ORIGINAL ARTICLE PML–Rara Initiates Leukemia by Conferring Properties of Self-Renewal to Committed Promyelocytic Progenitors

PML–RARa initiates leukemia by conferring properties of self-renewal to committed promyelocytic progenitors

S Wojiski1,2, FC Guibal3, T Kindler1,2, BH Lee1,2, JL Jesneck4,5, A Fabian1,2, DG Tenen3,6 and DG Gilliland1,2,6

1Division of Hematology, Brigham and Women’s Hospital, Boston, MA, USA; 2Department of Genetics, Howard Hughes Medical Institute, Harvard Medical School, Boston, MA, USA; 3Department of Hematology/Oncology, Harvard Institutes of Medicine, Boston, MA, USA; 4Cancer Program, Broad Institute of Harvard and MIT, Cambridge, MA, USA; 5Department of Pediatric Oncology, Dana-Farber Cancer Institute, Harvard Medical School, Boston, MA, USA and 6Harvard Stem Cell Institute, Harvard University, Boston, MA, USA

Acute promyelocytic leukemia (APL) is characterized by the leukemia stem cells (LSCs) are rare cells contained within hyperproliferation of promyelocytes, progenitors that are the more primitive CD34 þ CD38À population, resembling committed to terminal differentiation into granulocytes, making it an ideal disease in which to study the transforming potential transformed hematopoietic stem cells (HSCs). LSC activity was of less primitive cell types. We utilized a murine model of APL in shown by these investigators in all AML subtypes tested, but which the PML–RARa oncogene is expressed from the were not shown in acute promyelocytic leukemia (APL) endogenous cathepsin G promoter to test the hypothesis that associated with the expression of the PML–RARa fusion as a leukemia stem cell (LSC) activity resides within the differen- consequence of an acquired t(15;17). These observations tiated promyelocyte compartment. We prospectively purified suggest that either LSCs were not present in this differentiated promyelocytes from transgenic mice at various stages of disease and observed that PML–RARa-expressing promyelo- subtype of leukemia, or were resident in another hematopoietic cytes from young preleukemic mice had acquired properties of compartment. self-renewal both in vitro and in vivo. Progression to acute Although the cell of origin for LSCs is not known, studies in leukemia was associated with an expansion of the promyelo- mice have indicated that HSCs may be targets of transformation cyte compartment at the expense of other stem, progenitor and in leukemogenesis.4–8 This is an attractive hypothesis in that terminally differentiated populations. Leukemic promyelocytes normal tissue stem cells have inherent long-term self-renewal exhibited properties of self-renewal, and were capable of engendering leukemia in secondary recipient mice. These data potential that is requisite for tumorigenesis. However, a recent indicate that PML–RARa alone can confer properties of self- body of evidence suggests that more mature progenitor cells, renewal to committed hematopoietic progenitors before that normally lack any potential for self-renewal, may be an the onset of disease. These findings are consistent with the origin of LSCs. In this model, committed progenitors reacquire hypothesis that cancer stem cells may arise from committed properties of self-renewal mediated by the respective leukemia progenitors that lack stem cell properties, provided that the oncogene.9,10 For example, myeloid progenitors of the granu- initiating mutation in cancer progression activates programs that confer properties of self-renewal. locyte-monocyte lineage (GMPs) may acquire leukemia-initia- Leukemia (2009) 23, 1462–1471; doi:10.1038/leu.2009.63; ting potential in mice after retroviral transduction of the 8 11 published online 26 March 2009 leukemia-associated fusion proteins MLL-ENL, MOZ-TIF2 or Keywords: self-renewal; progenitor; promyelocyte; APL MLL-AF9.12 These leukemia-initiating progenitors can be distinguished from their normal counterparts in that they share certain characteristics of normal HSCs, including properties of self-renewal and the capacity to differentiate into progeny that Introduction lack self-renewing potential. These models are not mutually exclusive; LSCs may poten- The cancer stem cell hypothesis posits the existence of a rare tially arise either from the stem cell or from the progenitor population of tumor cells that is responsible for propagation and compartment. However, hematopoietic progenitors in the maintenance of a tumor phenotype. Furthermore, similar to myeloid lineage are short lived, and AML, like most cancers, normal developmental hierarchies, it has been proposed that requires multiple mutations. Thus, for an LSC to arise from a cancer stem cell, but not their clonogenic progeny, alone committed progenitor, one must posit that the initiating possess properties of long-term self-renewal. Thus, identification mutation itself confers self-renewal potential to a committed and characterization of these leukemia-initiating cells could progenitor. The retroviral transduction models noted above do translate into the development of more effective treatment not explicitly address this possibility. Tumors that arise in this modalities. context are typically mono- or oligoclonal, suggesting that There is strong experimental support for the existence of secondary mutations may be involved, and that integration site cancer stem cell in human acute myeloid leukemias (AMLs), effects may also contribute to the disease phenotype. initially from experiments using a NOD-SCID xenotransplanta- To better understand the role of leukemia oncogenes in tion model of human AML cells.1–3 These studies indicated that committed progenitor populations, we took advantage of a mouse model of leukemia developed by Westervelt and Correspondence: Dr DG Gilliland, Professor of Medicine, Harvard colleagues13 in which (i) PML–RARa is expressed from the Medical School and Brigham and Women’s Hospital, Division of endogenous murine cathepsin G promoter that obviates con- Hematology, 1 Blackfan Circle, Boston, MA 02115, USA. E-mail: [email protected] tribution to phenotype from retroviral integration sites, (ii) Received 21 February 2009; accepted 24 February 2009; published leukemia develops in a terminally differentiating hematopoietic online 26 March 2009 compartment, the promyelocyte, that has no potential for Self-renewal of PML–RARa-expressing promyelocytes S Wojiski et al 1463 self-renewal, and (iii) there is a long latency before development (MEPs) as