Abstract
Lysine to methionine mutations at position 27 (K27M) in the histone H3 (H3.3 and H3.1) are highly prevalent in pediatric high-grade gliomas (HGG) that arise in the midline of the central nervous system. H3K27M perturbs the activity of polycomb repressor complex 2 and correlates with DNA hypomethylation; however, the pathways whereby H3K27M drives the development of pediatric HGG remain poorly understood. To understand the mechanism of pediatric HGG development driven by H3.3K27M and discover potential therapeutic targets or biomarkers, we established pediatric glioma cell model systems harboring H3.3K27M and performed microarray analysis. H3.3K27M caused the upregulation of multiple cancer/testis (CT) antigens, such as ADAMTS1, ADAM23, SPANXA1, SPANXB1/2, IL13RA2, VCY, and VCX3A, in pediatric glioma cells. Chromatin immunoprecipitation analysis from H3.3K27M cells revealed decreased H3K27me3 levels and increased H3K4me3 levels on the VCX3A promoter. Knockdown of VCX3A by siRNA significantly inhibited the growth of pediatric glioma cells harboring H3.3K27M. Overexpression of VCX3A/B genes stimulated the expression of several HLA genes, including HLA-A, HLA-B, HLA-E, HLA-F, and HLA-G. The expression of VCX3A in pediatric HGG was confirmed using a tissue microarray. Gene set enrichment analysis revealed that CT antigens are enriched in pediatric HGG clinical specimens with H3.3K27M, with the upregulation of IL13RA2 contributing to the enrichment significantly. These results indicate that the upregulation of CT antigens, such as VCX3A and IL13RA2, correlates with pediatric gliomagenesis. Mol Cancer Res; 16(4); 623–33. ©2018 AACR.
Introduction
Brain and central nervous system tumors are the leading causes of cancer-related death and the second most common cancers in children and adolescents aged birth to 19 years old (1). Among various childhood brain tumors, pediatric high-grade gliomas (HGG) is the deadliest type. Even with a combination of the most advanced treatments, few patients achieve long-term survival (1). Thus, understanding the molecular mechanisms of pediatric HGG and developing new therapeutic agents for pediatric HGG are of prime importance. Recently, epigenetic changes, mutations in or altered expression of epigenetic machinery, have been implicated in the development of various cancers including pediatric HGG (2, 3). Significantly, sequencing of pediatric HGG tumors revealed the c.83A>T mutations in H3F3A or HIST1H3B/C, which result in the lysine 27 to methionine (K27M) missense mutations in histone H3.3 or H3.1, respectively (4–6). The H3K27M mutations are enriched in pediatric HGG residing in the midline structures, such as pontine, brainstem, thalamus, and spinal cord (7–12). Significantly, approximately 80% of pediatric diffuse intrinsic pontine gliomas (DIPG) harbor the H3K27M mutations (4–6, 13). H3K27M mutations were also detected in adult HGG, albeit at a much lower frequency (14, 15), indicating there are substantial differences in the genetic and epigenetic mechanisms underlying the development of pediatric and adult HGG.
Because the H3K27M mutation is highly prevalent in pediatric HGG and occurs at the target site of polycomb repressive complex 2 (PRC2), the discovery attracted considerable attention. Follow-up investigations showed that H3K27M held the PRC2 activity in check, behaved dominant-negatively, and caused a global decrease of the H3K27me3 level (16, 17). DNA hypomethylation (decreased 5-methylcytosine, 5-mC) was observed in pediatric HGG by multiple studies (13, 18–22). A global reduction of the H3K27me3 level and DNA hypomethylation potentially act together to drive gliomagenesis; however, the detailed mechanisms whereby H3K27M drives gliomagenesis remains poorly understood.
Cancer/testis (CT) antigens are characterized by a unique class of tumor antigens, which are aberrantly expressed in a wide variety of tumors and are silent in normal tissues, except for the immune-privileged male germ cells (23–26). Due to their tumor-restricted pattern of expression and robust immunogenicity, CT antigens are considered to be ideal targets for cancer biomarkers and immunotherapy. So far, more than 250 CT antigens have been identified (27). Variable charge, X-linked/Y-linked (VCX/Y) genes, which are primate-specific genes and encode positively charged proteins of largely unknown function, are newly identified CT antigens in lung cancers (28). VCX/Y family proteins include six members, VCX3A, VCX, VCX2, VCX3B, VCY, and VCY1B. They share a highly homologous N-terminal region, and their C-terminal regions are composed of different numbers of copies of a ten-amino-acid repeat. The X-linked members are organized in tandem on a region of chromosome Xp22 and interspersed by other genes; the region could undergo nonallelic homologous recombination and other complex rearrangements, potentially resulting in X-linked ichthyosis and cognitive impairment (29–31).
