ORIGINAL ARTICLE

MicroRNA gene expression during retinoic acid-induced differentiation of

human acute promyelocytic leukemia

R Garzon1,5, F Pichiorri1,5, T Palumbo1,2, M Visentini3, R Aqeilan1, A Cimmino1, H Wang1,H Sun1, S Volinia1, H Alder1, GA Calin1, C-G Liu1, M Andreeff4 and CM Croce1

1Department of Molecular Virology, Immunology and Medical Genetics and Comprehensive Cancer Center, Ohio State University,Columbus, OH, USA; 2Department of Experimental and Clinical Pharmacology, University of Catania, Catania, Italy; 3Division ofClinical Immunology, University of Rome ‘La Sapienza’, Rome, Italy and 4Department of Blood and Bone marrow Transplantation,Section of Molecular Hematology and Therapy, The University of Texas MD Anderson Cancer Center, Houston, TX, USA

MicroRNAs (miRNAs) are small non-coding RNAs of19–25 nucleotides that are involved in the regulation ofcritical cell processes such as apoptosis, cell proliferationand differentiation. However, little is known about therole of miRNAs in granulopoiesis. Here, we report theexpression of miRNAs in acute promyelocytic leukemiapatients and cell lines during all-trans-retinoic acid(ATRA) treatment by using a miRNA microarraysplatform and quantitative real time–polymerase chainreaction (qRT–PCR). We found upregulation of miR-15a,miR-15b, miR-16-1, let-7a-3, let-7c, let-7d, miR-223,miR-342 and miR-107, whereas miR-181b was down-regulated. Among the upregulated miRNAs, miR-107 ispredicted to target NFI-A, a gene that has been involvedin a regulatory loop involving miR-223 and C/EBPaduring granulocytic differentiation. Indeed, we haveconfirmed that miR-107 targets NF1-A. To get insightsabout ATRA regulation of miRNAs, we searched forATRA-modulated transcription factors binding sites inthe upstream genomic region of the let-7a-3/let-7b clusterand identified several putative nuclear factor-kappa B(NF-jB) consensus elements. The use of reporter geneassays, chromatin immunoprecipitation and site-directedmutagenesis revealed that one proximal NF-jB bindingsite is essential for the transactivation of the let-7a-3/let-7bcluster. Finally, we show that ATRA downregulation ofRAS and Bcl2 correlate with the activation of knownmiRNA regulators of those proteins, let-7a and miR-15a/miR-16-1, respectively.Oncogene (2007) 26, 4148–4157; doi:10.1038/sj.onc.1210186;published online 29 January 2007

Keywords: microRNAs; promyelocytic leukemia; NF-kB

Introduction

Acute promyelocytic leukemia (APL) is a subtype ofacute myelogenous leukemia characterized by maturationarrest at the promyelocytic stage of development andcaused by a novel fusion protein resulting from thereciprocal translocation involving the retinoic acidreceptor alpha (RARa) on chromosome 17 with thepromyelocytic leukemia gene (PML) on chromosome 15(Melnick and Licht, 1999). This oncogenic protein(PML-RARa) interferes with the process of myeloiddifferentiation by transcriptional repression of retinoicacid (RA) responsive genes (Melnick and Licht, 1999). Inthe absence or at physiological levels of RA, PML-RARfusion protein recruits a corepressor complex that bindsto target promoters and represses transcription. Pharma-cological doses of all-trans-retinoic acid (ATRA) reversethe dominant-negative effect of PML-RARa fusionprotein and induce gene transcription via RA responsiveelements that are present in the promoters of RA-inducedgenes (Melnick and Licht, 1999).MicroRNAs (miRNAs) are small, non-protein-coding

RNAs of 19–25 nucleotides that regulate gene expres-sion by targeting messenger RNAs (mRNA) in asequence-specific manner inducing translational repres-sion or mRNA degradation (Bartel, 2004). MiRNAsplay important roles in the development, differentiation,metabolism, cell growth and death in animals and plants(Poy et al., 2004; Cheng et al., 2005; Dresios et al., 2005).There is also evidence that supports a role for miRNAsin hematopoiesis. In mice, miR-181a is detected in earlyprogenitor cells and dynamically upregulated duringB-cell commitment (Chen et al., 2004). Furthermore, theectopic expression of miR-181a in murine hematopoieticprogenitor cells led to a proliferation in the B-cellcompartment (Chen et al., 2004). More recently, it hasbeen reported that miR-223 is a key member of aregulatory circuit involving C/EBPa and NFI-A thatcontrols granulocytic differentiation in ATRA-treatedAPL cell lines (Fazi et al., 2005). Little is known,however, about other miRNAs expression in ATRA-treated APL cells.In the present study, we report the miRNA gene

expression profile of an APL cell line (NB4) uponReceived 14 February 2006; revised 18 October 2006; accepted 30 October2006; published online 29 January 2007

Correspondence: Professor CM Croce, Department of MolecularVirology, Immunology and Medical Genetics and ComprehensiveCancer Center, Ohio State University, Wiseman Hall, Room 445C, 40012th Avenue, Columbus, OH 43210, USA.E-mail: Carlo.Croce@osumc.edu5These two authors contributed equally to this work.

Oncogene (2007) 26, 4148–4157& 2007 Nature Publishing Group All rights reserved 0950-9232/07 $30.00

www.nature.com/onc

granulocytic differentiation with ATRA. The resultswere also confirmed in HL-60 cells, another mycloidleukemia cell line that undergoes granulocytic differen-tiation upon ATRA treatment and in primary APLpatient samples treated in vitro with ATRA.

