FGFR4 Inhibitor BLU9931 Attenuates Pancreatic Cancer Cell Proliferation and Invasion While Inducing Senescence: Evidence for Senolytic Therapy Potential in Pancreatic Cancer

Simple Summary Pancreatic ductal adenocarcinoma (PDAC) is a deadly malignancy that is projected to become the leading cause of cancer death by 2050. Fibroblast growth factor receptor 4 (FGFR4) is a transmembrane receptor that is overexpressed in half of PDACs. We determined that its expression in PDAC positively correlated with larger tumor size and more advanced tumor stage, and that BLU9931, a selective FGFR4 inhibitor, reduced PDAC cell proliferation and invasion while promoting their senescence. Quercetin, a senolytic drug, induced cell death in BLU9931-treated cells. We propose that targeting FGFR4 in combination with senolysis could provide a novel therapeutic strategy in patients whose PDAC expresses high FGFR4 levels. Abstract Fibroblast growth factor receptor 4 (FGFR4), one of four tyrosine kinase receptors for FGFs, is involved in diverse cellular processes. Activation of FGF19/FGFR4 signaling is closely associated with cancer development and progression. In this study, we examined the expression and roles of FGF19/FGFR4 signaling in human pancreatic ductal adenocarcinoma (PDAC). In human PDAC cases, FGFR4 expression positively correlated with larger primary tumors and more advanced stages. Among eight PDAC cell lines, FGFR4 was expressed at the highest levels in PK-1 cells, in which single-nucleotide polymorphism G388R in FGFR4 was detected. For inhibition of autocrine/paracrine FGF19/FGFR4 signaling, we used BLU9931, a highly selective FGFR4 inhibitor. Inhibition of signal transduction through ERK, AKT, and STAT3 pathways by BLU9931 reduced proliferation in FGF19/FGFR4 signaling-activated PDAC cells. By contrast, BLU9931 did not alter stemness features, including stemness marker expression, anticancer drug resistance, and sphere-forming ability. However, BLU9931 inhibited cell invasion, in part, by downregulating membrane-type matrix metalloproteinase-1 in FGF19/FGFR4 signaling-activated PDAC cells. Furthermore, downregulation of SIRT1 and SIRT6 by BLU9931 contributed to senescence induction, priming these cells for quercetin-induced death, a process termed senolysis. Thus, we propose that BLU9931 is a promising therapeutic agent in FGFR4-positive PDAC, especially when combined with senolysis (195/200).

BLU9931 is an irreversible and highly selective small molecule inhibitor for FGFR4 [35]. BLU9931 makes a covalent bond with Cys552 within the ATP-binding pocket of FGFR4, which is not present in FGFR1-3. BLU9931 inhibited gastric cancer cell growth in a transplanted mouse model [36]. To date, however, there have been no reports about the effects of FGFR4 inhibitors in PDAC. In the present study, therefore, we delineated FGFR4 expression in PDAC, and examined the actions of BLU9931 on FGF19/FGFR4 signaling and biological processes in PDAC cells.

Expression of FGFR4 in Human PDAC Tissues and PDAC Cell Lines
To examine the expression of FGFR4 in human PDAC tissues, we performed immunohistochemical analysis of FGFR4 using pancreas tissue microarrays. In normal human pancreatic tissues, FGFR4 was weakly localized in normal exocrine and endocrine pancreatic tissues ( Figure 1(A-I)). FGFR4 was observed in the cytoplasm and cell membrane of human PDAC cells (Figure 1(A-II) and Figure S1). Strong FGFR4 immunoreactivity was seen in the cancer cells in 67 of 136 (49%) PDAC samples and high FGFR4 expression correlated with larger primary tumor size and more advanced stages of the cancer (p < 0.001, Table 1).
FGFR4 mRNA was expressed in all eight PDAC cell lines examined ( Figure 1B). FGFR4 mRNA was highest in T3M-4 and MIA PaCa-2 cells and lowest in PK-45P cells. By contrast, FGFR4 protein levels were highest in PK-1 and T3M-4 cells and lowest in MIA PaCa-2 and PK-45P cells ( Figure 1C). Fluorescence-activated cell sorting revealed that cell surface FGFR4 levels were highest in PK-1 and T3M-4 cells ( Figure 1D,E). Given that SNP Arg388 of the FGFR4 gene may be associated with decreased survival in certain cancers, we next examined this SNP in the above cell lines, which genotyped as follows: MIA PaCa-2 and PK-8 cells as Gly/Gly; PK-1, PANC-1, and PK-45P cells as Gly/Arg; and PK-59, T3M-4, and KP4 cells as Arg/Arg ( Figure S2 and Table 2). Inasmuch as FGF19 is the specific and sole ligand for FGFR4, we next performed real-time quantitative PCR (qPCR) for this ligand. FGF19 mRNA levels were relatively elevated in PK-1, PK-45P, and T3M-4 cells, and lowest in other cell lines ( Figure 1F).   The FGF19 gene is occasionally amplified in hepatocellular carcinomas and breast cancer as a consequence of the presence of an amplicon on chromosome 11q13.3 [37,38]. PDAC also exhibits regions of genomic amplification, including the 11q13.3 region in chromosome 11 [39]. To determine whether such an amplicon harbors an FGF19 amplification, we examined gene expression data in The Cancer Genome Atlas (TCGA), a publicly available data base [40,41]. We used the University of Texas Southwestern (UTSW) data set since it was based on high-quality cancer cell-enriched samples [42], which revealed the presence of an amplified FGF19 gene in 11 of the 109 samples in this cohort (Figure 2A). In addition, other genes within this amplicon were also amplified, including FGF4, FGF3, CCND1, RAD9A, RPS6KB2, GAB2, and PAK1 ( Figure 2A). In some cases, there was also co-amplification of FGF19 with FGFR4 ( Figure 2B), even though the FGFR4 gene is located on chromosome 5q25. Only 7 of the 109 samples exhibited concomitant upregulation of the KLB gene (encoding β-Klotho) and FGFR4 gene ( Figure 2B). Only three and two cases exhibited FGFR1 and FGFR2 gene amplification, respectively, whereas seven cases exhibited FGFR3 amplification, including one case that did not have a KRAS mutation. In three patients, there was concomitant overexpression of FGFR3 and FGFR4. Intriguingly, among the eight PDAC cases that did not harbor a KRAS mutation, four cases had either amplification or increased expression of the FGF19, FGF4, FGF3, and CCND1 genes, and six cases had RAD9A and RPS6KB2 gene overexpression, two of which were due to amplification (Figure 2A), raising the possibility that a portion of the 11q13.3 amplicon helps drive PDAC development in the absence of mutated KRAS. Additional real-time qPCR analysis of RNA from the above cell lines for FGF4, FGF3, CCND1, RAD9A, RPS6KB2, GAB2, PAK1, and KLB revealed

