Background
Rhabdomyosarcoma (RMS) represents the most frequent soft tissue sarcoma in pediatric patients. The two main histological subtypes of RMS tumors, alveolar RMS (ARMS) and embryonal RMS (ERMS), have distinct molecular and clinical profiles. The former, in fact, is characterized by more aggressive behavior and a higher tendency to present with signs of metastatic disease at diagnosis and to relapse after treatment [
1]. Approximately 80 % of ARMS harbor the reciprocal chromosomal translocation t(2;13) (q35;q14) or the less common variant translocation t(1;13)(p36;q14) in which
PAX3 and
FOXO1, or
PAX7 and
FOXO1 genes, respectively, are juxtaposed [
2]. The latter subtype, instead, is not characterized by specific genetic aberrations except for a loss of heterozygosity at 11p15, which could mean that this region contains tumor suppressor genes.
Over the past decade many genome-wide studies have demonstrated that fusion-positive and negative RMS present different gene expression signatures [
3,
4]. Despite the low rate of gene mutations shown by RMS, recent genomic studies have revealed that recurrent mutations in several key genes characterize different RMS subtypes. In particular, mutations in receptor tyrosine kinase/
RAS/PIK3CA and
FGFR signaling predominately affect fusion negative tumors [
5]. The presence of metastasis at diagnosis represents the strongest predictor of poor outcome, and the 5-year survival rate for patients with metastatic disease is approximately 30 % [
6].
The characterization of specific de-regulated genes in metastatic samples may help to define the tumor’s metastatic potential at a molecular level and to monitor disease progression as well as its response to therapy. Growing evidence indicates that normal DNA methylation patterns are altered in cancer cells as there is an overall decrease in the genomic content in 5-methylcytosine and frequent hypermethylation and inactivation of tumor suppressor genes [
7]. Aberrant DNA methylation in candidate genes such as
FGFR1 [
8],
JUP [
9],
MYOD1 [
10],
PAX3 [
11]
, RASSF1 [
12]
, BMP2 [
13] and
CAV1 [
14] has also been described in RMS.
Microarray and novel sequencing techniques have facilitated the comprehensive analysis of the genome and have paved the way for genome-wide scanning of DNA methylation states [
15]. Epigenetic information such as DNA methylation profiling could, in fact, help to identify tumor subtypes and lead to more accurate diagnoses [
16‐
18]. Several genome-wide studies, which have demonstrated that distinct methylation patterns are found in ARMS vs ERMS and fusion-positive vs fusion-negative tumors [
19‐
21], have shown that
PTEN and
EMILIN1 are differentially expressed genes that may be regulated by DNA methylation.
The current study aimed to examine methylation patterns in alveolar and embryonal samples and to explore epigenetic changes in different RMS subtypes at various clinical stages. We delineated, for the first time, the association between metastatic phenotype and DNA methylation pattern. Study results also uncovered a novel gene whose expression is lowered by DNA methylation, suggesting that epigenetic therapy could be utilized to improve current treatment protocols of rhabdomyosarcoma.
Methods
Cell culture
Human ARMS (RH4 and RH30) and human ERMS cells (RD and RH36) were maintained in Dulbecco’s modified Eagle’s medium containing 10 % fetal calf serum, penicillin (100 U/mL), and streptomycin (100 ug/mL) (Life Technologies, Carlsbad, CA) at 37 °C in 5 % CO
2 in a humidified incubator. RH30 and RD cells were obtained from American Type Culture Collection (Manassas, VA); RH4 were gift from Prof. Pier Luigi Lollini (Dept. Medicina Specialistica, Diagnostica e Sperimentale, University of Bologna, Italy) [
22]. RH36 were obtained from Dr. Maria Tsokos (National Cancer Institute, Bethesda, MD) [
23]. A summary of RMS cell line features is available in Additional file
1.
Tumor samples and ethics approval
Specimens were obtained from the Italian Association of Pediatric Hematology and Oncology Soft Tissue Sarcoma Bank at the Department of Women’s and Children’s Health, University of Padova (Padova, Italy). The study, part of a clinical trial carried out in association with the Association Italiana Ematologia Pediatrica AIEOP (Italian Association of Pediatric Hematology and Oncology), was approved by the local ethics committee. Selected clinical parameters of RMS patients used in the analysis are available in the Additional file
2.
