 and58 weeks (mean 47.3 weeks; Fig. 4a). Multifocal liver masses with splenicinvasion and peritoneal metastases were observed and demonstratedto be IHCC by histopathological analysis, CK19 staining, and lack ofreactivity for the hepatocyte/HCC marker Hep Par1 (Fig. 4b). By con-trast, only one out of seven Alb-Cre;LSL-KrasG12D (KrasG12D) mice sus-tained a tumour by 70 weeks (mean survival 5 81.6 weeks), and solelyHCCs were detected (Extended Data Fig. 8d). Tumour 2HG levels inIDH2R172K;KrasG12D mice were comparable to those in IDH-mutanthuman IHCC1 (Extended Data Fig. 8e).

a c

3

2

1

0

WT

Tet-R140Q

Hepatocyte markers Biliary markers

Rela

tive e

xp

ressio

n

Hnf4a

Adh1

Alb

AldobArg1

C2

C7

C9

Ttr

ApobKrt7

Krt19

Sprr1a

Onecut1

4

* * * * * * * * *

WT/Tet-R140Q

DDC

diet

DOX

Normal

diet Analysis Analysis

WT Tet-R140Q

HNF-4α(1–6)

Actin

1.0 0.85 1.04 0.32 0.25 0.32

e CK19/Ki-67/HNF-4α

1.2

1.0

0.8

0.6

0.4

0.2

0

*

HN

F-4α/

DA

PI (a

.u.)

WT

Tet-R140Q

WT Tet-R140Q

PV

PV

fWT IDH2R172K

SO

X9

SO

X9/R

172K

/HN

F-4α WT

IDH2R172K

SO

X9

+ c

ells

in

zo

nes 2

/3

*

b

No

DD

CD

DC

(d

ay 2

1)

WT Tet-R140Q

No

DDC

DDC

(day

21)

0

20

40

60

80

Ki-

67-p

ositiv

e c

ells

*

d

WT

Tet-R140Q40

30

20

10

0

CK19

+

HNF-

4α+

CK19

–

HNF-

4α–

Ki-

67-p

ositiv

e c

ells

*

*

g

WT

Tet-R140Q

WT

Tet-

R140Q

217Day –7 –5 0

PV PV

10

5

0

Figure 3 | Mutant IDH inhibits hepatocyte differentiation and quiescenceof liver progenitors. a, Schematic of DDC study in Tet-R140Q (Tet-R140Q;Alb-Cre;Rosa26-LSL-rtTA) and wild-type (WT) littermate controls(Alb-Cre;Rosa26-LSL-rtTA). b–f, Livers at day 21. b, Immunoblot (HNF-4a(1–6):actin is quantified). c, qRT–PCR. d, Ki-67 staining. Chart shows Ki-671 cellsper 20 high-powered fields. e, f, Immunofluorescence analysis. Graph depictsmean fluorescence intensity of HNF-4a:49,6-diamidino-2-phenylindole(DAPI) (125 cells per group were scored). Inset: 33 magnification of boxedregions. a.u., arbitrary units. f, Quantification of Ki-671 cells co-staining for the

indicated markers (N 5 3 mice per group, 5 high-powered fields per mouse).g, Immunohistochemistry (top) and immunofluorescence (bottom) ofwild-type (Alb-Cre) and IDH2R172K (Alb-Cre;LSL-IDH2R172K) livers at 20months. Note accumulation of SOX91 cells located .25mm away from bileduct or portal structures (dashed line), which express IDH2(R172K) andlack HNF-4a. Inset: 33 magnification. Chart shows quantification. N 5 3 miceper group; 4 high-powered images per mouse were scored. PV, portal vein.Error bars show mean 6 standard error of the mean (s.e.m.). Scale bars, 20 mm(d), 50mm (e, g).

LETTER RESEARCH

0 0 M O N T H 2 0 1 4 | V O L 0 0 0 | N A T U R E | 3

Macmillan Publishers Limited. All rights reserved©2014

In humans, IHCC is thought to develop from precursor lesions, in-cluding biliary intraepithelial neoplasias2. Aberrant oval cell prolifera-tion also precedes liver cancer development in murine models27. Notably,in all IDH2R172K;KrasG12D mice analysed (N 5 6), the adjacent liver ex-hibited oval cell expansion and a graded series of biliary intraepithelialneoplasia-like lesions, which were SOX91/CK191, and expressed theIDH2R172K transgene at physiological levels (Fig. 4c and Extended DataFigs 9a, b, 10a). Importantly, neither KrasG12D nor KrasG12D;p53Lox/1

livers26 showed oval cell expansion, and biliary intraepithelial neoplasiaswere found in only two of eight KrasG12D;p53Lox/1 mice. Thus, IDH2(R172K) and Kras(G12D) cooperatively incite activation of hepaticprogenitors and multistage IHCC progression.

We present a novel genetically engineered mouse model driven bymutant IDH that exhibits the hallmarks of human IHCC. Mutant IDHblocks hepatocyte differentiation from progenitors in vitro, and regu-lates hepatic regeneration, progenitor cell expansion, and IHCC patho-genesis in vivo. Collectively, our findings are consistent with a modelwhereby mutant IDH subverts the HNF-4a-mediated hepatocyte differ-entiation/quiescence program in proliferating hepatocytes or bipoten-tial progenitors, creating a persistent pre-neoplastic state primed fortransformation by additional oncogenic mutations, and leading to ade-nocarcinoma specifically with a biliary phenotype (Fig. 4d). While lineage-tracing studies are required to fully define the impact of IDH mutationson different liver cell types, this model predicts that IDH-mutant IHCCmay have progenitor-like features, and that IDH mutations occur earlyin disease pathogenesis as observed in glioblastoma and acute myeloidleukaemia28,29. Accordingly, GSEA of human IHCCs revealed that theIDH-mutant subset (17/107 cases) have strong enrichment of a hepaticstem cell signature (Extended Data Fig. 10b).

Defining the biological functions of mutant IDH is of immediateimportance as IHCCs are resistant to current treatment. Our genetic-ally engineered mouse model provides an autochthonous context for

‘co-clinical trials’ to help determine the best way to deploy IDH inhi-bitors and identify biomarkers of response, thereby informing clinicaldecision-making in this deadly disease.

METHODS SUMMARYCell culture. Hepatoblasts were prepared from wild-type mice at embryonic day14 and subjected to hepatocyte and bile duct differentiation as described13. For pro-liferation assays, cells were plated in duplicate on collagen-coated or uncoated 6-wellplates. Cells were trypsinized and counted by trypan-blue exclusion.Mice. Mice were housed in pathogen-free animal facilities. Studies were approvedby the Subcommittee on Research Animal Care at Massachusetts General Hospital(protocol 2005N000148). The Tet-R140Q, Tet-R172K and IDH2R172K transgenicmice and other strains are described in Methods. For DDC experiments, mice weremaintained on 200mg ml21 doxycycline-containing water. Mice were fed with 0.1%DDC-containing chow (F4643, Bio-serv) from days 2–7. Mice were then switchedto normal chow until harvest. DEN experiments were performed as described20.Statistics. Results are expressed as mean 6 s.d. unless otherwise specified. Signifi-cance was analysed using two-tailed Student’s t-test. A P value of less than 0.05 wasconsidered statistically significant.

Online Content Methods, along with any additional Extended Data display itemsandSourceData, are available in the online version of the paper; references uniqueto these sections appear only in the online paper.

Received 25 September 2013; accepted 2 May 2014.

Published online 2 July 2014.

1. Borger, D. R.et al.Frequent mutation of isocitrate dehydrogenase (IDH)1and IDH2in cholangiocarcinoma identified through broad-based tumor genotyping.Oncologist 17, 72–79 (2012).

