Everolimus-Induced Loss of Exosomal miR-7-5p Activates MNK/eIF4E Axis in Non-Small Cell Lung Cancer

Background: Everolimus is a kind of mTOR inhibitors. Activated mitogen-activated protein kinase interacting kinases/eukaryotic translation initiation factor 4E (MNK/eIF4E) axis plays a crucial role in resistance to Everolimus in non-small cell lung cancer cells (NSCLC). Typically, eIF4E phosphorylation increased by mTOR inhibitors was mainly mediated by MNKs. But the mechanisms are poorly understood. Recently, extensive reprogramming of miRNA proles has also been found after long-term mTOR inhibitor exposing. Our previous studies have found that tumor suppressor miR-7-5p was loss in A549 cells after treatment of Everolimus. Exactly, MNK1 is the target of miR-7-5p. Here, we analyzed levels and functions of miR-7-5p in the usage of Everolimus in NSCLC. Methods: miR-7-5p level and expression of main markers of MNK/eIF4E axis was evaluated by qRT-PCR, in situ hybridization, and immunohistochemistry on human NPC samples, and by RT-PCR, western blot on NPC cell lines. Proliferation, migration and invasion of NPC cells in culture were assessed by Colony formation, CCK-8, Wound healing and Transwell assays. NPC cell tumorigenicity was assessed by xenotransplants in nude mice. Targeted binding of miR-7-5p to MNK1 was conrmed by the Dual-luciferase reporter assay. And the isolation and identication of exosomes were invested by Invitrogen™ Total Exosome RNA Isolation Kit, western blot, transmission electron microscopy and Zetasizer Nano ZS90 instrument. Results: Everolimus stimulatedtherelease of miR-7-5p loaded exosomes from NSCLC cellsinRab27A and Rab27B dependent manners, thereby reducing the intracellular tumor suppressor miR-7-5p to attenuate the inhibition of MNK1 and promote MNK-dependent eIF4E phosphorylation. Of note, both lower miR-7-5p and positive MNK1 could act as independent poor prognostic biomarkers for NSCLC. TSG101, CD63 and CD9 expression were measured using western blot analysis, and GAPDH as an internal reference. Isolated in for washed with deionized Ten µL of was Images were obtained by TEM 46


Background
The mammalian target of rapamycin (mTOR) is a serine/threonine kinase that can regulate cell growth and proliferation under physiological and pathological conditions. Dysregulation of the mTOR signaling pathway can be seen in many cancers including non-small cell lung cancer (NSCLC). Its functions mainly exercise by forming two different complexes, named mTOR complex 1 (mTORC1) and mTOR complex 2 (mTORC2) 1 . Everolimus (RAD001) is an effective mTORC1 inhibitor to be a cutting-edge drug for targeted therapy, but has poor e cacy as a monotherapy 2 . Similar to other anti-tumor treatments, long-term Everolimus treatment can lead to secondary drug resistance caused by genomic instability 3 . Activation of other regulatory proteins or survival cascades are also involved in resistance to Everolimus, like mitogen-activated protein kinase (MAPK)-interacting kinases (MNK) 1 and 2.
The activated MNKs are phosphorylated, which continuously phosphorylate eIF4E, and play a crucial role in mediating resistance to rapamycin in NSCLC 4 . Moreover, eIF4E phosphorylation increased by mTOR inhibitors was mainly mediated by MNKs 5 . However, the mechanisms by which Everolimus induced phosphorylation of MNK/eIF4E axis are poorly understood.
Exosomes derived from tumor cells carry the components of membranes and cells fall off into the extracellular space and cause the loss of intracellular proteins, miRNAs and lncRNAs via the secretion. It re ects the phenotype of the original cells, the effect of drug treatments on cells and the process of speci c sorting of these molecules by tumor cells. Therefore, it is of great value for the early diagnosis of disease, the assessment of pathogen burden or disease stage, and the monitoring of the response to treatment 6 . This secretion-dependent mechanism, is regulated by mTORC1 7 . So, the occurrence of Everolimus resistance mechanism might relate to the release of exosomes.
Totary-Jain et al. demonstrated that long-term treatment of mTOR inhibitor rapamycin showed extensive reprogramming of miRNA pro les, characterized by down-regulation of tumor suppressor miRNAs. And delivery of tumor suppressor miRNAs would restore the sensitivity to rapamycin 8 . Our previous studies have found that when Everolimusposed onNSCLC cells, its tumor suppressor effect stabilized at a concentration of 25nM, accompanied by activation of the MNK/eIF4E axis 9 . Besides,we carried miRNA array out to detect changes of the miRNA pro les in A549 cells acceptingEverolimus treatment 10 . The down-regulated miR-7-5p has attracted our attention. For the reasons that numerous reports con rmed miR-7 was a tumor suppressor in NSCLC 11 , as well as miR-7 regulatedmany genes of the mTOR pathway, like MNK, eIF4E and 70 kDa ribosomal protein S6 kinase (p70S6K)in adult pancreatic βcells 12 .
Above all suggested that Everolimustargeting mTORC1might alter the level of miR-7-5p in NSCLC in an exosomedependent manner, thereby activating MNK /eIF4E axis, which might be involved in the resistance of Everolimus, but have not been studied yet. Therefore, in our study, we focused on the new mechanism of miR-7-5p in Everolimus resistanceand looked for new treatment options in NSCLC. Here, we discovered that when the activity of mTOR was inhibited by Everolimus in NSCLC, it would promote the secretion of exosomes derived from tumor cells, which loaded miR-7-5p. Consequently, MNK/eIF4E axis was activated followed byintracellular miR-7-5p declining,therebylimiting the tumor suppressor effect of Everolimus. Hopefully, endogenous miR-7-5p or exo-miR-7-5p could synergistically enhance the e cacy of Everolimus by inducing apoptosis.

