Identification of target genes of PAX3-FOXO1 in alveolar rhabdomyosarcoma

Rhabdomyosarcoma (RMS) is a soft tissue sarcoma categorized into two major subtypes: alveolar RMS (ARMS) and embryonal RMS (ERMS). Most ARMS express the PAX3-FOXO1 (P3F) fusion oncoprotein generated by the 2;13 chromosomal translocation. In the present study, the downstream target genes of P3F were identified by analyzing two independent sets of gene expression profiles: primary RMS tumors and RD ERMS cells transduced with inducible P3F constructs. We found 34 potential target genes (27 upregulated and 7 downregulated) that were significantly and differentially expressed between P3F-positive and P3F-negative categories, both in primary RMS tumors and in the inducible P3F cell culture system. Gene ontology analysis of microarray data of the inducible P3F cell culture system employed indicated apoptosis, cell death, development, and signal transduction as overrepresented significant functional categories found in both upregulated and downregulated genes. Therefore, among the 34 potential target genes, the expression of cell death-related [Gremlin1, cysteine knot superfamily 1, BMP antagonist 1 (GREM1) and death-associated protein kinase 1 (DAPK1)] and development-related [myogenic differentiation 1 (MYOD1) and hairy/enhancer-of-split related with YRPW motif 1 (HEY1)] genes were further investigated. The differential expression of GREM1, DAPK1, MYOD1 and HEY1 was confirmed in independent tumors and inducible cell culture systems. The expression of GREM1, DAPK1 and MYOD1 were significantly upregulated; HEY1 was significantly downregulated in independent P3F-positive ARMS tumors and transcriptionally active P3F cells, compared to those in ERMS tumors and transcriptionally inactive P3F cells. This study identified target genes of P3F and suggested that four downstream targets (GREM1, DAPK1, MYOD1 and HEY1) can contribute to the biological activities of P3F involved in growth suppression or cell death and myogenic differentiation.


Introduction
Rhabdomyosarcoma (RMS), the most common type of pediatric soft tissue sarcoma, is associated with the skeletal muscle lineage. Pediatric RMS is categorized into two major subtypes: embryonal RMS (ERMS) and alveolar RMS (ARMS) based on their histologic appearance (1). ERMS shows a more favorable prognosis, while ARMS is more aggressive with a high frequency of metastases at initial diagnosis and a worse prognosis (2,3).
ERMS has not been found to be associated with any diagnostic genetic alterations, but loss of heterozygosity at 11p15 is a common finding (4,5). In contrast, ARMS is associated with recurrent chromosomal translocations. The translocation t(2;13)(q35;q14) leading to the PAX3-FOXO1 (P3F) gene fusion was found to be present in 55% of ARMS cases, while the translocation t(1;13)(q36;q14) leading to the PAX7-FOXO1 gene fusion was present in 22% of cases, and 23% of ARMS were fusion-negative (2). ARMS with the 2;13 translocation is characterized as having overexpression of P3F relative to wild-type PAX3, at both the RNA and protein levels (6). In both PAX3 and P3F, an alternative splice occurs at the intron 2/exon 3 junction resulting in inclusion or exclusion of a glutamine (Q) residue [PAX3(+Q) or (-Q) forms, respectively] (7). As a result, PAX3(+Q)-FOXO1 and PAX3(-Q)-FOXO1 are coexpressed in P3F-positive ARMS tumors.
Among RMS tumors, PAX3(+/-Q)-FOXO1-positive ARMS is the most clinically intractable fusion subtype of pediatric RMS (1,2,8,9). However, the histologic classification of RMS into ARMS and ERMS may be difficult in some cases (9) and there are no specific drugs for treating specific histologic or fusion subtypes. Therefore, there is substantial impetus to elucidate target genes of P3F, which can be used to identify therapeutic targets and markers for use in RMS diagnosis and management.
Although several recent studies utilized gene expression profiles to classify RMS and/or identify target genes of P3F and PAX7-FOXO1 or P3F only in ARMS (10)(11)(12)(13)(14)(15)(16), more research is needed to validate the function of genes that are biologically relevant in ARMS development.
