Early-Stage Induction of SWI/SNF Mutations during Esophageal Squamous Cell Carcinogenesis

The SWI/SNF chromatin remodeling complex is frequently inactivated by somatic mutations of its various components in various types of cancers, and also by aberrant DNA methylation. However, its somatic mutations and aberrant methylation in esophageal squamous cell carcinomas (ESCCs) have not been fully analyzed. In this study, we aimed to clarify in ESCC, what components of the SWI/SNF complex have somatic mutations and aberrant methylation, and when somatic mutations of the SWI/SNF complex occur. Deep sequencing of components of the SWI/SNF complex using a bench-top next generation sequencer revealed that eight of 92 ESCCs (8.7%) had 11 somatic mutations of 7 genes, ARID1A, ARID2, ATRX, PBRM1, SMARCA4, SMARCAL1, and SMARCC1. The SMARCA4 mutations were located in the Forkhead (85Ser>Leu) and SNF2 family N-terminal (882Glu>Lys) domains. The PBRM1 mutations were located in a bromodomain (80Asn>Ser) and an HMG-box domain (1,377Glu>Lys). For most mutations, their mutant allele frequency was 31–77% (mean 61%) of the fraction of cancer cells in the same samples, indicating that most of the cancer cells in individual ESCC samples had the SWI/SNF mutations on one allele, when present. In addition, a BeadChip array analysis revealed that a component of the SWI/SNF complex, ACTL6B, had aberrant methylation at its promoter CpG island in 18 of 52 ESCCs (34.6%). These results showed that genetic and epigenetic alterations of the SWI/SNF complex are present in ESCCs, and suggested that genetic alterations are induced at an early stage of esophageal squamous cell carcinogenesis.

The components of the SWI/SNF complex are also inactivated by aberrant DNA methylation of promoter CpG islands [13,19], which is known to be involved in the repression of gene transcription. Components of the SWI/SNF complex, ACTL6B, SMARCA2, and SMARCD3, and those of the other types of chromatin remodeling complex, ATRX and SMARCA1, are aberrantly methylated in gastric cancers [13]; ARID1A in invasive breast cancers [19]; ARID1B in pancreatic cancers [20], and ACTL6B in hepatocellular carcinomas (HCCs) [21]. However, the presence of aberrant methylation of the components of the SWI/SNF complex in ESCCs is still unclear.
In this study, we aimed to clarify, in ESCC, 1) what components of the SWI/SNF complex have somatic mutations by deep sequencing using a bench-top next generation sequencer to overcome the intrinsic limitation in the reading depth of exome-sequencing, 2) what components have aberrant methylation, and 3) when somatic mutations of the SWI/SNF complex occur. It was found that genetic and epigenetic alterations of the SWI/SNF complex are present in ESCCs, and it was suggested that genetic alterations are induced at an early stage of esophageal squamous cell carcinogenesis.

Clinical samples
Ninety-two primary ESCC samples and their corresponding non-cancerous tissue samples were endoscopically collected from ESCC patients with written informed consents. The collected samples were stored in RNAlater (Life Technologies, Carlsbad, CA, USA) at -80°C until the extraction of genomic DNA. Clinical information of the 92 ESCCs is listed in Table 1. The study was approved by the Institutional Review Boards of the National Cancer Center. Genomic DNA was extracted from ESCC samples by the standard phenol/chloroform method, and was quantified using a Quant-iT PicoGreen dsDNA Assay Kit (Life Technologies).
The sequences obtained were mapped onto the human reference genome (hg19). Somatic mutations in individual ESCC samples were identified by subtraction of the sequence variations also detected in the corresponding non-cancerous tissue of the cancer sample. Somatic mutations identified using the Ion Proton Sequencer were confirmed by Sanger sequencing of amplified DNA using the primers listed in S1 Table. As for mutations with a low frequency, amplified DNA was cloned into pGEM-T Easy vector (Promega, Madison, WI, USA), and sequences were confirmed by analysis of 10 pools of four clones (40 clones).

Analysis of a cancer cell fraction in an ESCC sample
The cancer cell fraction of an ESCC sample with mutation(s) of the SWI/SNF complex was analyzed by measuring DNA methylation levels of three genomic regions, TFAP2B, ARHGEF4, and RAPGEFL1, which are specifically methylated in ESCC cells [23]. The highest methylation level of the three genomic regions was defined as the cancer cell fraction, as described previously [23]. The eight ESCC samples had cancer cell fractions of 23-87% (mean 54%) (S3 Table).