2. Hezel, A. F., Deshpande, V. & Zhu, A. X. Genetics of biliary tract cancers andemerging targeted therapies. J. Clin. Oncol. 28, 3531–3540 (2010).

3. Razumilava, N. & Gores, G. J. Cholangiocarcinoma. Lancet http://dx.doi.org/10.1016/S0140-6736(13)61903-0 (2014).

4. Voss, J. S. et al. Molecular profiling of cholangiocarcinoma shows potential fortargeted therapy treatment decisions. Hum. Pathol. 44, 1216–1222(2013).

Quiescent

hepatocyte

c

ba

BilIN I BilIN II BilIN IIIOval cell expansion

SO

X9

CK

19

H&

E

KrasG12D IDHR172K

SO

X9/H

NF

-4α/

DA

PI

IDHR172K;KrasG12D

*P < 0.0005

0

100

75

50

25

0 50 100

Control (N = 7)

IDH2R172K (N = 7)

KrasG12D (N = 7)

IDH2R172K;KrasG12D (N = 6)

Surv

ival (%

)

Time (weeks)

Cell turnover/

injury

Quiescent

hepatocyte

2HG

+HitsRecovery

IDH mutantWT

Progenitor

cellHNF-4α BilllN IHCCOval cell

expansion

Cell turnover/

injury

IDH mutantProgenitor

cellHNF-4α

d

Hep Par1

CK19

H&EFigure 4 | Mutant IDH cooperates withKras(G12D) to drive liver progenitor cellexpansion and multi-step IHCC pathogenesis.a, Kaplan–Meier analysis showing time until signsof illness necessitated euthanasia. All animalseuthanized had liver tumours. b, Top left,representative IDH2R172K;KrasG12D (Alb-Cre;LSL-IDH2R172K;LSL-KrasG12D) tumour, and peritonealand spleen metastases (insets). Top right,haematoxylin and eosin (H&E) staining showingIHCC histology. Bottom, the tumour is Hep Par12

and CK191 while adjacent hepatocytes stainHep Par11 and CK192. c, The livers ofIDH2R172K;KrasG12D animals exhibit oval cellexpansion and increasing grades of biliaryintraepithelial neoplasias (BilIN), which stainSOX91. CK19 levels increase with higher gradelesions. Immunofluorescence analysis reveals focalaccumulation of SOX91 oval cells in IDH2R172K

livers and pronounced oval cell expansion inIDH2R172K;KrasG12D livers. d, Model for mutantIDH in IHCC pathogenesis. Scale bars, 1 cm (b, topleft), 50mm (b, c).

RESEARCH LETTER

4 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 4

Macmillan Publishers Limited. All rights reserved©2014

5. Wang, P. et al. Mutations in isocitrate dehydrogenase 1 and 2 occur frequently inintrahepatic cholangiocarcinomas and share hypermethylation targets withglioblastomas. Oncogene 32, 3091–3100 (2013).

6. Dang, L. et al. Cancer-associated IDH1 mutations produce 2-hydroxyglutarate.Nature 462, 739–744 (2009).

7. Losman, J. A. & Kaelin, W. G. Jr. What a difference a hydroxyl makes: mutant IDH,(R)-2-hydroxyglutarate, and cancer. Genes Dev. 27, 836–852 (2013).

8. Lu, C. et al. IDH mutation impairs histone demethylation and results in a block tocell differentiation. Nature 483, 474–478 (2012).

9. Sasaki, M. et al. IDH1(R132H) mutation increases murine haematopoieticprogenitors and alters epigenetics. Nature 488, 656–659 (2012).

10. Xu, W. et al. Oncometabolite 2-hydroxyglutarate is a competitive inhibitor ofa-ketoglutarate-dependent dioxygenases. Cancer Cell 19, 17–30 (2011).

11. Michalopoulos, G. K. Principles of liver regeneration and growth homeostasis.Compr Physiol 3, 485–513 (2013).

12. Si-Tayeb,K., Lemaigre, F.P.&Duncan,S.A.Organogenesisanddevelopmentof theliver. Dev. Cell 18, 175–189 (2010).

13. Strick-Marchand, H. & Weiss, M. C. Inducible differentiation and morphogenesis ofbipotential liver cell lines from wild-type mouseembryos.Hepatology36, 794–804(2002).

14. Davis, M. I. et al. Biochemical, cellular and biophysical characterization of a potentinhibitor of mutant isocitrate dehydrogenase IDH1. J. Biol. Chem. 289,13717–13725 (2014).

15. Kuo, C. J. et al. A transcriptional hierarchy involved in mammalian cell-typespecification. Nature 355, 457–461 (1992).

16. Torres-Padilla, M. E., Fougere-Deschatrette, C. & Weiss, M. C. Expression of HNF4aisoforms in mouse liver development is regulated by sequential promoter usageand constitutive 39 end splicing. Mech. Dev. 109, 183–193 (2001).

17. Bluteau, O. et al. Bi-allelic inactivation of TCF1 in hepatic adenomas. Nature Genet.32, 312–315 (2002).

18. Bonzo, J. A., Ferry, C. H., Matsubara, T., Kim, J. H. & Gonzalez, F. J. Suppression ofhepatocyte proliferation by hepatocyte nuclear factor 4a in adult mice. J. Biol.Chem. 287, 7345–7356 (2012).

19. Servitja, J. M. et al. Hnf1a (MODY3) controls tissue-specific transcriptionalprograms and exerts opposed effects on cell growth in pancreatic islets and liver.Mol. Cell. Biol. 29, 2945–2959 (2009).

20. Walesky, C. et al. Hepatocyte nuclear factor 4a deletion promotesdiethylnitrosamine-induced hepatocellular carcinoma in rodents. Hepatology 57,2480–2490 (2013).

21. Yanger, K. et al. Robust cellular reprogramming occurs spontaneously during liverregeneration. Genes Dev. 27, 719–724 (2013).

22. Preisegger, K. H. et al. Atypical ductular proliferation and its inhibition bytransforming growth factor beta1 in the 3,5-diethoxycarbonyl-1,4-dihydrocollidine mouse model for chronic alcoholic liver disease. Lab. Invest. 79,103–109 (1999).

23. Malato, Y. et al. Fate tracing of mature hepatocytes in mouse liver homeostasis andregeneration. J. Clin. Invest. 121, 4850–4860 (2011).

24. Shin, S. et al. Foxl1-Cre-marked adult hepatic progenitors have clonogenic andbilineage differentiation potential. Genes Dev. 25, 1185–1192 (2011).

25. Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia.N. Engl. J. Med. 368, 2059–2074 (2013).

26. O’Dell, M. R. et al. Kras(G12D) and p53 mutation cause primary intrahepaticcholangiocarcinoma. Cancer Res. 72, 1557–1567 (2012).

27. Avruch, J., Zhou, D., Fitamant, J. & Bardeesy, N. Mst1/2 signalling to Yap:gatekeeper for liver size and tumour development. Br. J. Cancer 104, 24–32(2011).

28. Tefferi, A. et al. IDH1 and IDH2 mutation studies in 1473 patients with chronic-,fibrotic- or blast-phase essential thrombocythemia, polycythemia vera ormyelofibrosis. Leukemia 24, 1302–1309 (2010).

29. Watanabe, T., Nobusawa, S., Kleihues, P. & Ohgaki, H. IDH1 mutations are earlyevents in the development of astrocytomas and oligodendrogliomas. Am. J. Pathol.174, 1149–1153 (2009).

Supplementary Information is available in the online version of the paper.