Total RNA and DNA isolation
Genomic DNA was isolated from RMS cell lines and from RMS tumor biopsies using Trizol® Reagent (Life Technologies) after RNA extraction following the manufacturer’s instructions. The commercially available Qiamp DNA mini Kit (Qiagen) was used to purify the DNA. Total DNA was quantified using the ND-1000 spectrophotometer (Nanodrop, Wilmington, DE).
Genome-wide DNA methylation profiles
Four μg of genomic DNA was fragmented by sonication and purified using Mini-Elute columns (Qiagen Co., Hilden Germany), and the amount of double-stranded DNA (dsDNA) was measured using the Qubit instrument (Invitrogen, Life Technologies Co., Carlsbad, CA, USA). The success of fragmentation was evaluated using the Agilent Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA). The MethylMiner Methylated DNA enrichment kit (Invitrogen, Life Technologies Co., Carlsbad, CA, USA) was used to enrich the fraction of methylated dsDNA, starting from 2 μg of fragmented whole genomic DNA. Ten ng of methylated dsDNA for each sample was amplified using Whole Genome Amplification (WGA, Sigma-Aldrich Co., St. Louis, MO, USA). Genomic DNA was used as the control for each sample. DNA methylation profiling was carried out in RMS tumor samples using the Human DNA Methylation Microarray (Agilent Technologies, Santa Clara, CA, USA) consisting of about 244,000 (60-mer) probes designed to interrogate about 27,000 known CpG islands. The control genomic DNA and methylated dsDNA were labeled with Cy3 and Cy5 dye respectively using Agilent Genomic DNA labeling kit PLUS (Agilent Technologies, Santa Clara, CA, USA) and competitively hybridized to Human DNA Methylation microarrays platforms (GEO ID: GPL10878). The hybridization was carried out at 67 °C for 40 h in a hybridization oven rotator (Agilent Technologies, Santa Clara, CA, USA). The arrays were washed with Agilent ChiP-on-chip wash buffers as suggested by the supplier. Slides were scanned on an Agilent microarray scanner (model G2565CA), and Agilent Feature Extraction software version 10.7.3.1 was used for image analysis.
Availability of data and materials
Raw data are available on the GEO website using accession number GSE67201, and processed data are presented as Additional files
1,
2 and
3.
Statistical analysis of DNA methylation data
Intra-array normalization of methylation levels was performed with linear and lowess normalization. Inter-array normalization was performed with quantile normalization [
24] in order to correct experimental distortions. The normalization function was applied to the methylation data of all the experiments. Feature Extraction Software (Agilent Technologies, Santa Clara, CA, USA) provided spot quality measures with regard to methylation expression data in order to evaluate the quality and the liability of the hybridization data. In particular, flag “glsFound” and “rlsFound” (set to 1 if the spot had an intensity value that was significantly different from the local background or to 0 in any other cases) were used to filter out unreliable probes: flag equal to 0 was to be noted as “not available (NA)”. Probes with a high proportion of NA values (more than 25 %) were removed from the dataset to ensure more robust, unbiased statistical analyses. When twenty-five percent of NA was used as the threshold in the filtering process, a total of 90.591 probes were obtained. The microarray data were analyzed using the iChip R bioconductor Package. The microarray data were processed in accordance with the instructions contained in the package vignette (
www.bioconductor.org/packages/release/bioc/vignettes/iChip/inst/doc/iChip.pdf). Briefly, after normalization we computed the enrichment measure using the lmtstat function (a wrapper function of the empirical Bayes t-statistic from limma package) provided by iChip package. Specifically, we used the iChip2 function that implements the high order hidden Ising model described in [
25]. The iChip2 function was called with
b = 1 following the specifications for low resolution arrays, while the other parameters were left at the default value. iChip2 function Enriched regions were called using an FRD cutoff of 0.2 and maxGap = 500 bp.
The genes associated to DMRs identified using iChip algorithm were functionally analyzed using Gene Ontology (GO) implemented by the Database for Annotation, Visualization and Integrated Discovery (DAVID) tool [
26]. The significantly enriched biological categories were identified using a Modified Fisher Exact
p-value < 0.05.
