Distinct roles for miR-1 and miR-133a in the proliferation and differentiation of rhabdomyosarcoma cells

Abstract

Rhabdomyosarcoma is the most common soft tissue sarcoma in the pediatric population. As this tumor has an undifferentiated myogenic phenotype, agents that promote differentiation hold particular promise as part of a novel therapeutic approach to combat this type of cancer. In this report, we focus on the contribution of two microRNAs (miRNAs) in rhabdomyosarcomas. Levels of miR-1 and miR-133a are drastically reduced in representative cell lines from each major rhabdomyosarcoma subtype (embryonal and alveolar). Introduction of miR-1 and miR-133a into an embryonal rhabdomyosarcoma-derived cell line is cytostatic, thereby suggesting a tumor suppressor-like role for these myogenic miRNAs. Transcriptional profiling of cells after miR-1 and miR-133a expression reveals that miR-1 (but not miR-133a) exerts a strong promyogenic influence on these poorly differentiated tumor cells. We identify mRNAs that are down-regulated by these miRNAs and propose roles for miR-1 and miR-133a in repressing isoforms of genes that are normally not expressed in muscle. Finally, we show that mRNA targets of miR-1 and miR-133a are up-regulated in rhabdomyosarcomas, suggesting a causative role for these miRNAs in the development of rhabdomyosarcomas. More important, these results point to the promise of enhancing rhabdomyosarcoma therapy using miRNAs as agents that mediate cytostasis and promote muscle differentiation.—Rao, P. K., Missiaglia, E., Shields, L., Hyde, G., Yuan, B., Shepherd, C. J., Shipley, J., Lodish, H. F. Distinct roles for Mir-1 and Mir-133a in the proliferation and differentiation of rhabdomyosarcoma cells.

Rhabdomyosarcomas (RMS) are the most common soft tissue sarcomas in children and are classified into two major subtypes based on histology (1). Embryonal rhabdomyosarcomas (ERMS) occur more frequently and can be associated with both loss of heterozygosity and imprinting (in region 11p15.5) and chromosomal gain (regions of chromosomes 2, 8, 12, and 13) (2). These results implicate multiple genetic mechanisms underlying the development and progression of ERMS. In contrast, a majority of alveolar rhabdomyosarcomas (ARMS) cases can be attributed to the expression of a novel oncogenic fusion protein generated by the translocation of the genomic region encoding PAX3 or PAX7 to that encoding FOXO1(2). Although the exact identity of the cell that transforms to give rise to RMS has not been unequivocally identified, by definition, they show features of myogenic development. For example, RMS cells express skeletal markers, such as myosin and α actin, and also exhibit striations (3). Observations like these have led to the consensus opinion that RMS are tumors that have failed to complete the myogenic program. Consequently, agents that promote myogenic differentiation hold promise as novel therapeutic agents for this type of tumor. Clearly, novel therapeutic approaches are needed for RMS therapy, as there has been little change in 5-yr survival rates in recent decades (SEER Pediatric Monograph on Soft Tissue Sarcomas; http://seer.cancer.gov/publications/childhood/).

MicroRNAs (miRNAs) are small RNA-based regulators that usually repress gene expression by targeting mRNAs (4, 5). Recent evidence has implicated miRNAs as effective modulators of skeletal muscle development and homeostasis (6, 7). miRNAs achieve repression by binding to sequences in the 3′ untranslated regions of target mRNAs, promoting their degradation or inhibiting their ability to be translated. Nucleotides 2–8 (the seed region) of the miRNA are critical in determining target specificity (8). By computing the number of mRNAs with conserved sequences complementary to the seed sequences, it has been estimated that up to one-third of mammalian mRNAs are susceptible to miRNA-mediated regulation (9). Several miRNAs encoded within the mammalian genome exhibit exquisite tissue specificity. Specifically, miR-1-1, miR-1-2, miR-133a-1, miR-133a-2 are expressed almost exclusively in striated muscle tissue (cardiac and skeletal muscle) and in some cases, the reasons underlying such specificity have been described (10,11,12,13,14,15,16). Moreover, miR-133b and miR-206 are exclusively expressed in skeletal muscle and have seed sequences that are identical to miR-133a and miR-1, respectively. Since the seed sequence is critically important in determining target specificity and functionality, skeletal muscle, in essence, expresses three functional copies of miR-1 and miR-133.

We recently analyzed a set of 163 RMS patients for the expression of all muscle-specific miRNAs (unpublished results). This study and others (17,18,19,20) show a lower expression of muscle-specific miRNAs compared to normal skeletal muscle. However, we found their expression in RMS samples to be, on average, higher than that observed in other normal tissues, suggesting their potential involvement in RMS muscle-like phenotypes. In addition, the levels of miR-1, miR-133a were much lower in ERMS than ARMS, both in the cell lines and in primary tumor samples. This is consistent with the notion that the ARMS resemble a point later in the course of myogenic differentiation (21).

In this report, we provide a novel glimpse into miR-133a function and provide additional insights into miR-1 function. By inhibiting the cell cycle and fine-tuning isoform expression, miR-133a clearly promotes muscle cell identity. These results are in contrast to the prevailing view that miR-133a only inhibits muscle differentiation. Furthermore, our transcriptome analyses expand the scope of miRNA-mediated isoform regulation by providing concrete examples of genes that are down-regulated on miR-133a and miR-1 transfection. In concurrence with recently published data on RMS cell lines (17), we observe that miR-1 also exerts an instructive role in promoting myogenic differentiation. Collectively, our data suggest that miR-1 and miR-133a have both similar and distinct functions in promoting myogenic differentiation. Furthermore, by profiling miR-133a- and miR-1-induced changes in the transcriptome of RMS cells, we identify putative target mRNAs that are repressed directly by miR-133a and miR-1. Because miR-133a and miR-1 are down-regulated in RMS, we hypothesized that the high levels of expression of some miR-133a and miR-1 target genes in RMS could be a direct consequence of the low levels of miR-1 and miR-133a expression in these tumors (unpublished results and refs. 17,18,19,20). By intersecting our miR-1 and miR-133a target set with the set of genes overexpressed in RMS (relative to normal skeletal muscle), we identify a miRNA-regulated gene set that could contribute to RMS genesis. Further investigation into this gene set can provide mechanistic insights into the role of miR-1 and miR-133a in the prevention of tumorigenesis and, in conjunction with other studies (17, 18, 20), serve as guides in assessing the potential therapeutic applications of miRNAs in RMS.

