ORIGINAL ARTICLE

TGF-b1 suppression of microRNA-450b-5p expression: a novelmechanism for blocking myogenic differentiation ofrhabdomyosarcomaMM Sun1,5, JF Li2,5, LL Guo3, HT Xiao4, L Dong3, F Wang3, FB Huang3, D Cao3, T Qin1, XH Yin1, JM Li3 and SL Wang3

Transforming growth factor beta 1 (TGF-b1) is the most potent inhibitor of myogenic differentiation (MyoD) of rhabdomyosarcoma(RMS); however, the underlying mechanisms of this inhibition remain unclear. In this study, we identified novel TGF-b1-relatedmicroRNAs (miRNAs); among these, miR-450b-5p is significantly regulated by TGF-b1. We provide evidence that TGF-b1 exerts itfunction by suppressing miR-450b-5p. Both in cultured cells and tumor implants, miR-450b-5p significantly arrested the growth ofRMS and promoted its MyoD. Utilizing a bioinformatics approach, we identified miR-450b-5p target mRNAs. Among thesecandidates, only the expression of ecto-NOX disulfide-thiol exchanger 2 (ENOX2) and paired box 9 (PAX9) was augmented by miR-450b-5p knockdown examined by western blot; the engineered inhibition antagonized TGF-b1-mediated differentiation inhibition.Furthermore, we found that the Smad3 and Smad4 pathways, but not Smad2, are the principal mediator of TGF-b1 suppression ofmiR-450b-5p. Taken together, these results suggest that disrupting the TGF-b1 suppression of miR-450b-5p, or knockdown ofENOX2 and PAX9, are effective approaches in inducing RMS MyoD.

Oncogene advance online publication, 13 May 2013; doi:10.1038/onc.2013.165

Keywords: rhabdomyosarcoma; TGF-b1; miR-450b-5p; differentiation

INTRODUCTIONRhabdomyosarcoma (RMS) is the most common soft tissuesarcoma in pediatric patients and young adults. The two majorhistological types of RMS tumors are alveolar RMS (ARMS) andembryonal RMS (ERMS). Although most RMS is characterized byexpression of myogenic-promoting transcription factors, such asmyogenic differentiation (MyoD) and myogenin, there is failure ofterminal muscle differentiation.1,2 Therefore, strategies aimed atstoring the myogenic program and reverse RMS cell-malignantbehavior could potentially aid in therapeutic intervention. Recentevidence has demonstrated the role of microRNAs (miRNAs) in thedevelopment and progression of RMS. Several reports have shownthat the expression of miR-1 and miR-133a is markedly decreasedin RMS cell lines compared with differentiated myoblasts andskeletal muscle tissues. In particular, re-expression of miR-1 or miR-206 restores the MyoD programs of RMS cells and blocks thetumorigenic phenotype in vivo, raising the possibility that miRNAre-expression potentially represents an effective differentiationtherapy for RMS.3 However, miRNA therapy presents severalchallenges,4,5 such as delivery limitations, instability and sideeffects. In particular, miRNA re-expression influences theexpression of hundreds of genes involved in cellular pathwaysand exhibits potentially unpredictable side effects. Thus, a moredetailed understanding of the molecular events governingmyogenesis is warranted to identify the myogenesis-relatedmiRNA biogenesis, and modulation in which cellular pathwaysparticipate.

Transforming growth factor beta 1 (TGF-b1) is known to be acritical modulator of skeletal muscle differentiation.6,7 Ourprevious report revealed that inhibition of TGF-b1 signaling inhuman RMS cell line, RD, promotes tumor cell MyoD,8–10 but theunderlying mechanism remains unclear. In recent years, severalreports have shown that TGF-b1 signaling promotes themetastasis and invasive potential of cancer cells through themodulation of biosynthesis of oncogenic miRNAs such asmiR-21,11 miR-15512 and miR-181b.13 Moreover, reports havedemonstrated that TGF-b1 signaling also has an important role inskeletal muscle differentiation via regulation of miR-24;7 a recentreport demonstrated that TGF-b regulates miR-206 and miR-29 tocontrol MyoD of the C2C12 cell line. These findings revealed anewly discovered mechanism in the TGFb1 autocrine/paracrinepathway concerning the development of malignancy; thus, thereis a need for further elucidation of TGF-b1-regulated miRNAs thathave critical roles in the regulation of RMS.

To investigate the role of TGF-b1-regulated miRNAs in theprogression of RMS, we performed comprehensive miRNAmicroarray analysis on RNA derived from typical RMS cell linesand TGF-b1 knockdown cell lines. We identified a novel set of TGF-b1-related miRNAs that included miR-450b-5p, one of the mostsignificantly expressed miRNAs; we confirmed this finding in RMStissues. As we were interested in determining whether thesemiRNAs could potentially contribute to TGF-b1-mediated differ-ential inhibition, we performed sequence complementary analysisof these miRNAs against proliferation or myogenic regulators; our

1Department of Anatomy, Soochow University School of Medicine, Suzhou, China; 2Department of Pathology, the Second Affiliated Hospital of Zhejiang University School ofMedicine, Hangzhou, China; 3Department of Pathology, Soochow University School of Medicine, Suzhou, China and 4Department of Burn and Plastic Surgery, West ChinaHospital, Sichuan University, Chengdu, China. Correspondence: Professor SL Wang or Professor JM Li, Department of Pathology, Soochow University School of Medicine, Suzhou215123, China.E-mail: wangsoly112@hotmail.com or jianmingli@suda.edu.cn5These authors contributed equally to this work.Received 4 October 2012; revised 11 March 2013; accepted 25 March 2013

Oncogene (2013), 1–12& 2013 Macmillan Publishers Limited All rights reserved 0950-9232/13

www.nature.com/onc

findings revealed that only induction of miR-450b-5p conferredMyoD in RMS cells. In this study, we provide evidence that ecto-NOX disulfide-thiol exchanger 2 (ENOX2) and paired box 9 (PAX9)are targets of miR-450b-5p, and we show that ENOX2 and PAX9mRNAs and miR-450b-5p are inversely expressed in human RMStissues. Moreover, our findings indicate that miR-450b-5p isnegatively regulated by Smad4 and Smad3 in addition toTGF-b1, and that restoration of miR-450b-5p occurs via RNAinterference of these genes.

RESULTSTGF-b1 regulates miRNA in human RMS cell linesTo identify TGF-b1-related miRNAs, we conducted comprehensivelocked nucleic acid microarray analyses to obtain miRNAexpression profiles independently in TGF-b1 knockdown cell lines(RD, SMS-CTR and RH28) and control cell lines. Figure 1a displaysstatistically significant expression changes of the miRNAs in theserepresentative RMS cells (fold change X1.5, Pp0.05). The fivemiRNAs with overexpression levels in TGF-b1 knockdown RMS celllines were: miR-4275, miR-411, miR-493* and miR-450b-5p. OnlymiR-2113 expression was significantly downregulated in TGF-b1knockdown RMS cell lines versus controls. Raw microarray data areaccessible through Gene Expression Omnibus Series accessionnumber GSE40843.

To validate the expression profiles of these miRNAs indepen-dently, we conducted real-time RT–PCR assays using total RNA

isolated from TGF-b1 knockdown RMS cell lines (ERMS: RD, RH36and SMS-CTR; ARMS: RH30, RH28 and RH3) and control cells. Ofinterest, nearly all the TGF-b1 knockdown RMS cell lines over-expressed miR-450b-5p (4/6) (Figure 1b). Among other identifiedmiRNAs by microarray analyses, miR-411-5p, miR-411-3p and miR-493-5p were significantly upregulated (Po0.05) in RNA derivedfrom TGF-b1 knockdown RMS cell lines (see SupplementaryInformation, Supplementary Figure S1).

TGF-b1 and miR-450b-5p exhibited a correlation in human RMStissuesThe above work demonstrated that differential expression ofmiRNAs in TGF-b1-knocked down RMS cell lines versus controls,thus, we hypothesized that the augmented miRNAs identified inTGF-b1 knocked down RMS cells would also be overexpressed inTGF-b1-low-expression RMS tissues compared with TGF-b1-high-expression RMS tissues. To explore this possibility, we examinedthese miRNAs by real-time RT–PCR using total RNA isolated frompaired human RMS tissues that exhibited low–high expression ofTGF-b1. Based on the 68 cases of RMS that we previouslydescribed9 and 12 cases collected recently, the immuno-histochemical-staining findings for TGF-b1 revealed that 54 ofthe 80 cases of RMS tissues exhibited TGF-b1-high expression (seeSupplementary Information, Supplementary Table S1). Amongthese miRNAs, miR-450b-5p exhibited the most significantincrease in expression (3.2-fold) in TGF-b1-low-expression RMStissues (Figure 2a), and these findings were confirmed by northernblot and RT–PCR analysis independently (Figure 2b). In addition,we explored another six paired human RMS tissues having low–high expression of TGF-b1, with RT–PCR and our findings revealingthat the majority (4/6) of TGF-b1 low-RMS tissues exhibited highermiR-450b-5p expression compared with TGF-b1 high-RMS tissues(Figure 2c). In contrast, the other miRNAs that were identified asdifferentially overexpressed in TGF-b1-knocked down RMS cells(shown in Figure 1) were identified as being overexpressed in onlya minority of TGF-b1-low RMS tissues (data not shown). Of theseTGF-b1-deregulated miRNAs, our findings suggested that miR-450b-5p acted as an important regulator, and thus was selectedfor further study.

