HDAC6 involves in regulating the lncRNA-microRNA-mRNA network to promote the proliferation of glioblastoma cells

Background Glioblastoma (GBM) is the most aggressive and lethal brain tumor. Although the histone deacetylase (HDAC)/transcription factor axis promotes growth in GBM, whether HDACs including HDAC6 are involved in modulating long non-coding RNAs (lncRNAs) to affect GBM malignancy remains obscure. Methods Integrative analysis of microarray and RNA-seq was performed to identify lncRNAs governed by HDAC6. Half-life measurement and RNA-protein pull-down assay combined with isobaric tags for relative and absolute quantitation (iTRAQ)-based proteomic analysis were conducted to identify RNA modulators. The effect of LINC00461 on GBM malignancy was evaluated using animal models and cell proliferation-related assays. Functional analysis of the LINC00461 downstream networks was performed comprehensively using ingenuity pathway analysis and public databases. Results We identified a lncRNA, LINC00461, which was substantially increased in stem-like/treatment-resistant GBM cells. LINC00461 was inversely correlated with the survival of mice-bearing GBM and it was stabilized by the interaction between HDAC6 and RNA-binding proteins (RBPs) such as carbon catabolite repression—negative on TATA-less (CCR4-NOT) core exoribonuclease subunit 6 and fused in sarcoma. Targeting LINC00461 using azaindolylsulfonamide, an HDAC6 inhibitor, decreased cell-division-related proteins via the lncRNA-microRNA (miRNA)-mRNA networks and caused cell-cycle arrest, thereby suppressing proliferation in parental and drug-resistant GBM cells and prolonging the survival of mice-bearing GBM. Conclusions This study sheds light on the role of LINC00461 in GBM malignancy and provides a novel therapeutic strategy for targeting the HDAC6/RBP/LINC00461 axis and its downstream effectors in patients with GBM. Supplementary Information The online version contains supplementary material available at 10.1186/s13046-022-02257-w.


Background
Glioblastoma (GBM) has a poor prognosis and remains incurable, despite aggressive treatment options including surgery, radiation therapy, and first-line chemotherapy with the drug temozolomide (TMZ). Less than 5% of the patients with GBM survive for 5 years following diagnosis. Approximately 90% of patients suffer from disease relapse within 2 years of treatment, regardless of their initial response to prior therapy [1]. Previous studies have revealed that GBM frequently harbors different gene expression patterns, causing patients to exhibit diverse clinical characteristics, survival times, and responses to treatment [2,3]. Therefore, understanding the molecular mechanisms of GBM and developing novel strategies for treating the disease are urgent.
Although GBM is characterized by chromosome 10q loss, aberrant expression of proteins, such as epidermal growth factor receptor amplification and p16 INK4a deletion, and gene mutations, such as TP53, IDH1, and PTEN, the molecular basis of the onset and progression of GBM malignancy is not fully understood [4,5]. Because protein-coding genes account for < 2% of the whole genome, examining the impact of non-coding genes on the regulation of the glioma phenotype appears essential [6]. Long non-coding RNAs (lncRNAs) are defined as transcripts exceeding 200 base pairs in length; these molecules have a wide variety of biological functions [7]. Accumulating evidence has revealed that many dysregulated lncRNAs contribute to chemoresistance and poor prognostic features of GBM [8]. As a result, identification of lncRNA as a potential target for treating patients with GBM is urgently needed.
The structure of lncRNA is comparable to that of mRNA, with a 7mGpppG 5′ cap and 3′-polyadenosine tail. Similar to mRNA decay, lncRNA stabilization and degradation are regulated by microRNAs, decay-promoting RNA-binding proteins (RBPs), decapping enzymes, and deadenylases [9]. Our previous study has demonstrated that azaindolylsulfonamide (MPT0B291), a histone deacetylase (HDAC) inhibitor selective for HDAC6 (IC 50 ≈ 0.0052 μM) [10,11], induces cellular senescence in stem-like GBM cells and prolongs the survival of mice-bearing temozolomide (TMZ)-resistant xenografts through downregulation of Sp1 and its targeted genes associated with drug resistance [10]. Although HDAC6 is known to maintain the acetylation balance of histones and nonhistone substrates, such as α-tubulin, cortactin, HSP90, and a few transcriptional factors [12], whether it participates in the regulation of lncRNAs that affect tumorigenesis remains obscure.
In the current study, we used integrative microarray and RNA-seq analysis to investigate whether HDAC6 modulates lncRNAs. A highly conserved lncRNA, LINC00461, which functions as an essential regulator in glioma formation [13] and regulates the expressions of genes such as DNA topoisomerase II Alpha (TOP2A) [13], B-cell lymphoma 2 (BCL2) [14], and integrin β3 (ITGB3) [15], was identified to be remarkably downregulated by HDAC6 depletion. Our data highlight LINC00461-associated regulatory networks in GBM malignancies, including treatment resistance and cancer stemness, and provide a novel insight into targeting the HDAC6/RBP/LINC00461 axis as a therapeutic approach for patients with GBM.

