Autophagy modulates glioblastoma cell sensitivity to Selinexor-mediated XPO1 inhibition

Abstract

Background

Selinexor is a selective inhibitor of exportin-1 (XPO1), a key mediator of the nucleocytoplasmic transport for molecules critical to tumor cell survival. Selinexor’s lethality is generally associated with the induction of apoptosis, and in some cases, with autophagy-induced apoptosis. We performed this study to determine Selinexor’s action in glioblastoma (GBM) cells, which are notoriously resistant to apoptosis.

Methods

Patient-derived GBM cells were treated with Selinexor, and drug response and autophagy levels were monitored. Homozygous C528S XPO1 mutant GBM43 cells were generated by CRISPR/Cas9 editing. Single Selinexor or combination treatment with autophagy inhibitors was evaluated. In addition, bulk-tissue, single-cell, and spatial transcriptome were analyzed, and molecular docking was performed.

Results

Although all cell lines exhibited a dose- and time-dependent reduction of cell viability, the most profound molecular response to Selinexor was induction of autophagy instead of apoptosis. Selinexor-induced autophagy was an on-target consequence of XPO1 inhibition, and could be mitigated by expression of a mutant, Selinexor-resistant form of XPO1, and Selinexor-induced autophagy was related at least in part to nuclear trapping of the transcription factor TFEB. Furthermore, genetic or pharmacologic suppression of autophagy sensitized the cells to Selinexor-induced toxicity in association with the induction of apoptosis. Finally, in intracranial PDX studies, the combination of Selinexor with the autophagy inhibitor chloroquine significantly impeded tumor growth and extended mouse survival relative to single-agent treatment.

Conclusion

These results suggest that activation of autophagy confers a protective mechanism against Selinexor in GBM cells, and that the combination of Selinexor with autophagy inhibitors may serve as a viable means to enhance Selinexor-induced cell death.

Graphical Abstract

Autophagy provides a defense mechanism against the lethality of Selinexor and the combination of Selinexor with inhibitors of autophagy may be a viable means to increase cell death in glioblastoma cells caused by Selinexor.

Open in new tabDownload slide

The proper functioning of human cells relies on the coordinated transport of molecules across the nuclear envelope, which is essential for maintaining cellular homeostasis. Small molecules can diffuse through nuclear pore complexes, whereas larger macromolecules are actively transported by karyopherin receptor transporter proteins.¹ One of the most important and well-studied karyopherins is exportin 1 (XPO1), also known as CRM1.² XPO1 mediates the nuclear export of various RNAs and more than 200 proteins.^3,4 Since several XPO1 cargo proteins, such as TERT, p53, and survivin, are crucial for regulating cell growth and survival,^3,5–7 XPO1 makes a significant contribution to normal cell homeostasis. XPO1, however, is often over-expressed in a variety of cancers,⁸ and its overexpression has been associated with histological grade, increased tumor size, and worse patient prognosis.^9–12 These observations stimulated the development of small molecules that inhibit XPO1 and may have applications in targeted cancer therapy. Of these, the selective inhibitors of nuclear export (SINE) class of XPO1 inhibitors have proven to be of particular interest.

SINE compounds, such as KPT-185, KPT-8602, and KPT-330 (Selinexor), were rationally designed based on knowledge of XPO1 function.^13–15 In the nucleus, XPO1 interaction with active GTP-bound RAN exposes a nuclear export sequence (NES) binding site in XPO1.¹⁶ This NES binding site in turn interacts with the NES domains of cargo proteins to form a complex, which subsequently associates with the nuclear pore.^17,18 After the RanGTP-XPO1-cargo complex passes through the nuclear pore, GTP hydrolysis leads to the release of the cargo protein into the cytoplasm and the relocation of XPO1 to the nucleus.¹⁹ SINE compounds were designed to bind covalently but reversibly to cysteine 528 in the NES binding site of XPO1. This interference disrupts cargo protein binding and nuclear export, ultimately impeding tumor cell growth.^20,21

SINE compounds have been extensively studied in preclinical settings, with Selinexor demonstrating notable single-agent efficacy in both hematopoietic and solid tumors.¹³ These promising results led to multiple clinical trials and, ultimately, FDA approval of Selinexor either alone or in combination with bortezomib and dexamethasone for relapsed multiple myeloma.^22–24 Clinical trials of Selinexor in solid tumors have also been initiated. In glioblastoma (GBM), the most common and lethal form of primary brain tumor in adults, Selinexor was well-tolerated and achieved clinically relevant intratumoral concentrations.^25,26

Although the binding of XPO1 to SINE compounds such as Selinexor is well defined, the mechanism by which this action leads to selective tumor cell death remains less clear. Given that XPO1 cargos include tumor suppressor proteins, the growth-suppressive effect of XPO1 inhibitors may be linked to the retention of specific proteins in the nucleus.^5,6 Alternatively, XPO1 inhibitor-mediated retention of cargos in general may be sufficient to disrupt cell homeostasis lethally.^15,27 Nonetheless, in most cases, the lethality of XPO1 inhibitors is associated with apoptosis, particularly in lymphoid malignancies.^28–30 Furthermore, several studies have suggested that agents that synergize with Selinexor, such as bcl-2 inhibitors or proteasome inhibitors, also increase apoptotic potential.^31,32 Therefore, Selinexor efficacy has been associated with the extent of drug-induced apoptosis.