Effects of BLU9931 on Growth and Cell Cycle of PDAC Cells
To determine whether there was a functional FGF19/FGFR4 signaling axis in human PDAC cells, we cultured PDAC cells with or without the highly specific FGFR4 inhibitor, BLU9931. PK-1 cells exhibited a dose-dependent decrease in proliferation in the presence of BLU9931 ( Figure 3A). An inhibitory effect on proliferation was also observed in T3M-4 cells ( Figure S3A). By contrast, BLU9931 did no inhibit proliferation in PK-45P cells ( Figure 3A), which expressed very low FGFR4 levels ( Figure 1). Next, we examined whether BLU9931 exerts toxic effects in PDAC cells. Incubation of PK-1 cells with 2 μM BLU9931 followed by fluorescence activated cell sorter (FACS) analysis using annexin V and propidium iodide (PI) showed that ~2% of the cells underwent apoptosis, necrosis, or cell injury ( Figure 3B). Higher concentration (>10 μM) of BLU9931 revealed dead and floating cells suggestive of a toxic effect at these very high concentrations of the drug.

Effects of BLU9931 on Growth and Cell Cycle of PDAC Cells
To determine whether there was a functional FGF19/FGFR4 signaling axis in human PDAC cells, we cultured PDAC cells with or without the highly specific FGFR4 inhibitor, BLU9931. PK-1 cells exhibited a dose-dependent decrease in proliferation in the presence of BLU9931 ( Figure 3A). An inhibitory effect on proliferation was also observed in T3M-4 cells ( Figure S3A). By contrast, BLU9931 did no inhibit proliferation in PK-45P cells ( Figure 3A), which expressed very low FGFR4 levels ( Figure 1). Next, we examined whether BLU9931 exerts toxic effects in PDAC cells. Incubation of PK-1 cells with 2 µM BLU9931 followed by fluorescence activated cell sorter (FACS) analysis using annexin V and propidium iodide (PI) showed that~2% of the cells underwent apoptosis, necrosis, or cell injury ( Figure 3B). Higher concentration (>10 µM) of BLU9931 revealed dead and floating cells suggestive of a toxic effect at these very high concentrations of the drug. cells, and that FGFR4 downstream signaling can be inhibited by BLU9931 in certain PDAC cells. Furthermore, BLU9931 reduced FGF19 mRNA levels but not FGFR4 and KLB mRNA levels ( Figures  3F and S3E), suggesting that autocrine/paracrine FGF19/FGFR4 signaling upregulates FGF19 expression through the FGF19/FGFR4 signaling feedback loop. Taken together, these results indicate that the inhibitory effect of 2 μM BLU9931 on PDAC cell proliferation is caused by inhibition of autocrine/paracrine FGF19/FGFR4 signaling and not through induction of apoptosis. Results are presented as means ± SD from three independent experiments. (E) Western blot analysis for FGF19/FGFR4 signaling was performed in PK-1 or PK-45P cells that were incubated with or without 2 μM BLU9931 for 3 days. The expression of each band is shown under the blot. The histograms show mean densitometric readings ± SD for the phosphorylated proteins normalized to those of the loading controls. Results are the means ± SD from three independent experiments. (F) Results are presented as means ± SD from three independent experiments. (E) Western blot analysis for FGF19/FGFR4 signaling was performed in PK-1 or PK-45P cells that were incubated with or without 2 µM BLU9931 for 3 days. The expression of each band is shown under the blot. The histograms show mean densitometric readings ± SD for the phosphorylated proteins normalized to those of the loading controls. Results are the means ± SD from three independent experiments. (F) Real-time qPCR analysis of FGFR4 or FGF19 or KLB in PK-1 cells that were incubated with or without 2 µM BLU9931 for 3 days. Results are normalized to values obtained for control cells (value = 1), and are the as means ± SD from three independent experiments. * p < 0.05, ** p < 0.01. Control (Ctr): Control cells were incubated with dimethyl sulfoxide. We next examined the effects of BLU9931 on the cell cycle by FACS analysis. There was a higher percentage of BLU9931-treated PK-1 or T3M-4 cells in the G0/G1 phase (70.1% vs. 60.7% and 66.4% vs. 58.0%, respectively) and lower percentage in the S phase (13.5% vs. 20.7% and 27.3% vs. 29.9%, respectively) and G2/M phase (16.4% vs. 18.6% and 6.3% vs. 12.1%, respectively) relative to control cells ( Figure 3C and Figure S3B). These results point to a G1 to S phase cell cycle slowdown in BLU9931-treated cells.
Next, we examined the expression of cell cycle related-genes. Real-time qPCR analysis showed that the expression of p21, P27, and E2F1 was upregulated whereas CDK1 expression was downregulated in BLU9931-treated cells ( Figure 3D and Figure S3C). We also examined autocrine/paracrine FGF19/FGFR4 signaling by immunoblotting for downstream targets (ERK, AKT, and STAT3). As shown in Figure 3E, reduced phosphorylation of ERK, AKT, and STAT3 was observed in BLU9931-treated PK-1 cells. Similar reductions in phosphorylation in these downstream signaling molecules were observed in T3M-4 cells ( Figure S3D). By contrast, phosphorylation of these proteins was not altered by the same concentration of BLU9931 (2 µM) in PK-45P cells. These results indicate that autocrine/paracrine FGF19/FGFR4 signaling is functional in PK-1 and T3M-4 cells but not in PK-45P cells, and that FGFR4 downstream signaling can be inhibited by BLU9931 in certain PDAC cells. Furthermore, BLU9931 reduced FGF19 mRNA levels but not FGFR4 and KLB mRNA levels ( Figure 3F and Figure S3E), suggesting that autocrine/paracrine FGF19/FGFR4 signaling upregulates FGF19 expression through the FGF19/FGFR4 signaling feedback loop. Taken together, these results indicate that the inhibitory effect of 2 µM BLU9931 on PDAC cell proliferation is caused by inhibition of autocrine/paracrine FGF19/FGFR4 signaling and not through induction of apoptosis.