Results

MiRNAs gene expression in APL cell lines and primarypatient blasts treated with ATRAThe treatment of NB4 cells with 100 nM of ATRA for96 h induced granulocytic differentiation as evidencedby morphological changes and by increased expressionof the surface antigen CD11b and CD15 (Figure 1). Weinitially compared the miRNA expression profiles ofATRA-treated cells with those of untreated cells at thesame time point using t-test pair analysis within thesignificant analysis of microarrays (SAM) (Tusher et al.,

2001). Using this method, we identified eight miRNAsupregulated and one miRNAs downregulated (Table 1).We then compared miRNA expression in ATRA-

treated vs untreated cells by using quantitative analysisin SAM, which computes the linear regression coeffi-cient of each gene on the outcome. Untreated cells wereset to 0 and ATRA-treated cells to the number of daysof treatment (i.e., 1, 2, 3 and 4). We thus selected themiRNAs, which would show continuous positive ornegative trends upon ATRA treatments. We identifiedtwo additional miRNAs that were upregulated, miR-147and miR-107 (Table 1). Northern blots and miRNAquantitative real time–polymerase chain reaction (qRT–PCR) confirmed the array data (Figure 2). In addition,we found similar miR-223 and let-7a expression inHL-60 cells treated with ATRA (Figure 2e).We then asked whether the results from cell lines

could be reproduced in primary APL patient blastscultured with or without ATRA in vitro. As shownin Figure 3, we found similar miRNA expression in

a

11%

M1

M2

untreated 11b again.028

FL3-H

020

40

Cou

nts

6080

100

NB4 Untreated CD11

100 101 102 103 104

NB4 ATRA CD11

76%

M1

M2

Atra 11b.024

FL3-H

020

40

Cou

nts

6080

100

100 101 102 103 104

FL1-H

ATRACD 15.022

M1

M2

040

80Cou

nts

120

160

200

93%

NB4 ATRA CD15

100 101 102 103 104

b

9%

100 101 102 103 104

M1

M2

untreated CD 15.025

FL2-H

040

80Cou

nts

120

160

200

NB4 Untreated CD15

Figure 1 Morphological and flow cytometry analysis of differentiated NB4 cells upon ATRA treatment. (a) Morphology ofdifferentiated NB4 cells. May-Grunwald Giemsa stain of cytospin preparations of NB4 cells after ATRA treatment for 96 h. (b) Flowcytometry analysis of differentiated NB4 cells. As shown in Figure 1b, a substantial increase in the staining for anti-human CD11b andanti-human CD15 antibodies were observed in ATRA-treated NB4 cells at 96 h.

MicroRNA expression in ATRA treated APL cellsR Garzon et al

4149

Oncogene

primary blast cells from three APL patients uponATRA treatment, except for miR-181b (data notshown).

MiR-107 modulates NFI-AA recent study showed that two transcription factors,NFI-A and C/EBPa compete for binding to thepromoter of miR-223. Although NFI-A maintains miR-223 at low levels, its replacement by C/EBPa followingRA differentiation, upregulates miR-223 (Fazi et al.,2005). Consistent with these data, we confirmed thatmiR-223 is induced significantly upon treatment withATRA. Interestingly, we found that miR-107, which ispredicted to target NFI-A, (Figure 4a) was upregulatedas well after ATRA treatment (Figure 2d). To confirm adirect interaction between these two genes, we clonedthe NFI-A 30 untranslated region (UTRs) predicted tointeract with miR-107 into a pGL3 luciferase reportervector. When the miR-107 precursor oligonucleotidewas transfected into the NB4 cells along with thereporter plasmid containing the NFI-A 30UTR sequenceby nucleoporation, a 40% reduction of normalizedluciferase activity was observed compared with thecontrol scrambled oligonucleotide (Figure 4b).We alsoperformed a control experiment using a mutated NFI-Aconstruct lacking four bases in the miR-107 seed match.As expected the deletion completely abolished theinteraction between miR-107 and NFI-A (Figure 4b).

Regulation of the cluster miR-15a/miR-16-1 andBcl-2 by ATRAWe then asked whether there is a negative correlationbetween miR-15a and miR-16-1 expression and Bcl-2during ATRA-induced granulocytic differentiation ofNB4 cells. The expression of Bcl-2 mRNA and protein

were analysed during ATRA treatment and correlatedwith qRT–PCR expression of miR-15a and miR-16-1,which are known regulators of this protein (Cimminoet al., 2005). As shown in Figure 5a and b, both mRNAand protein levels were downregulated upon ATRAtreatment. In sharp contrast, the expressions of miR-15aand miR-16-1 were increased (Table 1). It has beenreported that ATRA induced downmodulation of Bcl-2mRNA and protein expression in several myeloid celllines, including APL cell lines (Bruel et al., 1995).However, there is no clear evidence about the mechan-ism of this downregulation. To confirm the in vivointeraction between the miR-15a/miR-16-1 cluster andBcl-2 in NB4 cells, we transfected the precursor miR-16-1 (pre-miR-16-1) oligonucleotide or scrambled oligonu-cleotides into NB4 cells by using nucleoporation andobtained total RNA at 24 h and proteins from whole celllysates at 48 h. As shown in Figure 5c, there was asignificant reduction of the Bcl-2 transcripts by qRT–PCR after transfection with pre-miR-16-1. Likewise,transfection of the pre-miR-16-1, but not the control,resulted in a strong Bcl-2 protein reduction by Westernblots at 48 h (Figure 5e).

ATRA modulation of RAS is inversed to let-7 expressionRAS proteins family has been established as the keyregulators of diverse cellular processes (Kolch, 2005).Activating mutations have been described in AML(Bowen et al., 2005) and, recently, it has been demon-strated that K-ras cooperates with PML-RARa toinduce APL like disease in a mouse model with highpenetrance and short latency (Chan et al., 2006). Thelet-7 miRNA family have been shown to regulate theRAS oncogenes through post-transcriptional repression(Johnson et al., 2005). Our above data indicated that