LinÀSca-1Àc-Kit þ CD34ÀFcgRII/IIIlo. For promyelo- of leukemia in which animals are phenotypically normal, cyte and mature granulocyte identification, staining was carried allowing for the analysis of the consequences of PML–RARa out as described with the following changes: Sca-1 antibody was expression during the earliest stages of malignant transformation. added to the lineage depletion cocktail, and Gr-1 antibody was Our studies show that the acquisition of self-renewal excluded from the lineage depletion cocktail. Promyelocytes capability in the promyelocyte compartment is mediated by were distinguished as c-Kit þ CD34 þ Gr-1 þ and mature PML–RARa as an initiating step in the pathogenesis of APL, granulocytes as c-KitÀCD34ÀGr-1 þ . Dead cells were excluded providing further evidence that committed progenitors can in from all analyses by propidium-iodide staining. Cells were fact possess leukemia-initiating activity. Furthermore, the use of sorted into 30% fetal bovine serum/phosphate-buffered a knock-in model in which the PML–RARa fusion is expressed saline for subsequent assays. Analysis and purification was from the endogenous cathepsin G promoter allows for performed using an FACSAria cytometer (Becton Dickinson) and transplantation of syngeneic tissues to show the leukemogenic FACSDiva or FlowJo (Treestar, San Carlos, CA, USA) software potential of transformed committed progenitors, providing a programs. platform for addressing mechanisms of the regulation of self-renewal capability in the LSC compartment. Quantitative real-time PCR (RT-PCR) RNA was extracted from cells using the RNeasy Micro or Mini Materials and methods Kits (Qiagen, Valencia, CA, USA) per the manufacturer’s instructions, incorporating a DNAse1 digest to remove any Mice contaminating genomic DNA. cDNA was prepared from RNA Cathepsin-G-PML–RARa knock-in mice13 were backcrossed at using Taqman reverse transcription reagents (Applied Biosys- least eight generations into the C57BL/6 background, and this tems, Foster City, CA, USA). Taqman gene expression assays strain was used for all subsequent experiments. Genotyping was were used to analyze the expression of the following murine performed as described previously utilizing genomic tail DNA genes: cathepsin G, neutrophil elastase, gelatinase B, myeloper- as a PCR template.13,14 All animals were housed in micro- oxidase and glyceraldehyde 3-phosphate dehydrogenase. isolator cages under pathogen-free conditions, and all experi- Detection of expression of the PML–RARa fusion transcript ments were conducted with the ethical approval of the was carried out using a custom made Taqman probe specific to Children’s Hospital Animal Care and Use Committee. Animals the fusion junction and flanking primers. PCR reactions were were observed on a weekly basis for the development of performed and analyzed using the 7300 Real Time PCR System hematopoietic malignancy and were killed when moribund. and SDS Software (Applied Biosystems). Expression values were Automated total and differential blood cell counts were normalized to GAPDH. obtained using a Hemavet 950 (Drew Scientific, CT, USA). Upon killing, spleen weights were recorded, and all organs were collected and stored in 10% neutral-buffered formalin (Sigma- Colony-forming assays and bone marrow Aldrich, St. Louis, MO, USA). Single cell suspensions were transplantation prepared from spleen and bone marrow, and cells were Purified populations of stem or progenitor cells were plated in subsequently frozen in 10% dimethylsulfoxide (Sigma-Aldrich)/ MethoCult GF M3434 methylcellulose culture medium (StemCell 90% fetal bovine serum for further analysis. Technologies, Vancouver, BC, Canada). For initial platings of purified populations, 500 HSCs, 1000 CMPs or GMPs, and 10 000 promyelocytes were seeded in duplicate cultures. Colonies were Histopathologic analysis counted after 7–9 days. Individual colonies or pooled colonies Tissues were fixed for at least 72 h in 10% neutral buffered from an entire culture dish were harvested for cytospin prepara- formalin (Sigma-Aldrich), dehydrated in ethanol, cleared in tions and subsequent staining with Wright–Giemsa. Serial xylene, and infiltrated with paraffin on an automated processor replating was performed every 7 days by amalgamating all cells (Leica Bannockburn, IL, USA). In all, 4 mm thick tissue sections from duplicate dishes and reseeding of new duplicate cultures at a were placed on charged slides, deparaffinized with xylene, density of 10 000 cells per plate. For all-trans-retinoic acid (ATRA; rehydrated through graded alcohol washes, and stained with Sigma-Aldrich) differentiation experiments, cells were plated in hematoxylin and eosin. M3434 media as described in the presence or absence of 1 mM ATRA. Cells were harvested after 7 days and stained with Wright– Giemsa. For transplantation, sorted promyelocytes or mature Flow cytometry and purification of hematopoietic stem granulocytes from healthy and leukemic PML-RAR\alpha hetero- and progenitor populations zygous mice were resuspended in Hank’s balanced salt solution. Cells were stained for immunophenotype analysis using the Cells were injected through the lateral tail vein into sublethally following antibodies: Allophycocyanin (APC)-conjugated Gr-1, irradiated (552 rad) syngeneic recipient mice (4–6 weeks of age). Phycoerythrin (PE)-conjugated Mac-1, APC-conjugated CD34 For in vivo reconstitution assays, 50 000 promyelocytes from and PE-conjugated c-Kit. Analysis was carried out using an healthy PML-RAR\alpha heterozygous animals were combined FACScalibur cytometer (Beckton Dickinson, Mountain View, CA, with 500 000 whole bone marrow cells from wild-type mice USA) and CELLQuest software. A minimum of 10 000 viable cells and injected into lethally irradiated (552 rad Â 2) syngeneic was analyzed by gating on 7-AAD-negative populations. Multi- recipient mice. parameter flow cytometry was used to isolate HSCs and myeloid progenitors as described previously.15 Briefly, HSCs were distinguished as LinÀSca-1 þ c-Kit þ (LSK fraction), common Microarray analysis myeloid progenitors (CMPs) as LinÀSca-1Àc-Kit þ CD34 þ FcgRII/ Total RNA was prepared from sorted promyelocytes isolated IIIlo, granulocyte-monocyte progenitors (GMPs) as LinÀSca-1Àc- from pooled bone marrow samples from wild-type, healthy Kit þ CD34 þ FcgRII/IIIhi, and megakaryocyte-erythrocyte progenitors PML-RAR\alpha heterozygous or leukemic PML-RAR\alpha