Effects of BLU9931 on Stemness Features of PDAC Cells
To clarify the correlation between autocrine/paracrine FGF19/FGFR4 signaling and stemness in PDAC cells, we examined stemness features, including stemness marker expression, anticancer drug resistance, and sphere-forming ability. Real-time qPCR showed that two of seven examined stemness markers (Sox2 and Nestin) were expressed at lower levels in 2 µM BLU9931-treated PK-1 cells ( Figure 4A), whereas one of the seven markers (CD24) was expressed at a higher level in BLU9931-treated PK-1 than in control cells ( Figure 4A). These results suggest that autocrine/paracrine FGF19/FGFR4 signaling is not definitively involved in stemness marker expression in PDAC cells. To assess anticancer drug resistance, we used three anti-pancreatic cancer drugs, gemcitabine, 5-FU, and abraxane. Cell survival rates following addition of gemcitabine, 5-FU, and abraxane (all at 100 µM) were approximately 60%, 50%, and 30%, respectively ( Figure 4B). However, survival rates were not significantly different in 2 µM BLU9931-treated PK-1 cells following incubation with the above anticancer drugs at either 10 or 100 µM ( Figure 4B). Furthermore, we examined the expression levels of four potential anticancer drug transporters. Real-time qPCR analysis revealed that the expression of ABCG2, ABCB1, ABCC1, and ABCC2 was not significantly different between control and 2 µM BLU9931-treated PK-1 cells ( Figure 4C). Thus, autocrine/paracrine FGF19/FGFR4 signaling is not involved in regulating the expression of anticancer drug transporters and the resistance toward anticancer drugs in PDAC cells. Next, we examined sphere-forming ability, which is viewed as an indicator of self-renewal and stemness [43][44][45]. Sphere-forming assays showed that the number of spheres was not different between control and 2 µM BLU9931-treated PK-1 cells ( Figure 4D). Furthermore, in sphere cells, the expression levels of FGFR4 and FGF19 were markedly lower than those in adherent cells ( Figure 4E). Thus, autocrine/paracrine FGF19/FGFR4 signaling is not involved in sphere formation of PDAC cells.

Figure 4. Effects of BLU9931 on stemness features of PDAC cells. (A)
Real-time qPCR analysis of stemness markers in PK-1 cells that were incubated with or without 2 μM BLU9931 for 3 days. Results shown are normalized to values obtained for control cells (value = 1). (B) Anticancer drug resistance assay in PK-1 cells that were incubated with or without 2 μM BLU9931. The dose response (10 or 100 μM) of PK-1 cells to gemcitabine, 5-FU, and abraxane was determined using the ATP assay. (C) Realtime qPCR analysis of transporters in PK-1 cells that were incubated with or without 2 μM BLU9931 for 3 days. Results shown are normalized to values obtained for control cells (value = 1). (D) Sphereforming assays performed in PK-1 cells that were incubated with or without 2 μM BLU9931. (E) Realtime qPCR analysis of FGFR4 or FGF19 in PK-1 cells cultured under adherent or 3-D conditions. Results shown are normalized to values obtained for adherent-cultured cells (value = 1). All histograms are presented as means ± SD from three independent experiments. * p < 0.05, ** p < 0.01. Control (Ctr): Control cells were incubated with dimethyl sulfoxide.

Effects of BLU9931 on Migration and Invasion of PDAC Cells
We next examined the inhibitory effects of FGF19/FGFR4 signaling on migration and invasion of PDAC cells. In wound healing assays, there were no significant differences in the migration rates between control and BLU9931-treated PK-1 cells ( Figure 5A). Furthermore, boyden chamber assays showed that BLU9931 did not alter the migration of PK-1 cells ( Figure 5B). Thus, these results indicate that FGF19/FGFR4 signaling does not regulate PDAC cell migration. By contrast, invasion assays showed that numbers of invaded cells were significantly reduced in 2 μM BLU9931-treated PK-1 cells compared with control cells ( Figure 5C), indicating that autocrine/paracrine FGF19/FGFR4 signaling promotes invasion. One of key features of invasion is increased production of matrix metalloproteinases (MMPs) [46,47] and MMP2, MMP9, and MT1-MMP are known to be expressed in PDAC cells [48]. Therefore, we hypothesized that FGF19/FGFR4 signaling might promote MMP expression and activity. Real-time qPCR revealed that MT1-MMP mRNA levels were significantly reduced in BLU9931-treated PK-1 cells, while MMP2 mRNA levels were upregulated ( Figure 5D). Furthermore, Western blot analysis showed that protein levels of MT1-MMP were significantly reduced in BLU9931-treated PK-1 cells ( Figure 5E). MT1-MMP is well known to induce activated Results shown are normalized to values obtained for control cells (value = 1). (B) Anticancer drug resistance assay in PK-1 cells that were incubated with or without 2 µM BLU9931. The dose response (10 or 100 µM) of PK-1 cells to gemcitabine, 5-FU, and abraxane was determined using the ATP assay. (C) Real-time qPCR analysis of transporters in PK-1 cells that were incubated with or without 2 µM BLU9931 for 3 days. Results shown are normalized to values obtained for control cells (value = 1). (D) Sphere-forming assays performed in PK-1 cells that were incubated with or without 2 µM BLU9931. (E) Real-time qPCR analysis of FGFR4 or FGF19 in PK-1 cells cultured under adherent or 3-D conditions. Results shown are normalized to values obtained for adherent-cultured cells (value = 1). All histograms are presented as means ± SD from three independent experiments. * p < 0.05, ** p < 0.01. Control (Ctr): Control cells were incubated with dimethyl sulfoxide.

