MondoA mediates transcriptional coordination between the MYC network and the integrated stress response in pancreatic cancer

Significance

This report investigates mechanisms underlying the dependence of MYC-amplified pancreatic cancer cells on the MYC network member MondoA which, as a heterodimer with MLX, is a nutrient-sensing transcription factor. We show this dependency is linked to genomic crosstalk between MYC, components of the proximal MYC network, and the master regulator of the integrated stress response, ATF4. Moreover, we find that small molecule inhibitors of MondoA-MLX transcriptional activity abrogate survival of MYC-amplified pancreatic ductal adenocarcinoma (PDAC) lines and patient derived organoids. The significance of this work is its elucidation of a unique vulnerability intrinsic to MYC, an oncogenic driver associated with a wide range of cancers, which is considered to be “undruggable.”

Abstract

MYC amplification contributes to poor survival and outcome in pancreatic ductal adenocarcinoma (PDAC). Here we show that in PDAC cell lines with amplified MYC, MondoA is required for viability, facilitating proliferation while suppressing apoptosis in vitro and in vivo. Transcriptional and genomic profiling demonstrates that loss of MondoA leads to altered expression of direct MondoA targets as well as MYC target genes and is accompanied by shifts in genomic occupancy of MYC, MNT, and the MondoA paralog ChREBP. This altered genomic binding by MYC network members is associated with transcriptional perturbation of multiple metabolic and stress pathways, as well as global changes in N6-methyladenosine modification (m⁶A) of messenger RNA (mRNA). MondoA inhibition disrupts coordination between MYC network members and the Integrated Stress Response (ISR), resulting in decreased translation of ATF4 mRNA, discordant gene regulation of shared targets of MYC and ATF4 and, ultimately, apoptosis. Reestablishing ATF4 protein expression rescues the diminished viability due to loss of MondoA expression or activity, providing direct evidence of a link between deregulated MYC and the transcriptional machinery of the ISR. Last, we find that small-molecule inhibition of MondoA is lethal in a subset of PDAC cell lines, including patient-derived organoids, suggesting that the ability to target MYC via chemical inhibition of MondoA transcriptional activity may have broad efficacy.

Metabolic sufficiency during normal development is required for biomass accumulation, genome replication, and subsequent cell division induced by growth factors. These linked mitogenic and biosynthetic pathways are targets of mutation (e.g., PI3K-AKT, RAS, MTOR) and can be activated or inhibited in malignancies. Components of the MYC network (reviewed in ref. 1) consist of interconnected mitogenic (MYC-MAX) and nutrient-responsive (MondoA/ChREBP-MLX) nodes of heterodimeric transcriptional activators and repressors. These transcriptional regulators, while acting to coordinate normal growth and metabolism during development, are also implicated in the pathological trajectories of the same pathways in diabetes, obesity, and cancer. Mitogenic stimulation of MYC leads to cell cycle progression and concurrent activation of biosynthetic pathways required for cell division. Deregulation of MYC in cancer drives a “selfish” biomass accumulation program coupled with sensitization to metabolic apoptotic stimuli (2). The MYC network exhibits complex interactions and cell-type specific functions, including both context-dependent oncogenic or tumor suppressive roles for individual members. Alterations to the network have been found to elicit new dependencies and synthetic lethal interactions between the MAX and MLX arms of the network. Examples include MYC/MYCN-amplified tumors requiring MondoA-MLX (3), and MAX-null tumors being sensitive to chemical inhibition of MondoA (4). Thus, the nutrient-sensing MLX arm of the network is responsive to alterations in the mitogen-responsive MAX arm.

Cellular metabolic state is monitored by a variety of nutrient-sensors that provide feedback to activate or inhibit both mitogenic and metabolic pathways. This includes Mondo proteins (MondoA and ChREBP, reviewed in ref. 5), which respond to flux through the major glucose and lipid metabolic pathways (e.g., glucose induces Mondo-dependent TXNIP, which negatively regulates glucose uptake). Prolonged metabolic stress (either deficiency or overabundance) can lead to activation of the Integrated Stress Response (ISR) and its transcriptional effector, ATF4, to either rebalance cellular metabolism or, alternatively, undergo apoptosis (reviewed in ref. 6).

The ISR integrates disparate signals (e.g., glucose, lipid, amino acid, ROS levels) with transcriptional programs that can be remedial (pro-survival) or terminal (pro-apoptotic), dependent upon duration and intensity of the signal. A fundamental output of the ISR is inhibition of global protein synthesis coupled with a switch to selective translation of uORF-containing transcripts, such as ATF4 (reviewed in ref. 7). Deregulated oncogenes, such as MYC or KRAS, are known to induce the ISR through multiple pathways, such as metabolic stress leading to ROS production, increased translation leading to ER Stress (ERS) via protein misfolding, and the Unfolded Protein Response (UPR). MYC requires components of the UPR/ISR (reviewed in ref. 8) and cooperatively shares targets and promoter co-occupancy with ATF4 for critical transcriptional crosstalk between the MYC network and the ISR (9). The MondoA/ChREBP-MLX target TXNIP is a component of this apoptotic cascade, as deletion of either ChREBP or TXNIP prevents glucotoxicity-induced beta-islet cell death in diabetes (10, 11). Taken together, the literature suggests crosstalk between multiple components of the MYC network and the ISR is critical for cell function.

