PAX7 is a required target for microRNA-206-induced differentiation of fusion-negative rhabdomyosarcoma

Rhabdomyosarcoma (RMS) is the most common soft tissue sarcoma of childhood. RMS can be parsed based on clinical outcome into two subtypes, fusion-positive RMS (FP-RMS) or fusion-negative RMS (FN-RMS) based on the presence or absence of either PAX3-FOXO1 or PAX7-FOXO1 gene fusions. In both RMS subtypes, tumor cells show histology and a gene expression pattern resembling that of developmentally arrested skeletal muscle. Differentiation therapy is an attractive approach to embryonal tumors of childhood including RMS; however, agents to drive RMS differentiation have not entered the clinic and their mechanisms remain unclear. MicroRNA-206 (miR-206) expression increases through normal muscle development and has decreased levels in RMS compared with normal skeletal muscle. Increasing miR-206 expression drives differentiation of RMS, but the target genes responsible for the relief of the development arrest are largely unknown. Using a combinatorial approach with gene and proteomic profiling coupled with genetic rescue, we identified key miR-206 targets responsible for the FN-RMS differentiation blockade, PAX7, PAX3, NOTCH3, and CCND2. Specifically, we determined that PAX7 downregulation is necessary for miR-206-induced cell cycle exit and myogenic differentiation in FN-RMS but not in FP-RMS. Gene knockdown of targets necessary for miR-206-induced differentiation alone or in combination was not sufficient to phenocopy the differentiation phenotype from miR-206, thus illustrating that miR-206 replacement offers the ability to modulate a complex network of genes responsible for the developmental arrest in FN-RMS. Genetic deletion of miR-206 in a mouse model of FN-RMS accelerated and exacerbated tumor development, indicating that both in vitro and in vivo miR-206 acts as a tumor suppressor in FN-RMS at least partially through downregulation of PAX7. Collectively, our results illustrate that miR-206 relieves the differentiation arrest in FN-RMS and suggests that miR-206 replacement could be a potential therapeutic differentiation strategy.

maintain high expression of PAX7 and PAX3, transcription factors that promote proliferation and self-renewal in myogenic satellite cells. 10,11 However, the full constellation of factors contributing to the differentiation arrest in RMS remains elusive. 12 MicroRNAs (miRNAs) are non-coding RNAs that reduce gene expression through binding complementary sequences in 3′ untranslated regions (UTR) of target mRNAs resulting in transcript degradation. 13 miR-206 is a member of a miRNA family with miR-1-1 and miR-1-2 that share an identical seed sequence while differing at four base pairs outside of the seed sequence in the mature miRNA. While miR-1 is expressed more abundantly in cardiac muscle, miR-206 is expressed nearly exclusively in mature skeletal muscle with increasing expression during myogenesis driven by MyoD1 and Myogenin. [14][15][16] Genetic deletion of miR-206 in mice has revealed a role of miR-206 in the regeneration of the neuromuscular synapsis and skeletal muscle regeneration following injury. [17][18][19] In both FN-and FP-RMS, decreased miR-206 expression has been demonstrated in patient tumors compared with normal skeletal muscle. 20,21 Higher miR-206 expression correlated to increased patient survival in FN-RMS but not in FP-RMS. 21 To gain insight into the biological relevance of miR-206 in RMS, several groups overexpressed miR-206 in RMS cell lines and illustrated decreased proliferation and migration as well as an induction of differentiation. [21][22][23][24] Furthermore, viral expression of miR-206 in RMS cell line xenografts in mice decreased tumor growth. 22,25 This recent work has highlighted a few exciting targets of miR-206 in RMS; 26 however, the necessity and/or sufficiency of these putative miR-206 target genes in mediating RMS differentiation remained unexplored. Using a combinatorial approach with microarrays, large-scale proteomics, and prediction algorithms, we identified the crucial miR-206 targets responsible for the developmental arrest in RMS.