At least two studies, however, have suggested part of the cellular response to Selinexor and Selinexor combinations is an induction of autophagy, which may also serve as a precursor to apoptosis.^29,33 Autophagy is a process by which cells initiate self-proteolysis by engulfing cellular constituents in macroautophagosomes in response to stress.^34,35 Autophagy is initiated by the activation of unc-51 like autophagy activating kinase 1 (ULK1) and the formation of phagophores,³⁶ a process typically regulated by mTOR. The maturation of the phagophore into the autophagosome is regulated by ATG7,³⁷ which recruits additional ATG proteins to the maturing phagosome, cleaves the LC3 precursor protein to generate cytosolic microtubule-associated protein light chain 3 LC3 (LC3B-I), and facilitates its conjugation with phosphatidylethanolamine to form LC3B-II. LC3B-II is required for autophagosome formation,³⁸ and its accumulation, along with that of autophagic vacuoles, are key indicators of autophagy. The fusion of newly formed autophagosomes with lysosomes leads to the degradation of autophagosomal contents and the release of essential nutrients, which provide an alternate energy source and support cell survival.³⁹ Given the relative resistance of GBM cells to apoptosis,^40–42 we investigated whether these cells initiate autophagy in response to Selinexor, and if so, whether autophagy contributes to or limits Selinexor-induced toxicity.

Our study reveals that in multiple GBM cell lines, Selinexor predominantly induces autophagy in a dose- and time-dependent manner. This effect is both an on-target effect of the drug and correlated with reduced cell viability. Furthermore, both genetic and pharmacologic suppression of autophagy reduced Selinexor IC50, and enhanced the effectiveness of Selinexor in vitro and in vivo. This suggests that, in GBM, disrupting the autophagy-mediated cellular defense could be an effective strategy to enhance Selinexor-induced cell death, and to achieve this at a lower, therapeutically achievable doses.

Materials and Methods

Cell Lines and Cell Culture

GBM cell lines LN229, T98, and SF188 were obtained from the University of California San Francisco (UCSF) Brain Tumor Research Center Tissue Core and Xiangya Hospital, Central South University. Patient-derived xenograft (PDX) GBM43 and GBM39 cells were provided by Jann Sarkaria, Mayo Clinic. Genetic characterization of these cell lines is detailed in Supplementary Table S1. Cell lines were maintained at 37°C in a humidified 5% CO₂ incubator in high glucose Dulbecco’s Modified Eagle Media supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin, and 1% fungizone. These cells were chosen to represent a cross-section of GBM cultures available and included tumor cells derived from both males and females, adult and pediatric patients, established lines and PDXs, cells with a range of c-myc expression, and both MGMT-proficient and MGMT-deficient cultures.^43,44

For experiments involving serum starvation, cell culture plates were rinsed twice with PBS and incubated for 4 h in HBSS media with Ca and Mg supplemented with 10 mM HEPES. All cell lines were validated for the absence of mycoplasma, and identities were confirmed by short tandem repeat analysis upon receipt and every 6 months thereafter (Supplementary Figure S7).

Generation of Homozygous C528S XPO1 Mutant GBM43 Cells by CRISPR/Cas9 Editing

GBM43 cells were co-transfected with plasmids expressing Cas9-2NLS and a single guide RNA targeting XPO1 (5ʹ- GGATTATGTGAACAGAAAAGAGG -3ʹ) using the CRISPR Gene Knockout Kit (SYNTHEGO). The transfection included a 135-base single-stranded oligodeoxynucleotide repair donor template containing the TGT to TCA mutation, along with 3 silent mutations (5ʹ-GCTAAATAAGTATTATGTTGTTACAATAAATAATACAAATTTGTCTTATTTACAGGATCTATTAGGATTATCAGAACAGAAgcGcGGCAAAGATAATAAAGCTATTATTGCATCAAATATCATGTACATAGTAGG-3ʹ (lowercase indicates additional silent mutations) obtained from Integrated DNA Technologies (IDT). Electroporation was performed using the Nucleofector-4D system and the SE Cell Line 4-D Nucleofector X kit (Lonza). The following day, Selinexor (1 μM) was added, and after sufficient growth, cells were trypsinized and distributed into 96-well plates to generate single-cell derived colonies. Colonies were then harvested, and genomic DNA was extracted for Sanger sequencing. Complete replacement of wildtype (WT) XPO1 with C528S mutant (mut) XPO1 was confirmed using Synthego’s Inference of CRISPR Edits (ICE) software tool (Synthego Performance Analysis, ICE Analysis, v3.0).

