Unveiling the conversion mechanism of glioblastoma from the mesenchymal to the proneural subtype driven by HDAC1/p-SMAD3-TP53I11 axis

Highlights

• •HDAC1 is highly expressed in MES GBM cells and silencing it weakens the MES characteristics of GBM.
• •HDAC1 combined with bevacizumab significantly inhibit the proliferation and invasion of MES GBM cells.
• •HDAC1/p-SMAD3 complex plays a role in transcriptional regulation through histone acetylation.
• •Inhibition of HDAC1 enhances the p-SMAD3 binding genome and thus promotes TP53I11 expression.

Abstract

Background

Mesenchymal glioblastoma (MES GBM) is characterized by rapid proliferation, extensive invasion, and formidable treatment resistance. We aimed to find the MES GBM subtype conversion mechanism.

Methods

HDAC1 expression was examined across various GBM subtypes through in vitro and in vivo experiments. The impact of HDAC1 inhibitors and bevacizumab on the phenotypic characteristics of MES cells was also assessed. Co-immunoprecipitation (Co-IP) and immunofluorescence techniques elucidated the epigenetic mechanism of HDAC1. Chromatin immunoprecipitation sequencing (ChIP-seq) and RNA-seq identified downstream transcribed genes, followed by the experiments to investigate their effects on MES GBM cells.

Results

Firstly, we observed overexpression of HDAC1 in MES-type cells and its positive correlation with MES representative genes. Inhibition or knockdown of HDAC1 transformed MES characteristics into proneural (PN) characteristics, prolonged survival in patient-derived xenograft (PDX) models, and suppressed in vitro cell proliferation and invasion. Additionally, bevacizumab significantly inhibited PN subtype GBM cells, and when combined with RG2833 (an HDAC1/3 inhibitor), it also inhibited the growth and proliferation of MES subtype GBM. RG2833 was found to enhance histone acetylation, promoting the binding of the transcription factor p-SMAD3 (Ser423 and Ser425) to the genome. Immunoprecipitation experiments revealed an interaction between p-SMAD3 and HDAC1. RNA-seq and ChIP-seq data analysis from MES cell lines before and after RG2833 treatment identified Tumor Protein P53 Inducible Protein 11 (TP53I11) as a downstream gene. Knocking down TP53I11 transformed PN subtype GBM into MES characteristics.

Conclusion

The study indicates that by intervening HDAC1/p-SMAD3-TP53I11, HDAC1 can serve as a promising therapeutic target for the treatment of mesenchymal glioblastoma.

Graphical abstract

Introduction

The standard therapeutic approach for adult glioblastoma (GBM) patients involves maximal safe resection, complemented by radiotherapy, chemotherapy, targeted pharmacotherapy, and tumor-treating fields (TTFields). Despite these interventions, the median survival time for adult patients with GBM remains just 14.7 months [1,2]. Intratumoral transcriptomic heterogeneity significantly contributes to the refractoriness of GBM. Advances in transcriptome analysis have led to the classification of GBM tumor cells into three distinct subtypes: classical (CL), mesenchymal (MES), and proneural (PN). The PN subtype exhibits a better prognosis and therapeutic responsiveness compared to the MES subtype [3,4]. Each subtype shows significant consistency in gene copy numbers, with the PN subtype displaying specific IDH1 and OLIG2 mutations. Most tumors with +7/-10 alterations belong to the CL subtype, while the MES subtype typically harbors NF1 mutations and PTEN loss [5]. These consistent genomic alterations likely account for the similarities in transcriptomes and phenotypes within each subtype. Similar to epithelial-mesenchymal transition (EMT), proneural-mesenchymal transition (PMT) facilitates invasion, proliferation, and treatment resistance in GBM [6,7]. Given the heterogeneous nature of tumors wherein almost all GBMs contain three distinct cell types whose proportions determine patient prognosis [3], it is hypothesized that promoting transformation of MES cells into PN cells within tumor tissues may extend overall patient survival.

The unstructured N-termini of core histones are characterized by a high abundance of lysine residues, which naturally carry a positive charge and exhibit an affinity for the negatively charged backbone of DNA. Acetylation neutralizes this positive charge, reducing histone-DNA interaction strength and enhancing chromatin accessibility for transcription, which is crucial in epigenetics [8]. Histone deacetylases (HDACs) form a class of co-regulated proteins that function as complexes. typically interact with transcription factors to exert its role in transcriptional regulation. These complexes enable HDACs to bind specific genomic sites, modulating the process of gene expression from DNA to mRNA [[9], [10], [11], [12]]. HDAC1, a key member, is extensively involved in gene repression through hypoacetylation of local chromatin domains. Epigenetic therapy strategies targeting HDAC1 have been widely reported, and numerous molecular inhibitors against HDAC1 have been developed and assessed in experimental studies across various cancers [13,14]. Clinical trials have shown therapeutic effects in treating lymphoma, hematological tumors, breast cancer, and gastrointestinal tumors [15,16]. HDAC1 has been found to promote the malignant progression of glioma and participate in the EMT process [[17], [18], [19], [20]]. However, its specific role in GBM molecular subtypes requires further investigation.

