© 2006 Nature Publishing Group http://www.nature.com/naturegenetics

ARTICLES

Essential role of Jun family transcription factors in PU.1 knockdown–induced leukemic stem cells Ulrich Steidl1, Frank Rosenbauer1,2, Roel G W Verhaak3, Xuesong Gu4, Alexander Ebralidze1, Hasan H Otu4,5, Steffen Klippel1, Christian Steidl6, Ingmar Bruns7, Daniel B Costa1, Katharina Wagner1, Manuel Aivado4, Guido Kobbe7, Peter J M Valk3, Emmanuelle Passegue´8, Towia A Libermann4, Ruud Delwel3 & Daniel G Tenen1 Knockdown of the transcription factor PU.1 (encoded by Sfpi1) leads to acute myeloid leukemia (AML) in mice. We examined the transcriptome of preleukemic hematopoietic stem cells (HSCs) in which PU.1 was knocked down (referred to as ‘PU.1knockdown HSCs’) to identify transcriptional changes preceding malignant transformation. Transcription factors c-Jun and JunB were among the top-downregulated targets. Restoration of c-Jun expression in preleukemic cells rescued the PU.1 knockdown– initiated myelomonocytic differentiation block. Lentiviral restoration of JunB at the leukemic stage led to loss of leukemic selfrenewal capacity and prevented leukemia in NOD-SCID mice into which leukemic PU.1-knockdown cells were transplanted. Examination of human individuals with AML confirmed the correlation between PU.1 and JunB downregulation. These results delineate a transcriptional pattern that precedes leukemic transformation in PU.1-knockdown HSCs and demonstrate that decreased levels of c-Jun and JunB contribute to the development of PU.1 knockdown–induced AML by blocking differentiation and increasing self-renewal. Therefore, examination of disturbed gene expression in HSCs can identify genes whose dysregulation is essential for leukemic stem cell function and that are targets for therapeutic interventions.

The transcription factor PU.1 (encoded by Sfpi1) is indispensable for myelomonocytic differentiation during normal hematopoiesis1. Several reports suggest that reduced function of PU.1 might also have a central role in AML, a disease entity characterized by disturbed myeloid development2,3. Recent experimental evidence proposes a model of AML pathogenesis in which two major molecular events are required for the development of malignant cells. One causes a differentiation arrest, and a second confers self-renewal properties on the cells, thereby ultimately leading to the formation of a pool of leukemic stem cells (LSCs)4. We have shown previously that the transcriptional control of Sfpi1 gene expression is mediated by a distal upstream regulatory element (URE) that is highly conserved among multiple species, including mice and humans5,6. We also demonstrated that knockout of this distal enhancer of Sfpi1, which reduces PU.1 expression levels by 80% in the bone marrow, leads to the development of AML in mice7,8. The course of disease includes a preleukemic phase with an accumulation of immature myelomonocytic cells in the marrow but still normal peripheral blood counts, followed by a leukemic phase with high numbers of malignant immature cells in the blood and marrow. However, the

molecular mechanisms underlying the malignant transformation are poorly understood. A widely used method to uncover oncogenic pathways is the examination of tumor cells, but this approach is hampered by at least two conceptual challenges: first, the bulk of tumor cells appear at the final stage of disease and hence are likely to show many nonspecific alterations beside the primary oncogenic events that lead to cancerous stem cell formation, and second, an adequate cellular control population is not available. This is particularly true for AML, in which a developmental block is a hallmark of the disease and results in a bulk tumor population of immature cells that cannot be easily compared with normal blood or bone marrow cells9. In this study, we carried out genome-wide transcriptional analysis of HSCs isolated from mice in which PU.1 was knocked down (referred to as ‘PU.1-knockdown mice’) at the preleukemic stage to identify prospective pathways that lead to LSC development. We delineate a transcriptional pattern in HSCs that precedes leukemic transformation and demonstrate that dysregulation of distinct identified targets is essential for LSC function, a finding that could be therapeutically useful in treatment of this disease.

1Harvard Institutes of Medicine, Harvard Medical School and Harvard Stem Cell Institute, Boston, Massachusetts 02115, USA. 2Max-Delbru ¨ ck-Center for Molecular Medicine, 13125 Berlin, Germany. 3Department of Hematology, Erasmus University Medical Center, 3015GE Rotterdam, The Netherlands. 4Beth Israel Deaconess Medical Center Genomics Center and Bioinformatics Core and Beth Israel Deaconess Medical Center, Boston, Massachusetts 02215, USA. 5Department of Genetics and Bioengineering, Yeditepe University, Istanbul 34755, Turkey. 6Department of Hematology and Oncology, University of Goettingen, 37075 Goettingen, Germany. 7Department of Hematology, Oncology and Clinical Immunology, University of Duesseldorf, 40225 Duesseldorf, Germany. 8Developmental and Stem Cell Biology Program, University of California, San Francisco, California 94314, USA. Correspondence should be addressed to D.G.T. ([email protected]).

Received 24 July; accepted 7 September; published online 15 October 2006; doi:10.1038/ng1898

NATURE GENETICS VOLUME 38

[

NUMBER 11

[

NOVEMBER 2006

1269

ARTICLES

1

10

100 100

102 1

10

1

10

2

10

3

10

Sca-1 FITC

4

10

100 100

103

2.84%

102 1

10

1

10

2

10

3

10

Sca-1 FITC

4

10

100 100

1

10

10

2

3

10

Sca-1 FITC

4

10

c 3 2.5 2

1.2 Control

1

1.5 1 0.5 0

0.8 0.6

HSC

0.4 0.2 0

BM 0

20

40 60 Time (d)

80

100

Figure 1 Examination of the phenotypic stem cell compartment and identification of LSCs in PU.1-knockdown mice. (a) Propidium iodide–negative Lin– c-kit+ Sca-1+ HSCs in wild-type (WT) and preleukemic and leukemic PU.1-knockdown (KD) mice. (b) Average percentage of HSCs in wild-type, preleukemic and leukemic PU.1-knockdown mice (n ¼ 3 in each group). Error bars indicate s.d. (c) Kaplan-Meier Plot of overall survival of NOD-SCID recipient mice after transplantation of bone marrow (BM; dotted line) or HSCs (solid line) of leukemic PU.1-knockdown mice (n ¼ 5 in each group). Mice injected with physiologic NaCl solution are displayed as a control (dashed line) (n ¼ 3).

Transcriptome differs in PU.1-knockdown and wild-type HSCs Linear amplification of RNA obtained from FACS-sorted HSCs and subsequent gene expression profiling showed that B21,500 out of 45,000 transcripts were expressed in PU.1-knockdown HSCs. Presence calls showed little variation (mean 47.7%, s.d. 1.27%). Scaling factors ranged between 0.86 and 1.09 (mean 1.02, s.d. 0.09). 3¢:5¢ ratios of Gapd and b-actin were 2.04 (s.d. 0.19) and 2.98 (s.d. 0.6), respectively, indicative of an efficient and robust amplification. We found that 225 transcripts were downregulated and 97 were upregulated in PU.1-knockdown HSCs (transcripts were considered upregulated or Figure 2 Expression profiling in PU.1-knockdown HSCs. (a) Scatter plot of expression of 45,000 transcripts (arbitrary units, Affymetrix Mouse Genome 430 2.0 Array) in three independent samples of wild-type HSCs versus three independent samples of PU.1-knockdown HSCs. Each individual sample contained HSCs pooled from three mice. Expression of the vast majority of genes was not changed (black dots). More than two-thirds of the differentially expressed genes are downregulated in PU.1-knockdown HSCs (red dots); fewer than one-third are upregulated (blue dots). PU.1 is indicated by the arrow. (b) Hierarchical cluster analysis using a correlationbased centroid-linkage algorithm clearly distinguishes wild-type and PU.1-knockdown HSCs. Relative expression values are color-coded (blue: low expression, red: high expression). The dendrogram visualizes the degree of similarity between individual samples.