Acknowledgements We thank R. Mostoslavsky, L. Ellisen, A. Kimmelman, andmembers of the Bardeesy laboratory for valuable input. We also thank S. Thorgeirssonand J. Andersen for sharing unpublished data sets. This work was supported bygrants from TargetCancer Foundation and the National Institutes of Health(R01CA136567-02 and P50CA1270003) to N.B. N.B. holds the Gallagher EndowedChair in Gastrointestinal Cancer Research at Massachusetts General Hospital. S.K.S. isthe recipient of a Cholangiocarcinoma Foundation/Conquer Cancer Foundation ofASCO Young Investigator Award, and an American Cancer Society PostdoctoralFellowship (PF-13-294-01-TBG). C.A.P. is the recipient of a Canadian Institutes ofHealth Research postdoctoral fellowship. N.B., J.M.L. and D.S are members of theSamuel Waxman Cancer Research Foundation Institute Without Walls. J.M.L. and D.S.are supported by the Asociacion Espanola para el Estudio del Cancer.

Author Contributions S.K.S. and C.A.P. contributed equally to the study. S.K.S., C.A.P.,K.S.G., M.S.N. and S.G. carried out the experiments involving hepatoblasts and mousemodels. K.N.R. andS.R.performedcomputational analysisongeneexpressiondata. J.F.performed immunohistochemistry on tissue sections. E.A.A. and K.-K.W. assisted withthe generation of the IDH mutant mice. V.D. analysed the histology from the murineliver specimens. C.W. and U.A. carried out the experiments involving the HNF-4aknockout mice. D.S., H.C., O.M. and J.M.L. performed GSEA analysis on human IHCCsamples. A.X.Z., A.F.H. and C.R.F. were involved in the study design. K.E.Y., K.S.S., J.T.,J.P.-M. and C.G. developed and provided the AGI-5027 compound and measured 2HGin our samples. S.K.S., C.A.P. and N.B. designed the experiments and wrote the paper.N.B. supervised the studies. All authors discussed the results and commented on themanuscript.

Author Information Affymetrix Mouse 420Av2 DNA microarray data have beendeposited in the Gene Expression Omnibus under accession number GSE57002.Reprints and permissions information is available at www.nature.com/reprints. Theauthors declare competing financial interests: details are available in the online versionof the paper. Readers are welcome to comment on the online version of the paper.Correspondence and requests for materials should be addressed to N.B.(Bardeesy.Nabeel@mgh.harvard.edu).

LETTER RESEARCH

0 0 M O N T H 2 0 1 4 | V O L 0 0 0 | N A T U R E | 5

Macmillan Publishers Limited. All rights reserved©2014

METHODSCell culture. Hepatoblasts were prepared from wild-type mice at embryonic day14 and immortalized by plating at clonal density13. Hepatoblasts were maintainedin HB media (DMEM/F-12 (Gibco Life Technologies) containing 10% fetal bovineserum, 1% penicillin-streptomycin, 50 ng ml21 epidermal growth factor, 30 ng ml21

insulin-like growth factor II (PeproTech), 10 mg ml21 insulin (Roche)) on platescoated with rat tail collagen (BD Biosciences) in a humidified atmosphere with5% CO2 at 37 uC. For IDH1 experiments, culture media was supplemented with25 ng ml21 of doxycycline (d9891, Sigma-Aldrich). For hepatocyte differentiationassays, 5 3 106 hepatoblast cells were cultured on uncoated 10 cm tissue culturedishes in HB cell medium for up to 5 days. Pictures were taken 2–5 days after plat-ing. For bile duct differentiation (matrigel) assays, 0.5 ml of Basement MembraneMatrix (BD) was coated onto 6 cm tissue culture dishes and allowed to set for 1 h.Then 0.5 3 106 hepatoblast cells were added to the plate in HB cell media supple-mented with 100 ng ml21 recombinant mouse hepatocyte growth factor (HGF)(R&D Systems). Tubules were counted after 24 h. For gene expression analysis, cellswere isolated after 10 days. For proliferation assays, cells were plated in duplicate oncollagen-coated 6-well dishes (1 3 104 per well) or uncoated 6-well plates (1 3 105

per well) in culture medium. Adherent and non-adherent cells were harvested,trypsinized and counted by trypan-blue exclusion using a Countess AutomatedCell Counter (Invitrogen) 24, 72 or 120 h later. Negative mycoplasma contamina-tion status of all cell lines and primary cells used in the study was established usingLookOut Mycoplasma PCR Kit (Sigma, MP0035).AGI-5027. AGI-5027 (ML309) is a small molecule (phenyl-glycine scaffold) inhib-itor of IDH1 R132C and R132H (IC50 5 30 nm and 68 nm, respectively) that doesnot affect wild-type IDH1 activity (IC50 . 36mM and was developed by NIH Chem-ical Genomics Center and Agios Pharmaceuticals14.Mice. Mice were housed in pathogen-free animal facilities. All experiments wereconducted under protocol 2005N000148 approved by the Subcommittee on ResearchAnimal Care at Massachusetts General Hospital. Mice were maintained on a mixed129SV/C57BL/6 background. Data presented include both male and female mice.All mice included in the survival analysis were euthanized when criteria for diseaseburden were reached.The Alb-Cre strain30 harbours a transgene in which the mouse albumin enhancer/promoter drives Cre recombinase. Cre activity is inefficient at birth, but leads toprogressive deletion of floxed sequences in all liver lineages by 4–6 weeks of age. Inthe Rosa26-LSL-rtTA strain31, the doxycycline-responsive reverse tetracycline trans-activator (rtTA) is knocked into the transcriptional start site at the ubiquitouslyexpressed Rosa26 locus and is preceded by a lox-stop-lox (LSL) casette. The LSL-KrasG12D strain26 consists of a mutant KrasG12D allele knocked into the endogenousKras locus, preceded by an LSL cassette. For IDH-mutant transgenic mouse mod-els, transgenic mice were generated as follows. The R140Q and R172K mutationswere introduced into human IDH2 cDNA (Origene, MD # SC319226, NCBI RefSeqaccession number NM_002168.3) using the Quickchange kit (Agilent Lexington,MA #200521) as per the manufacturer’s instructions. Two different transgenic tar-geting approaches were used. The first was used to generate doxycycline-induciblemodels. To generate the doxycycline-inducible Tet-R140Q and Tet-R172K strains(Extended Data Fig. 4), IDH2R140Q cDNA was cloned into the pBS31 Prime trans-genic targeting vector (Thermo Scientific Open Biosystems, #MES4487). A modi-fied version of pBS31 containing a floxed stop cassette was used for the Tet-R172Kmice. The targeting vectors were co-electroporated into C2 mouse embryonic stemcells (Thermo Scientific Open Biosystems, #MES4305) with a plasmid expressingFLPE recombinase (pCAGGS-FlpE, Thermo Scientific Open Biosystems, #MES4488)as described32. This system involves single copy integration of the transgene down-stream of the collagen-1 locus, and within the liver, results in expression in hepato-cytes but not in bile ducts32. Embryonic stem (ES) cells were screened for integrationof the transgene by PCR and correctly targeted ES cells were injected into C57BL/6blastocysts. Chimeras were crossed to the Alb-Cre strain30, and offspring were suc-cessively crossed with Rosa26-LSL-rtTA31 and LSL-KrasG12D mice26. We confirmedthe expected expression of the R140Q (Extended Data Fig. 4b, c) and R172K trans-genes (Extended Data Fig. 4d) in the adult hepatocytes and lack of expression inbile ducts.

