Over-expression of RARγ and esophagus tissue-specific genes in pancreatic cancer is associated with a poor patient prognosis
First, to explore the expression of RARs in pancreatic cancer, we compared the transcript levels of RARs between normal pancreas tissues (N) and pancreatic cancers (C) by analyzing two databases: the TCGA and GTEx. Our database analyses indicated that the transcript levels of RARα and RARγ in pancreatic cancer (C) were significantly higher than in normal pancreatic tissue (N) (RARα, mean TPM 41.1 vs. 5.6, p < 0.0001; RARγ, mean TPM 36.6 vs. 4.2, p < 0.0001) and that the transcript level of RARβ was low in both pancreatic cancer (C) and normal pancreatic tissue (N) (mean TPM 8.5 vs. 0.7) (Fig. 1a). We further examined the protein expression of RARγ in PDAC, PanIN (precancerous lesion of PDAC) and adjacent normal pancreatic ductal epithelium by IHC staining of surgical specimens (Fig. 1b). IHC results demonstrated that the IHC scores of RARγ were significantly higher in PDACs and high-grade PanINs than in normal pancreatic ductal epithelium (normal pancreatic ductal epithelium vs. high-grade PanIN, p < 0.01; normal pancreatic ductal epithelium vs. PDAC, p < 0.05) (Fig. 1b).
We next analyzed the TCGA-PAAD data to explore the association between the expression of RARs and the patient prognosis. In TCGA analyses, the expression of RARα and RARβ did not correlate with the prognosis of pancreatic cancer patients (S-Figs. 1a, b). However, the high RARγ-expression group had a significantly worse prognosis than the low-expression group in pancreatic cancer (median OS 15.7 vs. 24.6 months, p = 0.0011) (Fig. 1c). These results suggested that the activation of RARγ signaling might contribute to PDAC progression.
We previously reported that RARγ signaling determines the differentiation lineage into the esophageal epithelium [10]. To reveal whether or not the expression of esophagus tissue-specific genes is elevated in PDAC through the activation of RARγ signaling, we investigated tissue-specific genes whose protein was expressed in PDAC using the HPA. Our analysis revealed that PDAC expressed many esophagus tissue-specific genes (12 of 28 genes) as well as other tissue-specific genes (7 of 15 adipose tissue-specific genes, 5 of 32 adrenal-specific genes, 4 of 5 gallbladder-specific genes, 3 of 24 stomach-specific genes, 2 of 13 lung-specific genes, 1 of 6 duodenum-specific protein, and 0 of 1 small intestine-specific gene) (S-Fig. 1c and Table S4). We further explored whether or not the expression of esophagus tissue-specific genes was associated with the patient prognosis, similar to RARγ. TCGA analyses showed that among 12 esophagus tissue-specific genes expressed by PDAC, the increased expression of 7 significantly correlated with a poor prognosis of PDAC patients (ECM1; median OS 17.3 vs. 23.0 months, p = 0.0141, KRT13; median OS 15.8 vs. 37.7 months, p < 0.0001, KRT6A; median OS 16.2 vs. 23.4 months, p = 0.0141, ERO1L; median OS 17.7 vs. 23.2 months, p = 0.0177, FGFBP1; median OS 17.7 vs. 23.4 months, p = 0.0052, PADI1; median OS 17.3 vs. 30.4 months, p = 0.0022, GJB2; median OS 16.6 vs. 35.3 months, p = 0.0002; all data are listed in order of high-expression group vs. low-expression group) (S-Figs. 2a-l). These results also supported our hypothesis that the activation of RARγ signaling drove the progression of PDAC.
Blockage of RARγ signaling suppressed the proliferation of PDAC cells
We designed in vitro experiments using PDAC cell lines to elucidate the function of RARγ signaling in PDAC.
First, to check whether or not the RARγ-antagonist LY2955303 (RARγi-1) could block RARγ signaling in a PDAC cell line (PK-1), we examined the expression of FABP5, a known target gene of RA signaling [9], and the expression of KRT13, considered to lie downstream of RARγ [9, 25], using qPCR. RARγi-1 dose-dependently decreased the transcript levels of both FABP5 and KRT13 (Figs. 2a, b). These findings indicated that RARγi-1 dose-dependently blocked RARγ signaling.