In this study, we found H3.3K27M could activate the expression of multiple CT antigens in pediatric glioma cells, and demonstrated that IL13RA2 and members of the variable charge X/Y (VCX/Y) gene family were among the top upregulated genes in glioma cells stably expressing H3.3K27M. We further performed functional analysis of VCX3A/B, analyzed the epigenetic configurations at the VCX3A gene region, investigated the effects of VCX3A/B overexpression on the gene expression profiles of glioma cells, and examined the expression of VCX3A in pediatric glioma samples. In addition, we performed the gene set enrichment analysis (GSEA) using available data and revealed that H3K27M could indeed activate the expression of CT antigens in clinical samples.
Materials and Methods
Cell culture
Pediatric glioma cell line SF188 was obtained from Dr. Daphne Haas-Kogan (UCSF, San Francisco, CA), and Res259 was obtained from Dr. Michael Bobola (University of Washington, Seattle, WA). Both cell lines are cultured in high-glucose DMEM growth media (Gibco #11965) supplemented with 10% FBS. Cells were authenticated by short tandem repeat profiling (The Institute of Cancer Research, UK; ref. 32) and tested negative for mycoplasma contamination using the MycoAlert Mycoplasma Detection Kit (Lonza).
Generation of stable cell lines
Human H3F3A (C-terminal Myc-DDK–tagged) cloned into the pCMV6-Entry was purchased from OriGene. The H3.3K27M (c.83A>T) mutation was introduced by the QuikChange Lightning Site-Directed Mutagenesis Kit (Agilent Technologies) according to the manufacturer's protocol. Different isoforms of VCX3A/B were amplified using cDNA derived from Res259 cells as a template and cloned into pcDNA3 GFP LIC cloning vector (6D), which was a gift from Scott Gradia (Addgene plasmid # 30127). SF188 and Res259 cells were transfected using the Lipofectamine 2000 transfection reagent (Invitrogen; #11668019). Note that 1 mg/mL geneticin (G418; Thermo Fisher Scientific) was added to the culture medium for selecting stably transfected clones. The corresponding empty expression vectors were stably transfected into cells to serve as controls.
Cell proliferation and apoptosis assays
A total of 2 × 103 cells were plated in 96-well plates in 100 μL medium. Cell proliferation was analyzed by the CellTiter-Glo luminescent cell viability assay (Promega) for 5 consecutive days. For apoptosis assay, cells were grown to 70% to 80% confluence, harvested and stained using the Annexin V–FITC Apoptosis Detection Kit (Beyotime), and then analyzed by an FACScan flow cytometer (BD Biosciences). The data were analyzed with the FlowJo software (Treestar).
5-Methylcytosine dot blot assay
Genomic DNA was isolated with the DNeasy Blood & Tissue Kit (Qiagen; #69506), denatured in 0.4 mol/L NaOH, 10 mmol/L EDTA at 95°C for 10 minutes, and then neutralized with cold 2 mol/L ammonium acetate (pH 7.0). Two-fold serial dilutions of the denatured DNA samples were spotted onto a nitrocellulose membrane using the Bio-Dot apparatus (Bio-Rad). The membrane was washed with 2x saline-sodium citrate buffer, air-dried and cross-linked by UV irradiation, then blocked with 5% nonfat milk for 1 hour, and incubated with anti-5mC (Active Motif #39649, 1:10,000) overnight at 4°C. The membrane was visualized by enhanced chemiluminescence after incubating with horseradish peroxidase (HRP)–conjugated anti-mouse IgG secondary antibody. The same blot was subsequently stained with 0.01% methylene blue to verify equal loading.
Total RNA isolation
Total RNA was extracted using the TRIzol reagent (Invitrogen), genomic DNA was digested by RNase-Free DNase Set (QIAGEN; #79254), and the RNA was further purified using the RNeasy Mini Kit (Qiagen; #74106) following the manufacturer's instructions. Quality and concentration of RNA were determined by the Bioanalyzer 2100 (Agilent Technologies).
Gene expression analysis
Total RNA was amplified and labeled using the TargetAmp-Nano Labeling Kit for Illumina Expression BeadChip (Epicentre Biotechnologies; #TAN091096). Labeled cRNA was purified with the RNeasy mini Kit (QIAGEN; #74106) and hybridized on HumanHT-12 v4 Expression BeadChip microarrays (Illumina) according to the manufacturer's protocol. The hybridized arrays were scanned using the Illumina iScan (Illumina), and the image data were extracted using the Illumina GenomeStudio software. The raw data were deposited in the Gene Expression Omnibus data repository with accession number GSE102886.
Quantitative real-time reverse transcription PCR
Note that 1 μg of total RNA was used to synthesize the cDNA using the SuperScript III First-Strand Synthesis System (Invitrogen). qPCR was carried out using the LightCycler 480 SYBR Green I Master (Roche) in the CFX 96 thermocycler (Bio-Rad). The assays were performed in triplicate and repeated 3 times. The relative expression level of the gene of interest was normalized to GAPDH and calculated according to the 2−ddCt method (33). The primers used in this study can be found in Supplementary Table S1.