Leukemia Self-renewal of PML–RARa-expressing promyelocytes S Wojiski et al 1464 heterozygous animals using the RNeasy micro Kit (Qiagen). leukemic PML-RAR\alpha heterozygous animals were assessed Samples were prepared in triplicate for each genotype. Linear using a two-tailed unpaired t-test. amplification, biotin labeling and fragmentation of amplified cDNA were carried out using the Ovation RNA Amplification System V2 and the FL-Ovation cDNA Biotin Module V2 (NuGen Results Technologies, San Carlos, CA, USA) exactly according to the manufacturer’s instructions. Labeled probes were hybridized to PML–RARa confers properties of self-renewal to Affymetrix (San Jose, CA, USA) GeneChip Mouse Genome 430A hematopoietic progenitors 2.0 arrays. Raw gene expression values were processed with the We first assessed the effect of expression of PML–RARa from the 16 robust multiarray analysis algorithm using BioConductor endogenous cathepsin G promoter on hematopoietic stem and 17 software. progenitor compartments in bone marrow derived from 6–8- week-old heterozygous transgenic mice before the development of leukemia (hereafter referred to as ‘Healthy PR/ þ ’). We used Statistical analyses multiparameter flow cytometry to analyze the HSC compart- Statistical significance of differences in parameters measured ment (LinÀSca1 þ Kit þ , LSK cells), CMPs, GMP or MEP popula- between wild-type, healthy PML-RAR\alpha heterozygous and tions15 (see Materials and methods). We observed no differences

a LSKs b Progenitors c 1.3 1.2 LSKs 1.1 105 Progenitors 105 1.0 GMP 0.9 0.8 4 4 0.7 10 10 0.6 0.5 0.4 RII/III

c-kit 3 103 γ 10 MEP 0.3 CMP 0.2 Fc 0.1 0.0 2 2 10 10 wt Healthy Leukemic wt wt PR/+ PR/+ 102 103 104 105 102 103 104 105 35 Sca-1 CD34 30 25 20 105 Progenitors LSKs 5 10 GMP 15 10 4 4 10 10 5 0 RII/III

γ 3 MEP c-kit 103 10 wt Healthy Leukemic Fc CMP PR/+ PR/+ 80 2 102 70 10 Healthy PR/+ Healthy PR/+ 60 50 102 103 104 105 102 103 104 105 40 Sca-1 CD34 30 % GMPs 20 10 5 Progenitors LSKs 5 10 10 GMP 0 wt Healthy Leukemic PR/+ PR/+ 104 104 25 RII/III

c-kit 3 γ 3 MEP 20 10 10 CMP Fc 15 2 10 102 PR/+ leukemic PR/+ leukemic 10 % MEPs % CMPs % LSKs

2 3 4 5 5 10 10 10 10 102 103 104 105 Sca-1 CD34 0 wt Healthy Leukemic PR/+ PR/+

Figure 1 PML–RARa expression alters stem and progenitor compartments in the leukemic state. Whole bone marrow from wild-type (wt), healthy PR/ þ or leukemic PR/ þ animals was sorted into hematopoietic stem cell (LinÀ, Sca-1+, c-kit+, LSK fraction), common myeloid progenitor (CMP), granulocyte-monocyte progenitor (GMP) and megakaryocyte-erythrocyte progenitor (MEP) populations using multiparameter flow cytometry. (a) Representative sort profiles for LSKs. (b) Representative sort profiles for CMPs, GMPs and MEPs. (c) Graphical representation of the frequencies of each population in wild-type, healthy PR/ þ and leukemic PR/ þ animals (mean±s.e.m.). LSK frequencies are represented as the percentage of sca-1 þ c-Kit þ cells within the lineage-negative bone marrow fraction. Progenitor subpopulation frequencies are represented as the percentage of CD34 þ FcgRII/IIIhi (GMP), CD34 þ FcgRII/IIIlo (CMP) and CD34ÀFcgRII/IIIlo (MEP) cells within the lineage negative, c-Kit þ sca-1À bone marrow fraction.

Leukemia Self-renewal of PML–RARa-expressing promyelocytes S Wojiski et al 1465 in the proportion of these stem and progenitor compartments with immature myeloid morphology (Figures 2bii, 2biii and 2c). when comparing PR/ þ with wild-type littermate controls This is in contrast with the typical granulocyte-monocyte colony (Figures 1a and b, shown graphically in c). We also tested the morphology containing differentiated granulocytic and mono- LSK, CMP and GMP compartments for the expression of PML– cytic cell types observed in plates seeded with wild-type cells RARa and found low levels of expression of the transgene in (Figures 2biv–2bvi). Morphology was indistinguishable between earlier stem and progenitor compartments. In that the fusion is colonies formed from an initial plating of LSKs, CMPs or GMPs expressed from the endogenous cathepsin G locus, these (data not shown). These findings indicated that PML–RARa could findings suggest that the cathepsin G promoter is active, confer certain properties of self-renewal to early progenitor although at much lower levels, in other compartments within populations in vitro, even in the absence of frank leukemia. the myeloid lineage in addition to the promyelocyte compartment (See Supplementary Figure 1). These findings confirm and expand on previously reported data indicating a minimal Prospective identification and purification of phenotype in transgenic animals before the development of promyelocyte progenitors from mouse whole bone leukemia.13 marrow We then tested each of these populations of cells, respectively, Although we observed functional changes in the LSK, CMP and for serial replating potential in methylcellulose in the absence GMP compartments in healthy PR/ þ mice, we were most of stroma, an in vitro surrogate for self-renewing interested in assessing the effects of PML–RARa expression in potential.4,5,8,11,18,19 LSK, CMP or GMP derived from healthy more differentiated myeloid lineage cells, including promyelo- PR/ þ mice showed serial replating potential, whereas cells cytes and terminally differentiated granulocytes. We developed derived from wild-type mice did not (Figure 2a). Colonies were a novel multiparameter flow cytometry strategy with c-Kit, blast-like in appearance, with small compact size and rounded CD34 and Gr-1 staining that enabled prospective purification of edges (Figure 2bi), and contained c-Kit þ Gr-1 þ Mac-1 þ cells these cells (Figure 3a). After lineage and Sca-1 depletion, mature granulocytes were contained within the c-KitÀCD34ÀGr-1 þ a fraction of the bone marrow, whereas promyelocytes were 650 þ þ þ 600 Week 1 found within the c-Kit CD34 Gr-1 fraction (Figures 3a and b). 550 Week 2 500 Sorted granulocytes displayed characteristic morphologic fea- 450 Week 3 tures, including multilobed nuclei, whereas sorted pro- 400 Week 4 350 myelocytes were more immature in appearance, with a higher 300 Week 5 nuclear to cytoplasmic ratio and the presence of brightly 250 staining azurophilic granules (Figure 3b, arrows). The promye- 10,000 cells 200 150 locytic identity of the purified population was further confirmed 100 Number of colonies per by quantitative RT-PCR analysis of the expression of primary and 50 0 secondary granule proteins associated with granulocytic maturation (Figure 3c). As would be expected, promyelocytes expressed the primary granule genes for neutrophil elastase, WT LSK WT CMP WT GMP PR LSK PR CMP PR GMP cathepsin G and myeloperoxidase. In contrast, promyelocytes b did not express gelatinase B, which encodes a secondary i ii iii granule protein associated with the later myelocyte stage of granulocyte development.20 This expression profile was specific for the promyelocyte compartment, as GMPs expressed markedly lower levels of myeloperoxidase and cathepsin G iv v vi (Supplementary Figure 2). These data were corroborated by genome-wide expression analysis of purified promyelocytes performed using Affymetrix microarrays (Figure 3d).

