BET inhibition decreases HMGCS2 and sensitizes resistant pancreatic tumors to gemcitabine

Abstract

Efforts to develop targetable molecular bases for drug resistance for pancreatic ductal adenocarcinoma (PDAC) have been equivocally successful. Using RNA-seq and ingenuity pathway analysis we identified that the superpathway of cholesterol biosynthesis is upregulated in gemcitabine resistant (gemR) tumors using a unique PDAC PDX model with resistance to gemcitabine acquired in vivo. Analysis of additional in vitro and in vivo gemR PDAC models showed that HMG-CoA synthase 2 (HMGCS2), an enzyme involved in cholesterol biosynthesis and rate limiting in ketogenesis, is overexpressed in these models. Mechanistic data demonstrate the novel findings that HMGCS2 contributes to gemR and confers metastatic properties in PDAC models, and that HMGCS2 is BRD4 dependent. Further, BET inhibitor JQ1 decreases levels of HMGCS2, sensitizes PDAC cells to gemcitabine, and a combination of gemcitabine and JQ1 induced regressions of gemR tumors in vivo. Our data suggest that decreasing HMGCS2 may reverse gemR, and that HMGCS2 represents a useful therapeutic target for treating gemcitabine resistant PDAC.

Keywords: BET bromodomain inhibitor (BETi); Gemcitabine resistance (gemR); HMG-CoA synthase 2 (HMGCS2); Pancreatic cancer; Patient-derived xenograft (PDX) models.

Figure 1.. Gemcitabine resistant (gemR) PDAC cells harbor properties associated with a metastatic phenotype and metabolic changes, compared to parent cells.

(A) Panc1.gemR and MiaPaCa2.gemR cells are less sensitive to gemcitabine than parent counterparts. Cell viability was assessed by alamarBlue assays after a 96 h exposure to gemcitabine. N=3 independent experiments. (B) Panc1.gemR cells show greater migration than parent cells in Boyden chamber assays. 10% FBS was used as the chemo-attractant. ***P<0.001, two-tailed unpaired t-test. (C) Panc1.gemR cells adhered to collagen I more than Panc1 parent cells. ****P<0.0001, two-tailed unpaired t-test. N=3. (D) MiaPaCa2.gemR cells adhered to fibronectin more than MiaPaCa2 parent cells. ***P<0.001, two-tailed unpaired t-test. N=3. (E, F) gemR cells have higher levels of glycolysis as reflected by ECAR, and higher levels of mitochondrial oxygen phosphorylation as reflected by OCR, compared to parent cells. Assays were performed using Seahorse XF396. ****P<0.0001, two-way ANOVA. N=2.

Figure 2.. PDAC gemR cell line and PDX models have higher levels of HMGCS2 than parent counterparts.

(A) IPA of RNA-seq data identified superpathway of cholesterol biosynthesis as top pathway in PA16.gemR PDX tumors most upregulated, compared to PA16 parent tumors. (B) qRT-PCR data identified six gene products in the cholesterol biosynthesis pathway that were upregulated in Panc1.gemR cells compared to parent cells. (C) Immunoblot (IB) show that gemR cell lines have higher levels of HMGCS2 than parent cell lines. Relative differences in expression measured by densitometry are below each image. (D) Immunohistochemistry (IHC) staining for HMGCS2 shows gemR tumors express higher levels of HMGCS2 than parent counterparts. Abdominal metastases of PA16.gemR tumors (gemR-met) harvested on day 241 during the generation of this gemR model express levels of HMGCS2 equivalent to subcutaneous primary PA16.gemR tumors. Expression indices (EI, bottom left corner yellow boxes) were calculated as published. Scores range from 0–300 and represent the % of cells expressing HMGCS2 x staining intensity (0–3). Bar=10μm. (E) IB data show that PA10.gemR and PA16.gemR tumors had higher levels of HMGCS2 protein than parent counterparts. Relative differences in expression as measured by densitometry are below each image. (F) No HMGCS2 was detected in human normal pancreas. (G) IHC data demonstrate no difference in BRD4 levels between parent and gemR PDX tumors. Bar=10μm. Expression Indices are shown in left bottom corner of each image. (H) IB data show no difference in BRD4 levels between parent and gemR cells.

Figure 3.. Ectopic expression of HMGCS2 decreases sensitivity to gemcitabine and increases adhesion, migration, and anchorage-independent colony growth in soft agar by Panc1 cells.

(A) IB for HMGCS2 in Panc1.neo vs Panc1.HMGCS2 transfectants. (B) In clonogenic assays, HMGCS2 transfectants were less sensitive to a 72 hrs exposure to gemcitabine than control transfectants. N=3. **P<0.01, ***P<0.001, by two-way ANOVA. (C,D,E) Ectopically expressed HMGCS2 increased adhesion to collagen (C), migration (D) and number and size of colonies in soft agar (E), two-tailed unpaired t-test. N=3.

Figure 4.. HMGCS2-targeted shRNA sensitizes gemR cells to gemcitabine.