MiR-450b-5p arrests proliferation of RMS by inducing apoptosisIn order to investigate whether TGF-b1-related miRNAs wereresponsible for RMS proliferation, we transiently transfectedmimics and inhibitors of these miRNAs into RD (ERMS) andRH28 (ARMS) cells in vitro (Figure 3). Over 90% of cells from eachcell line were transfected (data not shown), and expression oftransfected miRNAs was confirmed by real-time RT–PCR analysis(Figure 3A). Our findings revealed that [3H] thymidine incorpora-tion of RMS cells transfected with miR-450b-5p mimic wasreduced (Figure 3B). Similar results were found in optical density,as demonstrated by a time-dependent decrease in [3H] thymidineincorporation. Significant differences in proliferation wereobserved after 4 (Po0.05) and 6 (Po0.01) days. No significantdifferences in proliferation were found for other miRNAs, such asmiR-4275, miR-493-5p and miR-493-3p, or control transfectants(data not shown). Next, to determine whether growth arrest was aconsequence of RMS cell apoptosis, we utilized caspase-3 andTUNEL (terminal deoxynucleotidyl transferase dUTP nick endlabeling). As expected, the number of caspase-3- and TUNEL-positive cells increased significantly after 4 days of miR-450b-5pmimic transfection. Notably, the apoptosis induced with miR-450b-5p mimic was enhanced by TGF-b1 small interfering RNA(siRNA) treatment (Figure 3C). Apoptosis is closely related to tumorcell differentiation and progression;14 thus, this suggests that miR-450b-5p is an anti-oncogene miRNA.

0

0.5

1

1.5

2

2.5

3

RD

miR

-450

b-5p

(fol

dch

ange

/GA

PD

H)

TGF-β1 knock down Control

** ****

*

Control

RD

-6.0

4.93

3448

2.46

6724

0.0

1 2 3 5 4 6 456123

3.0240307

1.5120153

0.0miR-450b-5pmiR-4298miR-411miR-4275miR-493*miR-2113

-3.0 -1.0

SMS-CTR RH28 RDSMS-CTR RH28

TGF-β1 knock down

RH3RH28RH30SMS-CTRRH36

Figure 1. Differentially expressed miRNAs in TGF-b1-knocked downRMS cell lines versus controls. (a) Locked nucleic acid microarrayanalyses to obtain miRNA expression profiles independently in TGF-b1-knocked down RD, SMS-CTR and RH28, and control cell lines. Theheat map diagram shows the results of the two-way hierarchicalclustering of miRNAs and samples. Each row represents a miRNAand each column represents a sample. The miRNA-clustering tree isshown on the left, and the sample-clustering tree appears at the top.As shown, expression of miR-450b-5p, miR-4298, miR-411, miR-4275and miR-493* were each significantly upregulated, and miR-2113was significantly downregulated in TGF-b1-knocked down RMS celllines (lines 4, 5, 6) versus controls (lines 1, 2, 3). (b) Furtherconfirmation of miR-450b-5p with other TGF-b1-knocked down RMScell lines such as RH36, RH30 and RH3, besides above RD, SMS-CTRand RH28. RT–PCR analysis using RNA isolated from TGF-b1-knockeddown RMS cell lines and control cell lines. Each assay was conductedfor at least three independent times. Error bars indicate SD. *Po0.05;**Po0.01; ***Po0.005.

TGF-b1 suppression of miRNA-450b-5p expressionMM Sun et al

2

Oncogene (2013), 1 – 12 & 2013 Macmillan Publishers Limited

MiR-450b-5p represses tumorigenicity in vivo by promoting MyoDTo further examine whether the observed growth arrest of RMSin vitro was associated with repression in vivo, we quantifiedtumorigenicity using an in vivo nude mouse animal model. Twoweeks after tumor cell inoculation, all nude mice developeddetectable tumors that reached B0.5 cm3. Next, tumors wereinjected with mimic, inhibitors of miR-450b-5p and negativecontrol oligonucleotides in transfection solution (see Materials andmethods) once weekly for 4 weeks. There was a significantdifference between the mimic-treated group and control groupafter 3 weeks (Po0.01), and after 4 weeks the significantdifference was even greater (Po0.005) (Figure 4A). This findingsuggests that miR-450b-5p mimic inhibits the tumorigenicity ofRD cells; this was further confirmed by Ki67 expression (Figure 4B,Po0.05).

It is well understood that the biological behavior of a tumor isrelated to the degree of differentiation of its cells, and a lowerdegree of differentiation generally correlates with greater tumorgrowth. As miR-450b-5p arrests in vitro growth and in vivotumorigenicity of RMS, we examined the role of miR-450b-5p indifferentiation of RD cells. As shown in Supplementary Information,(Supplementary Figure S2), after tumors were treated withmiR-450b-5p mimic once weekly for 2 weeks, tumor cells exhibitedtypical ultrastructural features of MyoD: convoluted nucleus and an

increase in organelles, degeneration of nucleus and cytoplasm, andincreased myofilament. The tumor cell control group typicallycontained a single nucleolus and few organelles, with no obviousevidence of MyoD (data not shown). Moreover, miR-450b-5p mimicinduced an increase in staining for myogenin (Po0.01) and myosin(Po0.05) (Figures 4B and C); this was further confirmed by real-timeRT–PCR analysis. As shown in Figure 4D, there was a significantdifference in the ratio of myogenin mRNA/GAPDH (Po0.005) andmyosin mRNA/GAPDH (Po0.005) between the miR-450b-5p mimic-treated group and control groups. In a parallel experiment, weinvestigated antagonistic effect of TGF-b1 on miR-450b-5p mimic-induced RMS cell MyoD. As anticipated, TGF-b1 counteracted theexpression of myogenin and myosin, and the differentiation-inducing effect of miR-450b-5p mimic on RMS tissues.

ENOX2 and PAX9 are direct targets of miR-450b-5pSix of the predicted miR-450b-5p targets overlapped between thethree bioinformatic algorithms, including calcium/calmodulin-dependent protein kinase II inhibitor 1, microtubule-associatedprotein 2, ankyrin 3, ENOX2, PAX9 and ATPase, class VI, type 11C.Among these candidates, only ENOX2 and PAX9 were repressedby induced miR-450b-5p overexpression and augmented by itsengineered knockdown in RMS cells, as examined by RT–PCR

miRNA TGF-β1 low (n=4) TGF-β1 high (n=4) Fold change P value

miRNA-450b-5p

miRNA-411-3p

miRNA-411-5p

miRNA-4275

miRNA-493-5p

miRNA-493-3p

miRNA-4298

miRNA-2113

1.90±0.12

1.48±0.29

1.77±0.20

1.80±0.57

1.50±0.35

1.33±0.43

1.40±0.46

0.83±0.25

0.60±0.12

0.60±0.14

0.80±0.24

1.00±0.14

1.00±0.14

0.93±0.16

1.08±0.26

1.28±0.29

3.2

2.5

2.2

1.8

1.5

1.4

1.3

-1.6

0.000

0.004

0.002

0.055

0.063

0.185

0.330

0.086

0

0.2

0.4

0.6

0.8

1

1.2

1.4

ERMS

miR

-450

b-5p

(fol

dch

ange

/GA

PD

H)

TGF-β1 low (IRS 0-4)

**

*

* *

207

Westem blot

Northem blot

RT-PCR

11051 7910 14736

TGF-β1

miR-450b-5p

U6

miR-450b-5p

GAPDH

β-actin2.1 2.8 1.9 0.4

1.1 0.8 2.1 3.8

0.5 0.2 0.8 1.2TGF-β1 high (IRS 6-9)

ARMSERMSARMS

Figure 2. TGF-b1 and miR-450b-5p correlated in human RMS tissue. (a) MiR-4275, miR-411-5p, miR-411-3p, miR-493-5p, miR-493-3p, miR-450b-5p, miR-4298 and miR-2113 expression correlated with TGF-b1expression, respectively, in two cases of ARMS and two cases of ERMS byreal-time RT–PCR analysis using RNA isolated from RMA tissues. Bars indicate RNA level normalized to GAPDH±s.d. SPSS analysis, comparisonof miRNA folds change between low–high TGF-b1 expression in RMS tissues using Student’s t-test. (b) Representative validation of miR-450b-5p expression correlated with TGF-b1 by western blot analysis TGF-b1, northern blot and RT–PCR analyses of miR-450b-5p. Protein and RNAisolated from RMS tissues with high and low TGF-b1 expression. The upper Arabic numerals are numbers of RMS tissues. (c) Real-time RT–PCRassays for miR-450b-5p were performed on RNA isolated from RMS tissues with high and low TGF-b1 expression. For the immunoreactivescore (IRS) of TGF-b1, see the Supplementary Information (Supplementary Table S1). Bars indicate RNA level normalized to GAPDH±s.d.MiR-450b-5p level in RMS with low versus high TGF-b1 expression. Each assay was conducted for at least three independent times. *Po0.05;**Po0.01.