PML–RARa confers properties of self-renewal to c 104 104 promyelocytes 3 3 10 10 We utilized our ﬂow cytometric approach to purify and 2 99% 2 10 10

Gr-1 compare promyelocytes and mature granulocytes from bone 101 CD34 101 30% marrow derived from wild-type or PR/ þ animals before 100 100 100 101 102 103 104 100 101 102 103 104 development of leukemia. Although there were no significant Mac1 c-kit differences in peripheral blood counts between wild-type and PR/ þ animals in this model (data not shown), there was a subtle a Figure 2 PML–RAR confers in vitro self-renewal ability to bone increase in granulocytes in the bone marrow of healthy PR/ þ marrow progenitors. (a) Numbers of colonies observed per 10 000 cells plated in serial replating assays from either wild-type (left) or animals compared with wild type (12.25 vs 3.58%, Figures 4a healthy PR/ þ (right) animals (mean±s.e.m.). LSKs, common myeloid and b, shown graphically in d). The number of promyelocytes progenitors (CMPs) and granulocyte-monocyte progenitors (GMPs) was also modestly increased (1.27 vs 0.46%, Figures 4a and b, refer to the origin of the cells plated in methylcellulose to initiate the shown graphically in d), although none of these differences experiment. (b) Colony morphology (i, iv, Â 40 magnification) and reached statistical significance. Expression of PML–RARa did cytospin morphology (ii, iii, v, vi, Wright–Giemsa stain, Â 600 not alter the phenotype of the promyelocyte compartment, as magnification) of individual week two colonies picked from methylcellulose plates of either PR/ þ cells (upper panels) or wild-type cells granule protein expression profiles were virtually indistinguish- (lower panels). (c) Representative immunophenotype of cells pooled able between wild-type and healthy PR/ þ animals (Figure 4e). from week two colonies and stained with the myeloid markers Gr-1 However, despite the similarity in granule protein expression and Mac-1 and the progenitor markers c-Kit and CD34. profiles between wild-type and PML–RARa-expressing promye-

Leukemia Self-renewal of PML–RARa-expressing promyelocytes S Wojiski et al 1466 104 a b Granulocytes Promyelocytes Lineage/Sca-1 103 Depletion sca-1, CD3, CD4, 102

Harvest bone CD8, B220, CD19, IL- Lineage 1 marrow 7Rα, Ter119 10 100 0 200 400 600 800 1000 Side Scatter

Lineage low fraction c-kit-CD34- fraction c-kit+CD34+ fraction 104 104 104

103 103 103

2 2 2

RII/III 10 10 10 RII/III

c-kit

1 Fc 101 10 101 0 100 10 0 0 1 2 3 4 10 100 101 102 103 104 10 10 10 10 10 100 101 102 103 104 CD34 Gr-1 Gr-1

GRANULOCYTES PROMYELOCYTES

c d 100000 100000

10000 10000

1000 1000

100 100

Log Relative Expression 10 10

Ela2 Mpo Ctsg Gel B Log Normalized Expression Ela2 Mpo Ctsg Gel B

Figure 3 Purified populations of granulocytes and promyelocyte progenitors can be obtained using multiparameter flow cytometry. (a) Sorting scheme for isolation and purification of mature granulocytes and promyelocyte progenitors from a wild-type mouse. (b) Cytospin preparations showing characteristic morphologic features of purified granulocytes and promyelocytes (Wright–Giemsa stain, Â 600 magnification). (c) Quantitative real-time PCR (RT-PCR) analysis of purified promyelocytes for the expression of neutrophil elastase (Ela2), myeloperoxidase (Mpo), cathepsin G (Ctsg) and gelatinase B (Gel B). Expression values are relative to Gapdh expression (mean±s.e.m.). (d) Affymetrix microarray analysis of expression of Ela2, Mpo, Ctsg and Gel B in purified promyelocyte populations.

