A hypusine-eIF5A-PEAK1 switch regulates the pathogenesis of pancreatic cancer

Cell lines, siRNAs and DNA constructs

Three PDAC cell lines (FG, PANC1, 779E) and HPNE (Human Pancreatic Nestin Expressing) cells were used in this study and maintained as described [30, 34]. 779E cells were established from a patient-derived tumor (moderately-to-poorly differentiated PDAC) by A. M. Lowy. FG cells were kindly gifted from Dr. David Cheresh (UCSD), while PANC1 cells and HPNE cells were obtained from American Type Culture Collection. The cell lines were regularly authenticated based on morphological and growth characteristics as well as analysis of KRas mutation, which was most recently conducted in March 2014. Details of siRNAs and DNA constructs are in Supplementary Methods.

Antibodies, chemicals and tissue microarrays

Antibodies directed at eIF5A1 (rabbit monoclonal (EP526Y): ab32443) and eIF5A2 (mouse monoclonal (1E7): H00056648-M01) were purchased from AbCam and Abnova, respectively. Anti-Src and phospho-Src (Tyr416) antibodies were from Cell Signaling Technology. Anti-GAPDH antibody was from Abcam, anti-survivin antibody was from Biolegend, anti-tubulin, DHPS and DOHH antibodies were from Sigma, anti-KRas antibody was from Santa Cruz Biotechnology (sc-30) and anti-PEAK1 antibody was from Millipore. The antibody specifically recognizing hypusinated eIF5A1 (NIH353) was a kind gift from Dr. Myung-Hee Park (NIH), which has been validated for use in western blotting [35] and immunohistochemistry [5, 36] in human as well as mouse tissues and cell lines. GC7 (N1-guanyl-1, 7-diaminoheptane) was purchased from Calbiochem, CPX (ciclopirox olamine) was from Santa Cruz Biotechnology. Cisplatin was purchased from Enzo Life Sciences, and gemcitabine was from Eli Lilly and used as described previously [30]. Human pancreatic cancer tissue arrays were purchased from US Biomax (PA2081a except for DOHH staining, where PA803 was used).

Immunohistochemistry

Human pancreatic cancer tissue array was purchased from US Biomax as described above. PDX-1-Cre:LSL-KRAS^G12D mouse pancreas sections were provided by A. M. Lowy. Samples from patients BK-17 and BK-19 were collected and processed as described previously [30], in accordance with UCSD (La Jolla, CA) Institutional Review Board (IRB) #071136X. Determination of eIF5A1, eIF5A2 and hypusinated eIF5A1 protein expression was performed as described previously using specific antibodies (ab32443 (Abcam) for eIF5A1, H00056648-M01 (Abnova) for eIF5A2 and NIH353 for hypusinated eIF5A1) [30, 35, 36]. Images were collected with a Leica DM2500 microscope using a 40x objective lens.

Western blotting and quantitative PCR (qPCR)

Western blotting and qPCR were performed as described previously [30]. Briefly, cells were lysed in RIPA (radioimmunoprecipitation assay) buffer (20 mM Tris-HCl pH7.4, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS) supplemented with Complete protease inhibitor cocktail (Roche), and equal amounts of cell extracts were resolved by SDS-PAGE, transferred to nitrocellulose membranes, and probed with appropriate primary and HRP-conjugated secondary antibodies (Jackson ImmunoResearch) for western blotting. For experiments in Fig. 2B and C, cells were serum-starved overnight prior to lysis to assess the effect of KRas activation without stimulation from growth factors present in the serum. For qPCR, mRNA was isolated using Trizol (Invitrogen) and RNeasy kit (Qiagen), and cDNA was synthesized using iScript cDNA synthesis kit (BioRad). All PCR reactions were carried out using GoTaq Flexi DNA Polymerase kit (Promega).

Cell proliferation assay, soft agar growth assay and clonogenic survival assay

Cell proliferation assay was performed using the CyQuant direct cell proliferation assay kit (Invitrogen) as described in Supplementary Methods. Soft agar growth assay was conducted as detailed in Supplementary Methods, and the number and size of colonies were analyzed using ImageJ software (NIH). Clonogenic survival assay was performed by plating cells without or with drug treatment into 6-well plates (50 cells/well) and allowing them to form colonies, as described in Supplementary Methods. Subsequently, cells were fixed and stained with crystal violet (Sigma-Aldrich), and the number of colonies was manually counted.