Table 1 MicroRNAs differentially expressed during ATRA treatment of NB4 cells

MicroRNA Fold change Sam scorea FDRb Putative targetsc

24 h 48 h 72 h 96 h

Upregulatedhsa-let-7a-3 1.87 1.55 1.56 2.3 4.39 0.01 RAS, MYCNhsa-mir-16-1 1.4 1.59 1.54 2.13 3.31 0.01 BCL2, WNT1hsa-mir-223 1.27 3.17 3.91 4.17 3.12 0.01 NFIA, HEXhsa-mir-15a 1.19 2.06 2.26 2.61 2.91 0.05 BCL2, WNT1hsa-mir-15b 1 10.02 11.05 6.55 2.66 0.09 BCL2, WNT1hsa-let-7c 2.13 1.51 1.21 1.48 2.49 0.09 RAS, MYCNhsa-let-7d 7.05 4.03 9.08 7.34 2.43 0.09 RAS, MYCNhsa-mir-342 0.94 3.59 5.42 3.87 2.41 0.09 MEIS1hsa-mir-107d 1.59 2.33 3.34 1.85 1.28 0.24 NFIAhsa-mir-147d 0 2.66 9.72 0.37 1.59 0 DNMT3A

Downregulatedhsa-mir-181b 0.48 0.77 0.99 0.52 �3.91 0.07 FOS, RUNX1

Abbreviations: ATRA, all-trans-retinoic acid; FDR, false discovery rate; Sam, significant analysis of microarrays. Four pair of samples wereanalysed for miRNA expression after 96 h of culture with or without ATRA. Raw data were normalized after background subtraction using globalmedian of the intensities. In this table, the results are presented as fold change of miRNA expression with respect to its control (untreated cells)at each time point (paired analysis). aGenes differentially expressed between untreated and ATRA-treated cells were identified using SAM, paired2-class analysis. bThe percentage of such genes identified by chance is the FDR. cTarget prediction according to Targetscan (www.genes.mit.edu/targetsacn) (Lewis et al., 2003), Miranda (www.MicroRNA.org/Miranda) (John et al., 2004) and Pictar (www.pictar.bio.nyu.edu) software (Kreket al., 2005). dMiRNAs selected by using quantitative SAM analysis.

MicroRNA expression in ATRA treated APL cellsR Garzon et al

4150

Oncogene

several let-7 members are upregulated by ATRA. Thus,we investigate the expression of RAS mRNA andprotein in ATRA-treated NB4 cells. As shown inFigure 5a and b, RAS transcript levels by qRT–PCRwere rather increased, whereas there was a decrease inthe RAS protein levels at 48h, indicating a post-transcrip-tional regulation mechanism, which is typical to that ofmiRNAs. Moreover, the levels of RAS protein wereinversed of that of let-7 family members, suggesting thatlet-7 could mediate ATRA downmodulation of RAS inNB4 cells (Table 1). Furthermore, we transfected theprecursor let-7a-3 or let-7d or scrambled oligonucleo-tides in NB4 cells by using nucleoporation. The ectopicexpression of let-7a-3 or let-7d in NB4 cells resulted inno change in the RAS mRNA levels, but a decrease ofthe RAS protein at 48 h, compared with the controltransfected cells (Figure 5d and e). These results suggestthat the modulation of RAS by ATRA could bemediated by let-7 family members.

ATRA treatment recruits nuclear factor-kappa B onthe let-7a-3/let-7b cluster promoterNuclear factor-kappa B (NF-kB) proteins are the keyregulators of various biological processes, includingcell proliferation and differentiation in many systems(Mathieu et al., 2005). Previous studies have shown theactivation of NF-kB upon ATRA-induced maturationof APL (Mathieu et al., 2005) and enrichment for NF-kB binding sites in RA target genes has been described(Meani et al., 2005). To investigate whether ATRA-induced NF-kB activation regulates miRNAs, we co-transfected NB4 cells with a potent NF-kB inhibitorexpression plasmid (IkBa-SR) or empty vector andtreated the cells with ATRA or dimethylsulphoxide(DMSO) (control). We then measured the maturemiRNA expression of all the differentially expressedmiRNAs identified by the microarrays analysis. Asshown in Figure 6a, the ATRA-induced upregulation oflet-7a was blocked by the inhibition of the NF-kB

b

7.2 7.1 8.3

0

Microarrayintensity

miR-15a0.5

0.4

0.3

0.2

0.1Line

ar e

xpre

ssio

n

U24 A24 U48 A48 U72 A72

7.5 7.0 8.1

8.3

c

0

0.2

0.4

0.6

0.8

1

1.2

Line

ar E

xpre

ssio

n

Microarrayintensity

Let-7a3

U24 A24 U48 A48 U72 A72

7..58.37.78.17.2

d

0

1

2

3

4

5

6

7

Line

ar E

xpre

ssio

n

12.1 12.1 13.9

U24 A24 U48 A48 U72 A72

Microarrayintensity

11.6 11.8 13.2

miR-10724 hs

miR-223

aU

48 hs 72 hs

ATRA U ATRA U ATRA

Microarrayintensity

12.4 12.7

U6

12.5 14.2 12.0 13.9

ATRA

U6

miR-223

Let-7-a3

eUntreated ATRA Untreated

24 hs 48 hs

Figure 2 Microarray data confirmation. (a) Northern blots were performed as described previously (Liu et al., 2004) using probeswith perfect complementarities to the mature sequence of miR-223 in NB4 cells. Loading control was performed with U6. Immediatelybelow the Northern blotting there is a corresponding normalized miRNA signal intensity (2logged) for each time point obtained by themiRNA microarrays. QRT–PCR of miR-15a (b), let-7a (c) and miR-107 (d) during granulocytic differentiation of NB4 cells induced byATRA. The miRNA expression is shown in untreated (U) and ATRA-treated cells (A) during the time course after 2(�D-Ct) calculationswith 18S normalization. Below each bar there is the normalized miRNA signal intensity (2 logged) for each sample obtained bymicroarrays. (e) Representative Northern blots for miR-223 and let-7a in HL-60 cells untreated or treated with ATRA for 48 h. Thetreatment of HL-60 was performed as described previously in the text for NB4 cells. Loading control was performed with U6.