Materials And Methods
Clinical data and tissue microarrays (TMA) In this study, a total of 318 cases of NSCLC and 91 cases of non-cancerous lung tissue (Non-CLT) were chosen randomly from the Second Xiangya Hospital of Central South University (Changsha, China). Complete clinical records and follow-up data were available for all patients. Besides, 40 pairs of primary NSCLC tissues and adjacent non-tumor lung tissues were collected from 40 cases of NSCLC. Protocols were approved by the Second Xiangya Hospital of Central South University Ethics Review Board (Scienti c and Research Ethics Committee, No. S039/2011). All patients involved in our study had written informed consent. The staging classi cation of this research was carried out based on the criteria of the 8th edition of the AJCC/UICC TNM staging system of lung cancer (2017). The con rmed histological diagnosis of NSCLC according to the World Health Organization histological classi cation of lung cancer.
We used the TMA technology designed and constructed high-throughput NSCLC TMAs according to rules previously described 47 .

Quantitative PCR analysis
We extracted total RNA using TRIzol reagent (Invitrogen). Two µg of total RNA was performed to synthesize cDNA using by Mir-X miRNA First-Strand Synthesis Kit (Takara) or RevertAid First-Strand cDNA Synthesis Kit (Thermo) according to the manufacturer's protocol. The RNA level was measured by qPCR with a TB Green premix Ex Taq II kit (Takara) or 2 × SYBR Green qPCR Master Mix (Bimake) on CFX96 Real-Time PCR Detection System (Bio-Rad).
Relative expression was determined with a U6 control or β-actin through the 2 −ΔΔ Ct method.

Western blotting
Whole-cell protein lysates preparation and Western blotting analysis were performed as described previously 45 .
Cytoplasmic and nuclear extracts were obtained by using the CEB-NEB kit (Applygen, China) according to the manufacturer's protocol.

Exosome isolation and identi cation
Before cell culture, FBS was depleted of exosomes by ultracentrifugation at 1 × 10 5 g at 4 ℃ for 12 h, then the supernatant was harvested and ltered with a 0.22 um lter (Millipore). After 24 h, the cell culture medium with exosome-free FBS was collected and removed oating cells and cellular debris from the supernatant by centrifuging at 3000 g for 30 min. The supernatant was then passed through a 0.22 µm lter. The resultant supernatant was forced through the membrane by centrifugal ltration at 3000 g for 1h with the ultra ltration device (UFC900396, Millipore). Finally, the total exosome isolation reagent (from cell culture media, Invitrogen) was added to the concentrated solution at a ratio of 1:2. Subsequent exosome extraction was conducted according to the manufacturer's protocol. The exosomes were used for the following experiments immediately or stored at − 80 ℃. Exosomal protein was measured by the BCA™ Protein Assay Kit (Thermo). The concentration of exosomal proteins was quanti ed using a BCA protein assay kit (Keygen Biotech, KGP902, China). HSP70, TSG101, CD63 and CD9 expression were measured using western blot analysis, and GAPDH as an internal reference.
Isolated exosomes were xed in 1% glutaraldehyde for 10 min, washed with deionized water. Ten µL of exosome suspension was placed on formvar carbon-coated 300-mesh copper electron microscopy grids, incubated for 5 min at room temperature and air-dried for 5 min. Images were obtained by TEM 46 .
In addition, exosomes were examined via injecting into the Zetasizer Nano ZS90 instrument. Phosphate buffered saline was used as controls. Data was analyzed by Zetasizer software.
The synthetic miRNA mimics, inhibitors, negative control (NC), siTSC1, siTSC2, siRab27A or siRab27B (GenePharma, Shanghai, China) were mixed with lipofectamine 3000 (Invitrogen) and then added to the cell culture medium according to the manufacturer's instructions. After 24 h or 48 h of transfection, total RNA and protein were extracted from cells, respectively.

Lentivirus transduction
The LV-miR-7-5p and the control virus LV-NC were purchased from GenePharma (Shanghai, China). Cells were infected with lentivirus for 2 days and then added to puromycin (2.5 mg/mL) for 2 weeks to generate stably transduced cell lines.