Cell culture
Human GBM cell lines, including U87MG (ATCC HTB-14; American Type Culture Collection, Manassas, VA, USA) and A172 (ATCC CRL-1620), and patient-derived primary GBM (Pt#3 and Pt#5) along with their TMZresistant and stem-like cells, were cultured in respective media as described previously [16]. Procedures utilized for establishing TMZ-resistant GBM cells were mentioned previously [16]. To maintain TMZ-resistant cells, 50 μM TMZ was added into the culture medium, and their resistance characteristics were confirmed using colony-formation assay. Informed consent obtained from the patients followed the protocols (Nos. 201,006,011 and 201,402,018) approved by the Joint Institutional Review Board (JIRB) of Taipei Medical University (Taipei, Taiwan).

RNA interference
Parental and TMZ-resistant GBM cells (U87MG, A172, and Pt#3) were transfected with 5, 10, or 20 nM specific gene siRNA or control non-targeting siRNA (ON-TARGETplus siRNA-SMARTpool/non-targeting pool, Dharmacon, Inc., Lafayette, CO, USA) using lipofectamine 3000 (Thermo Fisher Scientific). Seventy-two hours after the transfection of siRNA, the total RNA was extracted from GBM cells using TRIzol reagent (Thermo Fisher Scientific). The effects of depletion were validated by RT-qPCR. Targeted sequences for each gene are listed in Table S1.

Microarray analysis
Stem-like and TMZ-resistant U87MG cells were treated with 10 and 6 μM MPT0B291, respectively, for 1 day. U87MG cells were treated with si-Sp1 for 3 days. Gene expression analysis following total RNA extracted from these cells using TRIzol reagent (Thermo Fisher Scientific) was performed by Welgene Biotech (Taipei, Taiwan) using a SurePrint G3 Human Gene Expression 8x60K Microarray (Agilent Technologies, Inc., Santa Clara, CA, USA). The number of non-coding genes with the signalto-noise-ratio (SNR) threshold of > 1.0 was identified.

RNA-seq
RNA-seq was performed by Welgene Biotech. In brief, after HDAC6 or LINC00461 depletion, respectively, in TMZ-resistant U87MG and parental U87MG cells, the total RNA was extracted using Trizol Reagent (Thermo Fisher Scientific). Purified RNA was quantified at OD260 nm using an ND-1000 spectrophotometer (NanoDrop Technologies, Inc. Wilmington, DE, USA) and qualified using a Bioanalyzer 2100 (Agilent Technologies, Inc.) with RNA 6000 LabChip kit (Agilent Technologies, Inc.). The RNA-Seq libraries were constructed using Agilent's SureSelect Strand-Specific RNA Library Preparation Kit, followed by AMPure XP Beads (Beckman Coulter, Brea, CA, USA) size selection, and the sequence was directly determined using Illumina's sequencing-by-synthesis technology.

Quantitative reverse transcription PCR (RT-qPCR)
Total RNA was extracted from cells using TRIzol reagent (Thermo Fisher Scientific), followed by reverse transcription using SuperScript IV First-Strand Synthesis System (Thermo Fisher Scientific) and TaqMan ™ MicroRNA Reverse Transcription Kit (Thermo Fisher Scientific), respectively, for lncRNA or mRNA and miRNA. qPCR analysis was performed using iTaq Universal SYBR Green Supermix (Bio-Rad Laboratories, Inc., Hercules, CA, USA) and TaqMan ™ Small RNA Assays, respectively, for lncRNA or mRNA and miRNA on StepOnePlus Real-Time PCR System (Applied Biosystems Inc., Waltham, MA, USA). Relative expression levels of target genes were determined using the delta-delta Ct (2 -∆∆Ct ) method and normalized against the glyceraldehyde 3-phosphate dehydrogenase (GAPDH) mRNA and U6 snRNA levels, respectively, for lncRNA or mRNA and miRNA. Oligonucleotides used for TaqMan qPCR assays were purchased from Suu-Flower Co., Ltd. (Taichung City, Taiwan). The primer or probe sequences for each gene are listed in Table S2.