Results miR-206 mimic enforces differentiation of human RMS cell lines. To interrogate the mechanism and spectrum of miR-206 targets important for RMS pathogenesis, a miR-206 mimic or negative control (NC) mimic was transfected into human RMS cell lines. All RMS cell lines have lower miR-206 levels than human myotubes (Supplementary Figure S1a). Transfection of miR-206 mimic induced expression of myosin heavy chain (MHC), a differentiated muscle marker, in all RMS cell lines tested including the FN-RMS cell lines RD, SMS-CTR and Rh18 and the FP-RMS cell lines Rh30 and Rh41 (Figures 1a and b). Cell elongation and fusion was more evident in FN-RMS than in FP-RMS. In addition, miR-206 induced expression of terminal muscle differentiation markers evidenced by increased MYH1, ACTA1 and CKM mRNA (Supplementary Figures S1b-d). These changes were accomplished with a physiologic increase in miR-206 levels in cells maintained under high serum growth conditions still below the expression level of miR-206 in differentiated myotubes (Figure 1c and Supplementary Figure S1a). Low serum as used in normal myoblast differentiation experiments was not required for the miR-206induced RMS differentiation. In addition, increased miR-206 levels led to a decrease in Ki67 expression, a marker of proliferation, in all RMS cell lines but Rh41 (Figures 1d  and e). Thus, transfection with a miR-206 mimic relieved the differentiation blockade and decreased cell proliferation in RMS.
Prior to exploring the miR-206 target genes driving the differentiation phenotype, we determined the timing of expression of markers of terminal muscle differentiation following miR-206 transfection. RD cells were transfected with NC or miR-206 mimic and RNA was collected at 12 and 24 h, then daily for 7 days and ACTA1 and MYH1 expression was determined by qRT-PCR. As shown in Figure 1f, ACTA1 and MYH1 mRNA increased 3 days after transfection. We assumed this increased expression was a consequence of increased miR-206 levels. We chose 48 h and earlier post transfection with miR-206 mimic to identify direct miR-206 targets that relieve the differentiation blockade before the genetic or morphologic features of differentiation were evident.
Identification of miR-206 target genes. Next, we sought to identify the miR-206 target genes responsible for relieving the RMS differentiation arrest using a combinatorial approach with RNA microarrays, proteomic profiling and target prediction algorithms. RD cells were transfected with NC or miR-206 mimic and RNA was prepared after 48 h and whole-cell lysates were isolated after 24 and 48 h. Gene expression profiling measured the expression of over 24 000 genes and identified 245 genes with increased expression and 165 genes with decreased levels after transfection with miR-206 with a P-value o0.05 and decreased expression with Log 2 ratio (206/NC)o − 0.58 (Figure 2a and Supplementary Figures S2a and b). The top 20 genes with increased expression and the top 30 genes with decreased expression are illustrated in Figure 2b (GEO submitted).
Similarly, out of 8337 total proteins quantified by tandem mass tag (TMT) proteomic profiling, 191 proteins had increased and 272 proteins had decreased levels with a P-value of o0.05 and at least a 1.3-fold change after 48 h (Figure 2c). The top 20 proteins with increased levels and top 30 with decreased levels are shown in Figure 2d. Proteomic profiling quantitated 60% (475 of 790) of the predicted miR-206 target genes from TargetScan7.0 27 with 62 proteins having decreased levels following miR-206 transfection. As well, proteomics quantitated 62% (103 of 165) of genes significantly decreased by microarray (Figure 2e). Of the potential miR-206 targets downregulated by both microarray and proteomics, 23 contained predicted miR-206 sites by TargetScan. Gene ontology analysis of these 23 downregulated genes indicated the most significant genes participate in tissue regeneration and developmental growth (PAX7, NOTCH3 and TIMP3; Supplementary Table S1). In contrast, the same analysis on upregulated genes found no overlapping genes and proteins to have miR-206 predicted binding sites suggesting that the downregulated genes are enriched in miR-206 targets (Supplementary Figure S2c). Ninety-seven proteins had decreased levels 24 h after miR-206 transfection and 32 of these are predicted miR-206 targets (Supplementary Figures S2d-f). Previously documented miR-206 target genes not restricted to RMS, including CCND2, G6PD, GJA1, HDAC4, IGFBP5, NOTCH3 and PAX7, were similarly found to be regulated by either proteomics or microarray. 17,19,20,24,28,29 Decreased miR-206 expression has been shown in RMS compared with normal skeletal muscle. 21 Therefore, we hypothesized that the most relevant miR-206 targets for RMS pathogenesis and potentially the differentiation arrest of RMS would be upregulated in human RMS tumors (low miR-206) compared with normal muscle (high miR-206). We compared the expression of the group of 23 genes in a cohort of human FN-RMS tumors 1 to a cohort of human skeletal muscle. All but 2 of the 23 potential target genes had higher expression in FN-RMS tumors compared with skeletal muscle (Figure 2f and Supplementary Figure S2g). A similar pattern was observed comparing FP-RMS tumors with skeletal muscle; however, the ratio for PAX7 and NOTCH3 in FP-RMS compared with skeletal muscle was substantially less (Supplementary Figure S2h).