Pancreatic cancer progression, in transitioning from pancreatitis to pancreatic intraepithelial neoplasia (PanIN) and ultimately to pancreatic ductal adenocarcinoma (PDAC), also involves MYC-driven metabolism, RAS-induced senescence, and ISR components [see review of UPR/ISR in PDAC progression (12)]. MYC copy number gain has been shown to correlate with decreased PDAC patient survival (13) but can also activate or intensify putative targetable vulnerabilities [e.g., hypertranscription, metabolic and other downstream effectors such as hypertranslation, proteostasis (14) and ISR pathways]. PDAC has an abysmal outcome and survival rate, and actionable targets are required to develop effective therapeutics. We hypothesized that MYC network interactions could be critically important in sustaining oncogenesis in MYC-amplified PDAC. We therefore sought to determine the functional interactions between MYC and MondoA in this malignancy, as well as explore potential for targeting such interactions.

Results

MondoA Regulates Viability in MYC-Amplified PDAC Lines.

To determine whether pancreatic cancer cells require MondoA for growth and survival, we utilized validated small interfering RNAs (siRNAs) against MondoA and assessed relative growth rates in four PDAC lines compared to nontargeting siControl. As MYC genomic amplification correlates with poor patient outcome and survival (13), this cell line panel included two lines with MYC amplification (PSN-1 and DAN-G) and two without (PANC08.13 and SU.86.86). Basal levels of relevant proteins in these lines are shown in SI Appendix, Fig. S1A. Following MondoA knockdown, we observed a significant inhibition of growth only in MYC-amplified cells compared to non-MYC-amplified cells (Fig. 1 A and B). As monitored by western blot, the levels of MondoA protein knockdown were similar in the four lines while MYC levels were unaffected (Fig. 1C, quantification shown in SI Appendix, Fig. S1B). To directly test whether higher expression of MYC potentiates growth inhibition upon MondoA knockdown, we engineered a doxycycline-inducible (dox) MYC system in the nonamplified SU.86.86 cell line and monitored growth in the presence and absence of MondoA. As shown in Fig. 1 D and E, ectopic MYC increases cell number in the SU.86.86 cells, and this increase is abrogated upon MondoA knockdown. This suggests that deregulated MYC sensitizes these cells to loss of MondoA. If this is the case, then decreasing MYC levels should attenuate cell death upon siMondoA treatment. As shown in Fig. 1F, baseline caspase 3/7 activity is diminished significantly by siMYC treatment in all four lines, while it is increased significantly by siMondoA treatment in only the PSN-1 and DAN-G cells. This increase is mitigated by the combination of siMYC with siMondoA cotreatment. Therefore, MondoA loss promotes MYC-dependent apoptosis in MYC-amplified PDAC lines.

Fig. 1.

Functional interactions between MYC and MondoA in PDAC cell lines. (A) Incucyte growth curves of MYC-High PSN-1 and DAN-G (n = 4) or (B) MYC-Low PANC08.13 and SU.86.86 (n = 3) PDAC cell lines treated with siControl or siMondoA over 96 h. Data expressed as plate confluency and significant differences (two-tailed t test) as compared to final confluence of corresponding siControl. (C) Western blot for MondoA, MYC, and βActin in a panel of PDAC lines treated with either siControl or siMondoA for 72 h. Relative protein abundance expressed in relation to siControl treated PSN-1 cells and normalized to βActin. (D) SU.86.86 cells with a dox-inducible MYC vector were treated as indicated with or without dox (1 µg/mL) for 72 h. Data expressed as mean cell count (×10^5) and significant differences (n = 3, one-way ANOVA) as compared to siControl without dox. (E) Western blot corresponding to panel 1D. Relative protein abundance expressed in relation to siControls + dox normalized to βActin. Using a t test there was no significant difference in MondoA protein levels following MYC induction (n = 3, replicates not shown). (F) Caspase-Glo 3/7 system to assess apoptosis in a panel of PDAC lines treated as indicated for 72 h. Data expressed as mean luminescence and normalized within each line to siControl cells. Significant differences (n = 4, one-way ANOVA) as compared to siControl within each cell line. All data expressed as mean ± SEM and P-values expressed as ns = not significant; *P ≤ 0.05; **P ≤ 0.005, ***P ≤ 0.0005, ****P ≤ 0.0001.

MondoA Promotes Clonogenic and Xenograft Outgrowth of PSN-1 Cells.