locytes, a striking finding was observed in serial replating assays: the acquisition of secondary mutations.13,21 We assayed the promyelocytes derived from PR/ þ animals, but not wild-type effects of leukemic transformation on the stem and progenitor animals, showed serial replating activity in the absence of compartments using multiparameter flow cytometry, and stroma (Figure 5a). There was no replating activity in granulo- observed a marked reduction in the number of LSKs, CMPs cytes derived from either genotype, which may be due to the and MEPs compared with healthy PR/ þ or wild-type controls, fact that the expression level of PML–RARa is significantly with a concomitant expansion of the GMP compartment reduced in the granulocyte compartment compared with the (Figures 1a–c, shown graphically in d). Furthermore, in leukemic promyelocyte compartment (Figure 5b). These in vitro data animals (hereafter referred to as ‘Leukemic PR/+’), there was a suggested that PML–RARa expression in the promyelocyte dramatic expansion of the promyelocyte compartment that compartment may lead to enhanced self-renewal, as was comprised 25.42% of the cells after lineage depletion, observed in earlier progenitor compartments. To test this compared with 1.27% in healthy PR/ þ animals and 0.46% of hypothesis in vivo, we purified promyelocytes from healthy wild-type animals (Figures 4a–d, Po0.0001). The expansion of PR/ þ animals and transplanted 50 000 of these cells with helper the promyelocyte compartment was coupled with a reduction in bone marrow into syngeneic recipient animals. We then the number of mature granulocytes to normal levels when monitored for the expression of PML–RARa in the peripheral compared with healthy PR/ þ mice (Figure 4d). Promyelocytes blood of these animals over the course of 16 weeks. As shown in derived from leukemic animals retained their morphology and Figure 5c, PML–RARa expression was observed in the peripheral expression of primary granule proteins as determined by blood of transplanted mice at all time points tested, indicating cytospin and quantitative RT-PCR analyses (Figure 4e and data long-term reconstitution of recipient animals with PML–RARa- not shown). expressing promyelocytes. These findings suggest that this terminally differentiating population of cells had acquired potential for self-renewal before the development of leukemia Leukemic promyelocytes are enriched for as a consequence of expression of PML–RARa. leukemia-initiating activity and are retinoic acid responsive Given the significant expansion observed in the promyelocyte PML–RARa alters hematopoietic stem, progenitor and compartment of leukemic PR/ þ animals, we hypothesized that granulocyte compartments in the leukemic state this population of cells harbored leukemia-initiating activity. To PR/ þ animals ultimately develop a phenotype that is very test this hypothesis, we sorted promyelocytes and granulocytes similar to human APL, with a long latency that is associated with from leukemic and healthy PR/ þ animals, and transplanted

Leukemia Self-renewal of PML–RARa-expressing promyelocytes S Wojiski et al 1467 a Promyelocytes Granulocytes 104 104 104 d 3 3 3 p < 0.0001 10 I 10 10 3.58%

2 2 2 30

RII/II c-kit 10 10 10

RII/II

1 Fc 1 1

10 10 Fc 10 wt 0.46% 0 0 0 20 10 0 1 2 3 4 10 10 0 1 2 3 4 10 10 10 10 10 100 101 102 103 104 10 10 10 10 10 CD34 Gr-1 Gr-1 10 b 104 104 104 p = n.s.

3 3 3 12.25% % Promyelocytes

I 10 10 I 10

2 2 2 0

RII/II 10 10 RII/II 10

c-kit

γ Healthy γ wt Healthy Leukemic

1 Fc 1 1 10 10 1.27% Fc 10 PR/+ PR/+ PR/+ 100 100 100 100 101 102 103 104 100 101 102 103 104 100 101 102 103 104 p = n.s. CD34 Gr-1 Gr-1

c 104 104 104 p = n.s. 3 3 3 2.70% 17.5

I 10 I 10 10

2 2 2 15.0

RII/II 10 RII/II 10 10

γ Leukemic c-kit 12.5

1 Fc 1 1 10 10 25.42% 10 PR/+ 10.0 0 0 0 10 0 1 2 3 4 10 0 1 2 3 4 10 7.5 10 10 10 10 10 10 10 10 10 10 100 101 102 103 104 5.0 CD34 Gr-1 Gr-1 e % Granulocytes 2.5 1000000 0 wt 100000 wt Healthy Leukemic Healthy PR/+ PR/+ PR/+ 10000 Leukemic PR/+ 1000 100 10 Log Relative Expression 1 Ela2 Mpo Ctsg Gel B

Figure 4 Promyelocytes expressing PML–RARa are expanded in the leukemic state. Lineage/Sca-1-depleted bone marrow cells from wild-type (a), healthy PR/ þ (b) and leukemic PR/ þ (c) mice were gated into c-kit þ CD34 þ and c-kitÀCD34À fractions (left panels). Promyelocytes were gated based on c-kit þ CD34 þ Gr-1 þ immunophenotype (middle panels) and mature granulocytes were gated based on c-kitÀCD34ÀGr-1 þ immunophenotype (right panels). Expanded populations have bold gating. (d) Graphical representation of the frequencies of promyelocytes (upper panel) and granulocytes (lower panel) present in wild-type, healthy PR/ þ and leukemic PR/ þ animals (mean±s.e.m.). Promyelocyte frequencies are represented as the percentage of c-kit þ CD34 þ Gr-1 þ cells among all live cells, whereas granulocyte frequencies are represented as the percentage of c-kitÀCD34ÀGr-1À cells among all live cells. (e) Quantitative real-time PCR (RT-PCR) analysis of puriﬁed promyelocytes from wild- type, healthy PR/ þ and leukemic PR/ þ animals for the expression of neutrophil elastase (Ela2), myeloperoxidase (Mpo), cathepsin G (Ctsg) and gelatinase B (Gel B). Expression values are relative to Gapdh expression (mean±s.e.m.).

these purified populations into sublethally irradiated syngeneic (Figure 6e). After 7 days in methylcellulose culture, untreated recipient mice. As shown in Figure 6, promyelocytes from leukemic promyelocytes retained their immature morphology, leukemic donor mice were able to transfer disease in vivo. Mice whereas the same cells treated with 1 mM ATRA showed signs of receiving 50 000 flow-purified leukemic promyelocytes suc- differentiation and apoptosis. Thus, it can be inferred that ATRA cumbed to a rapidly fatal acute leukemia with a median disease is capable of reprograming promyelocytic LSCs to terminally latency of 42 days (Figure 6a). The disease recapitulated the differentiated cells that are incapable of self-renewal, and leukemic phenotype observed in primary animals with hyper- accounts in part for the disease-remitting activity of ATRA. cellular marrow predominated by myeloid cells that were c-Kit þ Gr1 þ Mac-1 þ (Figures 6b and c) with infiltration of liver and spleen (Figures 6b and c). Limit dilution transplant assays Discussion with purified promyelocytes indicated that the frequency of leukemia-initiating cells was B1:100 within the leukemic The cancer/LSC theory has recently been called into question,22 promyelocyte compartment (Figure 6d). In contrast, no disease mainly on the basis of the criticism of the xenograft transplanta- was observed when purified granulocytes derived from these tion model that has been used to lay the foundation for this area same animals were transplanted. These data indicate that of investigation in cancer research.1–3 Subsequent studies have promyelocytes, but not granulocytes, have efficient leukemia- extended the body of evidence that rare cells within the initiating activity, and that LSCs in this murine model of disease leukemic population drive tumor formation, and that these cells can be demarcated within the terminally differentiating myeloid could in fact be transformed progenitors.8,11,12 A caveat to these compartment. Of note, these promyelocytic leukemia-initiating analyses is that they rely on either human to mouse xenograft or cells were responsive to differentiation agent therapy with ATRA retroviral transduction of purified cell populations and sub-

