Identification of potential mutations and genomic alterations in the epithelial and spindle cell components of biphasic synovial sarcomas using a human exome SNP chip

Abstract

Background

Synovial sarcoma (SS) is one of the most aggressive soft-tissue sarcomas and is noted for late local recurrence and metastasis. It is of uncertain histological origin and exhibits a biphasic histopathological form involving both the mesenchyme and epithelium. Thus, its diagnosis and therapy remain a huge challenge for clinicians and pathologists. This study aimed to determine whether differential morphological-associated genomic changes could aid in ascertaining the histogenesis of SS and to determine whether these sarcomas showed some specific mutated genes between epithelial and spindle cells that would promote tumor invasion and metastasis.

Methods

We conducted a comprehensive genomic analysis of mesenchymal and epithelial components in 12 formalin-fixed paraffin-embedded biphasic SS samples using the Illumina human exon microarray. Exome capture sequencing was performed to validate the single nucleotide polymorphism (SNP)-chip data, and de novo data were generated using a whole-exome chip with the Illumina exon microarray. Fisher’s exact test based on PLINK analysis of the SNP-chip data.

Results

Here, the SNP-chip data showed that 336 SNPs had association P-values of less than 0.05 by chi-square test. We identified 23 significantly mutated genes between epithelial and spindle cell regions of SSs. Fifteen gene mutations were specific for the spindle cell component (65.2 %) and eight for the epithelial cell component (34.8 %). Most of these genes have not been previously reported in SS, and neuroguidin (NGDN), RAS protein activator like 3 (RASAL3), KLHL34 and MUM1L1 have not previously been linked to cancer; only one gene (EP300) has been reported in SS. Genomic analyses suggested that the differential SNPs in genes used for functional enrichment are mainly related to the inflammatory response pathway, adhesion, ECM–receptor interactions, TGF-β signaling, JAK–STAT signaling, phenylalanine metabolism, the intrinsic pathway and formation of fibrin.

Conclusions

This study investigated novel biological markers and tumorigenic pathways that would greatly improve therapeutic strategies for SS. The identified pathways may be closely correlated with the pathogenic mechanisms underlying SS, and SS development is associated with morphological features.

Background

Synovial sarcoma (SS) is a highly aggressive mesenchymal tumor with typical dual epithelial and mesenchymal differentiation [1]. It affects pediatric, adolescent, and adult populations and comprises approximately 10 % of all soft-tissue sarcomas; the age-standardized incidence rate per million individuals ranges from 0.5 to 1.3 [2]. SS has a wide spectrum that includes two main variants: monophasic fibrous SS, the most common variant and biphasic SS (BSS), which displays glandular epithelial differentiation architecture in a background of spindle cells. SS likely originates from undifferentiated mesenchymal tissue with variable epithelial differentiation and highly specific chromosomal translocation in more than 95 % of cases [3]. However, the molecular mechanisms underlying tumorigenesis and dual differentiation have remained elusive.

Much of the available literature on SS comprises case reports. Research on the pathogenesis of SS is mainly focused on the relationship between fusion genes and diagnosis or prognosis. However, the cell-of-origin, dual differentiation of SS and the histological transitional relationships between epithelial and mesenchymal differentiation are unclear.

Recent advances in genomic technologies have offered a great opportunity for identifying the complete biological characteristics of neoplastic tissues, resulting in improved diagnosis, treatment selection, rational classification based on molecular carcinogenesis, and identification of therapeutic targets. Genome-wide single nucleotide polymorphism (SNP) analysis is a powerful approach that allows for the investigation of genomes and transcriptomes using limited sample material, and identification of disease-causing genes; other advantages include its abundance in the human genome and ease of high-throughput typing therapy [4]. Thus, these methods have great potential for guiding therapy.

