Targeting MYC effector functions in pancreatic cancer by inhibiting the ATPase RUVBL1/2

Abstract

Objective: The hallmark oncogene MYC drives the progression of most tumours, but direct inhibition of MYC by a small-molecule drug has not reached clinical testing. MYC is a transcription factor that depends on several binding partners to function. We therefore explored the possibility of targeting MYC via its interactome in pancreatic ductal adenocarcinoma (PDAC).

Design: To identify the most suitable targets among all MYC binding partners, we constructed a targeted shRNA library and performed screens in cultured PDAC cells and tumours in mice.

Results: Unexpectedly, many MYC binding partners were found to be important for cultured PDAC cells but dispensable in vivo. However, some were also essential for tumours in their natural environment and, among these, the ATPases RUVBL1 and RUVBL2 ranked first. Degradation of RUVBL1 by the auxin-degron system led to the arrest of cultured PDAC cells but not untransformed cells and to complete tumour regression in mice, which was preceded by immune cell infiltration. Mechanistically, RUVBL1 was required for MYC to establish oncogenic and immunoevasive gene expression identifying the RUVBL1/2 complex as a druggable vulnerability in MYC-driven cancer.

Conclusion: One implication of our study is that PDAC cell dependencies are strongly influenced by the environment, so genetic screens should be performed in vitro and in vivo. Moreover, the auxin-degron system can be applied in a PDAC model, allowing target validation in living mice. Finally, by revealing the nuclear functions of the RUVBL1/2 complex, our study presents a pharmaceutical strategy to render pancreatic cancers potentially susceptible to immunotherapy.

Keywords: GENE REGULATION; GENE TARGETING; IMMUNOTHERAPY; ONCOGENES; PANCREATIC CANCER.

Figure 1

Genetic dropout screens reveal differential dependence on MYC binding partners in pancreatic ductal adenocarcinoma (PDAC) in vitro and in vivo. (A) Schematic of the in vitro (KPC culture) and in vivo (KPC allograft) dropout screens for 91 MYC binding partners. (B) Waterfall plot of the in vitro screen in KPC cells showing the doxycycline (Dox)-induced changes in abundance of 478 shRNAs (log2FC±SEM, n=3). shRNAs are grouped by target gene, sorted by the median change for the five shRNAs. Selected genes of interest, including MYC, are marked. The essentiality score is the percentage of human cancer cell lines in the DepMap portal that depend on each gene for viability. NTC1–4 correspond to groups of four or five non-targeting control shRNAs. (C) Volcano plot of the in vivo screen showing changes in abundance (enrichment) for 460 targeting shRNAs (black dots) and 18 non-targeting shRNAs (green dots), between five tumours from mice fed standard chow (here, ‘Vehicle’) and transplanted KPC cells (Input). Wald test p values for five mice. (D) Volcano plot of the in vivo screen showing changes in abundance (enrichment) for 460 targeting shRNAs (black dots) and 18 non-targeting shRNAs (green dots), between five tumours from mice fed doxycycline-containing chow (Dox) and five tumours from mice fed standard chow (Vehicle). Wald test p values. (E) Waterfall plot of the in vivo screen comparing integrated shRNA abundance in tumours of five mice fed doxycycline-containing chow (Dox) and five mice fed standard chow (Vehicle). Bars represent the median log2FC for five shRNAs per gene, and error bars represent the SEM. (F) Heatmap of the in vivo screen. The chromatic scale indicates the median change in integrated shRNA abundance (log2FC) per gene relative to vehicle-treated mice. Each column corresponds to one mouse fed either standard chow (Vehicle) or doxycycline-containing chow (Dox). Genes are sorted by the median change for five Dox-treated mice. (G) Scatter plot comparing the results of the in vitro (cultured KPC cells) and in vivo (allografted KPC cells) screens. Dots represent the median change (log2FC) of five shRNAs per gene and are colour coded according to the essentiality score. NTC1–4, groups of four or five NTC shRNAs. (H) Scatter plot as in panel G. Red, TIP60 complex components. Green, NTC shRNAs. (I) Scatter plot as in panel G. Red, genes that are more essential in KPC cells than in NIH3T3 cells (values are reported in online supplemental table 2). Green, NTC shRNAs. See also online supplemental figure S1.