MATERIALS AND METHODS

miRNA analysis

The ERMS (RD) and ARMS (Rh30) cell lines were obtained from the American Type Culture Collection (ATCC; Manassas, VA, USA) and Dr. Peter Houghton (St. Jude Children’s Research Hospital, Memphis, TN), respectively, and maintained in growth medium (GM; DMEM supplemented with 10% FCS, antibiotics and glutamine). For the Taqman low-density array (TLDA) described in Fig. 1A, RNA was extracted from nonconfluent plates grown in GM. Multiplex RT reactions and singleplex PCR reactions were performed as suggested by the manufacturer (Applied Biosystems, Foster City, CA) for the miRNA TLDA. Software provided by the manufacturer was employed for detecting differentially expressed miRNAs. For ΔC_t measurements presented in Supplemental Table 1, C_t values obtained for RNU48 were used for normalization across the three samples. A detailed description of the analyses of the primary tumor is described elsewhere (unpublished results). Briefly, RNA from pathologically reviewed RMS tumors (66 ERMS and 90 ARMS) and 13 normal skeletal muscle samples were subject to RT-PCR analyses. Ethical approval was obtained as outlined in Local Research Ethics Committee protocol No. 1836, Multi-Regional Research Ethics Committee/98/4/023 and, where required, consent had been obtained.

Figure 1.

Levels of miRNAs in representative RMS cell lines and correlation of miR-1 and miR-133a expressions in RMS. A) An miRNA-TLDA array was used to perform semimultiplexed RT and individual q-PCR reactions on 365 miRNAs. Scatterplot comparing ΔC_t values obtained from normal skeletal muscle (x axis) and alveolar (Rh30, blue dots, y-axis) and embryonal (RD, red squares, y axis) cell lines. ΔC_ts were obtained by subtracting the miRNA C_t value from the RNU48 C_t value. For clarity, data point symbols are exaggerated for the myogenic miRNAs (miR-1, miR-133a, miR-133b, and miR-206), and the origin is set to nonzero values. Low expression values are characterized by low ΔC_ts that fall to the left and the bottom portions of the graph. Conversely, high expression values are characterized by high ΔC_ts (including negative values) that fall to the right and the top portions of the graph. Long gray arrows next to the axes indicate this trend. If the level of a particular miRNA is the same in the cell line and skeletal muscle, then the symbol corresponding to that miRNA would fall on the diagonal line. Raw C_t values, as well as ΔC_t values, are also presented in Supplemental Table 1. Data obtained represent single determinations for each miRNA in a given RNA sample extracted from RD, Rh30, or skeletal muscle tissue. B) Levels of miR-1 and miR-133a were tested individually in ERMS (n=66), ARMS (n=90), and skeletal muscle samples (n=3). ΔC_t values (ΔC_tmiR-1,x axis; ΔC_tmiR-133a, y axis) were plotted as a scatterplot. ΔC_ts were obtained by subtracting the miRNA C_ts from the average expression of two endogenous control (RNU48 and RNU6B) C_ts. A highly statistical significant correlation was observed (Pearson correlation 0.91, CI 95% 0.88–0.93; P<0.001) between levels of miR-1 and miR-133a, demonstrating coordinate expression of these two miRNAs in normal and transformed tissue.

Cell culture

For differentiation assays described in Fig. 3, RD cells were initially plated in GM in 6-well plates at 7.5e4 cells/plate. Cells were transfected (see below) the next day in growth medium, and 24 h after transfection, the medium was again changed to fresh growth medium. Twenty-four hours after this point was considered “day 0” (d0) and hence represents a point at which the RD cells were exposed to miRNAs for 48 h. Differentiation was induced starting at d0 by replacement of growth medium with differentiation medium [DM; 10% FCS with 2% horse serum (HS) supplemented with 100 nM TPA, according to the previously published protocol for RD cells; ref. 22]. Differentiation medium was added to a parallel set and harvested 24 h later (d1). A similar protocol was followed for d2 and d3 RNA isolation. During these experiments, differentiation medium was changed every 24 h to ensure that nutrient depletion was not a complicating factor in our analyses. Real-time primers used for quantifying MyL1 and myogenin have been described previously (23); GAPDH primers were 5′-gggtgtcgctgttgaagtca-3′ and 5′tgggctacactgagcaccag-3′.

Figure 2.

Growth inhibition by miR-1 and miR-133a. A) ¹⁴C-thymidine incorporation during a 48-h time period was measured after transfection of miR-1 and miR-133a mimics into RD cells (open bars) and Rh30 cells (solid bars). Comparisons (one sample t test) were made to the amount of ¹⁴C-thymidine incorporation that resulted when these same cells were transfected with a control mimic (value set to 100%). P values (shown above bars) were significant for RD cells transfected with miR-1 and miR-133a. B) Representative example of a real-time thymidine incorporation experiment in RD cells that was used to obtain the data in A. C) Cell death (as measured by live-cell PI inclusion) in RD cells (open bars) and Rh30 cells (solid bars) after miR-1 and miR-133a transfection. Comparisons (one sample t test) were made to the amount of cells that stained positive for PI when these same cells were transfected with a control mimic (whose value was set to 1). P values (shown for RD cells) were not significant for either the RD or Rh30 cells after miR-1 or miR-133a transfection. D) Cell cycle analysis was performed using PI staining in fixed RD (left) and fixed Rh30 cells (right) 48 h after control, miR-1, and miR-133a transfections. P values are shown for RD cells after performing t tests comparing the G1, S, and G2/M percentages in the miR-1- or miR-133a-transfected population and the G1, S, and G2/M percentages in the control-transfected population. No P values are shown for Rh30 cells in A, C, D, as none of the t tests showed significant (P<0.05) differences.