Validation of putative miR-206 target genes. The 23 potential miR-206 target genes identified as downregulated by both microarray and proteomics as well as IGFBP5, PAX3 and HDAC4 (previously described miR-206 targets in RMS) 17,19,20 were validated by qRT-PCR ( Figure 3a) and a subset by immunoblotting ( Figure 3b). All predicted miR-206 targets were significantly downregulated upon transfection with miR-206. In contrast, EZH1 that does not contain miR-206 recognition sites was unaffected by miR-206 overexpression. Potential targets were validated using 3′ UTR luciferase reporter assays with regulation by miR-206 observed for every target tested compared with the empty psiCHECK2 reporter and the EZH1 3′ UTR negative control ( Figure 3c and Supplementary Figure S3a). Specificity of miR-206 regulation was confirmed with dose-dependence as shown for CCND2, NOTCH3, PAX3 and PAX7 Using a combinatorial approach with microarray, TMT proteomic profiling and bioinformatic prediction algorithm all predicted miR-206 target genes identified validated; however, the relevance of these miR-206 target genes to the RMS differentiation phenotype had not been determined.
Genetic rescue identifies miR-206 target genes required for differentiation. To interrogate miR-206 target genes critical for the differentiation arrest in RMS, we attempted to rescue the miR-206-induced phenotype by expressing mutant target genes lacking the 3′ UTR containing miR-206binding sites with the hypothesis that key targets if unable to be regulated by miR-206 would block miR-206-induced differentiation. We generated lentiviruses to express mutant cDNAs for 9 of the 23 target genes enriched in FN-RMS compared with muscle including the top seven as well as NOTCH3 and G6PD ( Figure 2f). In addition, previously described miR-206 targets, IGFBP5, HDAC4 and PAX3, were stably transduced in RD cells. After selection and validation of cDNA overexpression (Supplementary Figure S4a), cells were transfected with NC or miR-206 mimic. While RD cells transduced with the empty vector control retained decreased target gene expression with miR-206 transfection, RD cells overexpressing target gene cDNA lacking the 3′ UTR were not responsive to miR-206 (Supplementary Figures S4a and b).
Muscle differentiation was assessed with MHC immunocytochemistry (ICC) and a full range of phenotypes was observed from enhanced differentiation with EML4, partial inhibition with CCND2, significant block of differentiation with PAX3, and near complete inhibition of differentiation with NOTCH3 and PAX7 (Figures 4a and b Figure S5d). Although CCND2 partially blocked miR-206-induced differentiation by ICC, CCND2 did not block miR-206-induced CKM expression. G6PD (previously shown to delay differentiation in RD18 cells, a subclone of RD) 24 did have a subtle but significant decrease in CKM expression but no detectable difference in MHC expression by ICC (Supplementary Figures  S5a-d). These results suggest that downregulation of PAX7, PAX3, CCND2 and NOTCH3 is necessary for the miR-206induced differentiation of FN-RMS cells.