Animal models
For the subcutaneous tumor model, 1 × 10 6 GBM cells (U87MG and Pt#3) were implanted into both dorsal flanks of female nonobese diabetic/severe combined immunodeficiency (NOD/SCID) mice (8 weeks old, BioLASCO Co., Ltd., Taipei, Taiwan). Intraperitoneal administration with MPT0B291 (10 mg/kg) or vehicle (DMSO) every other day was initiated when the tumor volume reached a size of 10-20 mm 3 after tumor implantation. Tumor volumes were measured using a caliper three times a week and calculated following the modified ellipsoidal formula: tumor volume = 1/2 (length × width 2 ) [17]. Tumor weights were recorded at the end of the experiments. For the orthotopic tumor model, 5 × 10 5 Pt#3 cells (with or without stable LINC00461 expression) were injected into the right frontal brain area of male Bagg albino (BALB)/c nude mice (9 weeks old, BioLASCO Co., Ltd.) according to detailed procedures reported previously [18]. Intraperitoneal administration with MPT0B291 (10 mg/kg) or vehicle (DMSO) every other day was initiated 5 days after tumor inoculation. Mice weights were recorded three times a week. All protocols for animal experiments were approved by the Institutional Animal Care and Use Committee of the National Health Research Institute (IACUC, NHRI, Tainan, Taiwan). Animal experiments were conducted under the registration number (NHRI-IACUC-106010).

RNA-protein pull-down assay
Circular plasmid, pcDNA3.1-LINC00461, was converted to a linear template using EcoRV-HF digestion (New England Biolabs Inc., Ipswich, MA, USA) for in vitro LINC00461 synthesis (HiScribe ™ T7 Quick High Yield RNA Synthesis Kit; New England Biolabs). The pulldown of RNA-protein complexes was conducted using Pierce ™ Magnetic RNA-Protein Pull-Down Kit (Thermo Fisher Scientific) based on the manufacturer's instructions. Briefly, 50 pmol desthiobiotin-labeled (Pierce ™ RNA 3′ End Biotinylation Kit, Thermo Fisher Scientific) androgen receptor (AR) 3′-untranslated regions (UTR) and LINC00461 were captured by 50 μL streptavidin magnetic beads and then interacted with 200 μg lysate extracted from parental and TMZ-resistant Pt#3 cells. Elution of RNA-binding protein complexes was applied for western blotting and isobaric tags for relative and absolute quantitation (iTRAQ)-based proteomic analysis (Biotools Co. Ltd., New Taipei City, Taiwan). 3′-UTR of AR RNA, which contains UC-rich regions for human antigen R binding, is the control system for the pulldown assay.

Cell-cycle analysis
The cell cycle was evaluated by flow cytometry. Briefly, parental and TMZ-resistant U87MG cells at a density of 7.5 × 10 4 cells per well were seeded in 6-cm petri dishes. After downregulation of LINC00461 using siRNA, cells were collected and fixed with 70% ethanol at 4 °C overnight. and HRP-linked anti-rabbit IgG (sc-2005, 1:5000, Santa Cruz Biotechnology, Inc.) were used as secondary antibodies. The signals were amplified using an enhanced chemiluminescence reagent (ECL, GE Healthcare, Chicago, IL, USA) and captured using the ChemiDoc Touch Imaging System (Bio-Rad Laboratories, Inc.). The intensity of each band was determined by Image Lab, and the relative target protein levels were normalized against the tubulin level.

Single-cell RNA sequencing (scRNA-seq)
Fresh tumor tissues from four patients with GBM were prepared for scRNA-seq following the protocol established by 10x Genomics ® Single Cell (Pleasanton, CA, USA). Data analysis was performed using Cell Ranger (10x Genomics ® ) and Seurat (29608179). Informed consent obtained from the patients followed a protocol (IRB No. EC1080202) approved by the Research Ethics Committee of the National Health Research Institute (NHRI, Miaoli, Taiwan).