Figure 3.

Promyogenic effects of miR-1 in RD cells. Expression levels of myogenin (left panel) and skeletal myosin light chain (MyL1-right panel) were assessed after transfection of the indicated miRNAs and during a differentiation time course. Transfections of miRNAs, as indicated, were performed 2 d prior to initiation of differentiation. Differentiation was carried out for 3 d, and samples were harvested at d1–3 after commencement of differentiation. Day-0 samples were incubated only in growth medium. miR-1 promotes the precocious expression of differentiation markers in the absence of differentiation cues (compare levels at d0) and dramatically enhances the levels of these markers during differentiation. miR-133a consistently inhibits the expression of these markers. One representative experiment (of 3) is shown.

Thymidine incorporation assays were carried out using Cytostar-T scintillation plates (GE Healthcare-Biosciences Corp, Piscataway, NJ) using the protocol provided by the manufacturer. Briefly, miRNA-transfected cells were plated 1 d after transfection into 96-well Cytostar-T plates (7500/well). After allowing the cells to settle overnight, ¹⁴C-labeled thymidine was added, and background readings were obtained using a Wallac β counter (Perkin-Elmer, Waltham, MA, USA). Subsequent readings were taken at 24-h intervals. For cell cycle analysis, cells were fixed 2 d after transfection with miRNAs, and the DNA content was analyzed using propidium iodide (PI). The percentage of cells in each phase was estimated on cells within the 2n–4n window using Flow-Jo (Tree-Star Inc, Ashland, OR, USA). For cell death analyses, live cells (2 d post-transfection) were stained live using PI inclusion to quantify dead cells. All experiments measuring cell growth, cell death, and cell cycle distribution were carried out in growth medium.

Array experiments

For microarray analyses, we performed 3 biological replicates for each transfection (miR-1, miR-133a, and control) and harvested RNA that corresponded to the time point listed as d0 above. HG U133 Plus2 microarrays from Affymetrix (Santa Clara, CA, USA) were processed with robust multichip average (RMA) algorithm, and absent/present calls for each probe set were determined with the standard Affymetrix algorithm, both as implemented in Bioconductor. Probe sets that were absent in all samples were removed for subsequent analysis. Differential expression was determined by a moderated t test with the “limma” package in R (corrected for false discovery rate) as <0.005. Where a gene was represented by multiple probe sets (based on annotation from Affymetrix), average of gene expression log ratios was used; 545 of 584 genes predicted as miR-1 targets by TargetScan 5.1 were represented on the array. Similarly, 457 of 503 genes predicted as miR-133 targets by TargetScan 5.1 were represented on the array. Gene Ontology (GO) analysis was performed using Database for Annotation, Visualization and Integrated Discovery (DAVID) tools (http://david.abcc.ncifcrf.gov/). The lists of probes found differentially expressed were converted to unique Entrez GeneIDs and use as identifier for the analysis. The background for the comparison was generated using all the probe sets retained for the differentially expression analysis, also converted to their unique Entrez GeneIDs. Functional annotation charts were generated using default settings. Expression profiling of 132 primary RMS tumors and 32 normal skeletal muscle samples was obtained from publicly available databases and analyzed as described previously (24) using the samr package. A total of 2667 and 1365 unique genes were found to be significantly overexpressed or underexpressed compared to skeletal muscle, showing a false discovery rate equal to 0. These lists overlapped with the lists of genes found differentially expressed on miR-1 and miR-133a transfection in the RD cell line. The significance of the number of genes found in common was tested using a bootstrapping method. In detail, a random set of unique genes of equal size was generated sampling from the full list of genes contained in the Affymetrix chip. The overlap between this randomly generated list and the list of genes found differentially expressed in RMS vs. normal skeletal muscle was recorded. This procedure was repeated 1000 times to estimate a random distribution. The overlap experimentally observed was then compared vs. its random distribution to test its significance.

RESULTS

miR-1/133a levels in tumors/cell lines

In order to assess the relevance of miRNAs in the rhabdomyosarcomagenesis, we interrogated the levels of 365 miRNAs in RMS-derived cell lines and skeletal muscle using quantitative PCR in an array format (Taqman low-density array) (Supplemental Table 1). Levels of myogenic miRNAs (mir-1, miR-133a, miR-133b, and miR-206) in a representative cell line for each type of RMS (RD, embryonal; Rh30, alveolar) are drastically reduced (Fig. 1A) when compared with normal skeletal muscle. Note that 3 skeletal muscle-specific miRNAs—miR-1, miR-133, and miR-206—are shown as exaggerated symbols in Fig. 1, blue dots denoting expression in Rh30 cells relative to skeletal muscle and red squares denoting relative expression in RD cells.