Next, to clarify the function of RARγ signaling in PDAC, we evaluated the effect of RARγ inhibition on the proliferation of a PDAC cell line (PK-1) in vitro. Cell proliferation assays revealed that the proliferation of PDAC cells significantly decreased in the presence of RARγi-1 compared to in its absence at 24 and 72 h after seeding (p < 0.05 at 24 h, n = 3; p < 0.01 at 72 h, n = 3) (Figs. 2c, d). In addition, the proliferation suppressive effect was dependent on the concentration of RARγi-1 (S-Fig. 3a). To exclude the possibility that some off-target effect of RARγi-1 contributed to the results of the experiments, we re-evaluated the effect of RARγ inhibition using another RARγ-antagonist MM11253 (RARγi-2) and knockdown of RARγ. RARγi-2 suppressed the proliferation of PDAC cells (p < 0.001, n = 3) (Figs. 2d, e), and its effect depended on the concentration of RARγi-2, similar to RARγi-1 (S-Fig. 3b). We used small interfering RNA (siRNA) to knock down RARγ. Transfection with si-RARγ #1 or #2 decreased the mRNA and protein expression of RARγ (S-Figs. 3c, d). RARγ knockdown (si-RARγ #1 or #2) also significantly suppressed the proliferation of PDAC cells (si-Control vs. si-RARγ #1, p < 0.05, n = 3; si-Control vs. si-RARγ #2, p < 0.05, n = 3) (Fig. 2f).
To confirm whether or not the effect of RARγ inhibition was applied to other PDAC cells, we examined the expression of RARγ in various pancreatic cancer cell lines. Some lines (Panc-1, MIAPaca2 and BxPC-3) had a high expression of RARγ as well as PK-1, and other lines (PK-8 and KLM-1) had a low expression of RARγ (S-Figs. 3e, f). We used another line with a high RARγ expression (Panc-1) and carried out the same experiments as described above. The proliferation of Panc-1 cells significantly decreased in the presence of either RARγi-1 or RARγi-2 compared to those in their absence at 72 h after seeding (Panc-1: RARγi-1 vs. control p < 0.01, RARγi-2 vs. control, p < 0.01, n = 3) (Fig. 2d, S-Fig. 3g). Our data indicated that the activation of RARγ signaling was involved in proliferation in PDAC.
Activation of RARα signaling suppressed the proliferation in PDAC
One previous study showed that an increased expression of RARα was associated with a good prognosis in PDAC patients [13], suggesting that RARα signaling may be tumor-suppressive in PDAC. However, there is no established consensus concerning the function of RARα in PDAC. Our database analyses described above indicated that the transcript level of RARα in pancreatic cancer (C) was higher than in normal pancreatic tissue (N), similar to RARγ (Fig. 1a), but did not correlate with the patient prognosis (S-Fig. 1a).
Furthermore, to clarify whether or not RARα signaling contributes to the proliferation of PDAC cells, we evaluated the effect of RARα inhibition and stimulation on the proliferation of a PDAC cell line (PK-1) in vitro. We initially examined whether or not the RARα-antagonist BMS195614 (RARαi) could block RA signaling. At least 1 µM of RARαi significantly decreased the expression of FABP5, a target gene of RA signaling (Fig. 3a). Cell proliferation assays revealed that the proliferation of PDAC cells did not decrease in the presence of RARαi, unlike RARγi-1 or RARγi-2 (RARαi vs. control, not significant, n = 3) (Fig. 3b). Next, we confirmed that at least 0.1 µM of the RARα-agonist AM580 increased the expression of FABP5, albeit without statistical significance (Fig. 3c). The proliferation of PDAC cells significantly decreased by approximately 20%-30% in the presence of RARα-agonist (Fig. 3d). These results of our in vitro experiments supported the notion that RARα signaling might be tumor-suppressive in PDAC.
Blockage of RARγ signaling induced the cell cycle arrest of the G1 phase without causing cell death in PDAC cells
To identify the mechanism underlying the growth suppression by RARγ inhibition in PDAC cells, we performed a cell cycle analysis. In PK-1 cells, RARγ inhibition by either RARγi-1 or RARγi-2 induced an increase in G0/G1-phase cells (RARγi-1-treated 73.1% vs. control 47.2%, p < 0.01, n = 3; RARγi-2-treated 71.3% vs. control 47.2%, p < 0.05, n = 3) and a decrease of S-phase cells (RARγi-1-treated 7.5% vs. control 26.2%, p < 0.01, n = 3, RARγi-2-treated 7.1% vs. control 26.2%, p < 0.01, n = 3) (Fig. 2g). In Panc-1 cells, RARγ inhibition by either RARγi-1 or RARγi-2 induced a similar result to PK-1 (G0/1-phase cells: RARγi-1-treated 69.2% vs. control 44.3%, p < 0.05, n = 3; RARγi-2-treated 62.0% vs. control 44.3%, not significant, n = 3; S-phase cells: RARγi-1-treated 5.5% vs. control 22.9%, p < 0.01, n = 3; RARγi-2-treated 9.3% vs. control 22.9%, p < 0.01, n = 3) (S-Fig. 3h). These results indicated that blockage of RARγ signaling induced cell cycle arrest of the G1 phase in PDAC cells.