MicroRNA expression in ATRA treated APL cellsR Garzon et al

4151

Oncogene

pathway, suggesting that this miRNA is regulated byNF-kB. Because let-7a-3 is located at chromosome22q13 in cluster with let-7b, we measured the maturelet-7b transcripts levels by qRT–PCR in NB4 cells aftertreatment with ATRA and the IkBa-SR and similarresults were found (Figure 6a). However, the inhibitionof NF-kB pathway did not block the ATRA-inducedupregulation of other miRNAs (data not shown).Successful NF-kB pathway inhibition was confirmedby decreased NF-kB reporter luciferase activity(Figure 6b).Having shown that the mature forms of let-7a and

let-7b are transcriptionally regulated by NF-kB, we thensearched for NF-kB binding sites, 5000 bp upstreamfrom the precursor sequence of let-7a-3/let-7b clusterby using a positional weight matrix for NF-kB andMATCH (Kel et al., 2003) program. We identified 31putative NF-kB-binding sites (data not shown). Toinvestigate whether NF-kB is binding to any of thesepredicted sites, we performed chromatin immunopreci-pitation (ChiP) experiments in six of the 31 predictedsites. As shown in Figure 6c and d, we found that theNF-kB (p50) subunit is associated strongly with oneof the predicted sites at �833 nucleotides from the50 precursor let-7a-3 upon ATRA treatment. The bindingof NF-kB (p65) subunit was less convincing for thissite (data not shown). Further ChiPs experimentsusing c-Rel and Rel-B antibodies were performed usingthe oligonucleotides corresponding to this site.However, no binding was detected. Likewise, therewas no binding for the p50 or p65 subunits in theother five predicted sites that were tested (data notshown).

NF-kB activates the transcription of Let-7a-3 and let-7bTo confirm that the p50 binding to the NF-kB consensussite in the let-7a-3/let 7b promoter is associated withtranscriptional activation of this cluster, we cloned thegenomic region of the let-7a-3/let-7b promoter contain-ing this site upstream to the luciferase gene in the pGL3enhancer vector. Next, we co-transfected this reporterwith a NF-kB (p50) mammalian expression vector(pCMV2) or empty vector into K562 cells by nucleo-poration. As shown in Figure 7a, the luciferase/renillaratios were increased sixfold compared with the emptyvector. The increased luciferase activity was significantlyreduced, when a mutated pGL3-let-7a-3/let-7b reportervector (lacking three bases in the consensus NF-kBbinding site) was used. Consistent with the ChiP resultsfor this site, co-transfection of the NF-kB (p65) and thePGL3-let-7a-3/let-7b promoter reporter in K562 cells

0

5

10

15

20

25

30

35

40

Fol

d C

hang

e

APL patients

miR

-15a

miR

-16-

1

miR

-223

let-7

-a

let-7

d

miR

-107

Figure 3 MiRNAs expression in APL patient blasts by qRT–PCR. Total RNA was obtained from three APL patient blaststreated with ATRA or DMSO (control) for 48 h. qRT–PCR oflet-7d, let-7a-3, miR-223, miR107, miR-15a and miR-16-1 wasperformed as described. Relative expression was calculated usingthe comparative Ct method. The results are presented as foldchange in the miRNA expression between ATRA-treated cells withrespect to the control group (DMSO).

a

b

0

1

NF1-A +Scrambled

NF1-A + miR-107

NF1-Amut+miR-107

Rel

ativ

e lu

cife

rase

act

ivity

1.4

1.2

0.8

0.6

0.4

0.2

Figure 4 Hsa-miR-107 targets NFI-A. (a) The interaction betweenmiR-107 (green) with the 30UTR of NFI-A (red) is shown aspredicted by PICTAR (Krek et al., 2005). (b) miR-107 interact withNFI-A. The NFI-A 30-UTR segments containing the target site formiR-107 was cloned into the pGL3 control vector. The results areshown as a relative change of the miR-107þpGL3–NFI-A andmutated pGL3-NFI-A luciferase/renilla activity ratios with respectto that of the negative control (scrambled RNA) transfected cells.

MicroRNA expression in ATRA treated APL cellsR Garzon et al

4152

Oncogene

did not reveal increased luciferase activity. We per-formed further confirmatory experiments in NB4 cellsco-transfected transiently with the PGL3-let-7a-3/let-7breporter and the IkBa-SR construct (potent and specificinhibitor of NF-kB) or empty vector. As expected, theincreased activity of luciferase induced by ATRA wasblocked, when the NF-kB pathway was inhibited by theIkBa-SR, which is a potent and specific NF-kB inhibitor(Figure 7b). To discover a potential role for NF-kB(p65) regulation of the let-7a/let-7b cluster, we transfectNB4 cells with antisense oligonucleotides to the NF-kB(p50) and (p65) subunits, each alone and both combined,and treated the cells with ATRA for 48 h. As controls,we used scrambled oligonucleotides. We obtained totalRNA at 48 h and performed qRT–PCR for miR-let-7a.As shown in Figure 7c, the inhibition of NF-kB (p65) or/

and (p50), resulted in decrease levels of let-7a comparedwith the controls, suggesting that additional NF-kBbinding sites for p65 may exist.

Discussion

In the present study, we characterized the expression ofmiRNAs during ATRA-induced granulocytic differen-tiation of APL patients and cell lines. Using a miRNAmicroarray platform, we identified a small number ofmiRNAs upregulated upon ATRA treatment of APLcell lines. More importantly, these results were con-sistent with the profiling of APL patients blastsafter ATRA treatment by using a different platform

a

0

1

2

3

4

5

6

Line

ar E

xpre

ssio

n

U24 A24 U48 A48

BCL-2

0

1

2

3

4

5

6

Line

ar E

xpre

ssio

n

U24 A24 U48 A48

RAS

BCL-2

b U24 A24 U48 A48

GAPDH

RAS

eBCL-2

�-actin

RAS

GAPDH

Scrambled Let-7-a3

miR-16-1Scrambled

c

0

Line

ar E

xpre

ssio

n

BCL-20.25

0.2

0.15

0.1

0.05

Scrambled miR-16

d

0

1

Let-7a

Line

ar E

xpre

ssio

n

RAS

Scrambled

0.80.60.40.2

1.81.61.41.2

Figure 5 RAS and Bcl-2 downmodulation by ATRA. (a) Quantitative expression of Bcl-2 and RAS mRNA levels by TaqMan real-time PCR in untreated and ATRA-treated NB4 cells after 2(�D-Ct) calculations normalized with 18S. (b) Representative Western blotanalysis of Bcl-2 and Ras proteins during ATRA treatment of NB4 cells. Protein loading control was done with GAPDH (Santa Cruz).(c) Quantitative expression of Bcl-2 mRNA levels in NB4 cells 24 h after transfection with pre-miR-16-1 oligonucleotides (Ambion) orscrambled oligonucleotides (negative control-1). (d) Quantitative expression of RAS mRNA levels in NB4 cells 24 h after transfectionwith pre-let-7a oligonucleotides (Ambion) or scrambled oligonucleotides (negative control-1). (e) Western blot analysis of Bcl-2 andRAS in whole cell extracts obtained at 48 h from NB4 cells transfected with pre-miR-16-1 or pre-let-7a-3 oligonucleotides. Loadingcontrol was performed with b-actin and GAPDH (Santa Cruz).

