FRAX597, a PAK1 inhibitor, synergistically reduces pancreatic cancer growth when combined with gemcitabine

Background
Pancreatic ductal adenocarcinoma remains one of the most lethal of all solid tumours. It is the fourth leading cause of cancer-related mortality in Australia and the United States and is expected to become the second leading cause of cancer-related deaths by 2030, based on current management with no significant treatment im- provements [1]. The 5-year survival rate is less than 5 % and has not improved over the past few decades [2]. Although combinational chemotherapies exist such as FOLFIRINOX and gemcitabine with nab-paclitaxel, as a single agent, gemcitabine remains the standard of care for the treatment of pancreatic cancer in most countries [3]. Gemcitabine targets rapidly replicating cells by inhibiting DNA synthesis, but intrinsic and acquired chemoresistance is common. The limited treatment regi- mens and predicted increase in cancer-related mortality highlight the urgent need for the development of effect- ive therapies based on our understanding of the molecu- lar mechanisms involved in pancreatic cancer.
The most frequent and earliest mutation in pancreatic a number of signalling pathways, including other p21 proteins such as Cdc42 and Rac, through direct and in- direct mechanisms [5]. These p21 proteins can then acti- vate the p21-activated kinases (PAKs). Although there is evidence for the activation of PAKs by Kras-driven path- ways, other non-Kras-driven pathways or indirect Kras mechanisms may also activate PAKs.PAKs are a family of non-receptor serine/threonine kinases, which mediate many effector functions from cell cycle and DNA transcription to cell adhesion and motil- ity [6]. There are six isoforms of PAKs, divided into two groups: Group 1 contains PAK1, 2 and 3, and Group 2 contains PAK4, 5 and 6. Of the six isoforms, PAK1 is the best documented and has been found to be up-regulated in a number of cancers [7], including pancreatic cancer [8]. PAK1 is also up-regulated in pancreatic cancer cell lines when expression of MUC13, a transmembrane mucin, is increased [9]. We have previously found that a non-selective PAK inhibitor, glaucarubinone, reduced pancreatic cancer growth, and that treatment in combin- ation with gemcitabine resulted in synergistic inhibition [10]. The role of PAK1 in pancreatic cancer and its therapeutic potential have not been fully elucidated.

FRAX597 is a small-molecule pyridopyrimidinone that targets group 1 PAKs through binding to the ATP- binding site [11]. Although it has been found to prefer- entially target group 1 PAKs, FRAX597 also inhibits other kinases such as RET, YES1, TEK, and CSF1R [12]. Of the group 1 PAKs, FRAX597 selectively inhibits PAK1 with a kinase IC50 of 8 nM, compared to 13 nM and 19 nM with PAK2 and PAK3, respectively. No inhib- ition of the group 2 PAKs was observed [11]. FRAX597 inhibits proliferation in neurofibromatosis type 2 (NF2)- associated schwannomas [12], but has not been tested previously in pancreatic cancer. This study aimed to elu- cidate the role of PAK1 in pancreatic cancer, by examin- ing the effects of reduction of PAK1 expression by shRNA knock-down, or PAK1 activity with the select- ive inhibitor FRAX597, on the growth and migration/ invasion of pancreatic cancer cell lines in vitro, and in orthotopic murine models in vivo, alone and in combination with gemcitabine.The human PANC-1, MiaPaCa-2 and BxPC-3 pancreatic cancer cell lines (American Type Culture Collection, Manassas, VA) and the murine Pan02 (Division of Cancer Treatment and Diagnosis Tumor Repository, NCI, Frederick, MD), and LM-P (obtained from Andrew Lowy (Moores Cancer Center, University of California, San Diego, CA) [13]) pancreatic cancer cell lines were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) supplemented with 10 % FBS (fetal bovine serum: Hyclone Laboratories Inc., Scoresby, Australia). Normal immortalised human pancreatic duct epithelial (HPDE) cells (obtained from M.S Tsao (Ontario Cancer Institute, Ontario, Canada)) were cultured in Keratinocyte serum-free medium supplemented with bovine pituitary extract (BPE) and epidermal growth factor (EGF). All cells were cultured in a 37 °C incubator with 5 % CO2. Cells were tested regularly for mycoplasma contamination and were not passaged more than 30 times or for more than 6 months after resuscitation. FRAX597 was purchased from SYNthesis (Parkville, Australia) and gemcitabine was purchased from Sigma- Aldrich (Castle Hill, Australia).

