Small molecule-mediated disruption of ribosome biogenesis synergizes with FGFR inhibitors to suppress glioma cell growth

Abstract

Background

High-grade gliomas are malignant brain tumors characterized by aggressiveness and resistance to chemotherapy. Prognosis remains dismal, highlighting the need to identify novel molecular dependencies and targets. Ribosome biogenesis (RiBi), taking place in the nucleolus, represents a promising target as several cancer types rely on high RiBi rates to sustain proliferation. Publicly available transcriptomics data of glioma patients revealed a positive correlation between RiBi rates and histological grades. We, therefore, hypothesized that glioma cells could be susceptible to RiBi inhibition.

Methods

Transcriptomics data from glioma patients were analyzed for RiBi-related processes. BMH-21, a small molecule inhibitor of RNA pol I transcription, was tested in adult and pediatric high-grade glioma cell lines and a zebrafish transplant model. Cellular phenotypes were evaluated by transcriptomics, cell cycle analysis, and viability assays. A chemical synergy screen was performed to identify drugs potentiating BMH-21-mediated effects.

Results

BMH-21 reduced glioma cell viability, induced apoptosis, and impaired the growth of transplanted glioma cells in zebrafish. Combining BMH-21 with TMZ potentiated cytotoxic effects. Moreover, BMH-21 synergized with Fibroblast Growth Factor Receptor (FGFR) inhibitor (FGFRi) Erdafitinib, a top hit in the chemical synergy screen. RiBi inhibition using BMH-21, POLR1A siRNA, or Actinomycin D revealed engagement of the FGFR–FGF2 pathway. BMH-21 downregulated FGFR1 and SOX2 levels, whereas FGF2 was induced and released from the nucleolus.

Conclusions

This study conceptualizes the implementation of RiBi inhibition as a viable future therapeutic strategy for glioma and reveals an FGFR connection to the cellular response upon RiBi inhibition with potential translational value.

Key Points.

• Elevated ribosome biogenesis correlates with worse clinical outcomes in glioma.

• RNA Pol I inhibitor, BMH-21 inhibits glioma cell growth and synergizes with temozolomide.

• FGFR1 inhibitors synergize with BMH-21, revealing a RiBi-FGFR1/FGF2 crosstalk.

Importance of the Study.

High-grade gliomas (HGG), including glioblastoma multiforme (GBM), are the most frequent intracranial tumors in adults and are considered incurable, lacking treatment options that come with long-term survival, which for GBM rarely exceeds 14 months. This study explores the role of ribosome biogenesis (RiBi) in glioma malignancy and its contribution to clinical outcomes. We thoroughly characterize the previously unexplored biological effects of RiBi blockade in HGG using the rDNA transcription inhibitor BMH-21, which showed remarkable cytotoxic activity against glioma cells in vitro and orthotopic zebrafish glioma models, highlighting the RiBi potential as a therapeutic target for HGG. A synergy screen, performed to identify drugs that potentiate RiBi blockade and overcome the inherent HGG resistance to monotherapies, pinpointed Fibroblast Growth Factor Receptor (FGFR) inhibitors as the top hits. Mechanistic implications behind the synergy observed unfolded an FGFR–FGF2-mediated cellular response to RiBi inhibition with prospective translational value.

Gliomas are malignant primary brain tumors classified into 4 histological grades; high-grade gliomas (HGG), including glioblastoma multiforme (GBM), are the most frequent and aggressive intracranial tumors in adults.¹ GBM lacks treatment options that come with long-term survival, with the standard of care being surgical resection followed by radiation and temozolomide (TMZ) chemotherapy.

HGGs frequently exhibit upregulated receptor tyrosine kinase (RTK) activity due to genetic alterations in growth signaling pathways, including PDGFR and EGFR. Combined with PI3K mutations and/or PTEN loss, these result in PI3K/AKT/mTOR pathway overactivation,² rendering its inhibition a rational therapeutic approach against glioma. Nevertheless, the expected clinical outcomes were not met, partly due to compensatory pathways, resulting in resistance development.³

An emerging strategy to target PI3K/AKT/mTOR pathway is the perturbation of one of its downstream processes, Ribosome Biogenesis (RiBi). RiBi fuels the production of ribosomes, critical for protein synthesis, while it is upregulated in several cancers, enabling abnormal cell growth.⁴ It is executed in the nucleoli, which emerge at the sites of actively transcribed rRNA genes, known as nucleolar organizer regions (NORs).

RiBi perturbation triggers impaired ribosome biogenesis checkpoint (IRBC)/p53-mediated alongside p53-independent cell cycle arrest and death.⁴ Commonly used chemotherapeutic drugs were shown to interfere with RiBi at multiple steps, establishing the pharmacological importance of RiBi in cancer.⁵ The emerging class of RiBi-inhibitory drugs, including the Pol I transcription inhibitor BMH-21,⁶ has been shown to inhibit cancer cell growth, including neuroblastoma,⁷ and prostate,⁸ revealing new cancer dependencies. In brain tumors, the importance of RiBi remains elusive. We sought to evaluate its role and potential as a targeted therapeutic approach against HGG.

Materials and Methods

See Supplementary Materials and Methods for detailed experimental procedures.

Ethics Statement

Animal experiments were performed per the national ethical guidelines and regulations (N207/14). Experiments on adult zebrafish were covered by ethical permit no.14049-2019, while embryos younger than 5 days do not require an ethical permit. U4 and U18 cell lines were derived under permit from the research ethics committee in Uppsala (UpS98415).⁹ Pediatric glioma lines were covered by ethical permit Dnr-604-12 issued by Gothenburg’s regional ethical review board.