Dual-luciferase reporter assay
The full length of MNK1 3'-UTR were ampli ed and subcloned into the pmirGLO dual-luciferase miRNA target expression vector. The wild type and mutant sites in MNK1 3'-UTR or eIF4E 3'-UTR were produced by GeneCreate Biotech (Wuhan, China). Luciferase activity assays were performed with the Dual-Luciferase Reporter Assay System (Promega, Madison, WI).

Xenograft Mouse Tumor Models
For the subcutaneous tumor model, the LV-miR-7-5p and LV-NC A549 cells (1 × 10 7 cells) in 200 µL serum-free medium were injected subcutaneously into female BALB/c nude mice, respectively. As soon as the tumor volume reached 50-100 mm 3 , mice would be divided into four groups (n = 6/group) named as the LV-NC group, the LV-miR-7-5p group, the LV-NC combined Everolimus (3 mg/kg/day, oral gavage) group and the LV-miR-7-5p combined Everolimus group. The four groups received indicated treatments. The size of the tumors and the weight of the mice were monitored every 2 days. After 28 days, the mice were euthanized and the tumors were obtained. Tumor volume was calculated with the formula: V = (length × width 2 ) /2. Finally, all tumors were excised, weighed, and xed with formalin.
As for the intraperitoneal tumor model, single-cell suspensions of luciferase tagged A549-cells (1 × 10 7 cells) were injected into the abdominal of female BALB/c nude mice. And the baseline level of tumor cells was monitored by live imaging technology immediately after injection. And then, the survival and mental state of the nude mice were monitored every two days. After one week, we con rmed the survival of tumor tissue, randomly divided them into 6 groups (n = 3/group), and received PBS, exo-NC, exo-miR-7-5p, Everolimus, exo-NC combined with Everolimus and exo-miR-7-5p combined with everolimus treatment, respectively. The development of tumor burden was monitored by bioluminescence imaging (BLI) of anesthetized mice that were intraperitoneally injected with 75 mg/kg D-luciferin (Thermo).
Colony formation assay, CCK-8 assay, Wound healing and Transwell migration assays These procedures were performed as previously described 10 .

Flow cytometry
For cell apoptosis detection, cells were plated into 6-well plates and transfected with miRNA mimics for 24 h and treated with/without 25 nM Everolimus after an additional 48 h. The number of apoptotic cells were detected using a FITC-Annexin V apoptosis detection kit (BD Pharmingen) and a ow cytometer following the manufacturer's instructions.

Statistical analysis
Statistical analyses were determined by chi-square test, Spearman correlation test, multivariate Cox regression analysis, Pearson correlation, Wilcoxon rank-sum test, or student's t-test as appropriate using SPSS 19.0 and GraphPad Prism 5.0. Error bars indicated the standard deviation in all the gures. *P < 0.05, **P < 0.01, ***P < 0.001. And P value < 0.05 was considered signi cant.