Human tissue collection was approved by the Austin Health Human Research Ethics Committee (H2013/ 04953) and informed consent was obtained from all par- ticipants. Samples of 10 human pancreatic cancers and adjacent normal pancreas were collected from patients undergoing pancreatic cancer resection at Austin Health, and confirmed to have pancreatic ductal adenocarcin- oma by two independent pathologists. For PAK1 IHC, sample sections were incubated with 3 % hydrogen per- oxide in methanol for 10 min at room temperature to quench endogenous peroxidase activity. Antigens were retrieved by incubation in 10 mM citrate buffer and blocked in 5 % horse serum. Sections were incubated with antibody against PAK1 (Santa Cruz Biotechnology, Dallas, TX) or IgG. Sections were visualised using an ENvision Plus polymer-based detection kit (Dako, Botany, Australia). The slides were then counter-stained and images were taken with a NIKON Coolscope (Coherent Scientific, Hilton, Australia).To obtain PAK1 knock-down (KD) clones, PANC-1 and MiaPaCa-2 cells were transfected with SureSilencing shRNA plasmids for human PAK1 (SABioscience, Doncaster, Australia), or with a scrambled sequence as a negative control (NC), using Lipofectamine2000 (Invitrogen, Mulgrave, Australia), according to the manufacturer’s instructions. Stable clones were selected with geneticin (G418; 1 mg/ml). PAK1 protein expression was detected by western blot.Proteins in cell lysates were detected with antibodies against phospho-PAK1 (Santa Cruz Biotechnology), PAK1, phospho-AKT, AKT, HIF1α (BD Biosciences, North Ryde, Australia), and GAPDH. Antibodies were from Cell Signalling Technology (Arundel, Australia), unless otherwise stated. Bound antibodies were visualized using ECL reagents (GE Healthcare, Amersham, UK), and the density of each band was analysed using Multigauge computer software (Berthold, Bundoora, Australia). HIF1α expression was determined in cells cultured under nor- moxia or hypoxia (1 % O2).

Cell proliferation and survival was measured using 3H- thymidine incorporation and withdrawal assays, respect- ively, as previously described [14]. Growth curves were fitted based on a log-scale using MATLAB (MathWorks, Natick, MA) and differences in proliferation were evalu- ated by comparison of growth rates (expressed as %/h). Assessment of proliferation with FRAX597 and the com- bination of FRAX597 with gemcitabine was measured as previously described [10]. For assessment of cell survival with FRAX597, cells were seeded with increasing con- centrations of FRAX597 for 24 h without serum.The combined effects of FRAX597 and gemcitabine were evaluated using the Chou-Talalay method [15] as previously described [10]. The CalcuSyn program (Biosoft, Cambridge, UK) was used to calculate the combination index (CI) for each drug affected fraction (Fa). The CI value is interpreted as: <1, synergistic; =1, additive; >1, antagonistic.Cell migration/invasion was measured using the Transwell Boyden chamber assay as previously de- scribed [14]. Cells were seeded into the upper chambers of the inserts (ThinCert™, 8 μm pore size; Greiner Bio-One, Frickenhausen, Germany) with increasing concentrations of FRAX597. After 24 h, membranes were fixed and stained with Quick-Dip (Fronine, Riverstone, Australia) and 24 fields were counted at 40 times magnification using a NIKON Coolscope.All mice experiments were approved by the Austin Health Animal Research Ethics Committee (A2013/ 04898). Pan02 cells were implanted orthotopically in the pancreatic head or tail as previously described [16].

For assessment of tumour growth, 28 mice were implanted with cells in the pancreatic tail and monitored for 30 days. 7 mice per treatment group were randomly allo- cated to the four treatment groups: control, intraperito- neal (i.p.) injection of saline every other day; FRAX597 alone, FRAX597 (3 mg/kg) i.p. every other day; gemcita- bine alone, gemcitabine (40 mg/kg) i.p. twice weekly; and combination of FRAX597 and gemcitabine, follow- ing the individual treatments as described above. A sin- gle investigator measured the dimensions of all tumours, at the endpoint, using micro-calipers, in a double- blinded manner. Tumour volume was calculated usingthe formula for ellipsoid tumours: V = L x W x H x (π/6) where L was the longest distance from right to left; W, the largest dorsal/ventral diameter; and H, the largest rostral/ caudal diameter. For assessment of survival, 54 mice were implanted with cells in the head of the pancreas. Mice were monitored based on health score for up to 45 days and euthanased when a poor health score was reached. Mice were treated with either control, gemcitabine alone, or combination of FRAX597 and gemcitabine, as de- scribed above. An initial study was undertaken with 24 mice, with 12 mice per group for control or gemcitabine treatment alone. A second study was undertaken with 30 mice, with 13 mice per group for gemcitabine alone or combination treatment, and the remaining 4 mice allo- cated to the control group. A collated Kaplan-Meier sur- vival curve was plotted, and the two studies were analysed together using stratified Cox regression analysis (SPSS; IBM, New York, NY).All values are expressed as means ± standard error. Ex- periments were done in duplicate and data collated from three independent experiments. Results were analysed using student’s t-test or one-way ANOVA (SPSS). Differ- ences between two means with p < 0.05 were considered significant.

