5. Manuscript II
5.3.5 BET-dependent genes respond differentially to low and high doses of JQ1
Given the ability of low concentrations of JQ1 to sensitize cells to paclitaxel, we investigated the effects of JQ1 on gene transcription upon treatment for 24 hours with 100 nM and 500 nM JQ1.
We have already shown that 100 nM inhibits the growth of approximately 20% of cells while 500 nM affects around 70% of cell growth (Figure 26B). Upon studying the gene transcriptional changes of regulated genes at the two concentrations, we identified 16 patterns of gene transcriptional modulation. Notably, the most significant cluster of genes included 222 genes which were downregulated at 100 nM and further downregulated with 500 nM. Most interestingly, other genes showed preference in their response to one condition where they were downregulated more profoundly by 100 nM and others by 500 nM (Figure 30A, 30B, 30D).
Remarkably, a cluster comprised of 45 histone genes were shown to be upregulated in the case of high concentrations of JQ1 (Figure 30C). This is probably due to perturbed processing of the poly adenylated histone mRNAs as BET inhibition decreases the recruitment of CDK9 which was shown to play a significant role in 3' end processing of replication-dependent histones [201, 440-442].
Notably, the cluster of genes which responded more to 100 nM showed significantly lower occupation of BRD4 at their TSS compared to the clusters which showed no preference or were more affected by 500 nM JQ1 (Figure 30E). Gene ontology for genes responsive to lower concentrations showed that already at lower concentration, JQ1 can attenuate inflammation and regulation of gene cycle explaining the ability of lower concentration to sensitize cells to paclitaxel (Figure 30F, 30G).
Figure 30 : BET-dependent genes respond differentially to low and high doses of JQ1. (A-D) Expression plots showing different clusters of genes sharing response patterns to low and high concentrations of JQ1. x-axis shows concentration of JQ1 and y-axis stand for fold change compared to cells treated with DMSO. (E) Box-plot showing the occupancy of BRD4 at the TSS of the genes of different clusters. *** stands for p-value < 0.0001 for unpaired t-test. (F,G) Gene ontology plot for clusters of genes that are more responsive to low concentration (F) and high concentration (G).
5.4 Discussion
We report in this study the sensitization of pancreatic cancer cells to paclitaxel by low concentrations of the BET inhibitor JQ1. Combining JQ1 with chemotherapeutic agents has already been under investigation and showed promising results in various types of malignancies [443-447]. In particular, anti-microtubule drugs exhibited the most potent synergism with BET inhibitors in a screen of hundreds of agents in neuroblastoma [448]. While the anti-proliferative effect is usually used for JQ1 as a readout for efficiency, we showed that JQ1 sensitization can be exhibited in very low concentrations that have modest effects on growth. This indicates that sensitization can be rather due the transcriptional modulatory effects of BET inhibition.
Although the changes accompanying and causing paclitaxel resistance in vitro are stochastic in nature, we observed a transcriptional activation of a pro-inflammatory program that was reported in other systems upon paclitaxel treatment and resistance [420, 422, 423]. BET inhibition was reported to attenuate inflammation in acute pancreatitis [449] , heart failure [450], and a subset of chronic obstructive pulmonary disease (COPD) patients [451]. It is not clear whether activation of inflammation is predisposing or accompanying to resistance. In any case, the fact that TNFα inhibition synergizes with BET inhibition in sensitizing pancreatic cancer to paclitaxel implies a causative role of inflammation in the development of resistance [426].
Additionally, the reversal of the enrichment for a migratory phenotype coupled with increased
expression for the mutation inducers from the APOBEC family in resistant cells further justifies the investigation of BET inhibitors as an adjuvant therapy.
Interestingly, we observed a high association between BRD4 occupancy at putative enhancers and activation of associated genes in resistant cells. This was accompanied by the observation that some of the gained BRD4 regions at enhancers were associated with an open chromatin state in normal cells. Enhancers associated with open chromatin regions that are activated upon metastasis were reported in pancreatic cancer [154]. We also report here that enhancers associated with resistance may also be open before activation in case of resistance. This may imply that these regions are primed for activation and may explain why pancreatic cancer frequently displays innate or acquired resistance in a shorter period compared to other tumor types. It would be of interest to identify such regions and evaluate the potential of using their
“openness” to predict the development of resistance and potentially the sensitization with BET inhibition. Moreover, enhancer RNA (eRNA) transcribed at these particular enhancers can also provide us with unconventional biomarkers for prognosis and therapy [103].
Given that the gain of BRD4 has also resulted in the identification of resistance-specific super enhancers driving genes with poorer prognosis, other therapies that specifically target this special class of enhancers can be highly beneficial in resistant pancreatic cancer. For example, inhibition of Cyclin Dependent Kinase 7 (CDK7), a subunit of TFIIH, has been shown to extensively affect the activity of super enhancers in neuroblastoma and ovarian cancer [452, 453]. Additionally, histone deactylases inhibitors (HDACs) were reported to affect super enhancer-driven transcription [331]. Accordingly, these and other inhibitors can potentially be used to manage patients with resistance to de-activate SE-driven transcriptional activation and improve prognosis.
To evaluate the effect of low concentrations of JQ1 in sensitization, we compared the gene transcriptional profile of normal cells upon treatment with low and high concentrations of JQ1.
Interestingly, various clusters of genes showed different patterns of responsiveness to the different concentrations. So far, dose escalation studies for BET inhibitors only takes into consideration the anti-proliferative and side effects of those agents and not their transcriptional modulatory effects. Thus, the rationale for variant sensitivity of genes to different concentrations of BET inhibitors is not yet known. As JQ1 targets both the BD1 and BD2 of the BET family members, we propose a bivalent pattern of inhibition which can be mirrored in dose-dependence. For example, if only one of the bromodomains is bound at certain genes, they would respond to lower concentrations. On the other hand, if the two bromodomains are required to be occupied by acetylation at certain regions, higher concentrations of JQ1 may be needed for ejection of the BET family member from these regions (scheme in Figure 31A). This hypothesis can be tested by using a monovalent selective BET inhibitor for each bromodomain and investigating the profile of each. If this theory is true, the genes which respond to high concentrations would be resistant to either monovalent inhibitor but responsive to co-treatment.
Another explanation for various dose responses can be due to the various occupancies of transcription factors at these regions. It can be possible that other transcription factors which are not BET-dependent can compensate for the absence of BET members at some regions and ensure the activation of the genes irrespective of the concentration of BET inhibitor (Figure 31B). Additionally, genes that are responsive to high concentrations may be highly occupied by BRD4, requiring a higher concentration for an effect to take place (Figure 31C). This is further supported by our observation that highly responsive genes have lower occupancy of BRD4 at their TSS.
Figure 31 : Graphical representation for hypotheses explaining dose-dependent response for BET inhibitors. (A) Genes occupied in a bivalent manner require higher concentrations to be affected.
(B) Other activating transcription factors can compensate for absence of BRD4. (C) Occupancy of BRD4 can affect response with lower occupancy leading to high response.
In conclusion, as BET inhibitors are under investigation for their role as therapeutic management options for patients, insights into their mechanism of action and the optimal approaches to use them are currently lacking. We propose BET inhibitors as good options for sensitization to paclitaxel in both innate and acquired settings. These combinations however should be approached with the goal of leveraging the gene transcription modulating effects of these inhibitors rather than primarily their anti-proliferative effects in isolation.