Immunofluorescence Analysis of Protein Sub-cellular Localization

Control and experimental cells growing in 4-well chambered slides were fixed in 4% paraformaldehyde (room temperature, RT, 10 min), rinsed 3 times in PBS, immersed in blocking buffer (5% normal goat serum and 0.1% Triton X-100 in PBS, 60 min, RT), then incubated with TFEB (CST #91767) (1:200) or RanBP1 (Abcam, ab97659) (1:200) primary antibodies in 2% goat serum and 0.1% Triton X-100 in PBS (16–18 h, 4 °C). After washing, slides were incubated with secondary antibody (Alexa Fluor 488 conjugate or Alexa Fluor 647 conjugate, 1:500, 2 h, Invitrogen) appropriate for the host species of the primary antibody. Slides were then washed (3 times in PBS), incubated with DAPI for 5 min, washed again, and mounted with VECTASHIELD Antifade Mounting Medium. Images were processed on a Nikon Eclipse Ti2 microscope. Nuclear (DAPI-stained) and cytoplasmic compartments were distinguished and manually outlined, after which the average pixel intensities of each were quantitated using Nikon NIS-Elements AR software.

In Vivo Studies

Immunodeficient mice (nu/nu; Charles River Laboratories) were injected intracranially with 3 × 10⁵ luciferase-expressing GBM39 human PDX cells or LN229 cells as described.⁴⁵ Seven days later the animals were randomized into 4 groups (5 animals per group) and treated twice weekly with vehicle (DMSO in PBS), Chloroquine (CQ) (50 mg/kg intraperitoneally in PBS), Selinexor (10 mg/kg orally in 2% DMSO plus 40% PEG300 and 5% Tween 80), or both CQ and Selinexor. Tumor growth was monitored twice a week by treating mice with d-luciferin (150 mg/kg ip; Gold Biotechnology) and measuring bioluminescence using a Xenogen IVIS Bioluminescence imaging station (Caliper). Tumor growth was calculated by normalizing luminescence measurements to day 6 post-injection values. Animals were euthanized when they reached the end point in the animal protocol, after which Kaplan–Meier survival curves were generated.

Bioinformatics Analysis

The detailed methods for bioinformatics analyses, such as bulk-tissue RNA-seq analysis, single-cell RNA sequencing (scRNA-seq) analysis, spatially resolved RNA sequencing (stRNA-seq analysis), and molecular docking analysis, can be found in the Supplementary Materials.

Ethics

The animal experiments in this study were approved by the UCSF’s Institutional Animal Care Use Committee (IACUC) (No. AN198176-00A) and the Ethics Committee of Xiangya Hospital (No. 202408158). The glioma sequencing datasets from the Cancer Genome Atlas (TCGA), Chinese Glioma Genome Atlas (CGGA), Tumor Immune Single-cell Hub 2 (TISCH2), and 10 × Genomics databases were publicly available, and therefore no approval from the local ethics committee was required.

Quantification and Statistical Analysis

Data were analyzed using Prism 9 software (GraphPad, La Jolla, CA, USA) and R 4.3.1 software (http://www.r-project.org) and are expressed as means ± standard deviations. Differences of continuous variables between two or more groups were estimated using the Student-t test or 1-way analysis of variance (ANOVA). Differences of categorial variables between two groups were evaluated using Fisher’s exact test. A P-value of < .05 was considered statistically significant.

Results

Selinexor’s Target XPO1 Has Glioma Clinical Significance

We first assessed the clinical significance of XPO1, the target of Selinexor, in gliomas. Data from GEPIA2.0 revealed that XPO1 was overexpressed in low-grade glioma (LGG) and GBM compared to the adjacent normal tissues (Figure 1A). Analysis of TCGA data showed that higher expression of XPO1 in glioma correlated with worse survival rates in patients (P < .01), across various survival metrics, including overall survival (OS), disease-specific survival (DSS), progression-free interval (PFI), and progression-free survival (PFS) (Figure 1B, Supplementary Figure S1A). Furthermore, XPO1 expression was positively associated with the degree of glioma grades, and significantly up-regulated in more aggressive gliomas, like 1q19q non-codeletion, classic (CL) subtype, and mesenchymal (MES) subtype in both TCGA and CGGA cohorts (P < .05) (Figure 1C, Supplementary Figure S1B).

We then localized XPO1 expression in GBM using the scRNA-seq data. Five GBM datasets from TISCH2 were integrated (Supplementary Figure S1C). XPO1 expression was dramatically increased in GBM cells relative to that in normal cells in the tumor microenvironment (Figure 1D). Spatial data from GBMseq indicated that XPO1 exhibited high expression in both the tumor core and periphery, with slightly higher expression noted in the periphery (Figure 1E). These data support the idea that XPO1 may be a clinically relevant target in GBM.

Glioblastoma Cells Exhibit the Selinexor Sensitivity

Subsequently, we explored the action of Selinexor in GBM cell lines. The interaction sites of molecular docking between multiple XPO1 inhibitors (Selinexor, KPT-185, and KPT-8602) and XPO1 are presented (Figure 1F, Supplementary Figure S1D). Five GBM cell lines of varying genetic composition were incubated, including PDX-derived GBM43 and GBM39 cells with Selinexor (1–10,000 nM, 72 h), after which viability was measured by Cell Titer Glo and used to generate Selinexor dose- and time-response curves. All cell lines exhibited loss of viability after Selinexor exposure, with IC50 values ranging from 85 to 550 nM (Figure 1G). The extent of loss of viability was also time-dependent as increased duration to an IC50 Selinexor exposure, as determined in Figure 1G, also led to increased loss of viability in all cells examined (Figure 1H). A comparable dose- and time-dependent loss of viability was also noted using two other structurally related XPO1 inhibitors (KPT-185 and KPT-8602) (Supplementary Figure S1E–S1G).