Because of the relative rarity of SS and the poor prognosis, it is not easy to collect an adequate number of SS samples for a genome-wide association study. However, new cost-effective technologies such as microarray SNP-chip typing and whole exome sequencing have allowed us to identify candidate genes, mutations and potential targets for further study and validation [5]. Here, we integrated the resulting data for 319 target genes with those from existing open-source resources with the following aims: detection of somatic variants between the epithelial and mesenchymal components in SS by genome-wide SNP analysis; linkage of target genes with additional biologically significant pathways and gene sets; and to provide insights into our previous mechanistic studies, thereby providing the basis for further experiments for characterizing the genetic factors that are involved in pathogenesis and that could aid in future targeted therapies for SS.

Methods

Patients and tissue specimens

Twelve formalin-fixed, paraffin-embedded (FFPE) BSS samples that were confirmed to have a fusion gene and obtained from patients treated at the Department of Pathology, First Affiliated Hospital, Shihezi University, School of Medicine, between 1980 and 2011, were included in this series. Clinical and demographic data were obtained from the medical charts. The diagnosis of SS was confirmed by histological and immunohistochemical analyses and the presence of a fusion gene was detected by One Step RT-PCR (QIAGEN, Venlo, The Netherlands). The characteristics of the 12 BSS patients enrolled in this study are shown in Table 1. Clinical staging was performed according to the National Comprehensive Cancer Network 2012 guidelines for soft-tissue tumors. Follow-up surveys were conducted.

Table 1 Clinical Description of BSSs Patients for Genome-Wide SNP

Written informed consent was obtained from all patients regarding the use of the collected samples in research studies. The patient records and information were anonymized and de-identified before analysis. Human subjects in this study provided informed consent for use of their tissues for research purposes following procedures approved by the Clinical Research Ethics board of the First Affiliated Hospital, Shihezi University School of Medicine.

Tissue preparation and microscopic separation of the epithelial and spindle cell components of BSSs

Laser capture microdissection (LCM) and tissue chips were used for separation of the epithelial and spindle cell components in12 cases of biphasic SS. For LCM, paraffin sections were deparaffinized in xylene, rehydrated through a graded alcohol series to distilled water, and then stained with hematoxylin. Epithelial cells and spindle cells were isolated from the slides using LCM (PixCell II Laser Capture Microdissection System; Arcturus Engineering Inc. Mountain View, CA, USA). The cells were captured using a 30-μm pulse to focally melt a thermoplastic membrane attached to a transparent flat cap (PixCell II Laser Capture Microdissection System). After LCM, the cap containing the captured tissue was placed on a 0.5-mL standard Eppendorf microfuge tube.

For tissue chip separation, regions of the epithelial cell and spindle cell components in the paraffin block were selected by comparing the results of hematoxylin eosin staining. A hollow needle (diameter, 1.0 mm) was used to puncture the selected area to a new small wax block.

DNA extraction and SNP array

Genomic DNA was isolated from the FFPE tumor samples using the QIAamp DNA Micro Kit (Qiagen Inc., Valencia, CA, USA) in accordance with the manufacturer’s instructions. The lowest amount of genomic DNA for which the genomic SNP chip analysis was successful was 1 μg. The DNA was extracted from 24 FFPE tissue samples (12 patients), with quality metrics of A260/280 = 1.7–2.0 and A260/230 > 1.6. DNA targets from patients were prepared and hybridized to Infinium HD Assay Super chips following the manufacturer’s recommendations. Quality analysis was performed by importing the CEL files into Illumina Human Exome-12v1.1 according to the “Quality Control Assessment in Genotyping Console”. Association analysis was performed for control versus case with the PLINK (http://pngu.mgh.harvard.edu/~purcell/plink) software, using the Fisher’s exact test, chi-square test and estimated odds ratio. Pearson chi-square values and P-values (chi-square and Fisher’s exact test) were calculated using Haploview 4.2. PLINK data.

Illumina exon microarray detection

A custom microarray with exon level resolution was developed to identify gross deletions and duplications. Study samples were processed using Illumina Human Exome-12v1.1 (Beijing Compass Biological Technology Co. Ltd; http://www.kangpusen.com/index.html). The microarray contains approximately 60,000 integrated oligonucleotide probes that have been annotated against the human genome assembly build. The probes’ density was increased in the exons and 300 bp flanking intronic sequence. In addition, probes were placed at every 2.5 Kb of the intronic sequence and heavily tiled in promoter regions. After validation runs, only probes with optimal performance were selected for the final array design.