Expression analysis
Genome-wide gene expression analysis was conducted using a GeneChip Human Genome U133 Plus 2.0 expression microarray (Affymetrix, Santa Clara, CA), as described previously [25,26]. Obtained signal intensity of an individual probe was normalized so that mean signal intensity of all the probes would be 500. Mean signal intensity of all the probes in an individual gene was defined as its transcription level, and genes with 250 or more of signal intensities were defined as expressed genes [25].
Gene-specific expression of ACTL6B in ESCC cell lines and non-cancerous esophageal tissues was analyzed by quantitative RT-PCR as described previously [25], using primers listed in S2 Table. IMR-32 and KELLY were used as positive controls with ACTL6B expression based upon the findings in the Cancer Cell Line Encyclopedia (CCLE) [27].

Statistical analysis
The association between SWI/SNF alterations, namely SWI/SNF mutations and ACTL6B methylation, and tumor characteristics, namely clinical T stage, clinical N stage, and clinical M stage, was evaluated by the Fisher exact test.
The somatic mutations were located in various functional domains (Fig 1B). The SMARCA4 mutations were located in the Forkhead (85Ser>Leu) and SNF2 family N-terminal (882Glu>Lys) domains. The PBRM1 mutations were located in the bromodomain (80Asn>Ser) and the HMGbox domain (1377Glu>Lys). The presence of these somatic mutations was confirmed by Sanger sequencing. These results showed that various genes encoding components of the SWI/SNF complex were mutated in ESCCs.

Somatic mutations were present in most cancer cells in individual ESCCs
To analyze the timing of the somatic mutations of the SWI/SNF complex, a cancer cell fraction was estimated for each of the eight ESCC samples with mutation(s) of the SWI/SNF complex, and the association between the fraction and mutant allele frequency was analyzed. Theoretically, in the case that all the cancer cells in an ESCC sample have a somatic mutation on one allele of a specific gene and allelic imbalance of the region is absent, a mutant allele frequency is expected to be 50% of a cancer cell fraction (Fig 2A). The mutant allele frequency of five of the eight ESCC samples (#85, #89, #94, #127, and #169) was lower than their cancer cell fraction in the same samples, and ranged from 31 to 77% (mean 61%) of the cancer cell fraction (Fig 2B). In contrast, the mutant allele frequency of the other three ESCC samples (#20, #126, and #176) was higher than their cancer cell fraction, and ranged from 107% to 145% (mean 121%) of the cancer cell fraction (Fig 2B). This result showed that most of the cancer cells in some ESCC samples had SWI/SNF mutations on one allele, and suggested that somatic mutations of the SWI/SNF complex are induced at an early stage of esophageal cell carcinogenesis.

Aberrant DNA methylation of ACTL6B was present in ESCCs
DNA methylation data were available from our previous study for 52 of 92 ESCCs [23]. Eighteen of the 52 ESCCs (34.6%) had aberrant methylation of ACTL6B at its promoter CpG island, but normal esophageal sample and non-cancerous tissue sample did not (Fig 3A and 3B). As for the other components of the SWI/SNF complex, ACTL6A, ARID1A, ARID2, PBRM1, SMARCA2, SMARCA4, SMARCAL1, SMARCB1, SMARCC1, SMARCC2, SMARCD1, SMARCD3, and SMARCE1, none of the 52 ESCCs had their aberrant methylation (Fig 3A). Three ESCC cell lines, KYSE30, KYSE140, and KYSE220, had a completely methylated ACTL6B promoter (Fig 3C), and ACTL6B was not expressed in these cell lines (Fig 3D). In contrast, two neuroblastoma cell lines, IMR-32 and KELLY, had an unmethylated ACTL6B promoter, and ACTL6B was expressed. These results supported that ACTL6B methylation could be involved in its silencing in tissues where it is expressed.
To assess the role of aberrant DNA methylation of ACTL6B in esophageal squamous cell carcinogenesis, ACTL6B methylation and expression were analyzed in non-cancerous  esophageal tissues. ACTL6B was unmethylated, but was not expressed (Fig 3A and 3D). This result suggested that ACTL6B methylation was a passenger in esophageal squamous cell carcinogenesis.
The association between alcohol/smoking exposure and aberrant DNA methylation of ACTL6B was analyzed in non-cancerous esophageal tissues. ACTL6B was not aberrantly methylated in non-cancerous tissues, regardless of alcohol/smoking exposure (Fig 3E).

SWI/SNF alterations were not associated with characteristics of ESCCs
The association between somatic mutations of the SWI/SNF complex, also ACTL6B methylation, and tumor characteristics was analyzed. Neither somatic mutations of the SWI/SNF complex nor aberrant ACTL6B methylation was associated with clinical T stage, clinical N stage, and clinical M stage (Table 3). This result showed that SWI/SNF alterations were not associated with characteristics of ESCCs.