Selinexor Induces a Dose-and Time-dependent Autophagy in Glioblastoma Cells

To investigate the cellular dysfunction resulting from XPO1 inhibition in Selinexor-treated GBM cells, we first contrasted XPO1^high and XPO1^low GBM cells using the scRNA-seq data from TISCH2. Nine cell types were collated and annotated, including GBM cells, CD8 T, astrocytes, and others (Supplementary Figure S2A). XPO1^high and XPO1^low GBM cells were divided according to the 50% median of XPO1 expression (Figure 2A). Next, the differential analysis identified 656 feature genes in XPO1^high GBM cells (log2FC > 0.5 and FDR < 0.05) and 246 feature genes in XPO1^low GBM cells (log2FC < -0.5 and FDR < 0.05) (Supplementary Figure S2B). The feature genes of the XPO1^low GBM cells showed enrichment in apoptosis and autophagy (P < .05, Figure 2B), the latter of which has been reported to involve drug resistance and cause treatment failure in multiple cancers.⁴⁶

Next, we examined the correlation between autophagy and XPO1 by analyzing scRNA-seq data from TISCH2 and stRNA-seq data from 10 × Genomics. The autophagy level was quantified by analyzing a gene set compiled from Zou et al.‘s research,⁴⁷ using the algorithm “singscore.”⁴⁸ scRNA analysis revealed a significant negative correlation between autophagy level and XPO1 expression (r = −0.57 and P < .001) (Supplementary Figure S2C). Analysis of stRNA also showed a highly consistent spatial distribution between autophagy^high GBM cells and XPO1^low GBM cells (r = 0.79 and P < .001) (Supplementary Figure S2D–S2F). In addition, MES-like subtype was previously reported to be responsible for therapy resistance in GBM.⁴⁹ Our stRNA analyses found that autophagy^high GBM cells and XPO1^low GBM cells manifested higher expression levels in the MES-like GBM cell enriched region, when compared separately to autophagy^low GBM cells and XPO1^high GBM cells (Supplementary Figure S2E and S2G). These results suggest that XPO1 inhibition is associated with autophagy increase, indicating a potential relationship between Selinexor treatment and autophagy in GBM cells.

Subsequently, to verify whether autophagy occurs in Selinexor-treated GBM cells and whether this contributes to, or limits, Selinexor-induced toxicity, we exposed GBM cells to a range of Selinexor concentrations for 48 h, or a fixed dose of Selinexor (250 nM in LN229 and SF188 cells, 750 nM in GBM43, T98G, and GBM39 cells) for varying times (0–72 h), after which effects on measures of autophagy were determined. Western blot analysis showed that all cells exhibited a dose-dependent increase in the ratio of the PE-conjugated form of LC3B (LC3B-II) to LC3B-I (Figure 2C, Supplementary Figure S3A, S3B, and S8A), and a significant increase in the LC3B-II/LC3B-I ratio in response to increasing duration of Selinexor exposure (Figure 2D, Supplementary Figure S3C, 3D, and S8B). Consistent with these findings, LN229 and SF188 cells also exhibited a dose- and time-dependent decrease in levels of p-mTOR, and a dose- and time-dependent increase in levels of pAMPKα (Figure 2E and 2F, Supplementary Figure S3E, S3F, and S9), both consistent with changes noted in cells undergoing autophagy.⁵⁰

Finally, to verify that Selinexor exposure induced changes in autophagic flux, cells were stained with a cationic amphiphilic tracer dye (Abcam,⁵¹) that selectively labels autophagic vacuoles, after which dye accumulation was measured by immunofluorescence in control and Selinexor-treated cells. Although autophagic vacuoles were rare in control cells, exposure to Selinexor significantly increased the total area of autophagic vacuoles in all cells (Figure 2G, Supplementary Figure S3G and S10A). Exposure of the same cells to the autophagy inhibitor CQ, which blocks autophagosome-lysosome fusion, similarly increased the total area of autophagic vacuoles, and combined exposure of cells to both Selinexor and CQ further increased the total area of autophagic vacuoles (Figure 2G, Supplementary Figure S3G and S10A). Increases in the LC3B-II/LC3B-I ratio were also noted using a second inhibitor of autophagosome-lysosome fusion Bafilomycin A (Figure 2H, Supplementary Figure S3H, S3I, and S10B), additionally suggesting that Selinexor triggers an increase in LC3B-II and autophagic flux by inducing autophagy.