TGF-b1 suppression of miRNA-450b-5p expressionMM Sun et al

3

& 2013 Macmillan Publishers Limited Oncogene (2013), 1 – 12

(Figure 5a). All six cell lines (ERMS: RD, RH36 and SMS-CTR;ARMS: RH30, RH28 and RH3) were independently examinedin these analyses, and yielded concordant results. Next,we confirmed the effects of miR-450b-5p-mimic/inhibitoron the protein levels of ENOX2/PAX9 in RD cells. As shown inFigure 5b, downregulated protein levels of ENOX2 and PAX9were determined in RD cells 24 h following miR-450b-5p-mimictreatment. However, higher protein levels of ENOX2 and PAX9were found in RD cells treated with miR-450b-5p-inhibitor; inaddition, there was a more significant increase in PAX9 (Po0.05).

To confirm the effects of TGF-b1 treatment on the expressionof miR-450b-5p, ENOX2 and PAX9 in the RD cell line relative to

TGF-b1 knockdown, we investigated the mRNA levels of miR-450b-5p, ENOX2 and PAX9 in RD cells treated with 5 or 10 ng/mlTGF-b1, compared with TGF-b1-knocked down RD cells. As shownin Figure 5c, miR-450b-5p was increased in TGF-b1-knocked downRD cells, but decreased in TGF-b1-treated RD cells, particularly inthe 10-ng/ml TGF-b1 treatment group. Both ENOX2 and PAX9expression in TGF-b1-knocked down RD cells was decreased.However, TGF-b1 treatment led to a slight increase of ENOX2 in RDcells and a marked increase of PAX9.

Next, we explored whether the 30-untranslated repeat (UTR) ofENOX2 and PAX9 are functional targets of miR-450b-5p. As shownin Figure 5d, the reporter activities for wild-type 30-UTR sequences

0

1

2

3

4

miR

-450

b-5p

(fol

dch

ange

/GA

PD

H)

MNC M INC I

RD RH28

*

*

*

*

0

1000

2000

3000

4000

5000

6000

1

[3H

]inco

rpor

atio

n(c.

p.m

.)

MNC M INC I

*

**

RD

0

1000

2000

3000

4000

5000

6000

7000

[3H

]inco

rpor

atio

n(c.

p.m

.)

MNC M INC IRH28

****

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

OD

val

ues

MNC M INC I

*

**

RD

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

OD

val

ues

MNC M INC IRH28

**

***

*

MNC MM+

TGF-β1siRNA

0

10

20

30

40

Caspase3

Pos

itive

cel

l con

tent

(%)

MNC M M+ TGF-β1 siRNA

* ** *

Cas

pase

3T

UN

EL

days642 1 days642

1 days642 1 days642

TUNEL

Figure 3. MiR-450b-5p arrests proliferation of RMS cells by inducing apoptosis. (A) Expression of miR-450b-5p in embryonal (RD) and aveolar(RH28) RMS cell lines, transfected with mimic or inhibitor of miRNA-450b-5p (M or I), and negative control of mimic or inhibitor (MNC or INC)(see Materials and methods) was confirmed by real-time RT–PCR analysis on 4 days of treatment. Bars indicate RNA levels normalized toGAPDH±s.d. (B) Proliferation of RD and RH28 cells, treated with oligonucleotides (as above) for a period of 6 days, was evaluated by [3H]thymidine incorporation assay and MTT (see Materials and methods). The amount of [3H] thymidine incorporated was measured with a liquidscintillation counter. The optical density (OD) at 557 nm was determined using a microplate reader. The mean values from triplicate wells foreach treatment were determined. (C) Apoptosis was measured by the number of caspase-3-positive cells and TUNEL-positive cells. RD cellswere treated with oligonucleotides (as above) for a period of 6 days, followed by immunofluorescence staining with anti-caspase-3 and TUNELcell apoptosis detection (FITC end-labeling of the fragmented DNA of the apoptotic RD cells). Representative images of transfected MNC or Mof miR-450b-5p to RD or TGF-b1-knocked down RD cells (� 200). Each assay (A–C) was conducted for at least three independent times. Errorbars indicate s.d. *Po0.05; **Po0.01; ***Po0.005.

TGF-b1 suppression of miRNA-450b-5p expressionMM Sun et al

4

Oncogene (2013), 1 – 12 & 2013 Macmillan Publishers Limited

of both ENOX2 and PAX9 were decreased by co-transfection withmiR-450b-5p mimic in RD cells and increased by co-transfectionwith miR-450b-5p inhibitors. However, the activity of reportconstruct mutated at the specific miR-450b-5p target site wasunaffected by simultaneous transfection.

To comprehensively explore the clinical association betweenTGF-b1, miR-450b-5p, and targets ENOX2 and PAX9, we examinedmRNA isolated from human RMS tissues with low–high expressionof TGF-b1. According to our previously optimized scoring systemof TGF-b1 expression in paraffin-embedded tissues of RMS,9 we

paired TGF-b1-low-expression (immunoreactive score 0–4) RMStissue with TGF-b1-high-expression (immunoreactive score 6–9)RMS tissue. As shown in Figure 5e, high ENOX2 and PAX9expression levels were associated with low miR-450b-5p expres-sion and high TGF-b1 expression. The statistical analysis demon-strated that miR-450b-5p was significantly expressed in themajority of paired tissues (4/6), ENOX2 was significantly expressedin half of the paired tissues (3/6) and ENOX2 was significantlyexpressed in more than half of the paired tissues (4/6). Theseresults are in concordance with findings obtained in RMS cell lines.

0

1000

2000

3000

4000

1 2 3 4 weeks

Tum

or v

olum

e (m

m3)

MNC M INC I

Treatment time

*

** *

Ki67 Myogenin Myosin Desmin

MN

C tr

eate

dM

trea

ted

0

2

4

6

8

Ki67 Myogenin Myosin Desmin

IHC

sco

re

MNC M

*

***

0

0.5

1

1.5

GAPDH

1 2Myogenin Myosin

3 1 2 3

1 2 3 1 2 3

Rat

io o

fdi

ffere

ntia

tion

mak

erm

RA

N/G

AP

DH

1, MNC; 2.M+TGF-β1; 3, M

*** ***

miR-450b-5pinhibitors

miR-450b-5pmimics

Figure 4. MiR-450b-5p inhibits tumorigenicity of RMS by promoting myogenic differentiation. (A) Tumor volume of RD xenografts wasexamined. All mice were injected subcutaneously in the right flank with 100 ml RD cell suspension (1� 107). Once tumors reached B0.5 cm3,intratumoral injection was performed with mimic or inhibitor of miRNA-450b-5p (M or I) and negative control of mimic or inhibitor (MNC orINC) (see Materials and methods), once weekly for 4 weeks. Tumor volume (V, mm3) was calculated as (L�W2)/2, where L is length (mm) and Wis width (mm). The picture shows a representative tumor sphere. (B, C) Immunohistochemical staining and quantification of sections of RMSxenografts based on RD cells injected intratumorally with miR-450b-5p mimic or negative control oligonucleotides for 4 weeks. Ki67-specificantibody was used as the marker for cell proliferation; Myogenin, Myosin and Desmin were used as makers for myogenic differentiated cells.Original magnification, � 400. (D) Representative quantitative RT–PCR analysis of Myogenin and Myosin in RMS xenografts tumors from miR-450b-5p mimic or negative control oligonucleotide-treated RD cells, and with TGF-b1 antagonist or not for 2 weeks. All means values (±s.d.)are from three independent experiments (A, C and D). *Po0.05; **Po0.01; ***Po0.005.