As a single-cycle difference in ΔC_t represents a 2-fold difference in initial RNA values, the fold change in miR-1 and miR-133a/b levels between cell lines and tissue varied from 2^15.94 (comparing miR-1 levels in RD vs. skeletal muscle) to 2^6.3 (comparing miR-133b levels in Rh30 cells vs. skeletal muscle). Although there were multiple miRNAs that exhibited differences in expression level, none were as dramatic as miR-1 and miR-133a, and we chose to focus on these. The reduced levels of expression of miR-1 and miR-133a in the cell lines was similar to that found in our previous study of 163 primary RMA samples (unpublished results). We were also interested in determining the correlation between miR-1 and miR-133a levels. An analysis of miR-1 and miR-133a expression levels in primary tumors obtained in this previous study reveals that their expression is highly correlative (Pearson coefficient=0.91), suggesting that the coordinate expression of these miRNAs seen in untransformed tissue is also maintained in primary transformed tissue (Fig. 1B). The levels of miR-1 and miR-133a miRNAs were also considerably higher in the alveolar Rh30 cells when compared to the embryonal RD cells.

miR-1/miR-133a inhibits growth of RD cells

The fact that reduced miR-1 and miR-133a levels in RD and Rh30 cell lines were representative of those in primary tumor samples justified our use of these cell lines to determine the effects of miR-1 and miR-133a in RMS. Accordingly, we transfected miR-1 and miR-133a into these cell lines and monitored their growth. The levels of mature miR-1 and miR-133a were analyzed after transfection and shown to be comparable to the levels seen in skeletal muscle tissue (data not shown). As seen in Fig. 2A, both miR-1 and miR-133 mimics caused a significant decrease in thymidine incorporation in embryonal RD cells. The difference was already evident 24 h after the addition of labeled thymidine (as shown for a representative experiment in Fig. 2B) and persisted for another day. In contrast to what was observed in RD cells, miR-1 and miR-133 transfection did not lead to a decrease in thymidine incorporation in the alveolar-derived Rh30 cells. (Fig. 2A). Since neither miR-1 nor miR-133a caused a significant increase in PI uptake in live cells (note values of P>0.05; Fig. 2C), cell death is unlikely to be the cause for the reduction in thymidine incorporation. Conversely, analysis of DNA content following miRNA transfection implicates G1-S arrest as a probable cause for the miR-1-mediated growth inhibition in RD cells (Fig. 2D). miR-1 transfection caused an increase in the G1 population and a corresponding decrease in percentage of cells transiting through S phase. miR-133a did not cause a significant change in the steady state population of RD cells in any phase of the cell cycle, although the increase in percentage of G1-phase cells after miR-133a transfection did approach significance (P=0.08). Hence, we conclude that miR-1 induces a specific G1-S arrest, whereas the growth-inhibitory effects of miR-133a are not specific to any phase of the cell cycle. The conclusion we have drawn regarding miR-1-induced cell death, is in contrast to the one reached by the authors of a recently published study (20). However, we note that the amount by which apoptosis increased in the report by Yan et al.(20) is nominal (11–12% increase over control transfection), and hence, our differences could be explained by subtle differences in cell line variation, culture conditions, or confluency of cells. The effect of miR-1 on cell cycle arrest is much more robust and has also been demonstrated by Yan et al.(20), and we favor G1-S arrest as a more plausible explanation for the decrease in thymidine incorporation. Consistent with the lack of an effect on thymidine incorporation, neither cell death nor cell cycle progression was affected by miR-1 or miR-133a in Rh30 cells.

miR-1 promotes ectopic muscle gene expression in RD cells

In order to assess whether miR-1 and miR-133 also promote differentiation in RD cells, we assessed the levels of two well-established markers of differentiation after individual introduction of miR-1 and miR-133a. If miR-1 and miR-133a are merely permissive, then we anticipated an enhancement of differentiation only under conditions conducive for differentiation. Alternatively, if miR-1 and miR-133a are sufficient to promote differentiation (i.e., their role is instructive), we expected to see an increase in myogenic marker expression, even in the absence of the differentiation medium.

The results shown in Fig. 3 clearly reveal that overexpression of miR-1 can play an instructive role and promote differentiation. In the absence of differentiation medium (time point 0 on x axis), the levels of myogenin and MyL1 are ∼5 times higher in miR-1-transfected cells when compared with the control-transfected cells. The effect is specific to miR-1 because the other miRNAs (control and miR-133a) did not have a similar effect. Moreover, the presence of miR-1 causes a dramatic enhancement in the levels of these markers during differentiation. These experiments suggest that miR-1 can exert a strong promyogenic influence on the poorly differentiated RD cells. In contrast to miR-1, transfection of miR-133a led to a reproducible decrease in marker gene expression. It is unlikely that the decrease in marker gene expression seen in miR-133a-transfected cells is due to nonspecific toxicity, as we did not detect significant differences in PI staining (Fig. 2C).

Effects of miR-1/miR-133a on genome-wide transcript expression

Global analyses of miR-1 and miR-133 targets revealed that the introduced miRNA mimics did specifically down-regulate mRNAs bearing sites complementary to the respective seed sequences (Fig. 4). Expression ratios (log₂ ratios obtained by comparing the intensity after miRNA transfection with that obtained after control mimic transfection) for each gene on the array were calculated, and then the genes were parsed into two different categories—those that had a conserved miRNA binding site (as defined by TargetScan 5.1) and those that did not. Cumulative distribution (on the y axis) as a function of the expression ratios (on the x axis) for these two groups of genes was plotted separately and is shown in Fig. 4. A shift to the left, for genes with miRNA binding sites (black line), compared to a normal distribution for genes without miRNA binding sites (gray line) indicates that the presence of a miRNA binding site makes it much more likely that a gene will be repressed (P=2.2×10⁻¹⁶ by 1-sided Kolmogorov-Smirnov test for both miR-1 and miR-133a targets). Of 545 genes with conserved binding sites for miR-1, 404 genes showed reduced expression after miR-1 mimic transfection. Similarly, out of 457 genes with conserved binding sites for miR-133, 308 genes were down-regulated by miR-133a. mRNAs without binding sites for miR-1 and miR-133 were also down-regulated, but the above analyses reveal that a mRNA with a seed match is much more likely to be repressed in the presence of the cognate miRNA (miR-1 or miR-133a).

Figure 4.