Results
Everolimus targets mTORC1 inducing NSCLC cells to secrete miR-7-5p-loaded exosomes in Rab27A and Rab27B dependent manners In multiple NSCLC cell lines, including A549, H358, H520 and SPC-A1, the level of miR-7-5p was signi cantly downregulated after Everolimus exposing. And it lasted more than 48 hours without changing of autophagy in NSCLC cell lines ( Fig. 1a and S1a). In order to nd the reason for the reduction of miR-7-5p, we further analyzed the lung squamous cell carcinoma mRNA data of TCGA, and divided the data into two groups named mTOR low group and mTOR high group according to the mean value of mTOR level in the samples. The GSEA on the mRNA pro les changes revealed negative associations between mTOR and gene sets involved in extracellular exosomes (Fig. 1b).
Consequently, we focused on the exosomes derived from NSCLC cells in the culture medium. They were isolated and identi ed by transmission electron microscope (TEM) (Fig. 1c) and NanoSight Analysis (Fig. 1d). The size and morphology of exosomes were similar between two groups, presented as 50-100nm round like. It was also found the uorescent labelled exosomal marker CD63 in the cytoplasm was signi cantly weakened after Everolimus treating ( Fig. 1e). Meanwhile, expression of exosomal markers HSP70, TSG101, CD63 and CD9 signi cantly enriched in exosomes accompanied by down-regulation of intracellular exosomal markers in A549 cells by Western blotting (Fig. 1f). To further determine whether there were association between the declination of intracellular miR-7-5p and elevation of release of exosomes induced by Everolimus, we used exosomes inhibitor GW4869 to block the release of exosomes. When the secretion of exosomes was inhibited by GW4869, miR-7-5p distribution was reversed (Fig. 1g), which indicated that Everolimus induced the secretion of miR-7-5p-loaded exosomes into the surrounding extracellular environment, resulting in the loss of miR-7-5p in intracellular.
Furthermore, we performed siTSC1/2 to active 13 but Everolimus to inhibit the activity of mTORC1. After con rming the expression of mTOR, TSC1 and TSC2 following indicated treatments ( Figure S1b), we discovered miR-7-5p was more enriched in exosomes than in intracellular when Everolimus was employed. On the contrary, miR-7-5p located in intracellular rather than in exosomes when siTSC1/2 was performed. When Everolimus was employed in siTSC1/2-A549 cells, miR-7-5p would be enriched in exosomes again (Fig. 1h). It con rmed that the activity of mTORC1 controlled the redistribution of miR-7-5p.
Rab27A and Rab27B are two closely related Rab small GTPases, which play key roles in the secretion of exosomes in many types of cells 14 . The knockdown of Rab27A and Rab27B by siRNA would reduce exosome secretion of HeLa cells. And Rapamycin targeting mTORC1 stimulated the release of exosome dependent on Rab27A 7 . Here, we found that the release of miR-7-5p-loaded exosomes was signi cantly reduced to almost undetectable when Rab27A or Rab27B was knocked down (Figure S1c), accompanied by the enrichment of intracellular miR-7-5p. When Everolimus were performed after Rab27A or Rab27B knocking down, the exocytosis of miR-7-5p loaded exosomes was also suppressed (Fig. 1i). This revealed that the release of miR-7-5p-loaded exosomes induced by Everolimus depended on Rab27A and Rab27B.
Loss of intracellular miR-7-5p induces phosphorylation of MNK/eIF4E axis, but supplement of extra exosomal miR-7-5p can reverse it With the gradual deepening of understanding, it had been found that there were feedbacks of multiple kinases during the employment of Everolimus, such as hyperphosphorylation of eIF4E at the Ser209 site, which led to the development of drug resistance 15 . The mTOR inhibitor Rapalog induced eIF4E phosphorylation was done by kinases termed MAP kinase-interacting kinases (Mnks) independently of the conventional Mnk-activating MAPK pathway 16 . Coincidentally, MNK1 was a potential targeted gene of miR-7-5p by bioinformatics analysis (Pictar, TargetScan, and Tarbase) (Fig. 2a). It could directly bind to the 3'UTR of mRNA of MNK1 by the luciferase reporter gene test (Fig. 2b). When miR-7-5p was elevated by mimics, the MNK1 would be down-regulated. On the contrary, inhibition of miR-7-5p by inhibitor could up-regulate the MNK1 as Figure S1d & 2c. The two NSCLC cell lines A549 and SPC-A1 were utilized to conduct the following experiments due to that miR-7-5p level was lower but MNK1 expression was higher compared with the human bronchial epithelial (HBE) cell and other NSCLC cell lines ( Figure S1e). Therefore, a lentivirus-based system (LV-miR-7-5p and LV-NC) was used to conduct stable miR-7-5p overexpressed A549 and SPC-A1 cell lines.
When Everolimus posed on stable miR-7-5p A549 and SPC-A1 cells, overexpressed miR-7-5p was able to inhibit Everolimus induced upregulation of MNK1 and p-eIF4E S209 without affection of downstream of mTOR, including p-S6 S235/236 and p-4EBP1 Thr37/46 (Fig. 2d). Simultaneously, Considering higher miR-7-5p can inhibit protein ubiquitinal degradation 17 , and MNK1 protein can be degraded by ubiquitination in addition to be regulated by phosphorylation 18 , the stable miR-7-5p and control A549 cells were treated with protein synthesis inhibitor cycloheximide (CHX) to calculate the half-life of MNK1 protein. As show in Fig. 2e and 2f, the MNK1 protein decreased by ~ 50% within 7.785 hours in the control A549 cells, and 7.278 hours in the stable miR-7-5p A549 cells. The half-life of the MNK1 protein between the two groups was relatively close, which indicated that miR-7-5p had only few impacts on MNK1 protein stability. So, it could completely inhibit the activation of the MNK/eIF4E axis induced by Everolimus. Hence, we tried to use miR-7-5p-loaded exosomes as a candidate therapy. Here, we used stable miR-7-5p overexpressed A549 cells derived exosomes. They were able to be internalized by the A549 cells (Fig. 2g), and the rate of internalization is greater than the rate of secretion induced by Everolimus (Fig. 2h). The same as stable upregulation of miR-7-5p, they suppressed the phosphorylation of MNK/eIF4E axis without activation of downstream of mTOR (Fig. 2i), which suggested that miR-7-5p-loaded exosomes derived from tumor cells might be served as a new mechanism to alleviate the resistance of Everolimus.
