Targeted Intervention of eIF4A1 Promotes EMT and Metastasis of Pancreatic Cancer Cells through c-MYC/miR-9 Signaling


 Background: Due to the lack of effective interference options, early metastasis remains a major cause of pancreatic ductal adenocarcinoma (PDAC) recurrence and mortality. However, the molecular mechanism of early metastasis is largely unknown. We characterize the function of eukaryotic translation initiation factors (eIFs) in Pancreatic cancer cell epithelial mesenchymal-transition (EMT) and metastasis, to investigate whether it is effective to inhibit EMT and metastasis by joint interference of eIFs and downstream c-MYC. Methods: We used the data of The Cancer Genome Atlas (TCGA) and Genome Tissue Expression (GTEx) to analyze the expression level of eIF4A1 in PDAC tissues, and further validated in a microarray containing 53 PDAC samples. Expression regulation and pharmacological inhibition of eIF4A1/c-MYC was performed to determine their role in migration, invasion, and metastasis in pancreatic cancer cells in vitro and in vivo.Results: Elevated expression of eIF4A1 was positively correlated with lymph node infiltration, tumor size, and indicated a poor prognosis. eIF4A1 decreased E-cadherin expression through c-MYC/miR-9 axis. Ablation of eIF4A1 and c-MYC decreased the EMT and metastasis capabilities of pancreatic cancer cells. Upregulation of eIF4A1 could attenuate the inhibition of EMT and metastasis induced by c-MYC downregulation. Single-use of eIF4A1 inhibitor Rocaglamide (RocA) or c-MYC inhibitor Mycro3 and joint intervention all significantly the EMT level of pancreatic cancer cells in vitro. However, the efficiency and safety of RocA single-use were not inferior to joint use in vivo. Conclusion: The results demonstrated that overexpression of eIF4A1 downregulated E-cadherin through c-MYC/miR-9 axis, which promoted EMT and metastasis of pancreatic cancer cells. Despite the potential loop between eIF4A1 and c-MYC existing, RocA single strategy was a promising therapy for the inhibition of eIF4A1 induced PDAC metastasis.