Trichostatin A and 5-aza-2′-deoxycytidine treatments
RMS cells (0.25x106 cells/mL) grown in 100 mm dishes were treated with demethylating agent 5-aza-2′- deoxycytidine (5-Aza-dC) (Selleck Chemicals, Houston; TX, USA), with TSA (Selleck Chemicals, Houston; TX, USA), or with a combinatorial treatment using both agents. Concentrations varying from 100nM to 2 μM of 5-Aza-dC for 72 h and 200 ng/ml of TSA for 16 h were used. Cells were harvested and processed for RNA or DNA extraction.
qRT-PCR for mRNA detection
For mRNA detection, 1 μg of total RNA was retrotranscribed with Superscript II (Life Technologies), and qRT-PCRs were carried out with gene-specific primers and the SYBR PCR Master Mix (Applied Biosystem, Life Technologies) using a ViiA 7 Real-Time PCR System.
GADPH was selected for the endogenous normalization of the gene expression analysis. The relative expression levels between samples were calculated using the comparative delta Ct (threshold cycle number) method (2
-ΔΔCt) [
27] implemented in the ViiA 7 Real-Time PCR System software. A 95 % confidence interval (IC) was calculated.
The relative expressions of non-clustered protocadherins (PCDHs) were simultaneously analyzed using the relative expression software tool (REST) which is able to identify significance differences between two groups of samples using a randomization test [
28]. Permutation or randomisation tests are useful alternatives to more standard parametric tests because despite the fact that they remain as powerful as more standard tests, they make no distributional assumptions about the data. The randomisation test repeatedly and randomly reallocates the observed values to the two groups and notes the apparent effect (expression ratio in our case) each time. A proportion of these effects, which are as great as those actually observed in the experiment, gives the
P-value of the test.
The statistical analysis of PCDHA4 expression levels, evaluated in an expanded cohort of samples, was performed using Prism6 software, and the Mann–Whitney U-test was used.
Sodium bisulfite treatment of DNA and bisulfite sequencing
One μg of genomic DNA was subjected to conversion with sodium bisulfite using EZ DNA Methylation-Gold ™ kit (Zymo Research, Orange, CA, USA), following the manufacturer’s instructions. One hundred ng of bisulfite-converted DNA was used as template for the amplification of candidate regions. Polymerase chain reaction (PCR) was performed using methylation-independent primers designed with the free online tool MethPrimer (http:/itasa.ucsf.edu/~urolab/methprimer;). The PCR products were purified using the QIAquick PCR purification kit (Qiagen Co., Hilden Germany) and subcloned into pSC-A-amp/kan vector using the StrataClone PCR Cloning Kit (Agilent Technologies, Santa Clara, CA, USA). Competent cells were transformed with ligation reaction product and grown in Luria Bertani (LB) agar plates supplemented with 40 μg/ml of X-Gal (Promega Co., Madison, WI, USA) and 50 μg/ml of ampicillin for 16 h at 37 °C. Blue-white screening permitted identification of recombinant bacteria. Selected clones were evaluated by colony PCR performed using M13R and T7 universal primers (Invitrogen, Life Technologies Co., Carlsbad, CA, USA). The PCR products were checked for the presence of inserts using agarose electrophoresis, and those corresponding to positive clones were purified using a QIAquick PCR purification kit (Qiagen Co., Hilden Germany) and then sequenced by 3500 Dx Genetic Analyzer sequencer (Applied Biosystems, Life Technologies Co., Carlsbad, CA, USA) using BigDye® Terminator v3.1 CycleSequencing Kit (Applied Biosystems, Life Technologies Co., Carlsbad, CA, USA) following the manufacturer’s instructions.
RNA interference
RH36 cells at 50 % to 70 % confluence were transfected with small-interfering RNA (siRNA) for target gene PCDHA4 (siPCDHA4) or with non-targeting siRNA (siCONTROL) using Lipofectamine2000 transfection reagent (Thermofisher Scientific). We performed preliminary experiments in the attempt to achieve the highest efficiency and reproducibility. The efficacy of gene knockdown was evaluated at the mRNA level using qRT-PCR analysis after 48 h of transfection.