Cell Culture

Glioma Tumor Cell Panel was obtained from ATCC (TCP-1018). U4 and U18 high-grade primary glioma cell lines were provided by D. Hägerstrand.⁹ Normal human astrocytes were purchased from Lonza. Pediatric high-grade glioma lines were isolated and provided by H. Carén and cultivated in Stem Cell media (DMEM-F12 supplemented with B27 (Gibco), N2 (Gibco), and EGF (10 ng/ml, Peprotech) as described.^10,11

Cell Viability Assay

Cells were seeded in 96-well plates (3–5 × 10³/well) and treated as indicated. Viability was determined by fluorescent-based (resazurin) or luminescent-based assays (Cell-TiterGlo, Promega) using Tecan Infinite M1000Pro or EnVision PerkinElmer.

Data Mining

RNASeq data and metadata for glioma patients were retrieved from the TCGA and CGGA repository¹² and processed in R.

Gene Silencing

Cells were transfected with 20–80 nM SMARTpool ON-TARGET oligonucleotides (D-001810, Dharmacon), or Flexitube GeneSolution siRNAs (Qiagen).

HTS Synergy Screen

The CBCS oncology drug collection (482 compounds) was tested on A172 cells with DMSO or BMH-21 for 72h. Cell viability results were normalized as %viability of DMSO and statistically analyzed using GraphPad (v.8.2.1). Assay robustness was evaluated by a Z-factor > 0.5. Compounds were ranked by cell viability difference between co-treated with DMSO/BMH-21 (%δAUC).

Immunoblotting

Immunoblotting and immunofluorescence were performed as previously described.¹³ Antibodies are listed in Supplementary Table S1. Human apoptosis array kit was performed per the manufacturer's instructions (ARY009, R&D Systems).

Immunofluorescence

Apoptotic cells were quantified following staining with CellEvent Caspase3/7 Green Detection Reagent (ThermoFisher Scientific) or Cleaved Cas3 antibody (Abcam). Cell cycle analysis was performed with Click-iT®-EdU, and RNA synthesis was quantified by Click-iT®-EU (Life Technologies). Cells were subjected to EdU (10 μM/30–60 min) or EU pulse-labeling (100 µM/60 min) before the treatment endpoint. Images were acquired by INCell Analyzer2200 (GE Healthcare).

Migration Assay

Cell migration was quantified using the CytoSelect 24-Well Cell Migration Assay (CellBiolabs Inc.). Migratory cells were stained, captured by ZEISS PrimoStar Microscope and Axiocam ERc5 camera (Zeiss), and quantified by measuring OD_560nm.

Quantitative RT-PCR

The TaqMan RNA-to-CT-1-Step Kit (ThermoFisher) in a QuantStudio 5 PCR System (Applied Biosystems) was used. Relative quantity was analyzed with the ΔΔCt method using GAPDH as an internal normalizer. TaqMan probes are listed in Supplementary Material.

RNA Pol I Transcription Assay

Cells were transiently co-transfected with the pHrD and pRLTK vectors in OptiMEM (Fisher Scientific) and seeded in 96-well plates. Following treatment, luminescent signal was measured after cell lysis using the Dual-Luciferase Assay (Promega).

Statistics and Reproducibility

RNAseq was performed in biological triplicates: statistical analysis followed the tests provided in the DE analysis packages in R (student’s t-test,**P < .01,****P < .001). All other experiments have been performed a minimum of 3 times with at least 3 technical and/or biological replicates. Statistical significance was determined using t-tests with GraphPad Prism. Data are shown as means ± SD; n = 3 technical replicates; P-values considered significant were: *P < .0332,**P < .0021,***P = .0002, and****P < .0001. P-values are shown at the graph corner if the same for all comparisons or for each separately if different among groups.

Transcriptomics

RNAseq was performed by the SciLifeLab National Genomics Infrastructure (NGI). Data preprocessing through nf-core/RNAseq (NGI). DE analysis utilized DESeq2¹⁴ (v1.24.0) and dose-dependent gene clustering with maSigPro¹⁵ (quadratic regression with FDR < 0.05 and backward stepwise regression with P < .05). Gene ontology (GO) and pathway enrichment analysis conducted as described.¹⁶ FASTQ files generated specifically for this study are available at NCBI’s BioProject repository (ID: PRJNA825234).

Zebrafish Transplantation, Cryosections, and BBB Analysis

tdTomato/luciferase stably expressing GBM18 and H4 cells, generated from U18/H4 lines with a lentivirus-based approach (#32904, Addgene), were injected into AB or fli:EGFP zebrafish embryos as described.¹⁷ Embryos were imaged using ImageXpress Nano (Molecular Devices) and lysed or processed for cryosectioning.¹⁷ Cryosections were stained per routine protocols (Supplementary Table S1). 
Images were taken with NikonTi2. For Blood-brain Barrier (BBB) analysis, adult zebrafish were treated for 4h and sacrificed by hypothermic shock. Following decapitation, brains were removed and snap-frozen. Brain tissue was homogenized using a bead-mill homogenizer (Fisherbrand) and subjected to LC/MS-MS. Detected drug concentrations were normalized against water-tank drug concentrations.