Introduction Pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal solid malignancies universally.
Despite a relatively low incidence, it remains the fourth leading cause of cancer-related deaths in developed countries (1,2). There were no signi cant changes in the mortality to incidence ratio during the past decades. The ve-year survival rate remains about 3%-15% (3). At the early stage of carcinogenesis, pancreatic cancer cells could metastasize to distant organs through epithelial-mesenchymal transition (EMT) (4). An overwhelming majority of PDAC preliminary diagnosis patients have lost the chance of surgical eradication due to early metastasis, which is also the key reason for post-operation recurrence. KRAS was the most common oncogenic mutations associated with PDAC (5). So far, all the attempts targeted common PDAC KRAS variants (e.g., G12D, G12V, G12R) and multiple KRAS downstream kinases (e.g., RAF, MEK, ERK, PI3K) failed in / clinical trials (6-9). Thus, novel therapeutics other than KRASassociated kinases targeted inhibitors are urgently needed for PDAC patients.
Uncontrolled protein production is a symbolic characteristic of cancer cells, and it is also necessary for EMT and metastasis (10). Therefore, the intervention of the hyperactive protein production becomes a possible therapeutic for PDAC. Translation initiation regulated by eukaryotic translation initiation factors (eIFs) is the most important rate-limiting procedure in translation (11,12). Dysregulation of eIFs is a hallmark of various types of cancers including PDAC, among which eIF4F heterotrimeric complex is the main factor to facilitate mRNA translation. Meanwhile, the activity of eIF4F is largely regulated by RAS signaling, which further indicated eIF could play an important role in PDAC (13,14). The eIF4F complex is composed of the scaffold protein eIF4G, cap-binding protein eIF4E, and ATP-dependent DEAD-box RNA helicase eIF4A. Previous studies typically chose eIF4E as the target to inhibit the EMT and metastasis of cancer cells, because eIF4E was generally overexpressed in multiple cancers. However, recent studies demonstrated that there existed eIF-independent yet eIF4A-dependent binding sites of eIF-downstream oncogenic mRNA including c-MYC, which was the possible reason why the trials targeted eIF4F did not work (15,16). eIF4A is the only regulatory enzyme-catalytic factor in eIFs, which facilitates unbinding the complex long-sequence helix (CLSH) in mRNA 5'-untranslated region (5'-UTR). And the CLSH is a typical signature of various eIF-downstream mRNA including c-MYC (17). The overexpression of c-MYC is a carcinogenesis driver for multiple cancers and c-MYC was the most activated Oncogenes (18,19). c-MYC was a crucial regulator for EMT and metastasis through promoting miR-9 expression (20,21). However, due to the structure of c-MYC, there were few c-MYC targeted inhibitors by now (22,23). Considering eIF4A being dispensable for c-MYC translation, we hypothesized that intervention of eIF4A could be an effective way for c-MYC inhibition.
Traditional Chinese medicinal herbs gained our attention as a novel source of anticancer remedies during the past few years. Rocagalamide (RocA) was a cyclopenta-b-benzofuran-type compound derived from traditional Chinese remedies genus Aglaia (family Meliaceae). Iwasake et al. carried out a survey in 2016 that RocA could attach eIF4A rmly at the 5'-UTR sequence and thus suppress the hyperactive protein production than the intervention of eIF4E (24). Subsequently, RocA was used in various hematologic malignancies (e.g., myeloma, T-cell lymphoma) and exerted a prominent anti-tumor e cacy without observed side effects in mouse model (25). The trials of RocA in solid malignancies were rare, and one of those was our previous study which demonstrated that RocA obviously repressed the EMT and metastasis of pancreatic cancer cells in the mouse model (26). Nevertheless, how eIF4A was involved in the EMT and metastasis of PDAC was still largely unknown. And taken the complex loop relation between eIF4A and c-MYC, whether eIF4A and c-MYC dual inhibition were superior to the single-therapeutic intervention deserved to be investigated. There are 3 identi ed isoforms eIF4A in human. eIF4A1 is the major type to participate in the assembly of eIF4F in cancer cells and eIF4A2, which is mainly expressed in low-level proliferating cells, is correlated with a good prognosis in multiple cancers (12,17). Therefore, we selected eIF4A1 as the target in follow-up studies.
Here, we reported that overexpression of eIF4A1 downregulated the expression of E-cadherin through c-MYC/miR-9 signaling axis, which promoted EMT and metastasis.

Immunohistochemistry (IHC)
IHC staining with the eIF4A1 antibody (Abcam ab31217) was performed to detect the protein expression level. Assessments of microarray IHC staining were performed by ImageJ (http://imagej.nih.gov/ij) IHC pro ler (http://sourceforge.net/projects/ ihcpro ler). The intensity of staining was scored on a scare as negative (0 points), weak-positive (1 point), positive (2 points), and strong-positive (4 points). The protein expression level was calculated by multiplying the staining intensity and corresponding positive staining extent n (n%: percentage of positive areas to the whole areas). Then we divided the patients into two groups (grade < 50, low expression; grade > 50, high expression) and performed subsequent survival analysis. The results of IHC staining were reassessed by two independent pathologists who were blinded to this study.

Immuno uorescence
Paraformaldehyde xed samples were washed 3 times with cold PBS for 3min each time and then incubated in 10% donkey serum in PBS for 20min. Subsequently, then the samples were incubated with the eIF4A1 antibody (Abcam ab31217) in PBS at 4°C overnight. After washing, uorochrome-conjugated secondary antibodies (1:400, Alexa Fluor®488 donkey anti-rabbit IgG) were used, and then the samples were incubated with DAPI. Fluorescence was visualized under an Olympus microscope.