Results
PAK1 staining of pancreatic ductal adenocarcinoma cells was observed in all 10 human pancreatic cancer samples tested (Fig. 1a). In corresponding normal pan- creas samples, islet cells stained for PAK1, however, staining was absent in acinar and ductal epithelial cells. Expression of PAK1, and of the phosphorylated, active form of PAK1, was detected in low levels in the normal HPDE cell line and was significantly lower when compared to all the pancreatic cancer cell lines. All human and murine pancreatic cancer cell lines tested expressed phosphorylated and total PAK1 (Fig. 1b).Inhibition of PAK1 by shRNA knock-down decreases proliferation and survival of pancreatic cancer cellsThe PAK1 protein concentrations in two PANC-1 PAK1 KD clones (2.05 and 2.10) were decreased to 22 % and 24 %, respectively, of the PAK1 protein concentrations of the corresponding NC cells, which had been trans- fected with scrambled sequences (Fig. 2a). Similarly, the PAK1 protein concentrations in two MiaPaCa-2 PAK1 KD clones (3.09 and 3.12) were decreased to 11 % and 9 %, respectively, of the PAK1 protein expression of the corresponding NC cells (Fig. 2b). The proliferation ratewas significantly reduced in both PANC-1 (Fig. 2c) and MiaPaCa-2 (Fig. 2d) PAK1 KD cells compared to the corresponding NC cells. The growth rate of two clones of PANC-1 PAK1 KD cells (1.9 %/h and 2.1 %/h) was significantly less than two clones of NC cells (2.4 %/h and 2.6 %/h) (Table 1). A similar difference was observed in the MiaPaCa-2 PAK1 KD cells.Proliferation of PANC-1 (Fig. 2e) and MiaPaCa-2 (Fig. 2f) PAK1 KD cells in the presence of gemcitabine at concentrations of 20 nM and 50 nM was inhibited to a greater extent than the corresponding NC cells.

The IC50 values of two clones of PANC-1 PAK1 KD cells (20 nM and 21 nM) (Table 1), were significantly less than the values for NC cells (26 nM and 39 nM). Similarly, the IC50 values of two clones of MiaPaCa-2 PAK1 KD cells (26 nM and 25 nM) (Table 1) weresignificantly less than the values for NC cells (29 nM and 28 nM).AKT activity was significantly reduced in two clones of PANC-1 PAK1 KD cells, by 22 % and 31 % (Fig. 3a), and in two clones of MiaPaCa-2 PAK1 KD cells by 24 % and 33 % (Fig. 3b). HIF1α expression was significantly re- duced in two clones of both PANC-1 (Fig. 3c) and MiaPaCa-2 (Fig. 3d) PAK1 KD cells compared to the NC cells under either normoxia or hypoxia.FRAX597 decreases proliferation and migration/invasion in pancreatic cancer cell linesFRAX597 inhibited proliferation in all pancreatic cancer cell lines in a dose-dependent manner (Fig. 4a), with IC50 values between 650 nM for BxPC-3 cells and2.0 μM for PANC-1 cells (Table 2). Similarly, FRAX597inhibited migration and invasion in all pancreatic cancer cell lines in a dose-dependent manner (Fig. 4b), with IC50 values between 105 nM for MiaPaCa-2 cells and 605 nM for Pan02 cells (Table 2). FRAX597 inhibited survival of LM-P cells in the absence of FBS in a dose- dependent manner with an IC50 value of 1.10 μM (Fig. 4c). Significant inhibition of survival of PANC-1, MiaPaCa-2, BxPC-3, and Pan02 cells was only observed at concentrations greater than 1 μM.FRAX597 synergises with gemcitabine in inhibiting pancreatic cancer cell growthGemcitabine alone inhibited proliferation in all pancre- atic cancer cell lines (Fig. 5a-e) in a dose-dependent manner, with IC50 values between 5 nM for BxPC-3 cells and 80 nM for Pan02 cells (Table 2). A further reduction in proliferation was observed in all pancreatic cancer cell lines (Fig. 5a-e) when FRAX597 was combined with gemcitabine, compared to gemcitabine alone.

