The U251 TMZ-resistant model was established as previously reported (23). Briefly, mice with flank U251 tumors were treated with escalating oral doses of TMZ starting with 20mg/kg/day for 3 days, and after the initial tumor regrowth, mice were treated with 66mg/kg/day for another 3 days. The growth of the resulting resistant tumors was not inhibited by the highest TMZ concentration tolerable by mice, which was 120mg/kg/day for 5 days. TMZ-resistant flank tumors (hereafter named U251TMZ) were minced to a single-cell suspension and cultured as monolayers in vitro.

Primary patient-derived xenograft (PDX) (GBM6, GBM43, GBM14 and GBM22) and the established GBM cells (U251, U251TMZ, and T98G) were cultured as previously described (24).

The TMZ-resistant U251 (U251TMZ) cells were used for the shRNA screening of genome-wide modulators of TMZ resistance in MGMT promoter hypermethylated GBM cells. All the next steps were performed as we previously reported (11, 12). Briefly, cells were transduced with a pooled lentiviral shRNA (kindly provided by Dr. Yuichi Machida, Department of Oncology Research, Mayo Clinic, Rochester, MN) at a multiplicity of infection (MOI) = 1. This library targets about 38,000 genes or 47,000 mRNAs, and each shRNA is barcoded with the corresponding gene probe sequence used on the U133 plus 2.0 array (Affymetrix, Santa Clara, CA). After a brief selection in puromycin, cells were divided into 2 groups (each in triplicate). One group was treated with 100 µM TMZ while the other (control) group received the vehicle DMSO and was then allowed to grow for 14 days. Total RNA was extracted from control and cells survived TMZ treatment, followed by shRNA amplification according to a protocol supplied by System Biosciences (SBI, Mountain View, CA). The amplified shRNAs were hybridized onto the U133 plus 2.0 array (Affymetrix, Santa Clara, CA). The shRNAs with a significantly higher enrichment in the DMSO group in comparison to the TMZ group were considered to represent genes that negatively control TMZ sensitivity, and shRNAs preferentially enriched within the TMZ-treated group were considered positive modulators of TMZ sensitivity.

Western blotting was performed according to the previously reported protocol (23). The primary antibodies included XPO1 (Exporitin-1/CRM1 (D6V7N) rabbit mAb, Cell signaling cat# 46249S), phospho-CREB^S133 (87G3) rabbit mAb (Cell Signaling Cat# 9198S), CREB (48H2) rabbit mAb (Cell Signaling Cat# 9197S), anti-vinculin (E1E9V) XP(R) rabbit mAb (Cell Signaling cat# 13901S), beta-actin rabbit Ab (Cell Signaling Cat# 4967S), and secondary antibody was anti-Rabbit IgG, HRP-linked antibody (Cell signaling Cat. # 7074S).

Primary and established cells were plated in 96-well plates (primary, 2000/well; established, 500/well) and maintained at standard humidified culture conditions (37°C and 5% CO₂) overnight. Then, cells were exposed to graded concentrations of Selinexor, TMZ, and combined Selinexor/TMZ. The control cells were treated with DMSO. The growth of the control and treated cells was live monitored using an Incucyte SX5 machine (Sartorius Inc., Ann Arbor, MI) or by the Cyquant Cell proliferation assay (Thermo Fisher Scientific, Cat# C7026). All experiments were conducted in triplicate and replicated three times.

Lentiviral shRNA constructs were used for gene knockdown. Briefly, the pLKO shRNA constructs were purchased from Millipore-Sigma (Danvers, MA). Lentiviral particles were packaged using the 293T cells. For this, the 293T cells were plated into 100mm culture dishes to be about 70-80% confluent the next day. After overnight attachment, cells were changed to antibiotic-free media, and the co-transfection of plasmids was done by diluting 54 μL of FuGENE6 (Roche) in 2 mL of serum-free DMEM and incubating for 5 minutes at room temperature. Then, Gag-Pol (3ug), VSG (3ug), and pLKO shRNA (4.5ug) plasmids were added, mixed by flicking, and incubated for 20 minutes at RT, and then added into cells. The day after the infection, 10 mL of antibiotic-free fresh media was added to the dishes, and 2 days later, the media were collected, briefly centrifuged to collect the cell debris, filtered through 22-um filters, aliquoted, and stored at -80°C until use. For MGMT re-expression experiments, MGMT cDNA was cloned into a GIPZ lentiviral plasmid (Horizon Discovery, Lafayette, Colorado) to replace the GFP-cDNA. Both empty vector (GIPZ-GFP) and GIPZ-MGMT plasmids were packaged using the above-described protocol for shRNA constructs. The packaging and handling of the lentiviral particles were done following the University of Minnesota Biosafety Committee guidelines.