TGF-b1 suppression of miRNA-450b-5p expressionMM Sun et al

5

& 2013 Macmillan Publishers Limited Oncogene (2013), 1 – 12

MiR-450b-5p downregulates ENOX2 and PAX9 proteins, andpromotes MyoD of RMSWe examined the mRNA levels of ENOX2 and PAX9 in siRNA-mediated ENOX2 or PAX9-knocked down RD cells, respectively,48 h after initial miR-450b-5p-mimic transfection to determine

whether the effects of miR-450b-5p would be enhanced. Asshown in Figure 6A, ENOX2-targeting siRNAs and PAX9-targetingsiRNAs each repressed the respective mRNAs to 0.17- and0.33-fold change/GAPDH, compared with 1.93- and 1.83-foldchange/GAPDH in the control. As anticipated, miR-450b-5p mimic

0

100

200

300

mR

NA

(%

con

trol

)MNC M INC IENOX2

*

*** **

* *

*

RD

0

50

100

150

200

250

mR

NA

(%

con

trol

)

MNC M INC IPAX9

*

*

*

** *

*

***

0

50

100

150

200

250

miR-450-5P

mR

NA

(%

cont

rol)

ControlTGF-β1 knock downTGF-β1 treatment (5ng/ml)TGF-β1 treatment (10ng/ml)

*

* * *

*

0

100

200

300

Rel

ativ

e si

gnal

(%co

ntro

l)

*

*

ENOX2

Wt. UTRMut.UTR

miR-Ctrl

miR-M

miR-I

0

100

200

300

Rel

ativ

e si

gnal

(%co

ntro

l)

PAX9

**

**

Wt. UTRMut.UTR

miR-Ctrl

miR-M

miR-I

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

mR

NA

s le

vel

(fol

d ch

ange

/GA

PD

H)

miR-450b-5p ENOX2 PAX9

235

ERMS

TGF-β1 low (IRS 0-4)

**

* *

*

*

* *

* *

* **

miR-450b-5p-M

CtrlENOX2

PAX9

β-actin

ENOX2

**

*

PAX93

2

1

0Ctrl 12h

miR-450b-5p-M miR-450b-5p-I

24h Ctrl 12h 24h

Rat

io o

f EN

OX

2an

d P

AX

9/β-

actin

(%

)

24h 48h Ctrl 24h 48h

miR-450b-5p-I

RH

3

RH

28

RH

30

SM

S-

CT

R

RH

36 RD

RH

3

RH

28

RH

30

SM

S-

CT

R

RH

36

PAX9ENOX2

+

+++++

+

+ + +

+

+

+ +

-

-

--

-

--

---

-

---

-

-

+

+++++

+

+ + +

+

+

+ +

-

-

--

-

--

---

-

---

-

-

115261456777896322223541549876562728791123458

ARMSERMSARMS

TGF-β1 high (IRS 6-9)

TGF-b1 suppression of miRNA-450b-5p expressionMM Sun et al

6

Oncogene (2013), 1 – 12 & 2013 Macmillan Publishers Limited

intensified the effects of ENOX2 siRNA and PAX9 siRNA knock-down on mRNA levels of ENOX2 and PAX9. Next, to study whetherENOX2 and PAX9 were involved in miR-450b-5p deregulation andmediated the block of RMS MyoD, we individually knocked downthese target genes using siRNAs in RD cells, and the number ofcaspase-3 and myosin heavy chain (MHC)-positive cells wasanalyzed using rabbit anti-caspase-3 and MHC. As shown inFigure 6B, siRNA-mediated knockdown of either ENOX2 or PAX9induced apoptosis and MyoD, as well as miR-450b-5p mimic.Furthermore, we observed that siRNA-mediated knockdown ofeither ENOX2 or PAX9 promoted apoptosis and MyoD caused byengineered miR-450b-5p derepression (Figure 6C). Similar assayswere conducted with RH28 cells and similar results wereindependently obtained (see Supplementary Information,Supplementary Figure S3).

TGF-b1 signal pathway regulates miR-450b-5pTo investigate whether the downstream proteins of the TGF-b1signal pathway have a key role in the regulation of miR-450b-5pand ENOX2/PAX9, we examined the expression levels of miR-450b-5p and ENOX2/PAX9 in TGF-b1-, Smad4-, Smad2- andSmad3-knocked down RMS cell lines. To determine whether thesesignal proteins exhibited complete knockdown, we examinedSmad7, a TGF-b1-inducible antagonist of TGF-b1 signaling in TGF-b1 signal-gene-knocked down RD and RH28 cells treated withTGF-b1 5 ng/ml. As shown in the Supplementary Information,Supplementary Figure S4, Smad7 expression increased signifi-cantly in the TGF-b-treated TGF-b1-knocked down group, and notin Smad4-, Smad2- and Smad3-knocked down RMS cell lines.

We observed greater expression of pre-miR-450b-5p andmature-miR-450b-5p in TGF-b1 siRNA-treated RD cells and RH28cells compared with controls (Figure 7a). Moreover, the signal ofENOX2 and PAX9 was decreased in TGF-b1-, Smad4- and Smad3-knocked down RD cells, but not in Smad2-knocked down RD cellsexamined with ENOX2- and PAX9-30-UTR-containing reporter(Figure 7b); a similar finding was observed in RH28 cells(Supplementary Information, Supplementary Figure S5). Theseresults suggest that a key role is played by Smad4/3 on miR-450b-5p/ENOX2/PAX9 regulation.

As miR-450b-5p is in a miRNA cluster with miR-450a/b-3p/b-5p,and TGF-b1 regulates miR-450b-5p, it may also regulate othermiRNA clusters. To test this hypothesis, we examined miR-450a,miR-450b-5p and miR-450b-3p in TGF-b1 signal-protein gene-silenced RD cells individually. As shown in Figure 7c, miR-450a,miR-450b-5p and miR-450b-3p expression was more or lessinduced in TGF-b1 signal-protein gene-silenced RD cells. Interest-ingly, exogenous TGF-b1 antagonized miR-450b-5p and miR-450b-3p levels in TGF-b1- and Smad2-knocked down RD cells, but not inSmad4- and Smad3-knocked down RD cells. These findings

revealed that Smad3, not Smad2, has an important role in controlprocessing of miR-450b-5p expression, as well as the commonSmad and Smad4. This observation provides additional evidencethat TGF-b1 inhibits muscle differentiation through Smad3.15

DISCUSSIONThe TGF-b1 signaling pathway has a vital role in the developmentand homeostasis of normal tissues, and abnormal function of thispathway contributes to the initiation and progression of diversecancer types, including RMS.9,16,17 Our results demonstrated thatmiR-450b-5p acts as a tumor suppressor, has a key role in MyoD ofRMS by targeting of ENOX2 and PAX9, and is negatively regulatedby the TGF-b1/Smad pathway.

Previous studies have shown that TGF-b1 regulates theexpression of several miRNAs. These studies suggest that theseTGF-b1-upregulated miRNAs function as oncogenes; miR-21targets programmed cell death 4,11 miR-155 contributes toepithelial cell plasticity by targeting RhoA,12 and miR-181btargets the tissue inhibitor of metalloprotease 3.13 In this study,we performed a genome-wide analysis of differentially expressedmiRNAs in TGF-b1-knocked down or not RMS cell lines, andidentified TGF-b1-related miRNAs in RMS for the first time. Wedemonstrated that a set of miRNAs was prominently upregulatedin TGF-b1-knocked down RMS cells.

In recent years, miRNAs have emerged as prominent factors incancer. With respect to cancer, miRNAs are often located inunstable genomic regions, and therefore are typically down-regulated in tumors;18 moreover, inhibition of miRNA biogenesistends to enchance tumorigenesis.19 Several studies havedemonstrated that miRNA maturation pathways crosstalk withintracellular signaling molecules, Smad proteins,11 estrogenreceptor20 and p53.21,22 To determine whether TGF-b1downregulates miR-450b-5p as a tumor suppressor or not, weengineered re-expression of miR-450b-5p-repressed proliferationof both ERMAS cell line RD and ARMS cell line RH28; however,modest or no significant growth inhibition was observed inmiR-450b-5p inhibitor-treated RMS cell lines (Figure 3). Engineeredre-expression of miR-450b-5p also reduced in vivo tumorigenicity.Notably, this repression of in vivo tumorigenicity depended ontime-dependent treatment of miR-450b-5p mimic (Figure 4).