Briefly, the standard DNA plate was prepared first, and then the standard QNT plate with diluted PicoGreen and the sample QNT plate with PicoGreen and DNA were prepared. Subsequently, the DNA samples were shifted to the MSA1 plate. The samples were denatured and neutralized, and then prepared for amplification by overnight incubation. The next day, the DNA was enzymatically fragmented using end-point fragmentation to avoid over-fragmentation. Then, the DNA samples were treated with 2-propanol and PM1 to precipitate MSA1, and the precipitated DNA was resuspended using RA1. Next, the fragmented resuspended DNA samples were dispensed on BeadChips. The BeadChips were incubated in the Illumina hybridization oven to hybridize the samples onto the BeadChips. The incubation time was appoximately16–24 h. The next day, the BeadChips were prepared for the staining process, and the unhybridized and non-specifically hybridized DNA was washed off. Labeled nucleotides were added to extend the primers hybridized to the DNA. The primers were stained, the flow-through chambers were disassembled, and the BeadChips were coated for protection. The iScan System and the Illumina BeadArray Reader were using by the Illumina GenomeStudio Genotyping Module. There are eight categories of internal control in an exon array Illumina SNP microarray experiment for each sample: staining, extension, hybridization, stringency, target removal, restoration, non-specific binding and non-polymorphic controls. Quality analysis of the Illumina HumanExome-12v1.1 report samples showed that the call rate was 0.7526.

Functional annotation

WebGestalt (WEB-based GEne SeT AnaLysis Toolkit; http://bioinfo.vanderbilt.edu/webgestalt) [6], which involves the KEGG database, Pathway Commons and WikiPathways Analysis, was used for functional enrichment analysis and hierarchical clustering of SNPs detected by SNP Chip. P-values were calculated at medium classification stringency using a modified Fisher’s exact test (EASE score), and terms with P-values of < 0.05 were considered significantly enriched. The enriched protein interaction network modules were visualized using String network online analysis (http://www.string-db.org/newstring_cgi/show_input_page.pl?UserId=nXWzWNEgLwwF&sessionId=Gfs9PMrT96ZB).

Results

Patients’ characteristics

The study cohort included 12 patients with definite epithelial and spindle cells in BSSs. The patients were predominantly male (seven [58.3 %] versus five female patients [41.7 %]), with the age at diagnosis ranging from 10 months to 64 years (mean age, 39 years). The tumors had a wide anatomical distribution; however, most arose in the upper and lower limbs (6, 38 %), and the head and neck region (5, 31 %). A total of five tumors (25 %) occurred in the retroperitoneum, hip, spinal canal, lung and thoracic wall. Clinical follow-up was available for 11 cases. The mortality rate was very high, that is, 63.6 % (7 of 11 patients) patients died of the disease or of subsequent metastasis. Some of the surviving patients experienced local recurrence or liver and lung metastases. The data are listed in Table 1.

Exome-chip analysis

We performed exome capture sequencing to validate the SNP chip data and to identify SNPs not present on the array. Approximately 249,101 single-nucleotide variants and small insertion and deletion changes were discovered for the epithelial and spindle cells components in the genomes of the 12 patients on comparison with the current reference haploid human genome sequence (Fig. 1a). The Fisher’s exact test based on PLINK analysis of the SNP-chip data showed that 336 SNPs had association P-values of less than 0.05 (P < 0.05, chi square test) Additional file 1. Because our studies were limited by the small number of patients with BSSs, we set an initial threshold for genome-wide significance of less than 10–3 by the chi-square test and further analyzed variants meeting this criterion. Analysis with the PLINK and Haploview software showed that a total of 23 SNPs in 23 genes (CTH, RBM44, LARS, HNRNPA, ADAMTS2, F13A, TNXB, ADAP1, PIK3CG, GCN1L1, NGDN, AHNAK2, ACAN, PRR14, DSEL, ETV2, DERL3, CELSR1 and BCORL1) had association P-values of less than 10–3 (Fig. 1b). All these SNPs were located in exonic regions. Mutations in 15 genes were specific for the spindle cell component (15/23, 65.2 %) (CTH, RBM44, LARS, HNRNPA, F13A, PIK3CG, GCN1L1, NGDN, AHNAK2, PRR14, RASAL3, DERL3, CELSR1, KLHL34 and EP300), and those in eight genes were specific for the epithelial cell component (8/23, 34.8 %) (ADAMTS2, TNXB, ADAP1, ACAN, DSEL, ETV2, MUM1L and BCORL1).