Effects of BLU9931 on Migration and Invasion of PDAC Cells
We next examined the inhibitory effects of FGF19/FGFR4 signaling on migration and invasion of PDAC cells. In wound healing assays, there were no significant differences in the migration rates between control and BLU9931-treated PK-1 cells ( Figure 5A). Furthermore, boyden chamber assays showed that BLU9931 did not alter the migration of PK-1 cells ( Figure 5B). Thus, these results indicate that FGF19/FGFR4 signaling does not regulate PDAC cell migration. By contrast, invasion assays showed that numbers of invaded cells were significantly reduced in 2 µM BLU9931-treated PK-1 cells compared with control cells (Figure 5C), indicating that autocrine/paracrine FGF19/FGFR4 signaling promotes invasion. One of key features of invasion is increased production of matrix metalloproteinases (MMPs) [46,47] and MMP2, MMP9, and MT1-MMP are known to be expressed in PDAC cells [48]. Therefore, we hypothesized that FGF19/FGFR4 signaling might promote MMP expression and activity. Real-time qPCR revealed that MT1-MMP mRNA levels were significantly reduced in BLU9931-treated PK-1 cells, while MMP2 mRNA levels were upregulated ( Figure 5D). Furthermore, Western blot analysis showed that protein levels of MT1-MMP were significantly reduced in BLU9931-treated PK-1 cells ( Figure 5E). MT1-MMP is well known to induce activated MMP2 by proteolytic cleavage of prodomain of proMMP2 [46,47]. Therefore, we next examined the activity of MMP2 released from BLU9931-treated PK-1 cells using gelatin zymography. As shown in Figure 5F, collagenase activity of MMP2 was reduced in BLU9931-treated PK-1 cells. A BLU9931-induced reduction in invasion, MT1-MMP expression, and MMP2 activity was also observed in T3M-4 cells ( Figure S4). Thus, autocrine/paracrine FGF19/FGFR4 signaling regulates expression of MT1-MMP and contributes to the activation of MMP2 in PDAC cells.
Cancers 2020, 12, x 10 of 24 MMP2 by proteolytic cleavage of prodomain of proMMP2 [46,47]. Therefore, we next examined the activity of MMP2 released from BLU9931-treated PK-1 cells using gelatin zymography. As shown in Figure 5F, collagenase activity of MMP2 was reduced in BLU9931-treated PK-1 cells. A BLU9931induced reduction in invasion, MT1-MMP expression, and MMP2 activity was also observed in T3M-4 cells ( Figure S4). Thus, autocrine/paracrine FGF19/FGFR4 signaling regulates expression of MT1-MMP and contributes to the activation of MMP2 in PDAC cells. . Results are the means ± SD from three independent experiments. (E) Western blot analysis of MT1-MMP was performed in PK-1 that were incubated with or without 2 μM BLU9931 for 3 days. The expression of each band is shown under the blot. The histograms show mean densitometric readings ± SD for MT1-MMP normalized to those of the loading controls. Results are the as means ± SD from three independent experiments. (F) Gelatin zymography was performed using culture supernatants from PK-1 cells that were incubated with or without 2 μM BLU9931 for 3 days. Relative band intensity is shown. * p < 0.05, ** p < 0.01. Control (Ctr): Control cells were incubated with dimethyl sulfoxide.

Induction of Senescence in PDAC Cells by Long-Term Treatment with BLU9931
During long-term culture (one week) with BLU9931, some populations of cells exhibited an enlarged and flattened morphology ( Figure 6A). Such changes may occur in senescent cells [49]. Therefore, we next examined the effects of BLU9931 on senescence in PK-1 cells by assaying for Results are the means ± SD from three independent experiments. (E) Western blot analysis of MT1-MMP was performed in PK-1 that were incubated with or without 2 µM BLU9931 for 3 days. The expression of each band is shown under the blot. The histograms show mean densitometric readings ± SD for MT1-MMP normalized to those of the loading controls. Results are the as means ± SD from three independent experiments. (F) Gelatin zymography was performed using culture supernatants from PK-1 cells that were incubated with or without 2 µM BLU9931 for 3 days. Relative band intensity is shown. * p < 0.05, ** p < 0.01. Control (Ctr): Control cells were incubated with dimethyl sulfoxide.

Induction of Senescence in PDAC Cells by Long-Term Treatment with BLU9931
During long-term culture (one week) with BLU9931, some populations of cells exhibited an enlarged and flattened morphology ( Figure 6A). Such changes may occur in senescent cells [49]. Therefore, we next examined the effects of BLU9931 on senescence in PK-1 cells by assaying for senescence-associated β-galactosidase (SA-β-Gal) activity, which is a well-known marker for senescent cells. BLU9931 increased the number of SA-β-Gal-positive cells compared with the control group ( Figure 6A,B). By contrast, a 3-day incubation with BLU9931 did not alter SA-β-Gal staining in PK-1 cells ( Figure S5A,B). Furthermore, phosphorylation of H2A.X (γH2A.X), which is an early sign of DNA damage as well as a senescence marker [50,51], was increased in PK-1 cells that had been incubated for 3 days with BLU9931 and further increased in long-term culture with BLU9931 ( Figure 6C). There were no differences in telomere length between control and long-term BLU9931-treated PK-1 cells ( Figure 6D). Taken together, these results suggest BLU9931 induces premature senescence through DNA damage, possibly through the ATM pathway.
Senescent cells may exhibit a senescence-associated secretory phenotype (SASP) [52], a process whereby high levels of SASP factors, including inflammatory cytokines, are secreted. In the present study, prolonged incubation of PK-1 cells with BLU9931 resulted in a significant increase in the levels of SASP-associated cytokines (IL-1α, IL-1β, IL-6, TNFα, and GM-CSF) when compared with the corresponding levels in control cells ( Figure 6E). In T3M-4 cells, the number of SA-β-Gal-positive cells accompanying enlarged and flattened morphology was also increased following prolonged incubation with BLU9931, but the percentage of SA-β-Gal-positive cells was lower than in PK-1 cells ( Figure S5C,D). Moreover, among the above SASP cytokines, only GM-CSF was significantly upregulated in these T3M-4 cells ( Figure S5E).
It is known that reduction of sirtuin1 (SIRT1) and SIRT6 leads to induction of premature senescence caused from attenuation of DNA damage repair [53]. We therefore examined the expression of SIRT1 and SIRT6 in relation to the induction of premature senescence by BLU9931. BLU9931 reduced the expression of SIRT1 and SIRT6 ( Figures 6F and S5F), suggesting that long-term BLU9931 treatment induces premature senescence in PDAC cells, in part, through downregulation of SIRT1 and SIRT6.  Senescent cells may exhibit a senescence-associated secretory phenotype (SASP) [52], a process whereby high levels of SASP factors, including inflammatory cytokines, are secreted. In the present study, prolonged incubation of PK-1 cells with BLU9931 resulted in a significant increase in the levels of SASP-associated cytokines (IL-1α, IL-1β, IL-6, TNFα, and GM-CSF) when compared with the corresponding levels in control cells ( Figure 6E). In T3M-4 cells, the number of SA-β-Gal-positive cells accompanying enlarged and flattened morphology was also increased following prolonged incubation with BLU9931, but the percentage of SA-β-Gal-positive cells was lower than in PK-1 cells (Figure S5C,D). Moreover, among the above SASP cytokines, only GM-CSF was significantly upregulated in these T3M-4 cells ( Figure S5E).
It is known that reduction of sirtuin1 (SIRT1) and SIRT6 leads to induction of premature senescence caused from attenuation of DNA damage repair [53]. We therefore examined the expression of SIRT1 and SIRT6 in relation to the induction of premature senescence by BLU9931. BLU9931 reduced the expression of SIRT1 and SIRT6 ( Figure 6F and Figure S5F), suggesting that long-term BLU9931 treatment induces premature senescence in PDAC cells, in part, through downregulation of SIRT1 and SIRT6.