Orthotopic implantation experiments

Orthotopic implantation experiments were performed as described previously [30] and detailed in Supplementary Methods. Briefly, 1× 10⁶ cells were orthotopically injected into the tail of the pancreas of 4-6 weeks old female athymic mice (Jackson Laboratory), which were subsequently sacrificed at the indicated time points to assess the weight of primary tumors.

Oncomine analyses

Normalized eIF5A1, eIF5A2, DHPS, DOHH and PEAK1 expression data were downloaded from Oncomine (Compendia Bioscience). Datasets with statistical significance (p-values< 0.05) were used for the analysis. Heatmaps were generated by Microsoft Excel.

Statistical analyses

All quantified data were plotted and analyzed in GraphPad Prism 6.0 with ANOVA, Student's t-test, or nonlinear regression analysis. Data are representative of at least 3 independent experiments and are reported as mean +/- SD, and * represents P-values < 0.05.

eIF5A1, eIF5A2 and hypusinated eIF5A are increased in human and mouse PDAC tissues in response to activated KRas

We first sought to determine the relative expression levels of hypusinated and total eIF5A proteins in normal and PDAC tissues. Tissue microarrays (TMAs) representing normal pancreatic and PDAC tissues were immunohistochemically stained using antibodies specific to eIF5A1, eIF5A2, and the active, hypusinated form of eIF5A1. Importantly, antibody specificity to each eIF5A isoform was validated by western blotting as shown in Fig. S1A. Normal pancreatic ducts showed only weak expression of eIF5A proteins or hypusinated eIF5A1, whereas robust eIF5A protein expression and hypusination was observed in the majority of PDAC tissues regardless of the differentiation status (Fig. 1A-C). These findings were confirmed in matched PDAC and adjacent uninvolved pancreatic tissues from the same patients (Fig. 1D-F). In contrast, we found that the expression of DHPS and DOHH are high in both normal pancreatic ducts and PDAC tissues (Fig. S1B, C). In fact, there was no significant difference in DHPS levels between normal ducts and PDAC tissues, while DOHH levels were slightly decreased in PDAC tissues (Fig. S1C). These results indicate that eIF5A proteins and hypusination levels are amplified in human PDAC compared to normal pancreatic duct tissues, without a concomitant increase in the expression of DHPS or DOHH.

Activating point mutations in the KRas proto-oncogene are the primary oncogenic drivers in human PDAC [29, 37]. The Pdx-1-Cre;LSL-KRAS^G12D transgenic mouse model faithfully recapitulates early stages of pancreatic intraepithelial neoplasia (PanIN) development, which are induced by KRas activation [38]. We isolated pancreatic tissues at various stages of PanIN development from these animals and stained them with eIF5A1 and hypusinated eIF5A1 antibodies. We were unable to determine the levels of eIF5A2, due to the lack of a suitable antibody to mouse eIF5A2. Interestingly, both eIF5A1 and hypusinated eIF5A1 are increased in PanIN tissues compared to normal pancreatic duct tissues (Fig. 2A). These findings suggest that activated eIF5A1 is upregulated during early stages of PDAC progression in response to KRas activation. In support of these findings, introduction of KRas^G12D into human pancreatic nestin-positive (HPNE) cells strongly increased eIF5A1/2 protein levels and hypusination (Fig. 2B). In contrast, knockdown of activated KRas in PANC1 cells significantly reduced eIF5A1/2 protein levels and hypusination (Fig. 2B). Quantitative PCR (qPCR) analysis showed a modest, but statistically significant, decrease of eIF5A1/2 mRNA levels in KRas-depleted PANC1 cells (Fig. 2C), indicating that eIF5A upregulation by KRas can occur transcriptionally. In support of these findings, Oncomine analyses revealed that eIF5A1/2 mRNA expression levels are increased in patient-derived lung adenocarcinoma tissues with activated KRas^G12D compared to those harboring wild-type KRas (Fig. 2D) [39]. Taken together, these results indicate that KRas activation upregulates eIF5A1/2 expression as well as hypusination during the early stages of PDAC progression.