Fig. 1

Genome-wide analysis of a total of 2,345,678 SNPs in epithelial and spindle components in 12 cases of biphasic synovial sarcomas. a Manhattan plot showing negative log-transformed P-values of the case–control allele frequency significance on the y-axis. The color scale of the x-axis denotes chromosome numbers. Gene names associated with individual dots indicate SNPs of greatest significance or with potential disease relevance. b Mutation types of 343 SNPs meeting the established criterion for genome-wide significance (P < 0.05, chi square test)

The most highly ranked SNPs associated with genes are presented in Table 2. We identified 23 significantly mutated genes between the epithelial and spindle cell areas of SS. Most of these genes have not been previously reported in SSs, and four genes, neuroguidin (NGDN), RAS protein activator like 3 (RASAL3), KLHL34 and MUM1L1, have not been previously linked to cancer. Only one gene (EP300) has been reported in SS. NGDN is an EIF4E (MIM 133440)-binding protein that interacts with CPEB and functions as a translational regulatory protein during the development of the vertebrate nervous system [7]. RASAL3, a member of the RAS signaling pathway, is located at 19p13.12 and contains 18 exons; it was identified in 2004 [8]. KLHL34 is a member of the Kelch-like (KLHL) gene family, which encodes a group of proteins that generally possess a BTB/POZ domain, a BACK domain and 5–6 Kelch motifs. KLHL genes are responsible for several Mendelian diseases and have been associated with cancer. Further investigation of this family of proteins will likely provide valuable insights into basic biology and human disease [9]. The melanoma-associated antigen (mutated) 1-like 1 gene (MUM1L1) has four transcripts (splice variants). This gene encodes a protein that contains a mutated melanoma-associated antigen 1 domain. Proteins that contain mutated antigens are probably expressed at high levels in certain types of cancers. Multiple alternatively-spliced variants, encoding the same protein, have been identified. EP300 somatic mutations were significantly different between the epithelial and spindle cell components, and were mainly located in the DNA-binding domain and the mutated gene encoding histone H3 lysine acetyltransferases. EP300 is a very important factor in the TGF-β signaling pathway. SNPs located in genes with unknown functions or without obvious importance to disease pathogenesis are not discussed here.

Table 2 Highly ranked SNPs associated with BSS

Functional annotation of SS-associated variants

We analyzed 319 genes containing 336 SNPs significantly associated with BSSs by SNP-chip analysis using the WebGestalt functional clustering algorithm (Table 3). GO analysis showed three aspects: biological processes (biological regulation [43.9 %, 140/319] and metabolic processes [41.1 %, 131/319]); cellular components (membrane [45.8 %, 146/319] and nucleus [23.8 %, 76/319]); molecular function (protein binding [46.4 %, 148/319] and ion binding [26.3 %, 84/319]) (Fig. 2). SNPs were detected in genes with potential relevance for epithelial and spindle cell components of BSSs by Commons Pathway Additional file 2 and Functional Enrichment Analysis (WikiPathways Additional file 3 and KEGG pathways Additional file 4), mainly in genes such as those involved in the inflammatory response pathway, focal adhesion, the TGF-β signaling pathway, ECM–receptor interactions, phenylalanine metabolism and the JAK–STAT signaling pathway (Table 3). STRING network online analysis was used for additional evaluation of the enriched protein interaction network. The results showed enrichment of functional categories including focal adhesion, cytokine–cytokine receptor interactions, the JAK–STAT signaling pathway, the ERB signaling pathway, the cell cycle, adherens junctions and the Wnt signaling pathway (Fig. 3).