The second approach was the stably expressed model. To generate mice with aconditionally active IDH2R172K allele (LSL-IDH2R172K), human IDH2R172K was clonedinto a modified version of the PGK-ATG-Frt vector (Open Biosystems, #MES4490)containing the CAGGS promoter for stable expression (Extended Data Fig. 7a)33.Gene targeting and chimaeric mouse generation was carried out as described earlier.This strain was successively crossed with Alb-Cre and LSL-KrasG12D animals. Thistransgene was specifically expressed in bile ducts and oval cells (Fig. 3g and ExtendedData Fig. 7b). We did not observe developmental anomalies in these mice, althoughthe inefficiency of recombination induced by the Alb-Cre strain in neonates andthe restricted activity of the transgene in this system precludes definitive assess-ment of potential effects of mutant IDH2 on the embryonic liver.

DDC treatment. For DDC experiments, 4-week-old mice were treated with200mg ml21 doxycycline-containing water. Two days later, mice were fed with 0.1%DDC-containing chow (F4643, Bio-serv) for 5 days. The mice were then switchedto normal chow and maintained on doxycycline-containing water until harvest 1or 3 weeks later. Randomization was done according to genotype and blinding wasapplied during histological analysis. No mice were excluded from the analysis.Sample sizes were chosen to achieve statistical significance yet use the fewest ani-mals possible to minimize suffering.Statistics. Results are expressed as mean 6 s.d. between two technical replicates,unless otherwise specified. Significance was analysed using two-tailed Student’st-test. A P value of less than 0.05 was considered statistically significant.Plasmids. Human wild-type IDH1 cDNA (NCBI RefSeq accession number NM_005896.3) was obtained from Origene and subcloned into pRetro-Puro using EcoRIand BamHI fragments. Human wild-type IDH2, R140Q and R172K cDNA (Origene)were subcloned into pRetro-puro and pMSCV-blast (derived by replacing the pur-omycin resistance gene from pMSCV-puro with a blasticidin resistance gene) (bothfrom Clontech) using EcoRI and XhoI. Human IDH1R132C was generated fromwild-type IDH1 using QuikChange Lightning kit (Agilent Technologies) andprimers 59-GATGGGTAAAACC TATCATCATAGGTTGTCATGCTTATGGGGATCAATAC-39 for sense and 59-GTATTGATCCCCATAAGCATGACAACCTATGATGATAGGTTTTACCCATC-39 for antisense. pLVX-Tet-On was obtainedfrom Clontech. Murine Hnf4a was obtained from Addgene and cloned into pRetro-blast (derived by replacing the puromycin resistance with blasticidin) as a BglII-EcoRI fragment. Lentiviral (pLKO.1) Hnf4a shRNA vectors TRCN0000026154,target sequence: 59-CGACAATGTGTGGTAGACAAA-39; and TRCN0000026216,target sequence: 59-GCAGATTGATGACAATGAATA-39 were used as shHnf4a#1and shHnf4a#2, respectively, and were obtained from the Broad Institute TRCshRNA library. pLKO.1 shRNA with target sequence 59-GCAAGCTGACCCTGAAGTTCAT-39 was used as negative control shRNA. Hnf4a resistant to shHnf4a#2was generated using QuikChange Lightning kit and primers 59-GCCCTTCCAAGAGCTGCAGA TCGACGATAACGAATATGCCTGCCTCAAAGCC-39 for senseand 59-GGCTTTGAGGCAGGCATATTCGTTATCGTCGATCTGCAGCTCTTGGAAGGGC-39 for antisense.Immunoblot analysis. Cell extracts were prepared in 13 RIPA buffer (150 mMNaCl, 1% IGEPAL, 0.1% SDS, 50 mM Tris, 0.5% DOC) supplemented with a pro-tease inhibitor cocktail (Complete, Roche Applied Science) and phosphatase inhi-bitors (Phosphatase Inhibitor Cocktail Sets I and II, Calbiochem) and quantifiedby BCA Protein Assay (Thermo Scientific). Thirty-micrograms protein was resolvedon 9% SDS–PAGE gels and transferred onto PVDF membranes (GE Healthcare LifeSciences). Membranes were blocked in TBS with 5% non-fat milk and 0.1% Tweenand probed with anti-IDH1 polyclonal antibody (3997S, Cell Signaling), anti-IDH2polyclonal antibody (ab84726, Abcam), anti-occludin (71-1500, Invitrogen), anti-HNF-4a(126) monoclonal antibody (ab41898, Abcam), anti-HNF-4a (126)/NR2A1monoclonal antibody (PP-K9218-00, R&D Systems), anti-HNF-4a(729) mono-clonal antibody (PP-H6939-00, R&D Systems), or anti-b-actin monoclonal anti-body (A5316, Sigma-Aldrich) as loading control. Bound proteins were detected withhorseradish-peroxidase-conjugated secondary antibodies (Vector Biolaboratories)and SuperSignal West Pico Luminol/Enhancer Solution (Thermo Scientific).RNA analysis. Total cellular RNA was extracted using RNeasy Mini Kit (Qiagen).cDNA synthesis was performed with the QuantiTect Reverse Transcription Kit(Qiagen) with 1mg total RNA. Sense and antisense for amplification of target genesincluded: Alb, 59-CATGCCAAATTAGTGCAGGA-39 and 59-GCTGGGGTTGTCATCTTTGT-39; Aldob, 59-TGTCTGGAGGTATGAGTGAGG-39 and 59-CTGGGTTGCCTTCTTGTTTGC-39; Adh1, 59-GTGACTTGTGTGAAACCAGGT-39

and 59-GCTACAAAAGTTGCTTTCCGGG-39; Ggt1, 59-CTTGTGCGAGGTGTTCTGC-39 and 59-GCATAGTCACCGTCTCTCCTT-39; Bgp, 59-CAAAAGGAGGCCTCTCAGAT-39 and 59-GCTGAGGGTTTGTGCTCTGT-39; Hnf4a(1–6),59-GGTAGGGGAGAATGCGACTC-39 and 59-AAACTCCAGGGTGGTGTAGG-39; Foxm1, 59-ATCGCTACTTGACATTGGACCA-39 and 59-GATTGGGTCGTTTCTGCTGTG-39; Krt19, 59-TGCTGGATGAGCTGACTCTG-39 and 59-AATCCACCTCCACACTGACC-39; Onecut1, 59-GGCAACGTGAGCGGTAGTTT-39 and 59-TTGCTGGGAGTTGTGAATGCT-39; Arg1, 59-TTGGGTGGATGCTCACACTG-39 and 59-GTACACGATGTCTTTGGCAGA-39; C2, 59-GGAACCCATTTGCCGACAG-39 and 59-GGCCCAAACTTTTTGTCAGAAG-39; C7, 59-GCAGGAAAGTGTTCAGCGG-39 and 59-CATGACCGTAAGTATTCCGTGAG-39; C9, 59-CATGCCGTGACCGAGTAGC-39 and 59-TCACAGAGTCCGTTGTAGAACT-39; Krt7, 59-CACCCGGAATGAGATTGCG-39 and 59-GCACGCTGGTTCTTCAAGGT-39; Sprr1a, 59-TTGTGCCCCCAAAACCAAG-39 and 59-GGCTCTGGTGCCTTAGGTTG-39; Ttr, 59-CACCAAATCGTACTGGAAGACA-39 and 59-GTCGTTGGCTGTGAAAACCAC-39; ApoB, 59-TGGCTCTGATCCCAAATCCCT-39 and 59-GCTGCTCCTTGGCAGTATTAAA-39; Cyp3a13, 59-GACCACAAGCAGTGCTCTTTC-39 and 59-GCAGGGTATCATAGGTGGCAG-39.