U251 cells expressing empty vector (U251V) and MGMT cDNA (U251 MGMT) were grown overnight on coverslips. Cells were treated with TMZ or vehicle DMSO and fixed with 4% paraformaldehyde at 0 and 72-hour time points. Irradiated cells (2-Gy) were used as a positive control for γ-H2AX. The immunofluorescence staining was conducted as previously described (11). The primary antibody used was P-Histone H2A.X (S139)(20E3) rabbit mAb (Cell Signaling Cat# 9718S), and the secondary antibody was Alexa Fluor™ 594 goat anti-rabbit IgG (H+L) (Invitrogen Cat# A11012). The staining was analyzed with a confocal microscope (Zeiss LSM 900 with Airyscan detector, Zeiss, Germany). For γ-H2AX foci quantification, at least 200 cells with ≥ 25 foci/nuclei were analyzed for each condition.

Chromatin immunoprecipitation (ChIP) was performed using the MAGNA-ChIP™ kit, Cat. # 17-10085(Millipore, Billerica, MA). Cells were crosslinked with 1% formaldehyde, and after quenching the formaldehyde reaction with 0.1 M glycine, the cells were processed according to the protocol supplied by the vendor (Millipore). Antibodies used were anti-acetyl-lysine 27 histone H3 (H3K27ac rabbit monoclonal antibody, diagenode Cat# C15210016) and anti-trimethyl-lysine 4 histone H3 (H3K4-me3,1, Epigentek Cat# P-2028-24). The distal promoter region that is critical for MGMT silencing by hypermethylation (25) was targeted by a quantitative PCR using human-specific primer sequences: 5′-GCCCCGGATATGCTGGGAC-3′ (forward) and 5′-GGGCAACACCTGGGAGGCAC-3′ (reverse). ChIP enrichment was evaluated relative to the input chromatin.

RNeasy Kit was used for total RNA extraction from GBM cells according to the protocol supplied by the vendor (Qiagen Inc., Germantown, MD). The reverse transcriptase was performed as previously reported (23, 25). The following primers were used to amplify MGMT 5’ TCT TCA CCA TCC CGT TTT CC-3’(forward) and 5’-CCG AAT TTC ACA ACC TTC AGC-3’ (reverse); XPO1 5’-CTA CAT CTG CCT CGT TGCT-3’ (forward) and 5’-CCA ATA CTT CCT CTG GTT TAG CC-3’ (reverse); GAPDH 5’-CTC TGC TCC TGT TCG AC-3’ (forward) and 5’-GCC CAA TAC GAC CAA ATC C-3’ (reverse). The PCR mix was prepared using PowerUP™ SYBR™ Green Master Mix (Applied Biosystems Cat# A25742), and the PCR reaction was performed at 95°C for 10 minutes, followed by 40 cycles of 95°C denaturing, 60°C annealing, 72°C extension, and final extension at 72°C for 10 minutes.

Linear mixed models were used to compare the rate of % confluence changes over time between treatment groups, followed by pairwise comparisons using the Tukey method. The differences in percent confluence in Selinexor with and without TMZ and the ChIP enrichment were analyzed using a two-sample t-test. The p-values <0.05 were considered statistically significant.