Leukemia Self-renewal of PML–RARa-expressing promyelocytes S Wojiski et al 1468 a 500 Week 1 We selected APL as a model to characterize LSCs for several reasons. First, we were able utilize a well-characterized Week 2 400 knock-in mouse model of the disease in which expression of Week 3 PML–RARa is targeted to the promyelocyte compartment. Week 4 Second, APL represents a relatively more differentiated form of 300 acute leukemia compared with other subtypes, and is therefore an attractive model in which to test the possibility that more 200 differentiated progenitors can possess leukemia-initiating activity. Of note, APL (French-American-British subtype M3) 100 was the only AML subtype that was unable to transfer disease to

No. of colonies per 10,000 cells NOD-SCID recipient mice, regardless of the fraction of cells 0 transplanted.2,3 This lends further support to the notion that the wt Pro Healthy PR/+ Pro leukemia-initiating population within APL is most likely not derived from an HSC. Finally, APL is the only leukemia with a b 1.1 clinically proven differentiation therapy in the form of ATRA. 1.0 Therefore, the effects of a known successful treatment for the 0.9 disease could be tested speciﬁcally on the leukemia-initiating 0.8 p = 0.036 population. 0.7 Our results show that in APL, a committed progenitor, the 0.6 promyelocyte has the capacity to transfer disease upon 0.5 transplantation, and therefore possesses LSC properties. This 0.4 data support the notion that LSCs need not be derived from 0.3

Relative Expression HSCs, and that indeed a committed progenitor can be 0.2 transformed into a cell capable of maintaining the leukemic 0.1 0.0 clone. The hallmark features of LSCs are the ability to both self- PR/+ pro PR/+ gran renew and to differentiate into all cell types of the primary disease. Experiments by others have indicated that PML–RARa c 12 PR/+ may possess the potential to confer self-renewal, but these 11 BMT 1 assays were solely conducted in vitro and relied on retroviral 10 transduction of PML–RARa into stem and progenitor popula- 9 BMT 2 tions.23,24 Here, we have shown that progenitor populations 8 from PML–RARa-expressing animals do in fact possess the 7 capacity to self-renew. Importantly, this ability is conferred to 6 5 progenitor populations in the absence of frank leukemia; the 4 mice used in these experiments were roughly 8 weeks of age, and appeared perfectly healthy. Despite the lack of leukemic Relative Expression 3 2 transformation, PML–RARa-expressing promyelocytes displayed 1 properties of self-renewal both in vitro and in vivo, a striking 0 finding considering promyelocytes are thought to be postmitotic week 4 week 6 week 8 week 12 week 16 cells that are committed to terminal differentiation into Time Post Transplant granulocytes, cells that normally survive only hours to days Figure 5 Promyelocytes-expressing PML-RARa display in vitro and in once in circulation. This suggests that PML–RARa may confer vivo properties of self-renewal. (a) Numbers of colonies after serial self-renewal ability to progenitor populations normally lacking replating of purified promyelocytes from either wild-type or healthy this capacity as an initiating step in the process of leukemogen- PR/ þ animals (mean±s.e.m.). (b) Quantitative real-type (RT-PCR) esis. Indeed, analysis of the stem and myeloid progenitor a analysis of the expression of PML–RAR in purified promyelocyte (left) compartments of PR/ þ animals revealed no differences in the and granulocyte (right) populations. Expression values are relative to Gapdh expression (mean±s.e.m.). (c) Quantitative RT–PCR analysis of frequencies of these populations in the healthy state, suggesting the expression of PML-RARa in the peripheral blood of mice at various that despite the capacity for self-renewal, these cells do not yet time points post-transplant. BMT1 and BMT2 refer to two individual possess the requisite prosurvival or proliferative advantage mice reconstituted with 50 000 promyelocytes derived from healthy required for leukemogenesis. PR/ þ animals, along with wild-type helper bone marrow. The fact that APL is a disease of promyelocytes prompted us to PR/ þ refers to a healthy PR/ þ knock-in mouse that served as a carefully explore this more differentiated stage of granulocyte positive control for PML–RARa expression. development. Our results show that the shift from the healthy to leukemic state is accompanied by a massive expansion of the promyelocyte (c-Kit þ CD34 þ Gr-1 þ ) compartment. Further- more, this compartment is capable of transferring disease to sequent transplantation into secondary recipient animals. Thus, secondary recipient animals, and is highly enriched for they confer disease through non-physiologic conditions, and leukemia initiating activity as determined by limiting dilution cannot address questions regarding the pathologic initiation and transplants. These leukemic promyelocytes possess true LSC development of disease due to progressive oncogenic insults. properties, as defined by self-renewal and differentiation. Their Here, we have addressed these issues by assessing the LSC capacity to serially transplant disease indicates that they must theory using a model system of disease that expresses an possess self-renewal ability to perpetuate the tumor. Addition- oncogene in the appropriate developmental compartment ally, flow cytometric and histologic analyses of tissues from without the need for retroviral transduction, thus more closely diseased animals confirm that the leukemic animals succumbed mimicking the human pathologic condition. to a disease that recapitulated that of the primary donor.