Chromatin immunoprecipitation
Chromatin immunoprecipitation (ChIP) was performed using the EZ-Magna ChIP A/G Chromatin Immunoprecipitation Kit (Millipore; #17-10086) following the manufacturer's instruction. Briefly, cells were cross-linked with 1% formaldehyde and sonicated to obtain DNA fragments between 0.3 and 1.0 kb. Chromatin was incubated with anti-H3K27me3 (Millipore; #07-449), anti-H3K4me3 (Millipore; #07-473) antibodies, and normal rabbit IgG (Cell Signaling Technology; #2729S) overnight, and the immune complexes were precipitated by protein A/G magnetic beads. DNA was extracted and used for ChIP-qPCR analysis. The enrichment levels are presented as a percentage of input chromatin.
Western blot
Total protein was extracted using RIPA buffer supplemented with proteinase inhibitor cocktail (Sigma-Aldrich). The extracted proteins were separated by SDS-PAGE and then transferred to nitrocellulose membrane (Pall Corporation). Membranes were blocked with 5% nonfat milk and then incubated with the following antibodies: anti-FLAG (Sigma; #F1804), anti-H3K27me3 (Millipore; #07-449), anti-H3K4me3 (Millipore; #07-473), anti-Histone H3 (Abcam; #ab1791), anti-Histone H3 (K27M mutant; Millipore; #ABE419), anti-VCX3A (Abnova; #H00051481-M01), anti-GAPDH (Cell Signaling Technology; #5174S), and anti–β-actin (Santa Cruz Biotechnology; sc-47778). Membranes were then incubated with HRP-conjugated secondary antibodies (Jackson ImmunoResearch), and the signals were detected by the Supersignal West Pico Chemiluminescent Substrate (Thermo Fisher Scientific).
VCX3A knockdown by siRNA
Two siRNAs targeting VCX3A (siVCX #1: Hs_VCX_8 FlexiTube siRNA, #SI04173568, and siVCX #2: Hs_VCX_10 FlexiTube siRNA, cat. #SI04187295; Qiagen) were transfected independently into cells at a final concentration of 25 nmol/L using Lipofectamine RNAiMAX (Life Technologies) as per the manufacturer's instruction. Forty-eight hours after transfection, total RNA and protein were harvested for analysis, and the CellTiter-Glo Luminescent Cell Viability Assay (Promega) was performed in triplicate for three independent experiments to examine cell growth.
DNA methylation analysis
Total genomic DNA was isolated from Res259 cells using the DNeasy Blood & Tissue Kit (Qiagen; #69506). The genomic DNA was subjected to bisulfite conversion using the EpiTect Bisulfite Kit (Qiagen; #59104). Primers targeting VCX3A or GAPDH for bisulfite sequencing were designed by MethPrimer (34), and the primer sequences are listed in Supplementary Table S1. The bisulfite-modified DNA was amplified via PCR and cloned into the T-Vector pMD20 (Takara; #3270). The plasmid DNA was sequenced to determine the CpG methylation status. Only sequences with higher than 99.5% bisulfite conversion rate were included in the analysis.
Immunohistochemical analysis
Tissue microarrays comprising 43 pediatric HGG were collated at the Institute of Cancer Research (London, UK) from multiple collaborating centers, all under approval from local ethical research committees. The slides were deparaffinized and hydrated, and antigen retrieval was performed using the 2100 retriever (Aptum Biologics) in R-Buffer A (pH 6.0). Intrinsic peroxidase activity was blocked using 3% hydrogen peroxide for 5 minutes. Slides were then incubated with the antibody against human VCX3A (Abnova; #H00051481-M01, 1:50) overnight at 4°C; then, the slides were stained using the SuperPicture 3rd Gen IHC Detection Kit (Thermo Fisher Scientific) according to the manufacturer's protocol.
GSEA
Publicly available pediatric HGG gene expression profiles (GSE34824, GSE36245, and GSE49822) were downloaded from Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/). The raw data were processed with Robust Multi-array Average (35) background correction and quantile normalization, combined and stratified for H3F3A status [H3F3A K27M (K27M), n = 14; H3F3A wild-type (WT), n = 51] as per the published sample annotation, and then subjected to the analysis. GSEA was performed using GSEA software (version 3.0; refs. 36, 37). The signature gene set consists of the full list of CT antigens deposited in CTdatabase (http://www.cta.lncc.br/), including 276 genes. The custom Gene Matrix Transposed (GMT) file containing the list of CT antigens that we constructed for GSEA analysis is attached as Supplementary File S2.
