Soft tissue sarcoma subtypes exhibit distinct patterns of acquired uniparental disomy

Background Soft tissue sarcomas (STS) are heterogeneous mesenchymal tumors with diverse subtypes. STS can be classified into two main categories according to the type of genomic alteration: recurrent translocation driven STS, and non-recurrent translocations. However, little has known about acquired uniparental disomy in STS. Methods In this study, we analyzed SNP microarray data to determine the frequency and distribution patterns of acquired uniparental disomy (aUPD) in major soft tissue sarcoma (STS) subtypes using CNAG and R softwares. Results We identified recurrent aUPD regions specific to alveolar rhabdomyosarcoma with the most frequent at 11p15.4, gastrointestinal stromal tumor at 1p36.11-p35.3, leiomyosarcoma at 17p13.3-p13.1, myxofibrosarcoma at 1p35.1-p34.2 and 16q23.3-q24.1, and pleomorphic liposarcoma at 13q13.2-q13.3 and 13q14.11-q14.2. In contrast, specific recurrent aUPD regions were not identified in dedifferentiated liposarcoma, Ewing sarcoma, myxoid/round cell liposarcoma, and synovial sarcoma. Strikingly total, centromeric and segmental aUPD regions are more frequent in STS that do not exhibit recurrent translocation events. Conclusions Our study yields a detailed map of aUPD across 9 diverse STS subtypes and suggests the potential location of several novel tumor suppressor genes and oncogenes.

Single nucleotide polymorphism (SNP) microarrays allow the detection of copy number alterations and acquired uniparental disomy (aUPD also known as copy number neutral loss of heterozygosity), which occurs when both copies of a chromosome originate from the same parent, in the most cases without a change in copy number. There are two major mechanisms leading to aUPD: mitotic recombination of sister chromatids, which results to segmental aUPD, or the loss of a complete chromosome followed by duplication resulting in whole chromosome aUPD.
aUPD regions may cause pre-existing abnormalities (mutation, deletion, amplification, methylation, histone modification, and/or imprinting) to become homozygous, which may lead to clonal selection and growth advantage in the cells. To date, aUPD has been described mostly in hematologic malignancies, [12][13][14][15] breast cancer [16][17][18] and colon cancer [19]. Barretina and colleagues have recently reported aUPD in a limited number of STS samples [11]. The purpose of this study was to determine the frequency, distribution of aUPD in 9 subtypes of STS and identify recurrent aUPD regions specific for each subtype in a large sample set of STS.
After quality control of the retrieved raw data (CEL files), we processed the CEL (intensity) files to generate CHP files by using GeneChip Genotyping Analysis (GTYPE, version 4.1) and Genotyping Console (GTC, version 3.0) software (Affymetrix, Santa Clara, CA). QC metrics was calculated as default in GTC. Then, microarray data were analyzed for determination of allele-specific copy numbers using CNAG (Copy number analysis for GeneChips) (version 3.4) software (http://genome.umin.jp) by using a Hidden Markov Model to predict the presence of aUPD regions as previously described [24]. Data from each of the array platforms were independently analyzed by using non self-controls with automatically selected sex-matched reference samples from HapMap data and from previously published, publicly available datasets; GSE14860 [25], GSE10922 [26], GSE11417 [27], GSE10092 [28], and GSE15097 [29]. Only GSE21124 data set was analyzed by matching normal samples. In the aUPD analyses both the genotype information and the intensity were used. Then all the data from each array were used to generate aUPD profiles for each tumor. The total aUPD was calculated by counting the all aUPD regions. The segmental aUPD was calculated by counting the aUPD at telomeric and centromeric regions, and whole chromosomal aUPD was considered if aUPD occurs in entire chromosome. If aUPD occurs with one mitotic recombination defined as telomeric, and if aUPD occurs via two or more mitotic recombination defined as centromeric. The NCBI Build 36/hg18 (http://genome.ucsc.edu) was used for identifying gene localization and function. Previously, aUPD was detected in limited number of STS by using GISTIC analysis, which is designed to identify copy number alterations, but not aUPD [11].

Statistical analyses
We performed non-parametric Kruskal-Wallis test to identify difference of frequency of aUPD regions between translocation and non-translocation groups of STS and aUPD regions between segmental and whole chromosome, telomeric and centromeric. Frequency of aUPD describes the number of aUPD per sample. Percentage of aUPD in each of groups or subtypes was calculated by the  tumors that had at least one aUPD region. A two-sided p value < 0.05 was considered to be statistically significant. Statistical analysis was performed using R software version 2.15.0 (http://www.r-project.org/).