Results

Ribosome Biogenesis Is Upregulated in HGG

Early studies on GBM reported an association between nucleolar size and patient survival.¹⁸ As nucleolar size is proportionally related to RiBi activity,¹⁹ we studied RiBi in HGGs to discover potential vulnerabilities. We compared RNASeq datasets from GBM patients (TCGA) to those of normal donors to explore any RiBi-associated gene signatures. A Differential expression (DE) analysis showed 8550 deregulated genes (Supplementary Figure S1A, Table S2A), with RiBi transcripts being particularly represented in the upregulated gene set (Figure 1A, B, Supplementary Table S2B).

Figure 1.

RiBi is a relevant target for high-grade gliomas. (A) Differential expression (DE) analysis of RNAseq data between GBM (TCGA) and normal samples, with downregulated/green and upregulated/orange genes (log2FC > 0.5, padj < 0.05). (B) Reactome pathway enrichment analysis of DE genes from (A). z-score shows the tendency of genes within a term to be up/downregulated. (C) DE genes z-score-heatmap among GBM-normal or stages II to IV samples. (D) RiBi score and glioma grading association in TCGA cohort. (E) Kaplan–Mayer plot showing survival advantage of low versus moderate and high RiBi GBMLGG patients in TCGA. (F) BMH-21 dose–response (72 h) in normal human astrocytes (NHA) and glioma cell lines, with corresponding GI50 values (G) rDNA transcription in A172 and U18, quantified using an rRNA-promoter reporter assay. (H) Representative immunofluorescence images of nucleolar proteins FBL and UBF in glioma cells. Nuclei are counterstained with DAPI. Scale bar, 10 μm. (I) Indicated protein levels in glioma cells; beta-actin served as loading control (J–K) Click-it EdU-based cell-cycle analysis representative dot plots (J) and stacked barplot (K) of cell-cycle distribution into G₁, S, and G₂.

We next performed DE analysis between grade II and IV patients (Supplementary FigureS1B, Table S2C) and compared the Reactome GO terms with those from the GBM-Normal group. The analysis revealed shared RiBi-associated terms, with “rRNA processing” being the most representative (Figure 1C, Supplementary Table S2D). Utilizing a 25-gene RiBi gene signature produced by univariate/multivariate analysis of the TCGA RNASeq data (Supplementary Figure S1C, Table S3A), we ascribed an RiBi score to each patient’s tumor and noticed that HGG showed elevated RiBi scores (Figure 1D; Supplementary Figure S1D), while patients with tumors classified as low-RiBi tended to survive longer (Figure 1E; Supplementary Figure S1E, Table S3B). Multivariate COX-regression analysis further showed that the prognostic value of RiBi status is independent of other important clinical features for gliomas, with patients characterized as low RiBi, low grade, with WT IDH and methylated MGMT promoter showing better prognosis compared to their counterparts (Supplementary Figure S1F, G). Our analysis indicates a gradient upregulation of RiBi genes in normal subjects to HGG patients and pinpoints the overlooked prognostic value of RiBi in gliomas.

RiBi Inhibition Reduces Glioma Cell Viability

We hypothesized that glioma cells might depend on high RiBi rates and, thus, be susceptible to RiBi inhibition. We performed a dose-response viability assay in 2 glioma cell lines, comparing cytotoxicity of BMH-21, CX-3543, and CX-5461, rDNA transcription inhibitors with diverse mechanisms.⁴ BMH-21 was the most potent, with GI₅₀s around 0.30 μM for both cell lines (Supplementary FigureS1H) and selected for further evaluation. Dose–response assays of 7 high-grade glioma cell lines confirmed the potent growth inhibitory effect with a mean GI₅₀ of 0.30 μM (Figure 1F). An 8-fold higher GI₅₀ was recorded for normal human astrocytes (NHA), indicating a potentially acceptable therapeutic index.

Effective Degradation of RPA194 by BMH-21

Chemical perturbation of RiBi is accompanied by nucleolar structure alterations and IRBC activation.⁴ In both p53-proficient and mutant glioma cell lines, BMH-21 showed a dose-dependent decrease of the rDNA promoter activity (Figure 1G), as quantified utilizing an rRNA promoter-luciferase reporter, while immunofluorescence (IF) against the nucleolar markers fibrillarin/FBL, UBF (Figure 1H), nucleolin/NCL, and RPA194 (Supplementary Figure S1I) showed nucleolar caps and NCL translocation, hallmarks of RiBi impairment.⁴ BMH-21 triggered the degradation of Pol I catalytic subunit RPA194 and induced cell death, shown by PARP1 cleavage; BMH-21 treatment of p53-proficient cell lines (A172, H4) displayed p53 stabilization and p21 upregulation (Figure 1I). These effects were less pronounced in NHA (Supplementary Figure S1J). Cell cycle analysis using 5-ethynyl-2′-deoxyuridine (EdU) incorporation and high-content microscopy showed that low BMH-21 doses induced a robust, dose-dependent G₁ arrest in A172 and H4, with a concomitant increase of the G₂ population (Figure 1J–K & Supplementary Figure S1K–L). Taken together, BMH-21 impairs rDNA transcription and causes cell cycle arrest and cell death.