We developed a resistant model of U251 cells to study the mechanisms modulating TMZ resistance in MGMT-promoter hypermethylated GBM cells. Unlike previous models created by treating cells with TMZ in vitro, our U251-resistant model was generated in vivo by treating mice harboring flank U251 xenografts with escalating doses of TMZ until reaching the maximum dose tolerated by mice, but without any further slowing down the tumor growth. As shown in Figure 1A , cells cultured from these flank tumors (referred to hereafter as U251TMZ) were significantly resistant to TMZ compared with the parental U251 cells (relative fluorescence (30 µM TMZ) = 1803 ± 91.00 (parental) vs. 6239 ± 465.5 (U251TMZ); p-value = 0.01). To uncover the molecular modulators of TMZ resistance in U251TMZ, we performed the whole genome shRNA library screening as graphically displayed in Figure 1B . Using this approach, we identified shRNA candidates that, when expressed, either enhance or suppress the sensitivity of U251TMZ cells to TMZ, as shown by a volcano plot. ( Figure 1C ). We focused on the shRNA candidates preferentially enriched in the control DMSO-treated over TMZ-treated cells, indicating that the genes targeted by these shRNA are negative modulators of TMZ sensitivity in U251TMZ cells. Interestingly, the nucleocytoplasmic protein XPO1 was among the top 25 genes with the shRNA that resensitized U251TMZ cells to TMZ, as indicated by increased enrichment of XPO1 shRNA in DMSO-treated cells ( Figure 1D ). Since the XPO1 inhibitor Selinexor is a brain-penetrant compound and a potential therapeutic agent for GBM therapy (22, 26), we evaluated whether this compound could overcome resistance in U251TMZ cells. As shown in Figure 1E , Selinexor significantly resensitized U251TMZ cells to TMZ (Relative fluorescence: TMZ (100 µM) alone = 5820 ± 392.8 and Selinexor (100 nM) alone = 3273 ± 341.1 vs. TMZ (100 µM)/Selinexor (100 nM) = 1828 ± 231.4; p-value < 0.001). Together, these findings suggest that XPO1 may contribute to the evolution of TMZ resistance in MGMT-promoter hypermethylated GBM cells, and Selinexor can potentially be used for overcoming TMZ resistance.

The above findings suggest that Selinexor may control the sensitivity of MGMT-methylated GBM cells to TMZ. Since MGMT expression plays a critical role in the sensitivity to TMZ (6, 8, 9, 27), we evaluated whether the expression of MGMT influences the sensitivity of GBM cells to the combined Selinexor/TMZ treatment. To this end, we used cells primarily cultured from two GBM PDXs (MGMT-unmethylated GBM6 and MGMT-methylated GBM22) and two established GBM cell lines (MGMT-unmethylated T98 and MGMT-methylated U251). Expectedly, the MGMT-unmethylated GBM6 cells were resistant to TMZ (DMSO vs. TMZ 100 µM p-value = 0.17) but highly sensitive to Selinexor alone (DMSO vs. Selinexor p<0.0001), and there was no additional growth suppression by combining TMZ with Selinexor (p-value = 0.16; Figure 2A ; upper panel). Similar results were observed in the established MGMT-unmethylated T98G cells ( Supplementary Figure S1 ). In contrast, the MGMT-methylated GBM22 cells were sensitive to TMZ (30 uM), and this sensitivity was significantly enhanced by combining Selinexor with TMZ (DMSO vs. TMZ p<0.0001; TMZ vs TMZ/Selinexor p<0.0001; Figure 2A ; lower panel). These findings closely resemble the data showing that siRNA silencing of XPO1 enhances the sensitivity of U251 cells to TMZ ( Supplementary Figure S2 ). Unexpectedly, MGMT-unmethylated GBM6 cells demonstrated a higher sensitivity to Selinexor ( Figure 2A ; upper panel) than the MGMT-methylated GBM22 cells, which initially showed a diminished proliferation in response to Selinexor alone but progressively recovered ( Figure 2A ; lower panel), suggesting that GBM22 cells may be less vulnerable to Selinexor inhibition than GBM6 cells. Indeed, we found that Selinexor (100–1000 nM) completely inhibited the XPO1 protein in GBM6 cells, while similar concentrations did not fully block the expression of XPO1 protein in GBM22 ( Supplementary Figure S3 ).