Bioinformatic analyses
The heatmap, which displayed the log2 (fold change) values of selected lncRNAs from microarray datasets, was generated using online software Morpheus (https:// softw are. broad insti tute. org/ morph eus). Functional analysis of data from microarray and RNA sequencing was performed using ingenuity pathway analysis (IPA, Qiagen). The overlap between molecules in our datasets and a particular function was calculated using the right-tailed Fisher's exact test. Prediction of activation or inhibition of pathways was examined using a Z-score. The heat map was generated using a MultiExperiment Viewer by the log2 (fold change) values of selected genes.

Statistics
The statistical significance of the difference between the two groups was analyzed using the unpaired Student's t-test with a two-tailed p-value. For more than two groups, one-way or two-way analysis of variance (ANOVA) was used to examine statistically significant differences depending on the number of variables. The survival curves of different groups were compared using the log-rank test. A p-value of < 0.05 was considered statistically significant. The statistical significance increased in the order: * p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001.

HDAC6 inhibition significantly decreases LINC00461 expression
To determine whether HDAC6 affects lncRNA regulation, we analyzed the expression levels of 9445 lncRNAs using microarray. Microarray gene expression data were applied independently on mRNA samples isolated from GBM stem-like cells (serum-free/ suspension-adapted tumorspheres) [21] and TMZresistant cells [18]. A comparison of MPT0B291treated with DMSO-treated conditions showed that 54 lncRNA probe sets represent reliable microarray signals (signal-to-noise ratio ≥ 1, Fig. 1A). Among these, only two lncRNAs (LINC00461 and LINC01559) were commonly and significantly altered in stem-like (S-U87MG) and TMZ-resistant (R-U87MG) GBM cells following MPT0B291 treatment (Fig. 1B, Fig. S1). LINC00461 levels were decreased 0.27-fold and 0.49fold in S-and R-U87MG cells, respectively, while LINC01559 levels were upregulated 2.70-fold and 3.64fold in S-and R-U87MG cells, respectively. To further confirm the effect of HDAC6 on lncRNA regulation, we depleted HDAC6 using siRNA and performed RNA-seq analysis. Four lncRNAs showing altered levels after HDAC6 knockdown were identified (Fig. 1C).
Interestingly, a comparison of datasets obtained from MPT0B291 treatment and HDAC6 knockdown revealed that LINC00461 was the only lncRNA with an obvious reduction after HDAC6 inhibition, suggesting that it is a downstream lncRNA target of HDAC6. Moreover, further confirmation using RT-qPCR analysis showed a significant reduction in LINC00461 levels by MPT0B291 treatment in both parental and TMZresistant U87MG cells (Fig. 1D).

Attenuation of LINC00461 by MPT0B291 remarkably extends the survival time of mice with GBM
Using U87MG xenografts and Pt#3 xenografts in mice, we observed a significant reduction in tumor growth after MPT0B291 treatment ( Fig. 2A-C). Moreover, we analyzed the expression of LINC00461 in MPT0B291treated and control xenografts by RNA ISH and RNA FISH. A decrease of approximately 40% in LINC00461 molecules was visualized within the cytoplasm of MPT0B291-treated xenograft tissues compared to that in control tumors (Fig. 2D-E), confirming the suppressive effect of MPT0B291 on LINC00461 expression in vivo. Furthermore, MPT0B291 treatment extended the survival of mice-bearing intracranial GBM (Fig. 2F). To understand the mechanism by which MPT0B291 treatment prolonged survival through downregulating LINC00461, protein expressions of proliferation (Ki-67) and apoptosis (cleaved caspase-3) marker were measured in xenograft tissues. A significant reduction in Ki-67 expression was observed within the nucleus of MPT0B291-treated xenograft tissues compared to that in control tumors (Fig. S2A). In contrast, the difference in cleaved caspase-3 expression was not significant in the nucleus between MPT0B291-treated and control xenograft tissues (Fig. S2A), suggesting that the extension of survival in mice-bearing GBM is mainly derived from the anti-proliferative effect of MPT0B291. We then evaluated the role of LINC00461 on survival outcomes using the same orthotopic mouse model. Pt#3 cells that stably express LINC00461 were constructed (Fig. 2G) and injected into the right cerebral hemisphere of mice. The LINC00461 high expression group's median survival was 24 d, while that of the LINC00461 low expression group was 54 d (Fig. 2H). Thus, LINC00461 plays an important role in GBM growth, but HDAC6 inhibition induced by MPT0B291 blocks the oncogenic effect of LINC00461.