Global down-regulation of miR-1 and miR-133 targets by the respective miRNAs in RD cells. Levels of mRNAs bearing miR-1 target sites (left panel, black line) and miR-133 sites (right panel, black line) as defined by TargetScan 5.1 were compared to mRNAs that did not bear seed matches (gray lines, both panels). These studies reveal that miR-1 and miR-133 promote a global down-regulation of their cognate targets in RD cells. All types of sites (8-mer, 7mer-m8, and 7mer-1A; for definitions, see http://www.targetscans. org) were used for these analyses.

A list the 50 most down-regulated mRNA targets provides a useful starting point for further investigation (Supplemental Table 2). On close manual inspection of the 50 most down-regulated targets for both miR-1 and miR-133a, we identified several genes that can be classified into two groups (Table 1). First, a number of genes that are repressed by miR-1 or miR-133a have other isoforms that are highly expressed in striated muscle (skeletal muscle or heart). We grouped these genes under “isoform regulation.” For example, CAP1 is repressed by miR-1, and CAP2 is a related isoform that is highly expressed in skeletal and cardiac muscle. This type of role for miRNAs, which we term isoform regulation, has been suggested before (25). Second, we note that a number of genes that are down-regulated have functions related to actin—as direct actin-binding proteins or as proteins whose substrate is actin (see Table 1); in vivo observations made in zebrafish suggest a role for miR-1 and miR-133a in the organization of sarcomeric actin (26). Notably, these genes could be direct targets, as they possess sites matching either miR-1 and/or miR-133 seed sequences.

TABLE 1.

Manually identified categories of genes identified by extraction of the top 50 downregulated genes in miR-1- and miR-133a-transfected RD cells

Isoform regulation		Actin related
Target repressed	Isoform expressed in striated muscle
TPM 4 (miR-133a)	TPM1, TPM3	TWF1 (miR-1)
MYH9 (miR-133a)	MYH1, MYH2, MYH3, MYH4, MYH6, MYH7	MSN (miR-133a)
BTBD3 (miR-133a)	BTBD1, BTBD2	CNN2 (miR-133a)
CAP1 (miR-1)	CAP2	MYH9 (miR-133a)
TXLNA (miR-133a)	TXLNB	TPM 4 (miR-133a)
CAPN5 (miR-133a)	CAPN3, CAPN6	LASP1 (miR-1 & miR-133a)
TWF1 (miR-1)	TWF2	SRXN1 (miR-1)
LASP1 (miR-1 & miR-133a)	LASP2	TAGLN2 (miR-1 & miR-133a)
SMARCD1 (miR-133a)	SMARCD3	CORO1C (miR-133a)
		CAP1 (miR-1)
		XPO6 (miR-1)

Table lists repressed genes involved in isoform regulation and the miRNAs that repress them (in parentheses), the corresponding isoforms of the genes that are expressed in striated muscle, and actin-related genes and miRNAs (in parentheses). Individual lists comprising the 50 most significantly (P≤0.005) downregulated targets for each miRNA (with their fold changes) can be found in Supplemental Table 2.

Histone acetylation/deacetylation and other chromatin modifications are epigenetic modifications that can lead to widespread changes in gene expression. In the list of the top 50 down-regulated targets for miR-1/miR-133a are two epigenetic regulators, HDAC4 and SMARCD1. HDAC4 plays a direct role in muscle differentiation by repressing myogenin induction and MEF2C function (27, 28) and has been previously characterized as a miR-1 target (6). SMARCD1 is a member of the SWI/SNF family of chromatin modifiers with ATPase activity; chromatin remodeling has also been shown to be important in regulating bHLH factors that are important for muscle differentiation (29, 30). Our results with HDAC4 manipulation suggest that pathways that incorporate multiple miRNA targets may be necessary to explain the phenotypic changes caused by a single miRNA (see Discussion).

GO analyses

Analyses of expresssion profiling data identified 568 and 498 unique genes that were up-regulated after miR-1 and miR-133a transfection, respectively (adjusted P < 0.005). Similarly, 594 and 288 unique genes were down-regulated after miR-1 and miR-133a transfection, respectively (adjusted P<0.005). Data mining analysis performed using the web tool DAVID (31, 32) showed that the genes induced by the miR-1 mimic were significantly enriched in terms associated with skeletal muscle structure and development (Supplemental Table 3). A significant number of down-regulated genes were located at the cell membrane and are involved in cell communication and morphogenesis. Although 196 and 69 genes were up-regulated and down-regulated, respectively, by both miR-1 and miR-133a ectopic expression, data mining analysis of genes altered by miR-133a showed a less clear structure. There was no evidence that the genes induced by miR-133a had a global role in establishing myogenic identity (Supplemental Table 3). We also performed GO analyses on the sets of genes that were similarly regulated (up or down) by both miR-1and miR-133a, and there was not a single category that appeared significantly enriched after adjusting for multiple testing (data not shown).

Comparison with primary RMS expression profile

In order to assess the hypothesis that the loss of miR-1 and miR-133a have a role in RMS tumorigenesis by altering gene expression, we analyzed publicly available expression profiles of primary RMS samples (33). Our hypothesis was tested by superimposing our list of genes up- or down- regulated by miR-1 and miR-133a with those found differentially expressed between 132 primary tumors and 32 normal skeletal muscle samples (see ref. 24). To verify the significance of these overlaps, we compared these numbers with those expected by chance using a bootstrapping method. Table 2 summarizes the comparison. Notably, more than one-third of the genes down-regulated on miR-1 or miR-133a transfection in RMS cells were also overexpressed in RMS tumors, a percentage far higher than that expected by chance (P<0.001). To complement this observation, the overlap of this gene set was lower than expected when compared with genes underexpressed in RMS. More detailed information about the genes can be found in Supplemental Tables 4 and 5. Conversely, the gene set induced by transfection of the miR mimics was enriched in mRNAs down-regulated in RMS compared to normal skeletal muscle. Overall, the data in Table 2 suggest that expression of these miRNAs is able to revert the tumorigenic behavior of the cells. In Table 3, we list predicted target genes that are down-regulated by miR-1 or miR-133a (in RD cells) and are up-regulated in RMS relative to normal muscle. We regard these genes as prime candidates whose actions are critical in mediating the antitumorigenic properties of miR-1 and miR-133a, as they 1) are repressed by miR-1 or miR-133a; 2) are up-regulated in RMS, and 3) are predicted targets of miR-1 or miR-133. In particular, we note the presence of PDGF-a in this set, as PDGF signaling has been previously implicated in RMS etiology (34).