RESEARCH LETTER

Macmillan Publishers Limited. All rights reserved©2014

Gene expression profiling and GSEA. Gene expression of hepatoblasts expres-sing IDH1 wild type, IDH1 R132C, IDH2 wild type, R172K and empty vector con-trols (N 5 2 cultures for each condition) grown on collagen-coated plates, as wellas IDH1 R132C and empty vector controls grown on uncoated plates, was evaluatedusing Affymetrix Mouse 430Av2 DNA microarrays (DFCI Microarray core facil-ity). Raw expression values were processed and normalized using RMA in the RBioconductor package. Differentially expressed genes in IDH1(R132C)- and IDH2(R172K)-expressing hepatoblasts were identified using the signal-to-noise metric.GSEA of the expression data was used to uncover the molecular program account-ing for the differentiation defects. GSEA was performed using two libraries fromversion 3.1 of the molecular signature database (http://www.broadinstitute.org/gsea/msigdb/index.jsp): the c2 curated gene sets from online pathway databases,PubMed publications, knowledge of domain experts and the c3 motif gene sets.Pairwise GSEA analysis such as shown in Fig. 2a was performed by creating rankedlists of genes using the log2 ratio of mutant IDH samples to vector or wild-typecontrol and running GSEA in the ranked list mode. The heat map in Fig. 2a pre-sents the gene sets that distinguish IDH1 R132C or IDH2 R172K as compared tocells expressing wild-type IDH1 or IDH2 or empty vector. Each column representspairwise analysis of hepatoblast cells expressing mutant IDH (IDH1 R132C or IDH2R172K) with either empty vector (left columns) or wild-type IDH (right columns)control cells for two independent replicates of each sample. All other results wereobtained using GSEA to rank genes according to signal-to-noise ratio between mu-tant IDH samples and controls. Statistical significance for GSEA results were obtainedfrom permuting gene sets (1,000 permutations). Gene sets of ‘hepatocyte genes’for the GSEA results shown in Extended Data Fig. 1g were created from expressiondata from Shin, et al. 201124. The expression data were downloaded from the GeneExpression Omnibus (accession number GSE28891) in series matrix form and loadedinto R. Undifferentiated, differentiated hepatocytes, and primary hepatocyte sampleswere selected from GSE28891 (4 samples of each type for a total of 12). Within thatsample subset, the normalized signal intensities were thresholded to a minimumvalue of 10 and features were filtered out if they had less than threefold differencebetween minimum and maximum values or had absolute differences less than 50between minimum and maximum values (leaving 21,277 features out of 37,908).A t-test was performed between the undifferentiated and the combined differen-tiated and primary hepatocyte samples and P values from the t-test were correctedfor multiple hypothesis testing using the Benjamini–Hochberg correction (in Rusing the Bioconductor multtest package). This left 206 features up and 407 downin the hepatocyte direction with an adjusted Benjamini–Hochberg P-value , 0.01,which are represented by 133 and 311 unique gene symbols, respectively. Ninety-six of these genes were represented on the murine Affymetrix array. The featurelist was made into a gene set labelled as ‘hepatocyte genes’ and is shown in Sup-plementary Table 1.Chromatin immunoprecipitation. 53 106 empty vector, IDH1 wild-type or R132C-expressing hepatoblasts were transferred from collagen-coated to uncoated platesfor 5 days and fixed in 1% formaldehyde for 10 min at 37 uC. Crosslinking wasterminated by the addition of glycine at a final concentration of 125 mM for 5 min.Cells were then harvested and washed twice with ice-cold PBS. Chromatin extrac-tion, DNA sonication, and immunoprecipitation were performed using a chro-matin immunoprecipitation (ChIP) Assay Kit (17-295, Millipore) and ChIPAb1

Trimethyl-Histone H3 (Lys4) (17-614, Millipore) or Tri-Methyl-Histone H3 (Lys4)Antibody (Cell Signaling, #9727S) validated antibody. Sense and antisense for am-plification of target genomic regions included: Hhex, 59-GAAGGAGCCTGACCCTTTCC-39 and 59-ATCAGCAGCGTGCACTACTC-39; Hoxa10, 59-AACAGTAAAGCTTCGCCGGA-39 and 59-AGTTCAGAAGGTCAGCCTGC-39; Hnf4aP1, 59-AACATGGCCCTATCTTCGGG-39 and 59-AAACGCACACCGCTATGTTG-39.Histology and immunostaining. Tissue samples were fixed overnight in 4% buf-fered formaldehyde, and then embedded in paraffin and sectioned (5 mm thick-ness) by the DF/HCC Research Pathology Core. Haematoxylin and eosin stainingwas performed using standard methods. For immunohistochemistry, unstainedslides were baked at 55 uC overnight, deparaffinized in xylenes (two treatments,6 min each), rehydrated sequentially in ethanol (5 min in 100%, 3 min in 95%, 3 minin 75%, and 3 min in 40%), and washed for 5 min in 0.3% Triton X-100/PBS (PBST)and 3 min in water. For antigen unmasking, specimens were cooked in a 2100 Anti-gen Retriever (Aptum Biologics) in 13 Antigen Unmasking Solution, Citric AcidBased (H-3300, Vector Laboratories), rinsed three times with PBST, incubated for10 min with 1% H2O2 at room temperature to block endogenous peroxidase ac-tivity, washed three times with PBST, and blocked with 5% goat serum in PBST for1 h. Primary antibodies were diluted in blocking solution as follows: anti-R172K(26163, New East Biosciences) 1:100; anti-R140Q (26165 New East Biosciences)1:100; anti-IDH2 (NBP2-22166, Novus Biologicals) 1:100; anti-HNF-4a mono-clonal antibody (PP-K9218-00, R&D Systems) 1:250; anti-CK19 (Troma3) 1:100;anti-SOX9 (ab5535, Millipore) 1:300; or anti-Ki-67 (ab15580, Abcam) 1:100, and