and Supplementary
During normal myogenesis, induction of the differentiation program is coupled with a terminal cell cycle exit. We found partial rescue of Ki67 levels in miR-206 mimic-transfected cells with CCND2, G6PD, KIF2A and PAX3 mutant cDNA overexpression ( Supplementary Figures S5a and c). Overexpression of the mutant PAX7 cDNA was the only miR-206 target that completely blocked miR-206-induced decreased proliferation (Figures 4e and f and Supplementary Figures S5a  and c). NOTCH3-ICD overexpression alone decreased proliferation of RD cells compared with empty control. Paradoxically, miR-206 mimic in NOTCH3-ICD-overexpressing cells increased proliferation. Given the proliferative phenotype with NOTCH3-ICD expression in the absence of miR-206, we focused on PAX7. In order to ensure that the observed phenotypes with mutant PAX7 were not just a result of supraphysiologic protein overexpression, the viral titer was decreased to reduce the protein level of exogenous mutant PAX7 (Supplementary Figure S6a). Stably transduced cells with the lowest titer (a 40-fold reduction) were then transfected with NC or miR-206 mimic. Similar to the original observation even with a subtle increase in PAX7 expression, differentiation was significantly reduced and proliferation was unchanged ( Supplementary Figures S6b-g). Insufficiency of miR-206 targets for differentiation induction. Because PAX7, PAX3, CCND2 and NOTCH3 downregulation was necessary for miR-206-induced differentiation, we sought to determine if knocking down these genes alone or in combination with siRNAs could phenocopy the miR-206induced differentiation of RMS. Three siRNAs for each gene were assessed and significant knockdown by qRT-PCR and immunoblotting was observed (Supplementary Figures  S8a and b). The best two siRNAs, designated A and B for each gene, were then assessed for their effect on RD cell differentiation (Figures 6a-d). Modest increases in MHC expression were observed with the best siRNAs alone and in a pooled combination of the best siRNA for each of the four genes but not to the same extent as miR-206 mimic (Figures 6c and d). Despite strong knockdown comparable to the downregulation observed with miR-206 mimic, only siRNA knockdown of NOTCH3, PAX3 and PAX7 resulted in a significantly increased MHC staining. Another independent group of siRNAs individually and in a pooled combination resulted in slight increases in MHC and differentiation index with only knockdown of PAX3 and PAX7 alone reaching statistical significance ( Supplementary Figures S8c and d). Thus, targeted knockdown of miR-206 targets CCND2, PAX3, PAX7 and NOTCH3 with siRNAs alone or in combination was insufficient to phenocopy miR-206-induced differentiation. However, knockdown of these genes was sufficient to decrease proliferation as measured by Ki67 ICC (Figure 6c and Supplementary Figures S8c and e), suggesting these targets may be contributing to maintaining a high proliferation rate of FN-RMS.

miR-206 participates in FN-RMS pathogenesis in vivo.
To determine if miR-206 contributes to the pathogenesis of RMS in vivo, we utilized miR-206 loss of function mice 17 in a previously described mouse model of FN-RMS. 32 We  Figure S9a). miR-206 KO mice had a significant decrease in latency to tumor onset with a median onset of 24 days compared with 34 days in miR-206 WT animals (Figure 7a). In addition, the penetrance of tumors increased to 92% in miR-206 KO   with 85% in WT animals. As expected, miR-206 KO animals expressed no detectable miR-206 and heterozygous miR-206 mice expressed an intermediate level of miR-206 ( Figure 7b). The frequency of multiple primary tumors per animal increased with miR-206 loss, suggesting miR-206 loss lowered the threshold for tumor formation (Figure 7c). The histology and IHC staining pattern as well as the expression of myogenic genes in miR-206 KO tumors were indistinguishable from the miR-206 WT tumors in regard to these myogenic features (Figure 7d and Supplementary Figures S9b-d). Interestingly, miR-206 target genes Notch3 and Pax7 were significantly upregulated in miR-206 KO tumors compared with WT tumors, thus confirming the in vitro target gene regulation identified in human RD cells (Figure 7e). Furthermore, Pax7 protein levels were increased in miR-206 KO tumors by immunofluorescent staining and immunoblot (Figures 7f-h). These results indicate that miR-206 loss accelerated FN-RMS tumor development, and miR-206 actively participates in the biology of FN-RMS.