Leukemia Self-renewal of PML–RARa-expressing promyelocytes S Wojiski et al 1469 a b BM Spleen 104 104 63% 7% 29% 110 103 103 100 102 102 90

Mac-1 80 Mac-1 101 101 70 62% 100 100 60 100 101 102 103 104 100 101 102 103 104 50 Leukemic Promyelocytes (n=7) Gr-1 Gr-1 40 4 4 30 Leukemic Granulocytes (n=4) 10 10 38% Percent survival 17% 20 Healthy Promyelocytes (n=5) 103 103 10 2 2 0 10 10 c-kit

c-kit 0 25 50 75 100 125 150 175 200 101 101 Days Post BMT 100 100 100 101 102 103 104 100 101 102 103 104 CD34 CD34 c BM Liver Spleen

d 110 e 100 50000 L-Pro (n=6) 90 no ATRA 80 10000 L-Pro (n=6) 1000 L-Pro (n=6)

rvival 70 60 100 L-Pro (n=5) t Su t 50 40

30 Percen 20 10 1μM ATRA 0 0 50 100 150 200 250 300 Days Post BMT

Figure 6 Leukemic promyelocytes are enriched for leukemia-initiating activity. (a) Kaplan–Meier survival curves for recipient mice transplanted with 50 000 leukemic promyelocytes (blue), 50 000 leukemic granulocytes (black) or 50 000 non-leukemic promyelocytes (red). Curves represent combined data from two separate transplant experiments. One mouse receiving non-leukemic promyelocytes succumbed to a non-donor cell- derived T-cell lymphoma. (b) Flow cytometry of bone marrow (BM; left panels) or spleen (right panels) of a representative leukemic mouse transplanted with 50 000 leukemic promyelocytes. Top panels show immunophenotyping for the myeloid markers Gr-1 and Mac-1, whereas the bottom panels show immunophenotyping for the progenitor markers c-kit and CD34. (c) Histology of BM, liver and spleen from a leukemic mouse; hematoxylin and eosin stain, Â 100 magniﬁcation. (d) Kaplan–Meier survival curves for recipient mice transplanted with leukemic promyelocytes (L-Pro) at limiting dilutions (50 000 to 100 cells). Results shown represent combined data from two separate transplants. (e) Cytospins of leukemic promyelocytes grown in methylcellulose for 7 days in the absence (upper panel) or presence (lower panel) of 1 mM all-trans-retinoic acid (ATRA); Wright–Giemsa stain, Â 600 magniﬁcation.

These results suggest a model for the development of APL in RARa leads to massive expansion of promyelocyte progenitors which PML–RARa plays a key role in initiating disease by and acute leukemia results. conferring self-renewal, one of the requisite features of LSCs, to It is important to note that our results do not formally committed progenitors (see Supplementary Figure 3). Although address the target cell of transformation in APL in which the under normal physiologic conditions of myeloid development, initial PML–RARa mutation occurs, a question that remains HSC is the only cell capable of self-renewal in the pathway of unanswered in the APL ﬁeld.25,26 Interesting results obtained differentiation leading to the formation of a mature granulocyte, from Lane and Ley27 suggest that proteolytic processing of upon expression of PML–RARa, promyelocytic progenitors PML–RARa by neutrophil elastase is an important event for acquire the ability to self-renew. The sustained maintenance leukemic transformation,28 and therefore, PML–RARa may be of these progenitors ‘primes’ them for the acquisition of expressed in early stem and progenitor compartments, but may additional mutations that will lead to frank leukemia. Upon not be active or oncogenic until it is cleaved by enzymes that acquisition of additional mutations that confer proliferative and/ are only present later in myeloid differentiation. Further analysis or prosurvival advantage within the promyelocyte compartment, in both murine models and human patient samples will be cooperativity between these secondary mutations and PML– needed to clarify this debate.