MicroRNA expression in ATRA treated APL cellsR Garzon et al

4153

Oncogene

(qRT–PCR). Among the upregulated miRNAs, let-7a-3is located at 22q13, about 860 bp upstream from let-7b.Although these two miRNAs are in a cluster and arelikely transcribed together, let-7b was not identified bythe microarray as differentially expressed after ATRAtreatment. One possible explanation is that cross-hybridization with let-7c has occurred owing to thesimilarities between these two mature miRNAs (theydiffer by only one nucleotide). Nevertheless, upregula-tion of let-7b after ATRA treatment of NB4 cells wasconfirmed by qRT–PCR.The present data indicate that many miRNAs with

strong experimental data supporting a role for tumorsuppressors such as miR-15a, miR-16-1 and several let-7family members are upregulated by ATRA, whereas theexpression of their targets (Bcl-2 and RAS) are down-regulated. The ectopic expression of miR-15a and miR-16-1 in NB4 cells induced downregulation of both Bcl-2mRNA and protein. Thus, although we previouslydescribed that miR-15a and miR-16-1 target Bcl-2 at

the translational level in MEG01 leukemic cells (Cim-mino et al., 2005), it is possible that these two miRNAsmight affect both transcription and translation of Bcl-2or could act indirectly through the modulation of othergenes that modulates BCL-2 expression in the system wediscussed here.We have demonstrated that ATRA negatively modu-

lates RAS by a post-translational mechanism. Thistype of regulation is typical for miRNAs. Furthermore,the enforced expression of let-7a-3 recapitulates theATRA effects on RAS mRNA and protein expression.We have shown that a small number of miRNAs are

induced upon ATRA treatment. However, the mechan-ism by which ATRA regulates miRNA expression islargely uncharacterized. We hypothesize that miRNAsexpression could be regulated in part by ATRA-modulated transcription factors like NF-kB. In fact,we demonstrated that NF-kB (p50/p50), but not p65/p50binds to the upstream sequence of the let-7a-3/let7bcluster and activates its transcription. Previous work by

Fol

d C

hang

e

0

1

2

3

4

a

b

Untreated

ATRA

Untreated

ATRA

• • •

c

d

Let-7a3

–833

Oligo 1

Let-7b22q13

0

0.2

0.4

0.6

0.8

1

1.2

ATRA +Scrambled

ATRA +IkBα-SR

Untreated ATRA ATRA + NF-�B inhibitor4.5

3.5

2.5

1.5

0.5

let-7a let-7b miR-15a

Rel

ativ

e lu

cife

rase

activ

ity

Oligo 2(control)

Oligo 1(NF-�B)

Inpu

t

Rabbit

IgG

.Ant

i-NF-�

B

No Ab

ggggagcccc (NF-�B-binding site)

Figure 6 ATRA treatment recruits NF-kB on the let-7a3/let 7b cluster promoter. (a) Representative qRT–PCR of let-7a, let-7b andmiR-15a from three different experiments was performed with total RNA obtained from NB4 cells treated with DMSO (control),ATRA plus empty vector (control 2) and ATRA plus IkBa-SR. The results are shown as relative fold change in the miRNA expressionwith respect to the control (untreated) NB4 cells miRNA levels, after normalization with 18 s and 2(�D-Ct) calculations as describedpreviously (Chen et al., 2004). (b) Confirmation of successful NF-kB inhibition. To confirm successful inhibition of the NF-kBpathway by IkBa-SR, we co-transfected 5� 106NB4 cells with 5mg of NF-kB activity reporter, 0.5 mg of renilla reporter pRL-TK(Promega) and 5mg of IkBa-SR or empty vector by using nucleoporation. The cells were then treated with ATRA for 24 h. Renilla andluciferase activities were measured consecutively using the dual luciferase assays (Promega). The results are shown as relative change ofthe dominant-negative IkBa-SR luciferase/Renilla ratios with respect to the empty vector-treated cells (calibrator). (c) NF-kBconsensus element in the promoter of let-7a-3 and let-7b. Putative NF-kB site in the let-7a-3 and let-7b promoter identified by using apositional weight matrix and MATCH (Kel et al., 2003) program. In this figure, the genomic region that was amplified by the primers(oligos 1) to perform the ChiP experiments is shown. (d) NF-kB (P50) binds to the promoter of let-7a-3/let-7b. In this ChiP experiment,performed as described in detail in Materials and methods, there is a clear binding of NF-kB (P50) to the predicted NF-kB-bindingsite.

MicroRNA expression in ATRA treated APL cellsR Garzon et al

4154

Oncogene

other groups indicates that p50/p50 homodimers aregenerally repressors of transcription (Mori et al., 1999;Udalova et al., 2000). However, there are several reportsthat support the ability of p50/p50 homodimers toactivate transcription (Dechend et al., 1999; Kurlandet al., 2001; Priel et al., 2005). On the basis of ourexperiments with antisense oligonucleotides against p65and p50, it is possible that p65/p50 heterodimers mayalso bind to a yet to be identified site of the let-7a-3/let7-cluster and activate its transcription.In summary, our results demonstrate that ATRA

treatment of APL patients and cell lines modulate a smallnumber of miRNAs, most of which have confirmedtargets involved in hematopoietic differentiation and

apoptosis. Interestingly, we have confirmed that alongwith miR-223, miR-107 negatively regulates NFI-A. Inaddition, we demonstrated that ATRA induced NF-kBbinds to the promoter of let-7a-3 and activates itstranscription. Finally, we showed that ATRA down-regulation of RAS and Bcl-2 correlated with theactivation of two confirmed miRNA regulators of thesegenes, let-7a- and miR-15a/miR-16-1, respectively.