1270

downregulated if the change in transcript level as a multiple of the control was 41.5, P o 0.1 and lower confidence bound 41.2), indicating that PU.1 does not just activate but also represses a substantial number of genes in HSCs (Fig. 2a and Supplementary Table 1 online). Hierarchical cluster analysis and principal component analysis clearly distinguished PU.1-knockdown HSCs and wild-type HSCs, indicating that B80% decreased PU.1 expression leads to a characteristic transcriptional pattern of disturbed gene expression in stem cells in vivo (Fig. 2b and Supplementary Fig. 1 online). Known targets of PU.1 are affected in HSCs in vivo Analyzing the expression of Sfpi1 and genes in its genomic neighborhood, we found that apart from Sfpi1, expression of genes upstream or downstream of the deleted –14 kb URE was unchanged (Supplementary Fig. 2 online). These data indicate that knockout of the URE leads to a specific inhibition of PU.1 expression in HSCs.

b PU.1 kd HSC 2 PU.1 kd HSC 1 PU.1 kd HSC 3 Wild type HSC 2 Wild type HSC 1 Wild type HSC 3

RESULTS HSC expansion and transplantability of AML by HSCs We monitored the HSC compartment during the disease course of PU.1-knockdown mice by determining the number of Lin– c-kit+ Sca-1+ HSCs in the bone marrow of wild-type mice, preleukemic PU.1-knockdown mice (aged 8–12 weeks) and PU.1-knockdown mice with AML (aged 4–6 months). The percentage of HSCs in preleukemic mice did not differ from that of wild-type littermates, whereas we found a threefold expansion of HSCs at the leukemic stage (Fig. 1a,b). The total number of bone marrow cells did not significantly differ among the three groups (data not shown). In previous studies, we demonstrated that we could induce the disease by transplanting total bone marrow cells into NOD-SCID recipients7. We now asked whether HSCs could confer AML and therefore represent LSCs. When we transplanted HSCs isolated from leukemic mice into NOD-SCID recipients, we observed development of AML after 9–12 weeks, indicating that LSCs can develop from the HSC compartment in PU.1 knockdown–induced AML (Fig. 1c). These findings prompted us to examine gene expression profiles of preleukemic PU.1-knockdown HSCs in order to identify early transcriptional changes underlying the malignant transformation during the course of disease.

a mRNA expression in PU.1 KD HSC

© 2006 Nature Publishing Group http://www.nature.com/naturegenetics

0.86%

b

KD/KD leukemic

Survival

102

103

104

W T/ W T pr K el D/ eu K ke D m ic K le D uk /K em D ic

0.90%

KD/KD preleukemic

HSC (% of Lin–)

103

104

c-kit APC

WT/WT

104

c-kit APC

c-kit APC

a

104

103

PU.1 102 102

103

104

mRNA expression in WT HSC

–3.0 –1.5 0

VOLUME 38

[

NUMBER 11

[

1.5 3.0

NOVEMBER 2006 NATURE GENETICS

ARTICLES

a

4

TEK

SOCS 2

RhoB

FosB

Evi5

Egr1

TACI

PAK1

Dnmt2

BAFF-R

1

PU.1

RAB19

–1

PDGFR

HLX

0 Wnt5b

mRNA Expression [log2(PU.1 kd HSC/ wt HSC)]

2

–2 –3 –4

a

b

Mouse Rat Dog Cow Human

NATURE GENETICS VOLUME 38

[

NUMBER 11

[

NOVEMBER 2006

pu

N

In

200 bp

d

+ + + – – –

+ + – – – –

+ + – + – –

+ + – – + –

+ + – – – +

o

t

c

PU.1

pxp2 basic

pxp2 Luciferase

pxp2 JunB promoter –1400 JunB promoter

pxp2 Luciferase

pxp2 PU.1 site mutant –1400 JunB promoter

pxp2 Luciferase

PU.1

0

8 4 0

JunB 2

2

1

1

0

0

.1

1

12

PU.1

k

2

16

3

PU

3

g

oc

4

U937 stable

M

Firefly/Renilla [RLU]

5

f

.1

U937 transient

si c pr Ju om nB PU ote .1 r m sit ut e an t

Firefly/Renilla [RLU]

e

Ba

Figure 4 PU.1 directs activity of the Junb promoter. (a) Sequence alignment of the homology region in the Junb promoter containing the PU.1 binding site of the indicated species. The PU.1 core-binding motif is highlighted. (b) EMSA showing the binding of PU.1 to the PU.1 site in the Junb promoter. Nuclear extracts of U937 cells were incubated with a 32P-labeled probe containing the Junb PU.1 site, as well as PU.1 antibody and unlabeled competitor oligonucleotides as indicated. (c) Chromatin immunoprecipitation shows PU.1 binding to the Junb promoter in U937 cells. The genomic region of the PU.1 binding site was PCR amplified after reverse crosslink of the immunoprecipitates. An input control and precipitates using a PU.1 antibody, no antibody or nonspecific control IgG are shown. PCR products were verified by sequencing. (d) Disruption of the PU.1 site represses Junb promoter activity. Schematics of reporter constructs used for transfections of U937 cells. Top: the basic pxp2 luciferase vector (‘pxp2 basic’). Center: the wildtype –1400 Junb promoter in the pxp2 reporter vector. Bottom: the –1400 JunB promoter with mutated PU.1 binding site in the pxp2 reporter vector. (e) Transient transfection showed a 51% reduction of promoter activity upon mutation of the PU.1 site in U937 cells. (f) Mutation of the PU.1 site resulted in a 66% reduction in promoter activity in stably transfected U937 cells. Results of three independent experiments are shown. Error bars indicate s.d. (g) Restoration of PU.1 induces Junb expression. PU.1-knockdown cells were transfected with the PU.1-expressing pECE PU.1 construct, and expression of PU.1 and JunB was assessed after 24 h by qRT-PCR.

U937 cells JunB PU.1 site Self PU.1 cons PU.1 cons mut PU.1 Ab

PU.1 site

k

c-Jun and JunB are downregulated in PU.1-knockdown HSCs To check differential expression of genes identified by linear amplification and array analysis, we analyzed 15 genes by quantitative RT-PCR (qRT-PCR). We confirmed the results of the array analysis in all cases and found that the extent of differential expression measured by qRT-PCR was, in 14 out of 15 cases, greater than or equal to that measured by the microarrays (Fig. 3a).

PU

A number of genes have been shown to be regulated by PU.1. We checked expression of several of them to address whether they are affected in HSCs. PU.1 itself was reduced by 71%. Several putative targets of PU.1, including Fes, Btk, Tcfec, Ebi3 and genes encoding subunits of the GM-CSF receptor were significantly downregulated in PU.1-knockdown HSCs (Supplementary Fig. 2). This shows that these genes are PU.1 dependent in HSCs in vivo and that their expression is sensitive to a 70%–80% reduction of PU.1 levels.

Relative RNA expression

GAPDH 0.1

M oc

JunB

In order to understand the mechanism of leukemic transformation, we analyzed our data set, focusing on genes that have been shown to be potent regulators of stem cell functions. In both PU.1-knockdown and wild-type HSCs, we found normal expression of important stem cell–regulating genes such as Ctnnb1, Notch1, Notch2, Bmi1, Shh or classic Hox genes. In contrast, an entire set of AP-1 transcription factors, including Jun (also known as c-Jun), Junb and Fosb, were among the most strongly downregulated genes in PU.1-knockdown HSCs. We confirmed c-Jun and JunB expression by qRT-PCR from amplified copy RNA (cRNA), direct qRT-PCR and RNA slot blotting, and we found a consistent downregulation in PU.1-knockdown HSCs (Fig. 3b,c). Junb recently has been shown to act as a tumor suppressor in HSCs, as disruption of JunB expression leads to a stem cell–derived myeloproliferative disease, demonstrating its role in malignant growth control10. c-Jun is a known cofactor of PU.1 with regard to transcriptional activation of the M-CSF receptor promoter and, hence, induction of monocytic differentiation11. Furthermore, a pathway analysis (using Iobion PathwayAssist software) suggested that c-Jun and JunB are central in the network of regulated genes in PU.1-knockdown

si c pr Ju om nB o PU ter .1 m sit ut e an t

KD/KD

c-Jun

JunB

c-Jun

JunB

c-Jun

1

JunB

WT/WT

A PU b .1 C Ab on tro lI gG

c

Ba

10

c-Jun

b mRNA Expression [log(PU.1 kd / wt)]

© 2006 Nature Publishing Group http://www.nature.com/naturegenetics

3

Figure 3 Quantitative real-time RT-PCR and RNA slot blotting corroborate differential gene expression. (a) Comparison of differential gene expression in PU.1-knockdown versus wild-type HSCs as determined by different methods (black bars: Affymetrix Mouse Genome 430 2.0 Array; gray bars: quantitative real-time RT-PCR from amplified RNA; white bars: direct qRT-PCR). Downward-pointing bars indicate decreased expression; upwardpointing bars indicated increased expression in PU.1-knockdown HSCs. Expression is displayed on a log2 scale. (b) RNA expression of c-Jun and JunB was consistently lower in PU.1-knockdown HSCs as measured by array technology (black), qRT-PCR after linear amplification (gray) and direct qRT-PCR (white). (c) Lower expression of c-Jun and JunB in PU.1knockdown stem cells as assessed by RNA slot blotting of amplified RNA. GAPDH expression served as a control.