This study extracted the Cancer Genome Atlas (TCGA) clinical data and substantiated that MES subtype patients exhibit a significantly shorter median survival time compared to PN subtype patients. Targeting HDAC1 demonstrated its efficacy in inhibiting the proliferation of MES glioblastoma cells both in vitro and in vivo. RG2833 competitively binds to HDAC1, facilitating the transcription factor p-SMAD3 to bind specific genomic DNA sites, thereby promoting TP53I11-mRNA expression. This process leads to a reduction in MES characteristics and transformation the MES cells into PN cells. Furthermore, the ability of RG2833 to penetrate the blood-brain barrier provides compelling evidence for its potential as a targeted therapy option in clinical settings.

Materials and methods

Cell lines, cell culture conditions and GBM samples

Human glioma cells U251, purchased from ATCC (American Type Culture Collection, Manassas, VA, USA), were cultured in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% heat-inactivated fetal bovine serum (FBS; Hyclone, Logan, UT, USA). The acquisition methods for the N9 cell line, TBD0207B, and TBD0220L have been described previously [21]. All cell lines were sequenced by whole genome with purity greater than 99% and no mycoplasma contamination. These cells were grown in DMEM/F12 (1:1; Gibco BRL, Rockville, MA) with 10% FBS. All cell cultures were maintained at 37°C in a 5% CO₂ atmosphere. Additionally, data from 78 GBM samples, including information on known mutations, mRNA expression levels, and subtype classification, were obtained from TCGA.

GBM tissues and patient-derived cerebrospinal fluid exosomes

After obtaining informed consent from patients and approval from the ethics committee in accordance with the ethical standards of the 2008 Helsinki Declaration, tumor tissue and cerebrospinal fluid samples were randomly selected from four patients with GBM who underwent surgical resection (Table S1). Tumor grades were determined according to the 2021 World Health Organization (WHO) classification of nervous system tumors.

Exosomes were isolated using a micro exosome total RNA extraction kit (Absin, China). Following the completion of washing, digestion, and elution processes, the total RNA solution was obtained through multiple rounds of vigorous shaking, centrifugation, and adsorption.

SiRNA, reagents, plasmids, lentivirus and transfection

The siHDAC1, siHDAC2, siHDAC3, siHDAC8, siSMAD3 and siTP53I11 was synthesized in Ribobio (Guangzhou, China) (Table S2). The open reading frames (ORFs) of HDAC1 and SMAD3 were cloned with a C-terminal Flag into pENTER (Vigenebio, China). We transfected them into cells using Lipofectamine 3000 (Invitrogen, USA) according to the manufacturer’s protocol. Lentiviral shRNA plasmids targeting TP53I11 were constructed from Genechem (Shanghai, China). Before transfection, adherent cells were seeded at a density of 1 × 10^5/well in 24-well plates 18-24 hours prior. The next day, 2 ml of fresh culture medium containing 6 μg/ml polybrene was added to replace the original culture medium, followed by the addition of an appropriate amount of virus suspension. After 24 hours, fresh culture medium was added to replace the medium containing the virus, and fluorescent expression could be seen 48 hours after transfection.

Single cell RNA-seq library construction and sequencing

Single-cell RNA-sequencing (ScRNA-seq) libraries were constructed using the SeekOne® Digital Droplet System (SeekOne®, Model: M001A) and SeekOne® Digital Droplet Single Cell 3′ Library Preparation Kit (SeekGene, Cat. No. K00202). Briefly, appropriate number of cells were mixed with reverse transcription reagent and then added to the sample well in SeekOne® chip S3. Subsequently Barcoded Hydrogel Beads (BHBs) and partitioning oil were dispensed into corresponding wells separately in chip S3. After emulsion droplet generation reverse transcription were performed at 42°C for 90 minutes and inactivated at 85°C for 5 minutes. Next, cDNA was purified from broken droplet and amplified in PCR reaction. The amplified cDNA product was then cleaned, fragmented, end repaired, A-tailed and ligated to sequencing adaptor. Finally, the indexed PCR was performed to amplify the DNA representing 3′ polyA part of expressing genes which also contained Cell Barcode and Unique Molecular Index. The indexed sequencing libraries were cleaned up with VAHTS DNA Clean Beads (Vazyme, N411-01), analyzed by Qubit (Thermo Fisher Scientific, Q33226) and Bio-Fragment Analyzer (Bioptic, Qsep400). The libraries were then sequenced on DNBSEQ-T7 platform with PE150 read length.