As miR-450b-5p arrests the growth of RMS, we examined therole of miR-450b-5p in the differentiation of RD cells. Weinvestigated apoptosis (caspase-3 and TUNEL), MyoD regulator(myogenin), makers of MyoD (myosin and desmin) and morphol-ogy changes in RD cells treated with miR-450b-5p mimic in vitroand in vivo. These findings revealed that miR-450b-5p promotesMyoD of RMS cells by inducing apoptosis, which suggests thatmiR-450b-5p is a tumor suppressor.

Figure 5. ENOX2 and PAX9 are targets for miR-450b-5p. (a) ENOX2 and PAX9 expression levels were each downregulated by miR-450b-5p-mimic (M), but upregulated by miR-450b-5p-inhibitor (I) transfection independently, compared with negative control of mimic or inhibitor(MNC or INC) treatments (see Materials and methods), performed in human embryonal (RD, RH36 and SMS-CTR) and alveolar (RH30, RH28 andRH3) histotype RMS cell lines. (b) Western blot confirmed the effects of miR-450b-5p mimic/inhibitor on the protein levels of ENOX2/PAX9 inRD cells. Downregulated protein levels of ENOX2 and PAX9 were explored in RD cells 24 h after being treated with miR-450b-5p mimic.However, the higher protein levels of ENOX2 and PAX9 were explored in RD cells 24 h after being treated with miR-450b-5p inhibitor, andPAX9 increased more significantly (Po0.05). (c) RT–PCR confirmed the effects of TGF-b1 treatment on the expression of miR-450b-5p, ENOX2and PAX9 in RD cell line, relative to TGF-b1 knockdown. MiR-450b-5p increased in TGF-b1-knocked down RD cells, but decreased in TGF-b1-treated RD, especially more significantly in 10 ng/ml TGF-b1 treatment. Both ENOX2 and PAX9 in TGF-b1-knocked down RD cells decreased.The TGF-b1 treatment leads to a slight increase of ENOX2, but a marked increase of PAX9. (d) Diagram of ENOX2- and PAX9-30-UTR-containingreporter constructs RD with transfection of Wt- or mut-report and miR-450b-5p mimic or inhibitor (miR-M or miR-I). The luciferase assaysconfirmed that miR-450b-5p binds to the wild-type 30-UTR sequences of ENOX2 and PAX9 in RD cells. Co-transfection of miR-450b-5p mimicsignificantly reduced the luciferase levels, and co-transfection of miR-450b-5p inhibitor increased luciferase levels. Mut: contains seven-basemutation at the miR-450b-5p target region. (e) Clinical association between TGF-b1, miR-450b-5p, and targets ENOX2 and PAX9 wereexamined by real-time PCR. 235, 458, 1123, 2879, 5627, 9876 versus 154, 2354, 63222, 7789, 14567, 11526, respectively (the Arabic numeralsare numbers of RMA tissues). Each assay (a–c) was conducted at least three independent times. Error bars indicate s.d. *Po0.05; **Po0.01;***Po0.005.

TGF-b1 suppression of miRNA-450b-5p expressionMM Sun et al

7

& 2013 Macmillan Publishers Limited Oncogene (2013), 1 – 12

We identified six miR-450b-5p targets, including ENOX2 andPAX9 by bioinformatic analyses, and functional as well as bindingassays. It is known that each miRNA can affect multiple targets viadistinct mechanisms.23 Thus, we focused on identification ofcandidate miR-450b-5p targets that could exert a miR-450b-5pgrowth inhibition- or differentiation-inducing effect. The mRNAsof ENOX2 and PAX9 were downregulated in RMS cell linestreated with miR-450b-5p-mimic, and induced differentiation withincreased numbers of TUNEL-positive cells and myogenic markerMHC-positive cells, reinforced by siRNA-mediated knockdownof either ENOX2 or PAX9 (Figure 6). Our investigation ofclinical associations demonstrated that there is an inverserelationship between miR-450b-5p and targets, ENOX2 andPAX9, and a similar inverse relationship between miR-450b-5p

and TGF-b1 expression in RMS tissues. Taken together, thesefindings are in concordance with previous reports that TGF-b1signal pathways contribute to the inhibition of differentiation inRMS9 by revealing specific miRNAs that are underexpressed inRMS. We propose that miR-450b-5p acts as an antioncomir in thiscontext by negatively regulating specific oncogenes. This view issupported by miR-450b-5p re-expression or knockdownexperiments, in which its targets ENOX2 and PAX9 exhibitedderepression or repression, respectively. Thus, the TGF-b1/miR-450b-5p/ENOX2/PAX9 pathway constitutes a previouslyunrecognized differentiation regulator of RMS. It is unclear howENOX2 and PAX9 are involved in the MyoD of RMS. Based onour finding that miR-450b-5p mimic induces apoptosis anddifferentiation, which is reinforced by knockdown of either

0

0.5

1

1.5

2

2.5

3

mR

NA

leve

l (

fold

cha

nge

/ GA

PD

H)

ENOX2 PAX9

MNC+siRNACtrl

M

ENOX2 siRNA

PAX 9 siRNA

* * **

** **

*** *** *** ***

MNC M M+ENOX2siRNA M+PAX9siRNA

TU

NE

LM

HC

0

10

20

30

40

Pos

itive

cel

l con

tent

(%)

TUNNEL MHC

MNC+ siRNACtrl

M

ENOX2 siRNA

PAX 9 siRNA

** *

**

**

*****

******

***

+

-

-

- -

-

- -

-

-

-

-

- -

-

-

-

+

+

+

+

+ +

+

+

-

-

- -

-

- -

-

-

-

-

- -

-

-

-

+

+

+

+

+ +

+

+

-

-

- -

-

- -

-

-

-

-

- -

-

-

-

+

+

+

+

+ +

+

+

-

-

- -

-

- -

-

-

-

-

- -

-

-

-

+

+

+

+

+ +

+

Figure 6. MiR-450-5b-5p promotes differentiation of RMS by targeting ENOX2 and PAX9. (A) The mRNA levels of ENOX2 and PAX9 arepresented for the indicated transfected of RD cells with miR-450-5b-5p mimic (M), negative control of mimic (MNC), ENOX2 siRNA, PAX9 siRNAand control siNRA (siRNACtrl). Bars indicate RNA levels normalized to GAPDH±s.d. (B) Representative images of TNNEL- and MHC-positive RDcells upon miR-450-5b-5p mimic, ENOX2 siRNA and PAX9 siRNA induction. RD cells treated with oligonucleotides (as above) for a period of 6days, immunofluorescence staining with TUNEL cell apoptosis detection kit and anti-MHC. (C) Quantity of TUNEL and MHC expression in miR-450-5b-5p mimic-treated RD cells, promoted by siRNA-mediated knockdown of either ENOX2 or PAX9 performed 48 h after miR-450-5b-5pmimic transfection. Each assay was conducted for at least three independent times. Error bars indicate s.d. *Po0.05; **Po0.01; ***Po0.005.

TGF-b1 suppression of miRNA-450b-5p expressionMM Sun et al

8

Oncogene (2013), 1 – 12 & 2013 Macmillan Publishers Limited

ENOX2 or PAX9 (Figure 6), and previous reports demonstratingthat ENOX2 enhances cell proliferation and PAX9 modulates tissuedifferentiation,24–26 we hypothesized that these interactionscaused RD cells to drop out of cell cycle and block MyoD inRMS cells; further studies are warranted to investigate thishypothesis.

Recently, several studies have demonstrated that the expressionof specific miRNAs that regulate skeletal muscle developmenthave been shown to be downregulated in RMS, raising thepossibility that miRNA re-expression could potentially serve as aneffective differentiation therapy for RMS. RMS most commonlyoriginates from skeletal muscle; however, a recent findingindicates that RMS also originates from adipocyte lineage,27

indicating that the complex origination is far beyond our present

understanding. Therefore, exploration of differentially expressedmiRNAs between RMS and muscle tissue is warranted. Wedetermined the differentially expressed miRNAs in a number ofrepresentative TGF-b1-knocked down ERMS and ARMS cell lines,based on the common biological behavior, the excretive TGF-b1,differentiation block and coloration, previously demonstrated byour reports and other reports. Whether or not there aredifferentiated miRNAs between ERMS and ARMS, andassociations between ERMS/ARMS and TGF-b1/miR-450b/ENOX2/PAX9 axis, requires further study.