TABLE 2.

Overlap between the genes altered by miR-1 and miR-133a transfection in RD cell line and genes found differentially expressed in RMS (primary tumors)

miR mimic	Gene	UP in RMS		DW in RMS
		N test	N exp	N test	N exp
Down-regulated genes
miR-1	594	205 (34%)*	108 (84–142)	22 (4%)*	50 (32–70)
miR-133	288	111 (38%)*	52 (33–75)	11 (4%)*	24 (12–40)
Up-regulated genes
miR-1	568	107 (19%)	103 (79–134)	84 (15%)*	48 (29–69)
miR-133	498	92 (18%)	90 (63–118)	66 (13%)*	42 (24–62)

Genes found up-regulated and down-regulated in RD cells after miRs mimic transfection. Expression profiling of 132 primary RMS tumors and 32 normal skeletal muscle samples obtained from publicly available databases were analyzed by SAM test. A total of 2667 and 1365 unique genes were found to be significantly overexpressed or underexpressed compared to skeletal muscle, showing a false discovery rate equal to 0. The total number of genes found differentially expressed upon miR-1 and miR-133a transfection in the RD cell line (Gene column) was overlapped with the list of 2667 genes found up-regulated in RMS (UP in RMS) or 1365 genes down-regulated in RMS (DW in RMS) compared to skeletal muscle samples. N test columns for UP in RMS and DW in RMS represent the number of overexpressed and underexpressed genes in common, respectively, obtained by such comparison; percentage (in parentheses) was computed considering the total number of genes altered by the specified mimic (that is, divided by the number in the Gene column). The significance of the number of genes found in common was tested using a bootstrapping method by sampling 1000 random sets of unique genes of equal size from the full list of genes contained in the Affymetrix chip. The overlap between this randomly generated list and the list of genes found differentially expressed in RMS vs. normal skeletal muscle observed at each sampling was recorded, and a random distribution was estimated. The overlap experimentally observed was then compared vs. its random distribution to test its significance. N exp columns indicate the median number and range (in parentheses) of genes expected by chance using the above-mentioned bootstrapping method.

^*P < 0.001.

TABLE 3.

Genes found down-regulated after 48-h mimic transfection in RD cell lines, which were also up-regulated in RMS tumors and are predicted targets for miR-1 and miR-133