Leukemia Self-renewal of PML–RARa-expressing promyelocytes S Wojiski et al 1470 These data provide several provocative insights into the 4 So CW, Karsunky H, Passegue E, Cozzio A, Weissman IL, Cleary contribution of the PML–RARa oncogene to leukemogenesis. ML. MLL-GAS7 transforms multipotent hematopoietic progenitors First, this experimental system enables the analysis of the and induces mixed lineage leukemias in mice. Cancer Cell 2003; physiologic consequences of oncogene expression in a com- 3: 161–171. 5 Lavau C, Szilvassy SJ, Slany R, Cleary ML. Immortalization and mitted progenitor before the onset of leukemia, and an ability to leukemic transformation of a myelomonocytic precursor by monitor changes with disease progression. Second, these retrovirally transduced HRX-ENL. EMBO J 1997; 16: 4226–4237. findings help to understand how a hematopoietic progenitor 6 Passegue E, Wagner EF, Weissman IL. JunB deficiency leads to a that is committed to differentiation and cell death may be a myeloproliferative disorder arising from hematopoietic stem cells. target for transformation. In the context of multiple mutations Cell 2004; 119: 431–443. that are required for leukemogenesis, it would be imperative 7 Neering SJ, Bushnell T, Sozer S, Ashton J, Rossi RM, Wang PY et al. Leukemia stem cells in a genetically defined murine model of that the initiating mutation enables self-renewal potential that blast-crisis CML. Blood 2007; 110: 2578–2585. would allow for acquisition of secondary mutations that confer 8 Cozzio A, Passegue E, Ayton PM, Karsunky H, Cleary ML, the overt leukemia phenotype. PML–RARa expression meets Weissman IL. Similar MLL-associated leukemias arising from self- these criteria in conferring long-term self-renewing potential renewing stem cells and short-lived myeloid progenitors. Genes to promyelocytes that are normally short lived. Third, it is Dev 2003; 17: 3029–3035. remarkable that the expression of a single leukemia onco- 9 Passegue E, Jamieson CHM, Ailles LE, Weissman IL. Normal and geneFPML–RARaFapparently confers such properties to a luekemic hematopoiesis: are leukemias a stem cell disorder or a reaquisition of stem cell characteristics? PNAS 2003; 100: cell that is primarily distinguished by its short half-life with 11842–11849. terminal differentiation. 10 Reya T, Morrison SJ, Clarke MF, Weissman IL. Stem cells, cancer, The identification and characterization of cancer stem cells and cancer stem cells. Nature 2001; 414: 105–111. has been catalyzed by the notion that current chemotherapeu- 11 Huntly BJ, Shigematsu H, Deguchi K, Lee BH, Mizuno S, Duclos N tics fail to eradicate cancer stem cells, leading to eventual et al. MOZ-TIF2, but not BCR-ABL, confers properties of leukemic disease relapse (reviewed by Reya et al.10). Therefore, there is stem cells to committed murine hematopoietic progenitors. Cancer Cell 2004; 6: 587–596. increasing interest in identifying biologically distinct features 12 Krivtsov AV, Twomey D, Feng Z, Stubbs MC, Wang Y, Faber J et al. of cancer stem cells that distinguish them from other cell types, Transformation from committed progenitor to leukaemia stem cell including normal stem cells, to design better therapies that initiated by MLL-AF9. Nature 2006; 442: 818–822. specifically target cancer stem cells. Recent data from hemato- 13 Westervelt P, Lane AA, Pollock JL, Oldfather K, Holt MS, Zimonjic logic malignancies provide evidence that differences do exist DB et al. High-penetrance mouse model of acute promyelocytic between normal HSCs and leukemia-initiating cells, providing leukemia with very low levels of PML-RARalpha expression. Blood 2003; 102: 1857–1865. hope that this strategy will be fruitful in eventually identifying 29–31 14 Lee BH, Tothova Z, Levine RL, Anderson K, Buza-Vidas N, Cullen targeted therapeutics that can eradicate LSCs. In the case of DE et al. FLT3 mutations confer enhanced proliferation and APL, our data suggest that the leukemia-initiating promyelocytes survival properties to multipotent progenitors in a murine model are susceptible to differentiation therapy mediated by ATRA of chronic myelomonocytic leukemia. Cancer Cell 2007; 12: therapy. However, particularly for the subpopulation of patients 367–380. who develop eventual resistance to ATRA therapy, it is 15 Akashi K, Traver D, Miyamoto T, Weissman IL. A clonogenic important to develop effective alternative therapies. common myeloid progenitor that gives rise to all myeloid lineages. Nature 2000; 404: 193–197. 16 Irizarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, Scherf U et al. Exploration, normalization, and summaries of high Conflict of interest density oligonucleotide array probe level data. Biostatistics 2003; 4: 249–264. The authors declare no conflict of interest. 17 Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S et al. Bioconductor: open software development for computational biology and bioinformatics. Genome Biol 2004; 5:R80. Acknowledgements 18 Higuchi M, O’Brien D, Kumaravelu P, Lenny N, Yeoh EJ, Downing JR. Expression of a conditional AML1-ETO We thank Maricel Gozo and Dena Leeman for technical support, oncogene bypasses embryonic lethality and establishes a murine model of human t(8;21) acute myeloid leukemia. Cancer Cell as well as Stefan Fro¨hling, Claudia Scholl and all the members of 2002; 1: 63–74. the Gilliland laboratory for helpful discussions and comments 19 Deguchi K, Ayton PM, Carapeti M, Kutok JL, Snyder CS, Williams regarding this study. DGG is a Doris Duke Foundation Distin- IR et al. MOZ-TIF2-induced acute myeloid leukemia requires the guished Clinical Scientist and a Howard Hughes Medical Institute MOZ nucleosome binding motif and TIF2-mediated recruitment of investigator. Supported in part by National Institute of Health CBP. Cancer Cell 2003; 3: 259–271. grants DK50654 and CA66996 and by the Leukemia and 20 Anderson KL, Smith KA, Pio F, Torbett BE, Maki RA. Neutrophils Lymphoma society to DGG. deficient in PU.1 do not terminally differentiate or become functionally competent. Blood 1998; 92: 1576–1585. 21 Zimonjic DB, Pollock JL, Westervelt P, Popescu NC, Ley TJ. Acquired, nonrandom chromosomal abnormalities associated with References the development of acute promyelocytic leukemia in transgenic mice. Proc Natl Acad Sci USA 2000; 97: 13306–13311. 1 Hope KJ, Jin L, Dick JE. Acute myeloid leukemia originates from a 22 Kelly PN, Dakic A, Adams JM, Nutt SL, Strasser A. Tumor growth hierarchy of leukemic stem cell classes that differ in self-renewal need not be driven by rare cancer stem cells. Science 2007; capacity. Nat Immunol 2004; 5: 738–743. 317: 337. 2 Bonnet D, Dick JE. Human acute myeloid leukemia is organized as 23 Minucci S, Monestiroli S, Giavara S, Ronzoni S, Marchesi F, a hierarchy that originates from a primitive hematopoietic cell. Insinga A et al. PML-RAR induces promyelocytic leukemias with Nat Med 1997; 3: 730–737. high efficiency following retroviral gene transfer into purified 3 Lapidot T, Sirard C, Vormoor J, Murdoch B, Hoang T, murine hematopoietic progenitors. Blood 2002; 100: 2989–2995. Caceres-Cortes J et al. A cell initiating human acute myeloid 24 Zheng X, Beissert T, Kukoc-Zivojnov N, Puccetti E, Altschmied J, leukaemia after transplantation into SCID mice. Nature 1994; 367: Strolz C et al. Gamma-catenin contributes to leukemogenesis 645–648. induced by AML-associated translocation products by increasing

Leukemia Self-renewal of PML–RARa-expressing promyelocytes S Wojiski et al 1471 the self-renewal of very primitive progenitor cells. Blood 2004; 28 Lane AA, Ley TJ. Neutrophil elastase is important for PML-retinoic 103: 3535–3543. acid receptor alpha activities in early myeloid cells. Mol Cell Biol 25 Turhan AG, Lemoine FM, Debert C, Bonnet ML, Baillou C, 2005; 25: 23–33. Picard F et al. Highly puriﬁed primitive hematopoietic stem cells 29 Jin L, Hope KJ, Zhai Q, Smadja-Joffe F, Dick JE. Targeting of CD44 are PML-RARA negative and generate nonclonal progenitors in eradicates human acute myeloid leukemic stem cells. Nat Med acute promyelocytic leukemia. Blood 1995; 85: 2154–2161. 2006; 12: 1167–1174. 26 Edwards RH, Wasik MA, Finan J, Rodriguez R, Moore J, Kamoun M 30 Yilmaz OH, Valdez R, Theisen BK, Guo W, Ferguson DO, Wu H et al. Evidence for early hematopoietic progenitor cell involvement in acute promyelocytic leukemia. Am J Clin Pathol 1999; 112: et al. Pten dependence distinguishes haematopoietic stem cells 819–827. from leukaemia-initiating cells. Nature 2006; 441: 475–482. 27 Lane AA, Ley TJ. Neutrophil elastase cleaves PML-RARalpha and 31 Krause DS, Lazarides K, von Andrian UH, Van Etten RA. is important for the development of acute promyelocytic leukemia Requirement for CD44 in homing and engraftment of BCR-ABL- in mice. Cell 2003; 115: 305–318. expressing leukemic stem cells. Nat Med 2006; 12: 1175–1180.

Supplementary Information accompanies the paper on the Leukemia website (http://www.nature.com/leu)

Leukemia