The with control or FRAX597 alone. Mice treated with gem- citabine alone had significantly reduced tumour volume when compared to control or FRAX597 alone, and a fur- ther significant reduction in tumour volume was ob- served for the mice treated with the combination of FRAX597 and gemcitabine (Fig. 7a). A similar trend was found when mice were evaluated for the presence of peritoneal carcinomatosis. 43 % of mice in the combined treatment group had peritoneal carcinomatosis com- pared to 71 % of mice in the gemcitabine treatment group, and 100 % of mice in the control and FRAX597 treatment groups (Fig. 7b).Survival of mice in the combination treatment group was significant increased compared to the control group,as assessed by a stratified Cox regression analysis (Fig. 7c). A rates ratio of 7 was calculated, indicating that mice in the control group had a mortality rate 7 times greater than mice in the combination treatment group (Table 3). The rates ratio of 2.7 for mice in the gemcitabine alone group, compared to mice in the combination treat- ment group, was not statistically significant (p = 0.09).

Discussion
Our finding that PAK1 is expressed in pancreatic cancer is in agreement with previous studies [8, 17]. We confirmed that PAK1 was not expressed in normal pan- creatic acinar or ductal cells, which are the likely pro- genitors of pancreatic cancer [18]. In contrast, PAK1 was expressed in the tumour tissue, in which the pancre- atic ductal adenocarcinomas cells stained positive (Fig. 1a). All pancreatic cancer cell lines showed upregu- lation of PAK1 compared to the normal pancreas cellReduction of PAK1 expression by shRNA knock-down (Fig. 2a-b) inhibited proliferation of the PANC-1 and MiaPaCa-2 pancreatic cancer cell lines (Fig. 2c-d), likely through modulation of the AKT pathway. These two hu- man cell lines were chosen based on the PAK1 activity where PANC-1 is considered ‘high’ activity whilst MiaPaCa-2 is considered ‘low’ activity (Fig. 1b). This dif- ference may contribute to the contrasting results in cell survival where a reduction was observed in PANC-1 PAK1 KD cells (Additional file 1: Figure S1A) but not in MiaPaCa-2 PAK1 KD cells (Additional file 1: Figure S1B). The suggestion that ‘high’ PAK1 expressing cells may be driving cell survival whereas ‘low’ PAK1 expressing cells may rely on other mechanisms to drive cell survival re- quires further investigation. The reduction in cell growthin PAK1 KD cells was associated with a decrease in AKT activity (Fig. 3a-b), but not in ERK activity (Additional file 1: Figure S1C-D). Our group has previously found that PAK1 mediated growth of colorectal cancer cell lines by both ERK and AKT pathways [14], while an- other group has found that PAK1 signalled preferen- tially through the ERK pathway to control skin cancer growth [11].

Thus, PAK1 signalling through AKT and ERK pathways is dependent on the cancer type, and our study suggests that PAK1 mediates pancreatic can- cer cell growth through the AKT pathway rather than the ERK pathway.PAK1 may play a role in the resistance of pancreatic cancer to hypoxia through regulation of HIF1α. The transcription factor HIF1α regulates oxygen delivery andmetabolic adaptation to hypoxia and has been found to be a prognostic marker for pancreatic cancer [21]. Pan- creatic tumours are known to be highly hypoxic, as they feature a dense desmoplastic reaction (stroma), which may contribute to pancreatic cancer invasion, metastasis, and resistance to therapy [22]. Thus, mediators of sur- vival in response to a hypoxic challenge are attractive therapeutic targets for pancreatic cancer. Although, as far as we are aware, this is the first study to examineHIF1α as a downstream effector of PAK1 in pancreatic cancer, PAK1 has previously been linked to HIF1α in colorectal cancer [23]. The ability of PAK1 to contribute to pancreatic carcinogenesis via multiple signalling path- ways enhances its potential as a therapeutic target.PAK1 knock-down also enhanced the sensitivity of PANC-1 and MiaPaCa-2 cells to gemcitabine (Fig. 2e-f), as revealed by comparison of the IC50 values for inhib- ition of proliferation between control and knock-downclones (Table 1). Although gemcitabine remains a stand- ard monotherapy treatment for pancreatic cancer pa- tients, combining treatments with gemcitabine with the goal of decreasing chemotherapy-associated cytotoxicity and chemo-resistance and increasing survival has had varied results [3]. Previous studies have found that the PAK1 downstream effectors AKT and HIF1α could play a role in gemcitabine resistance through NFκB which limits gemcitabine uptake by decreasing nucleoside transporters such as hENT and hCNT [3]. Furthermore, PAK1 has been shown to regulate NFκB transcription upstream of fibronectin regulation in pancreatic cancer [8].