HDAC6 and RNA-binding proteins are involved in managing LINC00461 expression
MPT0B291 is known to block the activation of the HDAC6/Sp1 axis [10]. Whether MPT0B291 suppresses LINC00461 expression by a Sp1-dependent mechanism was further investigated. The result showed that LINC00461 expression was substantially decreased by MPT0B291 treatment while it increased with Sp1 depletion (Fig. 3A, bars 1-3). Consistently, RT-qPCR detection after Sp1 knockdown confirmed elevated LINC00461 levels in parental and TMZ-resistant U87MG cells (Fig. 3B), suggesting that MPT0B291-mediated downregulation of LINC00461 is modulated via an Sp1-independent pathway. Furthermore, to explore the mechanism underlying the modulation of LINC00461 expression, we investigated the RNA stability of LINC00461 using dactinomycin to block de novo RNA transcription in cells and found that MPT0B291 treatment caused 2.75-fold and 1.75-fold reduction in the half-life of LINC00461 in TMZ-resistant Pt#3 and Pt#5 cells, respectively (Fig. 3C). Considering that MPT0B291 is a selective inhibitor of HDAC6 and based on HDAC6 knockdown reduction of LINC00461 levels as verified by RNA-seq analysis (Fig. 1C) and RT-qPCR detection (Fig. 3D), HDAC6 was presumed to contribute to the regulation of LINC00461 stability. Additionally, we found that LINC00461 expression was higher in TMZ-resistant GBM cells. Increased expression of LINC00461 may correspond with the malignant behavior of GBM cells.
HDACs are known to regulate post-transcriptional gene expression by altering deadenylase complex activation and affecting polyadenylate tail stability of RNA [22]. Thus, we examined HDAC6 interaction with the catalytic subunits of the deadenylase complex and identified that carbon catabolite repression-negative on TATA-less (CCR4-NOT) core exoribonuclease subunit 6 (CNOT6), a typical protein related to mRNA decay, is associated with HDAC6 in U87MG cells (Fig. 3F). Furthermore, treating cells with HDAC6 inhibitors, MP0B291 and trichostatin A, led to a decrease in the interaction of HDAC6 and CNOT6 in cells (Fig. S3A), suggesting that deacetylase activity is essential for HDAC6 to bind CNOT6, causing changes in the deadenylase activation. Apart from RNA-degrading enzymes that contribute to RNA degradation, RNA-protective proteins that contribute to the stability of LINC00461 were also investigated. We purified LINC00461 (Fig. 3G) and performed RNA-protein pull-down assay (Fig. S3B) combined with iTRAQ-based proteomic analysis, by which several LINC00461-binding proteins were identified (Fig. 3H). Further investigation of the proteomics data using IPA, two RNA-stabilizing proteins, FUS and vimentin, were discovered as HDAC6-targeted factors that functioned in regulating RNA stability. Interestingly, a previous study has indicated that FUS is associated with several RBPs in the vicinity of the mRNA 3′ end and it controls polyadenylate tail maintenance [23]. To investigate the role of HDAC6 on these RNA-regulating proteins, HDAC6 inhibition was induced by MPT0B291 treatment or HDAC6 knockdown and this caused a significant decrease in FUS expression in both parental and TMZ-resistant GBM cells and a significant increase in CNOT6 expression (Fig. 3I-J and Fig. S3C). Furthermore, the effect of FUS knockdown on the RNA stability of LINC00461 showed a 3-fold reduction in the half-life of LINC00461 in TMZresistant Pt#3 cells (Fig. 3K).

LINC00461 depletion suppresses GBM cell proliferation through cell-cycle arrest
We examined the effects of LINC00461 on GBM cell proliferation. LINC00461 knockdown in parental and TMZ-resistant GBM cells reduced LINC00461 expression approximately by 40-50% (Fig. 4A). Interestingly, LINC00461 depletion suppressed cell proliferation significantly in both parental and TMZ-resistant GBM cells (Fig. 4B). To further determine the anti-proliferative effect of LINC00461 silencing on GBM cells, we analyzed cell-cycle distribution by flow cytometry. LINC00461 knockdown induced a significant cell-cycle shift from G2/M to G1 phases in parental GBM cells and similarly induced a slight perturbation of cell-cycle progression into G2/M phase in TMZ-resistant cells (Fig. 4C).
Recently, topoisomerase II alpha (TOP2A), a key enzyme in DNA replication to maintain cell proliferation, was found to be upregulated by LINC00461 through sponging of miR-411-5p during gliomagenesis [13]. Thus, we examined TOP2A expression, including levels of RNA (Fig. 4D) and protein (Fig. 4E), in our experimental model. The results showed that LINC00461 knockdown significantly attenuated TOP2A expression at the transcriptional level in both parental and TMZ-resistant U87MG cells. Additionally, MPT0B291 treatment showed a similar trend with reduced mRNA and protein expression of TOP2A in these GBM cells (Fig. 4F-G).