RNA sequencing and ChIP sequencing

For RNA sequencing, total RNA was isolated from N9 cells treated with either DMSO or RG2833 (Selleck, Shanghai, China) using TRIzol (Takara, Japan) according to the manufacturer’s protocol. Subsequently, cDNA libraries were prepared and sequenced on the Illumina HiSeq4000 platform. The Hisat2 tool was used to align sequencing data to the human reference genome (hg19).

The species was identified as Homo sapiens_Ensembl_GRCm38.104 in ChIP-seq. The sequencing method employed was Illumina PE150, generating 9G of sequencing data. Samples were divided into six groups (control_1, control_2, control_input, RG2833_1, RG2833_2, and RG2833_input). Initially, nuclear DNA was crosslinked with proteins, followed by cell lysis and DNA fragmentation via sonication. The DNA was then immunoprecipitated with anti-p-SMAD3 (#9520, Cell Signaling Technology, USA) to specifically recognize the target protein. Finally, the DNA was purified, a second-generation library was constructed, and sequencing was performed.

Differentially expressed genes were identified using the DESeq2 R package. Gene Ontology (GO) analyses were conducted using DAVID (https://david.ncifcrf.gov/).

Animals

All animal experiments were conducted in accordance with the ethical obligations and animal care guidelines approved by Qingdao University. Based on the previous description [21], patient-derived xenografts (PDX) were constructed from TBD0220L cells. The PDX tissues were implanted into the cranial region of nude mice using skull-guided screws under stereotactic guidance, and the overall survival time of the mice was monitored. Starting from the second day after transplantation, the treatment group received oral administration of RG2833 at a dose of 150 mg/kg every other day until death. The growth of intracranial tumors was monitored using bioluminescence imaging on days 7, 14, and 21. After death, the brains were carefully removed and fixed in 10% formaldehyde.

Stable transfection of sh-TP53I11 and control lentivirus in U251 cells was achieved following the manufacturer’s protocol. These transfected cells were then implanted into the cranial region of nude mice using skull-guided screws. Twenty days later, the mice were euthanized, and the size and weight of the intracranial tumors were measured before being placed in 10% formaldehyde.

Statistical analysis

The figures and figure legends provide detailed statistical parameters, including definitions and precise values of n, the statistical tests used, and the level of statistical significance. Data are presented as mean ± standard errors of the mean (SEM). The chi-squared test assessed significant differences between two groups. Kaplan–Meier survival plots were used to generate survival curves, with significance determined by the log-rank test. Two-sided Pearson’s correlation tests evaluated correlations between tissues. Statistical significance for functional analyses was determined using Student’s t-test or ANOVA. Statistical analyses were conducted using Prism 8.0 software, with p-values <0.05 considered statistically significant.

Results

The expression of HDAC1 is significantly elevated in the MES subtype GBM and shows a positive correlation with the expression of MES-specific genes

To assess the variation in HDAC1 expression across different GBM subtypes, TCGA GBM data were analyzed, revealing significantly elevated mRNA expression in the MES subtype and reduced expression in the PN subtype (Fig. 1A). The same trend was not observed among other HDAC family members (Fig. S1A). Analysis of TCGA and CGGA GBM patient data indicated that patients with higher HDAC1-mRNA expression in tumor tissue had significantly lower survival times compared to those with lower HDAC1-mRNA expression (Fig. 1B, C). Previously validated MES subtype GBM cell line TBD0220L/N9 and PN subtype GBM cell line TBD0207B/U251 [21] were used in this study. HDAC1-mRNA expression was notably higher in TBD0220L than in TBD0207B (Fig. 1D) and in N9 compared to U251 (Fig. 1E), with similar results for HDAC1-protein expression (Fig. 1F). Characteristic genes for each subtype were evaluated using Verhaak RG classification data (Table S5). Cerebrospinal fluid specimens from four patients with GBM (Table S1) were analyzed to evaluate the correlation between HDAC1-mRNA and subtype marker gene. HDAC1 expression positively correlated with the MES subtype marker gene MMP7 and negatively with the PN subtype marker gene HOXD3 (Fig. 1G). Immunofluorescence detection confirmed a positive correlation between HDAC1 and MMP7 expression (Fig. 1H). To inhibit HDAC1, siHDAC1 was designed to prevent protein synthesis, and an HDAC1/3 inhibitor (RG2833) was used to block its biological function (Fig. S2). Both methods decreased MES subtype marker gene expression and increased PN subtype marker gene expression (Fig. 1I-J). But no similar trend was found in other HDACs (Fig. S1B). TCGA data analysis further revealed a positive correlation between HDAC1 and MES subtype marker genes and a negative correlation with PN subtype marker genes (Fig. S3).