Figure 2

Expression of MYC and RUVBL1 correlate in pancreatic ductal adenocarcinoma (PDAC) and high levels are associated with aggressive tumours. (A) Scatter plot comparing expression of RUVBL1 and MYC target gene expression (mean of all HALLMARK MYC TARGET V1 genes after scaling expression (FPKM) across all TCGA patients) in patients with human PDAC from the TCGA database (r, Pearson’s correlation coefficient; p value, unpaired t-test, n=159). (B) Exemplary immunohistochemistry of RUVBL1 from a tissue core from a tissue microarray containing 31 individual human PDAC specimens. A section with high RUVBL1 expression and a zoom-in with PDAC (P) and stromal tissue (S) are shown (scale: 200 µm). The panel is also shown as part of online supplemental figure S2C. (C) Quantification of RUVBL1 expression in a tissue microarray containing 31 sections of human PDAC specimens and 24 sections of benign acinar tissue as in panel B. RUVBL1 expression was scored as negative, low, medium and high in PDAC, adjacent stroma and non-malignant ductal and acinar tissue. The ratio of tissues with high RUVBL1 expression is shown. n, sample size. (D) Kaplan-Meier survival curves for patients with PDAC stratified into groups of low and high expression of RUVBL1 and MYC target genes (mean of all MYC TARGET V1 genes after scaling expression (FPKM) across all TCGA patients). P value, log-rank test. See also online supplemental figure S2.

Figure 3

RUVBL1 is essential for DNA replication and growth of pancreatic cancer cells. (A) Ruvbl1 knock-in strategy for auxin-inducible degron (AID) tagging showing the elements of the knock-in cassette and the architecture of Ruvbl1. Arrows indicate the position of primers used to identify recombined cell clones by PCR (see online supplemental figure S3B). (B) Immunoblots of wild-type KPC and KPC ^AID-Ruvbl1 cell lysates probed with antibodies against RUVBL1 or the AID tag. (C) Immunoblot of KPC ^{AID-Ruvbl1; TIR1} cells treated with various concentrations of auxin for 3 or 6 hours. Vinculin, loading control. (D) Immunoblot of KPC ^{AID-Ruvbl1; TIR1} cells treated with 1 µM auxin over time. TIR1^F74G was detected with an antibody against the MYC tag. Vinculin, loading control. (E) Quantification of immunoblots of RUVBL1 after 6 hours of 1 µM auxin treatment (n=5, mean±SD, unpaired t-test). rU, relative units. (F) Logarithmic growth curve of KPC ^AID-Ruvbl1 cells expressing or not expressing TIR1^F74G. Cells were treated daily with 1 µM auxin or vehicle, and growth was followed for 9 days in biological triplicates (n=3, mean±SD). (G) Immunoblot of KPC ^{AID-Ruvbl1; TIR1} cells treated with 1 µM auxin or 1 µM CB-6644 for indicated time points. Vinculin, loading control. (H) BrdU-PI flow cytometry scatter plots of KPC ^{AID-Ruvbl1; TIR1} cells after treatment with 1 µM auxin or 1 µM CB-6644 for 24 or 48 hours. Cells were labelled with BrdU for 1 hour. (I) Quantification of S-phase cells with high or low BrdU incorporation as shown in panel H. The experiment was performed in biological duplicates (n=2, mean). See also online supplemental figure S3.