Transient transfection
The cells were transinfected with nonsense siRNA or siRNA targeting eIF4A1 (siRNA1: Small animal imaging (Spectral Imaging LagoX) was performed every 3 days after cell injection and the mice were randomly divided into different groups for the following studies (n = 5/ group). The mice were sacri ced when the weight loss was greater than 20%. RocA (MCE, USA) was administrated by intraperitoneal injection (5mg/kg/d, 3mg RocA dissolved in 30µl DMSO, 600µl PEG300 and 75µl Tween-80 successively, then made the solution up to 1.5ml with normal saline) every day originally, then adjusted to 2.5mg/kg/d once on an alternate day. Mycro3 was administrated by intragastric (100mg/kg/d, 25mg mycro3 dissolved in 2.5ml 0.5% methylcellulose solution a liated by ultrasound). The control group mice were treated by intragastric 200µl methylcellulose solution every day and 100 methylcellulose solution intraperitoneal cosolvent injection once on an alternate day.

Subcutaneous xenografts in nude mice
Four-week nude mice were obtained from HFK Bioscience Ltd (Beijing, China) and maintained in SPF conditions. AsPC-1 cells (5.0 × 106) suspended in a 100µl mixture of equal volumes of medium and matrix-gel were implanted subcutaneously into the right ank of 6-week-old female nude mice. When the tumors had reached a volume of about 60-90 mm3, the mice were then randomly divided into two groups. The treatment group received an intraperitoneal injection of RocA (previously adjusted dose), whereas the control group received cosolvent injection alone (n = 4). These treatments were carried out once daily for 28 days. Tumor volumes and the body weight of animals were measured twice a week. Tumor volumes (mm3) were calculated with the following formula: V = LS2/2 (where L is the longest diameter and S is the shortest diameter). At the end of experiment, the mice were sacri ced and the tumors were harvested for western blot.

Statistical analysis
Data are representative of at least three independent experiments or multiple independent mice as indicated. The characteristics of patients were summarized as mean ± standard deviation (SD) for normally distributed continuous variables, median with interquartile range for continuous variables with skewed distribution, and frequency (percentage) for categorical variables. All the analyses were performed using R (http://www.R-project. org, version 3.5.2) with a two-sided signi cance threshold of P < 0.05. Results eIF4A1 was highly expressed in pancreatic ductal adenocarcinoma and predicted a poor prognosis Firstly, we analyzed the expression of eIF4A1 in pan-cancer using data from Gene Expression Pro ling Interactive Analysis (GEPIA, http://gepia.cancer-pku.cn/). eIF4A1 was highly expressed in multiple cancer types, including pancreatic adenocarcinoma, thymoma, glioblastoma multiforme, diffuse large B cell lymphoma, and testicular germ cell tumors (Fig. 1a). To evaluate the whole eIF family expression in pancreatic adenocarcinoma, we analyze the RNA-seq data from TCGA and GTEx. Two samples were deleted after data quality control, and nally, 179 cases of pancreatic tumor tissues and 169 cases of normal pancreatic tissues were obtained. We ranked differential expression of all the eIF (Fig. 1b) and the result showed that the expression level of eIF4A1 in pancreatic tumor tissues was signi cantly higher than in normal tissues (Fig. 1c).
Subsequently, we analyze the prognostic role of eIF4A1 in PDAC patients. eIF4A1 was mainly localized in the cytosol (Fig. 1d), which was consistent with the function of assisting translation initiation. We detected the eIF4A1 expression in a tissue microarray using immunohistochemistry from 53 PDAC patients con rmed by surgeries and pathologists from 2009 (Fig. 1e). The results indicated that patients with high eIF4A1 expression suggest a poor prognosis, the median OS of high eIF4A1 expression patients was signi cantly shortened compared with patients with low eIF4A1 (6.0 months VS 9.0 months, HR = 2.10, 95% CI: 1.44-5.24, P = 0.0061) (Fig. 1f). Furtherly, we examined the correlations between eIF4A1 expression and multiple clinical features (Table 1, * P < 0.05; χ 2 test or Fisher's exact test). Importantly, we found that a high level of eIF4A1 expression was signi cantly correlated with tumor size and lymph node metastasis. In summary, these results showed that eIF4A1 was highly expressed in pancreatic adenocarcinoma tissues, and high expression of eIF4A1 suggested a poor prognosis. The expression of eIF4A1 was positively correlated with lymph node metastasis which is a major way for cancer cell metastasis. To elucidate the mechanism underlying eIF4A1 regulating the biological behavior of tumor cells, we searched the GEO database for all the translational pro ling by ribosome pro ling that included pancreatic cancer cells, and ultimately, datasets GSE120159 was selected for further analysis. The data included 3 Panc-1 cell samples treated by rocaglate CR31B, a small-molecule inhibitor of the eIF4A helicase, and 3 Panc-1 cell samples treated by DMSO. A differential expression analysis using R package DEseq2 identi ed 179 differentially expressed proteins (DEP) between the CR31B-treated group and the DMSO-treated group, with a p. adjusted (FDR) value < 0.05 and |log 2 FC (fold change) |> 1 as the cut-offs.
Among these proteins, 128 were downregulated and 51 were upregulated. The heatmap displayed the top overexpressed and suppressed molecules (Fig. 2a). The expression of c-MYC was signi cantly downregulated (log 2 FC = -1.05, FDR = 0.00192) in the CR31B-treated group compared with the control group, ranked the top 0.6% in all the regulated gene lists.
To furtherly screen the key target of eIF4A to promote metastasis, we repressed the eIF4A1 expression (eIF4A1 siRNA) and analyzed the alternation of protein pro le expression compared to the control group. Notably, the protein abundance at 49kDa, which is the molecular weight of c-MYC remarkably decreased after the deletion of eIF4A1 (additional le 1: Figure S1).
Gene Ontology (GO) analysis revealed that the DEPs enriched in the GO terms for 497 biological processes (BPs), 56 cellular components (CCs), 48 molecular functions (MFs) with statistical signi cance ( Fig. 2b). Through the GO analysis, we can conclude that the DEPs enriched in metastasis relevant functions including signal transduction, cytoskeleton, lymph-angiogenesis, cell junction. Gene set enrichment analysis (GSEA) performed with ribosomal pro ling showed that the enrichment of EMTrelated gene sets reduced signi cantly after CR31B treatment (Fig. 2c). These results indicated that eIF4A1 could target c-MYC to regulate the biological behaviors of pancreatic cancer cells.