Materials and methods

Patients, cell lines and ATRA treatmentThe human promyelocytic cell line NB4, was a gift of GuidoMarcucci. The NB4 cells were maintained in Rosewell’s ParkMemorial Institute media (RPMI) with 20% fetal bovineserum (FBS) and antibiotics. The cell line HL-60 was obtainedfrom American Type Culture Collection (Manassas, VA,USA) and maintained in Dulbecco’s modified Eagle’s mediumwith 20% FBS and glutamine. NB4 and HL-60 cells weretreated with 100 nM of ATRA (SIGMA, St Louis, MO, USA)for 4 days. Bone marrow (2) and peripheral blood (1) leukemicblood samples were collected after informed consent fromthree patients with newly diagnosed APL, prepared by Ficoll-Hypaque (Nygaard) gradient centrifugation and cryopre-served. After thawing and counting the cells at 371C,5� 105 cells were cultured in AVIM media without serum(Invitrogen, Carlsbad, CA, USA) with ATRA 100 nM orwithout ATRA. Total RNA was extracted using Trizol(Invitrogen) only after 48 h of culture because of the reducednumber of initial cells and the lack of sustained growth overtime.

0

1

2

3

4

5

6

ATRA + ATRA +p50 siRNA

ATRA +p65 siRNA

ATRA +

Fol

d C

hang

e

c

scrambled p65/p50siRNA

a

0

1

2

3

4

5

6

7

8

9

10

WT let-7+ EV

WT let-7+ p65

WT let-7+ p50

Mutated Mutated

Rel

ativ

e lu

cife

rase

act

ivity

let7+ EV let7+ p50

0

1

2

3

4

5

6

7

8

9

Untreated ATRA +EV

Rel

ativ

e Lu

cife

rase

act

ivity

b

ATRA ATRA +IkBα-SR

Figure 7 NF-kB (p50) is able to activate the promoter of let-7a-3/let-7b. (a) K562 cells (5� 106) were co-transfected in 10ml platesusing nucleoporation (AMAXA) with 5 mg of the let-7a-3/let-7breporter (WT let-7 in the Figure), 0.5mg of renilla pRL-TK(Promega) and 5mg of pCMV2-p50 or pCMV2-p65 vectors orempty vector (EV). Firefly and Renilla luciferase activities weremeasured consecutively using the dual luciferase assays (Promega)24 h after transfection. The results are expressed as relativeluciferase/renilla ratios between p50 and p65 expression plasmidstransfected cells with respect to the empty vector (control). Weperformed an additional control experiment using a mutatedNF-kB let-7a-3/let-7b reporter. K562 cells were nucleoporated aspreviously described with 5mg of the mutated let-7a-3/let-7breporter and 5mg of pCMV2-p50 or empty vector (EV). Asexpected the mutation abolished the interaction with NF-kB (p50).(b) Let-7a-3/let-7b promoter characterization NB4 (5� 106) cellswere transfected with 5mg of the pGL-3 let-7a-3/let-7b promoterreporter plus 0.5mg of Renilla pRL-TK by using nucleoporationas described. Half of the cells were treated with ATRA or leftuntreated (DMSO). In addition, we co-transfected NB4 cells with5mg of IkBa-SR or EV plus 5mg of pGL3-let-7a-3/le-t7b reporterand 0.5 mg of Renilla pRL-TK and treated the cells with ATRA for24 h. Renilla luciferase activities were measured consecutively usingthe dual luciferase assays (Promega) 24 h after transfection. Thedata are presented as relative change of the luciferase/Renillaactivity ratio between ATRA-treated cells transfected with pGl-3let-7a-3/let-7b reporter (pGL3 let-7 ATRA) alone or IkBa-SR orempty vector with respect to pGL-3 let-7a-3/let-7b untreated(DMSO) cells (c). Interfering with p50 or p65 activity block theupregulation of let-7a induced by ATRA. QRT–PCR of let-7a inNB4 cells transfected with NF-kB (p50) or/and NF-kB (p65) smallinterfering RNAs (Santa Cruz) and non-specific Control Pool(negative control) and subsequently treated with ATRA or DMSO.The results are reported as fold change in the miRNA expressionfrom ATRA-treated samples with respect to that of untreated cells.

MicroRNA expression in ATRA treated APL cellsR Garzon et al

4155

Oncogene

Granulocytic differentiation and characterizationCytospin preparations of NB4 cells treated with ATRA oruntreated were performed and stained with May–GrunwaldGiemsa at different time points during the granulocyticdifferentiation induction. Flow cytometry analysis of differen-tiated NB4 cells was performed using peridine chlorophyllprotein conjugated mouse antihuman CD11b and phycoery-thrin conjugated mouse antihuman CD15 antibodies with theirrespective isotypes (BD Pharmingen, San Jose, CA, USA). Forsingle staining, 1 million of cells were resuspended in 20 ml ofthe previously indicated antibodies and were incubated for60min at 41 in the dark. After two washes with phosphate-buffered saline (PBS) the cells were re-suspended in 0.5ml of3% FBS-PBS and analysed by FACScan (BD) using cellquestsoftware (BD).

RNA extraction, Northern blots and microarray experimentsTotal RNA for microarray studies was obtained using Trizolreagent (Invitrogen) from NB4 cells after 24, 48, 72 and 96 h ofculture with or without ATRA-containing medium. Labeledtargets from 5mg of total RNA were used for hybridizationon each miRNA microarray chip-containing 368 probes intriplicate, corresponding to 245 human and mouse miRNAgenes (Liu et al., 2004). Northern blots were performed asdescribed in detail elsewhere (Liu et al., 2004).