incubated with the tissue sections at 4 uC overnight. Specimens were then washedthree times for 3 min each in PBST and incubated with biotinylated secondaryantibody (Vector Laboratories) in blocking solution for 1 h at room temperature.Then, specimens were washed three times in PBST and treated with ABC reagent(Vectastain ABC kit #PK-6100) for 30 min, followed by three washes for 3 mineach. Finally, slides were stained for peroxidase for 3 min with the DAB (di-amine-benzidine) substrate kit (SK-4100, Vector Laboratories), washed with water andcounterstained with haematoxylin. Stained slides were photographed with an Olym-pus DP72 microscope. For immunofluorescence studies of tissue specimens, antigenunmasking was performed in 13 Antigen Retrieval Buffer, Tris-EDTA Buffer, pH9.0 (ab93684, Abcam), rinsed briefly with dH2O, washed for 5 min in PBS, followedby 5 min in 0.05% Tween-20/TBS (TBST). Specimens were then blocked using 0.5%blocking reagent (FP1020, PerkinElmer) diluted in TBS (TNB) for 1 h and incu-bated in primary antibodies at 4 uC overnight in TNB, followed by 3 min washes inTBST and 1 h incubation in secondary antibodies (1:400) diluted in TNB. Finally,specimens were rinsed 3 3 5 min in TBST and mounted using mounting mediawith DAPI (H-1500, Vectashield Vector Laboratories). For cell culture staining,samples were fixed in 4% paraformaldehyde (Fisher Scientific) in PBS for 20 minfollowed by washing four times in PBS. Residual paraformaldehyde was removedwith three 5 min washes with 100 mM glycine in PBS. Cells were permeabilizedwith 0.3% Triton X-100/PBS and blocked for 30 min with blocking buffer (5%bovine serum albumin, 0.2% Triton X-100, 0.05% Tween 20, PBS). Coverslips wereincubated with primary (1:100) and secondary (1:1,000) antibodies diluted in block-ing buffer for 1 h and 40 min, respectively, at room temperature, and nuclei werecounterstained with DAPI. Coverslips were mounted with Immu-Mount (Thermo-Shandon). The following primary antibodies were used: rabbit anti-SOX9 (1:500;Millipore AB5535); goat anti-HNF-4a (SC6556, 1:250; Santa Cruz). Secondaryantibodies were conjugated to Alexa-488, 555 or 647 (Life Technologies). Confocalimages were captured by a scanning laser confocal microscope (Nikon Eclipse Ti;Nikon) using 340 or 360 oil lenses and captured using NIS-Elements software(Nikon). Images were processed using ImageJ and Adobe Photoshop CS4. For quan-tification of signal intensity ratios, channel intensities of individual cells were selectedfrom at least 60 cells per sample from at 3–5 images using the Measure RGB pluginof ImageJ and exported into Excel for further analysis.2HG production and measurement. Octyl-2HG esters were prepared from com-mercially available 2HG lactones (CAS# 21461-84-7 (S-) and 53558-93-3 (R-)).Esterification of lactone with 1-octanol and EDCI (CAS# 25952-53-8), followed byopening of the lactone with lithium hydroxide in tetrahydrofuran/water, yieldedthe 2HG-octyl esters (free carboxylic acid) upon neutralization and chromatog-raphic purification. The sodium salts of both octyl-2HG esters are available fromToronto Research Chemicals. Liquid chromatography coupled with tandem massspectrometry (LC-MS/MS) analysis was performed using an AB Sciex 4000 oper-ating in negative electrospray mode. MRM data were acquired for each compoundusing the following transitions: 2HG (146.9/128.8 amu), 13C5-2HG (151.9/133.8amu) and 3HMG (160.9/98.9 amu). Chromatographic separation was performedusing an ion-exchange column (Bio-Rad Fast Acid Analysis, 9mm, 7.8 3 100 mm;Bio-Rad). The flow rate was 1 ml min21 of 0.1% formic acid in water with a total runtime of 4 min. Samples were prepared as follows. Tissues were weighed out andhomogenized in appropriate volumes of PBS using Tissue Lyser (Qiagen). Ten-microlitres of homogenate were extracted by adding 200ml of methanol with200 ng ml21 3HMG as internal standard. Cell pellets containing 1 3 106 cells eachwere resuspended in specified volumes of 80:20 MeOH:water, centrifuged for 10 minat 14,000 r.p.m. Thirty-microlitres of supernatant were extracted by adding 170mlof methanol with 200 ng ml21 13C5-2HG as internal standard. All sample typeswere then vortexed, centrifuged at 4,000 r.p.m. at 5 uC and 150ml of supernatanttransferred to a clean 96-well plate. The samples were dried down and reconsti-tuted in 200ml 0.1% formic acid in water, and 10ml injected on a column.Human samples and genomic studies. An initial cohort of 149 formalin-fixed,paraffin-embedded (FFPE) samples was obtained from IHCC patients resected bet-ween 1995 and 2007 at three centres from the HCC Genomic Consortium: IRCCSIstituto Nazionale Tumori (Milan), Mount Sinai School of Medicine (New York),and Hospital Clinic (Barcelona). Pathological diagnosis of IHCC was confirmed bytwo independent liver pathologists (M.S. and S.T.). DNA and RNA extraction fromhuman tissue, whole-genome gene-expression profiling (DASL assay, Illumina),and the Infinium FFPE restoration kit (Illumina) are extensively described else-where35. The study protocol was approved by each centre’s Institutional ReviewBoard. A total of 127 samples had available data for both gene expression and DNAcopy number changes and were analysed for molecular characterization to definemolecular-based classes. IDH1 and 2 mutations occurring at arginine residue 132and 172 respectively were analysed by PCR using primers 59-GTGGCACGGTCTTCAGAGA-39 (forward primer) and 59-TGCTTAATGGGTGTAGATACCAA-39

(reverse primer), and 59-GCCCACACATTTGCACTCTA-39 (forward primer) and59-AAGGAAAGCCACGAGACAGA-39 (reverse primer). The PCR amplifications

LETTER RESEARCH

Macmillan Publishers Limited. All rights reserved©2014

were performed in a volume of 25ml reaction mixture containing 1.5 mM MgCl2,0.2 mM of each dNTP, 0.125 mM of each primer and 1 U of Platinum Taq DNAPolymerase (Invitrogen). PCR products were purified using the Qiaquick PCRpurification kit (Qiagen) and sequenced using an Applied Biosystems 3700 DNAsequencer (ABI PRISM 3730XL; Applied Biosystems). Good quality Sanger sequenc-ing data for both IDH1 and IDH2 was obtained for 107 IHCC DNA samples andhas been included in all the subsequent analyses.Human GSEA. Genes, molecular pathways, and gene expression signatures asso-ciated with the classes were evaluated using GSEA for Molecular Signature Data-base gene sets (MSigDB, http://www.broadinstitute.org/msigdb). Data analysis wasconducted using the GenePattern Analytical Toolkit34, whereas correlation withclinico-pathological parameters was performed with SPSS software (version 18) aspreviously described35.

30. Postic, C. & Magnuson, M. A. DNA excision in liver by an albumin-Cre transgeneoccurs progressively with age. Genesis 26, 149–150 (2000).

31. Belteki, G. et al. Conditional and inducible transgene expression in mice throughthe combinatorial use of Cre-mediated recombination and tetracyclineinduction. Nucleic Acids Res. 33, e51 (2005).

32. Beard, C., Hochedlinger, K., Plath, K., Wutz, A. & Jaenisch, R. Efficient method togenerate single-copy transgenic mice by site-specific integration in embryonicstem cells. Genesis 44, 23–28 (2006).

33. Giacometti, E., Luikenhuis, S., Beard, C. & Jaenisch, R. Partial rescue of MeCP2deficiency by postnatal activation of MeCP2. Proc. Natl Acad. Sci. USA 104,1931–1936 (2007).

34. Reich, M. et al. GenePattern 2.0. Nature Genet. 38, 500–501 (2006).35. Sia, D. et al. Integrative molecular analysis of intrahepatic cholangiocarcinoma

reveals 2 classes that have different outcomes. Gastroenterology144, 829–840 (2013).

36. Oishi, N. et al. Transcriptomic profiling reveals hepatic stem-like genesignatures and interplay of miR-200c and epithelial-mesenchymal transitionin intrahepatic cholangiocarcinoma. Hepatology 56, 1792–1803(2012).

37. Kang, J. S., Wanibuchi, H., Morimura, K., Gonzalez, F. J. & Fukushima, S. Role ofCYP2E1 in diethylnitrosamine-induced hepatocarcinogenesis in vivo. CancerRes. 67, 11141–11146 (2007).

38. Verna, L., Whysner, J. & Williams, G. M. N-nitrosodiethylamine mechanisticdata and risk assessment: bioactivation, DNA-adduct formation,mutagenicity, and tumor initiation. Pharmacol. Ther. 71, 57–81(1996).