Fig. 1.

Characteristics of HDAC1 in molecular subtypes of GBM.

A mRNA expression of HDAC1 analyzed in CL, MES, and PN subtypes of GBM using TCGA data. Kaplan-Meier survival analysis of patients with GBM exhibiting high and low HDAC1 expression using the TCGA (B) and CGGA (C) dataset. D mRNA expression of HDAC1 in TBD0220L and TBD0207B cell lines. E mRNA expression of HDAC1 in N9 and U251 cell lines. F Protein expression levels analyzed by Western blotting in N9, U251, TBD0220L, and TBD0207B cells (n=3). G Correlation analysis was performed to examine the relationship between the expression of HDAC1-mRNA and the expression of MMP7-mRNA or HOXD3-mRNA in extracellular vesicles within the cerebrospinal fluid of GBM patients by q-PCR (n=3). H Expression of HDAC1-protein and MMP7-protein detected by immunofluorescence assay in patient GBM tissues (scale bar, 10 μm). Correlation analysis performed between HDAC1 and MMP7 of% positive cells. I-J Evaluation of mRNA changes in MES subtype marker genes and PN subtype marker genes after siHDAC1 and RG2833 treatment in N9 cells using q-PCR. All q-PCR data are shown as means ± SEM. Significance is indicated as *p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001.

In vitro and in vivo experiments demonstrated that RG2833 and siHDAC1 were effective in treating MES subtype GBM, inducing the transformation of the MES subtype to the PN subtype

Our experiments demonstrated that siHDAC1 inhibited the proliferation and invasion of the MES subtype cell line N9 in vitro (Fig. 2A, B). After intracranial transplantation of heterologous MES GBM tissue in nude mice, RG2833 was administered by gavage, demonstrating its inhibitory effect on MES GBM proliferation and its ability to extend the survival time of nude mice (Fig. 2 C-E). Immunohistochemical experiments were conducted using paraffin sections of intracranial tumor tissue, and mRNA from the tumor tissue was extracted for q-PCR analysis. The HDAC1 inhibitor reduced the expression of MES subtype marker genes and increased the expression of PN subtype marker genes in vivo (Fig. 2 F-G). In the N9 GBM cells, we performed scRNA-seq and applied UMAP nonlinear dimensionality reduction, resolving the population into nine clusters (Fig. 2 H). By transfecting siHDAC1, we observed dynamic changes in cell type composition. Monocle2 pseudotime analysis mapped the developmental trajectory, defined the differentiation direction from MES cells toward PN cells (Fig. 2 I-J).

Fig. 2.

RG2833 promotes intracranial PDX tumor tissue transformation toward a PN-like subtype.

A The N9 cell line was treated with siHDAC1 and analyzed the invasion using a transwell assay (scale bar, 50 μm). The number of cells passing through the gel and pores was counted (n=3). B The CCK-8 assay detected the proliferation of the N9 cell line after siHDAC1 treatment (n=3). C The PDX model was divided into control and RG2833 treatment groups, with bioluminescence images collected at 7, 14, and 21 days (n=5). D Quantitative analysis of bioluminescence. E Comparison of survival time between the two groups. F Immunohistochemical staining analyzed protein expression of MES marker genes (COL1A2, BCL3, KYNU, MMP7) and PN marker genes (HOXD3, PDGFRA, ERBB3, SOX10), with quantitative analysis performed using ImageJ (n=5). Scale bar, 10 μm. G Q-PCR analysis of mRNA expression of these genes (n=5). H Cell types of single‐cell sequencing data in N9 cells (AC (astrocyte)-like, Cycling, MES, OPC (oligodendrocyte-progenitor)-like and PN). I The direction of pseudotime differentiation, from dark blue to light blue. J The distribution of MES and PN cells by pseudotime analyze. All q-PCR data are shown as means ± SEM. **, *** and **** indicate p<0.01, p<0.001, and p<0.0001, respectively.

The combination of RG2833 and bevacizumab inhibited the proliferation and invasion of MES GBM cells