Effects of Senolytic Drugs on BLU9931-Induced Senescent PDAC Cells
Quercetin and dasatinib are senolytic drugs [54] that can selectively eliminate senescent cells. Therefore, the dose-dependent effects of quercetin or dasatinib on the viability of PDAC cell lines (PK-1, PK-45P, T3M-4) were examined next. We added increasing concentrations of quercetin (1.56-50 µM) or dasatinib (7.8-500 nM) to PK-1, PK-45P, and T3M-4 cells. As shown in Figure S6, we identified the maximal concentration without cytotoxic effects as follows: quercetin (25 µM) and dasatinib (62.5 nM) for PK-1 cells, quercetin (25 µM) and dasatinib (7.8 nM) for PK-45P cells, and quercetin (12.5 µM) and dasatinib (7.8 nM) for T3M-4 cells. We then assessed the cytotoxic effects of senolytic drugs on BLU9931-induced senescent PDAC cells. Quercetin significantly decreased the viability of long-term BLU9931-treated PK-1 and T3M-4 cells ( Figure 7A and Figure S6) but not PK-45P cells ( Figure 7A), whereas dasatinib was not effective in any of the BLU9931-treated cells ( Figure 7A and Figure S7). Thus, the senolytic drug, quercetin, was effective in those PDAC cells in which BLU9931 induced senescent changes. As shown in Figure S8A, the poly ADP ribose polymerase (PARP) inhibitor olaparib inhibited the growth of PK-1 cells dose dependently, with an IC50 of 8 µM. At this concentration, olaparib did not significantly alter the number of apoptotic and necrotic/injured cells that had been incubated with BLU9931 ( Figure S8B). Next, we examined senescence and cytotoxic effects of BLU9931+quercetin+olaparib. BLU9931+olaparib enhanced the proportion of senescent cells compared with BLU9931 alone ( Figure S9A,B). This was associated with a profound decrease in the total number of cells, indicating that in addition to inducing senescence, olaparib inhibited proliferation and exerted a senolytic effect in these cells. However, there was no additive effect on PK-1 cell viability with quercetin addition in BLU9931+olaparib-treated cells ( Figure S9C). We speculate that senescent cells induced by olaparib treatment are not sensitive to quercetin. Klotho-β (KLB) binding to FGFR4 enables FGF19 binding to the FGFR4-KLB heterodimer and promotes FGFR4 dimerization, thereby activating downstream signaling, such as ERK, AKT, and STAT3. These pathways promote cell proliferation, upregulate membrane-type matrix metalloproteinase-1 (MT1-MMP) that contributes to enhanced cell invasion, and upregulate sirtuin1 (SIRT1) and SIRT6 to help suppress senescence. Additionally, FGF19 auto-induces FGF19 expression, thereby promoting an autocrine/paracrine loop for persistent FGF19-mediated activation of FGFR4. BLU9931, a highly specific and irreversible FGFR4 inhibitor, inhibits the tyrosine kinase activity of FGFR4 by interacting with cysteine residue 552 near FGFR4's ATP binding region, thereby suppressing downstream signaling and impeding upregulation of the above genes.

Discussion
Downstream pathways of FGF19/FGFR4 signaling, such as PI3K-AKT and MEK-ERK, lead to enhanced cell proliferation and survival in cancer cells. The selective FGFR4 inhibitor BLU9931 that was used in this study can inhibit FGF19/FGFR4-positive hepatocellular carcinoma tumors [35]. BLU9931 also exerts potent anticancer effects in head and neck cancers, breast cancer, colorectal cancer, and squamous cell carcinomas [34,[55][56][57]. Moreover, in breast cancer, autocrine FGF19/FGFR4 signaling is important for cell survival, and FGF19/FGFR4 co-expression was observed in primary breast tumors [56]. In the current study, we demonstrated that BLU9931 was also effective at inhibiting the proliferation of PDAC cells (Figures 3 and S3). Furthermore, we found that FGF19 expression was lowered following inhibition of FGF19/FGFR4 signaling ( Figures 3D and S3C), suggesting that in PDAC cells, the FGF19/FGFR4 axis controls FGF19 in a positive loop and that Klotho-β (KLB) binding to FGFR4 enables FGF19 binding to the FGFR4-KLB heterodimer and promotes FGFR4 dimerization, thereby activating downstream signaling, such as ERK, AKT, and STAT3. These pathways promote cell proliferation, upregulate membrane-type matrix metalloproteinase-1 (MT1-MMP) that contributes to enhanced cell invasion, and upregulate sirtuin1 (SIRT1) and SIRT6 to help suppress senescence. Additionally, FGF19 auto-induces FGF19 expression, thereby promoting an autocrine/paracrine loop for persistent FGF19-mediated activation of FGFR4. BLU9931, a highly specific and irreversible FGFR4 inhibitor, inhibits the tyrosine kinase activity of FGFR4 by interacting with cysteine residue 552 near FGFR4's ATP binding region, thereby suppressing downstream signaling and impeding upregulation of the above genes.