To extend our findings on the relationship between MYC levels in PDAC cells and MondoA inhibition, we assessed the growth of cells as foci in a clonogenic assay. As shown in Fig. 2A, siRNA against MondoA impairs clonogenic survival of the MYC-amplified PSN-1 cells, demonstrating a significant decrease in colony formation frequency upon plating at a range of low densities compared to siControl cells. To test the effect of stable knockdown of MondoA, we utilized short hairpin RNA (shRNA) expression via a dox-inducible viral vector and compared the effect of shControl vs. shMondoA in the presence and absence of doxycycline (SI Appendix, Fig. S1C). This inducible system recapitulates the phenotype of transient siRNA knockdown, as PSN-1 cells harboring the shMondoA construct were impaired for clonogenic outgrowth only in the presence of doxycycline when seeded at an identical cell number (Fig. 2B).

Fig. 2.

Effects of MondoA-MLX inhibition on MYC-driven PDAC in vitro and in vivo. (A) Clonogenicity assay of PSN-1 cells seeded at the indicated cells per well and treated with either siControl or siMondoA. Representative plate stained with crystal violet after 12 d (Left), colonies per well plotted (Right). Significant differences (n = 3, two-tailed t test) as compared to siControls. (B) PSN-1 clonogenicity assay with dox-inducible shMondoA or a control shRNA treated with or without dox and seeded at 100 cells per well. Colonies per well after 12 d plotted, with significant differences (n = 4, one-way ANOVA). (C) Mean tumor volume over time for PSN-1 subcutaneous xenograft with either shControl (left flank) or shMondoA (right flank). Significant differences (n = 10, two-tailed t test) expressed between vectors. (D) Final tumor weight for the 10 subcutaneous xenografts at 38 d post injection. Significant differences (two-tailed t test) shown. (E) Simplified schematic of the MYC transcription factor network. (F) Viability expressed as mean PSN-1 cell number following an siRNA screen of MYC network members for 72 h. Significant differences (n = 4, one-way ANOVA) as compared to siControl cells. (G) Clonogenicity assay of PSN-1 or DAN-G PDAC cells seeded at 50 cells and treated indicated concentrations of SBI-477 for 12 d. Mean colonies per well denoted, significant differences (n = 3, one-way ANOVA) as compared to DMSO treatment. Statistical significance analyzed as in Fig. 1.

We then used this system to probe the role of MondoA for in vivo xenograft growth. PSN-1 cells expressing either shControl (left flank) or shMondoA (right flank) were subcutaneously injected into the same mouse for pair-wise comparison. These mice were then provided with 200 g/L doxycycline-containing drinking water, and tumor growth and volume were monitored over time. As shown in Fig. 2 C and D, in vivo targeting of MondoA via shRNA significantly decreased both tumor volume and final tumor weight compared to shRNA control (representative mouse and tumor images in SI Appendix, Fig. S1 D and E, degree of MondoA knockdown shown in SI Appendix, Fig. S1 F and G). Therefore, MondoA is a viable target in MYC-amplified PDAC for blocking growth both in vitro and in vivo.

Multiple MYC Network Members Regulate PDAC Survival.

In order to define the role of MondoA interactors in the PSN-1 cells and directly test any requirement for distinct MLX-containing heterodimers (network diagramed in Fig. 2E), we employed siRNA knockdown to determine their potential role in PSN-1 cell survival. Knockdown of MYC, MondoA, MLX, or MNT, but not ChREBP, attenuates viability in PSN-1 cells (Fig. 2F). Sensitivity to MLX knockdown correlated with MYC amplification status, as viability was significantly decreased in amplified line PSN-1 but not nonamplified lines SU.86.86 or PANC08.13 (SI Appendix, Fig. S1H). This suggests that multiple components of the MYC network can support viability. The dependency upon MondoA, MLX, and MNT is also observed in other tumor types, such as B cell lymphoma cell lines (3), murine lymphoma models (15), male germ cell tumor lines (16), and a subset of neuroendocrine tumors (4). Consistent with these results, we also find that SBI-477, a small-molecule inhibitor of MondoA shown to block MondoA-MLX nuclear translocation (17), significantly attenuates clonogenic growth of both MYC-amplified PSN-1 and DAN-G cells (Fig. 2G), suggesting that chemical targeting of MondoA-MLX transcriptional activity in PDAC could provide a viable therapeutic option. We also observe similar results with SBI-993, a more bioactive derivative of SBI-477 that has been used to target liver cancer models (18) (SI Appendix, Fig. S1I), further supporting the therapeutic potential of MondoA. We confirmed increased sensitivity to SBI-993 with MYC induction in the dox-inducible SU.86.86 system and observed an SBI-993 dose-dependent increase in apoptotic marker BIM with MYC induction as compared to a dose-dependent BIM decrease in the absence of MYC induction (SI Appendix, Fig. S1 J and K).

MondoA Loss Leads to Altered Transcription and Genomic Occupancy of MYC Network Members.

To identify the transcriptional program maintained by MondoA in MYC-amplified PDAC, we performed RNA-seq on PSN-1 cells in which MondoA transcriptional activity was impaired by either siRNA knockdown of MondoA (siMondoA), siRNA knockdown of the MondoA heterodimerization partner MLX (siMLX), or chemical inhibition of MondoA-MLX via SBI-477 treatment. Each of these treatments were compared to either nontargeting siRNA (siControl) or the vehicle control for SBI-477 (DMSO). We identified significantly up- and down-regulated differentially expressed genes (denoted as DEGs, adjusted P-value < 0.05) in cells treated with either siMondoA (1640 up, 1489 down), siMLX (655 up, 806 down), or SBI-477 (1547 up, 1394 down) (Fig. 3 A and B and SI Appendix, Fig. S2 A–C and Tables S1 and S2).