eIF4A1 promoted EMT and metastasis through c-MYC/miR-9 signaling
Recent studies have demonstrated that c-MYC promoted EMT through upregulating the expression of miR-9, and miR-9 could competitively bind with E-cadherin encoding sequence which led to EMT occurrence (20,28). Before exploring the role of eIF4A1 and c-MYC in the regulation effect in pancreatic cells, we analyzed the expression of eIF4A1 in pancreatic cell lines Panc-1, Capan-2, AsPC-1, MiaPaca-2, and normal pancreatic ductal epithelial cell line HPDE. Western blot showed that eIF4A1 was notably higher expressed in the aggressive pancreatic cancer cell line AsPC-1 and relatively lower expressed in the indolent Capan-2 cell line and normal HPDE cell line (Fig. 3a). Therefore, we selected AsPC-1 and Capan-1 in the follow-up studies. We repressed the eIF4A1 or c-MYC expression in aggressive AsPC-1 cells (eIF4A1 siRNA, c-MYC siRNA) and overexpressed the eIF4A1 or c-MYC expression in indolent Capan-2 cells (pcDNA3.1-eIF4A1, pcDNA3.2-myc). Western blot and qRT-PCR showed that down-regulation of eIF4A1 in aggressive AsPC-1 decreased the expression of EMT-related gene (c-MYC, snail, and miR-9) and increased the expression of E-cadherin (Fig. 3b). Downregulation of c-MYC resulted in the same expression alternation (Figure, 3c), however, repressing the c-MYC expression did not in uence the expression of eIF4A1. Accordingly, the results of eIF4A1 and c-MYC upregulation in Capan-2 were consistent with the trends of downregulation experiments (Fig. 3d, 3e). Transwell migration and invasion assays showed that eIF4A1/c-MYC-downregulated AsPC-1 cells displayed signi cantly lower migratory and invasive abilities (Fig. 3f), and the abilities of eIF4A1/c-MYC upregulated Capan-2 cells increased signi cantly than control (Fig. 3g). Based on the changes of EMT-related molecule expression level and the changes of migratory and invasive capabilities, these results indicated that eIF4A1 could promote EMT through targeting c-MYC/miR-9 signaling.