Analysis of survival data from 52 patients with GBM using the TCGA database revealed that MES subtype patients had a poorer prognosis and shorter median survival compared to those with PN and CL subtypes (Fig. S4A-B). This finding prompted an exploration of therapeutic strategies aimed at transforming the poorly prognostic MES subtype into other subtypes. The PN subtype, for example, shows an effective therapeutic response to bevacizumab [22]. Our experiments demonstrated that bevacizumab inhibited the proliferation and invasion of the PN subtype cell line U251 in vitro (Fig. 3A–C). Although bevacizumab had no significant effect on MES cell lines, knockdown by siHDAC1/RG2833 significantly inhibited their proliferation and invasion. Additionally, combining RG2833 with bevacizumab had a more pronounced inhibitory effect on the proliferation and invasion of the N9 cell line (Fig. 3D–F). Western blotting analysis was conducted to investigate the protein expression of MES marker genes following bevacizumab treatment in N9 cells, revealing no significant changes. However, the combination of bevacizumab and RG2833 significantly reduced the protein expression of MES marker genes. These changes were mirrored at the mRNA level, showing a consistent trend with the protein expression (Fig. 3G-H). To investigate the molecular mechanism by which RG2833 induces subtype conversion, the N9 cell line was divided into a control group and an RG2833 treatment group, each with two replicates. Differential gene analysis via clustering identified the 40 genes with the most significant differences, all with p-values less than 0.0001 (Fig. 3I). The intersection of identified genes (p<0.05) between replicates revealed gene overlap rates of 83% and 87% in the control and RG2833 treatment groups, respectively (Fig. 3J). Volcano plot analysis of overlap genes indicated that RG2833 treatment significantly upregulated 833 genes, downregulated 575 genes, and left 15,340 genes unchanged (p-adjust value of 0.05 and log fold change (log FC) of 1) (Fig. 3K). GO analysis linked the differentially expressed genes to cell growth, death, and cancer-related pathways (Fig. 3L).

Fig. 3.

RG2833 combined with bevacizumab significantly inhibits proliferation and invasion of MES cell lines.

A The U251 cell line was treated with bevacizumab and analyzed the invasion using a transwell assay (scale bar, 50 μm). B The number of cells passing through the gel and pores was counted (n=3). C The CCK-8 assay detected the proliferation of the U251 cell line after bevacizumab treatment (n=3). D Invasion ability was examined using a transwell assay for control, bevacizumab, RG2833, and bevacizumab+RG2833 treatments in the N9 cell line (n=3, scale bar, 50 μm). E The number of cells passing through the gel and pores was counted (n=3). Control vs. RG2833 and RG2833 vs. bevacizumab+RG2833 (t-test). F The CCK-8 assay detected the proliferation of the N9 cell line with the same treatments (n=3). Control vs. RG2833 and RG2833 vs. bevacizumab+RG2833 (two-way ANOVA). G-H Western blotting analysis and q-PCR were used to detect the expression of MES marker proteins and mRNA in the N9 cell line after bevacizumab and bevacizumab+RG2833 treatments (n=3). I RNA-seq data analysis of the N9 cell line, with a heatmap displaying differential gene expression in control-1, control-2, RG2833-1, and RG2833-2. J Venn diagram illustrating the intersection of gene sets from the control repeated group and RG2833 repeated group. K Volcano plot analysis of differential genes identified from the intersected gene set. L GO analysis of the differentially expressed genes. All q-PCR data are shown as means ± SEM. *, *** and **** indicate p<0.05, p<0.001, and p<0.0001, respectively.

HDAC1 interacts with p-SMAD3, and RG2833 induces alterations in the subcellular localization of HDAC1

Immunofluorescence experiments using specific antibodies revealed significant co-localization of HDAC1 and p-SMAD3 in TBD0220L cells, indicating a strong correlation between these proteins (Fig. 4A). In the N9 cell line, Co-IP assays demonstrated a reciprocal recruitment interaction between HDAC1 and p-SMAD3 (Fig. 4B). Additionally, we used antibodies against H2AK5ac, H2BK5ac, H3K9ac, and H3K27ac to assess changes in histone acetylation following RG2833 treatment. The results showed increased acetylation levels within the nucleus (Fig. 4C). Furthermore, RG2833 treatment altered HDAC1-protein distribution, decreasing its presence in the nucleus while increasing its presence in the cytoplasm (Fig. 4D). Intracellular immunofluorescence staining of TBD0220L cells showed consistent alterations, confirming the observed trends (Fig. 4E).

Fig. 4.

HDAC1 and p-SMAD3 form a transcription complex.

A Immunofluorescence co-localization detection of HDAC1 and p-SMAD3 in TBD0220L cells (n=5). Scale bar, 5 μm. The position of stained points from the Olympus confocal microscope was used for correlation analysis. B N9 cells transfected with flag-SMAD3 or flag-HDAC1, with Co-IP experiments detecting the binding of p-SMAD3 and HDAC1 (n=3). C Expression of acetylation histones (H2AK5ac, H2BK5ac, H3K9ac, and H3K27ac) in N9 cells after RG2833 treatment (n=3). D Western blotting analysis detecting HDAC1 expression in the nucleus and cytoplasm of N9 cells post-RG2833 treatment (n=3). E Immunofluorescence detection of intracellular localization changes of p-SMAD3 and HDAC1 after RG2833 treatment in TBD0220L cells (n=5). Scale bar, 5 μm. The position of stained points from the Olympus confocal microscope was used for correlation analysis. * and ** indicate p<0.05 and p<0.01.