Flow cytometric analysis of the cell cycle
After transfection, PCDHA4 silenced cells (siPCDHA4) and control cells (siCONTROL) were harvested. For each sample, 1x106 cells were fixed with 70 % cold ethanol, washed in PBS, and incubated with propidium iodide (50 μg/mL) and RNase (100 μg/mL) for 60 min at 37 °C. Samples were run in a BD FACScan (Becton Dickinson, Labware, Bedford, MA); the data were analyzed with ModFitLT V3.0 software (Verity Software House, Topsham, ME). Two independent experiments were performed with three replicates for each. A 95 % Confidence interval (CI) was calculated.
Invasion Transwell Assay
Chemoinvasion was measured using 24- well BioCoat Matrigel invasion chambers (Becton Dickinson) with an 8-μm pore polycarbonate filter coated with Matrigel. The lower compartment contained 0.5 mL of 1 % serum medium conditioned by the NIH3T3 cell line as a chemoattractant or serum-free Dulbecco’s modified Eagle’s medium as a control. In the upper compartment, 1x10
4 RH36 cells per well were placed in triplicate wells and incubated for 18 h at 37 °C in a humidified incubator with a 5%CO2 atmosphere. After incubation, the cells on the filter’s upper surface were wiped off with a cotton swab; the cells on the lower surface were, instead, fixed in 2.5 % glutaraldehyde, stained with 0.2 % crystal violet in 20 % methanol, and then photographed using a stereomicroscope (model MZ16; Leica Microsystems) equipped with a charge-coupled device (CCD) camera. Images were processed using Corel-Draw software (Corel, Ottawa, Canada), and the area occupied by the migrated cells was measured using ImageJ software (
http://rsbweb.nih.gov/ij, last accessed September 4, 2009). A 95 % Confidence interval (CI) was calculated.
Discussion
During the current study, methylation profiling of RMS samples using a genome-wide approach uncovered differences in DNA methylation signatures of metastatic and localized RMS and highlighted that epigenetic alterations are peculiar to disseminated RMS. The findings demonstrated that DNA methylation can contribute to defining the molecular features of RMS subgroups and can thus increase the accuracy of RMS subtype classification. PCDHA4 was identified as a gene whose expression was decreased in metastatic with respect to non-metastatic RMS. Preliminary data also suggested that PCDHA4 may act as a tumor suppressor in RMS cells and that it may be partially inactivated by DNA methylation.
Several investigators including ourselves have examined the different behaviors and molecular features of alveolar and embryonal RMS, and over the past decade many studies have confirmed that gene expression profiles distinguish between alveolar
PAX3/FOXO1 positive ARMS and
PAX3/FOXO1 negative ARMS and ERMS [
3,
4]. Although the mechanisms underlying different gene regulation patterns are still under investigation, it is known that epigenetic modifications, and in particular DNA methylation, can modulate gene expression and represents a challenging area of cancer research. Recent studies have demonstrated that DNA methylation profiles distinguish between RMS fusion-positive and negative samples and could be used to improve the molecular classification of rhabdomyosarcoma [
19,
21].
Analysis of our microarray data confirm that alveolar
PAX3/FOXO1 positive samples have a different methylation signature with respect to RMS fusion-negative ones. It is important to remember that our analysis of DNA methylation differences in metastatic and non-metastatic tumors is based on the Intergroup Rhabdomyosarcoma Study (IRS) grouping which is highly predictive of tumor outcome. In particular, patients classified as IRS IV group, which is characterized by metastatic disease, have long-term failure-free survival (FFS) rates of <30 % [
6,
30]. When we compared non-metastatic (IRS I-II-II) with metastatic (IRS IV) RMS patients, we found quite distinct methylation signatures. This data, together with those produced by other genome-wide studies highlight methylation pattern changes in different RMS subtypes and seem to suggest that an epigenetic therapy could be appropriate to treat rhabdomyosarcoma. Indeed, DNA methylation is an excellent target for anti-cancer therapy as it involves a reversible process that does not affect the DNA sequence.
No epigenetic drugs are currently included in RMS clinical protocols, but over the past few years, several new anti-cancer drugs with epigenetic activities have received approval for clinical trials on other solid cancers leading to a detailed characterization of their mechanisms of action. Some FDA-approved drugs for the treatment of solid cancers are 5-Aza that target histone deacetylase (HDAC) in metastatic non-small cell lung cancer [
31] and inhibitors of EZH2, a subunit of Polycomb repressive complex in diffuse large B cell Lymphoma (DLBCL) [
32].