Fig. 3.

MondoA and MYC network genomic engagement and transcriptional output in PDAC. (A) Venn diagram of DEGs up-regulated or (B) down-regulated after 72 h of siMondoA, siMLX, or SBI-477 (20 µM) treatment compared to siControl or DMSO. (C) Volcano plot of DEGs after siMondoA treatment. Up-regulated genes (blue points), and down-regulated genes (green), with shared DEGs across all three conditions (red), considered as “core” (red dots). (D) Hallmark pathway analysis of “core” DEGs up-regulated in “core” gene-list. (E) BigWig tracks for selected genes that lose MondoA and/or exhibit altered MYC network binding with siMondoA. (F) Volcano plot of DEGs after siMondoA treatment with genes promoter-bound by MYC (red) and bound by MondoA (blue). Relevant genes bound by both MYC and MondoA are labeled. (G) Total count of promoter-bound genes by indicated network members after siControl or siMondoA treatment. Solid bars denote unique targets; the checker box denotes shared gene target with MondoA. (H) TNFα/UPR target promoter binding by indicated transcription factors, siControl (blue), siMondoA (red).

We identified sets of “core” up- and down-regulated genes in response to MondoA-MLX inhibition that were changed in all three datasets (Fig. 3 A–C and SI Appendix, Fig. S3 A and B and Tables S1 and S2). Importantly, these “core” signatures correlate with MLXIP, encoding MondoA, within the TCGA PAAD dataset with the down-regulated “core” positively correlating with MondoA (R = 0.42, P = 6.2e-08) and the up-regulated “core” negatively correlating with MondoA (R = -0.29, P = 2.6e-4) (SI Appendix, Fig. S3C). The 79 up-regulated “core” DEGs were enriched for inflammatory pathways, including TNFα signaling and the UPR (Fig. 3D), while the 128 down-regulated “core” DEGs were enriched for the DNA damage response and interferon signaling (SI Appendix, Fig. S3D). We previously observed similar gene enrichment affecting metabolic genes and stress effectors upon loss of either MondoA or MLX in tumor cell lines (3) or male germ cells (16), and the UPR was targeted by SBI-477 in MAX-null neuroendocrine tumor cells (4). Among the DEGs are transcription factors (CEBP, NFκB, FOS) and genes that regulate metabolism (PFAS, FASN, SREBF1), proteostasis (ATF4, PPP1R15A, encoding GADD34), and apoptosis (CDKN1A encoding p21, GADD45A, H2AX) (SI Appendix, Fig. S3E and Tables S1 and S2). Many of these components of the MondoA-regulated program are critical determinants of homeostasis and cell fate and are also important for MYC-dependent tumorigenesis.

To determine direct vs. indirect transcriptional effects of MondoA perturbation upon MYC network members in PSN-1 cells, we investigated the genomic occupancy of MondoA, MYC, MNT, and MondoA paralog, ChREBP, in the presence (siControl) and absence of MondoA (siMondoA) by CUT&RUN. In PSN-1 cells, MondoA is associated with 1,051 gene promoters (Fig. 3 E and F). MondoA occupancy was detected at previously identified target promoters, such as TXNIP and ARRDC4, as well as newly identified targets, including TEAD2 and TIMM44 (Fig. 3E). Upon siMondoA treatment, its binding was diminished at these gene promoters as expected. Genes identified as both up- and down-regulated in our RNA-seq analysis (including decreased TXNIP expression and increased GADD34 expression) were occupied by MondoA, supporting a dual role for MondoA as both transcriptional activator and repressor.

Interestingly, genes whose promoters were occupied by MYC were likely to be differentially expressed upon MondoA depletion (KS stat 10e-14, comparing RNA-seq data from untreated vs MondoA depleted cells), consistent with the dependence of these cells on MYC-driven transcription. This occurred even as only a small subset of MYC bound genes were occupied by MondoA (1,051 of 11,691 genes) in these cells (Fig. 3F). Thus, MondoA appears to act as a substantial, albeit indirect, effector in the MYC-driven transcriptional program.

We have previously shown that altered expression of individual MYC network members can shift the genomic occupancy of other members (4, 16). Since knockdown of MondoA alters stress responsive targets of the UPR and TNFα-NFκB pathway, we plotted MYC, MNT, and ChREBP genomic coverage in the presence (siControl) and absence of MondoA (siMondoA) over the promoters of genes comprising these pathways. siMondoA resulted in decreased MYC binding (~25%) at MYC target loci and TNFα/UPR pathway genes compared to control cells (Fig. 3 G and H). We also observed a decrease (~20%) in total MNT binding over these same genes (Fig. 3 G and H). By contrast, ChREBP exhibited a gain (~twofold) in occupation of these gene promoters upon MondoA loss (Fig. 3 G and H). This included increased binding at the CEBPB promoter, which is in both the TNFα and UPR Hallmark categories (Fig. 3E). Altered binding by MYC, MNT, and ChREBP occurred at both up- and down-regulated gene promoters. This suggests that perturbation of MondoA leads to shifts in genome occupancy by multiple network factors, thereby contributing to dysregulation of both unique and shared targets, including the UPR.

