BET protein inhibition disturbs DNA damage response signaling pathways in glioblastoma
In an effort to leverage clinically relevant pathways disturbed by BETi for druggable targets, we analyzed differential gene expression data obtained in a GBM-derived sphere line, LN-2683GS, treated with the tool drug JQ1. The cells treated with 1µM JQ1 over a time course of 48h underwent extensive transcriptome changes as reported 6. Significant association with JQ1-treatment was observed for 4 712 genes (adjusted p-value by Bonferroni correction, < 0.1 and log2(CPM + 1) > 1), whereof 169 were annotated as DDR genes as defined by Pearl et al. 9. To identify JQ1-response patterns we determined the optimal number of gene clusters using K-means and obtained 6 clusters, as visualized in a Calinski’s criterion graphic and a corresponding heatmap (Supplementary Fig. 1). The comparison of the original K-means and the re-clustering method exhibited a similar classification for the 500 training datasets. The percentage of good classification was equal to 95% with a kappa value of 0.94. Two clusters showed JQ1-induced gene expression over time (clusters 1 and 2), and two clusters displayed consistent downregulation (clusters 4 and 5), while two clusters displayed transient down- (cluster 3) or upregulation of expression (cluster 6), respectively. The 169 DDR genes (Supplementary Table 1) were distributed among all six JQ1-response patterns (cluster 1, 23 genes; cluster 2, 35 genes; cluster 3, 51 genes; cluster 4, 16 genes; cluster 5, 34 genes; cluster 6 and 10 genes) as visualized in a heatmap (Fig. 1). Enrichment analyses of DDR pathways by cluster, showed only one significant association between cluster 3 and MMR genes (adjusted p-value by Bonferroni correction < 0.001). The annotated list of all 169 retained DDR genes, their expression levels by treatment and time point, cluster affiliation, Fréchet distance, and pathway information, based on Pearl et al. 9, is available in Supplementary Table S2.
A direct effect of BETi may be expected among rapidly downregulated genes by stripping BET proteins from their binding sites. Inspection of the clusters 4 and 5 revealed MGMT among the consistently downregulated genes, following expression pattern 4, as visualized in Fig. 1. The consistent downregulation by JQ1, identified MGMT as a prime target which opens the opportunity to sensitize GBM with an unmethylated MGMT promoter to TMZ. Patients with MGMT unmethylated GBM basically show no benefit from TMZ therapy 10. Given the fundamental role of a fully functional MMR system for sensitivity of cells to O6-methylguanin lesions in MGMT-deficient cells, we paid attention to the modulation of key genes involved in the MMR system. As mentioned above, the MMR genes were significantly enriched in cluster 3, and comprised among others MSH6, MSH2 and MLH1 that have been associated with acquired treatment resistance to TMZ in recurrent gliomas when mutated or silenced otherwise 27. Even though the expression of MSH6, MSH2 and MLH1 was transiently downregulated upon JQ1 treatment at early time points (6h and 12h), it was restored to baseline 24h after initiation of JQ1-treatment, suggesting that BETi does not compromise MMR in this sphere line (Fig. 1, Supplemental Table S2).
BET protein inhibition reduces MGMT expression and prohibits its induction upon temozolomide treatment
Next, we monitored downregulation of MGMT in several GBM sphere and cell lines with endogenous MGMT expression on the RNA and protein level in response to treatment with JQ1, TMZ, or both. The JQ1 concentrations were adapted to the sensitivity of the individual cell and sphere lines (0.1-1µM), and TMZ was used at a clinically relevant dose of 100µM 26, 28. The MGMT expression was significantly affected by treatment over the time course of 48h for the cell lines LN-340, LN-2683GS and LN4372GS (p = 0.0051, p < 0.0001 and p = 0.0095 respectively, Fig. 2A, Supplementary Table S3).
The results confirmed rapid downregulation of MGMT expression upon JQ1 treatment as measured over a time course of 48h (Fig. 2A). Moreover, MGMT-induction generally observed upon TMZ treatment alone, was prohibited by JQ1, and the expression levels were kept significantly below the baseline over the time course. This behavior also translated to the protein level, although with a delay (Fig. 2B). Substantial MGMT depletion with JQ1 treatment alone was observed after 72h of treatment, with a more pronounced effect after 120h. TMZ treatment alone also showed a decrease of MGMT protein after 24h, compatible with the suicide reaction of MGMT after transfer of the methyl group that leads to ubiquitination and proteasome-mediated degradation, requiring de novo synthesis 29. These results were consistent across all the different GBM cell lines and sphere lines tested (Fig. 2). In the following experiments testing TMZ-related effects, cells were pretreated 120h with JQ1 (or DMSO, control) to allow for JQ1-mediated depletion of MGMT.