1271

ARTICLES

5′ LTR

IRES

GFP

5′ LTR

c-Jun

IRES

MSCV-IRES-GFP empty

GFP

3′ LTR

c

M

IG M em IG pt c- y Ju n

b

3′ LTR

c-Jun expression (% GAPDH)

a

80 kDa 37.9 kDa 30.5 kDa

c-Jun

6 4 2 0

MIG empty

d 140

MIG c-Jun

e

120 Colony number (%)

100 80 60

f

40 20 0

g

WT MIG c-Jun

WT MIG PU.1

KD MIG empty

KD MIG PU.1

h

1

0.5

0 100

KD MIG c-Jun

101

102

103

Mac-1 PE

104

1

Events (norm.)

WT MIG empty Events (norm.)

© 2006 Nature Publishing Group http://www.nature.com/naturegenetics

Tubulin β

MSCV-IRES-GFP c-Jun

0.5

0 100

101

102

103

104

Gr-1 PE

HSCs (Supplementary Fig. 3 online). This prompted us to focus on those two genes in order to understand their contribution to PU.1induced leukemic transformation. Junb is a direct transcriptional target gene of PU.1 To address whether Jun, Junb or Fosb are direct transcriptional targets of PU.1, we performed conserved sequence analysis of their promoter regions. Although we could not identify potential PU.1 binding sites in the promoters of Jun and Fosb, we found a perfect PU.1 binding motif in the so-called flanking evolutionary conserved sequence (FECS) II of Junb (Fig. 4a), a region about 1,400 bases upstream of the Junb transcriptional start site that had previously been shown to be important in transcriptional regulation of Junb12. Electrophoretic mobility shift assays (EMSA) and chromatin immunoprecipitation (ChIP) assays demonstrated that PU.1 binds to that site in vitro and in vivo (Fig. 4b,c). We could not detect PU.1 binding to the Jun or Fosb promoters by ChIP (data not shown). To test for functional relevance of the PU.1 site upstream of Junb, we performed reporter assays and found that disruption of that site reduced luciferase activity in transient as well as stable transfection assays in U937 cells (Fig. 4d–f). Also, when we re-expressed PU.1 in PU.1-knockdown cells, we observed an upregulation of JunB (Fig. 4g). Taken together, these data indicate that Junb is a direct transcriptional target of PU.1. c-Jun rescues the myelomonocytic differentiation block As c-Jun has previously been shown to be an activator of monocytic differentiation, we hypothesized that restoration of c-Jun levels might rescue the myelomonocytic differentiation block that is encountered in preleukemic PU.1-knockdown cells. To address this question, we used a retroviral transduction system (Fig. 5a,b). We transduced cells with either c-Jun–expressing MSCV-IRES-GFP (MIG) virus (MIG

1272

Figure 5 Restoration of c-Jun expression partially rescues myelomonocytic differentiation. (a) Diagram of the MSCV-IRES-GFP empty and MSCV-IRESGFP c-Jun retroviral constructs. (b) Increased protein expression of c-Jun in BOSC23 cells upon transfection with MIG c-Jun, as shown by protein blotting. (c) Restoration of c-Jun expression in primary bone marrow cells after transduction with MIG c-Jun retrovirus. Error bars indicate 1 s.d. (n ¼ 3). (d) Restoration of c-Jun expression rescues monocytic colony formation. Primary bone marrow cells of either wild-type or PU.1-knockdown mice were transduced with empty MIG (negative control), MIG PU.1 (positive control) or MIG c-Jun virus. GFP+ cells were sorted after 3 d, incubated in suspension culture in the presence of IL-3, IL-6, SCF and MCSF and then subjected to methylcellulose clonogenic assays in only MCSF–containing medium. Numbers of monocytic colonies (CFU-M) were assessed after 14 d. Average values of four independent experiments are shown. Error bars represent one s.d. (e) Colonies were GFP+ after 14 d. (f) Cells showed monocytic morphology after 14 d of culture in methylcellulose. (g) Flow cytometric analysis of Mac-1 expression of cells grown in suspension culture. Only a small fraction of PU.1-knockdown cells treated with MIG empty virus show expression of the monocytic marker Mac1 (red shading). Upon treatment with MIG c-Jun virus, the percentage of Mac-1+ monocytes increases (white shading). (h) Flow cytometric analysis of Gr-1 expression of PU.1-knockdown cells after treatment with MIG empty virus (red shading) or MIG c-Jun virus (white shading). PE, phycoerytherin.

c-Jun), empty MIG virus as a negative control or PU.1-expressing MIG virus (MIG PU.1) as a positive control. We subjected green fluorescent protein (GFP)-positive transduced cells to methylcellulose clonogenic assays supplemented with macrophage colony-stimulating factor (M-CSF) and suspension culture with interleukin (IL)-3, IL-6, stem cell factor (SCF) and M-CSF. PU.1-knockdown bone marrow cells gave rise to only 20% of monocytic colonies (CFU-M) when compared with wild-type controls. The empty virus had no effect on colony formation, whereas Jun-transduced cells showed an increase of monocytic colonies, reaching 70% of the colony-forming potential of wild-type cells or MIG PU.1–treated cells (Fig. 5c–f). Cells grown in suspension culture were analyzed by flow cytometry after 5 d. Whereas cells transduced with empty virus only showed a small fraction of cells positive for Gr-1 (10%) and Mac-1 (7%), this fraction markedly increased after treatment with MIG c-Jun (31% and 34%, respectively) (Fig. 5g,h). We obtained very similar results when we used granulocyte macrophage (GM)-CSF instead of M-CSF stimulation after retroviral treatment of PU.1-knockdown cells (Supplementary Fig. 4 online). This shows that the rescuing effect of c-Jun is not restricted to the M-CSF receptor pathway but apparently affects multiple differentiation pathways. Taken together, these results show that restoration of c-Jun expression rescues the myelomonocytic differentiation block in preleukemic PU.1-knockdown bone marrow cells, suggesting that c-Jun is a critical downstream target in PU.1-knockdown HSCs. JunB inhibits leukemogenic properties of PU.1-knockdown cells We next addressed whether diminished expression of c-Jun or JunB is relevant to the malignant properties of cells in the leukemic stage of PU.1-knockdown mice. To target leukemic cells efficiently, we used c-Jun– and JunB-expressing lentiviruses that have been described previously10 (Fig. 6a). Expression of c-Jun did not have an effect on growth in suspension culture (data not shown), but JunB led to a substantial inhibition of malignant proliferation (Fig. 6b,c). When we subjected transduced GFP+ leukemic cells to clonogenic assays, we observed an 80% inhibition of colony formation after lentiviral restoration of JunB expression but did not observe any significant effect after the use of c-Jun–expressing virus or empty control virus (Fig. 6d). Treatment of normal bone marrow cells did not have a significant effect on colony-forming capacity, demonstrating a specific