Data analysisExpression data were normalized after background subtrac-tion using global median normalization of the bioconductorpackage (www.bioconductor.org). In two class comparisons(i.e., untreated vs treated), differentially expressed miRNAswere identified by using the t-test procedure within the SAM(http://wwwstat.stanford.edu/Btibs/SAM/index.html)(Tusher et al., 2001). SAM identifies genes with statisticallysignificant changes in expression by assimilating a set of gene-specific scores (i.e., paired t-tests). Each gene is assigned ascore on the basis of its change in gene expression relative tothe standard deviation of repeated measurements for thatgene. Genes with scores greater than a threshold are deemedpotentially significant. The percentage of such genes identifiedby chance is the false discovery rate (FDR) (q-value). Toestimate the FDR, nonsense genes are identified by anszingpermutations of the measurements. The threshold can beadjusted to identify smaller or larger sets of genes, and FDRare calculated for each set. We also used SAM quantitativeanalysis, which computes the linear regression coefficient ofeach gene on the outcome.

Western blotsTotal cell extracts (50 mg) from untreated or treated NB4 cellsafter 48 h of culture were extracted using radioimmuno-precipitation assay buffer (Sigma, St Louis, MO, USA) andfractionated by electrophoresis on 10% sodium dodecylsulfate–polyacrylamide gel, electroblotted to polyvinylidenedifluoride membranes (BIORAD, Hercules, CA, USA) andreacted with anti-BCL-2 (Dako, Carpinteria, CA, USA) orRAS (ABCAM, Cambridge, MA, USA). Protein loadingcontrol was done with GAPDH (Santa Cruz, Santa Cruz, CA,USA). Appropriate secondary antibodies were used (SantaCruz).

ChiP experimentsNB4 cells were seeded at 1� 106 cells/ml and were grown for48 h with or without 1mM RA. After 48 h of treatment, the cellswere cross-linked by 1% formaldehyde for 10min at roomtemperature. Cell extracts were prepared and sonicated to

generate 200–1000 bp DNA fragments. DNA–chromatincomplex was immunoprecipitated with anti-p50, -p65, c-Rel,Rel-B antibody (Upstate, Chicago, IL, USA) and with normalrabbit IgG (Santa Cruz) used as internal control. The DNA–protein crosslinks were reversed by heating at 651C for 4 h. Therecovered DNA was PCR-amplified with the oligo 1 (forward(For) 50-TTT CCC CAG GAA GGT GGT AG-30; reverse(Rev) 50-GTC CAG TGC CTG TCC CTG CG-30) corre-sponding at NF-kB consensus binding site at �833 and withthe negative control oligo corresponding at KIAA66 (Avivasystem negative control primers). Control amplification wascarried out on input chromatin before immunoprecipitation(column input) and on mock-immunoprecipitated chromatin(column No-Ab).

Luciferase gene reporter assaysThe NFI-A 30UTR segments containing the target site for miR-107 was amplified by PCR from genomic DNA and insertedinto the pGL3 control vector (Promega, WI, USA), using theXbaI site immediately downstream from the stop codon ofluciferase. We used the following primers; For: 50 TAC TCTAGA CAA TTC CGA AAA TGG CAA AC 30 and Rev: 50

CAT TCT AGA CCA ATG AAA TTG CAC TTT ACT CA30. We also generated an insert with deletion of four bases fromthe site of perfect complementarity using the Quiagen XL-sitedirected mutagenesis kit (Stratagene, La Jolla, CA, USA).Wild-type and mutant inserts were confirmed by sequencing.The human NB4 cell line was grown in 10% FBS in RPMI-1640 at 371C in a humidified atmosphere of 5% CO2. The cellswere co-transfected in 10ml plates using nucleoporation withkit V (Amaxa, Gaithersburg, MD, USA) according to themanufacturer’s protocol using 5mg of the firefly luciferasereport vector and 0.5mg of the control vector containingRenilla luciferase, pRL–TK (Promega). For each plate, thepre-miR-107 oligonucleotides or the scrambled RNA (Ambion,Austin, TX, USA) were co-transfected. Firefly and Renillaluciferase activities were measured consecutively using the dualluciferase assays (Promega) 24 h after transfection. Theupstream sequence of the Let-7a-3/let7-b cluster containingthe NF-kB consensus site (-833) was amplified by PCR fromgenomic DNA and inserted into the pGL3 enhancer vector(Promega, WI, USA), using the XhoI and BglII sites upstreamfrom luciferase sequence. We used the following primers: For50CTT TCC CAG GAA GGT AG 30; Rev 50 CGC AGG GACAGG CAC TGG AC 30. A mutated construct, lacking threebases and with a mutation A-T in the NF-kB consensus site,was also generated by using the Quiagen XL-site directedmutagenesis kit (Stratagene). Wild-type and mutant insertwere confirmed by sequencing.

Small interfering RNA experimentsNB4 cells were seeded at 5� 106 and transfected with 5mg ofNF-kB (p50) or/and NF-kB (p65) small interfering RNAspurchased from Santa Cruz (Cat# SC_29410 and SC_29407,respectively), and non-specific control pool (negative controlCat# SC 37007) by using nucleoporation (Amaxa) kit V. TotalRNA were extracted by TRIZOL (Invitrogen) after 48 h oftreatment with ATRA. Mature let-7a expression was analysedby qRT–PCR as described previously. (Chen et al., 2004).Untreated cells transfected with negative control oligonucleo-tides were used as a calibrator.

Acknowledgements

This work was supported by National Cancer Institute/National Institutes of Health Grants, P01CA76259,

MicroRNA expression in ATRA treated APL cellsR Garzon et al

4156

Oncogene

P01CA16058 and P01CA81534 (CMC), P01 CA055164 andthe Paul and Mary Haas Chair in Genetics (MA), LauriStrauss Discovery grants awards (RG), Kimmel Foundationgrants awards (GAC and RA) and CLL Global Research

Foundation Grant (GAC). We thank Guido Marcucci forkindly providing us with NB4 cells and Dennis Guttridge forhis generous support with many reagents used to characterizethe NF-kB regulated miRNAs (The Ohio State University).