In the present study, we identified potential target genes of P3F by analyzing gene expression profiles from two independent systems: primary tumors (P3F-positive ARMS and fusion-negative ERMS) and a cell culture system expressing the inducible PAX3(+/-Q)-FOXO1-estrogen receptor (ER) ligand binding domain construct in the pBabe (pB) retroviral vector. Among the potential target genes of P3F, we focused on cell death or apoptosis-related [Gremlin1, cysteine knot superfamily 1, BMP antagonist 1 (GREM1), death-associated protein kinase 1 (DAPK1)] and development-related [myogenic differentiation 1 (MYOD1) and hairy/enhancer-of-split related with YRPW motif 1 (HEY1)] genes since apoptosis and development were significantly overrepresented functional categories in our gene expression profiles.

Materials and methods
Tumor samples. The tumor specimens (15 ERMS and 16 P3F-positive ARMS) used for microarray were previously described (15). Independent tumor specimens (20 ERMS and 17 P3F-positive ARMS) examined for quantitative reversetranscription PCR (qRT-PCR) in the validation studies were previously described (15). The presence of P3F in ARMS tumor specimens was determined and confirmed by RT-PCR and/or qRT-PCR (17).
RNA extraction and microarray data analysis. For microarray analysis, total RNA was extracted using the RNeasy Mini Kit (Qiagen, Valencia, CA, USA). For other studies including qRT-PCR and RT-PCR, total RNA was extracted using RNA STAT-60 (Tel-Test, Inc., Friendswood, TX, USA). Microarray analysis of RNA isolated from RMS tumors and RD cell culture systems expressing inducible PAX3(+/-Q)-FOXO1-ER in pB was performed on Affymetrix arrays. Microarray analysis of 31 RNA tumor samples (15 ERMS and 16 P3F-positive ARMS) was described in our previous publication (15).
For microarray analysis of inducible cell culture systems, three RD cell populations were independently transduced with each construct [PAX3(+Q)-FOXO1-ER in pB; PAX3(-Q)-FOXO1-ER in pB; pB vector alone] and treated without or with 30 nM Tmf for 24 h to generate a total of 18 samples. RNA isolated from these cells was hybridized onto Affymetrix GeneChip Human Genome U133 Plus v.2.0 (HG U133+ v.2.0). After the expression levels were determined using Affymetrix Microarray Suite 5.0, the expression levels were normalized in GeneSpring (Agilent Technology). The probe-sets, whose expression levels were detected in at least 2 of the 18 samples (in inducible cell culture system) or 1 of the 31 samples (in tumors), were selected for further analysis. To compare the gene expression profiles from tumors with those from the inducible cell culture systems, the probes on the HG U133A array were used for GeneSpring analysis.
The Significance Analysis of Microarrays (SAM) (23) with two-class paired condition was used as the common method to identify differentially expressed genes between 16 P3F positive-ARMS and 15 fusion-negative ERMS tumors as well as between the RD cells expressing PAX3(+/-Q)-FOXO1-ER in pB (treated with Tmf; 6 samples) and the RD cells expressing PAX3(+/-Q)-FOXO1-ER in pB (not treated with Tmf; 6 samples). Data was filtered by a false discovery rate (FDR) at <5% (FDR 4.94% for cells; FDR 4.91% for tumors) and ≥1.5 fold-change. The PAX3(+Q)-FOXO1-ER and PAX3(-Q)-FOXO1-ER gene expression profiles were pooled since SAM did not reveal significant differentially expressed genes between these two groups. The genes that were differentially expressed between transcriptionally active PAX3(+/-Q)-FOXO1-ER (Tmf-treated) and inactive PAX3(+/-Q)-FOXO1-ER (untreated control) were then examined.
Finally, common genes that were present in both tumors and the inducible PAX3(+/-Q)-FOXO1-ER cell culture system were identified by Venn Diagram analysis. After this analysis, the probe-sets that encoded the same gene with similar expression profiles and showed raw expression values at <150 were removed. Classifications and functional annotations of genes were analyzed via web-accessible programs: Expression Analysis Systematic Explorer (EASE) 2010 and Database for Annotation, Visualization and Integrated Discovery (DAVID) 2010 (24).