Figure 4

RUVBL1 redirects transcription from immune genes to growth genes. (A) Gene set enrichment analysis (GSEA) plots of selected RUVBL1-activated and RUVBL1-repressed gene sets. GSEA was performed on SLAM-seq data of KPC ^{AID-RUVBL1; TIR1} cells. Cells were treated with 1 µM auxin or DMSO (Vehicle) for 15 hours, followed by 800 µM 4sU for 2 hours. The normalised enrichment score (NES) is positive for gene sets activated by and negative for those repressed by RUVBL1 (FDR, false discovery rate). (B) GSEA of KPC ^{AID-Ruvbl1; TIR1} cells treated with 1 µM auxin or DMSO for 15 hours and 800 µM 4sU. The NES and the q-value are shown for all enriched gene sets (positive NES: gene sets activated by RUVBL1, negative NES: gene sets repressed by RUVBL1). (C) NES values for different hallmark gene sets and a gene set containing primary MYC targets defined in Muhar et al, compared between the 3 and 15-hour durations of RUVBL1 depletion. (D) Scatter plot comparing gene regulation after auxin-induced degradation and CB-6644 inhibition of RUVBL1. KPC ^{AID-Ruvbl1; TIR1} cells were treated with 1 µM auxin for 15 hours (n=3) or 1 µM CB-6644 for 24 hours (n=3). Gene expression was analysed by SLAM-seq. Changes (log2FC) in total RNA versus DMSO (vehicle)-treated cells are shown (r, Pearson’s correlation coefficient; p value, unpaired t-test). (E) Genome browser track of RUVBL1 chromatin immune precipitation (ChIP)-seq signal in KPC ^{AID-Ruvbl1; TIR1} cells treated with vehicle or 1 µM auxin for 15 hours. Binding to the Rpl8 gene is shown as spike-normalised reads and compared with the input signal as control. (F) Density plots of RUVBL1 ChIP-seq signals around transcription start sites (TSS). Averaged binding (RPKM) of RUVBL1 in KPC ^{AID-Ruvbl1; TIR1} cells treated with DMSO (vehicle) or 1 µM auxin for 15 hours compared with the input signal. (G) Bar graph of the genomic distribution of RUVBL1 and MYC peaks. Chromatin binding of RUVBL1 and MYC was analysed by ChIP-seq and was compared with a set of random genomic intervals in promoters (TSS− 3000 bp to TSS+ 3000 bp), gene bodies (TSS+ 3000 bp to transcription end site (TES)), regions downstream of genes (TES to TES+ 2000 bp) and intergenic regions. (H) Scatter plot comparing the promoter occupancy of MYC and RUVBL1 (spike-normalised reads) as measured by ChIP-seq. r, Pearson’s correlation coefficient; p value, unpaired t-test. (I) Bin plot comparing gene regulation due to acute RUVBL1 depletion with RUVBL1 and MYC binding (spike-normalised reads) at gene promoters. Genes were binned into eight equally distant bins of gene regulation on 15 hours of RUVBL1 depletion (‘Regulation’). Mean regulation per bin was plotted against mean promoter occupancy by MYC and RUVBL1 in unperturbed KPC ^{AID-Ruvbl1; TIR1} cells (mean±SEM). r, Pearson’s correlation coefficient; p value, unpaired t-test. NFkB, nuclear factor kappa B; TGFβ, transforming growth factor-beta; TNFα, tumour necrosis factor alpha. See also online supplemental figure S4.

Figure 5

RUVBL1 is an essential cofactor of MYC. (A) Genome browser tracks of RUVBL1 and MYC chromatin immune precipitation (ChIP)-seq signal in A375 ^MYC-AID cells treated with 1 µM auxin for 3 hours or vehicle control. RNA polymerase II (RNAPII) tracks of U2OS cells (GSE162264) are shown. Binding to the EIF4A1 and TBRG4 genes is shown as spike-normalised reads and compared with the input signal as control. In the TBRG4 promoter, one of two RUVBL1 peaks located at an MYC-negative region does not decrease on auxin-mediated depletion of MYC. (B) Rank plot of RUVBL1 and MYC ChIP-seq signal in A375 ^MYC-AID cells treated with 1 µM auxin for 3 hours or vehicle control. All MYC and RUVBL1-bound promoters are sorted for decrease in RUVBL1 chromatin binding on acute MYC depletion and plotted as log2FC (y-axis, black). Mean MYC binding is plotted for 15 equally sized bins containing the same genes (y-axis, blue). (C) Scatter plot of SLAM-seq data comparing MYC-induced gene expression changes in the absence or presence of CB-6644. KPC ^MYC-ER cells were treated with DMSO or 1 µM CB-6644 for 20 hours followed by ethanol or 200 nM 4-hydroxytamoxifen (OHT) for 4 hours. Changes in total RNA (log2FC) are shown (n=3). The slope (m) and p value (p) of the linear regression (blue) are indicated as is the Pearson’s correlation coefficient (r). A line with slope m=1 is shown in black. (D) Heatmap of SLAM-seq data from cells treated as in panel C. Biological replicates are labelled 1, 2 and 3. Changes in total RNA (log2FC) are shown. (E) Sucrose gradient ultracentrifugation demonstrating a physical interaction between MYC and RUVBL1/2. Purified recombinant His6-MBP-MYC^1-163 (MYC^1-163, top panel), RUVBL1/2 (middle panel) and all three proteins together (bottom panel) were subjected to sucrose gradient ultracentrifugation. Fractions were collected and analysed by SDS-PAGE, followed by Coomassie blue staining. AID, auxin-inducible degron. See also online supplemental figure S5.