Transcriptional Changes in Glioma Cells Treated With BMH-21

To address the transcriptional changes induced by BMH-21, we performed an RNAseq on H4 cells treated with low (BMH-21^0.15) or high (BMH-21^1.0) doses for 24 h (Supplementary Figure S2A). BMH-21^0.15 perturbed the expression of 6117 genes (Supplementary Figure S2C, Table S4A) versus 9049 upon BMH-21^1.0 (Supplementary Figure S2D, Supplementary Table S4B), with p53-signature-associated transcripts being particularly represented and, in most cases, monotonically upregulated in both concentrations, as shown by the dose-dependent DE analysis (Supplementary FigureS2E–F, Supplementary Table S4C). BMH-21 induced an overall suppression of genes involved in cell cycle, DNA replication, RNA metabolism, and translation, whereas genes involved in p53 signaling (e.g., CDKN1A, MDM2) were upregulated (Figure 2A–B, Supplementary Table S4D). BMH-21^0.15 induced the expression of genes related to “interferon signaling” such as STAT1/STAT2 (Supplementary Figure S2G, Supplementary Table S4E). RNAseq data were validated by qRT-PCR (Supplementary Figure S2H–I).

Figure 2.

BMH-21 induces glioma cell death and suppresses glioma cell growth in zebrafish model. (A) Heatmaps showing the most significant DE genes corresponding to Reactome terms p53-Signaling (left) and Cell Cycle (right) following indicated 24 h-treatments (μM). (B) Reactome PE analysis in DE genes (g: Profiler). (C) Venn diagram comparing GO-BP ontology terms among BMH-21^0.15 and BMH-21^1.0 (padj < 0.05) corresponding to cell death-related signatures. (D) Heatmap shows commonly enriched terms and gene perturbations corresponding to apoptotic signatures among BMH-21 concentrations. (E) Quantification of Cas3/7-positive nuclei (48h-treatment). (F) Quantification of the luminescent signal emitted from injected zebrafish embryo lysates (72h-treatment). VEH mean signal intensity = 100% luminescence. (G) representative images and area quantification of fluorescent tumor cell transplants 24 and 96 hpf. (H) Immunofluorescence images of zebrafish cryosections (72 h-treatment, stained for Ki67 and NPM1). Insets depict magnifications of the designated regions. Scale bar 100 μm. (I) Relative BMH-21 and TMZ quantification in dissected zebrafish brains following 4 h-treatment, suggesting BBB permeability.

BMH-21 Induces Apoptotic Cell Death and SOX2 Downregulation

By comparing biological process-associated gene ontology (GO-BP) terms of DE genes between BMH-21^0.15 and BMH-21^1.0, we identified 271 commonly enriched terms, referring mainly to cell death, with downregulation of anti-apoptotic genes like BCL-2, and induction of pro-apoptotic genes, i.e., FAS and BIRC3 (Figure 2C,D, Supplementary Table S4F,G). High-content apoptotic assays corroborated the efficacy of BMH-21 in inducing apoptosis (Figure 2E; Supplementary Figure 2SJ). Apoptosis induction was only slightly affected when knocking down p53 by siRNA (Supplementary Figure S2K–L), whereas it was attenuated by caspase inhibitors (Supplementary Figure S2Q–T).To detect p53-independent cell death signatures upon BMH-21 we performed a GO analysis by subtracting known p53 target genes from our RNAseq dataset, suggesting that multiple pathways are potentially involved (Supplementary Figure S2V).

Given the central role of cancer stem-like cells (CSCs) in drug resistance and tumor recurrence, we treated cells cultured in serum-free Neural Stem Cell (NSC) media to promote CSC-associated phenotypes, and quantified apoptosis; the cytotoxic effect of BMH-21 was preserved in both NSC-grown cell lines tested (Supplementary Figure S2M,N). To address the expression of CSC markers following BMH-21, we stained against SOX2 and quantified SOX2-expressing cells in both serum- and NSC-grown cells (Supplementary Figure S2O, P). SOX2-positive numbers in serum-grown cells were reduced dose-dependently, suggesting that inhibition of Pol I activity could promote differentiation. However, a higher dose was required in NSC-grown cells to exert a similar effect on the SOX2 population. Notably, NSC-grown cell numbers were drastically reduced in all BMH-21 doses, in line with the apoptotic assay; however, half of the surviving population treated with BMH-21^0.15 was found to be SOX2-positive, suggesting a possible SOX2-mediated mechanism of CSCs to overcome Pol I inhibition.

BMH-21 Impairs Growth of Transplanted Glioma Cells in Zebrafish

To evaluate the efficacy of BMH-21 in vivo, we utilized a previously published orthotopic glioma zebrafish model.¹⁷ Results showed a 50% reduction in signal intensity in GBM18- and H4-Luc-transplanted embryos treated with BMH-21^0.15 
(Figure 2F). Live imaging confirmed decreased signal intensity (Figure 2G). Immunofluorescence staining of cryosections for Ki67 and Nucleophosmin/NPM1 revealed reduced Ki67 signal and NPM1 relocation to the cytoplasm in the BMH-21 group (Figure 2H). To evaluate the overall acute toxicity of BMH-21, we treated 48hpf embryos for 48h. The maximum tolerated dose (MTD) was 3µM. No gross neurotoxic effects could be observed. Increasing concentrations slightly affected heartbeat (Supplementary Figure S2X–Z). BBB permeability was also assessed by analyzing drug accumulation in dissected brains of adult zebrafish using LC/MS-MS. We detected BMH-21 at relatively higher levels than the BBB-permeable TMZ (Figure 2I), in agreement with prediction algorithms and results obtained in the embryo transplant model.