Discussion
In this study, we demonstrate that miR-206 overexpression with a miRNA mimic led to induction of muscle differentiation coupled with decreased proliferation in FP-RMS and FN-RMS cell lines. An orthogonal approach using a combination of gene arrays, TMT proteomic profiling and miRNA target prediction algorithms identified a subset of 23 genes regulated by miR-206 at the RNA and protein levels. We also established the in vivo role of miR-206 in a genetically engineered mouse model of FN-RMS where miR-206 deletion led to an increase in penetrance, number of primary tumors and decreased tumor-free survival. The robust phenotype of miR-206induced differentiation offers an opportunity to comprehensively explore necessary miR-206 target genes, and thus define the roadblocks to myogenic differentiation in RMS. This study and that of others provide evidence for a model where miR-206 promotes cell cycle exit and differentiation through a collection of targets, including cell cycle regulators, myogenic transcription factors, cytoskeleton proteins as well as proteins involved in metabolism, signaling and epigenetic modification (Figure 8).
Our studies directed focus to CCND2, NOTCH3, PAX3 and PAX7 as key miR-206 targets in FN-RMS. CCND2 is the predominant D-type cyclin in FN-RMS. 33 Furthermore, miR-206 downregulates CCND2 in both FN-and FP-RMS potentially promoting cell cycle exit. 20 NOTCH signaling is also increased in RMS and NOTCH inhibition promotes RMS differentiation. 34  progenitors and function upstream of Myf5 and MyoD1 to promote myoblast and satellite cell proliferation and selfrenewal. 7 PAX3/PAX7 are normally downregulated in myogenic differentiation and their exogenous overexpression inhibits differentiation. 18,37,38 In addition, mouse satellite cells marked with higher expression of Pax7-nGFP are less likely to undergo myogenic commitment. 39 PAX3 and PAX7 are overexpressed in FN-RMS and are the targets of reciprocal chromosome translocation with FOXO1 that drives FP-RMS. In addition, PAX3-3′ UTR variants have been identified in RMS and different satellite cell populations, implicating the importance of regulation by miR-206. 20,40 Recent investigations of PAX7 in RMS have revealed that PAX7 promotes proliferation through downregulation of MyoD1 in a proteasome-dependent manner. 11 PAX7 promotes metastasis and invasiveness of RMS and pleomorphic sarcomas. 10,41 Surprisingly, the precise role of miR-206 regulation of PAX7 in FN-RMS has remained largely unexplored.
PAX7 was able to significantly inhibit both differentiation and decreased proliferation in response to miR-206 overexpression highlighting the dual roles of miR-206. While induction of differentiation is obviously critical for normal myogenesis, decreased proliferation and cell cycle exit are also important and likely contribute to the tumor-suppressing potential of miR-206. This study highlights the role of PAX7 in RMS. While expression of mutant PAX7 without miR-206-binding sites  9 Therefore, both the PAX3-and PAX7-FOXO1 lack the 3′ UTR of the PAX gene containing miR-206 recognition sites. This results in PAX3/7 target gene expression driven by the fusion protein uncoupled from miR-206 regulation. Despite the distinct difference in drivers of tumorigenesis, miR-206 is able to promote differentiation and cell cycle exit in both subtypes of RMS illustrating that miR-206 RMS phenotypes are secondary to more than PAX7 regulation alone.
To determine whether miR-206-induced differentiation could be explained by a single target gene, we assessed siRNA knockdown of targets. Gene knockdown of target gene alone or in combination only slightly induced differentiation compared with miR-206 treatment, despite knockdown levels comparable to their miR-206-directed downregulation. Knockdown of these targets reduced proliferation as previously shown. 11,20,24,34,35 This finding suggests that single targets and even a subset of targets are insufficient to induce differentiation to the extent induced by miR-206 overexpression. Thus, target identification collectively is a key to understanding the mechanism of miR-206-induced differentiation in RMS further suggesting the '1 miRNA-1 target' model to be insufficient to explain complex phenotypes induced by miRNAs. 42 Our results further substantiate the hypothesis that because a single miRNA can target hundreds of mRNAs, they may be more valuable therapeutically than directed agents against a single molecule or pathway. With the recent entry of miRNA mimics into clinical use (NCT01829971) as well as the use of miRNAs as prognostic and diagnostic biomarkers, elucidating the collective targets of a therapeutic miRNA is critical for understanding mechanism and identifying targets for efficacy analysis. The current barrier for broad clinical use of miRNA mimics and inhibitors is delivery, yet as these issues are overcome with new delivery strategies the use of therapeutic miRNAs will advance future targeted therapies in cancer.