By inhibiting HDAC1 to block histone acetylation leads to the recruitment of p-SMAD3 to the genome, thereby influencing transcription in MES GBM cells

The N9 cell line was used for ChIP-seq experiments, enriching proteins with the p-SMAD3 ChIP-grade antibody, precipitating genomic DNA, and conducting high-throughput sequencing analysis [23]. The p-SMAD3 binding sequence was identified (https://jaspar.elixir.no/) (Fig. S5). Cell samples were divided into a DMSO control group and an RG2833 treatment group, each with two replicates. Post-treatment comparisons revealed a significant increase in p-SMAD3 enrichment around the transcription start site (TSS) following RG2833 treatment (Fig. 5A). Analysis of p-SMAD3 distribution in different coding DNA regions showed an increased distribution ratio of p-SMAD3 in the promoter region, particularly the regions within 1kb base pairs of the actual TSS after RG2833 treatment (Fig. 5B). Comparing sequencing coding gene data from the control and treatment groups, we found 89% gene overlap in the control repeated group and 93% in the treatment repeated group (Fig. 5C). Combining differential genes between the control and treatment groups with genes showing significant binding changes, we performed clustering analysis to identify genes with p-adjust values < 0.0001. The results indicated that 18 genes showed a significant decrease in binding, while 31 genes showed a significant increase in binding (Fig. 5D). In summarizing the RNA-seq results before and after RG2833 treatment (Fig. 3K), we took the intersection to identify genes with increased mRNA expression (p-adjust value<0.05) and increased p-SMAD3 binding with ChIP-seq analysis (p-adjust value<0.05), and obtained 107 potential downstream target genes (Fig. 5E). Targeted gene GO analysis revealed associations with axon guidance, ECM-receptor interaction, and cell adhesion molecules (Fig. S6).

Fig. 5.

ChIP-seq analysis of HDAC1/p-SMAD3 complex binding sites.

A Enrichment analysis of p-SMAD3 near the transcription start site (TSS) in control_1, control_2, control_input, RG2833_1, RG2833_2, and RG2833_input groups. B Distribution ratio analysis of p-SMAD3 in various genomic functional regions within the control_1, control_2, control_input, RG2833_1, RG2833_2, and RG2833_input groups. C Intersection of peaks from control_1 and control_2 to obtain reliable sites, as well as intersection of peaks from RG2833_1 and RG2833_2 to obtain reliable sites. D Heatmap detection of differentially accessible regions (Padj ≤ 0.001). E Identification of 107 binding sites for the HDAC1/p-SMAD3 complex through the intersection of ChIP-seq differentially binding regions and RNA-seq differentially expressed genes (Padj ≤ 0.05).

TP53I11 maintains the characteristics of PN subtype GBM and inhibits the proliferation and invasion of MES GBM both in vivo and in vitro

TP53I11 emerged as a significant downstream regulated gene cluster from our previous analyses (Figs. 3I and 5D). ChIP-qPCR experiments revealed that the p-SMAD3 antibody effectively enriched TP53I11, with enrichment levels increasing following RG2833 and siHDAC1 treatment (Fig. 6A-B, Table S6). Analysis of TCGA and CGGA GBM patient data indicated that patients with higher TP53I11-mRNA expression in tumor tissue had significantly longer survival times compared to those with lower TP53I11-mRNA expression (Fig. 6 C-D). Knockdown of HDAC1 or SMAD3 in the N9 cell line led to a significant reduction in TP53I11 expression at both the mRNA and protein levels (Fig. 6 E-G).

Fig. 6.

TP53I11 plays a crucial role in the MES-to-PN transformation process.

A ChIP-PCR analysis showing changes in p-SMAD3 binding to the promoters of TP53I11 after RG2833 treatment in N9 cells (n=3). B ChIP-PCR analysis showing changes in p-SMAD3 binding to the promoters of TP53I11 after siHDAC1 treatment in N9 cells (n=3). Kaplan-Meier survival analysis of patients with GBM exhibiting high and low TP53I11 expression using the TCGA (C) and CGGA (D) dataset. The TP53I11-mRNA expression with siHDAC1 (E) and siSMAD3 (F) treated in N9 cells (n=3). G The TP53I11-protein expression with siHDAC1 and siSMAD3 treated in N9 cells (n=3). H mRNA expression analysis of TP53I11 in CL, MES, and PN subtypes of GBM using TCGA data. I Enrichment score level in Verhaak PN subtype package (high TP53I11 vs low TP53I11) analyzed by GSEA. J ROC curve evaluating the sensitivity of TP53I11 as a molecular marker of PN GBM in the TCGA dataset. K CCK-8 assay detecting proliferation of U251 cells after siTP53I11 treatment (n=3). L Transwell assay examining invasion ability of U251 cells with control and siTP53I11 treatment (n=3). Scale bar, 50 μm. M Number of cells passing through the gel and pores counted (n=3, t-test). N Images of intracranial tumors in nude mice injected with U251 cells transfected with shNC or shTP53I11 (n=4). O HE staining of representative brain tissue from each group. P-Q Tumor volume and weight measured on the 14th day after U251 cell injection. R-S Immunohistochemical staining analyzing Ki-67-protein expression with quantitative analysis using ImageJ (n=5). Scale bar, 10 μm. All q-PCR data are shown as means ± SEM. *, **, *** and **** indicate p<0.05, p<0.01, p<0.001, and p<0.0001, respectively.