The UPR Is Involved in the Response to MondoA Inhibition.

Because induction of the UPR has been shown to inhibit global protein synthesis, we surveyed translation rates in relation to MondoA and MYC in our MYC-low vs. MYC-amplified PDAC lines. We used puromycin incorporation to measure possible alterations in protein synthesis upon MondoA knockdown. As expected, the highest levels of puromycin incorporation were observed in the two MYC-amplified cell lines, and MondoA loss decreased puromycin incorporation in all four lines by ~50 to 65% (Fig. 4A). To examine sensitivity to UPR-induced inhibition of protein synthesis, we tested the dose-responsiveness of these lines to a UPR-inducing drug, Tunicamycin (Tm). We found that the MYC-amplified lines exhibited increased sensitivity to Tm treatment relative to the MYC-low lines (Fig. 4B), suggesting that high MYC expression poises the UPR for activation. This is likely due to increased proteostatic stress from enhanced biosynthetic demand driven by MYC (2).

Fig. 4.

Crosstalk between MondoA, MYC, and proteostasis via the UPR/ISR. (A) Puromycin incorporation assay for PDAC cell line panel treated with siControl or siMondoA for 48 h. Relative protein synthesis as detected by anti-puromycin western blot shown in relation to siControl treated cells and normalized to βActin within each cell line. (B) Cell viability after 48 h expressed as percent of DMSO treatment for a panel of PDAC cell lines treated with Tunicamycin. Significant differences (n = 3, one-way ANOVA) shown compared to DMSO treatment within lines. (C) Western blot of the three primary effector arms of the UPR in PSN-1 cells with siMondoA, siMYC, or combined, with or without UPR activation by Tunicamycin for 6 h. Relative protein abundance expressed in relation to Tunicamycin treated siControl cells and normalized to βActin. (D) Schematic of MYC Network interactions, as determined by promoter occupancy, with the ISR and ATF4. Green lines indicate activation, red lines repression. (E) ATF4 mRNA levels by RT-qPCR with and without siMondoA in MYC-amplified PDAC lines PSN-1 (n = 11) and DAN-G (n = 7). Data expressed as mRNA fold-change, significant differences (two-tailed t test) as compared to siControl within each line. (F) Time course western blot for ATF4 and its upstream activators under activation by Tunicamycin (2 μg/mL) with and without MondoA in PSN-1 cells. Relative protein abundance is compared to maximal tunicamycin-induction at 8 h in siControl treated cells (normalized to βActin). Statistical significance analyzed as in Fig. 1.

To further examine the MYC dependence on MondoA and the UPR, we assessed the three major UPR pathways in PSN-1 cells subsequent to Tm treatment and knockdown of MYC and MondoA, alone or in combination. While Tm induced all three expected downstream UPR transcriptional effectors (ATF4 protein translation, ATF6 cleavage, and XBP-1 splicing, reviewed in refs. 8 and 12), ATF4 alone required the combined presence of MYC and MondoA for maximal expression, even with Tm induction (Fig. 4C). This combinatorial requirement was also observed for the ATF4-downstream target CHOP, whereas GADD34, which is also an ATF4 target, was repressed directly by MondoA as seen by comparing the first and third lanes (Fig. 4C). These data support a coordinated output of the MYC network (shown in Fig. 4D), via the ATF4 arm of the UPR and the ISR, where MondoA plays an obligatory role in the induction of ATF4 under MYC-initiated stress.

To further test the function of the ISR in this system, we determined the effects of knocking down DEGs down-regulated (ARRDC4, TXNIP), or up-regulated (ATF4, PPP1R15A; encoding GADD34) following MondoA-MLX perturbation (Fig. 3) to gauge their effect on viability (SI Appendix, Fig. S4A). siRNAs against ARRDC4, TXNIP, or GADD34 had only nonsignificant or borderline effects on cell growth, while siRNA targeting ATF4 markedly compromised viability in PSN-1 cells to a similar extent as siMondoA. The ATF4 dependence of these cells is consistent with the increased reliance on the ISR when MYC is amplified. Taken together, our expression and biological data indicate that MondoA-MLX functions within the ISR by both repressing and activating ISR components to regulate metabolic adaptation to stress.

Loss of MondoA Leads to Decreased ATF4 Protein Expression.