LINC00461 is involved in HDAC6-mediated cell division and survival regulation
To characterize the role of LINC00461 in GBM cells, we performed a genome-wide analysis of gene expression in LINC00461-silenced U87MG cells, and identified that 115 genes were upregulated and 219 genes were downregulated (Fig. 5A). Interestingly, the 219 down-regulated genes included TOP2A and CCNA2, both containing a gene expressing cyclin A2 to maintain the G2/M transition phase of the cell cycle. A comprehensive analysis conducted by comparing the LINC00461-silenced dataset with the MPT0B291-treated dataset revealed 105 intersection genes (Fig. 5B). The expression of 88 genes showed the same trend in both treatments. Functional analysis by IPA revealed highly significant functional gene classes associated with "cell division (50 of (See figure on next page.) Fig. 3 Interaction of HDAC6 and RBPs in regulating LINC00461 stability. A Expression levels of LINC00461 from transcriptome data, including two microarray data as shown in Fig. 1A  88 genes)" and "survival (49 of 88 genes)" (Fig. 5C). Furthermore, we examined the activity of several pathways using IPA, and found that either LINC00461 depletion or MPT0B291 treatment significantly inhibited cell proliferation/survival and induced cell death/senescence in GBM (Fig. 5D). These data indicated that LINC00461 is involved in HDAC6-regulated GBM malignancy. To further confirm the clinical relevance of the 50 cell-divisionrelated genes appearing in the intersection dataset (88 genes), we analyzed the TCGA-GBM next-generation sequencing (NGS) datasets of 124 patients. A strong negative correlation was observed (PCC: − 0.452) (Fig. 5E).
The result indicated that abnormalities in the expression of the 50 cell-division-related genes are inverted following the HDAC6/LINC00461 inhibition in GBM. Moreover, survival analysis using clinical outcome data showed that several cell-division-related genes (CENPF, DLGAP5, GTSE1, HMMR, KIF14, MCM10, MELK, RAC-GAP1, TMPO, and TOP2A) have poor prognostic implications in high-grade gliomas ( Fig. 5F and Fig. S4A), suggesting that the reducing expressions of these genes through HDAC6/LINC00461 inhibition may improve outcomes in GBM patients. Additionally, the importance of the HDAC6/LINC00461-mediated regulation in clinical GBM was analyzed using single-cell transcriptome sequencing datasets. Surprisingly, higher expressions of LINC00461 and its associated downstream cell-divisionrelated genes mentioned in Fig. 5E were enriched in a specific cluster (Cluster 5) of GBM cells (Fig. 5G-H).
Consistently, functional analysis of the conserved markers in Cluster 5 by IPA indicated highly similar functional gene classes as that of the LINC00461-silenced and MPT0B291-treated datasets mentioned in Fig. 5C  (Fig. S5A, Table S4).