The declined miR-7-5p and elevated MNK1 are associated with poor prognosis of NSCLC To investigate the clinical signi cance of miR-7-5p and MNK1 in NSCLC specimens, we rstly used qPCR con rmed that the levels of miR-7-5p declined in cancer tissues compared to paired adjacent tissues (Fig. 3a), while the MNK1 levels signi cantly elevated in cancer tissues rather than paired adjacent tissues (Fig. 3b) in 34 pairs fresh tissues.
There was a negative correlation between the mRNA levels of miR-7-5p and MNK1 in NSCLC (Fig. 3c, R=-0.3004, P = 0.0128). Then, we performed the in-situ hybridization and immunohistochemistry in 318 cases of para n-embedded NSCLC tissues (including 161 cases of lung adenocarcinoma and 157 cases of lung squamous cell carcinoma) and 90 cases of adjacent tissues respectively to further detect miR-7-5p and MNK1. Positive miR-7-5p and MNK1 were mainly located in the cytoplasm (Fig. 3d). Strongly positive miR-7-5p was observed in normal alveolar epithelial cells, accompanied by weak staining of MNK1. In adenocarcinoma and squamous cell carcinoma tissues, miR-7-5p was declined, while MNK1 was elevated (Fig. 3e).
NSCLC patients tended to have poor prognosis if miR-7-5p was lower (Fig. 3g, upper), or MNK was positive (Fig. 3g,  middle). Optimal outcomes were observed if combined higher miR-7-5p and negative MNK1 was positive. But lower miR-7-5p and positive MNK1 coexist with a herald of the worst prognosis. And the prognosis was intermediate in other expression patterns with no signi cant statistical difference (Fig. 3g, lower). Multivariate regression analysis showed that besides LNM status, clinical stages, gender and pathological grades, miR-7-5p and MNK1 are also independent prognostic factors for NSCLC ( Table 2). Correlation analysis discovered that there was a signi cant negative correlation between the miR-7-5p and MNK1 (Table 3, r=-0.143, P = 0.011). The human NSCLC clinical data supported the concept that miR-7-5p appeared as a tumor suppressor, but MNK1 had an oncogenic function. The balance of the two biomarkers plays an important role in the pathogenesis of NSCLC. Therefore, restoring the levels of miR-7-5p not only has a tumor suppressor effect, but also reverses the phosphorylation of the MNK/eIF4E axis.  Abbreviations: CI, con dence interval; LNM, lymph node metastasis; NSCLC: non-small cell lung cancer. Exogenous miR-7-5p could supply the de ciency of it. Further contrast among the groups of Everolimus, exogenous miR-7-5p and combined of them, we discovered that Everolimus alone or stable miR-7-5p expression could inhibit tumor cell proliferation (Fig. 4a & S2a), Clone formation ( Fig. 4b & S2b) and 2D&3D migration capabilities (Fig. 4c- S2c-d, respectively). When Everolimus was carried out to pose on stable miR-7-5p cells, the inhibitory effect was the most signi cant.
As the functions of exosomes mediated by their contents are more widely recognized, they have been proposed as a potential cell-based alternative therapy which have received increasing attention 19 . Due to the direct sorting and packaging of nucleic acids into exosomes may not provide the functionally active contents into recipient cells effectively 20 . We tried to use exosomes derived from tumor cells directly to improve miR-7-5p in Everolimus treated NSCLC cells. Like exogenous miR-7-5p, exosomal miR-7-5p could inhibit the cell proliferation ( Fig. 4e & S2e), Clone formation ( Fig. 4f & S2f) and 2D&3D migration capabilities (Fig. 4g-h & S2g-h, respectively). And further usage of miR-7-5p-loaded exosomes combined with Everolimus showed a more obvious tumor suppressive effect, which indicated that exo-miR-7-5p could be internalized by NSCLC cells and cooperated with Everolimus to enhance the anticancer effect of it in vitro.
We used LV-miR-7-5p and LV-NC A549 cells to construct nude mouse subcutaneous tumor models respectively. Once the tumor volume reached 50-100mm 3 , they would be divided into four groups and received indicated treatments. The growth rate of the LV-miR-7-5p group was signi cantly slower than the control group, as well as the tumor volume. It was worth noting that although Everolimus had a strong inhibitory effect on tumor growth in the early stage, its growth rate tended to accelerate in the later period, resulting in the relatively bigger volume than LV-miR-7-5p combined Everolimus group. The proliferation rate of tumor cells and tumor volume always kept at a lowest level when LV-miR-7-5p was performed with Everolimus ( Fig. 5a-d). Further, in spite of that Everolimus could suppress the Ki-67 index of tumors than the control groups no matter combined with LV-NC or LV-miR-7-5p, the activation of MNK1/eIF4E axis was only observed in the LV-NC combined Everolimus group, (Fig. 5e), meant restoring the levels of miR-7-5p could be used as an effective method to reverse the activation of MNK/eIF4E axis induced by Everolimus.
Currently, the treatment of cancers highlights the need for more effective and innovative therapies. Because of safer biological behaviors, exosomes have the value of translational medicine for clinical treatment purposes 20 . Using tumor-derived exosomes will provide additional competitive advantages for the selective delivery of anticancer therapies not only to the primary tumor but also to the premetastatic niche and even to metastasis, owing to their intrinsic organotropic tumor-homing properties 21 . We further established the abdominal metastasis models accepting the indicated treatments to monitor tumors in vivo and the exosomal miR-7-5p in the blood. Tumor involvement was observed in all groups ( Fig. 5f & S3). Notably, the exo-miR-7-5p combined with Everolimus group had the weakest uorescence signal (Fig. 5g), and high miR-7-5p in the blood (Fig. 5h).