Table 3 Functional enrichment analysis of top ranked SNPs

Fig. 2

GO output enrichment analysis of differential SNPs for target genes. Distribution of significant genes in gene function (molecular function) and cell components (cellular component) categories, and those that participate in biological processes (in process)

Fig. 3

STRING network analysis reveals a database of known and predicted protein interactions. The interactions include direct (physical) and indirect (functional) associations. The network has been simplified for clear illustration of genes of interest. Stronger associations are represented by thicker lines. The significant genes were mainly TP53, AKT1, SRC, EGF, EGFR, PIK3CG, EP300 and FYN, and enriched functional categories included focal adhesion, cytokine–cytokine receptor interactions, the ERBB signaling pathway, the cell cycle, adherens junctions and the JAK–STAT signaling pathway

Discussion

SSs harbor a chromosomal translocation t(X; 18)(p11.2; q11.2), which produces the SS-specific fusion gene SYT–SSX. Although the precise function of SYT–SSX has not yet been studied, accumulating evidence suggests its role in gene regulation via epigenetic mechanisms, and the product of SYT–SSX target genes may serve as biomarkers of SS [10, 11]. However, lack of knowledge regarding the cell-of-origin of SS has hampered identification of its targets [12]. The clinical outcomes of SS are very poor and approximately 90 % of patients die because of metastasis. Understanding the process by which tumor cells destroy the basement membrane, invade and metastasize is essential for the advancement of SS treatment strategies and improvement of survival. An intriguing observation in SS is that the specific gene fusion (SYT–SSX1 vs. SYT–SSX2) correlates strongly with the tumor phenotype (monophasic vs. biphasic histology, as defined by the presence of glandular epithelial differentiation with lumen formation), and almost all BSS has been shown to harbor the SYT–SSX1 fusion gene [13, 14]. Our results showed that most cases are SYT–SSX1 fusion genes (8/11). Moreover, SYT–SSX fusion transcripts are found in both epithelial and spindle cell areas of BSS [15].

The recent advances in high-throughput sequencing technologies are rapidly becoming common practice. A large amount of data are generated using these methods and are becoming an important resource for deciphering the genotype underlying a given phenotype in clinical diagnosis, cancer research and molecular targeted therapies for various cancers [4, 16]. Because of the relative rarity of SS and its poor prognosis, it is not easy to collect adequate numbers of fresh samples of SS for a genome-wide association study. Therefore, it is very difficult to analyze subtle differences in pathogenesis and clinical benefits on the basis of the histological type. Therefore, future more in-depth studies should consider these issues in the study design when FFPE tissue samples are used. Samples stored in diagnostic pathology archives represent an invaluable biobank for retrospective clinical and molecular research. This is especially true in the case of SS, which is a rare malignant tumor, and FFPE tissue samples can alleviate or eliminate the need for the tedious collection and storage of cryopreserved clinical samples. It is more difficult to extract nucleic acids from FFPE tissue because of the need to remove the paraffin and to counteract covalent protein–DNA interactions that result from the fixation process [17]; however, Schweiger et al. [18], who first reported NGS-based analysis of DNA (DNA-Seq), had in fact used FFPE samples in their study. Recently, Eliezer et al. [19] also described an approach for generating high-quality whole-exome sequencing (WES) data from archival tumor material and validated WES data obtained from FFPE tumor samples by using corresponding WES data from frozen samples. The results showed that the ability to detect base mutations with sufficient power was equivalent regardless of whether frozen or FFPE tissue-derived genomic DNA was used for WES. The power of NGS (involving genomic SNP chips) for in-depth analysis of large numbers of short sequences potentially makes this an ideal technology in the case of the fragmented nucleic acids that are usually extracted from FFPE specimens [17]. Therefore, the development of reliable NGS-based methods for use with low-quality, FFPE tissue-derived nucleic acids would enable the use of samples from diagnostic pathology archives for high-throughput profiling, thereby facilitating extensive retrospective clinical studies. Our studies showed that these two methods can be used to extract good-quality DNA for detection with the exome SNP chip (call rate, 0.7526), using FFPE SS samples.