VOLUME 38

[

NUMBER 11

[

NOVEMBER 2006 NATURE GENETICS

ARTICLES

LV empty

hPGK

c-Jun IRES

GFP

LV c-Jun

hPGK

JunB

GFP

LV JunB

IRES

300 200 100 0

LV empty LV c-Jun LV JunB

4.0 3.0 2.0 1.0 0

350

0

1

2

3

4

LV empty LV c-Jun LV JunB

300 250 200 150 100 50 0

5

WT

f

LV empty, GFP 105

250 200

ym = 18,900

BrdU

100 50

10

103 102

0 10

2

10

3

4

10

LV JunB, GFP 105

+

1 plate

g

ym = 3,600

10

9.8 %

4

103 102

72.4 7.5

84.2 3.0

WT Leukemic PU.1 KD

2

nd

plate

LV JunB

100 13.6 %

4

150

+

Leukemic PU.1 KD

st

Survival (%)

e Counts

© 2006 Nature Publishing Group http://www.nature.com/naturegenetics

Time (d)

60 40

LV c-Jun

20

10

50 100 150 200 250

Annexin V - PE

50 100 150 200 250

LV empty

7-AAD

7-AAD

0

50 100 Time (d)

WT Leukemic PU.1 KD

rd

4th plate

3 plate

h

LV PU.1

80

0

5

WT Leukemic PU.1 KD

Leukemic NOD-SCID recipient

WT KD BM

IRES

Cell number (× 106)

GFP

hPGK

Gene expression (% Gapd)

400

d

Control

c

Li ve Sp r le en

b

Number of GFP+ colonies

a

150

Figure 6 Restoration of JunB expression in leukemic cells of PU.1-knockdown mice inhibits leukemogenesis. (a) Schematic diagram of lentiviral (LV) constructs. (b) Expression of c-Jun (filled bars) and JunB (open bars) in sorted GFP+ cells. Error bars indicate 1 s.d. (n ¼ 3). (c) Suspension culture after infection with JunB-GFP or empty GFP lentivirus. Number of JunB+ cells (open circles) and empty LV+ cells (filled squares) are shown ± s.d. (n ¼ 3). (d) Serial replating of lentivirally transduced and sorted GFP+ cells of leukemic PU.1-knockdown mice. LV empty, white bars; LV c-Jun, gray bars; LV JunB, black bars. Bone marrow cells of wild-type littermates served as control. Error bars indicate s.d. (n ¼ 4). (e) Apoptosis in GFP– (light gray) and GFP+ (dark gray) cells after infection with LV JunB. (f) Cell cycle analysis of cells after infection (LV empty (left); LV JunB (right)). Percentages of cells in G0/G1 (lower left gates), S (upper gates) and G2+M (lower right gates) phase of cell cycle are indicated. (g) Rescue of PU.1 knockdown–induced leukemia by restoration of JunB expression. Leukemic PU.1-knockdown cells were infected with LV empty (n ¼ 14), LV c-Jun (n ¼ 6), LV JunB (n ¼ 6) or LV PU.1 (n ¼ 8) lentiviruses, and transplanted into NOD-SCID recipients. Mice transplanted with leukemia cells treated with LV empty (dotted line) or LV c-Jun (solid line) died of leukemia after 2–4 months. After transplantation of leukemic cells infected with LV JunB (dashed line), NOD-SCID recipients did not develop leukemia. (h) Diagnostic DNA blot of EcoRV/SspI-digested genomic DNA. A probe recognizing a 9.7-kb wild-type fragment and a 6.8-kb fragment indicating the knockdown allele was used as previously described7.

effect on leukemic PU.1-knockdown cells. Furthermore, we examined in vitro LSC self-renewal properties by performing serial replating assays13,14. Leukemic cells treated with either control virus or c-Jun– expressing virus showed a marked increase of colony-forming potential with increasing number of replatings, whereas colony formation derived from normal bone marrow cells decreased in the same time period. Notably, lentiviral expression of JunB in leukemic PU.1knockdown cells led to a constant decrease of colony formation, indicating suppression of the LSC phenotype (Fig. 6d). Flow cytometric analysis of the leukemic colony-forming cells showed that the vast majority still coexpressed Gr-1 and Mac-1 after three replatings, whereas only a small subset expressed CD61 or Ter-119. Neither this immunophenotype nor the immature morphology of the malignant colony-forming cells changed in response to either c-Jun or JunB expression (data not shown). Apparently, in leukemic PU.1-knockdown cells, neither c-Jun nor JunB has an effect on differentiation; JunB acts as an inhibitor of LSC self-renewal in vitro. To elucidate the mechanism of action of JunB, we determined the apoptosis rate after transduction. JunB expression led to an increase of apoptotic cells from 12% to 34% (Fig. 6e). Cell cycle analysis showed that JunB expression led to reduced cycling of cells in S phase, a reduced number of cells in G2 and M phase of the cell cycle and an increase of cells in G0/G1 phase (Fig. 6f). To test the ability of JunB to disrupt LSC properties of PU.1knockdown cells in vivo, we transplanted lentivirally transduced, GFPsorted leukemic bone marrow cells of PU.1-knockdown mice into NOD-SCID recipients. The percentage of GFP+ cells in the peripheral blood ranged between 7% and 15% 4 weeks post-transplant (Supplementary Fig. 5 online). After transplantation of PU.1-knockdown cells infected with the empty control virus, NOD-SCID

NATURE GENETICS VOLUME 38

[

NUMBER 11

[

NOVEMBER 2006

recipient mice developed donor-derived leukemia after 2–4 months (Fig. 6g,h). Leukemic mice showed hepatosplenomegaly, spleen cells showed typical blast morphology and blasts were GFP+, Gr-1+ and Mac-1+ (data not shown). Infection of PU.1-knockdown cells with c-Jun-GFP–expressing lentivirus before transplantation did not prevent leukemia in NOD-SCID recipients (Fig. 6g). Notably, leukemic PU.1-knockdown cells that were infected with JunB-GFP–expressing lentivirus before transplantation did not cause leukemia, demonstrating that restoration of JunB expression disrupts leukemic potential of PU.1-knockdown cells in vivo. This suggests that JunB downregulation by PU.1 is critical to confer long-term self-renewal potential to transformed cells. JunB and PU.1 expression in individuals with AML To address the potential relevance of our findings for human disease, we analyzed gene expression data of leukemic bone marrow cells of 285 individuals with AML. There was no correlation of PU.1 and cJun (r ¼ 0.007, data not shown), but we found a correlation of PU.1 and JunB expression (r ¼ 0.52) that was more prominent in some subgroups of affected individuals (Fig. 7a and Supplementary Fig. 6 online). Individuals with French-American-British (FAB) subtypes M4 and M5 showed PU.1-JunB correlations of 0.62 and 0.56, respectively. Leukemic bone marrow cells of individuals with AML represent heterogeneous cellular populations, and thus the interpretation of expression data between affected individuals and in comparison with normal bone marrow is difficult. Furthermore, PU.1 is expressed at highest levels in granulocytes, and thus, presence of any mature cells in bone marrow samples will result in falsely higher levels of PU.1. Therefore, we sorted Lin– CD34+ CD38– Thy1low HSCs of 20 individuals with AML (AML-HSC) and five healthy controls and