The miR-7-5p combined with Everolimus can induce apoptosis via dual abrogation of MNK/eIF4E and mTOR in NSCLC
The function of eIF4E is relay on the subcellular locations of it. The initiation of the translation of most mRNAs is controlled by intracytoplasmic eIF4E. And if it locates in the nucleus, the export of certain mRNAs to the cytoplasm will depend on it 22 . In order to further explore the speci c value and mechanism of the combined targeting of the MNK/eIF4E axis and the mTOR pathway in the treatment of NSCLC, the rst synthetic small molecule MNK1 inhibitor CGP57380 was performed to block the MNK/eIF4E axis. We found p-MNK1 Thr197/202 was mainly distributed in the nucleus and was inhibited. In spite of p-eIF4E S209 was located both in the cytoplasm and nucleus, CGP57380 mainly suppressed the phosphorylation of intracytoplasmic eIF4E (Fig. 6a, Figure S4a). Besides, it still has characteristics such as poor speci city and relatively weak a nity, which limits its value in clinical treatment 23 . Fortunately, both endogenous and exosomal miR-7-5p could restore the intracellular tumor suppressor miR-7-5p and reverse the feedback activation of MNK/eIF4E axis more speci c and safer. When Everolimus was performed in stable LV-miR-7-5p A549 and SPC-A1 cells (Fig. 6b, Figure S4b) or combined with the miR-7-5p-loaded exosomes derived from A549 cells (Fig. 6c, Figure S4c), the activation of the MNK/eIF4E axis was eliminated.
Further ow cytometry analysis indicated the apoptosis rate of miR-7-5p mimics or Everolimus treatment alone was less than the combined group (Fig. 6d, Figure S4d). And representative microscopic images of liver metastatic lesions among all groups in previous abdominal metastasis models observed obvious necrosis of metastases in exo-miR-7-5p combined with Everolimus group (Fig. 6e), which suggested that the miR-7-5p in combination with Everolimus could signi cantly strengthen the therapeutic effect of Everolimus through inducing cell death. Mechanismly, caspase-3, cleaved caspase-3, DR4, DR5, pro-apoptotic proteins Bak, Bad and Bax were signi cantly up-regulated but anti-apoptotic protein Bcl-xL was down-regulated ( Fig. 6f-h, Figure S4f-h) when Everolimus posed on the stable LV-miR-7-5p NSCLC cells, indicating it could induce the co-activation of the inherent mitochondrial apoptosis pathway and the death receptor pathway.