Inhibition of eIF5A expression or hypusination suppresses PDAC cell growth, whereas overexpression of eIF5A enhances PDAC cell growth in vitro

Hypusination of eIF5A proteins is required for its translational activity in eukaryotic cells [11, 26]. GC7 and CPX are established inhibitors of eIF5A hypusination that target DHPS and DOHH, respectively [8, 18, 19, 26-28]. Previous work has shown that low micromolar doses of CPX induce growth arrest and apoptosis in various cancer cells, as evidenced by decreased survivin expression [40]. Treatment of FG and PANC1 cells with GC7 or CPX potently suppressed cell proliferation, eIF5A1 hypusination, and survivin expression in a dose-dependent manner. Together these findings indicate that eIF5A hypusination is necessary for PDAC cell growth in vitro (Fig. 3A-C).

To directly determine whether eIF5A protein expression is necessary for PDAC cell growth in vitro, we depleted eIF5A1 or eIF5A2 in PANC1 and 779E cells using specific shRNAs. Using this approach we were able to selectively deplete eIF5A1 by greater than 80% and eIF5A2 by greater than 90% in both cell lines (Fig. 4A). Importantly, eIF5A1 knockdown significantly suppressed cell growth by approximately 40% and 70% in PANC1 and 779E cells, respectively, whereas eIF5A2 depletion reduced cell proliferation by 40% in both cell lines. These results were confirmed using an independent set of shRNAs in PANC1 cells (Fig. S3A). In addition, combined depletion of eIF5A1 by shRNA and eIF5A2 by siRNA inhibited growth of PANC1 and 779E cells by approximately 60% and 90%, respectively (Fig. 4B). Furthermore, depletion of either eIF5A1 or eIF5A2 in PANC1 inhibited anchorage-independent growth in soft-agar as evidenced by the significantly reduced size of individual colonies (Fig. 4C). Finally, overexpression of eIF5A1 or eIF5A2 dramatically enhanced growth of PANC1 cells (Fig. 4D) and 779E cells (data not shown). Time course analyses confirmed that eIF5A1/2 knockdown or overexpression mainly affects cell proliferation, not cell death (Fig. S2). These results suggest that eIF5A proteins and hypusination are essential for anchorage-dependent and -independent growth in vitro.

eIF5A is essential for PDAC tumor growth in vivo

To determine whether eIF5A proteins are essential for PDAC tumor growth in vivo, PANC1 or 779E cells expressing control, eIF5A1 or eIF5A2 shRNAs together with GFP, were orthotopically implanted into the pancreas of nude mice. Changes in tumor development were assessed using fluorescence whole-body imaging and by measuring primary weight of resected tumors. eIF5A1 depletion in PANC1 and 779E cells reduced tumor weight by greater than 50% compared to control tumors (Fig. 5A, Fig. S3B). Similarly, eIF5A2 depletion in the 779E cells reduced tumor formation compared to control tumors (Fig. 5B, Fig. S3B). Also, eIF5A2 depletion in PANC1 cells reduced tumor formation, but did not reach statistical significance (Fig. 5A). The presence of eIF5A1 or eIF5A2 knockdown in the orthotopic tumor tissues was validated by qPCR (Fig. S3C and D). These findings suggest that eIF5A1 is essential for PDAC tumor growth, whereas eIF5A2 expression contributes to PDAC growth in vivo in a context-dependent manner.

To determine the effect of overexpression of eIF5A proteins for PDAC tumor growth in vivo, we assessed orthotopic tumor formation of PANC1 cells overexpressing eIF5A1 or eIF5A2. As shown in Fig. 5C, the average tumor mass of eIF5A1-overexpressing PANC1 cells was significantly increased compared to control tumors. eIF5A2 overexpression also enhanced tumor growth albeit to a lesser degree (Fig. 5C). The presence of eIF5A1 or eIF5A2 overexpression in the tumor tissues was determined as above (Fig. S3E). Taken together these findings underscore the importance of eIF5A1 and eIF5A2 in driving PDAC tumor growth in vitro and in vivo.