Results
H3.3K27M reduces the global level of trimethylation of H3K27
To determine the effects of H3.3K27M mutation on gliomagenesis in vitro, we first established two pediatric glioma cell lines stably carrying the mutation. Constructs encoding a FLAG-tagged WT or K27M mutant form of histone H3.3 were stably transfected into pediatric glioma cells, Res259 (WHO grade II) or SF188 (WHO grade IV) cells, whereas the empty vector was stably transfected into cells to serve as controls. Consistent with previous studies (16, 17), H3.3K27M significantly reduced the global levels of H3K27me3 in both cell lines, whereas did not affect the global levels of H3K4me3 (Fig. 1A). A modest reduction of global DNA methylation level was also observed by 5-mC dot blot assay in two independent clones of Res259 cells harboring the H3.3K27M mutation (Fig. 1B), consistent with the reduction of DNA methylation found in clinical pediatric glioma samples with the H3.3K27M mutation (13, 18–22). However, exogenous expression of H3.3K27M did not have an apparent effect on the cell proliferation and apoptosis in both cell lines (Fig. 1C and D).
Establishing and characterizing cell line models harboring the H3.3K27M mutation. A, Pediatric glioma cell lines, SF188 and Res259, were stably transfected with constructs encoding a C-terminal Myc-DDK–tagged WT (H3.3WT) form or a Lysine27Methionine (K27M)-mutant form of histone H3.3. The parental cell lines (Control) and cells stably transfected with the empty vector (Vector) were used as controls. Acid-extracted histones were subjected to the Western blot analysis with the indicated antibodies. B, Dot blot analysis using the 5-mC–specific antibody to detect global levels of 5-mC on genomic DNA from Res259 cells, which are stably transfected with the empty vector (Vector), or constructs encoding either H3.3WT, or Histone H3.3 harboring the K27M mutation (K27M). Two independent clones harboring H3.3K27M were examined. Methylene blue staining (bottom panels) was used to assure equal loading. C, The CellTiter-Glo (Promega) luminescent cell viability assays were performed to examine effects of H3.3K27M on cell growth. The assay was conducted in quadruplicate and repeated twice. Error bars represent ± SD of triplicates. D, The effects of H3.3K27M on cell apoptosis were examined by Annexin V labeling and propidium iodide staining.
H3.3K27M significantly upregulates multiple CT antigens
To gain insight into the molecular mechanisms underlying gliomagenesis driven by H3.3K27M, microarray analysis was carried out on Res259 with or without the K27M mutation. A total of 290 genes were found to be differentially expressed between Res259-K27M and Res259-vector control cells, of which 84 genes were downregulated (log2 fold change < –0.5, P < 0.05) and 206 genes were upregulated in Res259-K27M cell (log2 fold change > 0.5, P < 0.05; Fig. 2A). Kyoto Encyclopedia of Genes and Genomes analysis (38) revealed that the upregulated genes were enriched in pathways including transcriptional misregulation in cancer, MAPK signaling, cell cycle, p53 signaling, ErbB signaling, and Hippo signaling, whereas the downregulated genes were enriched in pathways including AMPK, metabolism, PPAR, and calcium signaling (Fig. 2B). Intriguingly, we found that 12 of the top 50 upregulated genes in cells with H3.3K27M are CT antigens (Fig. 2C). To validate the observations, the mRNA levels of selected CT antigens were examined by RT-qPCR in two independent H3.3K27M-expressing Res259 stable clones. As shown in Fig. 3A, H3.3K27M mutation significantly induced the expression of ADAM metallopeptidase domain 23 (ADAM23), ADAM metallopeptidase with thrombospondin type 1 motif 1 (ADAMTS1), interleukin 13 receptor subunit alpha 2 (IL13RA2), sperm protein associated with the nucleus, X-linked, family member A1, B1/B2 (SPANXA1, SPANXB1/2), variable charge, X-linked 3A (VCX3A), and variable charge, Y-linked (VCY) in Res259 cells, when compared with control cells stably transfected with the empty vector. A similar trend was also observed in SF-188 cells expressing H3.3K27M (Supplementary Fig. S1A). Significantly, it seems that positive stimulation of expression of CT antigens is K27M specific because ectopic expression of WT Histone H3 (H3.3WT) could not cause a similar change (Fig. 3A; Supplementary Fig. S1A). Among the CT antigens upregulated in H3.3K27M-carrying cells, genes of variable charge X/Y family, such as VCX3A and VCY, had the highest fold changes of expression. In addition, overexpression of VCX3A in H3.3K27M-carrying cells could be confirmed at the protein level by the Western blot analysis (Fig. 3B; Supplementary Fig. S1B). Therefore, we next focused on examining the potential involvement of VCX3A in gliomagenesis.
Impacts of H3.3K27M on gene expression in Res259 cells. A, Gene expression in Res259 cells was examined by HumanHT-12 v4 Expression BeadChip microarrays (Illumina). The volcano plot shows statistical significance (−log10 P value) plotted against log2 fold change of genes for Res259 cells harboring H3.3K27M against vector control cells. Differentially expressed genes (DEG) were selected by criteria of P < 0.05 (blue line) and absolute log2 fold change > 0.5 (red line). B, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of DEGs between Res259 cells harboring H3.3K27M and vector control cells. C, A pie chart indicates the percentage of CT antigens in the top 50 upregulated genes in Res259 cells carrying H3.3K27M as compared with vector control cells.