By utilizing the subtype-specific TCGA GBM data, we observed a significant elevation of TP53I11 expression in the PN subtype compared to the MES subtype at the mRNA level (Fig. 6H). Gene set enrichment analysis (GSEA) based on the Verhaak GBM PN gene set showed that patients with GBM exhibiting high TP53I11 expression had higher enrichment scores, while those with low TP53I11 expression had lower enrichment scores (Fig. 6I). Receiver operating characteristic (ROC) curve analysis of TCGA GBM data indicated that TP53I11 has high sensitivity in distinguishing PN patients from other subtypes, with an area under the curve (AUC) of 84% (Fig. 6J). TP53I11-specific interfering RNA (siRNA) was constructed and applied to the PN subtype GBM cell line U251, resulting in increased cell proliferation and invasion as shown by CCK-8 and Transwell assays (Fig. 6K-M). Additionally, a shTP53I11 lentiviral vector was developed, transfected into U251 cells, and then injected into the brains of mice to form tumors. Compared to the control group, the shTP53I11-treated group displayed larger tumor volumes and weights (Fig. 6N-Q). Enhanced Ki-67 positivity, detected by immunohistochemistry, indicated increased proliferative capacity in shTP53I11-treated tumors (Fig. 6R-S).

Discussion

GBM, comprising both newly diagnosed and recurrent cases, accounts for 60% of all gliomas [24]. Among these GBM patients, 98.03% are IDH wild type, while only 1.97% are IDH mutant type. In recent years, there have been significant advancements in the treatment of IDH mutation gliomas (including WHO grades 1-4), with options beyond surgery, radiation, chemotherapy, TTfields, targeted and immunotherapy becoming increasingly available [25,26]. However, effective breakthroughs in the treatment of IDH wild-type gliomas, particularly IDH wild-type GBM, remain elusive [27]. Based on transcriptome differences, GBM with IDH wild type was initially classified into four types, with the neural type later excluded, leaving PN, CL, and MES types [3,28]. These subtypes exhibit distinct genetic and phenotypic characteristics, with specific mRNA clusters influencing tumor development and prognosis [29,30]. By performing RNA-seq analysis on a particular tumor, we can determine its overall tendency towards a particular type. At a more microscopic level, we can obtain molecular subtypes for different parts of the same tumor, even different cells [31]. Our previous studies, the analysis of sequencing data from different parts of a single GBM patient's tumor revealed that in a single lesion from one patient, two of five tissue directions were classified as MES subtype, two as PN subtype, and one as CL subtype (Fig. S7A). Single cell sequencing data from four patients with glioma revealed significant variation in the proportion of cells in each subtype, with MES cells accounting for 80% in the highest case and 32% in the lowest (Fig. S7B). This observation highlights substantial mRNA heterogeneity present in glioma. This provides us with a basis for implementing treatment at the mRNA level to control the proliferation and invasion of GBM.

Analysis of TCGA GBM data and GBM tissues and cells revealed elevated HDAC1 expression in the MES subtype. This was accompanied by an increase in MES-related markers and a decrease in PN-related markers following HDAC1 downregulation. This led us to hypothesize that HDAC1 plays a pivotal role in distinguishing MES and PN types. Prior research has highlighted the effectiveness of HDAC inhibitors in treating hematological tumors [[32], [33], [34]]. Moreover, combining HDAC inhibitors with PD-1 monoclonal antibodies has improved overall survival in patients with solid tumors [[35], [36], [37]]. Building on these preliminary findings, comparative experiments using a nude mouse PDX model of mesenchymal tumor cells and selected RG2833—a potent HDAC1/3 inhibitor capable of crossing the blood-brain barrier [38]—were conducted. Results demonstrated significant inhibition of MES intracranial tumor growth by RG2833, along with reductions in MES-related marker expression.