miR-206 is potentially therapeutically valuable especially in RMS where it can both promote cell cycle exit and differentiation. miR-206 replacement therapy with a miRNA mimic would allow for regulation of these processes by targeting many nodes in the RMS developmental arrest as well as a mechanism to target transcription factors that have been historically difficult to target with small molecules. In addition, miR-206 has been shown to have tumor suppression capabilities in other cancer types such as breast cancer, lung cancer, hepatocellular carcinoma, pancreatic adenocarcinoma and others. 26 These studies provide further evidence for the role of miR-206 in decreasing proliferation independent of myogenic differentiation, suggesting miR-206 mimics may be of value as a broad cancer therapy.  Table S2). 43 Cells were maintained in DMEM (SH30243; HyClone, Logan, UT, USA) supplemented with 10% FBS (HyClone), antibiotic-antimycotic (A5955; Sigma-Aldrich, St Louis, MO, USA) and incubated at 37°C, 5% CO 2 .
Mouse strains. All mice used for this study were maintained in mixed genetic backgrounds; thus, littermate controls were used for all comparisons. All strains were described previously: aP2-Cre, 44 SmoM2 (#5130, Jackson Laboratories, Bar Harbor, ME, USA), 45 miR-206 − / − . 17 All experiments involving animal procedures were reviewed and approved by the SJCRH Institutional Animal Care and Use Committee. Normalized signal data were log 2 transformed in STATA/SE 11.2 (StataCorp, College Station, TX, USA). The log 2 transformed NC and miR-206 data were batch corrected and compared. Each probeset was compared by an unequal variance t-test (Partek Inc., Partek Genomics Suite 6.6, St Louis, MO, USA). Probesets with adjusted P-value less than 0.05, an FDRo0.05 and a fold change42 were Z-score normalized, hierarchically clustered and visualized in Spotfire Decision Site 9.2 (TIBCO, Boston, MA, USA). Previously described cohorts of human RMS 1 and normal skeletal muscle (GEO, GSE9103) were compared by taking the RMS/wt-skeletal muscle log ratios median normalized to center at zero.

Differentiation
TMT proteomics profiling by liquid chromatography and mass spectrometry (LC-MS/MS). Cell pellets from 24 and 48 h after NC or miR-206 mimic transfection were lysed, digested and labeled as described previously. 47 One hundred micrograms of each sample was digested with Lys-C (Wako, Richmond, VA, USA; 1:100 w/w) at room temperature for 2 h, diluted 4 × with 50 mM HEPES and further digested with trypsin (1:50 w/w; Promega, Madison, WI, USA) overnight at room temperature. Peptide samples were acidified by TFA, desalted with Sep-Pak C 18 cartridge (Waters, Milford, MA, USA), eluted by 60% ACN 1% formic acid and dried by speedvac. Samples were resuspended in 50 mM HEPES and labeled with 10-plex TMT reagents (Thermo Fisher Scientific). The labeled samples were equally pooled, desalted and dried by speedvac. Labeled peptides were analyzed based on an optimized LC-MS/MS platform on an Orbitrap Elite mass spectrometer (Thermo Fisher Scientific). 48 Acquired MS data were searched against a human protein database utilizing JUMP algorithm, a tag-based database search program, for protein identification (protein FDRo1%). 48 Quantification was achieved by analyzing the reporter ion intensities from the MS/MS spectra.
Immunoblotting. Immunoblotting was performed on whole-cell extracts as described previously. 46 Antibodies applied overnight as detailed in Supplementary  Table S3