Extraction of cellular proteins and isolation of secreted proteins from the medium of cultured cells and immunoblot
analysis. The cells were seeded at 10 6 /100-mm dish in phenol red-free DMEM medium containing 10% FBS, 1% P/S and 1% AM for 24 h prior to Tmf treatment. The cells expressing inducible PAX3(+/-Q)-FOXO1-ER were then treated with 30 nM Tmf for various times. Cellular proteins were lysed as described previously (20). To assay the secreted GREM1 protein, the medium containing 10% FBS was replaced with FBS-free DMEM after 24 h of Tmf treatment, and cells were cultured for an additional 24 or 48 h with Tmf treatment. The medium of the cultured cells was collected and concentrated to <1,000 µl using Amicon Ultra-4 with Ultracel-10K (UFC80-1024; Millipore, Billerica, MA, USA). Protein was quantified using Coomassie Plus Protein Assay (Pierce Biotechnology, Inc., Rockford, IL, USA).
Statistical analysis. The qRT-PCR data of tumors (20 fusionnegative ERMS and 17 P3(+Q)F-positive ARMS) were analyzed by the Mann-Whitney Test (Wilcoxon Rank Sum Test) and differences between the two tumor groups were considered significant at a P-value <0.05. Statistical analysis of gene expression profiles are described in 'RNA extraction and microarray data analysis'.

Development of inducible
or pB alone and selected with puromycin for 2 weeks. The transduction of RD cells with these pB-P3(+/-Q)F-ER constructs resulted in the expression of PAX3(+/-Q)-FOXO1-ER at both the mRNA (Fig. 1A) and protein levels (Fig. 1B).
In order to determine and confirm the inducible PAX3(+/-Q)-FOXO1 function, the RD-derived cells transduced with pB-P3(+/-Q)F-ER or pB were transiently transfected with a luciferase gene reporter construct containing PAX3 DNA binding sites and were treated with tamoxifen at various concentrations (0, 1, 3, 10, 30 and 100 nM) for 24 h (Fig. 2). The RD cells transduced with an empty vector pB alone ( Fig. 2A)   cells transduced with an inducible PAX3(+Q)-FOXO1-ER construct in the pK vector that enforces high expression of PAX3(+Q)-FOXO1-ER (16). In contrast, our present expression profiling study used the pB vector to express lower levels of inducible PAX3(+/-Q)-FOXO1-ER in RD cells. Second, a previous expression profiling analysis of the inducible cell system (16) was complicated by the confounding factor of a treatment-day effect. Thus, microarray data from inducible cells with PAX3(+Q)-FOXO1-ER in pK were analyzed by Mixed ANOVA, whereas primary RMS tumors were analyzed by SAM. In comparison, no such confounding factors were present in the present study and thus microarray data from both primary RMS tumors and inducible cells were both analyzed by SAM.
Many of the dif ferentially expressed genes in the PAX3(+/-Q)-FOXO1-ER cell systems are related to apoptosis and development. To investigate biological consequences of the PAX3(+/-Q)-FOXO1 expression signature, the 540 significantly upregulated or downregulated probes identified in RD cells expressing the inducible PAX3(+/-Q)-FOXO1-ER in pB on the HG-U133+ v.2.0 platform were analyzed using EASE and DAVID (24). Notably, 10 out of the 15 most significant functional categories analyzed by EASE for the upregulated genes were apoptosis, cell death, or negative regulation of cell proliferation (Fig. 3).
The most significant 20 overrepresented functional groups in the RD cells carrying inducible PAX3(+/-Q)-FOXO1-ER in pB identified by DAVID are listed in Table II. Among these 20 groups, five groups were commonly found in both upregulated and downregulated genes. Four out of these 5 functional groups were related to development (developmental process, multicellular organismal development, anatomical structure development, system development) and one group to signal transduction. MYOD1 (an upregulated gene) and HEY1 (a downregulated gene) were found in all of the 5 common functional groups.