Discussion
Although lncRNAs were considered non-functional and transcriptional noise initially, cumulative evidence over the past decade has unraveled that lncRNAs perform diverse biological functions [28]. Strategies for targeting lncRNAs in cancer treatment have gained widespread interest because they have vital roles in gene regulation and thus modulate hallmark processes of cancer [29].
Our previous studies demonstrated that HDAC inhibition using inhibitors represses stem-like properties in GBM that provide a potential therapeutic strategy to overcome drug resistance [10,30]. Nevertheless, those two studies focused on HDAC-regulated protein-coding genes associated with drug resistance, but the role of HDACs on the impact of dysregulated lncRNAs remains unclear and hence needs further investigation. Here, we investigated the involvement of HDAC6 in LINC00461 regulation by performing microarrays and RNA-seq analysis in stem-like and TMZ-resistant GBM cells with HDAC6 inhibition using MPT0B291 and siRNA. According to a transition in investigative approaches from lncRNA annotation and molecular characterization to tumorigenic mouse models needed for deciphering the physiological roles of cancer-associated lncRNAs [29], we examined the impact of LINC00461 on GBM growth and survival in vivo. Evidence gathered shows that MPT0B291 treatment significantly reduced subcutaneous tumor growth along with a decreased level of LINC00461 and a negative correlation between LINC00461 expression and survival of mice with intracerebral GBM, suggests a pivotal role for LINC00461 as an oncogenic factor in GBM malignancy. Indeed, a previous study has demonstrated that the expression of C130071C03Rik, the mouse ortholog of LINC00461, is increased not only in precancerous conditions but also during glioma progression [13]. Some characteristics suggest that LINC00461 is a potentially promising target for treating patients with GBM. First, previous literature on LINC00461 demonstrated its oncogenic functions across diverse cancer types [13,14,31,32]. Second, LINC00461 displays a brain-specific expression pattern, which implies reduced unintended toxic effects associated with their targeting. Third, unlike most intergenic lncRNAs that show poor sequence conservation, LINC00461 has a high degree of sequence homology across vertebrates [13], suggesting its essential role among such species [33].
The mechanisms underlying the regulation of mRNA stability have been extensively reviewed [34], whereas little is known about the modulation of lncRNA stability. Considering the inhibitory potency of MPT0B291 on HDAC1 and 2 (IC 50 ≈ 0.31 and 1.16 μM, respectively) [10,11], the impact of HDAC1 and HDAC2 on LINC00461 expression was also examined in GBM cells using siRNA-mediated depletion. Interestingly, the silencing of neither HDAC1 nor HDAC2 was involved in modulating LINC00461 expression (Fig. S7A), suggesting that a decrease in LINC00461 expression by MPT0B291 is mainly due to the inhibition of HDAC6 rather than that of HDAC1 or HDAC2. Although several aspects discriminate lncRNAs from protein-coding mRNA, they also share certain similarities [35]. Numerous lncRNAs are subjected to transcriptional editing, for example, splicing, polyadenylation, and 5′ capping, just like mRNAs [35]. Post-transcriptional regulation of lncRNAs is mediated by RBPs in mainly two aspects: stability and localization. Our results identified two HDAC6-associated proteins (CNOT6 and FUS) that might explain a shorter half-life of LINC00461 in response to MPT0B291 treatment (Fig. S7B). Because the enzyme activity of CNOT6 is required for deadenylation by the CCR4-NOT nuclease complex [36], and acetylation enhances CNOT6 activity inducing RNA degradation [22], we predict that high levels of HDAC6 in GBM cells inhibit CNOT6 acetylation, leading to LINC00461 stabilization, while MPT0B291 treatment induces CNOT6 acetylation, promoting LINC00461 degradation (Fig. S7B). FUS-containing 3′-end processing machinery facilitates mRNA stability [23], and lysine status regulated by cyclic adenosine monophosphate-response element binding protein/p300 and histone deacetylase families contribute to FUS function [37]. Therefore, in GBM cells, high levels of HDAC6 may increase the RNA-binding affinity of FUS, resulting in LINC00461 stabilization, whereas MPT0B291 treatment induces FUS acetylation, promoting LINC00461 degradation (Fig. S7B). Although our current results prove that HDAC6 inhibition increases CNOT6 and decreases FUS protein levels, further studies are needed to confirm whether HDAC6 directly regulates the lysine status of these two proteins.
Although recent studies have shown the oncogenic roles of LINC00461 in tumors [13,14,31,32], (See figure on next page.) Fig. 5 Involvement of LINC00461 in HDAC6-mediated cell division and survival in GBM cells. A Volcano plot visualizing the differentially expressed genes between si-NC and si-LINC00461 (si461) treatment. Green and red dots represent genes with significant downregulation and upregulation (p-value < 0.01), respectively, in si-LINC00461 compared with si-NC. B Venn diagram illustrating overlaps between the number of genes that were significantly altered in MPT0B291-treated and si-LINC00461-treated cells. C Top five molecular and cellular functions of genes altered in si-LINC00461-treated (334 genes), MPT0B291-treated (1709 genes), and intersection subset (88 genes) by IPA. The number of genes involved in the biological process is shown on the right side of the