H3.3K27M activates CT antigens. A, mRNA expression of selected CT antigens was examined by RT-qPCR. Res259 cells, which were stably transfected with the empty vector (Vector), or constructs encoding either H3.3WT or Histone H3.3 harboring the K27M mutation (K27M), were subjected to experiments. Two independent K27M clones were used to ensure reproducibility. mRNA levels were normalized to GAPDH. Shown are the representatives of three independent experiments. Error bars represent ± SD of triplicates. **, P < 0.01; #, a primer set detecting all VCX/Y family genes was used. B, Western blot analysis of VCX3A in nontransfected Res259 cells, or Res259 cells stably transfected with the empty vector (control), or constructs encoding H3.3 WT, or Histone H3.3 harboring the K27M mutation (H3.3K27M). GAPDH was used as a loading control.
H3.3K27M alters the epigenetic modifications of VCX3A genomic locus
To get an insight into the mechanisms by which H3.3K27M causes the upregulation of VCX3A expression, we first examined the changes of the epigenetic configurations of the VCX3A genomic locus (Fig. 4A) as a function of H3.3K27M mutation. By chromatin immunoprecipitation coupled with quantitative PCR (ChIP–qPCR), we found the level of H3K4me3 on the promoter of VCX3A increased, and the level of H3K27me3 on the promoter of VCX3A decreased in the Res259 cells harboring H3.3K27M (Fig. 4B).
H3.3K27M alters the epigenetic modifications of the VCX3A genomic locus. A, A map of the VCX3A gene showing positions of exons (gray rectangles), the intragenic CGI (orange diagonal stripes), and two sets of primers (arrows) used for the ChIP-qPCR analysis. TSS, transcription start site; F, forward primer; R, reverse primer. B, Alternations of the indicated histone modification marks on VCX3A genomic locus in Res259 cells carrying H3.3K27M were detected by ChIP followed by quantitative real-time PCR. The enrichment of marks is represented by the percentage of ChIP input. Res259 cells stably transfected with the empty vector (Vector) served as a control. Left: results from primers 1F and 1R; Right: results from primers 2F and 2R. *, P < 0.05 and **, P < 0.01; Student t test. Error bars represent ±SD of triplicates. C, Increased DNA methylation of an intragenic CGI of VCX3A in Res259 cells carrying H3.3K27M. DNA methylation patterns were determined by bisulfite sequencing. Res259 cells stably transfected with the empty vector served as a control. Black and white circles represent methylated and unmethylated cytosines, respectively.
Previous studies reported that H3.3K27M could cause not only the reduction of global H3K27me3 level, but also DNA hypomethylation (13, 18, 19). Thus, we next examined the DNA methylation status at the VCX3A locus of the Res259 cells. By bioinformatics analysis, we found there were no CpG islands (CGI) at the promoter of all four VCX genes. Instead, there were CGIs in the gene bodies of VCX genes (Supplementary Fig. S2A and S2B). We performed bisulfite sequencing targeting an intragenic CGI of VCX3A, which overlaps with the exon 2 of VCX3A (Fig. 4A). Surprisingly, we observed an increased DNA methylation at this CGI in the cells carrying H3.3K27M compared with empty vector–transfected control (Fig. 4C), whereas no changes of DNA methylation status were found in an intragenic CGI of the housekeeping gene GAPDH (Supplementary Fig. S3A and S3B). Methylation of intragenic CpG islands (iCGI) is reported to be positively correlated with gene transcription by unclear mechanisms (39, 40). Hypermethylation of the iCGI of VCX3A indicates the correlation between gene activation, methylation of iCGI remains intact in the mutant cells, and the observed hypermethylation of the iCGI of VCX3A might be a consequence of active VCX3A transcription. Together, these results support the changed epigenetic landscape caused by H3.3K27M that was associated with the upregulation of VCX3A.
Knockdown of VCX3A inhibits cell growth
We next set out to knock down VCX3A in Res259 (WHO grade II) and SF188 (WHO grade IV) cells harboring H3.3K27M to explore their function. Two siRNAs (Qiagen) were used to target VCX3A gene. Due to high similarity of VCX/Y family genes, Hs_VCX_8 FlexiTube siRNA (siVCX #1) targets all the VCX/Y genes, whereas Hs_VCX_10 FlexiTube siRNA (siVCX #2) targets VCX3A, VCX3B, and VCX, but not VCX2 and VCY. As shown in Fig. 5A, both siRNAs could knock down VCX3A efficiently. The knockdown of VCX3A significantly inhibited the proliferation of the two cell lines carrying the K27M mutation (Fig. 5B). However, the cell growth–inhibitory effect could not be observed in Res259 cells without the K27M mutation (Fig. 5B), which could be explained by the low expression of VCX3A in the vector control Res259 cells. In all, these results suggested that VCX3A might play an oncogenic role in pediatric HGG.