Discussion
Downstream pathways of FGF19/FGFR4 signaling, such as PI3K-AKT and MEK-ERK, lead to enhanced cell proliferation and survival in cancer cells. The selective FGFR4 inhibitor BLU9931 that was used in this study can inhibit FGF19/FGFR4-positive hepatocellular carcinoma tumors [35]. BLU9931 also exerts potent anticancer effects in head and neck cancers, breast cancer, colorectal cancer, and squamous cell carcinomas [34,[55][56][57]. Moreover, in breast cancer, autocrine FGF19/FGFR4 signaling is important for cell survival, and FGF19/FGFR4 co-expression was observed in primary breast tumors [56]. In the current study, we demonstrated that BLU9931 was also effective at inhibiting the proliferation of PDAC cells (Figure 3 and Figure S3). Furthermore, we found that FGF19 expression was lowered following inhibition of FGF19/FGFR4 signaling ( Figure 3D and Figure S3C), suggesting that in PDAC cells, the FGF19/FGFR4 axis controls FGF19 in a positive loop and that FGF19 has the potential to act in an autocrine/paracrine manner when produced and released by these cells. Indeed, in two of our tested PDAC cell lines, PK-1 and T3M-4, FGF19 ( Figure 1F) and FGFR4 ( Figure 1C-E) expression was relatively elevated. Moreover, activation of FGFR4 is associated with FGF19 gene amplification in several types of solid tumors [58,59].
In the present study, we also demonstrated that a subset of PDAC samples harbored an 11q13.3 amplicon and exhibited FGF19 gene amplification in conjunction with amplification of other genes within this amplicon, including FGF3, FGF4, CCND1, RAD9A, RPS6KB2, GAB2, and PAK1. In three cases, there was also co-amplification of FGF19 with FGFR4 even though the FGFR4 gene is located on chromosome 5q25. Altogether, 27 of the 109 samples in the UTSW-TCGA data set exhibited concomitant increases in FGF19 and FGFR4 expression. Moreover, in three of the cases, FGFR4 co-amplified with FGFR3, a gene located on 4p16.3. Importantly, in a genomic study of 4853 solid cancers, 7.1% were found to have FGFR gene alterations, 66% of which consisted of gene amplifications [59], underscoring the importance of FGFR signaling in cancer pathobiology. Taken together, these observations suggest that the FGF19/FGFR4 may activate a novel driver oncogene pathway in PDAC. Additional lines of evidence in support of this possibility derive from our observation that 49% of PDAC samples analyzed in the present study exhibited strong FGFR4 immunoreactivity, that autocrine/paracrine FGF19/FGFR4 signaling was functional in PK-1 and T3M-4 cells that expressed relatively high levels of FGF19 and FGFR4, and that FGFR4 downstream signaling was readily inhibited by BLU9931, a highly specific inhibitor of FGFR4 that does not target other FGFRs or other tyrosine kinase receptors.
The FGF19/FGFR4 signaling is known to promote invasion and metastasis in several different types of cancer [57]. In the current study, we demonstrated for the first time that FGF19/FGFR4 signaling is involved in PDAC cell invasion through the regulation of MT1-MMP expression and MMP2 activity in these cells ( Figure 5 and Figure S4). Our findings are thus consistent with the known strong correlation between MT1-MMP expression and cancer cell invasiveness [60]. Moreover, MMP2 activity and effects on cellular invasion are MT1-MMP dependent [61]. It is therefore possible that the BLU9931-mediated decrease in MT1-MMP levels led to the suppression of invasion, in part, as a result of the associated decrease in MMP2 activity. In as much as STAT3 directly controls expression of MT1-MMP [62], inhibition of the FGF19/FGFR4-activated STAT3 pathway by BLU9931, as evidenced by marked suppression of STAT3 phosphorylation, suggests that BLU9931 may also suppress metastasis in PDAC. Given the important role of the metastatic process in PDAC, it will be important to test this possibility in mouse models of PDAC metastasis.
PNETs overexpressing FGFR4 exhibit an increased propensity for forming hepatic metastases in mice [29] and hepatocellular carcinoma in humans is associated with excessive FGFR4-FGF19 signaling [31,35]. Our analysis of TCGA data revealed that 47 of 109 patients with PDAC overexpressed FGF19, and in 11 of these patients, this was due to FGF19 gene amplification (Figure 2A). It is therefore possible that high levels of FGF19 in~43% of patients with PDAC are released by their tumors and drain directly into the liver via the pancreatic vein, thereby facilitating the growth of metastases within the FGF19-enriched hepatic microenvironment. In this context, the potential of senolytic therapy with newer and safer agents, such as BLU554 [63], which is currently undergoing a large phase 1 clinical trial in hepatocellular carcinoma patients (NCT02508467), may provide an exciting therapeutic opportunity in patients with PDAC and could also help treat their hepatic metastases.
To date, the possible involvement of FGF19/FGFR4 signaling in cancer stem cells (CSCs) has not been definitively established. Our findings suggest that FGF19/FGFR4 signaling had little association with stemness in PDAC cells (Figure 4). Thus, FGFR4 and FGF19 expression was reduced in 3-D culture ( Figure 4E), which is an optimal culture condition for CSCs [43][44][45], suggesting that FGF19/FGFR4 activation was not contributing to CSC formation. FGFR1 and FGFR4 are known to compete via their interactions with KLB [58]. Moreover, in FGFR1-amplified non-small cell lung cancer cells, FGFR1 signaling contributes to maintenance of the CSC phenotype [64]. Furthermore, PD173074, a selective FGFR1 inhibitor, suppressed the CSC-like phenotype of PDAC cells [65]. Therefore, it is possible that another pathway, such as FGF21/FGFR1, may be involved in modulation of stemness through KLB in PDAC cells. Moreover, CSCs can undergo epithelial-mesenchymal transition (EMT), and FGF19/FGFR4 signaling has been implicated in this process in colorectal cancer and hepatocellular carcinoma [66,67]. However, the actions of endocrine FGFs are complex and may be context dependent [68]. For example, the pancreatic acinar cell expresses high levels of FGF21 that are suppressed by oncogenic Kras, and placing such mice on a high-fat diet promotes pancreatic inflammation and PanIN progression to PDAC, processes that are attenuated by the exogenous administration of FGF21 [69]. Moreover, acinar cells mostly express the 3-Ig form of FGFR1, whereas the PDAC cells express the 2-Ig form of FGFR1 [70], which leads to enhanced constitutive FGFR1 signaling [71]. Given this complexity, further studies are necessary to improve our understanding of the relationship between FGF19/FGFR4 signaling, CSCs, and EMT in PDAC.
Cellular senescence is a complex biological process, whereby cell proliferation is arrested in response to various stressors [52]. Senescence can also occur in cancer cells, which can exhibit context-dependent biological alterations, such as dormancy and suppression of tumor progression or exit from dormancy with CSC-like features and enhanced capacity to invade and metastasize [72][73][74][75]. Senescent tumor cells are also observed following chemotherapy or radiation therapy [76]. While the induction of cancer cell senescence, per se, has been proposed as a therapeutic strategy in tumor therapy, the persistent presence of senescent cancer cells within the tumor microenvironment (TME) can promote cancer recurrence due to TME remodeling that includes recruitment of myeloid-derived suppressor cells (MDSCs) and chronic SASP that promote senescence bypass mechanisms [77]. Thus, combining senescence-inducing therapy with senolytic drugs could be beneficial to shortand long-term outcomes in cancer patients [78,79]. Here, we showed that BLU9931, a highly selective inhibitor of FGF19/FGFR4 signaling, induces senescence in some cases, and further that the senolytic drug, quercetin, is effective on senescent PDAC cells ( Figure 6, Figure 7, Figures S5 and S7). Quercetin exerts multiple effects in different cancers, such as targeting BCL-2 family members, hypoxia inducible factor-1α, and PI3-kinase, and p21-related anti-apoptotic pathways [80]. By contrast, dasatinib exerts more limited effects, such as targeting Src family kinases and other key senescent cell anti-apoptotic pathway elements [44]. In addition to regulating different targets, quercetin and dasatinib may exert their effects more efficiently in different cell types. For example, quercetin targets senescent endothelial cells but not senescent adipocyte progenitors, while dasatinib targets senescent adipocyte progenitors but not senescent endothelial cells [54]. Thus, the effect of quercetin alone on PDAC cells may be due to the difference in the mechanisms of action between quercetin and dasatinib.
The sirtuins are members of a family of deacetylases (SIRT1 through SIRT7) that modulate numerous cellular processes [53]. SIRT1 and SIRT6 are nucleus-localized sirtuins that also modulate cellular senescence, and downregulation of SIRT1 and SIRT6 has been demonstrated to induce premature senescence in a variety of cell types [81][82][83]. In the present study, we determined that BLU9931 reduced SIRT1 and SIRT6 levels, leading to the induction of premature senescence in PDAC cells ( Figure 6). It is not clear, however, whether inhibition of SIRT1 or SIRT6, per se, was sufficient to cause the observed effects on senescence since rescue experiments to restore SIRT1 or SIRT6 activity and/or expression were not carried out in the present study. Nonetheless, quercetin but not dasatinib is known to inhibit SIRT6 activity [84]. It is possible, therefore, that due to its actions on SIRT6, quercetin was able to enhance the effects of BLU9931.
Our analysis of the effects of BLU9931 on cell cycle-regulating genes revealed that in responsive PDAC cells, BLU9931 upregulated p27 and p21, both of which slow down progression through the G1 phase of the cell cycle, and downregulated CDK1, which contributes to the cell's ability to progress through the G1/S and G2/M phases of the cell cycle. The net effect is that BLU9931 attenuates cell cycle progression as shown in Figure 3C. Moreover, BLU9931 upregulated E2F1 expression. Elevated E2F1 levels can promote cell cycle arrest and/or apoptosis [85], and the possibility that BLU9931 may activate this pathway warrants further studies.
Our current findings suggest that BLU9931 may have a therapeutic potential in patients with PDAC and that a precision medicine-based strategy could be devised based on FGF19/FGFR4 expression and downstream signaling analysis. In addition, therapy with BLU9931 could lead to novel treatment strategies in PDAC perhaps via the sequential addition of senolytic drugs, such as quercetin, its analogs, or the PARP inhibitor olaparib, all of which could have fewer side effects than anticancer drugs. Importantly, BLU9931 also upregulated the expression of GM-CSF, which can promote pancreatic cancer progression and metastasis [86,87]. However, FGFR4 can also activate pro-metastatic signaling [88] and can crosstalk with EGFR by promoting amphiregulin expression [89]. Moreover, EGFR activation is essential for oncogenic Kras-driven PanIN progression to PDAC [90]. It will be important, therefore, to determine whether the combination of BLU9931 and senolytic agents will abrogate the upregulation of GM-CSF expression and suppress the metastatic potential of PDAC cells, and to assess the need for inhibiting GM-CSF actions in any therapeutic trials for PDAC with BLU9931.