bar chart. D Heatmap representation of the top 20 predicted influential biological functions after treatment by IPA. Positive activation Z-score (red) represents activated pathways, and negative activation Z-score (green) represents inhibited pathways. E Relationships between si-LINC00461-treated NGS data (in horizontal) and TCGA-GBM NGS data (in vertical) by Pearson's correlation coefficient (PCC). Each dot represents the expression value of a cell-division-related gene that existed in the intersection subset. F Forest plots showing hazard ratios for risk of mortality in high-grade glioma patients with higher expressions of the indicated genes. Circles represent the hazard ratio, and bars on both sides denote their corresponding 95% confidence intervals (CI). The original survival curves were obtained from PROGgeneV2 (Fig. S4). Log-rank test. G Individual cells suspended/isolated directly from patient's fresh GBM sample were used for single-cell RNA sequencing. Each dot denotes one cell with the expression of LINC00461, and the depth of blue represents the amount of LINC00461 expression. the molecular mechanisms underlying LINC00461 regulation remain unclear. To understand the function of LINC00461 comprehensively, we compared the LINC00461-silenced dataset with the clinical TCGA-GBM dataset. Pathway enrichment analysis of intersected genes in these two datasets revealed cell growth, which is consistent with the findings that LINC00461 promoted GBM proliferation through the regulation of TOP2A expression. TOP2A is a nuclear enzyme that controls and alters the topologic states of DNA during transcription and mitosis, facilitating gene expression and mitotic progression, respectively, in tumor cells [38].
Here, we found that the HDAC6/LINC00461 axis inhibition downregulated TOP2A expression in parental and TMZ-resistant cells, thereby explaining the perturbation of cell-cycle progression into G2/M phase. In addition, a previous study has demonstrated that LINC00461 knockdown reduces expressions of a few cell-division-related genes, including cyclin D1, CDK4, cyclin A2, and cyclin E in U87MG [39]. Although genome-wide analysis with critical criteria for selecting potential LINC00461 downstream targets from altered genes in the LINC00461silencing dataset only identified cyclin A [39] and TOP2A [13] as presented previously, we measured the expressions of the other cell-division-related genes mentioned above in our GBM cell model. Significant decreases in the expression levels of these cell-division-related genes (cyclin D1, CDK4, cyclin A2, and cyclin E1) following LINC00461 knockdown were observed in parental and TMZ-resistant U87MG (Fig. S8A). Furthermore, LINC00461 knockdown significantly downregulated protein expressions of cyclin D1 and CDK1 (Fig. S8B), which is consistent with the findings of a previous report [39]. Except for the LINC00461-regulated cell-cycle-related genes, which have already been proved, numerous celldivision-related genes that are modulated by LINC00461 were revealed in this study for the first time (Figs. 5E and 6). In summary, LINC00461 displays a widespread regulation of cell-division-related proteins that are indispensable for GBM growth. Apart from exploring the function of LINC00461 using the public clinical dataset, we also performed transcriptome profiling of clinical samples at the single-cell level.
Recent advances in the development of NGS technologies have overcome conventional profiling methods that assess bulk populations and provide the opportunity to gain insights into the characterization of individual cells [40]. Considering the heterogeneity of tumor biology, we explored whether there is a cluster of cells overexpressing LINC00461. Surprisingly, our findings revealed that the cells in Cluster 5 showed highly elevated expressions of LINC00461 and cell-division-related genes simultaneously and also displayed biomarkers enriched in pathways associated with cell division (Fig. S5A, Table S1). Therefore, the cell characteristics in Cluster 5 are require further study to clarify the clinical malignancy of GBM and the development of treatment strategies.
Like proteins, appropriate subcellular localization patterns of lncRNAs allow for the primary determination of their molecular functions [41]. Our observations in xenograft tumors through both RNA-ISH and RNA-FISH staining demonstrated that LINC00461 localizes to the cytoplasm rather than to the nucleus. Cytoplasmic lncRNAs modulate mRNA fate by acting as miRNA sponges that compete for miRNA binding [42]. To dissect the regulatory mechanism between LINC00461 and cell-division-related proteins comprehensively, we used three reliable public datasets to predict lncRNA-miRNA-mRNA networks. Downregulation of MCM10 and MELK proteins by LINC00461 depletion combined with the induction of miR-485-3p by LINC00461 inhibition using MPT0B291 suggests that this strategy is effective for identifying the downstream regulatory networks of LINC00461.

Conclusions
Recent research has unraveled the indispensable roles of lncRNAs in regulating GBM malignancy, including cancer stem-like features [8]. Nevertheless, the mechanism by which cancer-associated lncRNAs are dysregulated remains unexplored. Here, our results provide an insight into lncRNA regulation by HDAC proteins and lead to a comprehensive understanding of LINC00461associated networks in GBM malignancy, by expounding the utility of the HDAC6/RBPs/LINC00461 axis as a potential therapeutic approach for treating patients with