Knockdown of VCX3A inhibits cell growth. A, Knockdown of VCX3A with siRNAs in Res259-H3.3K27M and SF188-H3.3K27M cells. The knockdown effects were examined by RT-qPCR (left) and Western blot analysis (right). GAPDH was used as a loading control [siControl, Negative Control siRNA (Qiagen); siVCX #1, Hs_VCX_8 FlexiTube siRNA; siVCX #2, Hs_VCX_10 FlexiTube siRNA (Qiagen); **, P < 0.01]. B, The effects of VCX3A knockdown on cell growth. Cells were transfected with negative control siRNA or two independent siRNAs against VCX3A. Forty-eight hours after transfection, cell growth was examined by a CellTiter-Glo kit (Promega) in quadruplicate. Three independent experiments were performed, and results are presented as relative luminescence unit (RLU) fold change compared with the value measured in cells transfected with control siRNA. The cell type is indicated at the top of each graph (**, P < 0.01).
VCX3A/B overexpression stimulates the expression of HLA
To perform functional analysis of VCX3A, we next set out to clone the VCX3A gene from the cDNA of Res259. We designed primers targeting the coding region of VCX3A; however, because the coding regions of VCX3A and VCX3B have identical 5′ and 3′ ends, the primers we used also targeted VCX3B unavoidably. Indeed, DNA sequencing revealed that three genes we cloned bear N-termini identical to the N-terminus of VCX3A (NCBI Ref Seq: NM_016379.3), which are followed by 10, 7, or 2 copies of a ten-amino-acid repeat, respectively, at their C-termini (Fig. 6A). Surprisingly, the composition and organization of the C-termini of the genes share higher similarity to VCX3B (NCBI Ref Seq: NM_001001888.3) than to VCX3A (Supplementary File S1), suggesting VCX gene family locus is polymorphic and subject to alternative splicing or recombination. However, further studies are required to clarify the precise mechanisms involved. In this study, we distinguished the VCX family genes we cloned by the number of the tandem repeats they bear, and named them as VCX3A/B-10R, VCX3A/B-7R, and VCX3A/B-2R, due to their resemblance to a hybrid of the reference sequences of VCX3A and VCX3B. Next, we generated Res259 cells stably expressing different isoforms of VCX3A/B fused with GFP. Consistent with the presence of a putative bipartite nuclear localization sequence (KRKSSSQPSPSDPKKKTT) at the N-terminus of VCX3A/B, all three isoforms of VCX3A/B-GFP were observed to be primarily localized to the nucleus (Supplementary Fig. S4), suggesting that VCX3A/B members are nuclear proteins.
A, Diagrams of VCX3A/B cloned from Res259 cells. The ten-amino-acid repeats with different variations are color coded. The genes were named VCX3A/B-10R, VCX3A/B-7R, and VCX3A/B-2R, respectively, according to the number of repeats (R) they have. B, Gene expression changes caused by VCX3A/B expression in Res259 cells were examined by HumanHT-12 v4 Expression BeadChip microarrays (Illumina). The volcano plot shows statistical significance (−log10 P value) plotted against log2 fold change of genes for Res259 cells expressing different isoforms of VCX3A/B-GFP versus cells overexpressing GFP alone. The downregulated genes are colored in blue (log2 fold change < −0.5, P < 0.05), and the upregulated genes are colored in red (log2 fold change > 0.5, P < 0.05). C, Gene ontology analysis of differentially expressed genes between RES259 cells expressing VCX3A/B-GFP and GFP only. D, Heatmap of upregulated genes in VCX3A/B-overexpressing Res259 cells involved in antigen processing and presentation of endogenous peptide antigen via MHC class I (GO:0002474). E, The mRNA expression levels of HLA-A, -B, -E, -F, and -G in Res259 cells ectopically expressing different isoforms of VCX3A/B-GFP (2R, 7R, and 10R) were evaluated by RT-qPCR, compared with cells expressing GFP alone. Shown are a representative of three independent experiments; error bars represent ±SD of triplicates (**, P < 0.01; *, P < 0.05).
To understand the function of VCX3A/B proteins and their potential involvement in gliomagenesis, we next examined the changes of gene expression profiles caused by VCX3A/3B overexpression by microarray analysis. Briefly, RNA was extracted from Res259 cells stably expressing the different isoforms of VCX3A/B-GFP, or GFP control, and subjected to an Illumina HT12.2 Bead CHIP array analysis. As shown in Fig. 6B and Supplementary Fig. S5, the different isoforms of VCX3A/B caused similar gene expression changes. Therefore, the datasets were combined for subsequent analysis. A total of 138 genes were found to be differentially expressed in VCX3A/B-overexpressing Res259 cells compared with GFP-control cells, of which 35 genes were downregulated (log2 fold change < –0.5, P < 0.05) and 103 genes were upregulated in Res259-VCX3A/B cells (log2 fold change > 0.5, P < 0.05). Gene ontology analysis revealed that the upregulated genes in VCX3A/B-overexpressing cells are primarily involved in immune response (Fig. 6C). Significantly, VCX3A/B overexpression caused the upregulation of MHC class I genes, including HLA-A, HLA-B, HLA-E, HLA-F, and HLA-G (Fig. 6D and E). MHC class I molecules are primarily involved in binding to and presenting antigens on the cell surface for recognition by cytotoxic T cells. Upregulation of MHC class I molecules by VCX3A/B suggests that VCX3A/B proteins are processed by MHC class I proteins in cells. We further examined the expression of HLA genes in Res259 cells carrying the H3.3K27M mutation by RT-qPCR. Indeed, minor upregulations of HLA genes, particularly of the HLA-B, were observed in cells carrying H3.3K27M (Supplementary Fig. S6).