Distribution of aUPD patterns in STS
We integrated genomic data from 5 different studies to allow us to interrogate a large number of samples encompassing different types of STS. As indicated in Additional file 2: Figure S1, aUPD is found across all chromosomes in STS. We identified aUPD in 47.9% (151/315) of tumor samples with a range between 0 and 37 regions (mean 2.3, median 0) with a total of 724 aUPD regions. Segmental aUPD (630/724; mean 2, median 0) was more frequent (P < 2.3E-16) than whole-chromosome aUPD (94/724; mean 0.3, median 0) (Figure 1), suggesting that mitotic recombination is a more common mechanism of aUPD generation in STS than is the loss of one chromosome and duplication of the remaining chromosome. In addition, we found that centromeric aUPD (441/724; mean 1.4, median 0) was significantly more common (P < 0.0002) than telomeric aUPD (189/724; mean 0.6, median 0) (Figure 1). This requires complex chromosomal rearrangements indicating that multiple mitotic recombination events occur frequently in soft tissue sarcoma tumorigenesis.
Strikingly, the patterns of aUPD varied markedly across STS subtypes (Figure 2A and 2B). Moreover, the proportion of patients with aUPD were found to vary in each subtype; 73% (27/37) Figure S2B). The frequencies of aUPD of these three subgroups are significantly different (P<9.43E-06, Figure 3B). Total, centromeric and segmental aUPD were significantly more frequent in non-recurrent translocation STS than recurrent translocation driven STS (P < 3.71E-04, P < 6.64E-06, P < 1.57E-04, respectively) ( Figure 4). We also identified statistically significant differences in total, centromeric and segmental aUPD between subtypes of (See figure on previous page.) Figure 2 Distribution of aUPD in (A) non-translocation and (B) translocation driven soft tissue sarcomas. (A) aUPD regions in nontranslocation driven soft tissue sarcomas; GIST, leiomyosarcoma, myxofibrosarcoma, pleomorphic liposarcoma, and dedifferentiated liposarcoma. (B) aUPD regions in translocation driven soft tissue sarcomas; myxoid/round cell liposarcoma, synovial sarcoma, Ewing sarcoma, and alveolar rhabdomyosarcoma. Each red line represents region of aUPD for each soft tissue sarcoma sample. Gene name in red represents most mutated genes and green represents imprinted genes that previously reported, which are mapped in the aUPD regions. liposarcoma when comparing pleomorphic liposarcoma to MRC (P < 9.26E-05, P < 6.23E-04, P < 2.12E-04, respectively), and also comparing both pleomorphic liposarcoma and dedifferentiated liposarcoma to MRC (P < 0.02, P < 0.03, P < 0.03). However, no statistically significant difference was observed between dedifferentiated liposarcoma and MRC (P < 0.28, P < 0.26, P < 0.35). The frequencies of total, telomeric, centromeric, segmental and whole-chromosome aUPD for each subtype of STS are summarized in Additional file 4: Figure S3 A-E.
In alveolar RMS, the most frequent aUPD region was at chromosome 11p (29.8%), with the minimal recurrent region at 11p15.4. Several potential cancer genes map to this region: TAF10, ILK, and EIF3F ( Figure 2B, Additional file 5: Table S2). aRMS is characterized by loss of imprinting in IGF2 and H19 [34][35][36]. Interestingly IGF2 is expressed from the paternal allele, which may lead to increased expression of IGF2 while H19 is maternal expressed, and may lead to suppressed expression of H19 [37]. Thus aUPD in these regions could result in decreased or increased expression of candidate cancer genes depending on which parental allele is duplicated. Figure 4 The comparison of frequency of total, telomeric, centromic, segmental and whole chromosome aUPD in nontranslocation and translocation driven soft tissue sarcomas.

Homozygous deletion and focal amplification at aUPD regions
Next, we identified aUPD regions with homozygous deletion or focal amplification. aUPD regions with homozygous deletion that could potentially harbor tumor suppressor genes are summarized in Additional file 6: Table S3. We found a focal amplification region at chromosome 11p15.3-p15.2 (RASSF10, RRAS2, and COPB1) in one tumor sample of aRMS, where the other aRMS samples harbor aUPD in the same region (Additional file 6: Table S3). The amplification at 11p15.3-p15.2 may increase the level of a gain-offunction allele in this region.

Conclusion
In conclusion, to our knowledge, our study encompasses the largest sample set available for the analysis of aUPD in soft tissue sarcoma subtypes. Our results yield a detailed map of aUPD across 9 diverse sarcoma subtypes. The frequency and distribution of aUPD is significantly higher in fusion-negative STS than translocation driven STS suggesting an alternative mechanism underlying tumor development. This study provides evidence for a basis for mutation screening with next-generation sequencing to identify potential mechanistic mediators and therapeutic targets for each subtype of STS and particularly for recurrent regions specifically associated with translocation negative STS.