BMH-21 Sensitizes Glioma Cells to TMZ

Next, we explored the synergistic potential of BMH-21 with TMZ by performing a cell viability-based synergistic analysis for 3 glioma cell lines. The most synergistic dose pair—0.15 μΜ BMH-21 and 300 μΜ TMZ (hereafter mentioned as CMB)—was validated and showed a 70%–90% viability inhibition in all 3 cell lines, but only a 20% inhibition in NHA (Figure 3A; Supplementary Figure S3A–C). The CMB synergy was further confirmed by performing order-of-addition analysis, adding BMH-21 24h before, after, or concomitantly with TMZ (Supplementary Figure S3D). CMB retained the DNA-damaging activity of TMZ and the nucleolar activity of BMH-21 (Figure 3B), while enhancing p53 stabilization in A172 and caspase activation both in serum and serum-free conditions (Figure 3C–D & Supplementary Figure S3F). CMB reduced slightly γ-H2AX while retaining high p53 levels compared to TMZ alone in A172 (Figure 3B). This reflects probably an intermediate, cell-line-specific effect of TMZ on the cell cycle as shown by our analysis (Figure 3E; Supplementary Figure S3E). Finally, CMB moderately inhibited cell migration and abrogated the colony-forming capacity of cells (Figure 3F-G; Supplementary Figure S3G). Collectively, CMB showed a promising effect on cell death and migratory features, even though BMH-21 monotherapy was almost as effective in these assays.

Figure 3.

BMH-21 synergizes with TMZ (A) Cell viability of glioma cell lines compared with NHA upon 72 h treatment. (B) Indicated protein levels following 48 h treatment; beta-actin served as a loading control. (C–D) IF-based Quantification of Cleaved Cas3-positive nuclei in (C) serum- and (D) NSC-grown glioma lines (48 h). (E) Cell-cycle distribution barplot of A172 and U18 upon 24 h treatment. (F) Migration assay showing cell-migratory capacity during 6 h-treatment (G) Colony formation assay of H4 and U18 following 72 h-treatment and then 15-day-culture in drug-free media.

BMH-21 Synergizes With Inhibitors of the FGFR Pathway

To identify drug classes that could potentiate RiBi inhibition, we performed a high-throughput synergy screening of 482 oncology drugs in A172 (Supplementary Figure S4A). Compounds were classified into non-synergistic (158), synergistic (147), antagonistic (24), or highly toxic (153) (Figure 4A–2; Supplementary Figure S4B). The results revealed 30 hit combinations (Figure 4A–3), whereas the highly toxic hits were re-evaluated in lower concentrations as monotherapy candidate hits (Figure 4A-4), indicating 70 potent drugs (Supplementary Table S5). Among the 30 synergistic hits, 37% were RTK/MAPK inhibitors; hits were distributed across all clinical development stages (Figure 4A-3-upper) with, only 7% being investigated against glioma.

Figure 4.

High-throughput screen reveals compound classes that synergize with BMH-21. A172 cells were treated with 482 compounds in 5 concentrations, +/−BMH-21^0.15 for 72 h. (A) Data analysis and hit selection pipeline. (1) Compound ranking based on %δAUC. Cell viability (%DMSO) for each compound (y-axis) when combined with DMSO or BMH-21^0.15, as well as the %δAUC; horizontal broken lines indicate DMSO/TMZ controls. (2) Ranked compounds assigned into 3 groups (A–C) based on dose-response profiles indicated by brackets. (3) Synergistic compounds validation (Group B). Hits shown in pie charts, classified based on clinical stage and MoA; final-lead compounds shown with their MoA., and (4) Reevaluation of the 70 monotherapy hits (MoA classification pie chart). (B) A172 Cell viability (72 h treatment) and (C) Cas3/7-positive nuclei (48 treatment) with final synergistic pairs. (D) Luminescence from zebrafish 72 hpf-embryos (48 h-treatment). VEH mean value = 100%luminescence.

Six lead compounds were selected following validation and filtering; half were Fibroblast Growth Factor Receptor (FGFR) inhibitors (Figure 4A-3-lower). AZD4547 and Erdafitinib (ERD) were chosen based on their efficacy and the computationally predicted blood-brain barrier permeability score (Supplementary Figure S4C) to explore further the synergy observed. ERD was tested across 4 additional glioma cell lines to address whether the synergistic effect is cell line-specific (Supplementary Figure S4D). Additive effects were observed among all lines tested; H4 cells were particularly sensitive to the combination. This sensitivity could be explained by the common mutational status with A172, bearing CDKN2A and PTEN mutations. The most synergistic pairs were further evaluated by viability (Figure 4B; Supplementary Figure S4E) and apoptotic assays (Figure 4C; Supplementary Figure S4F), demonstrating the synergistic effects; combination with ERD was superior to AZD4547. The efficacy of BMH-21 and ERD was tested in both serum- and NSC-grown H4 cells, and in the p53-deficient U118 and U18 (Supplementary Figure S4F–H); the latter cell line showed decreased levels of survivin and claspin, and increased levels of Bax and cleaved caspase-3 using an apoptosis antibody array (Supplementary Figure S2U). Zebrafish H4 cell transplants were also tested and showed a reduced luminescence of ~70% following 48 h-treatment with the combination (Figure 4D).