Overexpression of eIF4A1 increased the MYC-downregulated AsPC-1 cells' invasive, migratory, and metastatic capabilities in vitro and in vivo
The relation between eIF4A1 and c-MYC is not simply direct upstream and downstream, recent studies revealed that overexpressed c-MYC could increase the expression level of eIF4A1 reversely (15,29). To further explore the regulatory relations between eIF4A1 and c-MYC, lentivirus was used to regulate the eIFA1 and c-MYC expression. The setting groups were as follow : 1)  cells was between e-U group and M-D group, however, still higher than the level of vector group cells (Fig. 4a). Consistent with the trends of western blot results, Transwell migration and invasion assays showed that the migratory and invasive abilities of e-U-M-D sequential regulation group cells were superior to the M-D group cells, but inferior to the vector group and e-U group (Fig. 4b). In vivo, the metastatic potentials of the above group cells were examined using a mouse metastasis model via caudal vein injection. The results (Fig. 4c) also showed that the luminescence intensity of e-U-M-D group cells was signi cantly higher than M-D group cells (1.008e + 10 vs 5.387e + 9, P = 0.0349), accordingly weaker to the e-U group cells (1.008e + 10 vs 2.410e + 10, P = 0.2369). Collectively, both in vitro and in vitro results showed that the EMT level and metastatic capabilities of e-U-M-D sequential group cells were between e-U group and M-D group. These results indicated that overexpression of eIF4A1 expression could attenuate the inhibition of MYC-downregulated pancreatic cancer cells' capabilities of EMT and metastasis.

RocA alone was not inferior to RocA plus Mycro3 joint intervention to inhibit EMT and metastasis in vitro and in vivo
Our previous studies demonstrated that c-MYC was not the only target of eIF4A1 to promote EMT and metastasis. When c-MYC was repressed, eIF4A1 might expert pro-EMT effect through other pathways to compensate for the inhibition. Considering the complex loop relation between eIF4A1 and c-MYC, we adopted joint intervention of RocA (eIF4A1 inhibitor) plus Mycro3 (c-MYC inhibitor) to explore whether joint intervention is superior to the two inhibitors intervention alone.
To select the optimal drug concentration, we conducted series of drug concentration gradient experiments. The western blot results showed that the expression level of eIF4A1 decreased signi cantly at 100nM RocA and the expression level of c-MYC decreased signi cantly at 5000nM Mycro3 (Fig. 5a). There was crystal precipitation when concentration increasing. Thus, we selected 100nM RocA and 5000nM Mycro3 in the follow-up studies. To compare the safety and e ciency of different intervention methods, we set 4 groups: 1) RocA + Mycro3 group; 2) RocA group; 3) Myro3 group; 4) DMSO control group. Western blots and qRT-PCR showed that all 3 intervention methods signi cantly increased the Ecadherin expression level and decreased the c-MYC expression level of AsPC-1 cells (Fig. 5b). Compared with the control group, all 3 intervention methods also remarkably decreased the migratory and invasive abilities. All 3 intervention methods could decrease the EMT level of pancreatic cancer cells (Fig. 5c). However, there was no statistical signi cance among the 3 groups.
To examine the e ciency and safety of different intervention methods in vivo, we used a mouse metastasis model via caudal vein injection. We found that the mouse mortality rates of RocA group (40.0%) and joint intervention group (62.5%) were relatively high. The autopsy showed that there were multiple cases present with nonocclusive intestine dilation. Considering the possible high-dose and frequent RocA use induced intolerance, we decreased the dose intensity from 5mg/kg/d qd to 2.5mg/kg/d once on alternate day by intraperitoneal injection, and the dosage of mycro3 remains intragastric 100mg/kg/d qd. The mice were well-tolerant to the modi ed regimen without more death cases. The luminescence intensity from RocA single-dose group was signi cantly weaker than the control group (1.393e + 9 vs 2.707e + 9, P = 0.0474). However, there were no statistical differences in luminescence intensity between the control group and the myro3 single-dose group/ joint intervention group (Fig. 5d). Taken together, these data demonstrated that the modi ed regimen of RocA single-dose presented an obvious anti-metastasis effect, even superior to mycro3 or joint-intervention.
To further testify the e cacy of RocA, we used a subcutaneous xenograft nude mice model. The 2-way ANOVA analysis indicated that the tumor volumes of RocA group were signi cantly smaller than those of the control group (P < 0.0001) ( Figure. 5e) which indicated that RocA notably suppressed the growth of tumor. And western blot showed that RocA markedly decreased the expression of eIF4a1, c-MYC, and snail whereas increased the expression of E-cadherin ( Fig. 5f) in vivo.