Gene symbol	Gene name	Entrez Gene ID	Fold change	P value
Genes for miR-1
ADAR	Adenosine deaminase, RNA-specific	103	0.4	0.001
ANXA2	Annexin A2	302	0.5	0.000
ZFP36L1	Zinc finger protein 36, C3H type-like 1	677	0.6	0.002
CCND2	Cyclin D2	894	0.4	0.000
CNN3	Calponin 3, acidic	1266	0.6	0.004
DHX15	DEAH (Asp-Glu-Ala-His) box polypeptide 15	1665	0.7	0.004
E2F5	E2F transcription factor 5, p130-binding	1875	0.4	0.000
GJA1	Gap junction protein, α 1, 43 kDa	2697	0.3	0.001
H3F3A /// H3F3B	H3 histone, family 3B (H3.3B)	3021	0.5	0.000
HNRNPU	Heterogeneous nuclear ribonucleoprotein U (scaffold attachment factor A)	3192	0.4	0.001
KIF2A	Kinesin heavy chain member 2A	3796	0.5	0.001
LASP1	LIM and SH3 protein 1	3927	0.4	0.002
CAPRIN1	Cell cycle associated protein 1	4076	0.7	0.003
MTX1	Metaxin 1	4580	0.3	0.000
MYLK	Myosin light chain kinase	4638	0.4	0.001
PDGFA	Platelet-derived growth factor α polypeptide	5154	0.6	0.003
PFTK1	PFTAIRE protein kinase 1	5218	0.5	0.004
PPIB	Peptidylprolyl isomerase B (cyclophilin B)	5479	0.7	0.004
PPP2R1B	Protein phosphatase 2 (formerly 2A), regulatory subunit A, β isoform	5519	0.5	0.000
TWF1	Twinfilin, actin-binding protein, homolog 1 (Drosophila)	5756	0.2	0.000
RRBP1	Ribosome binding protein 1 homolog 180 kDa (dog)	6238	0.6	0.001
SFRP1	Secreted frizzled-related protein 1	6422	0.5	0.001
SNAI2	Snail homolog 2 (Drosophila)	6591	0.7	0.003
SMARCA4	SWI/SNF-related, matrix-associated, actin-dependent regulator of chromatin, subfamily a, member 4	6597	0.7	0.001
SOX9	SRY (sex determining region Y)-box 9	6662	0.6	0.001
HSP90B1	Heat shock protein 90 kDa β (Grp94), member 1	7184	0.6	0.003
TRPS1	Trichorhinophalangeal syndrome I	7227	0.4	0.000
ADAM12	ADAM metallopeptidase domain 12	8038	0.5	0.001
SFRS9	Splicing factor, arginine/serine-rich 9	8683	0.3	0.000
JOSD1	Josephin domain containing 1	9929	0.6	0.002
UST	Uronyl-2-sulfotransferase	10090	0.4	0.001
STX6	Syntaxin 6	10228	0.6	0.002
BET1	Blocked early in transport 1 homolog (Saccharomyces cerevisiae)	10282	0.7	0.003
CAP1	CAP, adenylate cyclase-associated protein 1 (yeast)	10487	0.4	0.001
ANP32B	Acidic (leucine-rich) nuclear phosphoprotein 32 family, member B	10541	0.6	0.002
GAS2L1	Growth arrest-specific 2 like 1	10634	0.6	0.003
NUP50	Nucleoporin 50 kDa	10762	0.5	0.000
AKAP11	A kinase (PRKA) anchor protein 11	11215	0.7	0.003
SULF1	Sulfatase 1	23213	0.5	0.001
XPO6	Exportin 6	23214	0.5	0.000
TRIM2	Tripartite motif-containing 2	23321	0.5	0.000
SERP1	Stress-associated endoplasmic reticulum protein 1	27230	0.5	0.000
PDCD4	Programmed cell death 4 (neoplastic transformation inhibitor)	27250	0.5	0.000
EML4	Echinoderm microtubule associated protein like 4	27436	0.5	0.004
SEC61A1	Sec61 α 1 subunit (S. cerevisiae)	29927	0.7	0.003
RNF138	Ring finger protein 138	51444	0.5	0.000
PTPLAD1	Protein tyrosine phosphatase-like A domain containing 1	51495	0.4	0.002
TH1L	TH1-like (Drosophila)	51497	0.5	0.000
GATAD2A	GATA zinc finger domain containing 2A	54815	0.5	0.002
C10orf26	Chromosome 10 open reading frame 26	54838	0.6	0.001
NXT2	Nuclear transport factor 2-like export factor 2	55916	0.4	0.000
MAN1C1	Mannosidase, α, class 1C, member 1	57134	0.3	0.000
POGK	Pogo transposable element with KRAB domain	57645	0.3	0.000
TNS3	Tensin 3	64759	0.6	0.002
C7orf23	Chromosome 7 open reading frame 23	79161	0.6	0.003
C11orf61	Chromosome 11 open reading frame 61	79684	0.6	0.002
TXNDC1	Thioredoxin-related transmembrane protein 1	81542	0.6	0.002
ADPGK	ADP-dependent glucokinase	83440	0.6	0.001
SH3BGRL3	SH3 domain binding glutamic acid-rich protein like 3	83442	0.5	0.003
UNC119B	Unc-119 homolog B (Caenorhabditis elegans)	84747	0.4	0.001
RNF38	Ring finger protein 38	152006	0.4	0.000
HNRNPA3	Heterogeneous nuclear ribonucleoprotein A3	220988	0.6	0.002
Genes for miR-133
TNFRSF10B	Tumor necrosis factor receptor superfamily, member 10b	8795	0.5	0.000
CORO1C	Coronin, actin binding protein, 1C	23603	0.2	0.000
LASS2	LAG1 homolog, ceramide synthase 2	29956	0.4	0.000
TBL1X	Transducin (β)-like 1X-linked	6907	0.5	0.000
MSN	Moesin	4478	0.5	0.000
CTBP2	C-terminal binding protein 2	1488	0.5	0.000
QKI	Quaking homolog, KH domain RNA binding (mouse)	9444	0.3	0.001
TPM4	Tropomyosin 4	7171	0.4	0.001
ARHGDIA	Rho GDP dissociation inhibitor (GDI) α	396	0.5	0.001
ATP6AP2	ATPase, H+ transporting, lysosomal accessory protein 2	10159	0.6	0.001
ELAVL1	ELAV (embryonic lethal, abnormal vision, Drosophila)-like 1 (Hu antigen R)	1994	0.5	0.001
YES1	V-yes-1 Yamaguchi sarcoma viral oncogene homolog 1	7525	0.4	0.001
BTBD3	BTB (POZ) domain containing 3	22903	0.6	0.001
ARL3	ADP-ribosylation factor-like 3	403	0.6	0.001
CDCA8	Cell division cycle associated 8	55143	0.6	0.002
CMTM6	CKLF-like MARVEL transmembrane domain containing 6	54918	0.5	0.002
FTL	Ferritin light polypeptide	2512	0.6	0.003
UBXN7	UBX domain protein 7	26043	0.6	0.003
FBN2	Fibrillin 2	2201	0.6	0.003
AK2	Adenylate kinase 2	204	0.6	0.003
LPGAT1	Lysophosphatidylglycerol acyltransferase 1	9926	0.6	0.003
PITPNB	Phosphatidylinositol transfer protein, β	23760	0.6	0.004
ARFIP2	ADP-ribosylation factor interacting protein 2	23647	0.6	0.004
LASP1	LIM and SH3 protein1	3927	0.5	0.004

Differential expression was determined by a moderated t test selecting probe sets with a value of P < 0.005. This list was overlapped with the list of probe sets found differentially expressed in 132 RMS samples compared to 32 normal skeletal muscle and the list of genes predicted as target for miR-1/206 or miR133 in conserved sites (among mammals) by TargetScan 5.1.

DISCUSSION

Abortive myogenesis is a hallmark of RMS. The reasons for a block in myogenesis are not clear, and we hypothesized that miRNAs may play a role in this process. Two myogenic miRNAs (miR-1 and miR-133a) are severely down-regulated in primary tumors, as well as in cell lines derived from RMS. This clearly raises the possibility that the loss of miRNAs could be causative in nature. Because the decrease is seen in primary tumors (unpublished results and refs. 18,19,20), it is unlikely that the down-regulation of miR-1 and miR133a observed in cell lines is solely a secondary effect due to adaptations required for in vitro growth. Notably, we provide evidence that both miR-1 and miR-133a inhibit the growth of RD cells. Studies in C2C12 cells have demonstrated a promyogenic role for miR-1 and an antimyogenic, proliferative role for miR-133a (6). While in agreement with the former observation regarding miR-1, our data also indicates that miR-133a can inhibit growth. The data presented here and elsewhere (unpublished results and refs. 18,19,20), namely, that miR-133a levels are lower in RMS cell lines and transformed tissue, supports a role for miR-133a in growth inhibition. If miR-133a is growth-stimulatory, then it should be expressed at higher levels in transformed cells and tissue (and it is not). Secondly, miR-133a and miR-1 are cotranscribed (15, 16), and hence, the direction of their regulation should be similar. There is a high degree of correlation between miR-1 and miR-133a levels in RMS samples (Fig. 1B). On the basis of these observations, we conclude that both miR-1 and miR-133a have common roles in the cessation of cell division that accompanies the formation of mature striated muscle tissue.