RESEARCH LETTER

Macmillan Publishers Limited. All rights reserved©2014

Extended Data Figure 1 | Impact of mutant IDH1 and IDH2 onhepatoblast cell differentiation. a, Primary hepatoblast cells were engineeredto express the indicated human IDH alleles or empty vector (EV) under adoxycycline (Dox)-inducible system. Lysates from hepatoblast cells cultured inthe presence of increasing doxycycline concentrations were analysed byimmunoblot using an antibody that recognizes both murine and human IDH1(top) or with an antibody specific to human IDH2 (lower panel); actin is theloading control. Note that 25 ng ml21 doxycycline induces physiological levelsof IDH1 expression and was used in the experiments shown in Figs 1 and 2.b, Photomicrographs of cells grown for 2 days on collagen-coated plates in thepresence of 25 ng ml21 doxycycline. WT, wild type. c, Hepatoblast cellscultured in the presence of increasing doxycycline concentrations wereanalysed by liquid chromatography/mass spectrometry (LC/MS) for levels ofintracellular 2HG. d, Growth curve of hepatoblasts cultivated on collagen-coated dishes. e, f, Hepatoblast cells from a were grown on collagen-coated or

uncoated plates and analysed for expression of hepatocyte markers by qRT–PCR. g, Enrichment plot showing downregulation of a hepatocyte gene set(Gene Expression Omnibus accession number GSE28892) in R132C-expressing cells. FDR, false discovery rate; NES, normalized enrichment score.h, 2HG levels in hepatoblast cells expressing the indicated IDH alleles or emptyvector and treated with AGI-5027 (1) or DMSO vehicle (2). i, j, Wild-typehepatoblasts were treated with 500mm octyl-(R) or -(S) enantiomers of 2HG orDMSO vehicle, and tested for hepatocyte differentiation upon transfer touncoated plates by hepatocyte sphere formation (i) or qRT–PCR(j). k, l, Hepatoblast cells expressing the indicated alleles were tested for biliarydifferentiation upon transfer to matrigel as assessed by qRT–PCR for theinduction of biliary makers Bgp and Ggt1. Scale bars, 100mm. Error barsindicate 6 s.d. between technical duplicates. *P , 0.05, Student’s t-test. Dataare representative of at least two independent experiments.

LETTER RESEARCH

Macmillan Publishers Limited. All rights reserved©2014

Extended Data Figure 2 | Mutant IDH represses the HNF-4a-mediatedhepatocyte differentiation program. a–c, Gene expression profiling ofhepatoblast cells expressing the indicated alleles and grown on collagen.a, b, Clustering analysis (a) and enrichment plot (GSEA) (b) showingdownregulation of HNF-4a targets in the IDH-mutant cells relative to controls.c, GSEA plots using the collection of cis regulatory elements from the MolecularSignatures Database (C3 collection) reveals strong downregulation of genescontaining consensus HNF-4a- and HNF1a-binding sites in the IDH-mutanthepatoblast cultures. d, qRT–PCR showing expression of Hnf4a and its targetgenes in hepatoblast cells expressing either empty vector (EV) or IDH2 R172Kand grown on collagen-coated plates. e, HNF-4a(7–9) expression(immunoblot) in hepatoblast cells grown on collagen-coated plates(quantification of HNF-4a:actin is indicated). Data are representative of twoindependent experiments. f, g, Immunofluorescence staining for HNF-4a(using an antibody that detects all isoforms) in hepatoblast cells expressingempty vector or IDH1 R132C and grown on collagen (f). The chart showsquantification of the ratio of the HNF-4a and DAPI signals. The specificity ofthe antibody was demonstrated by the diminished staining in cells with

knockdown of endogenous HNF-4a (g). h, Immunoblot showing that R132Csuppresses HNF-4a(1–6) and HNF-4a(7–9) isoforms on uncoated plates.i, j, AGI-5027 restores Hnf4a(1–6) induction in R132C-expressing hepatoblastsas shown by qRT–PCR (i) and immunoblot (j). DMSO 5 vehicle. k, Wild-typehepatoblasts were treated with 500mm octyl-(R) or -(S) enantiomers of 2HG,DMSO vehicle, or media alone. Hnf4a(1–6) levels were determined by qRT–PCR. l, m, Cultures of hepatoblast cells expressing the indicated alleles weregrown on uncoated plates for 5 days and subjected to chromatinimmunoprecipitation (ChIP) for H3K4me3 (l) or H3K27me3 (m). Enrichmentfor the promoter regions of Hhex (highly expressed gene in hepatoblast cells),Hoxa10 (transcriptionally silent in hepatoblast cells) and P1 Hnf4a wasmeasured by qPCR. n, o, Analysis of wild-type hepatoblasts expressing shRNAcontrol (shCTL) or targeting different HNF-4a sequences (shHnf4a#1,shHnf4a#2): hepatocyte marker expression (qRT–PCR) (n); HNF-4aimmunoblot (o). Error bars indicate 6 s.e.m. for f and 6 s.d. ford, i, k–n between technical duplicates. Scale bars, 50mm. *P , 0.05, Student’st-test.

RESEARCH LETTER

Macmillan Publishers Limited. All rights reserved©2014

b

a

HNF-4α(1-6)

Actin

EV

1 +

EV

2

EV

1 +

HN

F4α

R13

2C +

EV

2

R13

2C +

HN

F4α

shC

TL

shC

TL

shH

nf4α

shC

TL

shH

nf4α

shC

TL

shH

NF

4a

Rel

ativ

e ex

pres

sion

Collagen Matrigel

101

102

103

104

100

Bgp

shH

nf4α

shC

TL

shH

nf4α

Rel

ativ

e ex

pres

sion

Collagen Matrigel

101

102

103

100

Ggt1

Extended Data Figure 3 | HNF-4a is dispensable for biliary differentiation.a, Wild-type hepatoblast cells expressing shRNA control (shCTL) or targetingHNF-4a were tested for ability to undergo biliary differentiation upon transferto matrigel as assessed by tubule formation 24 h after transfer (left) and

qRT–PCR for induction of the biliary markers Bgp and Ggt1 10 days aftertransfer to matrigel (right). b, Immunoblot showing ectopic expression ofHNF-4a(1) in hepatoblast cells. Error bars indicate 6 s.d. between technicalreplicates. Scale bar, 100mm.

LETTER RESEARCH

Macmillan Publishers Limited. All rights reserved©2014

a

bTe

t-R

140Q

CK19HNF-4α IPAD/egreM)latot(IDH2-R140Q

WT

CK19HNF-4α IPAD/egreM)6-1( W

TTe

t-R

140Q

IDH2 (total)

IDH2-R172K HNF-4α (total) CK19 Merge/DAPI

Tet-

R17

2KW

T

d

Rosa26 locus

Col1A1 locus (3’ end) R140QCMV/TetO R172KCMV/TetO STOPloxP loxP

c

K271R-teTQ041R-teT

Col1A1 locus (3’ end)

Rosa26 locus STOP rtTAloxP loxP

Albumin CreAlbumin Cre

STOP rtTAloxP loxP

Extended Data Figure 4 | Analysis of genetically engineered mouse modelwith hepatocyte-specific mutant IDH expression. a, Schematic ofdoxycycline-inducible IDH2 mutant alleles (see Methods for details). Miceharbouring mutant human IDH2 alleles were crossed with Alb-Cre andRosa26-LSL-rtTA strains for liver-specific expression. b–d, Characterization ofexpression pattern of mutant IDH2 in Tet-R140Q and Tet-R172K geneticallyengineered models compared to control wild-type mice, after 4 weeks ofdoxycycline supplementation, reveals hepatocyte-specific expression,consistent with previous models using this transgenic targeting system36.b, Immunofluorescence analysis of Tet-R140Q and control wild-type livers

using an antibody that detects both endogenous and transgenic IDH2expression. Note that transgenic IDH2(R140Q) is expressed in the hepatocytesthat stain for HNF-4a(1–6), but not in the bile ducts that are marked by CK19.c, Immunofluorescence analysis of Tet-R140Q and control wild-type liversusing an antibody that is specific to IDH2(R140Q), showing an identicalpattern of transgene expression. d, Immunofluorescence analysis of Tet-R172K and control wild-type livers using an antibody that is specific toIDH2(R172K). This allele is also expressed in the hepatocytes, but shows morefocal expression compared to R140Q. In b and c an antibody to total HNF-4awas used to label hepatocytes. Scale bars, 50mm.