Bevacizumab, a monoclonal antibody targeting VEGF, is commonly used in combination with temozolomide or lomustine/carmustine for treating adults with progressive or recurrent GBM [1,39]. Our findings indicate that bevacizumab exhibits a stronger inhibitory effect on the proliferative invasion of the PN subtype GBM compared to the MES subtype, likely due to differential VEGF receptor expression between these subtypes [22]. The therapeutic effects of combining bevacizumab with RG2833 on MES GBM were investigated and achieved a good therapeutic response. Given heterogeneity of GBM, the strategy of drug combination provided a significant theoretical basis for targeted GBM treatment.

Subsequently, we conducted a prediction of potential protein interactions with HDAC1, focusing particularly on transcriptional regulatory factors. Experimental results revealed a significant interaction between HDAC1 and p-SMAD3, both highly enriched in the nuclei of GBM cells. SMAD3, a component of the transforming growth factor-beta (TGF-β) signaling pathway, transmits signals from the cell surface to the nucleus, regulating gene activity and cell proliferation by forming a complex with SMAD4 and binding DNA [[40], [41], [42]]. Our study identified the co-localization and binding of HDAC1 and p-SMAD3 within the nuclei of MES subtype GBM cell lines. RG2833 reduced the binding of HDAC1 and p-SMAD3 and the nuclear distribution of HDAC1. Moreover, targeting HDAC1 partially disrupted the function of the TGF-β pathway without inducing a global breakdown (Fig. S8).

Therefore, we investigated the impact on genome transcription by conducting RNA-seq and ChIP-seq [43,44]. Both experiments were conducted using the MES cell line N9. In the RNA-seq experiment, the application of RG2833 resulted in a marked increase in the number of upregulated mRNA genes (833 vs. 575) compared to downregulated genes, indicating that inhibition of HDAC1 enhances histone acetylation, allowing chromatin increased possibility to interact with transcription factors and initiate transcriptional activation. The distribution of ChIP-seq signals near the TSS is of particular interest due to the frequent binding of transcription factors or histone modifications that serve as regulatory "markers" in the promoter region, influencing gene expression [45]. Our findings revealed a significant increase in p-SMAD3 binding at the TSS region after RG2833 treatment, particularly within the 1kb promoter region. These results suggest that RG2833 induces a relaxation of the chromatin-histone interaction, thereby facilitating the binding of p-SMAD3 to chromatin and enabling its transcriptional regulatory function. RNA-seq detected changes in mRNA levels of coding genes, while ChIP-seq annotated peaks with genes, defining the associated gene of a peak as the gene represented by the TSS closest to the peak, and detecting the changes in the enrichment fold of associated gene peaks. Differential gene analysis between these methods identified intersecting gene sets, leading to the identification of TP53I11, a key gene potentially involved in the MES to PN subtype transition. TP53I11, also known as p53-Induced Gene 11 Protein (PIG11), located on chromosome 11p11.2 [46]. In 1997, Polyak et al. initially identified a significant induction of multiple genes by high p53 expression using Serial Analysis of Gene Expression technology during their investigation of the mechanism of p53-dependent apoptosis in human colon cancer cells, while TP53I11 was involved in apoptosis induced by p53 [[47], [48], [49]]. Moreover, TP53I11 also suppresses epithelial-mesenchymal transition and metastasis in breast cancer cells [50]. Our further validation through data analysis and in vivo/in vitro experiments confirmed that TP53I11 is overexpressed in PN subtype GBM, and knocking down TP53I11 in PN subtype cells and tumor tissues enhances GBM proliferation and invasion.

Conclusions

The investigation focused on the molecular heterogeneity of GBM multiforme, revealing HDAC1′s involvement in the subtype conversion and intrinsic epigenetic mechanisms of MES and PN subtypes. The study examined the impact of the HDAC1/p-SMAD3-TP53I11 axis on the proliferation and invasion of MES subtype GBM. This innovative approach, which targets histone acetylation modifications, offers a promising new avenue for the clinical treatment of MES subtype GBM.

Funding statement

Kai Zhao was supported by National Natural Science Foundation of China (No. 82002996). Chao Wang was supported by Youth Research Fund of the Affiliated Hospital of Qingdao University (Grant. No. QDFYQN2023114).

Declaration of generative AI use

None.

Consent for publication

All the authors agree to publish this paper.

Ethics statement

The acquisition of human tissue and its subsequent use were performed with the ethical obligations and animal care guidelines approved by Qingdao University (No. 20230114SD20231105028), and informed consent of the patient/family was obtained.

CRediT authorship contribution statement

Chao Wang: Writing – original draft, Methodology. Rui Shang: Writing – review & editing, Methodology, Data curation. Hua Liang: Writing – review & editing, Formal analysis. Meng Zhu: Writing – review & editing, Formal analysis. Wujun Chen: Validation, Software. Kai Zhao: Validation, Supervision, Project administration, Funding acquisition.

Declaration of competing interest

The authors declare no competing interests.