miR-1 and miR-133a also are important in down-regulating specific genes that are normally expressed at low levels in skeletal muscle. Although a role for up-regulated genes in determining muscle identity is widely recognized, miRNAs may play a significant role in the less appreciated function of down-regulating genes that may be detrimental to mature muscle function. Specifically, we have identified a broad role for miR-1 and miR133a in isoform regulation. A role for isoform regulation was first proposed by Cohen and coworkers (25), and our studies confirm this hypothesis and extend the observation to mammalian systems. Examples for candidate genes exhibiting isoform regulation include miR-1-mediated repression of a nonmuscle isoform of adenylate cyclase associated protein (CAP1) and miR-133a-mediated repression of a nonmuscle isoform of myosin (MYH9) and suggests a common role for miR-1 and miR-133a in promoting muscle differentiation. In addition, and as exemplified by these examples, we note that a number of targets repressed by miR-1 and miR-133 also have another shared feature as actin binders. Prior research has focused on the contrasting roles role of miR-133a and miR-1 in muscle differentiation (6); in contrast, our studies reveal that growth inhibition, isoform regulation, and a (yet to be identified) role in regulation of actin function are shared functions of miR-133a and miR-1 that could cooperate positively to lead to similar phenotypic endpoints that are important for muscle function. A simple model in which miR-133a inhibits the ability of miR-1 to promote differentiation (as defined by the global up-regulation of myogenic markers) fails to take into account the contributions made by mir-133a and miR-1 in inhibiting mRNAs that may be detrimental to muscle function. Hence, we favor a model in which both miR-1 and miR-133a positively regulate striated muscle identity through both shared and distinct mechanisms.

Our results reveal that ARMS and ERMS cell lines exhibit different sensitivities to the reintroduction of miR-1 and miR-133a. Growth of RD, an ERMS-derived cell line, is inhibited by miR-1 and miR-133a; Rh30, an ARMS-derived cell line, does not respond in a similar manner. The basal levels of miR-1 and miR-133a are higher in the alveolar tumors than in the embryonal ones; this relationship is maintained in comparisons between the RD (embryonically derived) and Rh30 (alveolarly derived) cell lines. It is possible that the Rh30 cell line is “desensitized” due to its relatively low but still substantial level of miR-1 and miR-133a expression. Second, oncogenic PAX3-FOXO1 expression in the Rh30 cell line (35) may override the growth inhibitory effects of miR-1 and miR-133a. Similarly, the ability of miR-1 to promote induction of myogenic genes may also be abrogated by PAX3-FOXO1, as we do not see a robust induction of marker myogenic genes in miR-1-transfected Rh30 cells. Alternatively, miR-1 may promote differentiation only to a certain point and that point may already be the basal state of the more differentiated cell line Rh30. Further experiments will be necessary to clarify the role of PAX3-FOXO1 in miR-1-mediated differentiation and miR-1/133a-mediated cytostasis.

Our array data expand on the marker gene assays to prove that miR-1 and miR-133a-mediated regulation is extensive. Clearly, miR-1 promotes the expression of a number of genes that are normally expressed at high levels in mature skeletal muscle. The underlying basis for such widespread changes could be the increase in myogenin and MEF2C expression and/or function. These two transcription factors collectively mediate the activation of a number of myogenic genes (36). The up-regulation of myogenin, in turn, can be linked to a decrease in the miR-1 target HDAC4, as this histone deacetylase gene represses myogenin (28) and MEF2 gene function (27). Moreover, myogenin and MEF2c synergistically transactivate and up-regulate each other’s expression (37, 38). Given its central role in myogenin and MEF2C regulation, we hypothesize that the miR-1 target HDAC4 (6) could lie at the core of a miR-1-regulated promyogenic network and play a significant role in the ability of miR-1 to promote differentiation of RMS cells. In addition, HDAC4 repression has already been implicated in explaining the promyogenic effects of miR-1 in C2C12 cells (6). We tested this hypothesis by expressing in RD cells modest levels of HDAC4 lacking its 3′ untranslated region and thus resistant to miRNA down-regulation. Although the basal levels of marker myogenic genes were depressed in these cells, HDAC4 expression did not inhibit the ability of miR-1 to enhance the expression of marker myogenic genes (data not shown). Hence, we conclude that HDAC4 repression cannot singularly account for miR-1’s ability to promote differentiation and implicate additional regulators of myogenesis in explaining the effects of miR-1. This is consistent with the notion that in most cases, miRNAs function by modulating the expression of not one, but a number of target genes.

Despite the lack of a complete understanding of its biological function, the ability of miR-1/miR-133a to promote different aspects of differentiation suggests their use as a prodifferentiation therapeutic in ERMS. The idea of using prodifferentiation therapy is not new and has been clinically established in another tumor type (39). Conceptually, prodifferentiation therapy is applicable to tumors that can be promoted to differentiate. Our study suggests that this is a feasible strategy to pursue using miRNAs as a prodifferentiation agent. While this article was being prepared, an independent report (17) demonstrated the feasibility of this idea using miR-206 in RMS. In addition, and as is the case with differentiation-sensitive acute promyelocytic leukemia, in which a prodifferentiation agent therapy (all trans retinoic acid and arsenic trioxide) is coupled with chemotherapy, miRNAs could also be considered as a supplement to the chemotherapeutic regimens in use today, and this may serve to reduce the toxicity associated with chemotherapy.