Several TGF-b1-regulated miRNAs have been shown to mediateTGF-b1 signaling to control the aggressiveness of cancer cellsassociated with metastasis and invasiveness.11–13 In this study, ofthe six significantly differentially expressed miRNAs, none of the

0

0.5

1

1.5

2

2.5

miR

-450

b-5p

leve

l(f

old

chan

ge/G

AP

DH

)

Ctrl pre-miR mature-miR

siRNA CtrlTGF-β1siRNA

****

*

**

RD RH28

0

20

40

60

80

100

120

Rel

ativ

e si

gnal

(%

Ctr

l)

ENOX2 PAX9

ENOX2-Wt.UTRPAX9-Wt.UTR

*

**

* * *

*

TGF-β1siRNA

Smad4siRNA

Smad2siRNA Smad3

siRNA

RD

0

0.5

1

1.5

2

2.5

miR

-450

a/b-

5p/b

-3p

(fol

d ch

ange

s/C

ontr

ol)

miR-450a miR-450b-5p miR-450b-3p

TGF-β1siRNA

TGF-β1siRNA

+TGF-β1

Smad4siRNA

Smad4siRNA

+TGF-β1

Smad2siRNA

Smad2siRNA

+TGF-β1

Smad3siRNA

+TGF-β1

Smad3siRNA

**

* ** *

---

---+

+ + ++

+-

-+

+-

-+

+-

-+

+-

-+

+

Figure 7. Smads-mediated TGF-b1 signal pathway regulates miR-450b-5p. (a) Real-time RT–PCR analysis on the effect of TGF-b1 siRNA on miR-450b-5p processing in RMS cell lines. The level of pre-miR-450b-5p and mature miR-450-5p increased in RD and RH28 cells treated with TGF-b1siRNA for a period of 6 days; moreover, mature miR-450-5p exhibited a more marked increase compared with pre-miR-450b-5p. Ctrl means thecontrol for mature miR-450-5p in RD and RH28 cells treated with TGF-b1 siRNA for a period of 24 h. (b) The luciferase assay analysis of miR-450b-5p targets ENOX2 and PAX9. The luciferase reporter confirmed that miR-450b-5p binds to the wild-type 30-UTR sequences of ENOX2 andPAX9 in RD cells (Figure 5). Signals of ENOX2 and PAX9 in TGF-b1-, Smad4-, Smad2- and Smad3-knocked down RD were examined after 48 h indifferentiation medium. Each assay was conducted at least three independent times. (c) Real-time RT–PCR analysis on expression of miR-450a,miR-450b-5p and miR-450b-3p in TGF-b1-, Smad4- and Smad3-knocked down RD cells treated with exogenous TGF-b1 5 ng/ml. Methods forTGF-b1-, Smad4-, Samd2- and Smad3-knocked down RD construction see Materials and methods, and Supplementary Information,Supplementary Figure S4. Error bars indicate s.d. *Po0.05; **Po0.01; ***Po0.005. (d) Schematic illustration of miR-450b-5p processing viaTGF-b1 signal pathway, block of RMS differentiation, miR-450b-5p targets ENOX2 and PAX9, and their possible relationship.

TGF-b1 suppression of miRNA-450b-5p expressionMM Sun et al

9

& 2013 Macmillan Publishers Limited Oncogene (2013), 1 – 12

miRNAs exhibited a change consistent with these results. Thisdiscrepancy may be due to the use of different cell types andinterference means to TGF-b1 signal for the miRNA array analysis.However, an important inconsistency between TGF-b1-mediatedupregulation of these miRNAs, such as miR-21,11 miR-18113 andmiR-155 expression,12 and TGF-b1-mediated downregulation ofmiR-450b-5p is the requirement of the common Smad, Smad4(Figure 7). Moreover, we demonstrated that Smad3 has animportant role in control processing of miR-450b-5p expression,as well as Smad4. Consistent with this finding, a previousstudy reported that TGF-b1 inhibited muscle differentiationthrough the Smad3-mediated repression of myogenic trans-cription factors.15,28,29 Further investigation is required todetermine the mechanism underlying the TGF-b1/Smadpathway downregulation of miR-450b-5p, although it is likelythat the regulation occurs post transcriptionally (Figure 7d).

This study revealed a novel mechanism for blocking MyoDof RMS, that is, TGF-b1 suppression of miRNA-450b-5p expression.Engineered re-expression of miR-450b-5p arrested growth ofRMS and promoted MyoD of RMS in vitro and in vivo. Notably,ENOX2 and PAX9 were targeted by miR-450b-5p, providing alikely mechanism responsible for the blocking of TGF-b1-mediatedRMS differentiation through previously unrecognized regulation oftumor-promoting pathways. It is intriguing to speculate that use ofmiR-450b-5p as a differentiation therapy for RMS could potentiallyprovide even prognostic information. Indeed, evidence indicatesthat clinical associations exist between miR-450b-5p and thehistological grade of RMS differentiation (data not shown), andmiR-450b-5p mimic promotes RMS differentiation in vitro andin vivo. Taken together, these findings indicate that miR-450b-5pacts as an anti-oncogenic miRNA in RMS by conferring repressionof specific oncogenes. Conceivably, targeting of miR-450b-5p byanti-miR oligonucleotides would constitute a novel approach toarrest RMS development.

MATERIALS AND METHODSCell culture and RMS tissue samplingThe human ERMS (RD, RH36 and SMS-CTR) and ARMS (RH30, RH28 andRH3) cell lines were obtained from the American Type Culture Collection(ATCC; Manassas, VA, USA) and were maintained in Dulbecco’smodified Eagle’s medium. The differentiation medium comprisedserum-free Dulbecco’s modified Eagle’s medium with 10 mg/ml insulinand transferrin.30,31 All samples were obtained following informed consentand with obscured identity corroding to the recommendations of theInstitutional Review Board. The histological material consisted of initialbiopsy or resection specimens and of post-chemotherapy biopsyspecimens from residual, recurrent or metastatic tumor sites.

Construction of TGF-b1 signal-protein-knocked down cell linesand western blotThe procedure for TGF-b1, Smad4, Smad2 and Smad3 gene silencing wasdescribed previously.8,9,32 These human TGF-b1 signal protein-specificoligonucleotide sequences were synthesized (Sangon Inc, Shanghai, China)and ligated into the mammalian expression vector, pSUPER gfp-neo(Catalog: VEC-PBS (phosphate-buffered solution)-0006, OligoEngine,Seattle, WA, USA) at the BglII and HindIII cloning sites. A neomycinresistance cassette was cloned into the pSUPER vector and transfected intoRMS cell lines using Effectent Transfection Reagent (Qiagen, Hilden,Germany). To test the efficiency of the gene silencing, western blot analysiswas utilized to explore the above signal proteins and conventional TGF-b1target Smad7. Membranes were incubated with specific antibodies againstTGF-b1, Smad4, Smad2, Smad3 and Smad7 (Santa Cruz Biotechnology,Santa Cruz, CA, USA) diluted at 1:1000. ENOX2 rabbit polyclonal antibodywas purchased from Abcam, Inc (Cambridge, MA, USA), and PAX9mouse polyclonal antibody was obtained from Abnova Inc (Taipei,Taiwan). Proteins were visualized with peroxidase-conjugated secondaryantibodies at 1:2000, using an enhanced chemiluminescence detectionsystem (Santa Cruz).

RNA isolation and miRNA arraysTotal RNA was isolated from TGF-b1-knocked down RD, SMS-CTR andRH28, respectively, using TRIzol reagent (Invitrogen, Carlsbad, CA, USA).After having passed RNA quantity measurement using the NanoDrop 1000,the samples were labeled using the miRCURY Hy3/Hy5 Power labeling kitand hybridized on the miRCURY LNA Array (v.16.0) (Exiqon, Vedbaek,Denmark). Following the washing steps, the slides were scanned using theAxon GenePix 4000B microarray scanner. Scanned images were thenimported into GenePix Pro 6.0 software (Axon Instruments, Foster City, CA,USA) for grid alignment and data extraction. Expressed data werenormalized using the Median normalization. Finally, hierarchical clusteringwas performed to determine distinguishable miRNA expression profilesamong the samples.

Real-time RT–PCR quantification of precursor and mature miRNAsTotal RNA was harvested from RMS cell lines and frozen tissues accordingto the manufacturer’s recommendations. Pre-miRNA sequences wereobtained from miBase (http://www.mirbase.org/index.shtml). The primersfor pre-miRNA were designed with the following criteria: (a) both forwardand reverse primers were designed to be located within the hairpinsequence of the miRNA precursors, (b) a maximal extension of fournucleotides in the 50 direction was allowed for each primer over thepresumed termini of the pre-miRNA, and (c) primers were designed with amaximal Tm difference between both primers of p2 1C and a primerlength between 18 and 24 nucleotides. The sequences of the miRNA-450b-5p precursor was as follows: 50-GCAGAAUUAU(UUUUGCAAUAUGUUCCUGAAUA)UGUAAUAUAAGUGUA(UUGGGAUCAUUUUGCAUCCAUA)GUUUUGUAU-30 . The underlined sequences are forward and reverse PCR primers formiRNA-450b-5p precursors. As the hairpin is contained within both the pri-miRNA and the pre-miRNA, primers amplify both RNAs. Mature miRNAanalysis was performed using the TaqMan miRNA assays, including RT andreal-time PCR. The RT reactions were performed using a single miRNA-specific stem-loop RT primer, as described by Chen33,34 (for primers, seeSupplementary Information, Supplementary Table S2).