Morphologically, the most outstanding characteristic of BSSs is the mixture of glandular solid epithelial sheets or cystic papillary epithelial components and spindle mesenchymal elements. The proportion of the epithelial component varies widely. Some biphasic components form glands with a highly atypical epithelium, and the spindle cell component is composed of uniform mildly hyperchromatic cells. The broad spectrum of histopathological and molecular sarcoma subtypes, as well as the clinical behavior of these tumors, is the reason for the delay in clinical studies and preclinical discoveries. Mertens et al. [20] reported that cytogenetic analysis has not only provided important information on the pathogenesis of soft-tissue tumors but, by identifying distinct chromosomal rearrangements in different histopathological entities, has also come to serve as a valuable diagnostic tool. Therefore, to isolate pure tumor cell populations, we demonstrate the efficacy of LCM coupled with a cored tissue chip for the isolation of DNA from epithelial and spindle components of SSs. LCM is a well-established method for the isolation of cells from biological mixtures, even when cells are found in low abundance [21]. In our previous study [22], we found significant differences between different histological subtypes and epithelial–mesenchymal composition with respect to the expression of the TGF-β1 pathway and EMT-related proteins in SS. Our functional annotation of SNP-chip data suggests that the differing morphologies of BSSs indicate different genomic changes. The differential mutations had 336 SNPs, which contained 319 genes (including TGF-β1 pathway-related genes). For example, exm1611770 in EP300, exm412589 in SPP1 and exm648916 in PIK3CG were found to be strongly associated with BSSs. Therefore, these data perhaps support the hypothesis that neoplastic EMT contributes to the transition between the spindle cell and epithelial morphologies of SS.

Although our SNP analysis points to disease associations with many gene variants, the contribution of hereditary factors to the SS phenotypes is currently unknown. WebGestalt [6] is a suite of tools for functional enrichment analysis in various biological contexts (including GO Slim classification, directed acyclic graph [DAG] structure, KEGG pathway, WikiPathways and Pathway Commons pathway). The WebGestalt results showed that the differential SNPs in genes for functional enrichment were mainly related to the inflammatory response pathway, adhesion, ECM–receptor interactions, the TGF-β signaling pathway, the JAK–STAT signaling pathway, phenylalanine metabolism, the intrinsic pathway and fibrin formation. Moreover, those pathways that closely correlated with tight junction genes served as a signature for epithelial-like cancer cells or some invasion- and metastasis-associated genes. Kohn et al. [23] demonstrated that some highly correlated genes, including genes involved in interactions at tight junctions, adhesion junctions, desmosomes, transcription regulation of cell–cell junction complexes, epithelial vesicle traffic and epithelial Ca(+2) signaling, were implicated in epithelial functions in NCI-60 human tumor cell lines and the CCLE cell lines of the Broad Institute. Saito [24] provided a good model for this epithelial differentiation, which shows a possible mechanism for the aberrant mesenchymal-to-epithelial transition (MET) of SS and suggested that it would be better to consider it as an epithelial-to-mesenchymal transition (EMT). Therefore, these findings suggested that epithelial-like cancer cells or aberrant MET were present by analyzing the regulation of networks, whether carcinoma or sarcoma. These molecular signatures are not only important for appropriate diagnosis but also contribute to the investigation of the tumorigenic and metastatic mechanisms of cancer [25]. Our results also suggested that differing morphologies are due to different genomic changes.

Conclusions

Clinical applications of genomic biomarkers have been rapidly expanded and developed. Here, we performed a genome-wide SNP analysis of the epithelial and spindle cell components of BSSs. We found some significant mutations in genes whose functional annotations suggested involvement in cell adhesion, ECM–ECM receptor interactions, the TGF-β signaling pathway and cell junctions and signaling. These findings closely correlate with our previous immunohistochemical studies on SS, wherein TGF-β1 was found to induce mesenchymal–epithelial transition and to regulate biphasic differentiation. Such annotations provide a framework and a new target for treatment along with strategies for continued research on the mechanism.

Availability of supporting data

The data discussed in this publication have been deposited in LabArchives and are accessible through dataset DOI: 10.6070/H4Z31WP8 (http://www.labarchives.com/).