The expression of VCX3A and other CT antigens in pediatric HGG
We next examined the expression of VCX3A protein with tissue microarrays comprising 43 pediatric and young adult HGG. VCX3A showed primarily nuclear staining and was strongly expressed in normal testis but was negative in normal brain (Fig. 7A). Of three samples harboring H3.3K27M, one showed strong expression of VCX3A, one showed moderate expression, and the third showed negative/equivocal expression. Among the 40 tumors with WT Histone H3, 1 showed strong expression of VCX3A (2.5%), and 3 showed moderate expression (7.5%; Fig. 7A and B; Supplementary Table S2), indicating the expression of VCX3A was not limited to tumors carrying the H3K27M mutation and other mechanisms also potentially activate the expression of VCX3A in pediatric HGG.
The expression of VCX3A and other CT antigens in pediatric HGG. A, Immunohistochemistry for VCX3A protein in pediatric HGG. The tissue microarrays were stained using SuperPicture 3rd Gen IHC Detection Kit (Thermo Fisher Scientific). Cell nuclei were counterstained with hematoxylin. The genotype of H3F3A was indicated on the top of each panel. K27M, histone H3 lysine27methionine mutation; scale bar, 50 μm. B, The bar plot of VCX3A expression in pediatric HGG stratified by H3F3A status, depicting the percentage of cases with the indicated signal grades. –, negative; +/–, equivocal; +, weak positive; ++, strong positive. C, GSEA of a signature gene set for CT antigens in pediatric HGG with H3F3AK27M mutation versus those with WT H3F3A. The signature gene set consists of the full list of CT antigens curated in CTdatabase (http://www.cta.lncc.br/). Publicly available pHGG gene expression profiles (GSE34824, GSE36245, and GSE49822) were processed using RMA (quantile normalization), combined and stratified for H3F3A status [H3F3A K27M (K27M), n = 14; H3F3A WT, n = 51] as per the published sample annotation, and then subjected to the analysis. The normalized enrichment score (NES), the nominal P value (NOM P-val), and the false discovery rate Q value (FDR q val) are shown at the upper right corner of the graph. D, Upregulation of IL13RA2 in pHGG with the H3.3K27M mutation. The gene expression data of IL13RA2 were retrieved from Gene Expression Omnibus (GEO) of NCBI (GSE34824, GSE36245, and GSE49822).
To further examine whether H3.3K27M mutation could activate CT antigens, we collected and combined the published microarray datasets of pediatric HGG deposited in the Gene Expression Omnibus repository of NCBI, which include 14 samples with H3.3K27M mutation and 51 samples with WT H3.3 (5, 13, 18). We next performed GSEA using the full list of CT antigens curated in the CTdatabase (http://www.cta.lncc.br) as the Gene Matrix, which contains 276 genes (27). As shown in Fig. 7C, CT antigens are enriched in tumors with H3.3K27M mutation compared with those without the mutation with an enrichment score of 0.38 (P = 0.037; FDR q value = 0.159), indicating that the activation of CT antigens might be one of the hallmarks of pediatric HGG harboring H3.3K27M. Notably, IL13RA2, which has been explored as a therapeutic target for adult glioblastoma (41), is one of the top upregulated CT antigens in pediatric HGG with H3.3K27M (Fig. 7D).
Discussion
In this study, we found that H3.3K27M could activate the expression of multiple CT antigens. Among them, VCX/Y family, which was recently proposed as novel CT antigens in lung cancers (28, 42), was the most upregulated. VCX/Y genes are specific to primates and absent from nonprimate mammals. Expression of members of the VCX/Y gene family is restricted to testis. VCX/Y gene family contains four paralogs (VCX3A, VCX, VCX2, and VCX3B) on X chromosome and two paralogs (VCY, VCY1B) on Y chromosome. Their functions remain largely unknown. Deletion of VCX3A was observed in X-linked nonspecific mental retardation patients (29, 31). Jiao and colleagues found that VCX3A bound the 5′ end of capped mRNAs to prevent mRNA decapping and decay (43), and inhibit mRNA translation (44). The author further proposed that VCX3A modulates neuritogenesis through selective binding to mRNAs (44). Also, the enrichment of VCX/Y in nucleoli and the putative interaction of VCX3A with RPLP0 (Ribosomal Protein Lateral Stalk Subunit P0) also suggest their potential involvement in ribosome biogenesis (45). Consistent with the results in lung cancer, we observed knockdown of VCX3A inhibited the growth of the pediatric glioma cells harboring H3.3K27M, suggesting its potential involvement in gliomagenesis. However, more detailed studies are needed to validate the proposed functions of VCX3A and explore the unknown.