Northern blot analysisFor quantitative northern blot analysis of miRNAs, 5 mg of total RNA wereelectrophoresed in a 15% polyacrylamide-urea gel and transferred byelectroblotting onto Hybond-Nþ membrane (Amersham Biosciences,Piscataway, NJ, USA). Hybridization was performed with 32P-labeled DNAoligos (see Supplementary Information, Supplementary Table S3). Anti-U6,50-TGTGCTGCCGAAGCGAGCAC-30 .

Transient transfectionThe sequences of mimics and inhibitors of miRNAs used in our studies arelisted in Supplementary Information, Supplementary Table S4. All of thenucleotides contain 20-OMe modifications at every base and a 30- C3-containing amino linker.35 To transfect, 100-pmol mimics or inhibitors ofmiRNA (miR-M or miR-I) and negative control of mimics or inhibitors (MNCor INC) in 200ml of serum-free medium were mixed with 5 ml ofLipofectamine 2000 transfection reagent (Invitrogen), dissolved in 200mlof the same medium. The resulting transfection solutions were then addedto each well containing 1.6 ml of the medium. Eight hours later, thecultures were replaced with differentiation medium. The specific siRNAs ofENOX2 and PAX9 were designed with BLOCK-iT RNAi Designer (Invitrogen)(ENOX2 siRNA: 50-GGAATTGCCGGACAACCAA-30 ; PAX9 siRNA: 50-GCGTGTGCGACAAGTACAA-30) (synthesized by Shanghai GeneCore BiotechnologiesCo, Ltd, Shanghai, China). The siRNA transfection was conducted followingidentical procedures as for miR at a final concentration of 30 nM.

ImmunohistochemistryParaffin-embedded tissue sections (3–5mm thick) were immunostained.After boiling in 10 mM citrate buffer for 10 min in a microwave oven, thetissue sections were incubated with 10% normal rabbit serum in buffer toblock nonspecific binding sites, and incubated with antibodies at 4 1Covernight. The slides were incubated with peroxidase-conjugated IgG(Genomics, Shanghai, China) and stained with diaminobenzidine–chromogen substrate mixture (DAKO, Carpinteria, CA, USA), and thencounterstained with hematoxylin. The expression of proteins wereclassified according to the following grading system: staining extensitywas categorized as 0 (no positive cells), 1 (p25% positive cells), 2 (425%and p50% positive cells ) or 3 (450% positive cells), and staining intensity

TGF-b1 suppression of miRNA-450b-5p expressionMM Sun et al

10

Oncogene (2013), 1 – 12 & 2013 Macmillan Publishers Limited

was categorized as 0 (negative), 1 (weak), 2 (moderate) or 3 (strong). Thetotal immunoreactive score was a summation of the individual categories.

Cell growth assaysThe embryonal (RD) and alveolar (RH28) RMS cell lines were plated at5� 104 cells per well in 96-well plates, with transfection solutionscontaining miR-450b-5p-M, miR-450b-5p-I or NC for 8 h. Then, the culturemedium was replaced with differentiation medium. The [3H] thymidineincorporation with 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazoliumbromide (MTT) was explored after each treatment for 1, 2, 4 and 6 days.During the last 4 h of each treatment, cells were pulsed with 3.7� 104 [3H]thymidine (1 mCi) per well or with MTT (Sigma, St Louis, MO, USA) 10 ml perwell. After cells were pulsed with [3H] thymidine, the cells were rinsed withPBS, and DNA was precipitated with 10% (wt/vol) trichloroacetic acid for30 min at 4 1C. The amount of [3H] thymidine incorporated was measuredwith a liquid scintillation counter. For the MTT assay, after the cells wereincubated at 37 1C for 4 h to allow MTT formazan formation, the mediumand MTT were replaced by 100ml DMSO to dissolve the formazan crystals.Then, the optical density at 557 nm was determined using a microplatereader (Model 550; Bio-Rad, Hercules, CA, USA). The mean value fromtriplicate wells for each treatment was determined.

Apoptosis assay and immunofluorescence stainingAfter cultivating on coverslips for 24 h in Dulbecco’s modified Eagle’smedium that contained 10% fetal calf serum, the coverslips were placed atthe bottom of each well. The 400ml of the above resulting transfectionsolutions were then added to each well containing 1.6 ml of medium. Eighthours later, the culture medium was replaced with differentiation medium.After 6 days, coverslips were then washed with iced PBS and fixed with 4%paraformaldehyde for 15 min at 4 1C. Briefly, the treated cells werepermeabilized with 0.1% Triton X-100 for fluorescein isothiocyanate (FITC)end-labeling of the fragmented DNA of the apoptotic RD cells, using aTUNEL cell apoptosis detection kit (Beyotime, Shanghai, China). Expressionof caspase-3 and MHC was analyzed using rabbit anti-caspase-3 and MHC(Boster, Wuhan, China). At least 200 cells per slide in random fields werecounted under a fluorescence microscope (BX2-FLB3-000; Olympus, Tokyo,Japan) to determine the percentage of FITC-positive cells.

Electron microscopyAfter RD cells were plated at 5� 104 cells per well in six-well plates withtransfection solution containing miR-450b-5p-M for 6 days, the cells wereharvested, fixed with 3% glutaraldehyde and osmium tetroxide, dehy-drated and embedded in Epon812. Thin sections were cut at a thickness of50 nm, stained with uranyl acetate and lead citrate, and examined underan H-600IV electron microscope (Hitachi Inc, Tokyo, Japan).

In vivo tumorigenicityAll animal procedures were approved by the Ethics Review Committee forAnimal Experimentation of Soochow University. A total of 100 male BALB/cnude mice (5–6 weeks old and 18 g in weight) were purchased form SLACLaboratory Animal Co Ltd (Shanghai, China). The mice were divided into sixgroups of 20 mice each. All mice were injected subcutaneously in the rightflank with 100-ml RD cell suspension (1� 107). For intratumoral transfec-tion, 100 pmol mimics or inhibitors of miRNA and negative controloligonucleotides in 100ml of PBS were mixed with 5 ml of Lipofectamine2000 transfection reagent (Invitrogen) dissolved in 100ml of the same PBS.Once tumors reached B0.5 cm3, the 200ml of the above resultingtransfection solutions were injected into the tumor once weekly for 4weeks. Tumor size was measured in a blinded manner weekly with calipers,and tumor volume was calculated. Histological differentiation wasevaluated by observing changes in morphology and expression of MyoDmarkers.

Bioinformatic prediction of miRNA targetsThe following bioinformatic algorithms were utilized to predictmiRNA target sites: target ccan 6.1 (http://targetscan.org), Diana microT4.0 (http://diana.cslab.ece.ntua.gr/DianaTools/index.php?r=microtv4/index)and MICRORNA.ORG (http://www.microrna.org/microrna/home.do).

Construction of 30-UTR luciferase plasmid and reporter assaysTo create ENOX2- and PAX9-30-UTR luciferase reporter constructs, the full-length 30-UTR of ENOX2 (1746 nucleotides) and PAX9(1026 nucleotides)was amplified using complementary DNA from RD (primers of ENOX2:forward -50-TGTGTGAGAGTCCAGCCTTG-30 , reverse -50-AAGTAAGTTCCGGGGTACATG-30 ; primers of PAX9: forward -50-GGTGCTCTCGCAAGGAGAAAGG-30 , reverse -50-GCGCTTATGAGTTAAACTGTGTC-30). The primers weredigested and cloned into the Xba1-site of pGL3 (Promega, Madison, WI,USA), checked for orientation, sequenced and named Wt. The miR-450b-5p-binding sites for ENOX2 and PAX9 are indicated in Figure 5. Themutagenesis within selected seeding sequence regions was conductedaccording to the following rules: A was changed to T and vice versa. Forreporter assays, RD and RH28 were transiently transfected with Wt ormutant reporter plasmid and miR-450b-5p mimic and inhibitor usingLipofectamine 2000 (Invitrogen). Reporter assays were performed 24 h posttransfection using the Dual-luciferase assay-system (Promega), andnormalized for transfection efficiency by co-transfected Renilla luciferase.The luciferase signal was read using a TD-20/20 Luminometer (TurnerBiosystems, Sunnyvale, CA, USA).