Discussion
Everolimus is a derivative of rapamycin, which could inhibit the activity of mTORC1 and be approved for the treatment of various cancers. Regrettably, the single drug activity in the cancers is limited 24 . Multiple studies have shown that targeted inhibition of mTOR is able to activate a variety of survival signaling pathways, including PI3K/Akt, MEK/ERK and Mnk/eIF4E, which will provide cancer cells with a survival advantage and ultimately enable rapamycin treatment 15 . In addition, long-term effects of rapamycin lead to changes of intracellular miRNA expression pro le. In rapamycin-resistant cells, oncogenic miRNAs were signi cantly up-regulated, while tumor suppressor miRNAs were down-regulated. The tumor suppressor miRNA-101 in the cell culture medium of two clear cell renal cell carcinoma cell lines increased by the usage of mTOR inhibitors, indicating that mTOR inhibitors promoted the excretion of it into the cellular microenvironment to neutralize its anti-tumor activity 25 . After comparing the expression pro les of miRNA between parental myoblasts and drug-resistant strains, there was down-regulation of miR-7a 8 . This is consistent with our research. We con rmed that Everolimus promoted the decreasing of miR-7-5p in NSCLC. The bioinformatics predictions and luciferase experiments validated that miR-7-5p could target the mRNA of MNK1. As the eIF4E phosphorylation was increased by mTOR inhibitors in a MNK-dependent manner 26 . We speculated that Everolimusmediated activation of MNK/eIF4E might be related to miR-7-5p, which had not been studied before. So, we elevated or suppressed the miR-7-5p by mimics or inhibitor, and combined with Everolimus or not to further test the phosphorylation of MNK/eIF4E axis and the activity of downstream molecules of mTOR pathway in NSCLC cells. It was discovered that up-regulation of miR-7-5p could indeed inhibit the phosphorylation of the MNK/eIF4E axis without feedback activating of mTOR pathway. However, miR-7-5p could inhibit protein ubiquitinated degradation 17 .
And MNK1 protein could be degraded by ubiquitination in addition to being regulated by phosphorylation 18 . If the upregulation of miR-7-5p inhibits the ubiquitinated degradation of MNK1, it would offset the tumor suppressor effect of miR-7-5p. Hence, we further veri ed that the targeted inhibitory effect of miR-7-5p on MNK protein was independent on the proteasome mediated degradation of MNK1. The above studies indicated that Everolimus targeted mTORC1 inducing the loss of intracellular miR-7-5p to activate MNK/eIF4E axis mediating Everolimus resistance. This has not been reported in other studies, and it is worthy of further research and exploration.
In recent years, many studies have shown that miR-7 may be a tumor suppressor, and are signi cantly downregulated.
But there were also some opposite reports 27 . In NSCLC, the role of miR-7 was also contradictory. It was found that EGFR promotes lung tumorigenesis by activating miR-7 28 . While miR-7 was signi cantly higher in serum exosomes from healthy controls than from patients with lung carcinoma, and higher miR-7 was associated with strong response to lung carcinoma patients receiving ge tinib treatment 29 . Therefore, further research is essential to identify whether miR-7 is a oncomiR or tumor suppressor in NSCLC. Here, we found that there was lower expression of miR-7-5p in multiple NSCLC cell lines, fresh tissues and para n-embedded tissues of NSCLC, which was related to the poor prognosis of patients. When miR-7-5p was overexpressed by lentivirus or exosomes, it could inhibit the proliferation, growth and migration of NSCLC in vivo and in vitro. Our research supported that miR-7-5p had tumor suppressor value in NSCLC. As for the apparently contradictory functions of miR-7-5p, it might be explained by the balance of the positive or negative feedback loops in different cancer types or stages.
The transfer of miRNAs by exosomes is considered to be a novel and important mechanism of genetic exchange between cells 30 . Through analyzing TCGA database, it was found that if the mTORC1 was low, the extracellular exosomes would be signi cantly enriched. Exactly, miR-7-5p was recruited in exosomes, which could be inhibited by the exosome inhibitor GW4869. Hence, we speculated Everolimus could stimulate NSCLC cells to excrete miR-7-5ploaded exosomes, thereby reducing intracellular miR-7-5p to weaken the anti-tumor activity of Everolimus. Previous studies believed that mTORC1 controlled cell metabolism by removing organelles and components, which was controlled by the main mechanism of autophagy. Recently, studies have found that the release of exosomes and autophagy were two related processes leading to the loss of cell membrane and diverse contents. The release of exosomes was also regulated by mTORC1 7 . Everolimus suppressed the activity of mTORC1 inducing the release of exosomes. In contrast the release of exosomes was signi cantly reduced after activation of mTORC1 by knockdown of TSC1/2, important negative regulators of mTORC1. Besides, the regulation of release of exosomes by mTORC1 was also depended on two crucial exosomes regulatory proteins, Rab27A and Rab27B. They might participate the fusion of vesicles and plasma membrane in the process of exosomes secretion 31 . Different from other research, Everolimus induced exosomes derived from NSCLC cells were showed up as enrichment of miR-7-5p. There might be a mechanism involved in the sorting process of miR-7-5p into exosomes derived from Everolimus treated NSCLC cells.
According to the report, the motif of "GUUG" was associated with indicated miRNAs into exosomes in SW620 cell 32 .
Coincidentally, the base sequence of miR-7-5p (UGGAAGACUAGUGAUUUUUGUUGU) contains the "GUUG" motif, which might be a candidate reason for miR-7-5p sorting process. But the possible mechanism needs further veri cation.
The abnormal miRNA pro le of tumor cells accepted Everolimus treatment also supported corresponding targeted therapy strategies. And therapeutic intervention might be aimed at restoring normal miRNA levels 33 . Therefore, restoration of miR-7-5p could sensitize the effect of Everolimus. The exosomes could be used as carriers, which had the value of low immunogenicity, low toxicity, and high stability via being modi ed to improve loading and targeting e ciency 34 . The miRNAs loaded in exosomes relied on the transcripts of parental cells 35 . Over-expression of miRNAs in parental cells will increase indicated miRNA in extracellular vesicles (EVs) 36 . Therefore, the e ux exosomes derived from miR-7-5p overexpressed NSCLC cells had a higher content of miR-7-5p, which could be internalized by A549 Therefore, based on the changes of miRNA pro le in NSCLC cells induced by Everolimus, here we used tumor cellderived exosomes to restore miR-7-5p in cells, and combined with Everolimus, which has an important clinical application value.
The compensation feedback mechanism links the AKT/mTOR pathway with the MNK/eIF4E axis. The downregulation of one pathway was associated with the activation of another one, which subsequently supported cell proliferation and cancer survival 40 . The complexity of cell signaling means that traditional single treatment methods were usually ineffective, further multiple targets were needed to be found 41 . The e cacy of combined targeted therapy was signi cantly better than that of single medication. The MNK deletion can reduce basal mTORC1 signaling and MNK activation contributes to rapamycin resistance in cancer cells by sustaining mTORC1 activity following rapamycin treatment 42 . The combination of MNK inhibitor CGP57380 with Everolimus overcomes Everolimus-mediated eIF4E phosphorylation and shows enhanced cytotoxicity against T-cell acute lymphoblastic leukemia cells 40 44 . We speculated the up-regulation of miR-7-5p was able to inhibit other targets which associated with death receptor pathway-mediated apoptosis. This apoptosis-based cancer treatment strategy could selectively activate death-induced molecules in the apoptosis program of transformed cells for tumor eradication, which provided important support for the clinical application of tumor suppressor miR-7-5p.
In conclusion, our results prove a hypothesis that Everolimus stimulates the release of miR-7-5p loaded exosomes from NSCLC cells in Rab27A and Rab27B dependent manners, thereby reducing the intracellular miR-7-5p levels, which attenuates the inhibition of MNK1 and subsequently promotes MNK-dependent eIF4E phosphorylation. Both endogenous and exosomal miR-7-5p all have the synergy with Everolimus, which not only inhibits the proliferation, migration, and metastasis of NSCLC in vitro and in vivo, but also promotes the apoptosis of NSCLC through dual targeting MNK/eIF4E axis and mTOR pathway. Besides, both low level of miR-7-5p and positive expression of MNK1 could act as independent poor prognostic biomarkers for NSCLC, which may provide a promising strategy for mTORtargeted cancer therapy and prognosis (Fig. 6h).
Both endogenous miR-7-5p or exo-miR-7-5p collaborated with Everolimus, not only inhibited the proliferation, migration, and metastasis of NSCLC in vitro and in vivo, but also promoted the apoptosis of NSCLC via targeting MNK/eIF4E axis. Therefore, miR-7-5p could act as a combined therapeutic strategy for mTOR-targeted cancer therapy and prognosis. Declarations Figure 1 Everolimus targets mTORC1 inducing NSCLC cells to secrete miR-7-5p-loaded exosomes in Rab27A and Rab27B dependent manners. a. The levels of miR-7-5p in Everolimus treated NSCLC cells (A549, H358, H520 and SPC-A1) were measured by qPCR for 16h, 24h and 48h. T-test. Each bar represents mean ± SD. b. GSEA analysis showed the difference gene set between mTROlow and mTORhigh. ES, enrichment score; NES, normalized enrichment score. c.
Transmission electron microscopy images for exosomes pointed by the red arrow (scale bar = 100 nm). d. Size distribution analysis of puri ed exosomes by Zetasizer Nano ZS90. e. Immuno uorescence was carried out to detect the expression of exosomal marker CD63 (green) in A549 and SPC-A1 cell lines with or without Everolimus treatment.
f. Exosomal markers including HSP70, TSG101, CD63 and CD9 were analyzed by Western blotting in A549 cells and A549-derived exosomes. DAPI (blue) was used for nuclear staining. g. The levels of intracellular and exosomal miR-7-5p were measured by qPCR. A549 cells were treated with 5 nM Everolimus combined with or without 10 μM GW4869 Exosomal miR-7-5p sensitizes the anti-cancer effect of Everolimus in vitro. Indicated A549 cells were treated with or without Everolimus at the settled time. The cell proliferative ability was determined by a. CCK8 assay and b. cell cloneformation assay. The cell migration and invasive ability was determined by c. cell scratch test and d. transwell matrigel assay. A549 cells were treated with indicated exosomes or Everolimus for settled times. The cell proliferative ability was determined by e. CCK8 assay and f. cell clone-formation assay. The cell migrating and invasive ability was determined by g. cell scratch test and h. transwell matrigel assay. Data are shown as mean ± SD. *P < 0.05, **P< 0.01, and ***P < 0.001 compared with control.