Immunohistochemistry
Paraffin-embedded sections from pancreas tissue microarrays (PA1002, PA2072; US Biomax, Inc. Rockville, MD, USA) were immunostained using Histofine Simple Stain MAX PO (R) kit. After deparaffinization, antigen retrieval was performed at 121 • C for 15 min in a sodium citrate buffer solution (pH 6.0). Then, endogenous peroxidase activity was blocked by incubating sections for 30 min with 0.3% hydrogen peroxide in methanol. The tissue sections were incubated overnight with the rabbit polyclonal anti-FGFR4 antibody (1:100 in dilution; Sigma-Aldrich, St. Louis, MO, USA). Bound antibodies were detected with Simple Stain MAX PO (R) reagent, using diaminobenzidine-tetrahydrochloride chromogen (DAB) as the substrate. Negative control studies were performed by omitting the primary antibodies. For the evaluation of the intensity of FGFR4 immunostaining in the tissue microarray, the following scale was employed: 0, no staining; 1+, mild staining; 2+, moderate staining; and 3+, intense staining. Two pathologists (H. Y. and Y. M.) independently evaluated the staining results. Scales 0 and 1 were low, and 2 and 3 were divided into FGFR4 highly localized groups. Correlation between FGFR4 expression and tumor grades were analyzed in 129 cases, and the other factors were analyzed in 136 cases, because of the provided clinical data from the tissue arrays. Microscopic images were taken using Mantra, multi-spectral microscopy (PerkinElmer, Waltham, MA, USA), and then images were loaded into inform software ver. 2.4 (PerkinElmer) to separate specific chromogens, DAB, and hematoxylin, by using the spectral compensation library built in the software according to the manufacturer's instruction.

Cell Culture
The human PDAC cell lines PK-1, PANC-1, PK-59, and MIA PaCa-2 were obtained from the Cell Resource Center for Biomedical Research, Institute of Development, Aging and Cancer, Tohoku University (Sendai, Japan). PK-45P, PK-8, T3M-4, and KP4 human PDAC cells were provided by the RIKEN BRC through the National Bio-Resource Project of the MEXT, Japan. Cells were grown in growth medium (RPMI 1640 medium containing 10% fetal bovine serum) at 37 • C under a humidified 5% CO 2 atmosphere. For 3-D culture, cells in growth medium were plated at 1.0 × 10 4 cells/well or 3.0 × 10 3 cells/well in 24-well ultra-low attachment plates (Corning Inc., Kennebunk, ME, USA) or 96-well ultra-low attachment plates (Thermo Fisher Scientific, Waltham, MA, USA), respectively. The spheres were aspirated after 7 days using micropipettes and placed in microcentrifuge tubes for use in further experiments. Cells were treated with BLU9931 (EMD Millipore Corp., Burlington, MA, USA) or 0.005% dimethyl sulfoxide as vehicle control for 3 days, and after further experiments were performed.

Genotyping
Genomic DNA was extracted from PDAC cell lines using DNeasy Blood & Tissue Kit (QIAGEN, Hilden, Germany) according to the manufacturer's instructions. Genotypes were determined according to the previous report [91]. Briefly, FGFR4 exon 9 was amplified using primers (Forward: 5 -GAC CGC AGC AGC GCC CGA GGC CAG-3 ; Reverse: 5 -AGA GGG AAG AGG GAG AGC TTC TG-3 ) and analyzed for Gly388Arg polymorphism. PCR products were digested with BstNI (New England Biolabs Inc., Ipswich, MA, USA). SNP Arg388 in FGFR4 gene was characterized by two distinctive fragments of 82 and 27 bp. On the other hand, the FGFR4 Gly388 wild-type allele was identified by a single fragment of 109bp.