Selinexor-induced Autophagy Is an on-target Consequence of XPO1 Inhibition

To directly address whether the autophagic response noted in cells exposed to Selinexor was a direct result of XPO1 inhibition, we used 3 approaches. First, we showed that the increased autophagic flux induced by Selinexor was not unique to this drug but was also noted following exposure of GBM43 cells to the structurally-related XPO1 inhibitors KPT-185 and KPT-8602 (Supplementary Figure S10C). Second, we created GBM cells expressing only a mutant, yet functional, form of XPO1 that was completely insensitive to inhibition by Selinexor, and then used these cells to determine whether the effects of Selinexor on viability and autophagy were purely on-target effects of the drug. To create these cells, a CRISPR-based knock-in strategy was used to replace both copies of the DNA encoding WT XPO1 in GBM43 cells with DNA encoding an XPO1 protein that, by virtue of a C528S mutation, could not bind Selinexor but could still function as a nuclear exporter (Figure 3A and Supplementary Figure S4A).⁵² Molecular docking studies showed the higher binding energy of mut XPO1 to Selinexor than WT XPO1, which confirmed the insensitivity of mut XPO1 to Selinexor (Figure 3B). Sequencing of clonal populations of CRISPR-modified GBM43 cells showed complete replacement of the WT XPO1 with C528S XPO1 (Supplementary Figure S4A). Although these CRISPR-modified GBM43 cells (here designated GBM43 mut) exhibited viability and growth similar to WT cells, they failed to undergo nuclear sequestration of the XPO1 substrate RanBP1⁵² in response to Selinexor exposure (Figure 3C), exhibited a Selinexor IC50 roughly 100× greater than that in GBM43 cells expressing endogenous WT XPO1 (Figure 3D), and were completely resistant to Selinexor in clonogenicity assays (Figure 3E, Supplementary Figure S4B). Furthermore, while exposure of GBM43 or GBM43 mut cells to the autophagy activator rapamycin significantly increased the LC3B-II/LC3B-I ratio, exposure of the cells to Selinexor resulted in an increase in the LC3B-II/LC3B-I ratio in GBM43 cells, but not GBM43 mut cells (Figure 3F, Supplementary Figure S4C and S11A). As such, the effects of Selinexor on autophagy are not an off-target effect but rather a direct effect of the drug on XPO1. Finally, to further verify this point, GBM43 and GBM43mut cells were incubated with a pooled siRNA targeting XPO1, and analyzed for effects on LC3B-II/LC3B-I ratio and cell viability. As predicted, suppression of XPO1 levels led to an increase in LCB3-II/LC3B-I ratio in both cell groups (Figure 3F, Supplementary Figure S4C and S11A), consistent with the onset of autophagy. Furthermore, siRNA-mediated suppression of XPO1 also led to a loss of cell viability regardless of whether the cells expressed WT or Selinexor-resistant XPO1 (Figure 3G). Together, these findings show that Selinexor induces autophagy in GBM cells by on-target inhibition of XPO1.

Selinexor-induced Nuclear Trapping of TFEB Contributes to Autophagy

As an inhibitor of XPO1 and nuclear export,⁵³ Selinexor traps a variety of proteins in the nucleus, including multiple transcription factors.²³ Accordingly, we asked whether the nuclear trapping of transcription factors was involved in the autophagy induced by XPO1 inhibition with Selinexor. Considering the activation of most transcription factors correlates to their translocation from the cytoplasm into the nucleus,⁵⁴ we first conducted SCENIC analysis on scRNA data to estimate the activity scores of autophagy transcription factors, collected from previous literature.⁵⁵ Analysis showed that activity scores of TFEB, FOXO3, and PPARA were significantly higher in XPO1^low GBM cells compared to XPO1^high GBM cells (P < .05, Supplementary Figure S4D). Among the 6 autophagy transcription factors analyzed, TFEB activity exhibited the most robust negative correlation with XPO1 expression in GBM cells (r = −0.20 and P < .01) (Figure 4A), indicating that TFEB was the most probable transcription factor to regulate XPO1 inhibition (Selinexor) induced autophagy in GBM cells.

To determine if nuclear trapping of TFEB contributed to Selinexor-induced autophagy in GBM, sub-cellular localization of TFEB was assessed by immunofluorescence in control and Selinexor-treated cells, and compared to the extent of autophagy. In control GBM43 and LN229 cells, TFEB was distributed throughout both the nucleus and the cytoplasm, as previously reported,⁵⁶ serum-starvation of these cells resulted in a significant sequestration of TFEB in the nucleus (Figure 4B and 4C, Supplementary Figure S11B). Notably, a similar increase in the nuclear/cytoplasmic ratio of TFEB was noted as early as 24 h after exposure of the cells to Selinexor (Figure 4B and 4C, Supplementary Figure S11B). To determine if there was a causal link between Selinexor-induced nuclear TFEB sequestration and induction of autophagy, cells were incubated with a pooled siRNA targeting TFEB, after which the effect of TFEB depletion on Selinexor-induced autophagy was determined. siRNA-mediated suppression of TFEB (Figure 4D) as expected decreased the ratio of LC3B-II/LC3B-I in all cells examined (Figure 4E, Supplementary Figure S4E and S11C), consistent with the role of TFEB in the induction of autophagy. Q-PCR analysis further showed that while an exposure of Selinexor shown to induce autophagy significantly increased levels of a variety of transcripts encoding proteins implicated in autophagy, this increase could be reversed by siRNA-mediated suppression of TFEB (Figure 4D). siRNA-mediated suppression of TFEB also lessened the Selinexor-induced increase in LC3B- II/I ratio (Figure 4E, Supplementary Figure S4E and S11C), and was associated with a small but statistically significant increase in Selinexor-induced toxicity (Figure 4F). Moreover, we performed the TFEB-binding CUT&RUN assay followed by next-generation sequencing in LN229 cells treated with Selinexor or DMSO controls. The CUT&RUN data showed that most TFEB binding occurred in the promoter (39.9%) (Supplementary Figure S4F). Metaplots and heatmaps depicted the changes in TFEB binding sites following Selinexor treatment (Supplementary Figure S4G), with the up-regulated binding sites showing significant enrichment of genes related to autophagy (Supplementary Figure S4H). These data therefore suggest that Selinexor-induced XPO1 inhibition is linked to autophagy at least in part by nuclear trapping of TFEB.