As ATF4 is the primary downstream stress responsive transcription factor cooperatively affected by perturbation of MYC and MondoA (Fig. 4C and SI Appendix, Fig. S4A), we further characterized their crosstalk with the ISR, which regulates ATF4. The ISR responds to a variety of cellular stressors and, through stress-induced downstream phosphorylation of eIF2α, inhibits global protein synthesis while permitting preferential translation of the master-regulator ATF4 through its normally inhibitory uORF2 (19) (Fig. 4D). In both the PSN-1 and DAN-G cell lines, we found ATF4 mRNA upregulated ~1.5-fold upon siMondoA treatment (Fig. 4E). Furthermore, eIF2α phosphorylation is readily induced following Tm treatment in both siControl and siMondoA treated PSN-1 cells (Fig. 4F). However, ATF4 protein levels were significantly reduced after Tm challenge in siMondoA cells relative to siControls (Fig. 4F, replicate quantification SI Appendix, Fig. S4B). To test whether this phenomenon contributes to the diminished viability phenotype seen with MondoA loss, we measured cell confluency with either individual siRNA knockdown of MondoA or ATF4, overexpression of ATF4, or a combination of MondoA knockdown and ATF4 overexpression (Fig. 5A). While loss of ATF4 leads to a similar decrease in growth as MondoA knockdown, ectopic overexpression of ATF4 was sufficient to rescue growth of siMondoA treated cells (Fig. 5A, statistics in SI Appendix, Fig. S4C). Rescue was also observed in a clonogenicity assay in ATF4 overexpressing PSN-1 cells treated with MondoA inhibitors SBI-477 or SBI-993 (Fig. 5 B and C). Together, these data indicate that proper induction of ATF4 protein and its subsequent downstream program is a significant driver of MondoA dependency in the PSN-1 cells.

Fig. 5.

Assessing the role of ATF4 as a MondoA effector. (A) Incucyte growth curves (over 72 h) for PSN-1 cells with siMondoA, siATF4, combination siMondoA and ATF4 overexpression vector (ATF4oe) (all n = 8), or ATF4oe alone (n = 7). Data plotted as mean percent plate confluency. Significance values (one-way ANOVA) shown in SI Appendix, Fig. S4C. (B) Clonogenicity assay after 12 d with control or ATF4oe vector seeded at 100 cells and treated with indicated amounts of SBI-477 in four wells or (C) with indicated amounts of SBI-993 in three wells. Data expressed as colonies per well, significant differences (one-way ANOVA) expressed as compared to DMSO treatment. For (B) comparison between vector control and ATF4oe at both 10 and 20 µM P < 0.0001. For (C) comparison between vector control and ATF4oe at 5 µM is P < 0.0001. (D) Viability assay for PDAC cells treated with DMSO or 50 µM Rhein for 72 h. Data expressed as mean cell count, significant differences (n = 4, one-way ANOVA) expressed as compared to DMSO treatment. (E) Viability assay for PSN-1 (n = 13) or DAN-G (n = 3) cells with siControl or siFTO treatment for 72 h read as percent of siControl cell count. Significant differences (two-tailed t test) shown compared to siControl within lines. (F) Relative global RNA m⁶A modification level as determined by immunofluorescent corrected total cell fluorescence (CTCF) for PSN-1 cells treated with indicated amounts of SBI-993 (Top) or siControl, siFTO, siMondoA (Bottom) for 72 h. Significant differences (one-way ANOVA) compared to siControl or DMSO. (G) Western blots to validate treatment target readout in cells treated with either siFTO or SBI-993 for 72 h. Relative protein abundance expressed compared to control treatment (siControl or DMSO) and normalized to γTubulin. Statistical significance analyzed as in Fig. 1.

In order to gain mechanistic insight into the connection between ATF4 dysregulation and MondoA loss, we surveyed our RNA-seq signatures for perturbation of pathways potentially regulating ATF4, focusing on m⁶A-dependent RNA methylation which has been previously implicated in ATF4 translation (20). N6-methyladenosine (m⁶A) modification of RNA is regulated by a methyltransferase complex (writers), demethylase enzymes (erasers), and various RNA binding proteins (readers). Dysregulation of this machinery has been shown to impact every stage of RNA processing in a wide range of tumor cells, including pancreatic cancer where dysregulation of the m⁶A pathway is linked to tumor progression and maintenance (reviewed in refs. 21 and 22). As shown in the box plots in SI Appendix, Fig. S4D, multiple components of the m⁶A pathway are altered upon inhibition of MondoA-MLX via siRNA and/or SBI-477, most notably down-regulation of the primary mRNA demethylase FTO in all three conditions. Inhibition of RNA demethylases ALKBH5 and FTO by the inhibitor Rhein (Fig. 5D) and siRNA knockdown of FTO (Fig. 5E) led to a significant loss of viability in the PSN-1 and DAN-G cells. We also observed a global 1.5- to 2-fold increase in m⁶A RNA modification by immunofluorescent staining in cells with siRNA knockdown of either FTO or MondoA, as well as a dose-dependent increase in cells treated with SBI-993 (Fig. 5F, representative images SI Appendix, Fig. S4E). As m⁶A levels in RNA have been shown to vary across the cell cycle (23), we surveyed global m⁶A modification via flow cytometry and observed a dose-dependent increase in RNA m⁶A modification independent of cell cycle stage when treated with SBI-993 (SI Appendix, Fig. S4F).To test whether perturbation of m⁶A machinery was sufficient to cause ATF4 protein loss, we determined ATF4 levels in PSN-1 cells and observed a significant loss of ATF4 protein both with siFTO and, in a dose-dependent manner, with SBI-993 treatment (Fig. 5G and SI Appendix, Fig. S4 G and H). We conclude that MondoA is necessary for the induction of the major ISR effector ATF4 at the protein level likely through crosstalk with the machinery of the mRNA m⁶A epitranscriptomic modification network.