Discussion
Various PDAC involved oncogenes including KRAS and downstream kinases cannot be targeted because of their speci c molecular structures (23,30). However, targeting the translational procedure has been an alternative option for cancer treatment. The protein production of PDAC was veri ed to be hyperactive both in vivo and in organoids (31). And eIF4A which participates in the assembly of eIF4F, is the nexus for translational regulation (13,32). Unlike KRAS, eIF4A could be targeted and the curative effects have been validated in several hematological malignancies (25,33). Thus, targeting eIF4A could be an alternative treatment for PDAC.
Our previous study demonstrated targeting eIF4A could signi cantly decrease the lung metastasis of pancreatic cancer cells in vivo (26). Here we further represent the mechanism of eIF4A promoting EMT and metastasis and veri ed the overexpression of eIF4A1 was correlated with poorer prognosis. Considering the complex long-sequence 5'-UTR structure of c-MYC mRNA, we demonstrated that the expression regulation of c-MYC was highly dependent on eIF4A. Depletion of eIF4A1 signi cantly decreased the expression level of c-MYC and pro-EMT molecules in vitro. Overexpression of eIF4A1 induced E-cadherin expression decreasing thorough c-MYC/miR-9 axis. Furthermore, depletion of eIF4A1 or c-MYC signi cantly decreased the EMT level and metastasis in vitro and in vivo, and vice versa. However, the regulation of translational regulation was extremely complicated. c-MYC cannot be the only target for eIF4A1 to promote EMT and metastasis and it was reported that c-MYC overexpression could reversely upregulate c-MYC (29,34). In this study, we found that overexpression of eIF4A1 rescued the impaired EMT and metastasis capabilities of c-MYC repressed pancreatic cancer cells in vitro and in vivo. Similar to KRAS, c-MYC was traditionally deemed as an "undruggable" target. Mycro3 was a novel anti-MYC compound to inhibit c-MYC activity through MYC-dimerization (23,35). As previously mentioned, the relation between eIF4A and c-MYC was not simply unidirectional, thus, adopted eIF4A and c-MYC joint intervention to testify the e cacy and safety. Our works demonstrated single-use of RocA or Mycro3, and joint remedies all notably repressed the EMT level in vitro. However, the joint intervention did not show any superiority compared with single-use. Similar to the in vitro result, RocA was the only reagent that showed a signi cant anti-metastasis effect in vivo among two single-use remedies and joint intervention.
The reason leading to this result is still unknown, but we can speculate from the preparatory experiments that joint intervention of RocA plus mycro3 increased the risk of death. The mice may not be tolerant of the original dose. But regardless, Roc single-use dramatically suppressed the eIF4A-mediated pancreatic cancer cell metastasis.

Conclusions
Collectively, we demonstrate that eIF4A1 overexpression downregulated E-cadherin expression through c-MYC/miR-9 axis, which promoted EMT and metastasis of pancreatic cancer cells in vitro and in vivo. The upregulation of eIF4A1 could partially rescue the c-MYC-depletion mediated inhibition of EMT and metastasis. Moreover, RocA single-use was superior to the joint remedies the at current dose possibly due to the intolerance. Our works indicated that eIF4A1 was a satisfying biomarker for PDAC prognosis, and intervention of eIF4A1 provide a promising therapeutic strategy for PDAC metastasis.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.