RiBi Inhibition Triggers FGF2 Upregulation via FGFR1

ERD is a pan-FGFR kinase inhibitor with a higher affinity for FGFR1-3, receptors with different roles and expression levels in glioma.² Given the association of FGFR1 expression with higher glioma grade, invasion, and irradiation resistance,²⁰ alongside its high expression levels and dependency scores observed among glioma cell lines (DepMap, Supplementary Figure S5Q), we questioned whether it mediates the synergistic effects observed. BMH-21 treatment of FGFR1 depletion or chemical inhibition enhanced the apoptotic effect of BMH-21 (Figure 5A; Supplementary Figure S5A, B, M). Functional redundancy among FGFR members cannot be excluded, with FGFR4 also showing a synergistic effect, even though its depletion appears to influence FGFR1 levels in A172 (Supplementary Figure S5R–U).

Figure 5.

RiBi inhibition in glioma cells triggers an FGF2/FGFR1 response. (A) Cas3/7-positive nuclei following FGFR1-knockdown (KD). (B–C) FGF2 protein levels quantification by IF. (D) Immunofluorescence images in H4 cells. Scale bar, 10μm. (E–G) FGF2 signal quantification by 
immunofluorescence following POLR1A (E) or FGFR1 (F–G) depletion. (H) Indicated protein levels in A172 and H4 (I) Indicated protein levels following time-course treatment with BMH-21 and/or ERD in H4. (J) SOX2 and FGFR1 (left) or FGF2 (right) protein levels were treated overnight following FGFR1- or FGF2-KD. Beta-actin or tubulin served as loading controls. (K) Quantification of SOX2 protein levels by immunofluorescence following BMH-21 and/or ERD w/wo FGF2 addition. Treatments’ duration in (A–K) was 24 h.

FGFR1 is expressed in CSCs, regulating crucial stem cell-associated transcription factors, including SOX2 and ZEB1, upon FGF2 stimuli.²⁰ We, therefore, focused on the FGF2-FGFR1 axis, starting with our RNAseq dataset, where FGF2 was consistently upregulated by BMH-21 (Supplementary Figure S5C).

FGF2 has been shown to localize in the nucleolus and promote rDNA transcription.²¹ In agreement, we found that FGF2 tends to accumulate in the nucleolus and is translocated to the nucleoplasm and cytoplasm upon nucleolar stress induced by BMH-21 or the Pol I inhibitor ActD (Figure 5D). TMZ, Hydroxyurea, or Erdafitinib did not induce translocation, indicating a previously unknown RiBi-FGF2 interplay. FGF2 signal intensity quantification showed increased protein levels upon BMH-21^0.15, ActD (Supplementary Figure S5D), and POLR1A/RPA194 depletion by siRNA (Figure 5E & Supplementary Figure S5E), corroborating transcriptomics data and confirming a Pol I inhibition-mediated response. Combining BMH-21^0.15 with Erdafitinb blunted FGF2 levels induction at the protein (Figure 5B–C; Supplementary Figure S5F–G) and mRNA level (Figure 5F), while by performing ELISA against FGF2 on the cell supernatants, we also detected increased secretion following BMH-21 (Supplementary Figure S5G).

Next, we questioned if FGF2 induction is FGFR1-mediated; FGFR1 knockdown attenuated FGF2 induction (Figure 5F–G). Total and phosphorylated FGFR1 levels were downregulated by BMH-21 in A172 and U18 cells but not in U118 (Supplementary Figure S5K). FGFRi treatment resulted in phosphorylation recurrence after 24h, an effect rescued by the combination in all cell lines tested (Figure 5H; Supplementary Figure S5K). FGFR1 downregulation was also observed following ActD (Supplementary Figure S5L) or siPOLR1A (Supplementary Figure S5E). To understand the kinetics of FGF2 induction and whether it precedes ERK phosphorylation—a downstream event of FGFR1-mediated RAS/MAPK activation²²—we performed a time-course treatment with BMH-21, ERD, or their combination for up to 24 h (Figure 5I). In BMH-21-treated cells, FGF2 levels increased after 4 h treatment and peaked after 12 h, followed by ERK1/2 reactivation. On the contrary, ERD induced FGF2 levels after 8 h, while ERK1/2 phosphorylation recurred within 12 h, possibly due to FGFR or alternative RTK signaling recurrence that may help glioma cells overcome FGFR inhibition.²³ Notably, RPA194 protein levels also increased after 12 h, suggesting upregulated Pol I transcription, which could be triggered directly by FGF2 or as a response to growth signals by ERK1/2.²⁴ Most importantly, ERK1/2 phosphorylation was completely inhibited by the combination. Our data hint toward an FGFR1-mediated glioma cell response that could potentially overcome nucleolar stress and promote cell growth, effects that are blunted by FGFR1 inhibition.