1273

ARTICLES Genes of various functional groups are differentially expressed in PU.1-knockdown 70 70 HSCs (Supplementary Table 1). Notably, 60 r = .72 3,000 60 r = .52 many transcription factors, including Jun, 50 50 Junb, Fosb, Egr1, Lmo1, Klf2, Klf4, Pbx1, 40 2,000 40 Zfp36 and Zfp185, show diminished expres30 30 20 1,000 20 sion upon PU.1 knockdown, indicating that 10 10 PU.1 acts as a key regulator of a complex 0 0 0 transcriptional program in HSCs. We have 0 50 100 150 0 200 400 600 800 1,000 Normal AML PU.1 expression PU.1 expression HSC HSC also found upregulation of a smaller number of transcription factors, including Maf, Figure 7 PU.1 and JunB expression in individuals with AML. (a) Correlation of PU.1 and JunB Hlx1, Mta3, Zfp251 and Tcf19, demonstrating expression in leukemic bone marrow of 285 individuals with AML. The scatter plot shows PU.1 and that PU.1 might act as a repressor of JunB mRNA expression in bone marrow cells of individuals with AML. Expression values are expressed distinct transcription factors in HSCs. in arbitrary units. A linear regression curve and the correlation coefficient are indicated. (b) Correlation However, whether those are primary targets of PU.1 and JunB expression in human HSCs. After sorting of Lin– CD34+ CD38– Thy1low HSCs of 20 individuals with AML (filled circles) and normal controls (open circles), PU.1 and JunB mRNA were of PU.1 has yet to be determined. Besides measured by quantitative real-time RT-PCR. A linear regression curve and the correlation coefficient are transcription factors, several protein kinases, indicated. (c) JunB mRNA expression is lower in HSCs (Lin– CD34+ CD38– Thy1low) from individuals including Tek, Kdr, Matk, Adrbk2 and Pak1, with AML than in normal HSCs. Medians are indicated by black lines. are downregulated in PU.1-knockdown stem cells, reflecting disturbed intracellular signaling pathways. Several other potentially measured PU.1, JunB and c-Jun expression in these purified and well- interesting genes are downregulated as well, including Dnmt2 and defined cellular populations. Sorted T cells and granulocytes of normal Socs2. However, their functional importance in HSCs has yet to donors served as controls (Supplementary Fig. 6). We did not see any be determined. Notably, several AP-1 transcription factors are downregulated in correlation of PU.1 and c-Jun in AML-HSCs (r ¼ 0.04), nor did c-Jun levels differ in comparison with normal HSCs (data not shown). PU.1-knockdown HSCs. c-Jun seemed of particular interest, as it can However, we found a higher correlation of PU.1 and JunB expression act as a coactivator of PU.1 with regard to the M-CSF receptor in AML-HSCs (r ¼ 0.72) (Fig. 7b). Average PU.1 expression in AML- promoter11. Furthermore, activation of the AP-1 pathway has a role HSCs was not significantly different from normal HSCs (P ¼ 0.54); in differentiation of immature myeloid cells18. Therefore, we hypothehowever, 9 out of 20 individuals with AML showed lower PU.1 sized that reduced c-Jun and PU.1 combined might contribute to the expression in HSCs compared with expression in normal HSCs. myelomonocytic differentiation block observed in preleukemic PU.1Average expression of JunB was 78% lower in AML-HSCs than in knockdown mice. In fact, retroviral restoration of c-Jun expression normal HSCs (P ¼ 0.00015) (Fig. 7c). All nine affected individuals restores monocytic differentiation, showing that the disturbed c-Jun with low PU.1 had also low JunB, and 17 out of 20 individuals pathway contributes to the differentiation block encountered in with AML showed reduced Junb expression compared with normal preleukemic PU.1-knockdown mice. Forced retroviral expression of HSCs. These data suggest that the correlation between PU.1 and c-Jun in normal bone marrow does not lead to an increase of JunB downregulation that we identified in PU.1-knockdown mouse monocytic colony formation, ruling out a nonspecific effect. HSCs has a role in human AML and that diminished JunB expression Recently, several genes have been identified as important stem could contribute to malignant self-renewal in LSCs of individuals cell regulators, including b-catenin, Bmi1, Shh, Notch and HOX with AML. genes19–23. Although expression of these genes was not altered in PU.1-knockdown HSCs, we detected downregulation of Junb. Junb has DISCUSSION recently been identified as a potent tumor suppressor in normal HSCs, Lineage-specific transcription factors are critical for hematopoiesis, as its knockdown leads to a myeloproliferative disease that originates HSC function and leukemic transformation8,9. Here, we demonstrate from the stem cell compartment10. Hence, we asked if reduced JunB that knockdown of the myeloid master regulator PU.1 leads to expression levels are important for the malignant properties of cells in transcriptional signatures in stem cells that precede leukemic trans- the leukemic stage of PU.1-knockdown mice. Leukemic cells from the formation. Furthermore, we show that transcriptional profiling of bone marrow of PU.1-knockdown mice formed a large number of highly enriched HSCs at a preleukemic stage can lead to the identi- malignant colonies in clonogenic assays, and this colony-forming fication of pathways that are functionally essential for the pool of potential increased in serial replating assays, indicating the successive leukemic stem cells that arise later during the course of disease. selection of a pool resembling LSCs with increased self-renewal Knockout of the URE of the Sfpi1 gene leads to knockdown of PU.1 capacity. Restoration of JunB expression led to a marked decrease of expression in HSCs but does not affect other genes in the genomic colony-forming potential of the leukemic cells and loss of serial neighborhood, indicating the specificity of the enhancer function in replating capability, indicative of a disruption of LSC-like properties. the HSC compartment. The majority of genes differentially expressed We found that JunB acts by inducing both increased apoptosis and in PU.1-knockdown HSCs were downregulated, thereby demonstrat- decreased cycling of malignant PU.1-knockdown cells. This is in line ing the general role of PU.1 as a transcriptional activator in HSCs. with previous findings of diminished p16-INK4a and increased Bcl2 Several studies have identified transcriptional targets of PU.1 and Bclx expression in Junb-deficient mice10. Lentiviral expression of (reviewed in ref. 15). We have found decreased expression of many JunB in leukemic PU.1-knockdown cells also prevented the developof them in PU.1-knockdown HSCs, demonstrating that these targets ment of leukemia after transplantation. Apparently, beyond its funcare affected in HSCs in vivo. Notably, our data suggest that PU.1 not tion as a tumor suppressor in normal HSCs, JunB acts as a suppressor only activates but also represses a number of genes16,17. of LSC function in PU.1 knockdown–induced leukemia.

© 2006 Nature Publishing Group http://www.nature.com/naturegenetics

1274

HSC (Lin–, CD34+, CD38–, Thy1low )

80

c 80 JunB expression

b

JunB expression

JunB expression

a 4,000

VOLUME 38

[

NUMBER 11

[

NOVEMBER 2006 NATURE GENETICS

© 2006 Nature Publishing Group http://www.nature.com/naturegenetics

ARTICLES Diminished expression of c-Jun and JunB after PU.1 knockdown has two major effects. Reduced c-Jun levels contribute to a block of myelomonocytic differentiation before the onset of leukemia, and decreased expression of the stem cell suppressor JunB confers increased self-renewal capacity. Both are essential for the formation of a pool of LSCs and the development of AML in PU.1-knockdown mice. c-Jun did not have a differentiating effect at the leukemic stage, suggesting that it might be more relevant in the preleukemic phase of the disease. However, JunB has a strong inhibitory effect on leukemic PU.1-knockdown cells, indicating its importance in the leukemic situation. Although reduction of JunB alone causes a myeloproliferative phenotype, in the PU.1-knockdown situation, the combination with a differentiation block seems to lead to an acute leukemic phenotype. When we analyzed gene expression in leukemic bone marrow cells and HSCs from individuals with AML, we found a correlation of PU.1 and JunB expression in total bone marrow and an even stronger one in purified HSCs, suggesting that a similar mechanism might have a role in human disease. Notably, PU.1 and JunB do not correlate to the same extent in normal HSCs, which might indicate that cofactors are required for PU.1 to exert its Junb-regulating function and that such cofactors are not present under normal conditions. PU.1 expression in HSCs was decreased in some individuals with AML, and all of these individuals showed also low JunB in HSCs. Further, we found that JunB expression is lower in AML-HSCs when compared with normal control HSCs, suggesting that diminished stem cell suppression by JunB could contribute to the pathogenesis of human AML. In contrast to JunB, we found neither downregulation of c-Jun nor correlation of c-Jun with PU.1 in individuals with AML, which might be explained by our finding that Junb is a direct transcriptional target of PU.1 in myeloid leukemic cells, but Jun is not. Although recent studies have shown that gene expression profiling of leukemic blast cells in human AML is of diagnostic importance24,25, such studies are difficult to interpret with regard to affected etiologic pathways because of the lack of an adequate cellular control population. This makes it challenging to decide if a gene might be overexpressed or underexpressed at the respective developmental stage of the leukemic bulk population. Furthermore, it is difficult to examine genes such as PU.1 or other developmental regulators whose expression changes during differentiation because of lineage and maturation differences between different AML subtypes. Finally, it is hard to accurately measure levels of PU.1 or other genes that are expressed at highest levels in mature cells because of the presence of such cells in total bone marrow samples26,27. Our study shows that HSC profiling provides a method to circumvent this challenge and complement the analysis of blasts, as it analyzes developmentally well-defined cells and thereby reduces experimental bias caused by variations of the cellular composition of samples. This notion is supported by our finding of a stronger PU.1-JunB correlation in HSCs than in unfractionated bone marrow samples of individuals with AML. In addition to these experimental considerations, gene expression profiling of HSCs might at the same time offer other advantages. Recent experimental evidence suggests the existence of LSCs as the cells of origin of various hematological malignancies28,29. This model assumes that after initial events in the stem cell compartment, additional transforming events occur either in HSCs or in progenitors during differentiation that ultimately lead to formation of LSCs, giving rise to a hierarchy of leukemic bulk cells30. As a consequence, treatment must be directed against those LSCs if one aims to cure the disease31,32. A prerequisite of this approach is the identification of pathways essential for LSC function. Our study shows that