References

Bartel DP. (2004). MicroRNAs: genomics, biogenesis, mechan-ism, and function. Cell 116: 281–297.Bowen DT, FrewME, Hills R, Gale RE, Weathley K, Groves Met al. (2005). RAS mutation in acute myeloid leukemia isassociated with distinct cytogenetic subgroups but does notinfluence outcome in patients younger than 60 years. Blood106: 2113–2119.Bruel A, Benoit G, De Nay D, Brown S, Lanotte M. (1995).Distinct apoptotic responses in maturation sensitive andresistant t (15; 17) acute promyelocytic leukemia NB4 cells.9cis retinoic acid induces apoptosis independent of matura-tion and Bcl-2 expression. Leukemia 9: 1173–1184.Chan IT, Kutok JL, Williams IR, Cohen S, Moore S,Shigematsu H et al. (2006). Oncogenic K-ras cooperateswith PML-RAR alpha to induce an acute promyelocyticleukemia-like disease. Blood 108: 1708–1715.Chen CZ, Li L, Lodish HF, Bartel DP. (2004). MicroRNAsmodulate hematopoietic lineage differentiation. Science 303:83–86.Cheng AM, Byrom MW, Shelton J, Ford LP. (2005).Antisense inhibition of human miRNAs and indications foran involvement of miRNA in cell growth and apoptosis.Nucleic Acids Res 33: 1290–1297.Cimmino A, Calin GA, Muller F, Iorio M, Ferracin M,Shimitzu M et al. (2005). Mir-15a and miR-16-1 inducedapoptosis by targeting BCL2. Proc Natl Acad Sci USA 102:13944–13949.Dechend R, Hirano F, Lehmann K, Heissmayer V, Ansieau S,Wulczyn F et al. (1999). The Bcl-3 oncoprotein acts as abridging factor between NF-kB/Rel and nuclear co-regula-tors. Oncogene 18: 3316–3323.Dresios J, Aschrafi A, Owens GC, Vanderkilsh PW, EdelmanGM,Mauro VP. (2005). Cold stress induced protein Rbm3 binds60S ribosomal subunits, alters microRNA levels, andenhances global protein synthesis. Proc Natl Acad Sci USA102: 1865–1870.Fazi F, Rosa A, Fatica A, Gelmetti V, Marchis ML, Nervi Cet al. (2005). A mini-circuitry comprising microRNA-223and transcription factors NFI-A and C/EBPa regulateshuman granulopoiesis. Cell 123: 819–831.John B, Enright J, Aravin A, Tuschl T, Sander C, Marks DS.(2004). Human microRNA targets. Plos biology 11: e363.Johnson SM, Grosshands H, Shingara J, Byrom M, Jarvis R,Cheng A et al. (2005). Ras is regulated by the let-7microRNA family. Cell 120: 635–647.Kel AE, Gossling E, Reuter I, Cheremushkin E, Kel-Margoulis OV, Wingender E. (2003). MATCH: a tool forsearching transcription factor binding sites in DNAsequences. Nucleic Acids Res 31: 3576–3579.

Kolch W. (2005). Coordinating ERK/MAPK signallingthrough scaffolds and inhibitors. Nat Rev Mol Cell Biol11: 827–837.Kurland JF, KodymR, StoryMD, Spurgers KB, Mc Donnell TJ,Meyn RE. (2001). NF-kappaB1 (p50) homodimers contri-bute to transcription of the bcl-2 oncogene. J Biochem Chem30: 45380–45386.Krek A, Grun D, Poy M, Wolf R, Rosenberg L, Epstein EJet al. (2005). Combinatorial microRNA target predictions.Nat Genet 37: 495–500.Lewis BP, Shih IH, Jones-Rhoades MW, Bartel P, Burge CB.(2003). Prediction of mammalian microRNA targets. Cell115: 787–798.Liu CG, Calin GA, Meloon B, Gamliel N, Sevignani C,Ferracin M et al. (2004). An oligonucleotide microchip forgenome-wide microRNA profiling in human and mousetissues. Proc Natl Acad Sci USA 101: 9740–9744.Mathieu J, Giraudier S, Lanotte M, Besancon F. (2005).Retinoic-induced activation of NF-kB in APL cells is notessential for granulocytic differentiation, but prolongs thelife span of mature cells. Oncogene 24: 7145–7155.Meani N, Minardi S, Licciulli S, Gelmetti V, Lo Coco F,Nervi C et al. (2005). Molecular signature of retinoic acidtreatment in acute promyelocytic leukemia. Oncogene 24:3358–3368.Melnick A, Licht JD. (1999). Deconstructing a disease:RARalpha, its fusion partners, and their roles in thepathogenesis of acute promyelocytic leukemia. Blood 93:3167–3215.Mori N, Fujii M, Ikeda S, Yamada Y, Tomonaga M,Ballard DW et al. (1999). Constitutive activation ofNF-kB in primary adult T-cell leukemia cells. Blood 93:2360–2368.Poy MN, Eliasson L, Krutzfeldt J, Kuwajima S, Ma X,McDonald PE et al. (2004). A pancreatic islet-specific microRNA regulates insulin secretion. Nature 432:226–230.Priel IP, Cai DH, Wang D, Kowalski J, Blackford A, Liu H.(2005). CCAAT/Enhancer binding protein a(EBPa) and C/EBPa myeloid oncoproteins induce Bcl-2 via interaction oftheir basic regions with nuclear factor-kB p50. Mol CancerReser 3: 585–596.Tusher VG, Tibshirani R, Chu G. (2001). Significance analysisof microarrays applied to the ionizing radiation response.Proc Natl Acad Sci USA 98: 5116–5121.Udalova IA, Richardson A, Denys A, Smith C, Ackerman H,Foxwell B et al. (2000). Functional consequences of apolymorphism affecting NF-kB p50-p50 binding to the TNFpromoter region. Mol Cell Biol 20: 9113–9119.

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

MicroRNA expression in ATRA treated APL cellsR Garzon et al

4157

Oncogene