RESEARCH LETTER

Macmillan Publishers Limited. All rights reserved©2014

Extended Data Figure 5 | Characterization of genetically engineered mousemodel with hepatocyte-specific mutant IDH expression in the presence orabsence of liver injury. a, Measurement of 2HG levels in liver lysates fromIDH2-mutant and control mice treated with doxycycline for 1 month.b, Uninjured wild-type and Tet-R140Q livers exhibit comparable expression ofthe hepatocyte markers Hnf4a, Adh1, Alb, and Aldob and the biliary markersSprr1a and Onecut1 by qRT–PCR. c, d, Tet-R140Q mice and littermate controls

(WT) receiving doxycycline were fed a DDC-containing diet for 5 days beforebeing switched to normal chow for 1 week (see Fig. 3a for schematic). MutantIDH2 does not provoke liver injury as reflected by comparable levels of serumAST, Tbili and ALT (c), and absence of cleaved caspase-3 staining in Tet-R140Q compared to wild-type mice (d). Duodenum from a TNF-a-treatedmouse was used as a positive control for cleaved caspase-3 staining. Scalebars, 50mm. NS, not significant. Error bars represent 6 s.e.m.

LETTER RESEARCH

Macmillan Publishers Limited. All rights reserved©2014

Extended Data Figure 6 | Characterization of response to liver injury ingenetically engineered mouse model with hepatocyte-specific mutant IDHexpression. a–f, Tet-R140Q or Tet-R172K mice and littermate controls (WT)receiving doxycycline were fed a DDC-containing diet for 5 days before beingswitched to normal chow for 7 or 21 days (see Fig. 3a for schematic).a, Representative H&E images of livers from wild-type and Tet-R140Q miceanalysed at day 7 and 21. b, Quantification of Ki-67-positive cells for indicatedmarkers as shown in Fig. 3e (N 5 3 mice per group, from at least 5 high-powered fields per mouse scored). Error bars show 6 s.e.m. Data from Fig. 3f(day 21) is reproduced here for comparison. c, d, Immunofluorescence analysis

of IDH2-mutant livers showing that Ki-67 co-localizes with HNF-4a1 (c) andIDH2(R140Q)-expressing cells (d). As shown in Fig. 3e, these cells expresslower levels of HNF-4a compared to wild-type hepatocytes. e, f, Tet-R172Kmice and littermate controls (WT) receiving doxycycline were fed a DDC-containing diet for 5 days before being switched to normal chow for 1 week.Immunofluorescence analysis of IDH2 mutant livers shows that Ki-67 co-localizes with HNF-4a1 and IDH2(R172K)-expressing cells (e) and greaternumbers of Ki-671, HNF-4a1 cells in Tet-R172K compared to wild-type mice(f). Chart shows quantification. Error bars show 6 s.e.m. between 3 mice. Scalebars, 50mm. *P , 0.05, Student’s t-test.

RESEARCH LETTER

Macmillan Publishers Limited. All rights reserved©2014

Extended Data Figure 7 | Characterization of IDH2R172K geneticallyengineered mouse model. a, Schematic of LSL-IDH2R172K geneticallyengineered mouse model (see Methods for details). Mice harbouring the LSL-IDH2R172K allele were crossed with the Alb-Cre strain to target expression tothe liver. b, Immunofluorescence analysis revealed specific expression of

IDH2(R172K) in CK191 biliary cells, whereas hepatocytes were negative.c, Immunofluorescence analysis showing comparable low levels of proliferation(Ki-67) in IDH2R172K and littermate control livers (WT) at 3 months of age.Scale bars, 50mm.

LETTER RESEARCH

Macmillan Publishers Limited. All rights reserved©2014

Extended Data Figure 8 | The impact of HNF-4a ablation on IHCCpathogenesis in vivo. a–c, Alb-CreERT2;Hnf4afl/fl (HNF-4afl/fl;Cre) mice weretreated with DEN on postnatal day 15, and subsequently administeredtamoxifen (TAM) or corn oil (CO) control at 8 months of age as described20.DEN is activated to its carcinogenic form by cytochrome P450 (CYP) enzymes,including CYP2E1, specifically in hepatocytes37,38. Livers were harvested 2months later and analysed by H&E staining, immunofluorescence andimmunohistochemistry. a, b, In HNF-4afl/fl;Cre 1 TAM livers, HCC arisesfrom HNF-4a1 cells that have escaped HNF-4a ablation (a), while CK191

IHCC stains negative for HNF-4a (b, right). b, Left, normal liver from control

corn-oil-treated livers is shown. c, Representative H&E staining (top) andimmunofluorescence images (bottom) show marked expansion of SOX91 ovalcells in TAM-injected mice. Surrounding these oval cells are HNF-4a1

hepatocytes that escaped Hnf4a ablation in TAM-injected mice. d, H&E stainconfirming that the liver tumour arising in a KrasG12D (Alb-Cre;LSL-KrasG12D)mouse is an HCC. e, Measurement of 2HG in liver tumours fromIDH2R172K;KrasG12D (Alb-Cre;LSL-IDH2R172K;LSL-KrasG12D) mice comparedto normal liver from control mice (N 5 2). Scale bars, 50mm. Error barsshow 6 s.d. *P , 0.05, Student’s t-test.

RESEARCH LETTER

Macmillan Publishers Limited. All rights reserved©2014

a

Kra

sG12

D

HNF-4α(1) CK19 Merge

IDH

2R17

2K; K

rasG

12D

IDH2 (total)

HNF-4α (total) CK19 MergeIDH2 (R172K)

Kra

sG12

DID

H2R

172K

; Kra

sG12

D

b

Extended Data Figure 9 | Expression of mutant IDH2 in the IDH2R172K

genetically engineered mouse model. a, b, Immunofluorescence analysis tocharacterize expression of IDH2 in the liver of IDH2R172K;KrasG12D compoundmice. a, Staining with an antibody that recognizes both endogenous wild-typeand mutant IDH2 shows that endogenous IDH2 is expressed in the bile duct innon-diseased liver from KrasG12D mice (top). In IDH2R172K;KrasG12D mice,

total IDH2 staining is at near endogenous levels in the normal bile ducts(middle) and in the biliary intraepithelial neoplasia lesions (bottom).b, Immunofluorescence using an antibody specific to IDH2(R172K) showingspecific expression of the R172K transgene in CK191 bile ducts (middle) andbiliary intraepithelial neoplasias (bottom) in IDH2R172K;KrasG12D animals.Scale bars, 50mm.

LETTER RESEARCH

Macmillan Publishers Limited. All rights reserved©2014

aWT

IDH2R172K; KrasG12D

oval cell expansion IHCCIDH2R172K

R17

2K/S

OX

9/H

NF

-4a/

DA

PI

IHCC with hepatic stem cell/progenitor cell features

WT IDH1/2 mutant

Enr

ichm

ent s

core

(E

S)

-0.50

-0.30

0.00

-0.10

-0.40

-0.20

NES = -2.66FDR< 0.001p-val <0.001

b

Extended Data Figure 10 | Stem cell features in IDH-mutant murine andhuman IHCC pathogenesis. a, IDH2(R172K) expression using mutant-specific antibodies in livers from IDH2R172K and IDH2R172K;KrasG12D

compound mice showing specific expression of the R172K transgene in SOX91

oval cells and IHCC. Scale bars, 50mm. b, Normalized enrichment plot of a

cohort of 127 human IHCC samples genotyped for IDH status, showing thatthe subset of tumours with IDH1/IDH2 mutations exhibits enrichment of agene signature36 that identifies IHCCs with hepatic stem cell/progenitorfeatures.

RESEARCH LETTER

Macmillan Publishers Limited. All rights reserved©2014