NATURE GENETICS VOLUME 38

[

NUMBER 11

[

NOVEMBER 2006

phenotypic HSCs can carry early alterations in gene expression that are essential for transformation and maintenance of LSCs during the course of disease and that those alterations can be detected by stem cell gene expression profiling and have the potential to be exploited therapeutically. Hence, genes that are deregulated in HSCs of individuals with AML might be suitable targets for LSC-directed therapeutic approaches. Our results demonstrate that gene expression profiling of HSCs is useful for the analysis of transcription factor–defective models, as it directly measures transcriptional effects as a function of transcription factor levels. In the future, genome-wide transcriptional analysis of HSCs deficient for other transcription factors might ultimately lead to the delineation of the entire transcriptional network effective in hematopoietic stem cells, which would help to understand their function in a physiological setting as well as in disease. METHODS Cells. PU.1-knockdown mice with targeted disruption of the distal enhancer (upstream regulatory element) –14 kb upstream of the PU.1 gene have been previously described7. Mouse experiments and analysis of mouse cells were approved by the Beth Israel Deaconess Medical Center Institutional Animal Care and Use Committee (protocol # 115-2004). We analyzed human cells after obtaining written informed consent, and the procedures were approved by the Institutional Review Board of Beth Israel Deaconess Medical Center (protocol # 2001P-000567/E-85-0001-FB). Flow cytometry and sorting of HSCs. Mouse bone marrow and spleen cells were analyzed on a FACScan cytometer (Becton Dickinson) after gating on viable cells. After we lysed erythrocytes, we performed lineage depletion of bone marrow cells using antibodies directed against CD3, CD4, CD8a, CD19, Ly-6G, Ter119 and CD45R antigens. Flow cytometric analysis and sorting of Lin– c-kit+ Sca-1+ hematopoietic stem cells using a double laser (488 nm/350 nm Enterprise II +647 Spectrum) high-speed cell sorter (MoFlo-MLS, Cytomation) were performed as described previously33. Assessment of the differentiation state of transduced cells was performed with phycoerythrin-conjugated antibodies to Gr-1 and Mac-1 (BD Pharmingen). Human HSCs were isolated from bone marrow as reported previously34,35. In brief, bone marrow mononuclear cells were enriched for CD34+ cells using immunomagnetic beads as described36. CD34+ cells were then stained with Tricolor-conjugated antibodies against lineage antigens CD34-APC, Thy1-FITC and CD38-APC-Cy7. Viable Lin– CD34+ CD38– Thy1low cells were sorted by a MoFlo-MLS cell sorter. Linear amplification of RNA from HSCs and array expression analysis. RNA of 5,000 HSCs of PU.1-knockdown and wild-type animals (three pools of three animals each) were isolated using 20 ng of bacterial ribosomal carrier RNA (Roche Diagnostics) and denaturing buffer RLT containing guanidine isothiocyanate (Qiagen) without b-mercaptoethanol. We extracted RNA according to the RNeasy Micro method (Qiagen) optimized for small amounts of RNA. For linear amplification of RNA, we applied a strategy of two rounds of reverse transcription followed by T7 promoter–dependent in vitro transcription. We used a modified version of protocols previously described37,38. In brief, we used T7 promoter oligodT primers and SuperScript III reverse transcriptase for reverse transcription. After second-strand cDNA synthesis using Escherichia coli DNA polymerase and ligase, we performed in vitro transcription using T7 polymerase (Ambion) according to the manufacturer’s instructions. The first round of amplification was finished by a cRNA cleanup using a silica gel–based membrane method (RNeasy Micro, Qiagen). After a second round of amplification, we hybridized 10 mg of the resultant biotinylated cRNA to Affymetrix Mouse Genome 430 2.0 arrays covering about 45,000 transcripts. We hybridized, washed, stained and scanned the Affymetrix arrays according to standard protocols (Affymetrix). Array data and statistical analysis. Array data were analyzed with the dChip software39. After smoothing spline normalization, expression values were calculated applying the perfect match–mismatch difference model algorithm (dChip)39. Array data are available at http://www.ncbi.nlm.nih.gov/geo

1275

© 2006 Nature Publishing Group http://www.nature.com/naturegenetics

ARTICLES (see below). When we compared PU.1-knockdown HSCs with wild-type HSCs, we considered a change in expression significantly different if the change in transcript level as a multiple of the control was 41.5, with P o 0.1 (Student’s t test) and a lower confidence bound (LCB) 41.2. LCB is a stringent estimate of the change in expression as a multiple of the control and has been shown to be a superior ranking statistic40. The use of LCB as a cutoff provides a 90% confidence that the actual change in expression as a multiple of the control is some value above the reported LCB. Studies exploring the accuracy and calibration of Affymetrix chips using custom arrays and quantitative reverse transcriptase real-time PCR assays have further shown that chip analyses generally underestimate differences in gene expression and that an LCB of 41.2 corresponds to genes with an at least a 300% change in gene expression41. Hierarchical cluster analysis was performed using the correlation-based centroid-linkage algorithm (dChip)39. As a dimension-reduction method, we used principal components analysis (PCA) to explain the variation in the original data set with fewer variables, as previously described42. We represented the multidimensional gene expression data with a two-dimensional visualization using dChip. This plot visualizes the closeness between the samples as they are projected along the two main axes (Supplementary Fig. 1). We used PathwayAssist 5.2 software (Iobion) to analyze pathways affected in PU.1-knockdown HSCs compared with wild-type HSCs. We performed ‘direct interaction’ analysis, including all differentially expressed genes as previously described43. Quantitative real-time RT-PCR. We extracted total RNA from FACS-sorted hematopoietic stem cells using RLT buffer and 20 ng bacterial carrier RNA (Roche Diagnostics) per sample utilizing the RNeasy Micro protocol (Qiagen). RNA was treated with DNase I according to the manufacturer’s instructions. For some experiments, we used amplified RNA that had undergone one cycle of reverse transcription and T7 promoter–based in vitro transcription. After reverse transcription, we amplified the resultant cDNA using an AbiPrism 7700 Sequence Detector (Applied Biosystems) with 40 cycles of 95 1C (15 s), 60 1C (1 min) and 72 1C (1 min), using SYBR green for detection (for primer sequences, see Supplementary Table 2 online). PCR products were checked by melting curve analysis and sequencing. For detection and quantification of human PU.1, JunB and c-Jun, we used commercially available TaqMan assays (Applied Biosystems), including Gapd as a control. Stem cell cRNA slot blotting. We extracted total RNA from FACS-sorted HSCs and performed two rounds of reverse transcription and T7 promoter–based in vitro transcription as described. We blotted 1 mg of cRNA and detected c-Jun RNA with 32P-labeled mouse c-Jun cDNA, JunB RNA with mouse JunB cDNA and Gapd RNA with a 1.3-kb PstI fragment of the rat cDNA. Slot blots were quantified using ImageQuant densitometry software (Amersham). Protein blot assays. We extracted total cell lysates as described44. Proteins were resolved by SDS-PAGE and electrotransferred to a nitrocellulose membrane (Bio-Rad Laboratories). We used polyclonal rabbit antibody to c-Jun (Santa Cruz) and monoclonal mouse antibody to tubulin b (Sigma). We detected immunoreactive proteins using HPRT-conjugated antibodies to mouse or rabbit (Santa Cruz) and the ECL system (Amersham). Bands were quantified using ImageQuant densitometry software (Amersham). Electrophoretic mobility shift assays. Electrophoretic mobility shift assays (EMSA) were performed as previously described6. We annealed 23-mer oligonucleotides containing the PU.1 site, labeled them with [g32P]ATP by the use of T4 polynucleotide kinase and gel-purified them using 10% polyacrylamide gels. Probes were incubated with nuclear extracts of U937 cells in 10 mM HEPES (pH 7.8), 50 mM KCl, 1 mM dithiothreitol, 1 mM EDTA and 5% glycerol (vol/vol) for 30 min. A PU.1 consensus oligonucleotide (ActiveMotif) and a mutated consensus oligonucleotide (ActiveMotif) were used as competitors as indicated. Reaction mixtures were separated with 6% polyacrylamide gels in 0.5
TBE buffer at 4 1C. PU.1 antibody (Santa Cruz) was used for supershift. Chromatin immunoprecipitation. Chromatin immunoprecipitation (ChIP) experiments were performed as previously described6. Chromatin was isolated from U937 cells and sonicated three times for 10 s with a 90% duty cycle and