Targeting MondoA in Patient-Derived Organoids (PDOs).

PDAC PDOs recapitulate many features of the original disease and are suitable for testing personalized medicine and drug sensitivity (24). Compared to normal pancreatic organoids, PDAC PDOs exhibit a gene expression signature that includes activation of MYC and E2F targets, concomitant with suppression of the TNFα pathway (25). As these pathways are affected by MondoA inhibition in the PSN-1 cells, we investigated the feasibility of targeting MondoA via SBI-993 in PDAC PDOs. Six of the seven tested PDOs are sensitive to SBI-993 treatment (Fig. 6A, all images SI Appendix, Fig. S5A) with faster growing PDOs exhibiting higher sensitivity to MondoA inhibition. Sensitivity to SBI-993 did not correlate either positively or negatively with protein expression levels of MYC, MondoA, ChREBP, or ATF4 (quantification in SI Appendix, Table S7). Representative insensitive and sensitive organoid images and growth curves are shown in Fig. 6 B and C and SI Appendix, Fig. S5B. Importantly, we confirmed target engagement in one of the sensitive lines, as the MondoA target TXNIP is decreased in a dose-dependent manner by SBI-993 (Fig. 6D).

Fig. 6.

Effect of MondoA inhibition in PDOs and relationship to TCGA patient data. (A) Final organoid area µm²±SEM for PDAC PDOs after indicated doses of SBI-993 for 7 d. Significant differences (two-way ANOVA) expressed for 20 µM dose compared to DMSO treatment within PDOs. (B) Representative Incucyte final images for SBI-993 insensitive PDAC PDO 174 and (C) SBI-993 sensitive PDAC PDO 154 with 20 µM of SBI-993. (Scale bar, 200 µm.) (D) Western blot to confirm SBI-993 target engagement as determined by expression of the MondoA target TXNIP in PDO 154 with SBI-993 for 3 d. Relative protein abundance shown is relative to DMSO treatment and normalized to γTubulin. Significant differences (n = 3, one-way ANOVA) are compared within doses of SBI-993. (E) Intersection between TCGA signature of DEGs in MYC-amplified PDAC and RNA-seq datasets following siMondoA or SBI-477 treatment as generated in Fig. 3 A–C. (F) Kaplan-Meier survival curve comparing patients with low (red) vs. high (blue) expression of the combined gene set generated in Fig. 6E. (G) Kaplan-Meier survival curves comparing patients with low (red) vs. high (blue) expression of the combined gene set generated in Fig. 6E. Statistical significance analyzed as in Fig. 1.

Translational Relevance of MondoA in PDAC Clinical Data.

To further establish the relevance of MondoA-regulated pathways and targets in MYC-amplified PDAC, we queried the TCGA for expression data in patients with MYC amplification. We juxtaposed our RNA-seq datasets of either siRNA against MondoA or chemical inhibition of MondoA in the PSN-1 cells against the significantly enriched transcripts up- or down-regulated with MYC amplification in the patient data (Fig. 6E). This yielded signatures for MondoA-dependent putative MYC-regulated targets of interest: those of which are up-regulated with MYC amplification, but down-regulated by MondoA inhibition, as well as those repressed by MondoA, and enriched in the non-MYC-amplified patients (SI Appendix, Table S3). These “signatures” were then assessed for survival differences in the patient data of the pdacR database (26). As shown in Fig. 6 F and G (Kaplan–Meier plots), these four signatures predict overall survival, with high expression of MondoA/MYC-amplified positively regulated genes correlating with poor survival, and high levels of the repressed genes correlating with better survival. These data are consistent with a role for MondoA in maintaining the MYC-amplified transcriptional program by enabling proper activation and repression via MYC in patients.

Discussion

In this report, we have characterized functional coordination between MondoA and other components of the MYC network which directly contribute to MYC-driven tumorigenesis and survival in PDAC. We have found that MYC amplification or ectopic expression is synthetically lethal with loss of MondoA-MLX. Suppression of MondoA expression, or abrogation of MondoA transcriptional activity by means of a small molecule inhibitor, leads to decreased protein expression of the critical stress responsive transcription factor ATF4, altered MYC-ATF4 shared target gene expression, and cell death. Importantly, viability is rescued by the enforced expression of ATF4 in cells lacking MondoA expression or transcriptional activity. This work highlights a critical point of intersection and dependency between MYC, MondoA, and the ATF4-driven ISR.