To check whether or not the antineoplastic effect of RARγ inhibition was mediated by apoptosis, we performed annexin V staining. In PK-1, the percentage of annexin V-positive cells among cells treated with either RARγi-1 or RARγi-2 was not higher than that among untreated cells (control 12.0% vs. RARγi-1-treated 11.5% vs. RARγi-2-treated 16.1%, not significant, n = 3) (Fig. 2h). In Panc-1, RARγ inhibition by either RARγi-1 or RARγi-2 did not significantly increase annexin V-positive cells, similar to PK-1 (control 13.0% vs. RARγi-1-treated 11.9% vs. RARγi-2-treated 21.9%, not significant, n = 3) (S-Fig. 3i). This result indicated that blockage of RARγ signaling did not induce apoptosis in PDAC.
RARγ signaling did not cross-talk with the mitogen-activated protein kinase (MAPK) pathway
The RAS-RAF-MEK-MAPK signaling pathway is a core signaling pathway genetically altered in most PDAC and strongly involves the proliferation in PDAC [3, 26]. A previous study reported that the inhibition of MEK, an essential effector of the MAPK pathway [27], caused the cell cycle arrest of the G1 phase, similar to our findings concerning RARγ inhibition [28].
To clarify whether or not RARγ signaling cross-talked with the MAPK pathway, we examined the changes in the expression and phosphorylation of ERK1 and ERK2 (ERK1/2), which are signal-regulated kinases activated through the MAPK pathway [27], by RARγ inhibition using Western blotting in PDAC cells (PK-1 and Panc-1) (S-Fig. 4a). The expression and phosphorylation of ERK1/2 in cells treated by either RARγi-1 or RARγi-2 were not significantly decreased compared to untreated cells (PK-1, not significant, n = 3; Panc-1, not significant, n = 3) (S-Figs. 4b-e). These results suggested that RARγ signaling involved the proliferation of PDACs independently of the MAPK pathway.
Over-activation of RARγ signaling did not promote the proliferation of PDAC cells
To examine whether or not further activation of RARγ signaling promoted the proliferation of PDAC cells, we evaluated the effect of a RARγ-agonist BMS961 on the proliferation of PDAC cells (PK-1) in vitro. RARγ-agonist dose-dependently increased the expression of FABP5, a target gene of RA signaling (S-Fig. 5a). However, a cell proliferation assay revealed that the proliferation of PDAC cells did not increase in the presence of RARγ-agonist (S-Fig. 5b). These results suggested that RARγ signaling was necessary for the cell cycle progression of the G1-S phase but that its over-activation was not beneficial for cell proliferation in PDAC.
Blockage of RARγ signaling broadly downregulated the gene expression associated with the cell cycle progression of the G1-S phase and DNA synthesis in PDAC cells
To further understand the molecular mechanism by which RARγ inhibition affects cell proliferation, including the cell cycle process, we carried out RNA-seq using RARγi-1- or RARγi-2-treated cells (PK-1 and Panc-1).
First, we compared the gene expression between the control and RARγi-1-treated cells or between the control and RARγi-2-treated cells in each cell line. In PK-1 and Panc-1, we identified 826 and 1002 entities, respectively, that were commonly more than 2-fold down-regulated in RARγi-1- and RARγi-2-treated cells (S-Fig. 6a). Next, to narrow down the specific genes and pathways associated with RARγ signaling, we focused on 328 entities that were downregulated in 2 cell lines (PK-1 and Panc-1) (Fig. 4a) and performed a pathway analysis. A WikiPathway analysis revealed that RARγ inhibition significantly downregulated many pathways related to the cell cycle and DNA repair, such as G1 to S cell cycle control (WP45, p < 1.00E-12), Cell Cycle (WP179, p < 1.00E-12), Mitotic G1-G1-S phases (WP1858, p < 1.00E-12), Regulation of DNA Replication (WP1898, p < 1.00E-12) and DNA Damage Response (WP707, p < 1.00E-12) (Fig. 4b and Table S5). A further query of genes included in these pathways indicated that the gene expression associated with the cell cycle progression of the G1-S phase and DNA synthesis was broadly downregulated (Fig. 4c, Mitotic G1-G1-S phases [WP1858]). In addition, a GSEA also confirmed that gene sets related to the cell cycle or DNA replication were downregulated in RARγi-1- and RARγi-2-treated cells (S-Figs. 6b, c). These findings were consistent with our experimental results that blockage of RARγ signaling caused the cell cycle arrest of the G1 phase.