Figure 5
Exosomal miR-7-5p sensitizes the anti-cancer effect of Everolimus in vivo. Five-week-old male nude mice were randomly divided into two groups to establish LV-NC and LV-miR-7-5p A549 cells subcutaneous xenograft nude mice models. As soon as the tumor volume reached 50-100mm3, they would be divided into four groups named as 1). the LV-NC group, 2). the LV-miR-7-5p group, 3). the LV-NC combined Everolimus group and 4). the LV-miR-7-5p combined 6 groups for receiving PBS, exo-NC, exo-miR-7-5p, Everolimus, exo-NC combined with Everolimus and exo-miR-7-5p combined with Everolimus treatment, respectively. f. Upper: Schematic illustration of abdominal tumor xenograft nude mice models. Lower: Luminescence signals of intraperitoneal A549-luciferase cells with different treatment groups at the indicated week. Results are shown as the mean ± SD. Also see in the Figure S3. g. Analysis of intraperitoneal A549-luciferase cells accepted different treatment groups at the indicated week. h. The levels of exo-miR-7-5p from the plasma of nude mice receiving treatments above were measured by qPCR. * P < 0.05, ** P < 0.01, ***P < 0.001.

Figure 6
The miR-7-5p combined with Everolimus can induce apoptosis via dual abrogation of MNK/eIF4E and mTOR in NSCLC Western blotting analysis to detect the proteins of p-MNK1 Thr197/202 and p-eIF4ES209 respectively extracted from cytosolic or nuclear in the indicated A549 cells. GAPDH was used as a loading cytoplasmic control.
And Histone H3 was used as a loading nuclear control. a. A549 cells were exposure to MNK inhibitor CGP57380 or