BET protein inhibition reduces BRD4 occupancy at the MGMT promoter region
To investigate whether MGMT expression is directly regulated by BRD4, we performed chromatin immuno precipitation followed by quantitative PCR (ChIP-qPCR) analysis for BRD4 binding in the MGMT promoter region. T98G cells that exert relatively high levels of endogenous MGMT on the RNA and the protein level (Supplemental Figure S2), were treated for 2h with JQ1 [1µM], and relative BRD4 occupancy was determined by ChIP-qPCR using previously described primer sets 19, 25. The two regions interogated are located within the CpG island of the promoter and the first exon and the analysis demonstrated a significant difference in BRD4 occupancy at both MGMT promoter regions tested (both locations, F2, F3, p < 0.5, two-tailed ratio paired t-test, with correction for variance heterogeneity, Fig. 3). The JQ1-associated decrease in BRD4 binding was supportive of a direct regulatory effect on MGMT expression. Furthermore, we also evaluated associated changes of RNA polymerase II (Pol II) occupancy, a marker for active transcription. In line with the observed decrease in MGMT expression, we detected a significant difference in Pol II occupancy (F2, p < 0.0001; F3, p = 0.0377), with decreased binding at both MGMT promoter regions tested, suggesting attenuation of the MGMT transcription process.
Overall, our results support that MGMT transcription is attenuated upon JQ1 treatment due to BRD4 depletion at the promoter region.
BET protein inhibition modulates TMZ-induced DNA damage in an MGMT-dependent manner
In light of the postulated direct downregulation of MGMT upon BETi in GBM lines, we investigated the role of BETi in modulating the TMZ-induced DNA damage response. We treated the MGMT expressing GBM cell line LN-340 with TMZ alone or in combination with JQ1 [0.1µM], and monitored γ-H2AX levels, a marker for DNA double-strand breaks (DSBs)30. The formation of γ-H2AX foci is among the first steps that initiates the recruitment of DNA repair proteins. Cells were pretreated for 120h with JQ1 or DMSO to allow for JQ1-mediated depletion of the MGMT protein, and γ-H2AX was measured 48h after treatment with TMZ, quantifying immunofluorescence determined by confocal microscopy. In absence of JQ1, we observed that LN-340 showed no difference in γ-H2AX levels upon TMZ treatment alone, as compared to DMSO control, while a significant difference was observed upon treatment with the MGMT-specific pharmacologic inhibitor O6-benzylguanine (O6BG) (adjusted p-value < 0.01, Dunnett T3 multiple comparisons tests following one-way ANOVA), restoring sensitivity to TMZ (Fig. 4A, Supplementary Table S4). Indeed, LN-340 cells treated with O6BG in combination with TMZ showed a robust increase in the rate of DSBs as compared to TMZ treatment alone. This is in line with the MGMT-mediated resistance in this cell line. In contrast, when treating the cells in combination with JQ1, a significant difference (adjusted p < 0.01) of γ-H2AX levels was observed upon TMZ treatment, as compared to TMZ alone, indicating that depletion of MGMT upon JQ1 treatment led to an increase in DSB formation following TMZ treatment (Fig. 4A). However, no interaction was observed between O6BG and JQ1, hence the addition of O6BG in JQ1 treated cells did not further sensitize cells to TMZ, suggesting that MGMT protein levels were already low from JQ1 treatment.