As in the method described above, we also investigated the genes upregulated by RARγ inhibition. We focused on 460 commonly up-regulated entities in 2 RARγi-1- and RARγi-2-treated cell lines (PK-1 and Panc-1) (S-Figs. 7a, b) and performed a pathway analysis. A WikiPathway analysis revealed that RARγ inhibition significantly up-regulated many pathways related to the unfolded protein response (UPR), such as ATF4 activates genes (WP2753, p < 1.00E-12), XBP1(S) activates chaperone genes (WP3472, p < 1.00E-12), the NRF2 pathway (WP2884, p < 1.00E-12) and ATF6 (ATF6-alpha) activates chaperone genes (WP2655, p = 8.37E-08) (S-Figs. 7c, d). These findings indicated that blockage of RARγ signaling activated UPR pathways in PDAC cells, suggesting that the activation of the UPR might also partially contribute to the tumor-suppressive effects of RARγ inhibition, as ER stress and subsequent activation of the UPR are reported to cause G1 arrest [29].
Next, we examined the expression and phosphorylation of proteins associated with the cell cycle progression of the G1-S phase using Western blotting. We identified a significant increase in the expression of endogenous cyclin-dependent kinase (CDK) inhibitor p21 and p27 and a significant decrease in the phosphorylation of CDK2 and the expression of CDK4 and CDK6 (Fig. 4d, S-Figs. 8a-g). These findings also confirmed our experimental results.
Furthermore, to explore whether or not RARγ inhibition affected the expression of esophagus tissue-specific genes in PDAC cells, we created a heatmap of their expression using RNA-seq data. A heatmap analysis showed that RARγ inhibition reduced the expression of some esophagus-tissue-specific genes (S-Figs. 9a, b). This result supported the notion that RARγ signaling regulated the expression of some esophagus tissue-specific genes.
RARγ signaling underlay the proliferation of patient-derived PDAC organoids
To investigate whether or not the findings obtained from PDAC cell lines could be applied to patient-derived PDAC, we tested the effect of RARγ inhibition using established patient-derived PDAC organoids (KYK070, KYK002, KYK023, KYK090 and KYK093).
First, to clarify whether or not RARγ-antagonists could block RARγ signaling in PDAC organoids, we examined the expression of FABP5 and KRT13 in KYK070 and KYK002 using qPCR. Both RARγi-1 and RARγi-2 significantly decreased the transcript levels of FABP5 and KRT13 (KYK070; Figs. 5a, b and KYK002; S-Fig. 10a). These findings indicated that RARγ-antagonists blocked RARγ signaling in PDAC organoids. Next, we evaluated the effect of RARγ inhibition on the proliferation of PDAC organoids. Cell proliferation assays revealed that the proliferation of all PDAC organoids (KYK070, KYK002, KYK023, KYK090 and KYK093) was significantly decreased in the presence of RARγi-1 or RARγi-2 compared to their absence at 72 h after seeding (Figs. 5c, d, S-Fig. 10b). Furthermore, treating lumen-forming KYK070 organoids with RARγi-1 or RARγi-2 significantly reduced the Ki67+ cell ratio without disrupting their lumen formation (Fig. 5e). These results supported the notion that activation of RARγ signaling underlay proliferation by promoting cell cycle progression in patient-derived PDAC.
Blockage of RARγ signaling synergized with chemotherapy to suppress the proliferation of PDAC cells and patient-derived PDAC organoids
To investigate whether or not RARγ inhibition synergizes with Gem, a critical anticancer drug for PDAC chemotherapy [30], we tested the combined effect of RARγi-1 and Gem on PDAC cells (PK-1) and patient-derived PDAC organoids (KYK070).
In PK-1, RARγi-1 and Gem alone inhibited tumor growth, and their combination significantly enhanced the effects (RARγi-1 vs. Gem + RARγi-1, p < 0.05; Gem vs. Gem + RARγi-1, p < 0.01) (Fig. 5f). Next, we confirmed the dose-dependent effect of RARγi-1 on KYK070 and decided to use 10 µM RARγi-1 for the assay (S-Fig. 10c). In KYK070, the combined effect of RARγi-1 and Gem was also observed, although there was no significant difference between the effects of Gem and Gem + RARγi-1 (RARγi-1 vs. Gem + RARγi-1, p < 0.05; Gem vs. Gem + RARγi-1, not significant) (Fig. 5g). These findings suggested that blockage of RARγ signaling might substantially impact PDAC therapy.