Figure 6

RUVBL1 is required for the maintenance and progression of pancreatic cancer. (A) Immunoblot of pancreatic tumour lysates. Native KPC cells and KPC ^{AID-Ruvbl1; TIR1} cells were transplanted into one and seven mice, respectively. After 18 days, mice with KPC ^{AID-Ruvbl1; TIR1} cell allografts were treated with 20 mg/kg auxin for up to 72 hours or with vehicle for 72 hours. Lysates of tumours were analysed using an anti-RUVBL1 antibody. Lysates of untreated and auxin-treated cultured KPC ^{AID-Ruvbl1; TIR1} cells were loaded for comparison. The band between WT and AID-tagged RUVBL1 can be attributed to murine immunoglobulin G (IgG) in the tissue. Vinculin, loading control. (B) Schematic of the in vivo RUVBL1 depletion experiment. (C) Bioluminescence images of mice with pancreatic tumours. Luciferase-expressing KPC ^{AID-Ruvbl1; TIR1} cells were transplanted into mice, and tumour size on day 7 was used to form pairs of animals with similar size tumours. Then, mice were treated daily with 20 mg/kg auxin (n=7) or vehicle (n=7). Tumour growth was assessed by bioluminescence imaging at the indicated time points. (D) Line plot of the relative volume of murine pancreatic tumours, estimated from bioluminescence as in panel C. Vehicle, n=7. Auxin, n=7. Values are mean±SEM. *n=6. (E) Kaplan-Meier survival curves for mice with pancreatic tumours treated with auxin (n=7) or vehicle (n=7). P value, Log-rank test. (F) Immunoblot of TIR1^F74G in lysates of tumours excised from mice that reached the humane endpoint. After KPC ^{AID-Ruvbl1; TIR1} cell engraftment, mice were treated daily with 20 mg/kg auxin or vehicle for up to 28 days. TIR1^F74G was detected with an antibody against the MYC tag. GAPDH, loading control. See also online supplemental figure S6.

Figure 7

RUVBL1 promotes immune evasion in pancreatic ductal adenocarcinoma (PDAC). (A) Immunohistochemical staining of RUVBL1, KI67, BrdU and CD3 in sections of KPC ^{AID-Ruvbl1; TIR1} tumours from mice treated with vehicle (n=4) or 20 mg/kg auxin (n=4) for 5 days. Mice received an injection of BrdU and 2.5 hours later were killed to harvest the tumours. Quantification of positively stained cells is depicted below (n=4, unpaired t-test). (B) Kaplan-Meier survival curves of mice bearing KPC ^{AID-Ruvbl1; TIR1} tumours treated with 10 mg/kg anti-PD-1 antibody (αPD-1) two times per week, alone (n=10) or with 20 mg/kg auxin daily (n=11). Vehicle-treated group (n=7) from figure 6E. (C) Dose–response curves of CB-6644 on cell viability (resazurin assay) in native KPC cells and KPC cells overexpressing RUVBL1^WT or RUVBL1^A62T. Cells were treated with CB-6644 for 72 hours (n=3, mean±SD). (D) Immunoblot of MYC and RUVBL1 in a panel of murine PDAC cell lines with the indicated genetic mutations representing different PDAC subtypes. (E) Scatter plot of MYC and RUVBL1 protein levels as in panel D (n=2; r, Pearson’s correlation coefficient; p value, unpaired t-test). (F) Scatter plot of MYC/RUVBL1 protein levels as in panel D and sensitivity to treatment with CB-6644 for 72 hours (n=2; r, Pearson’s correlation coefficient; p value, unpaired t-test). (G) Bioluminescence images of mice with pancreatic tumours. Luciferase-expressing KPC ^{AID-Ruvbl1; TIR1} cells were transplanted into mice, and tumour size on day 7 was used to form pairs of animals with similar size tumours. Then, mice were treated two times per day with 25 mg/kg CB-6644 (n=5) or vehicle (n=5). Tumour growth was assessed by bioluminescence imaging at the indicated times. (H) Line plot of the relative volume of pancreatic tumours in mice treated with CB-6644 (n=5) or vehicle (n=5), as in panel G. Values are mean±SEM. (I) Kaplan-Meier survival curves for mice with pancreatic tumours treated with CB-6644 (n=5) or vehicle (n=5), as in panel G. Log-rank test. (J) Line plot of the relative volume (mean±SEM) of pancreatic tumours in mice treated with CB-6644 (n=8), αPD-1 (n=8), a combination of αPD-1 and CB-6644 (n=8) or vehicle (n=8). KPC cells were transplanted into mice, and tumour size on day 7 was used to form pairs of animals with similar size tumours. Mice were treated two times per day with 25 mg/kg CB-6644, two times per week with 10 mg/kg αPD-1, a combination of both or vehicle for up to 28 days. (K) Kaplan-Meier survival curves for mice with pancreatic tumours treated with CB-6644 (n=8), αPD-1 (n=8) or a combination of both (n=8), as in panel J. Log-rank test. See also online supplemental figure S7.