FGF2 Upregulation Mediates SOX2 Levels

FGF2 regulates glioma cell growth, vascularization, and CSC self-renewal, by inducing stem cell-associated transcription factors, such as SOX2.²⁵ To address whether FGF2 upregulation could affect BMH-21 susceptibility, we performed an apoptotic assay in EGF-containing only, NSC-grown H4 cells by concomitantly adding recombinant FGF2. FGF2 appeared to have a protective role, partly rescuing cell death (Supplementary Figure S5I). To elucidate the effect of BMH-21-mediated FGF2 induction on SOX2, we knocked down FGF2 or FGFR1 and performed an IB against SOX2 (Figure 5J. We observed that in serum-grown H4 cells, BMH-21 downregulates SOX2, and this effect is likely FGFR1- and FGF2-independent; however, the basal SOX2 expression was reduced in both cases, supporting the previously described FGFR1-FGF2-SOX2 positive feedback loop.²⁰ Next, we asked whether these findings could explain the SOX2 upregulation by BMH-21, observed only in the FGF2-supplemented NSC cultures, where the FGFR1/FGF2 axis is enhanced (Supplementary Figure S2P). We stained for SOX2 in serum-grown A172 and H4 cells following BMH-21 and/or ERD treatment, with or without recombinant FGF2 addition. We observed that a 24h-incubation with FGF2 is sufficient to upregulate SOX2 levels, which is more pronounced when combined with BMH-21 (Supplementary Figure S5K). Protein levels of stem cell-associated markers such as N-cadherin, SOX2, and AKT phosphorylation, were also upregulated upon FGF2 addition and BMH-21, and that was prevented by ERD; importantly, p53 stabilization and PARP1 cleavage levels remained the same, indicating that the cell death-inducing capacity is retained (Supplementary Figure S5J). Other RiBi inhibitors had the same effect on SOX2 levels in NSC media, suggesting a common response following nucleolar stress; nevertheless, the compounds effectively reduced cell counts, suggesting that FGF2 and/or SOX2 upregulation may not be enough to bypass RiBi inhibition (Supplementary Figure S5I). In summary, glioma cells activate the FGF2-FGFR1 signaling axis to maintain SOX2 and promote the expression of stem cell-associated genes to overcome RiBi inhibition, underlining, at least in part, the mechanistic basis of the synergy observed.

Synergistic Induction of Cell Death in Pediatric Glioma Cell Lines

The functional importance of FGFR1 in glioma stem cells has been described recently.²⁰ FGFR1 is also a top-ranked vulnerability in pediatric gliomas of DIPG type²⁶ and is often mutated or amplified in pediatric high-grade gliomas.²⁷ To further validate and explore the translational potential of our findings, we tested BMH-21 combined with ERD in a panel of pediatric high-grade glioma cell lines, isolated and cultured in NSC. These cells have been thoroughly characterized and showed stem cell properties and tumorigenicity in mice^10,11 (Supplementary Figure S6A). Dose–response assays confirmed the potent growth inhibitory effect of BMH-21 in a 4-cell line panel, with a mean GI₅₀ of ~1.0 μM (Figure 6A). While the calculated GI₅₀s are higher than those in serum-cultured cells, and responses to BMH-21 monotherapy varied among the panel, we observed a clear, synergistic induction of apoptosis following combinatorial treatment in all lines tested (Figure 6B–C; Supplementary Figures S6B–C, S6E–H). Cell cycle analysis on 2 lines with differential responses to BMH-21 showed a dose-dependent G1 and G2 arrest of GU-pBT-28 cells, whereas, for GU-pBT-7, we observed an initial increase in the S-phase population (Figure 6D), but increasing concentrations eventually resulted in G1 arrest. BMH-21 induced nucleolar stress in both lines, as shown by NCL translocation and nucleolar cap formation (Figure 6E; Supplementary Figure S6D), RPA194 degradation, and RNA synthesis downregulation in GU-pBT-28. Interestingly, FGFR inhibition resulted in RPA194 upregulation, consistent with our findings on the established lines (Supplementary Figure S6O–P). Importantly, immunofluorescence confirmed FGF2 induction following BMH-21 among all lines (Figure 6G–H; Supplementary Figure S6I–J), FGF2 relocalization (Figure 6K; Supplementary Figure S6D), and SOX2 downregulation following the combinatorial treatment (Figure 6I–J; Supplementary Figure S6K–L). These findings underscore the BMH-21/ERD potential in perturbing RiBi and FGFR signaling, downregulating SOX2, and inducing cell death (Figure 6L).

Figure 6.

FGFR inhibition potentiates BMH-21 across a pediatric patient-derived cell line panel. (A) BMH-21 dose–response (72 h) in 4 cell lines derived from pediatric GBM tumors with corresponding GI₅₀ values. (B–C) Cas3/7-positive nuclei of GU-pBT-19 and GU-pBT-28 cells. (D) Cell cycle distribution barplot of GU-pBT-7 and GU-pBT-28 following Click-it-EdU-labeling. (E) Representative immunofluorescence images of GU-pBT-28 stained against RPA194&NCL, Scale bar 10 µm (F) RPA194 protein levels quantification by immunofluorescence and RNA synthesis quantified by Click-it-EU-labeling in GU-pBT-28 treated with BMH-21. (G–H) FGF2 and (I–J) SOX2 protein levels quantification by immunofluorescence in GU-pBT-19 and GU-pBT-28 cells. (K) Representative immunofluorescence images of GU-pBT-28 stained against FBL and FGF2, Scale bar 10 µm. The treatment duration was 24 h for (B-F, K) and 48 h for (G–J). (L) Proposed model of synergy: (1) BMH-21 blocks Pol I and downregulates FGFR1 and SOX2., (2) FGF2 is upregulated, released from the nucleolus and secreted, feeding the FGFR1/ERK1/2 feedback loop and rescuing stemness-associated genes to support self-renewal and survival under stress., and (3) Combination of Pol I Inhibition with FGFRi blocks the feedback loop, inhibits the expression of stem cell-associated phenotypes and synergistically induces cell death. Created with Biorender.