The expression of CT antigens is regulated epigenetically by DNA methylation within the promoter region and histone modifications (23, 24, 46, 47). Perhaps, the strongest evidence for the involvement of epigenetics in the regulation of CT antigens comes from the induction of the expression of CT antigens by chemical inhibitors of DNA methylation, or inhibitors of histone deacetylases and histone methyltransferases (23, 24, 46, 47). In this study, we provided evidence that the changed epigenetic landscape contributes to the upregulation VCX3A in the cells carrying H3.3K27M. Tissue microarray–based analysis showed that VCX3A was expressed in pediatric HGG, but the expression was not limited only to tumors harboring H3.3K27M, suggesting multiple pathways are involved in activation of VCX3A. The H3.3K27M mutation could cause the global reduction of H3K27me3 and DNA hypomethylation (13, 16–19, 21, 22), both of which would potentially affect the expression of CT antigens. Consequently, other than VCX/Y, we also observed the overexpression of CT antigens including ADAMTS1, ADAM23, SPANXA1, SPANXB1/2, and IL13RA2 in pediatric glioma cells carrying H3.3K27M.
In general, CT antigens are digested by the proteasome into small peptides, then transported into the endoplasmic reticulum (ER), and presented on the cell surface by MHC class I molecules. Mutations in or downregulation of MHC class I molecules is the mechanism most frequently exploited by tumor cells to escape from immune surveillance (48). In this study, we found overexpression of VCX3A/B in pediatric glioma cells strongly stimulates the expression of HLA-A, HLA-B, HLA-E, HLA-F, and HLA-G, indicating VCX3A/B are probably presented by the MHC class I–mediated pathway as antigens. Once VCX3A/B are presented, the tumor cells bearing them could potentially be recognized and eliminated by the host immune system. This could explain the observation that, although H3.3K27M activates the expression of VCX/Y in cell culture models, only part of the clinical samples with H3.3K27M has VCX/Y overexpression. However, the underlying mechanisms of upregulation of MHC molecules driven by VCX3A/B and its potential significance remain to be explored.
Immunotherapeutic approaches for treating brain tumors, including pediatric gliomas, have been actively explored (49–51). For example, a trial using the H3.3K27M peptide as a vaccine for the treatment of HLA-A2+ H3.3K27M-positive gliomas is ongoing (NCT02960230). Considering testis does not express MHC class I or II molecules and therefore being immune-privileged, CT antigens are proposed to be ideal targets for cancer immunotherapy because of the tumor-restricted pattern of expression and their strong immunogenicity in vivo (23, 26). Accordingly, clinical trials have utilized CT antigens as targets for adoptive T-cell therapy, or as vaccines against tumors, including gliomas (24, 25, 41, 52). Significantly, an adult glioblastoma patient had tumor regression after receiving chimeric antigen receptor–engineered T cells targeting IL13RA2 (41). In this study, we observed the upregulation of IL13RA2 in pediatric glioma cells carrying H3.3K27M and further confirmed its upregulation in pediatric HGG with H3.3K27M using published datasets.
In summary, we found that H3.3K27M could cause upregulation of multiple CT antigens in pediatric glioma cells, including IL13RA2 and VCX3A/B. Our data also indicate that VCX3A/B might be oncogenic. Thus, it is worth exploring the potential use of VCX3A/B and IL13RA2 as immunotherapeutic targets for pediatric HGG.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: H. Deng, G. Li
Development of methodology: H. Deng, L. Gong
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Deng, L. Gong, E. Cheung, C. Jones
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Deng, J. Zeng, L. Gong, G. Li
Writing, review, and/or revision of the manuscript: H. Deng, G. Li
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Deng, T. Zhang, L. Gong, H. Zhang
Study supervision: G. Li
Acknowledgments
This work was supported by the Science and Technology Development Fund of Macau (137/2014/A3 and 095/2015/A3) and the Research & Development Administration Office of the University of Macau (SRG201400015, MYRG201500232, and MYRG201700099).
The authors thank the iPSC Core of the University of Macau and Professor Guokai Chen for help with the mycoplasma detection assay.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Footnotes
Note: Supplementary data for this article are available at Molecular Cancer Research Online (http://mcr.aacrjournals.org/).
- Received August 23, 2017.
- Revision received December 7, 2017.
- Accepted January 16, 2018.
- Published first February 16, 2018.
- ©2018 American Association for Cancer Research.