Autophagy Protects Glioblastoma Cells from Selinexor-induced Cytotoxicity

Autophagy has been reported to contribute to, or limit, cell survival,³⁹ and even after exposure to Selinexor, the consequences of autophagy appear to be situationally dependent.²⁹ To determine if Selinexor-induced autophagy contributes to, or limits, survival of GBM cells, we used both genetic and pharmacologic approaches. First, to assess the effect of genetic suppression of autophagy on Selinexor-induced toxicity, cells were incubated with a pooled siRNA targeting ATG7, a pro-autophagy protein, after which they were exposed to Selinexor and assayed for viability. Suppression of ATG7 levels alone decreased the LC3B-II/LC3B-1 ratio and also prevented the increase in LC3B-II/LC3B-1 ratio noted after Selinexor exposure (Figure 5A, Supplementary Figure S5A and S12A). The combined suppression of ATG7 with Selinexor also significantly reduced cell viability more than either treatment alone (Supplementary Figure S5B).

Next, to determine if pharmacologic inhibition of autophagy could similarly sensitize cells to Selinexor-induced toxicity, cells were incubated with Selinexor, with concentrations of CQ shown to inhibit autophagy (Figure 2G,⁵⁷), or both, after which effects on cell viability were determined. Although continuous exposure to Selinexor resulted in the expected loss of viability in all cell lines, the additional exposure of cells to 1.5, 3, or 6 µM CQ resulted in a synergistic loss of cell viability (as determined by HSA Synergy Finder analysis) specifically across the 2–4 log Selinexor exposure range that induced the largest loss of viability alone (Figure 5B—5D, Supplementary Figure S5C and S5D). Similar results were noted in colony formation assays in which the combined exposure of cells to Selinexor and CQ induced a greater loss of clonogenicity than either drug alone (Figure 5E, Supplementary Figure S5E). These results show that in GBM cells, autophagy provides a defense mechanism against the lethality of Selinexor, and that cells can be sensitized by genetic or pharmacologic suppression of autophagy.

Cell death following Selinexor exposure appears to occur through various pathways, the most prominent of which is apoptosis in lymphoid cells.^29,30 In some GBM cells, we noted that Selinexor exposure also triggered an apoptotic response, as indicated by the appearance of cells exhibiting annexin V staining, although this response was highly variable (Figure 5F, Supplementary Figure S5F). Although exposure of cells to CQ alone had little effect on levels of apoptosis, and in most cases reduced levels of apoptosis, the combination of CQ and Selinexor led to significantly higher levels of apoptosis than either drug alone (Figure 5F and 5G, Supplementary Figure S5F). Similar results were also noted in the combination of ATG7 suppression and Selinexor, which exhibited markedly elevated levels of apoptosis compared to the individual treatments (Supplementary Figure S5G and S5H). These results suggest that in GBM cells the process of autophagy limits the induction of apoptosis, and in doing so provides a degree of protection from Selinexor-induced toxicity.

Autophagy Increases the Resistance of Glioblastoma Cells to Selinexor

Selinexor-induced cytotoxicity reflects the cellular sensitivity to Selinexor.⁵⁸ To verify the impact of autophagy on GBM cell sensitivity to Selinexor, we assessed the Selinexor therapy responses based on the gene expression profiles of scRNA-seq datasets from TISCH2 and drug sensitivity profiles from PRISM. Selinexor-sensitive and Selinexor-resistant GBM cells were divided according to the 50% median value of AUC (Figure 5H). Differential gene analysis identified 1803 feature genes in Selinexor-sensitive GBM cells (log2FC < −0.2 and P < .05) and 3001 feature genes, which encompassed autophagy activation genes like ATG14, ATG10, OPTN, and ULK1, in Selinexor-resistant GBM cells (log2FC > 0.2 and P < .05) (Supplementary Figure S13A). These differentially expressed genes exhibited enrichment in autophagy regulation (P < .05) (Supplementary Figure S13B). Furthermore, gene set enrichment analysis (GSEA) revealed that genes of “positive regulation of autophagy (GO:0010508)” were enriched in Selinexor-resist GBM cells but not in Selinexor-sensitive GBM cells (NES = 0.412 and P = .047) (Supplementary Figure S5I). Gene Set Variation Analysis analysis showed that higher AUC values of Selinexor, indicating increased resistance to Selinexor, were observed in high autophagy GBM cells compared to those with low autophagy levels (P < .001, Supplementary Figure S5J). Autophagy levels were quantified using the gene set compiled from the study by Zou et al.⁴⁷ These findings collectively suggest a positive correlation between Selinexor resistance and autophagy.