Real-Time qPCR
Total RNA was isolated from cells using the RNeasy plus mini kit (QIAGEN) and subsequently reverse-transcribed using the ReverTra Ace ® qPCR RT Kit (Toyobo, Osaka, Japan) according to the manufacturer's instructions. Real-time qPCR was performed using the Power Sybr ® Green kit (Applied Biosystems, Foster City, CA, USA) and a StepOnePlus ™ real-time PCR system (Applied Biosystems). β-actin was amplified as an internal control. Primer sets for real-time qPCR are listed in Table S1. Real-time qPCR analysis for FGFR4 was performed using TaqMan Fast Universal PCR Master Mix (Life Technologies Corporation, Carlsbad, CA, USA) and TaqMan Gene Expression Assays (Life Technologies), with 18S rRNA serving as an internal control.

Cell Proliferation Assay
Cells were cultured in growth medium at a density of 3 × 10 3 cells/well in 96-well plates followed by incubation for 72 h. ATP assays were used to examine proliferation using the CellTiter-Glo ® 2.0 Assay (Promega, Madison, WI, USA) according to the manufacturer's protocol.

Apoptosis Assay
Measurement of the percent of apoptotic cells was performed using an Annexin V-Biotin Apoptosis Detection kit (BioVision Inc., Milpitas, CA, USA) per the manufacturer's instructions. Cells were analyzed using a FACSAria™ Cell Sorter.

Cell Cycle Assay
Cells were washed in PBS, resuspended in PBS, and stained with cell cycle assay solution (deep red; Dojindo Molecular Technologies, Inc., Rockville, MD, USA) at 37 • C for 15 min. The cell cycle profiles were obtained using a FACSAria™ Cell Sorter at 640 nm. Data were analyzed using FlowJo software (Becton Dickinson).

Anti-Drug Resistance Assay
Cells (3.0 × 10 3 cells/well) were plated in 96-well culture plates with growth medium with or without BLU9931. Each anticancer drug was administered at the indicated concentration after 3 days culture and cell growth rates were measured by ATP assays 4 days after treatment of anticancer drug. Cell viability was calculated as the percentage of luminescence in drug-treated cells relative to non-treated control cells.

Wound Healing/Cell Scratch Assay
Cells were seeded at a density of 7 × 10 3 cells on each side of an Ibidi culture insert (Ibidi GmbH, Martinsried, Germany), with a 500-µm separation between each side of the well, and allowed to grow for 24 h in growth medium. After removal of the insert, cells were incubated in the same medium. Cells were photographed using phase contrast microscopy at insert removal (0 h) and following 6 and 24 h of incubation. Accurate measurements of the wounds were taken during the time course to calculate the migration rate according to the equation: Percentage of wound healing = [(wound area at 0 h) − (wound area at 6 or 24 h)]/(wound length at 0 h) × 100.

Boyden Chamber Assay
Cell culture inserts (8-µm pore size and 6 mm in diameter) were used according to the manufacturer's instructions. Cells were plated at a density of 1 × 10 5 cells/500 µL on the upper chamber of the inserts with serum-free medium, and medium containing 10% serum was used in the lower chamber. Six hour later, cells that had migrated through the membrane pores were fixed and stained with a Diff-Quick staining kit (Polysciences, Inc., Warrington, PA, USA), and counted under a light microscopy.

Invasion Assay
Invasion assays were performed using Corning Matrigel invasion chambers (pore size: 8 µm, Discovery Labware Inc., Bedford, MA, USA). Cells were plated at a density of 1 × 10 5 cells/500 µL on the upper surface of the inserts, and, 16 or 24 h later, cells that had migrated through the membrane to the lower surface of the filter were fixed and stained with a Diff-Quick staining kit (Polysciences, Inc.) and counted under a light microscope.

SA-β-Gal Assay
SA-β-Gal activity was assayed using a senescence detection kit (BioVision Inc., Milpitas, CA, USA) as described in our previous reports [92]. Briefly, cells were washed twice with PBS, exposed to fixation solution for 10 min, and then incubated overnight in freshly prepared staining solution. After staining, cells were counterstained with DAPI and then four fields of view were examined microscopically. By counting the number of SA-β-Gal-positive cells based on their blue color and the total number of cells stained with DAPI using ImageJ software, the percentage of SA-β-Gal-positive cells was calculated to estimate the percentage of senescent cells.

Telomere Length Assay
Telomere length was quantified by real-time PCR using a Telomere Length Quantification qPCR Assay Kit (ScienCell Laboratory, Carlsbad, CA, USA) according to the manufacturer's protocol. Genomic DNA samples were extracted as described in Section 4.3.

Immunostaining
Cells were fixed with 4% (w/v) paraformaldehyde and washed. Subsequently, cells were permeabilized with 0.1% [v/v] Triton ™ X-100 and blocked with PBS containing 1% (w/v) BSA and 5% (v/v) normal goat serum. After washing, cells were incubated with monoclonal mouse anti-γH2A.X (sc-517348; Santa Cruz Biotechnology, Dallas, TX, USA) at 4 • C overnight. After washing, cells were stained with Alexa Fluor ® 488-conjugated secondary antibodies (Molecular Probes), and then counterstained with DAPI. Immunofluorescence images were acquired using a confocal laser scanning microscope (Leica Microsystems, Wetzlar, Germany). For the quantification of γH2A.X-positive cells, the numbers of γH2A.X-positive cells with green color and the total numbers of cells stained with DAPI from four fields were counted and then the percentage of γH2A.X-positive cells was calculated.

Senolytic Drug Sensitivity Assay
Cells (3.0 × 10 3 cells/well) that were incubated in the presence or absence of BLU9931 for 7 days were plated in 96-well culture plates containing the respective media. Thus, only cells previously incubated with BLU9931 contained the FGFR4 inhibitor. Each senolytic drug, Quercetin (Cayman Chemical, Ann Arbor, MI, USA) or Dasanitib (Cayman Chemical), was then added as indicated and cell growth rates were determined by the ATP assay 4 days later. Cell viability was calculated as the percentage of luminescence in drug-treated cells relative to non-treated control cells.

Statistical Analysis
Results were shown as means ± SD, and differences between two groups were compared using Student's t test or Welch's t test. One-way ANOVA was performed when comparing multiple groups. Chi-square and Fisher's exact tests were used to analyze correlations between FGFR4 expression and clinicopathological features.

Conclusions
In this study, we demonstrated that inhibition of signal transduction pathways through ERK, AKT, and STAT3 pathways by the FGFR4 inhibitor BLU9931 inhibited PDAC cell proliferation and invasion, in part by downregulating MT1-MMP in autocrine/paracrine FGF19/FGFR4 signaling-positive PDAC cells ( Figure 7B). Furthermore, downregulation of SIRT1 and SIRT6 by BLU9931 may have contributed to senolysis in those cells ( Figure 7B). Thus, we propose that BLU9931 is a promising drug for the therapy of FGFR4-positive PDAC.

Conflicts of Interest:
The authors declare no conflict of interest.