GREM1, DAPK1, MYOD1 and HEY1 are significantly and differentially expressed in both independent tumors and inducible PAX3-FOXO1-ER cell systems. Since gene ontology analysis EASE or DAVID indicated apoptosis (Fig. 3) and development (Table II) as significant functional categories, we validated the expression of apoptosis-related genes (GREM1, DAPK1) and development-related genes (MYOD1, HEY1) that were identified among the 34 potential target genes (Table I) differentially expressed in both primary tumors and the inducible cell culture system. These genes were analyzed in an independent panel of primary tumors (20 fusion-negative ERMS and 17 PAX3-FOXO1-positive ARMS) ( Fig. 4; Table III) and RD cells carrying inducible PAX3(+/-Q)-FOXO1-ER in pB or pK (Tmf-treated versus untreated control) (Fig. 5; Table III). The validation was carried out at the RNA level using qRT-PCR for the expression of GREM1  (Figs. 4D and 5D). Furthermore, the upregulation of GREM1 and DAPK1 at the protein level was validated using immunoblot analysis (Fig. 5E). These validation studies of GREM1, DAPK1, MYOD1 and HEY1 (Figs. 4 and 5) consistently demonstrated significant differential expression, which was comparable to that noted in the microarray expression signature (Table III).

Discussion
In the present study, we developed the RD ERMS cell culture system expressing the inducible PAX3(+/-Q)-FOXO1-ER    in pB, which is useful for determining the early short-term effects of PAX3(+/-Q)-FOXO1 and for regulating the strength and duration of transcriptional activity of PAX3-FOXO1. We analyzed two independent sets of gene expression profiles: primary RMS tumors and RD ERMS cells transduced with inducible PAX3(+/-Q)-FOXO1 constructs. We found 34 poten- The  (16) constructs. Constitutive PAX3-FOXO1 cell system represents relatively longer (later) effects of PAX3-FOXO1 compared to the inducible PAX3-FOXO1-ER cell systems. These two studies (13,16) reported 81 and 39 target genes, respectively. Compared with these two previous studies, 11 and 12 target genes, respectively, were found in common with the present study (Table IV). Six genes (KCNN3, MCAM,   Table I) from those of the ERMS, respectively. MYOD1, TCF7L2, TM4SF10, ZFP36L2) were identified as targets in all three studies (Table IV) (13,16, present study).
The oncogenic activity of PAX3-FOXO1 is characterized in part by its stimulatory effects on cell proliferation (33)(34)(35) and cell survival/anti-apoptosis (36,37) as well as its inhibitory role on terminal myogenic differentiation (31,32). However, in addition to these effects, PAX3-FOXO1 can also cause growth suppression and cell death in other settings (20,21,38).    in pBabe (pB) or P3(+Q)F-ER in pK1 (pK) (◼) or retroviral vectors (◻) alone (pB or pK), were treated without or with 4-hydroxytamoxifen (Tmf) at 30 nM for 24 h. Total RNA was isolated and the expression of these genes was determined by qRT-PCR. The relative expression levels of the 4 genes were normalized to 18S rRNA (mean ± SD; n=3). (E) Cells were cultured for 48 h with 30 nM Tmf. Total proteins were extracted and protein expression was determined by immunoblot analysis. The protein expression levels of GREM1 and DAPK1 were normalized by β-actin protein expression for densitometry graphs. ND, not determined.
These paradoxical features of PAX3-FOXO1 being both oncogenic and growth-suppressive were demonstrated in previous studies with immortalized murine fibroblasts (20,21) and human myoblasts (38). We identified and validated that tumor-suppressor genes, GREM1 (39) and DAPK1 (40), are upregulated at both RNA and protein levels in PAX3-FOXO1positive ARMS tumors and PAX3-FOXO1-ER inducible cell culture systems. Our study suggests that GREM1 and Table IV. Comparison of the potential target genes of PAX3-FOXO1 in three studies (13,16, present study The potential target genes of P3F were identified by microarray analysis in both primary RMS tumors and cells transduced with constitutive PAX3/PAX7-FOXO1 (13) or inducible P3F-ER (16, present study) constructs. Unless noted as (↓) for downregulated genes, all other genes were upregulated genes. DAPK1 tumor suppressor genes can be potential target genes contributing to this growth-suppressive activity of high PAX3-FOXO1 expression.
In conclusion, we identified 34 potential downstream target genes of PAX3(+/-Q)-FOXO1 by analyzing two independent sets of gene expression profiles: primary RMS tumors and RD ERMS cells transduced with inducible PAX3-FOXO1 constructs. Our study can serve as a basis to propose the 4 genes (GREM1, DAPK1, MYOD1 and HEY1) as targets that function in growth suppression or myogenic differentiation downstream of PAX3-FOXO1 in ARMS (Fig. 6).