Our initial report of MYCN-MondoA cooperation in neuroblastoma suggested that MondoA is broadly required for MYCN function at a significant proportion of shared target genes, many of which regulate metabolic processes (3). Our present study using PDAC cells implicates both direct coordination at a limited subset of shared promoters (e.g., TXNIP, GADD34) as well as a dependency upon MondoA for full genomic engagement of MYC and its extended network, even at promoters not normally bound directly by MondoA. This is consistent with a report of overlapping MYC and MondoA genomic binding in B-ALL, in which loss of MondoA profoundly affects MYC genomic occupancy, despite normal occupation of MondoA at only a fraction (<20%) of total MYC-bound promoters (27). Such indirect regulation could be due not just to the loss of MondoA but also to the lack of ATF4 or other MondoA-regulated proteins that act as transcriptional cofactors with MYC at shared targets (9). In addition, we observed a reconfiguration of MNT and ChREBP genomic occupation in response to loss of MondoA, changes that may be either secondary to, or directly contribute to, altered MYC binding.

Among the genes whose function is regulated by MondoA is FTO, an important metabolic regulator in obesity and diabetes (28), as well as a modulator of chemotherapy-response and bZIP transcription factor activity in cancer (29). FTO is a therapeutically targetable oncogene in PDAC, where loss of function by knockdown or chemical inhibition impairs growth in vitro and in vivo (22, 28, 30). In those studies the role of MYC in the sensitivity of PDAC cells to loss of FTO expression or activity was not addressed. Nonetheless, our work would suggest that MYC amplification does not lead to resistance to such insults. The work presented here also implicates ATF4 as an important downstream effector of FTO in PDAC. It joins a host of other factors associated with adaptive response and cellular plasticity such as hypoxia, EMT, and pro-survival pathways mediating tumor progression and therapy resistance, all targeted by FTO and m⁶A in pancreatic cancer (22, 30).

This work supports a model in which the MLX arm of the MYC network modulates MYC transcriptional output, mediating both hypo- and hyper-activation of MYC targets. MondoA influences MYC and MNT co-occupancy at both shared and nonshared genomic targets, most likely through heterodimerization with MLX, thereby limiting the availability of MLX for dimerization with MNT or ChREBP and affecting cooperative vs. antagonistic interactions with MYC-MAX. Thus the balance between MLX- and MAX-containing heterodimers may act as a context-dependent source of gene regulation by the MYC network. The ability to modulate MYC levels (through changes in protein abundance as well as genomic ecDNA copy number) has been shown to drive PDAC plasticity (31). We surmise that functional interactions within the network contribute to inherent plasticity at the genomic level, where the ability to constrain MYC output can act as a survival advantage. In MYC-amplified cells, proteostatic stress must be curbed by the remedial ISR to accommodate diminished nutrient availability or lack of metabolic sufficiency. Thus, MondoA’s role in fine-tuning MYC network dynamics and permitting ATF4 protein expression for the remediation of stress is critical for cell survival. A diagram summarizing these roles for MondoA in MYC-amplified PDAC is shown in Fig. 7.

Fig. 7.

A model for MYC network transcriptional and genomic coordination with ATF4 in MYC-amplified PDAC. Loss of MondoA leads to alterations in expression of direct MondoA-bound targets, as well as secondary effects on the genomic engagement of MYC network members, even at nonbound promoters. MYC and MNT binding at UPR/TNFα pathway targets is decreased, whereas the MondoA-paralog ChREBP is enhanced. This co-occurs with loss of ATF4 protein, which may contribute to lack of MYC and MNT engagement at critical shared targets. This ultimately unbalances the tumorigenic and apoptosis-promoting effects of amplified MYC, resulting in nonremediated stress and cell death.

Our findings raise the possibility that inhibition of MondoA by means of a chemical inhibitor such as SBI-477/SBI-993 might be used to arrest the growth of PDAC and other neoplasms addicted to high levels of MYC while sparing cells with tonic MYC expression. Indeed MYC protein is present in both classical and basal PDAC as well as in normal pancreatic epithelium. A precision medicine approach to MondoA inhibition will require a deeper understanding of what levels and in what contexts MYC protein abundance activates the ISR to impose dependence on MondoA and its downstream effectors. A related approach would be to target factors downstream of MondoA critical to the maintenance of dysregulated growth of tumor cells, a concept supported by our findings that the MYC-regulated/ MondoA-dependent gene expression signature significantly correlates with patient survival. In summary, our work indicates that MYC’s role in driving PDAC occurs within the context of the extended MYC network and that further studies on the interactions and dynamics of the network are warranted in order to more fully understand and control pancreatic neoplasia.

Materials and Methods

All mice used in the study were housed and treated according to the guidelines provided by the Fred Hutch Institutional Animal Care and Use Committee (IACUC) protocol number: PROTO201900024. Complete Materials and Methods are included as a SI Appendix. Standard molecular biology and cell culture techniques were used, and kits and reagents were utilized following the manufacturer’s recommendation. Statistical comparisons and the tests for significance used for each experiment are indicated in the corresponding figure legends.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

RNA-seq; Cut&Run Genomic occupancy data have been deposited in GEO (GSE316429; GSE316427; and GSE316426) (32–34).