To further support our findings of a MGMT-dependent effect of BETi and to confirm the key role of MGMT in conferring resistance to TMZ in these GBM models, we used a Dox-inducible Tet-On system for MGMT in the GBM line LN-229. LN-229 does not express endogenous MGMT, due to promoter methylation 19, and is known to be highly sensitive to TMZ treatment 31, 32, 33. Induction of MGMT with Dox at 100ng/mL yielded MGMT protein levels comparable to MGMT-proficient cell lines (Supplementary Figure S3). We observed that LN-229MGMTind_C12 cells acquired a strong TMZ resistance phenotype upon MGMT induction (Fig. 4B, Supplementary Table S4). Expectedly, the use of the pharmacologic MGMT inhibitor (MGMTi) O6BG had a significant effect on γ-H2AX levels (p < 0.01), restoring TMZ sensitivity, reflected by increased DSBs. In contrast, no effect (adjusted p > 0.5) was observed with JQ1 treatment, hence, not sensitizing Dox-treated LN-229MGMTind_C12 cells to TMZ treatment in this context. This suggested that JQ1 was not able to interfere with ectopic MGMT expression, which is controlled by the Dox-inducible Tet-On promoter. Therefore, BETi did not influence ectopic MGMT expression or sensitivity to TMZ induced DSBs, whereas pharmacologic depletion of MGMT by O6BG treatment reversed the acquired TMZ resistance.
Altogether, our data have shown that JQ1 induces more DNA DSBs in TMZ treated GBM cells expressing endogenous MGMT as compared to TMZ alone.
BET inhibition attenuates glioblastoma viability upon TMZ treatment
The observed increase of DSBs suggested that treatment with JQ1 may reduce the viability of GBM cell lines with endogenous MGMT expression, in response to TMZ treatment. We treated LN-340 and T98G with JQ1 or TMZ alone or combined with JQ1, while both single agent treatments had a significant effect on cell viability in both cell lines (for both lines and both treatments, p < 0.0001, fixed effect from mixed model with interactions between treatments, Supplementary Table S5), we observed that the addition of JQ1 significantly sensitized cells to TMZ treatment (Fig. 5A), reflected in the significant interaction effect between JQ1 and TMZ (P = 0.006 and p < 0.0001, respectively). The specificity of the MGMT-mediated effect of JQ1-treatment was further tested using the pharmacologic inhibitor O6BG in the experiments, with or without JQ1, respectively. The addition of O6BG on its own had no effect on cell viability, whereas it sensitized the cells in combination with TMZ. However, O6BG did not further sensitize the cells to TMZ in presence of JQ1 (no significant interaction between O6BG and JQ1, p = 0.2746. In accordance, cell cycle analysis revealed that TMZ or JQ1 treatment alone did not alter the cell cycle profile compared to untreated cells. However, combinatorial treatment of JQ1 and TMZ increased S phase and G2/M phase cell cycle arrest in GBM cells as compared to controls (Fig. 5B). And the addition of O6BG alone, or in combination with JQ1 had no effect on the cell cycle, while in combination with TMZ it increased the proportion of cells in S and G2/M phase (Fig. 5B).
BET protein inhibition does not compromise the MMR system in glioblastoma
As aforementioned, a compromised MMR system generates resistance to TMZ treatment as it is essential for the cytotoxic effect O6meG lesions that remain unrepaired in absence of MGMT. Our differential gene expression analysis in LN-2683GS had shown that key MMR genes, were only transiently modulated by JQ1 treatment and their expression was restored after 12h (Fig. 1). Treatment with JQ1 alone or in combination with TMZ did not significantly alter RNA or protein expression levels of MSH6 and MSH2 in LN-340 (Fig. 6A, 6B, Supplementary Table S6) or T98G (data not shown).
To determine the effect of a non-functional MMR pathway in conferring TMZ resistance in our experimental model, we transduced LN-340 with a Dox-inducible Tet-On shRNA against MSH6 using 2 distinct sequences (Fig. 6D). Depletion of MSH6 confirmed that TMZ resistance was independent of MGMT in this scenario, as pharmacologic inhibition of endogenous MGMT with O6BG was ineffective in restoring sensitivity to TMZ as measured by cell viability (adjusted p-values = 0.0022 and 0.0002 for the 2 sequences, three-way ANOVA including interaction between the treatments, TMZ, OB6G and JQ1, corrected for multiple testing by Tukey’s test). On the contrary, cells transduced with the non-targeting Dox-inducible Tet-On shRNA did not change behavior upon doxycycline exposure and remained sensitized to TMZ upon O6BG treatment (Fig. 6C, Supplementary Table S7) (adjusted p-value = 0.127).
Finally, we demonstrated that the use of BETi in GBM cells does not negatively impact the MMR system that would result an undesirable TMZ resistance.