Next, we used Beyondcell,⁵⁹ an algorithm targeting cancer therapeutic heterogeneity, to investigate the therapeutic response of Selinexor-resistant or Selinexor-sensitive GBM cells to the autophagy inhibitor CQ. In the scRNA data, Selinexor-resistant GBM cells exhibited increased sensitivity to CQ (Figure 5I), and in the stRNA data, Selinexor-resistant GBM cells showed a similar spatial distribution to CQ-sensitive GBM cells (Supplementary Figure S5K), which were divided according to the 50% median value of bcscore. Specifically, both Selinexor-resistant GBM cells and CQ-sensitive GBM cells manifested elevated expression levels in the MES-like GBM cell enriched region that was previously identified as being associated with therapy resistance in GBM⁴⁹ (Supplementary Figure S5K and S5L). Therefore, these results suggest that Selinexor resistance may at least partly be attributed to autophagy, and that CQ-mediated inhibition of autophagy may be a promising means to increase Selinexor sensitivity in otherwise resistant GBM cells.

Combining Chloroquine with Selinexor Increases Apoptosis, Improves the Survival of GBM-bearing Mice, and Improves the Predicted Prognosis of Glioma Patients

To examine the effects of the Selinexor/CQ combination in vivo, mice bearing established GBM39 and LN229 intracranially PDX tumors were separately randomized into 4 groups and treated with vehicle, Selinexor, CQ or the Selinexor/CQ combination (Figure 6A, and Supplementary Figure S6A). Both GBM39 and LN229 tumors in the vehicle-treated group grew rapidly, separately necessitating the sacrifice of half of the animals at roughly 26 days and 32 days post-implantation (Figure 6B–D, and Supplementary Figure S6B–S6D). GBM39 and LN229 tumors in the groups treated with either Selinexor or CQ grew at a rate that was not significantly different from vehicle groups, and animal survival was also not significantly extended. GBM39 tumors and LN229 in the Selinexor plus CQ groups, however, grew significantly slower than those in single agent groups, and Selinexor plus CQ exposure significantly extended the survival of these animals relative to single agent and control groups (Figure 6B–C, and Supplementary Figure S6B–S6C). Of note, fixed sections of GBM39 tumors from sacrificed animals treated with Selinexor plus CQ exhibited significant decreases in Ki67 staining and significant increases in cleaved caspase 3 staining relative to either control or single agent groups (Figure 6E), consistent with the ability of the Selinexor/CQ combination to slow growth and induce apoptosis in these cells.

Currently, there are no clinical trials on Selinexor/CQ combination. To preliminarily explore the treatment effect of Selinexor/CQ combination in glioma patients, we used open-source scRNA-seq data from TISCH2 and bulk-tissue RNA-seq data from TCGA and CGGA. Deconvolution was performed to map the Selinexor and CQ sensitivity from scRNA-seq data to bulk RNA-seq data. Patients were divided into high and low Selinexor sensitivity groups based on the 50% median of the Selinexor sensitivity index. Covariates were balanced between the two groups in both the TCGA and CGGA datasets through propensity score matching (PSM) (Supplementary Tables S2 and S3), with matching parameters including age, sex, tumor grade, IDH mutation, 1p19q co-deletion, and MGMT methylation. Analysis of TCGA and CGGA data after PSM showed that glioma patients with low Selinexor sensitivity exhibited significantly higher CQ sensitivity compared to those with high Selinexor sensitivity (P < .001 in both TCGA and CGGA) and a strong negative correlation between Selinexor sensitivity and CQ sensitivity (r = −0.81 and P < .001 in TCGA; r = −0.95 and P < .001 in CGGA) (Figure 6F and 6G), indicating that glioma patients resistant to Selinexor might show sensitivity to CQ. Furthermore, 5 machine learning algorithms, including StepCox, RSF, CoxBoost, GBM, and LASSO, were combined using 10-fold cross-validation to assess the differences in OS prediction when considering Selinexor/CQ sensitivity combination and single drug sensitivity in the TCGA training cohort and CGGA validation cohort. The results indicated that the combination of Selinexor and CQ sensitivities exhibited improved performance in predicting the OS of glioma patients, with the C-index increasing from 0.700 to 0.761, in comparison to single drug sensitivity, whose C-index ranged from 0.583 to 0.673 (Figure 6H). These data, therefore, indicate that the Selinexor/CQ combination may be a promising therapeutic strategy for glioma patients.

Discussion

The results of the present study show that in response to Selinexor exposure, GBM cells undergo autophagy. This autophagy is not a precursor to cell death but rather a pro-survival event that also appears to limit Selinexor-induced apoptosis.