1276

output setting 4 on a Branson Sonifier 450 apparatus. Immunoprecipitation was performed with 10 mg of PU.1 antibody (Santa Cruz) or 10 mg of normal rabbit immunoglobulin G (IgG) (Santa Cruz). Promoter activity assays. The –1.4 kb Junb promoter, including the PU.1 binding site, was cloned into the pxp2 luciferase vector. Constructs with a disrupted PU.1 binding site (CTGGGAAGAGGAACCT to CTGCTACGA TCTGCCT) were manufactured by PCR mutagenesis as previously described45. For transient assays, U937 cells were cotransfected with pxp2 vectors and Renilla luciferase in a CMV vector for standardization. Stable transformants were obtained by cotransfection with a construct expressing a puromycin resistance gene and selection by 2 mg/ml of puromycin for 4 weeks. Luciferase assays were performed using the Dual Luciferase Reporter Assay System (Promega) according to the manufacturer’s instructions. Retroviral and lentiviral rescue experiments. We created a c-Jun–expressing MIG virus by inserting Jun in the multiple cloning site of the MIGR1 virus46. We then treated bone marrow cells with MIG, MIG PU.1 and MIG c-Jun retroviruses as previously described47. We cultured bone marrow cells in Roswell Park Memorial Institute (RPMI) 1640 medium containing FCS, IL-3, IL-6, and SCF for 2–3 d and infected them by adding cell-free retroviral supernatants for 24 h in the presence of 8 mg ml–1 polybrene. After three washing steps in complete medium, and 48 h post-transduction, we sorted GFP+ cells using a high-speed cell sorter (MoFlo-MLS, Cytomation) and resuspended the cells in cytokine-containing RPMI for 3 d before plating them in methylcellulose. Transduction efficiencies ranged from 15%–24%. For the transduction of leukemic cells, we used c-Jun– and JunB-expressing lentiviruses as previously described10. We grew leukemic bone marrow cells in Myelocult M5300 (Stem Cell Technologies) containing 10% FCS and 5% WEHI supernatant for 3 d and transduced them by addition of concentrated lentiviral supernatants at an multiplicity of infection of 10 for 48 h. Afterwards, transduced GFP+ cells were sorted by a MoFlo high-speed sorter. Apoptosis and cell-cycle assays. Apoptotic cells were identified by use of Annexin V staining as previously described48. Cell cycle analysis was performed by means of bromodeoxyuridine (BrdU) assays as described49. In brief, cells were labeled with BrdU (10 mM) for 2 h and then permeabilized and stained with fluorochrome-conjugated antibodies to BrdU and 7-amino-actinomycin (7-AAD) followed by flow cytometric analysis. Colony-forming assays and serial replating assays. Clonogenic assays to assess myelomonocytic colony formation were carried out in Methocult M3231 (Stem Cell Technologies) supplemented with M-CSF or GM-CSF. GFP+ colonies were scored after 12–14 d by the use of an AxioVert200M/Apotome fluorescence microscope (Zeiss). To test the clonogenic capacity of leukemic cells, we used Methocult M3434 medium (Stem Cell Technologies). We assessed GFP+ colony formation per 5,000 plated cells after 8–10 d. We performed serial replating assays as previously described13,14. In brief, we washed the cells with PBS after each scoring, resorted GFP+ cells, replated them onto a new plate and scored them again after 8–10 d. Transplantation of NOD-SCID mice. We transplanted 1,000 viable, propidium iodide–negative, FACS-sorted HSCs into NOD-SCID mice by tail vein injection as previously described7. For viral rescue experiments, bone marrow cells of leukemic PU.1-knockdown mice were infected with GFP lentiviruses. We sorted viable GFP+ cells by a high-speed cell sorter (MoFlo-MLS) 2 d after infection and subsequently transplanted them into NOD-SCID recipients. Mice that developed leukemia were killed and spleen and liver weight determined; leukemic bone marrow and spleen cells were subjected to cytologic and immunocytologic analysis, and DNA blotting was performed to demonstrate that leukemic cells were derived from the donor PU.1 enhancer knockout mice. Array expression analysis of individuals with AML. To investigate expression levels of PU.1 and JunB, and c-Jun in humans, we used the MADEx application to analyze gene expression profiles of bone marrow samples of 285 individuals with de novo AML. The MADEx tool has been described previously50. Expression profiles were generated using the Affymetrix HG-U133a platform as recently described24.

VOLUME 38

[

NUMBER 11

[

NOVEMBER 2006 NATURE GENETICS

ARTICLES Accession codes. GEO: array data, GSE5654 (http://www.ncbi.nlm.nih.gov/ geo/query/acc.cgi?acc¼gse5654).

© 2006 Nature Publishing Group http://www.nature.com/naturegenetics

Note: Supplementary information is available on the Nature Genetics website. ACKNOWLEDGMENTS We thank K. Martens for excellent assistance with mouse husbandry, M. Joseph for help with array hybridization, T. Dajaram and C. Hetherington for quantitative real-time RT-PCR analysis and J. Tigges and V. Toxavidis for expert assistance with multicolor flow cytometry and high-speed cell sorting. U.S. thanks S. Steidl for invaluable support and advice. U.S. also thanks R. Kronenwett and R. Haas for long-term support and mentorship. This work was supported by US National Institutes of Health grant CA41456 to D.G.T. and by fellowships of the Dr. Mildred Scheel Foundation for Cancer Research to U.S. (D/03/41221) and the Lymphoma Research Foundation to F.R. COMPETING INTERESTS STATEMENT The authors declare that they have no competing financial interests. Published online at http://www.nature.com/naturegenetics Reprints and permissions information is available online at http://npg.nature.com/ reprintsandpermissions/