Discussion
Components of the SWI/SNF complex, ARID1A, ARID2, ATRX, PBRM1, SMARCA4, SMAR-CAL1, and SMARCC1, were mutated in ESCCs. Among these, somatic mutations of SMAR-CAL1, and SMARCC1 were identified for the first time in ESCCs. Somatic mutations with low allele frequencies were successfully detected by deep sequencing. This suggested that deep sequencing focusing on specific sets of genes is useful to detect somatic mutations with low allele frequencies, which are generally difficult to detect by whole-exome sequencing.
Early-stage induction of alterations of the SWI/SNF complex during carcinogenesis has also been suggested for cancers other than ESCCs. During esophageal adenocarcinoma (EAC) development, somatic mutations of ARID1A and SMARCA4 are already present in benign metaplastic never-dysplastic Barrett's esophagus (NDBE) [28]. During gastric carcinogenesis, aberrant methylation of an ISWI component, SMARCA1, was detected in normal gastric tissues of people infected with Helicobacter pylori [13], a potent gastric cancer inducer. These early induction of genetic and epigenetic alterations of chromatin remodeling factors in multiple types of cancers suggested that their inactivation may be involved in predisposition to cancers (the formation of a field for cancerization [29]).
The mutant allele frequencies of three ESCCs (#20, #126, and #176) were higher than their cancer cell fractions. Theoretically, in the case that all the cancer cells in an ESCC sample have a somatic mutation on one allele of a specific gene and allelic imbalance of the region is absent, a mutant allele frequency is expected to be 50% of a cancer cell fraction (Fig 2A). Therefore, these three ESCCs might have a copy number loss of the wild type allele and this might result in the higher mutant allele frequency than cancer cell fractions.
Aberrant DNA methylation of promoter CpG islands is generally known to cause silencing of their downstream genes [30]. Regarding ACTL6B, aberrant methylation was found in its promoter CpG island, and the island was methylated in ESCCs. Expression analysis in cell panel. (B) DNA methylation status of ACTL6B around TSS. Aberrant methylation was induced around TSS of ACTL6B. (C) DNA methylation of ACTL6B in ESCC cell lines. Three ESCC cell lines, KYSE30, KYSE140, and KYSE220, had complete methylation of ACTL6B. (D) Expression levels of ACTL6B in ESCC cell lines and non-cancerous esophageal tissues. ACTL6B was not expressed in ESCC cell lines with its complete methylation (KYSE30, KYSE140, and KYSE220), and was expressed in neuroblastoma cell lines (NB) without its methylation (IMR-32 and KELLY). This supported that ACTL6B methylation could be involved in its silencing in tissues where it is expressed. At the same time, ACTL6B was not expressed in non-cancerous esophageal tissues without its methylation. This suggested that ACTL6B methylation was a passenger in esophageal squamous cell carcinogenesis. DNA methylation status of noncancerous tissues and cell lines was analyzed by qMSP and an Infinium HumanMethylation450 BeadChip array, respectively. (E) DNA methylation of ACTL6B in non-cancerous esophageal tissues. ACTL6B was not aberrantly methylated in non-cancerous tissues regardless of alcohol/smoking exposure.
doi:10.1371/journal.pone.0147372.g003 lines supported that ACTL6B could be silenced by aberrant methylation of its promoter CpG island. At the same time, ACTL6B was not expressed in non-cancerous esophageal tissues, which had an unmethylated ACTL6B promoter. Therefore, ACTL6B methylation was considered to be a passenger in esophageal squamous cell carcinogenesis. In contrast, somatic mutations of other components of the SWI/SNF complex were likely to be drivers because the genes with the mutations were expressed in non-cancerous esophageal tissues (Fig 3A). Mechanistically, disruption of the SWI/SNF complex has been reported to repress cell growth in other types of cancers [6,13]. Therefore, it is likely that inactivation of the SWI/SNF complex is involved in esophageal squamous cell carcinogenesis by promoting cell growth rate. At the same time, the SWI/SNF complex is known physiologically to regulate a large number of genes that are involved in a wide variety of cancer-related pathways, including the Wnt pathway, the p53 pathway, the MAPK pathway, DNA repair, cell cycle regulation, and apoptosis [31]. Therefore, complicated combinations of disruption of multiple cancer-related pathways might be alternative mechanisms of esophageal squamous cell carcinogenesis.
In conclusion, genetic and epigenetic alterations of the SWI/SNF complex are present in ESCCs, and genetic alterations were suggested to have been induced at an early stage of esophageal squamous cell carcinogenesis.
Supporting Information S1