A-E: MiaPaCa2.gemR transfectants; F-I: Panc1.gemR transfectants. (A) IB demonstrated that shHMGCS2 transfectants expressed lower levels of HMGCS2 than control transfectants. (B) alamarBlue and (C) clonogenic assays showed that shHMGCS2 transfectants were more sensitive to gemcitabine than shGFP transfectants. ****P<0.0001, two-way ANOVA. N=3. (D, E) shRNA-mediated downregulation of HMGCS2 in MiaPaCa2.gemR cells (D) inhibited anchorage-independent growth in soft agar, and (E) decreased migration in Scratch assays (****P<0.0001). Both methods are interpreted to reflect metastatic properties in vitro. Two-tailed t test. N=3 (mean ± SEM). (F) IB data show lower levels of HMGCS2 in Panc1.gemR cells transfected with shHMGCS2 compared to shGFP control transfectants. (G) alamarBlue and (H) clonogenic assays showed that transfection with shHMGCS2 sensitized gemR cells to gemcitabine. ****P<0.0001, two-way ANOVA. N=3. (I) shHMGCS2 transfectants adhered less to collagen I than control transfectants. ****P<0.0001, unpaired t-test. N=3.

Figure 5.. HMGCS2 expression is BRD4-dependent.

(A, B) ChIP of cell lysates with anti-BRD4 show that BRD4 interacts with the promoter sequence of HMGCS2 in BxPC3 (A) and Panc1.gemR (B) cells. JQ1 (1μm x 24 h) decreased this interaction. DNA that co-precipitated with BRD4 was quantitated using qPCR with primers that flanked the HMGCS2 promoter locus. Data were analyzed relative to the percent input and normalized to DMSO control. Rabbit IgG was a negative control. Data are shown as mean ± S.D. of precipitated DNA/input. One-way ANOVA, ****P<0.0001, N=3. (C, D) shRNA targeting BRD4 simultaneously decreased expression of BRD4 and HMGCS2 in MiaPaCa2.gemR (C) and Panc1.gemR (D) cells. (E, F) HMGCS2 co-precipitated with BRD4. Cell lysates were harvested, immunoprecipitated with BRD4 antibody and precipitates analyzed by immunoblot using an anti-HMGCS2 antibody. WCL: whole cell lysate as input.

Figure 6.. JQ1 inhibits clonogenicity and decreases HMGCS2 levels in parent and gemR pancreatic cancer cells in vitro.

(A,C) JQ1 inhibits colony growth in clonogenic assays. Cells were exposed to JQ1 for 72 h followed by culture in drug-free medium for another 7 days. **P<0.01, ***P<0.001, ****P<0.0001. One-way ANOVA followed by Tukey’s post test (Prism). Each bar represents >4 replicates in N=3 independent experiments. VC: vehicle control (DMSO). (B, D) Representative images of clonogenic assays for Panc1, MiaPaCa2 parent and gemR cells. (E, F) IB shows a dose-dependent inhibition of expression of HMGCS2 by JQ1 in Panc1, Panc1.gemR, MiaPaCa2 and MiaPaCa2.gemR cells. Cells were exposed to drug for 48 h. Whole cell lysates were used to perform IB for HMGCS2. DMSO (0.01%) was used as a control.

Figure 7.. JQ1 + gemcitabine (gem) induces regressions of PA16.gemR and PA10.gemR PDAC PDX tumors.

All mice received 100 mg/kg gemcitabine weekly until the combination study began. The combination regimen of JQ1 + gemcitabine began when tumor volumes reached ~250mm³ (A) Comparison of PA16.gemR tumor volumes on day 31 (dotted vertical lines indicate the start of JQ1 treatment) vs 46 (end of treatment) shows that JQ1 + gem induced tumor regressions (P<0.05), and that JQ1 + gemcitabine was more effective than gem alone (P<0.0001). N=9 tumors/group. (B) RNA-seq data showing the fold-decrease in genes associated with metastasis that are downregulated >5-fold by JQ1+gem compared to gem in PA16.gemR PDXs. Data were generated using tumors harvested on day 47 (24 hr after the last treatment) in the experiment in Panel A. (C) IHC staining for HMGCS2 in PA16.gemR tumors harvested on day 47 in panel A shows that JQ1 inhibits HMGCS2 expression in vivo. Bar=10μm. Expression Indices are shown in left bottom corner of each image. (D) IHC staining for HMGCS2 in PA10.gemR tumors harvested on day 64 (end of treatment) in panel G shows that JQ1 inhibits HMGCS2 expression in vivo. Bar=10μm. (E) IB comparing levels of HMGCS2 in PA16.gemR or PA10.gemR tumors exposed in vivo to gemcitabine compared to JQ1 + gemcitabine. PA16.gemR and PA10.gemR tumors were harvested day 47 and day 64, respectively, after initiation of gemcitabine treatment. Relative differences in expression measured by densitometry are below each image. (F) Comparison of tumor volumes on day 46 of experiment in panel A. Data were analyzed using two-way ANOVA or two-tailed t-test (Prism). (G) JQ1 + gemcitabine induced PA10.gemR tumor regressions, reflected by comparison of tumor volumes on day 50 (start of JQ1 treatment) vs day 64 (end of treatment) (P<0.01). N=9 tumors for JQ1 + gemcitabine group; N =3 tumors for gemcitabine group. (H) Comparison of quantitated tumor volumes on day 64 of experiment in panel G. Tumor volumes (mm³) = mean ± S.E.M. Analyses for in vivo study were done with two-way ANOVA or two-tailed t-test (Prism). (I, J) IHC data demonstrating that PA16.gemR (I) and PA10.gemR (J) tumors harvested from mice treated with JQ1 + gemcitabine had lower levels of the cell proliferation marker Ki67 than tumors of mice treated with gemcitabine alone. Data were generated with formalin-fixed paraffin-embedded (FFPE) tumor tissue. Left panels: IHC; right panels: bar graphs quantitating data in left panels (mean ± S.E.M.) ****P<0.0001, unpaired t-test (Prism).