1. McKercher, S.R. et al. Targeted disruption of the PU.1 gene results in multiple hematopoietic abnormalities. EMBO J. 15, 5647–5658 (1996). 2. Mueller, B.U. et al. Heterozygous PU.1 mutations are associated with acute myeloid leukemia. Blood 100, 998–1007 (2002). 3. Vangala, R.K. et al. The myeloid master regulator transcription factor PU.1 is inactivated by AML1-ETO in t(8;21) myeloid leukemia. Blood 101, 270–277 (2003). 4. Gilliland, D.G. & Tallman, M.S. Focus on acute leukemias. Cancer Cell 1, 417–420 (2002). 5. Li, Y. et al. Regulation of the PU.1 gene by distal elements. Blood 98, 2958–2965 (2001). 6. Okuno, Y. et al. Potential autoregulation of transcription factor PU.1 by an upstream regulatory element. Mol. Cell. Biol. 25, 2832–2845 (2005). 7. Rosenbauer, F. et al. Acute myeloid leukemia induced by graded reduction of a lineagespecific transcription factor, PU.1. Nat. Genet. 36, 624–630 (2004). 8. Rosenbauer, F., Koschmieder, S., Steidl, U. & Tenen, D.G. Effect of transcription-factor concentrations on leukemic stem cells. Blood 106, 1519–1524 (2005). 9. Tenen, D.G. Disruption of differentiation in human cancer: AML shows the way. Nat. Rev. Cancer 3, 89–101 (2003). 10. Passegue, E., Wagner, E.F. & Weissman, I.L. JunB deficiency leads to a myeloproliferative disorder arising from hematopoietic stem cells. Cell 119, 431–443 (2004). 11. Behre, G. et al. c-Jun is a JNK-independent coactivator of the PU.1 transcription factor. J. Biol. Chem. 274, 4939–4946 (1999). 12. Phinney, D.G., Tseng, S.W., Hall, B. & Ryder, K. Chromosomal integration dependent induction of junB by growth factors requires multiple flanking evolutionarily conserved sequences. Oncogene 13, 1875–1883 (1996). 13. Higuchi, M. et al. Expression of a conditional AML1-ETO oncogene bypasses embryonic lethality and establishes a murine model of human t(8;21) acute myeloid leukemia. Cancer Cell 1, 63–74 (2002). 14. Huntly, B.J. et al. MOZ-TIF2, but not BCR-ABL, confers properties of leukemic stem cells to committed murine hematopoietic progenitors. Cancer Cell 6, 587–596 (2004). 15. Tenen, D.G., Hromas, R., Licht, J.D. & Zhang, D.E. Transcription factors, normal myeloid development, and leukemia. Blood 90, 489–519 (1997). 16. Borras, F.E., Lloberas, J., Maki, R.A. & Celada, A. Repression of I-A beta gene expression by the transcription factor PU.1. J. Biol. Chem. 270, 24385–24391 (1995). 17. Bellon, T., Perrotti, D. & Calabretta, B. Granulocytic differentiation of normal hematopoietic precursor cells induced by transcription factor PU.1 correlates with negative regulation of the c-myb promoter. Blood 90, 1828–1839 (1997). 18. Wang, Q., Salman, H., Danilenko, M. & Studzinski, G.P. Cooperation between antioxidants and 1,25-dihydroxyvitamin D3 in induction of leukemia HL60 cell differentiation through the JNK/AP-1/Egr-1 pathway. J. Cell. Physiol. 204, 964–974 (2005). 19. Bhardwaj, G. et al. Sonic hedgehog induces the proliferation of primitive human hematopoietic cells via BMP regulation. Nat. Immunol. 2, 172–180 (2001). 20. Antonchuk, J., Sauvageau, G. & Humphries, R.K. HOXB4-induced expansion of adult hematopoietic stem cells ex vivo. Cell 109, 39–45 (2002). 21. Reya, T. et al. A role for Wnt signalling in self-renewal of haematopoietic stem cells. Nature 423, 409–414 (2003).

NATURE GENETICS VOLUME 38

[

NUMBER 11

[

NOVEMBER 2006

22. Iwama, A. et al. Enhanced self-renewal of hematopoietic stem cells mediated by the polycomb gene product Bmi-1. Immunity 21, 843–851 (2004). 23. Duncan, A.W. et al. Integration of Notch and Wnt signaling in hematopoietic stem cell maintenance. Nat. Immunol. 6, 314–322 (2005). 24. Valk, P.J. et al. Prognostically useful gene-expression profiles in acute myeloid leukemia. N. Engl. J. Med. 350, 1617–1628 (2004). 25. Bullinger, L. et al. Use of gene-expression profiling to identify prognostic subclasses in adult acute myeloid leukemia. N. Engl. J. Med. 350, 1605–1616 (2004). 26. Bjerregaard, M.D., Jurlander, J., Klausen, P., Borregaard, N. & Cowland, J.B. The in vivo profile of transcription factors during neutrophil differentiation in human bone marrow. Blood 101, 4322–4332 (2003). 27. Rosenbauer, F. et al. Lymphoid cell growth and transformation are suppressed by a key regulatory element of the gene encoding PU.1. Nat. Genet. 38, 27–37 (2006). 28. Reya, T., Morrison, S.J., Clarke, M.F. & Weissman, I.L. Stem cells, cancer, and cancer stem cells. Nature 414, 105–111 (2001). 29. Passegue, E., Jamieson, C.H., Ailles, L.E. & Weissman, I.L. Normal and leukemic hematopoiesis: are leukemias a stem cell disorder or a reacquisition of stem cell characteristics? Proc. Natl. Acad. Sci. USA 100 (Suppl.) 11842–11849 (2003). 30. Hope, K.J., Jin, L. & Dick, J.E. Acute myeloid leukemia originates from a hierarchy of leukemic stem cell classes that differ in self-renewal capacity. Nat. Immunol. 5, 738–743 (2004). 31. Michor, F. et al. Dynamics of chronic myeloid leukaemia. Nature 435, 1267–1270 (2005). 32. Wang, J.C. & Dick, J.E. Cancer stem cells: lessons from leukemia. Trends Cell Biol. 15, 494–501 (2005). 33. Akashi, K., Traver, D., Miyamoto, T. & Weissman, I.L. A clonogenic common myeloid progenitor that gives rise to all myeloid lineages. Nature 404, 193–197 (2000). 34. Manz, M.G., Miyamoto, T., Akashi, K. & Weissman, I.L. Prospective isolation of human clonogenic common myeloid progenitors. Proc. Natl. Acad. Sci. USA 99, 11872– 11877 (2002). 35. Jamieson, C.H. et al. Granulocyte-macrophage progenitors as candidate leukemic stem cells in blast-crisis CML. N. Engl. J. Med. 351, 657–667 (2004). 36. Steidl, U. et al. Primary human CD34+ hematopoietic stem and progenitor cells express functionally active receptors of neuromediators. Blood 104, 81–88 (2004). 37. Ivanova, N.B. et al. A stem cell molecular signature. Science 298, 601–604 (2002). 38. Kiel, M.J. et al. SLAM family receptors distinguish hematopoietic stem and progenitor cells and reveal endothelial niches for stem cells. Cell 121, 1109–1121 (2005). 39. Li, C. & Wong, W.H. Model-based analysis of oligonucleotide arrays: expression index computation and outlier detection. Proc. Natl. Acad. Sci. USA 98, 31–36 (2001). 40. Li, C. & Hung, W.W. Model-based analysis of oligonucleotide arrays: model validation, design issues and standard error application. Genome Biol. 2 RESEARCH0032 (2001). 41. Ramalho-Santos, M., Yoon, S., Matsuzaki, Y., Mulligan, R.C. & Melton, D.A. ‘‘Stemness’’: transcriptional profiling of embryonic and adult stem cells. Science 298, 597–600 (2002). 42. Landgrebe, J., Wurst, W. & Welzl, G. Permutation-validated principal components analysis of microarray data. Genome Biol. 3 RESEARCH0019 (2002). 43. Donninger, H. et al. Whole genome expression profiling of advance stage papillary serous ovarian cancer reveals activated pathways. Oncogene 23, 8065–8077 (2004). 44. Kobayashi, S. et al. Calpain-mediated X-linked inhibitor of apoptosis degradation in neutrophil apoptosis and its impairment in chronic neutrophilic leukemia. J. Biol. Chem. 277, 33968–33977 (2002). 45. Kobayashi, S. et al. EGFR mutation and resistance of non-small-cell lung cancer to gefitinib. N. Engl. J. Med. 352, 786–792 (2005). 46. Pear, W.S. et al. Efficient and rapid induction of a chronic myelogenous leukemia-like myeloproliferative disease in mice receiving P210 bcr/abl-transduced bone marrow. Blood 92, 3780–3792 (1998). 47. DeKoter, R.P., Walsh, J.C. & Singh, H.P.U. 1 regulates both cytokine-dependent proliferation and differentiation of granulocyte/macrophage progenitors. EMBO J. 17, 4456–4468 (1998). 48. Steidl, U. et al. Gene expression profiling identifies significant differences between the molecular phenotypes of bone marrow-derived and circulating human CD34+ hematopoietic stem cells. Blood 99, 2037–2044 (2002). 49. Dolbeare, F., Gratzner, H., Pallavicini, M.G. & Gray, J.W. Flow cytometric measurement of total DNA content and incorporated bromodeoxyuridine. Proc. Natl. Acad. Sci. USA 80, 5573–5577 (1983). 50. Verhaak, R.G. et al. Mutations in nucleophosmin (NPM1) in acute myeloid leukemia (AML): association with other gene abnormalities and previously established gene expression signatures and their favorable prognostic significance. Blood 106, 3747–3754 (2005).

1277