Although further investigation is required, the data presented herein supports the use of a PAK1 inhibitor combined with gemcitabine to limit gemcitabine cyto- toxicity and chemo-resistance.The overall statistics for the stratified Cox regression analysis were:χ2 (2) = 9.9, p = 0.007The group 1 PAK-selective inhibitor FRAX597 inhib- ited proliferation, migration/invasion, and survival of all pancreatic cancer cell lines tested (Fig. 4a-c). Although FRAX597 also inhibits other kinases such as RET, YES1, TEK, and CSF1R [12], the similar results obtained in the PAK1 knock-down experiments suggest that in this case PAK1 is indeed the relevant target. Furthermore, the IC50 values for proliferation are similar to the value ob- served in NF2-null Schwann cells [12]. However, the IC50 values for either proliferation or migration/invasion did not significantly correlate with the amount of active PAK1 in the pancreatic cancer cells (data not shown). This observation suggests that there may be a barrier (e.g. uptake at the cell membrane) that prevents realisa- tion of the full potential for inhibition in intact cells. The presence of such a barrier could have contributed to the failure to detect any difference in tumour volume between the FRAX597-treated mice and the control mice in our in vivo study. Furthermore, the dense des- moplastic reaction may have also prevented the drug’s uptake by the tumour. These observations illustrate the importance of the microenvironment in assessment of a drug’s efficacy, as the in vitro cell culture conditions may not fully mimic the clinical setting.Combination of the PAK1 inhibitor, FRAX597, with gemcitabine resulted in increased inhibition of PAK1activity in some, but not all, of the pancreatic cancer cell lines tested (Fig. 6a-e). In all the pancreatic cancer cell lines tested, PAK1 activity was significantly decreased after treatment with FRAX597 alone, but no change in activity was observed after treatment with gemcitabine alone.

Thus, combined treatment with FRAX597 and gemcitabine might be expected to inhibit PAK1 to the same extent as FRAX597 treatment alone, as was ob- served for PANC-1, Pan02 and LM-P cells. The signifi- cantly greater inhibition observed in MiaPaCa-2 and BxPC-3 cells after combination treatment provided clear evidence for synergy, although the mechanism for this is unclear. Interestingly, these two pancreatic cancer cell lines had the lowest phospho-PAK1 expression of all the pancreatic cancer cell lines tested. This observation sug- gests that phospho-PAK1 may be a predictive marker for gemcitabine response, as has recently been shown for PAK4 in pancreatic cancer [24].Treatment with FRAX597 combined with gemcitabine significantly decreased tumour volume in vivo (Fig. 7a) and revealed a promising trend towards decreasing me- tastasis (Fig. 7b) and increasing survival (Fig. 7c). Fur- thermore, Ki67 staining of the tumours indicated that the difference in tumour volume was due to inhibition of proliferation (Additional file 2: Figure S2). Although liver metastasis is often observed in the orthotopic pan- creatic tail murine model, a total of only three mice, from control and FRAX treatment groups, had liver me- tastases at sacrifice, so no comparison could be under- taken [16]. However, peritoneal carcinomatosis, or peritoneal spread, was present and was compared. As a difference in tumour volume was observed between ani- mals treated with gemcitabine alone or with the combin- ation of FRAX597 and gemcitabine, a decrease in peritoneal carcinomatosis and an increase in survival was expected, but significance was not reached. This may be due to the fact that the study was stopped early, before all mice were euthanised because of tumour- related illness. Although the potential clinical value of FRAX597 and the likely therapeutic benefit of targeting PAK1 are clearly established by the data in Fig. 4, longer studies are needed for a complete picture of the possible survival benefits of combination treatment.

Conclusion
PAK1 is upregulated in human pancreatic cancer. Knock-down experiments indicated that PAK1 is re- quired for proliferation and survival of human pancre- atic cancer cell lines through AKT- and/or HIF1α- dependent pathway(s). Furthermore, PAK1 knock-down sensitised pancreatic cancer cells to gemcitabine. A group 1 PAK-specific inhibitor, FRAX597, inhibited pro- liferation, migration/invasion, and survival of human pancreatic cancer cell lines. When combined with gemcitabine, FRAX597 synergistically inhibited pancre- atic cancer growth in vitro and in vivo. This study sug- gests the promise of inhibiting PAK1 function and defines areas for further investigation to clarify its poten- tial value as a target for pancreatic cancer therapy.Human ethics approval was obtained from the Austin Health Human Research Ethics Committee (H2013/ 04953) and all participants gave consent to participate and for publication. All mice experiments were approved by the Austin Health Animal Research Ethics FRAX597 Committee (A2013/04898).