eIF5A regulates the expression of PEAK1, a novel non-receptor tyrosine kinase

PEAK1 is a newly identified non-receptor tyrosine kinase that modulates Src kinase activity and plays an essential role in driving PDAC malignancy [30-33]. Like eIF5A, PEAK1 is amplified in PanINs and PDAC patient tissues in response to KRas activation [30]. Interestingly, PEAK1 possesses a polyproline motif suggesting that it could be translationally regulated by eIF5A [12-14]. This prompted us to determine if eIF5A is an upstream regulator of PEAK1 expression. Depletion eIF5A1 or eIF5A2 significantly reduced PEAK1 protein levels and Src activity in PANC1 and 779E cells (Fig. 6A, Fig. S4A, B). These results were confirmed by eIF5A1 or eIF5A2 knockdown using an independent set of shRNAs (Fig. S3A). In addition, simultaneous depletion of eIF5A1 by shRNA and eIF5A2 by siRNA, as well as treatment with GC7 or CPX, also robustly inhibited PEAK1 expression and Src activity compared to depletion of individual isoforms (Fig. 6B, C). On the other hand, overexpression of eIF5A1 or eIF5A2 significantly increased PEAK1 protein levels and Src activity (Fig. 6D and data not shown). Collectively, these results demonstrate that eIF5A is sufficient and necessary for PEAK1 protein expression in PDAC cells.

eIF5A-PEAK1 axis is essential to drive PDAC cell proliferation in vitro

To determine whether PEAK1 is the downstream effector of eIF5A in driving PDAC cell proliferation, we employed two independent approaches. First, we investigated whether PEAK1 knockdown can offset the enhanced PDAC cell proliferation induced by eIF5A overexpression. As shown in Fig. 6E, PEAK1 knockdown reduced the proliferation of eIF5A1- or eIF5A2-overexpressing cells by 50%. Secondly, we investigated whether overexpression of PEAK1 could rescue the anti-proliferative effect of eIF5A1 or eIF5A2 depletion. PANC1 cells expressing shRNAs were transiently transfected with plasmids encoding GFP or GFP-tagged PEAK1, and subjected to western blotting and cell proliferation analyses (Fig. 6F). PEAK1 was successfully overexpressed in PANC1 cells containing eIF5A1 or eIF5A2 shRNAs, which rescued the anti-proliferative effect of eIF5A1 or eIF5A2 depletion. Thus, exogenous introduction of PEAK1 was sufficient to overcome downregulation of PEAK1 in these cells to drive cell proliferation. PEAK1 overexpression did not affect cell proliferation in control PANC1 cells in a statistically significant manner, most likely because PEAK1 is expressed at high levels in these cells and forced expression of PEAK1 above the elevated endogenous levels may not be sufficient to promote PANC1 cell growth. However, it significantly enhanced cell proliferation in PANC1 cells with eIF5A1 or eIF5A2 knockdown (Fig. 6F). Altogether, these results indicate that PEAK1 is a necessary downstream effector of eIF5A proteins in enhancing PDAC cell proliferation.

Targeting eIF5A hypusination increases gemcitabine sensitivity in PDAC cells in vitro

In our previous work, we demonstrated that knockdown of PEAK1 renders PDAC cells sensitive to gemcitabine, the current first-line chemotherapy for PDAC treatment [30]. Having determined the role of eIF5A in regulating PEAK1 levels, we hypothesized that targeting eIF5A would increase sensitivity to gemcitabine in both drug-resistant (PANC1) and -sensitive (FG) PDAC cell lines [41]. As shown in Fig. 7A-D and Fig. S4C, inhibiting eIF5A hypusination with CPX dramatically enhanced gemcitabine sensitivity in both lines. In fact, IC₅₀ value of gemcitabine decreased from 165 nM to 1.74 nM and 12.4 nM to 21.3 pM upon co-treatment with CPX in PANC1 cells and FG cells, respectively. Moreover, we found that knockdown of eIF5A1, eIF5A2 or PEAK1 renders PANC1 or FG cells significantly more susceptible to gemcitabine (Fig. 7E and Fig. S4D-F). Interestingly, we found that gemcitabine treatment of resistant PANC1 cells robustly increased eIF5A1 hypusination and PEAK1 protein expression in a dose-dependent manner (Fig. 7F). This response was not observed in gemcitabine-sensitive FG cells (Fig. 7F) and it was specific to gemcitabine, as treatment with the DNA damaging and cytotoxic agent, cisplatin, did not significantly alter eIF5A1 hypusination or PEAK1 expression (Fig. 7G). These findings support a mechanism by which eIF5A-mediated PEAK1 expression contributes to gemcitabine sensitivity in PDAC cells, and suggests that combination therapies involving gemcitabine and CPX could benefit PDAC patients.