In the present study, GBM cells exhibited variable but low levels of apoptosis, but consistently elevated levels of autophagy in response to Selinexor. The low and variable levels of apoptosis appear consistent with the general resistance of GBM cells to multiple apoptotic stimuli⁴⁰ as well as the fact that 85% of GBM cells (including all those used here) have defects in p53 signaling.⁶⁰ In contrast, all cell lines examined underwent autophagy after exposure to Selinexor. This Selinexor-induced autophagic response was comparable to that noted following siRNA-mediated suppression of XPO1, and was a direct on-target effect of Selinexor as autophagy could be induced by other selective XPO1 inhibitors and could be completely reversed by providing cells with a form of mutant XPO1 that could not bind or be inhibited by Selinexor. Our results also suggest that at least part of the ability of Selinexor to induce autophagy is related to nuclear trapping of the transcription factor TFEB, a process which has previously been shown to be associated with the induction of autophagy.⁵⁶ This nuclear trapping of TFEB could also be related to the Selinexor-induced decrease in p-mTOR levels noted in the study as phosphorylation of TFEB by mTOR contributes to TFEB nuclear retention,⁵⁶ although this was not directly examined. Near complete suppression of TFEB levels, however, did not completely reverse Selinexor-induced autophagy, perhaps consistent with the fact that TFEB is only one of a member the MiT/TFE family of transcription factors,⁶¹ all of which could regulate autophagic potential. Alternatively, Selinexor also likely generates a generalized stress response that can also trigger autophagy independently of effects on TFEB.

The Selinexor-induced autophagic response noted in this study also appeared to limit apoptosis rather than being a precursor to apoptotic cell death. All cells exhibited a significant increase in Selinexor-induced toxicity when concomitantly exposed to CQ. Furthermore, although CQ is a relatively non-specific inhibitor of autophagy,⁵⁷ a similar sensitization was noted following siRNA-mediated suppression of the pro-autophagic molecule ATG7. A direct link has been noted between the induction of autophagy and apoptosis in gallbladder cancer cells and myeloma cells,^29,33 which may relate to the higher apoptotic potential and intact p53 status of both cell types. The inverse linkage noted here, however, is similar to that noted in lung and pancreatic cells in which autophagy provides protection from EGFR- and ERK- inhibitor toxicity, respectively.^62,63 A similar inverse linkage was also noted in sub-groups of breast cancer cells with a baseline dependence on autophagy for survival,⁶⁴ although it should be noted that the GBM cells in this study were minimally effected by suppression of ATG7 and did not appear to exhibit a baseline dependence on autophagy for survival. The exact means by which suppression of autophagy enhances apoptosis in otherwise apoptotic-resistant GBM cells, however, remains unclear, although many proteins that regulate apoptosis are XPO1 substrates and could be impacted. It also remains to be determined if the autophagic responses of cells to Selinexor limit apoptotic cell death in cells other than the GBM cells analyzed here.

Finally, the studies presented here suggest that GBM cells reponed to Selinexor by inducing a self-protective autophagy, and that inhibition of this autophagic response may in turn enhance apoptotic cell death and XPO1 inhibitor efficacy. While the idea of combing Selinexor with agents that enhance apoptosis is not new, the possibility of inhibiting autophagy to do so is novel. Furthermore, because the autophagy inhibitor CQ is well-studied and has a low toxicity profile in humans, combining CQ with XPO1 inhibitors such as Selinexor may warrant further clinical investigation.

Supplementary material

Supplementary material is available online at Neuro-Oncology (https://dbpia.nl.go.kr/neuro-oncology).

Acknowledgments

We express our gratitude to Nick Stevers and Chibo Hong from the Costello lab for their technical assistance. We also acknowledge the support received from: (1) Professor Zhixiong Liu and all colleagues at the Neurosurgery Department of Xiangya Hospital, (2) Professor Tao Song and all colleagues at Xiangya Jiangxi Hospital, (3) Professor Pan Chen of the Experimental Animal Department of Hunan Cancer Hospital, and (4) the Glioma Clinical Diagnosis and Treatment Center at Xiangya Hospital.

Conflict of interest statement

The authors have no conflicts of interest to declare.

Authorship Statement

Conceptualization: Yongjian Tang, Russell O Pieper, Bo Chen and Liyang Zhang. Methodology: Yongjian Tang, Lisa Sprinzen, Yukinori Terada, Bo Chen and Liyang Zhang. Formal Analysis: Yongjian Tang, Lisa Sprinzen, Yukinori Terada, Bo Chen. Original Draft Preparation: Yongjian Tang, Jianzhong Zhang, Russell O Pieper, Bo Chen, Liyang Zhang. Writing—Review & Editing: Yongjian Tang, Lisa Sprinzen, Yukinori Terada, Karrie M. Kiang, Chuntao Li, Fangkun Liu, Hongshu Zhou, Xisong Liang, Jianzhong Zhang, Russell O Pieper, Bo Chen, and Liyang Zhang. Supervision: Liyang Zhang, Bo Chen, Russell O Pieper, Jianzhong Zhang. Project Administration: Liyang Zhang, Bo Chen, Russell O Pieper.

Funding

Russell O Pieper, Liyang Zhang, Chuntao Li, Karrie M. Kiang, and Jianzhong Zhang. The Panattoni Foundation and the loglio Project (to Russell Pieper); The Natural Science Foundation of Hunan Province (Nos. 2023JJ0897 to Chuntao Li; 2023JJ30972 to Liyang Zhang); The National Natural Science Foundation of China (Nos. 82403157 to Karrie M. Kiang); The Jiangxi Provincial Health Technology Project (Nos.202310113 to Jianzhong Zhang); The Jiangxi Provincial Natural Science Foundation (Nos. 20242BAB25471 to Jianzhong Zhang).

Data Availability

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.

References