Statistical analysisAll data were presented as mean and s.d. Analysis was conducted usingSPSS 10.0 (Chicago, IL, USA). Student’s unpaired t-test was used tocompare values of test and control samples. A value of Po0.05 wasconsidered to indicate a statistically significant result.

CONFLICT OF INTERESTThe authors declare no conflict of interest.

ACKNOWLEDGEMENTSWe would like to thank the West China Hospital of Sichuan University for providing uswith the tumor specimens. This work was supported partially by the Chinese NatureScience Foundation (81072186, 81272738), Jiangsu Provincial Nature ScienceFoundation (SBK20110743), Jiangsu Provincial Higher Institution Nature ScienceFoundation (10KJB320018), Suzhou Applied Basic Research Programs (SYS201207),and Sponsorship of Jiangsu Overseas Research and Training Program for UniversityProminent Young and Middle-aged Teachers and Presidents.

REFERENCES1 Sebire NJ, Malone M. Myogenin and MyoD1 expression in paediatric rhabdo-

myosarcomas. J Clin Pathol 2003; 56: 412–416.2 Tapscott SJ, Thayer MJ, Weintraub H. Deficiency in rhabdomyosarcomas of a

factor required for MyoD activity and myogenesis. Science 1993; 259: 1450–1453.3 Mishra PJ, Merlino G. MicroRNA reexpression as differentiation therapy in cancer.

J Clin Invest 2009; 119: 2119–2123.4 Wurdinger T, Costa FF. Molecular therapy in the microRNA era. Pharmacogenomics

J 2007; 7: 297–304.5 Wang V, Wu W. MicroRNA-based therapeutics for cancer. BioDrugs 2009; 23:

15–23.6 Li Y, Foster W, Deasy BM, Chan Y, Prisk V, Tang Y et al. Transforming growth

factor-beta1 induces the differentiation of myogenic cells into fibrotic cells ininjured skeletal muscle: a key event in muscle fibrogenesis. Am J Pathol 2004; 164:1007–1019.

7 Sun Q, Zhang Y, Yang G, Chen X, Cao G, Wang J et al. Transforming growth factor-beta-regulated miR-24 promotes skeletal muscle differentiation. Nucleic Acids Res2008; 36: 2690–2699.

8 Wang SL, Yao HH, Guo LL, Dong L, Li SG, Gu YP et al. Selection of optimal sites forTGFB1 gene silencing by chitosan-TPP nanoparticle-mediated delivery of shRNA.Cancer Genet Cytogenet 2009; 190: 8–14.

9 Wang S, Guo L, Dong L, Li S, Zhang J, Sun M. TGF-beta1 signal pathway maycontribute to rhabdomyosarcoma development by inhibiting differentiation.Cancer Sci 2010; 101: 1108–1116.

10 Ye L, Zhang HY, Wang H, Yang GH, Bu H, Zhang L et al. Effects of transforminggrowth factor beta 1 on the growth of rhabdomyosarcoma cell line RD. Chin Med J(Engl) 2005; 118: 678–686.

11 Davis BN, Hilyard AC, Lagna G, Hata A. SMAD proteins control DROSHA-mediatedmicroRNA maturation. Nature 2008; 454: 56–61.

12 Kong W, Yang H, He L, Zhao JJ, Coppola D, Dalton WS et al. MicroRNA-155 isregulated by the transforming growth factor beta/Smad pathway and contributesto epithelial cell plasticity by targeting RhoA. Mol Cell Biol 2008; 28: 6773–6784.

TGF-b1 suppression of miRNA-450b-5p expressionMM Sun et al

11

& 2013 Macmillan Publishers Limited Oncogene (2013), 1 – 12

13 Wang B, Hsu SH, Majumder S, Kutay H, Huang W, Jacob ST et al. TGFbeta-mediated upregulation of hepatic miR-181b promotes hepatocarcinogenesis bytargeting TIMP3. Oncogene 2010; 29: 1787–1797.

14 Shoji Y, Saegusa M, Takano Y, Ohbu M, Okayasu I. Correlation of apoptosiswith tumour cell differentiation, progression, and HPV infection in cervical car-cinoma. J Clin Pathol 1996; 49: 134–138.

15 Liu D, Black BL, Derynck R. TGF-beta inhibits muscle differentiation throughfunctional repression of myogenic transcription factors by Smad3. Genes Dev2001; 15: 2950–2966.

16 Massague J. TGFbeta in cancer. Cell 2008; 134: 215–230.17 Bouche M, Canipari R, Melchionna R, Willems D, Senni MI, Molinaro M. TGF-beta

autocrine loop regulates cell growth and myogenic differentiation in humanrhabdomyosarcoma cells. FASEB J 2000; 14: 1147–1158.

18 Calin GA, Croce CM. MicroRNA signatures in human cancers. Nat Rev Cancer 2006;6: 857–866.

19 Kumar MS, Lu J, Mercer KL, Golub TR, Jacks T. Impaired microRNA processingenhances cellular transformation and tumorigenesis. Nat Genet 2007; 39:673–677.

20 Yamagata K, Fujiyama S, Ito S, Ueda T, Murata T, Naitou M et al. Maturationof microRNA is hormonally regulated by a nuclear receptor. Mol Cell 2009; 36:340–347.

21 Suzuki HI, Yamagata K, Sugimoto K, Iwamoto T, Kato S, Miyazono K. Modulation ofmicroRNA processing by p53. Nature 2009; 460: 529–533.

22 Suzuki HI, Miyazono K. Emerging complexity of microRNA generation cascades.J Biochem 2011; 149: 15–25.

23 Bartel DP. MicroRNAs: target recognition and regulatory functions. Cell 2009; 136:215–233.

24 Liu NC, Hsieh PF, Hsieh MK, Zeng ZM, Cheng HL, Liao JW et al. Capsaicin-mediatedtNOX (ENOX2) up-regulation enhances cell proliferation and migration in vitroand in vivo. J Agric Food Chem 2012; 60: 2758–2765.

25 Rodrigo I, Hill RE, Balling R, Munsterberg A, Imai K. Pax1 and Pax9 activate Bapx1to induce chondrogenic differentiation in the sclerotome. Development 2003; 130:473–482.

26 Zeng ZM, Chuang SM, Chang TC, Hong CW, Chou JC, Yang JJ et al. Phosphor-ylation of serine-504 of tNOX (ENOX2) modulates cell proliferation and migrationin cancer cells. Exp Cell Res 2012; 318: 1759–1766.

27 Hatley ME, Tang W, Garcia MR, Finkelstein D, Millay DP, Liu N et al. A mouse modelof rhabdomyosarcoma originating from the adipocyte lineage. Cancer Cell 2012;22: 536–546.

28 Liu D, Kang JS, Derynck R. TGF-beta-activated Smad3 represses MEF2-dependenttranscription in myogenic differentiation. EMBO J 2004; 23: 1557–1566.

29 Ge X, McFarlane C, Vajjala A, Lokireddy S, Ng ZH, Tan CK et al. Smad3 signaling isrequired for satellite cell function and myogenic differentiation of myoblasts. CellRes 2011; 21: 1591–1604.

30 Segev H, Fishman B, Ziskind A, Shulman M, Itskovitz-Eldor J. Differentiation ofhuman embryonic stem cells into insulin-producing clusters. Stem Cells 2004; 22:265–274.

31 Yang Z, MacQuarrie KL, Analau E, Tyler AE, Dilworth FJ, Cao Y et al. MyoD andE-protein heterodimers switch rhabdomyosarcoma cells from an arrested myo-blast phase to a differentiated state. Genes Dev 2009; 23: 694–707.

32 Guo H, Zhang HY, Wang SL, Ye L, Yang GH, Bu H. Smad4 and ERK2 stimulated bytransforming growth factor beta1 in rhabdomyosarcoma. Chin Med J (Engl) 2007;120: 515–521.

33 Chen C, Ridzon DA, Broomer AJ, Zhou Z, Lee DH, Nguyen JT et al. Real-timequantification of microRNAs by stem-loop RT-PCR. Nucleic Acids Res 2005; 33: e179.

34 Schmittgen TD, Lee EJ, Jiang J, Sarkar A, Yang L, Elton TS et al. Real-time PCRquantification of precursor and mature microRNA. Methods 2008; 44: 31–38.

35 Cheng AM, Byrom MW, Shelton J, Ford LP. Antisense inhibition of human miRNAsand indications for an involvement of miRNA in cell growth and apoptosis. NucleicAcids Res 2005; 33: 1290–1297.

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

TGF-b1 suppression of miRNA-450b-5p expressionMM Sun et al

12

Oncogene (2013), 1 – 12 & 2013 Macmillan Publishers Limited