Discussion

The nucleolar size was associated with glioma malignancy 30 years ago,¹⁸ but the relative importance of RiBi in glioma biology remains elusive. Recent studies showed that GBM cells rely on enzymes responsible for the de novo nucleotide biosynthesis to maintain elevated rDNA transcription rates, such as IMPDH2²⁸ and DHODH.²⁹ A third study highlighted the role of the RiBi-associated protein WDR12 in GSC lines.³⁰ In line with this, we conducted analyses of gene expression data from TCGA and CGGA patient cohorts, where RiBi rate was shown to correlate positively with histological grades and worse clinical outcomes. We thus sought to assess the role of RiBi in glioma malignancy by using BMH-21, a small molecule inhibitor that targets the rate-limiting step of rDNA transcription. Both serum- or NSC-grown glioma cells of adult and pediatric origin were highly susceptible to BMH-21—treatment with BMH-21-induced cell cycle arrest, apoptosis, and sensitized cells to TMZ. Importantly, BMH-21 impaired the growth of transplanted glioma cells in zebrafish by 50%, while showing a favorable profile in the early pre-clinical evaluation performed. Zebrafish is a robust pre-clinical model with increasing popularity, as in contrast to the resourceful toxicological testing in rodents, it provides a fast and straightforward alternative platform to accelerate drug discovery and bridge in vitro and in vivo testing.^31,32

GBM is characterized by the inherent resistance to radiation and chemotherapy, a hallmark considered to be fueled by the existence of glioma CSCs.³³ The role of RiBi in the CSC population remains largely unexplored. Interestingly, using a high-throughput screening in patient-derived spheroids, ActD was found to be highly cytotoxic in most of the glioma stem-like cell lines tested and downregulate Oct4 and SOX2 expression.³⁴ Likewise, we show that BMH-21 downregulated SOX2 levels, even though a higher dose was required in NSC-grown cells compared with serum-based cultures, implying the involvement of SOX2-related mechanisms that could render the cells less susceptible to lower doses of BMH-21.

To identify drugs that potentiate RiBi blockade, we performed a synergy screen, which, apart from the synergy hits, revealed a set of highly cytotoxic monotherapy drugs worthy of further investigation. Several hits overlapped with those of a recent patient-derived cell-based screen, ³⁵ which revealed several protein synthesis-targeting compounds among the most cytotoxic hits (e.g., homoharringtonine), together with the RiBi-inhibitory compounds ActD and quinacrine. Further evidence supporting the potential of RiBi inhibition in HGG was provided in a study evaluating BBB-permeable 9-amino acridine (9-AA) analogs that effectively prolonged the survival of orthotopic GBM mouse models.³⁶ 9-AA is structurally related to BMH-21 and inhibits RiBi by targeting rRNA synthesis and processing.³⁷

The characterization of the synergies illuminated by our screen identified the FGF2-FGFR1 signaling axis as an important mediator of the observed cellular effects. This may constitute a novel response following nucleolar stress involving FGF2 translocation and upregulation. Earlier studies have shown FGF2 upregulation following radiation³⁸ or hypoxia,³⁹ while certain FGF2 isoforms have been reported to localize in the nucleolus and shown to directly regulate rDNA transcription.²¹ However, to the best of our knowledge, this is the first instance such effects are reported following nucleolar stress. The mitogenic effect of FGF2 and its multifaceted role in glioma malignancy are widely described,²⁵ including the maintenance of SOX2 levels, and promoting the expression of CSC-associated phenotype via FGFR1. This role could explain the maintained SOX2 levels observed under low BMH-21 doses in NSC-grown glioma cells and the potentiated cytotoxic effect observed by combining BMH-21 with FGFR inhibitors, which could prevent proneural to mesenchymal transition and resistance development.

Finally, our data illustrated the redundancy in RTK signaling and the clinical limitations of this therapeutic strategy related to resistance and tumor recurrence, as observed for different classes of RTK inhibitors.^2,20,23 In detail, we detected recurrence of FGFR and ERK1/2 phosphorylation and FGF2 levels 12–24 h following monotherapy with FGFRi, in parallel with upregulation of RPA194 protein levels, possibly mediating the positive regulation of Pol I transcription and RiBi in response to growth signals. Combining BMH-21, abrogated ERK1/2 phosphorylation and FGF2 upregulation for the treatment duration tested, underlining a potentially important role of RiBi to overcome RTK inhibition.

Our findings shed light on previously unknown cellular responses following RiBi inhibition and highlight its translational therapeutic potential in HGG and relevance as a combinatorial target with FGFRi. The clinical development of Pol I inhibitors with enhanced specificity and optimized pharmacokinetic properties currently in the pre-clinical stage is expected to enable the prospective clinical translation and provide more therapeutic options and, hopefully, prolonged survival rates for glioma patients.

Funding

This work was funded by the following grants: King Gustaf V:s Jubileumsfond (grant nos.164102 and 134082) to M.S.L.; the Cancerfonden to H.C.; the Swedish Cancer Society (no. 170176), and the Vetenskapsrådet (no. VR-MH 2014-46602-117891-30) to J.B.

Conflict of Interest

The authors declare no conflict of interest

Authorship Statement

Conceptualization:A.Z., M.S.L; Evolution of conceptualization:A.Z, M.S.L, D.K., J.B.; Methodology:A.Z., M.S.L., D.K., S.M., L.B.; Formal analysis:A.Z., D.K., S.M. Investigation: all authors; Resources:M.S.L., H.C., S.M., and J.B.; Data Curation:A.Z., D.K., S.M., I.K.; Visualization: A.Z., D.K., S.M.; Supervision:M.S.L, J.B.; Project administration:A.Z., M.S.L.; Funding Acquisition:M.S.L., J.B. All authors contributed to writing, have read and approved the manuscript.