When we analyzed the distribution in the genome of the differentially methylated regions (DMRs) identified by microarray analysis, we found that only 25 % of them overlap with promoter regions (regions defined as 2 kb upstream and 1 kb downstream of RefSeq transcription star site) while the others map inside gene bodies or are localized in intergenic regions distal to known genes. Advances in genome-wide approaches have demonstrated that methylation varies depending upon the specific genomic context. Although the majority of studies have focused on methylation in the promoter region adjacent to the transcription start site (TSS), methylation of the gene body or intergenic regions seems to have a functional role and contributes to defining the whole picture of the methylation status [
33].
Our microarray analysis revealed an abnormal methylation pattern in promoters of protocadherins (PCDHs). PCDHs, which are a group of transmembrane proteins, constitute the largest subfamily of the cadherin cell-adhesion molecules. In mammals, PCDHs are organized in clusters (α,β,γ) or are scattered throughout the genome [
29,
34]. The methylation value [the enrichment value expressed as moderated t-statistics computed using the eBayes function (limma-t)] of DMRs associated to protocadherins was that of a common hypermethylation level of promoter regions in metastatic compared to non-metastatic samples. Using qRT-PCR assays, we analyzed the expression levels of some PCDHs and observed that the correlation between promoter hypermethylation and downregulation of genes is very low, indicating that epigenetic alterations may have an alternative effect with respect to typical gene regulation. Other genome wide studies have reported the same result suggesting that the correlation between DNA methylation and mRNA expression does not always conform to the paradigmatic inverse correlation between the two processes [
21].
Our findings highlighted
PCDHA4 as an example of a gene in RMS whose expression differs in metastatic and non-metastatic RMS samples. We hypothesize that the decrease in
PCDHA4 expression depended on DNA methylation. Interestingly, a recent methylation profiling study of cervical cancer samples revealed a methylation silencing of many clustered protocadherins in the cancer with respect to the control cells. The study also reported that there was a positive correlation between methylation frequency of
PCDHA4 and
PCDHA13 and tumor severity, highlighting the role of
PCDHA4 silencing in cancer progression [
35]. Other studies have demonstrated the involvement of several protocadherins in tumor processes. It has been shown that protocadherins behave as tumor suppressor genes in many solid cancers such as non-small-cell lung cancer, gastric and prostate cancer. It has also been demonstrated that their involvement is due to aberrant DNA methylation that determines an altered expression pattern [
36‐
38].
In the light of these findings, we performed some in vitro functional studies on RMS cells. These preliminary data uncovered the potential role of PCDHA4 as a tumor suppressor in RMS. In fact, we observed that PCDHA4-silenced cells acquire a more aggressive phenotype, as demonstrated by the increase in proliferation rate and invasiveness that was found. Further functional studies are warranted to clarify the involvement of PCDHA4 in rhabdomyosarcoma.
Treatment with 5-Aza-dC and/or TSA of RMS cell lines suggested that there is epigenetic control of
PCDHA4 expression. Although the restoration of
PCDHA4 expression was noted only in one alveolar cell line (RH30), no changes in
PCDHA4 expression levels were observed in embryonal cell lines (RD, RH36). Bisulfite sequencing thus confirmed the different methylation status of the
PCDHA4 promoter region in alveolar and embryonal cell lines. Taken together, these data suggest that DNA methylation probably decreases the transcription of
PCDHA4 selectively in a RMS subgroup. Our experiments also demonstrated that a combination of 5-aza-2′-deoxycytine and trichostatin A drugs in RMS cells act synergistically to restore
PCDHA4 expression confirming the potential utility of a combination therapy. The combination of demethylating molecules with drugs that target histone modifications may enhance the efficacy of treatment as already demonstrated by other studies [
39,
40].
Acknowledgements
The authors wish to thank MicroCribi Microarray Service, C.R.I.B.I., University of Padova, Italy for assistance with the microarray experiments. Appreciation is expressed to Matteo Zampini (Department of Biology, University of Padova, Italy) for assistance in revising the manuscript.