To confirm that MGMT status determines the sensitivity to Selinexor/TMZ treatment, we reexpressed the exogenous MGMT in promoter methylated U251 (U251MGMT) or the control empty vector-expressing U251V ( Figure 2C , top panel). Interestingly, the control U251V cells were sensitive to either single-agent TMZ treatment (10 µM) or Selinexor (30 nM). A significant increase in sensitivity was observed when TMZ was combined with Selinexor (p<0.001; Figure 2C , middle panel). In contrast, U251MGMT cells were resistant to TMZ and slightly sensitive to Selinexor alone, but no additional effect on proliferation was observed by combined TMZ/Selinexor treatment ( Figure 2C , lower panel). To further demonstrate the role of MGMT in selinexor/TMZ sensitivity, we knocked down MGMT in promoter-unmethylated GBM6 cells ( Figure 2B , upper panel). MGMT-depleted GBM6 cells showed a relatively increased sensitivity to Selinexor, which was further potentiated when combined with TMZ (DMSO vs. Selinexor p<0.01, DMSO vs. TMZ, p<0.001; TMZ vs. TMZ/Selinexor, p<0.001, Selinexor vs. TMZ/Selinexor, p<0.0001; Figure 2B , lower panel) compared to control shNT-expressing cells ( Figure 2B ; middle panels), further highlighting the potential role of MGMT in modulating responsiveness to Selinexor/TMZ treatment.

In parallel with suppressing the sensitivity to selinexor/TMZ treatment, exogenous MGMT significantly diminished the Selinexor-mediated potentiation of TMZ-induced DNA double-strand breaks (DSBs) as evidenced by decreased ɣ-H2AX foci both in TMZ- and Selinexor/TMZ-treated U251MGMT compared with the control U251V ( Figure 3 , left panel). The number of nuclei with ≥ 25 ɣ-H2AX foci in Selinexor/TMZ-treated U251MGMT was 6.30 ± 1.415, whereas for Selinexor/TMZ-treated control U251V was 20.10 ± 2.036, a statistically significant difference (p<0.0001). Consistent with the increased TMZ-induced DNA damage, Selinexor-potentiated TMZ-induced PARP-cleavage indicated an increased cell death through apoptosis in U251V but not in U251MGMT ( Supplementary Figure S4 ). Together, these findings support the notion that MGMT may influence the sensitivity to combined TMZ/Selinexor treatment.

As an initial step toward understanding the mechanism through which MGMT may control sensitivity to TMZ/Selinexor in GBM cells, we evaluated the effect of Selinexor on MGMT expression in unmethylated GBM cells. Interestingly, Selinexor strongly induced MGMT expression in unmethylated T98G and GBM14 cells. At the same time, less induction was observed in unmethylated low-MGMT GBM43 cells ( Figure 4A ). A qRT-PCR analysis showed increased MGMT transcripts and a rebound XPO1 mRNA in Selinexor-treated GBM6 ( Figure 4D ) and GBM14 cells ( Supplementary Figure S5 ). Since the PKA-CREB pathway controls MGMT expression (28, 29), we tested whether Selinexor induces MGMT expression by activating this pathway. Indeed, Selinexor treatment blocked XPO1 protein in a concentration-dependent manner and was accompanied by increased phospho-CREB^S133 in unmethylated GBM cells ( Figure 4B ). To confirm the role of the PKA-CREB pathway in mediating Selinexor-induced MGMT expression, we treated the unmethylated GBM6 cells with Selinexor with and without a PKA inhibitor, H89. As expected, Selinexor but not H89 suppressed XPO1 protein in GBM6 cells and was accompanied by increased MGMT and phospho-CREB^S133 ( Figure 4C ). The PKA-CREB inhibitor H89 suppressed the basal and Selinexor-induced MGMT expression. Moreover, H89 significantly blocked the Selinexor-induced phospho-CREB^S133 ( Figure 4C ). Next, we performed a qPCR to demonstrate whether Selinexor controls MGMT at the transcription level. As shown in Figure 4D , Selinexor treatment led to a 6-fold increase in MGMT mRNA in GBM6 cells. Since XPO1 mRNA was previously shown to be increased in Selinexor-treated tumors (30), XPO1 mRNA level was used as a positive control. Finally, we performed ChIP to establish the impact of Selinexor on the MGMT promoter chromatin accessible-histone status. As shown in Figure 4E , Selinexor increased the promoter decoration with trimethylated lysine 4 of histone H3 (H3K4me3) compared with the control DMSO-treated cells. Since the MGMT promoter is not decorated with H3K27Ac (https://genome.ucsc.edu/), we used this mark as a negative control. These findings suggest that selinexor induces MGMT transcription through PKA-CREB signaling-mediated promoter activation.