CARM1 Promotes Gastric Cancer Progression by Regulating TFE3 Mediated Autophagy Enhancement via Cytoplasmic AMPK-mTOR and Nuclear CARM1-TFE3 Signaling Pathways

Background: The role of CARM1 in tumors is contradictory. It acts as an oncogene in most kinds of cancers while inhibits the progression of liver and pancreatic cancers. CARM1 has recently been reported to regulate autophagy, which is also context-dependent. However, the effect of CARM1 on gastric cancer has not been studied. We aimed to explore whether CARM1 was involved in the progression of gastric cancer by regulating autophagy. Methods: The clinical values of CARM1 and autophagy in gastric cancer were determined by immunohistochemistry and qRT-PCR. Transmission electron microscopy, immunouorescence and western blotting were applied to recognize autophagy. The role of CARM1 in gastric cancer was investigated by CCK8, colony formation and ow cytometry assays in vitro and xenograft model in vivo. Immunoprecipitation assay was performed to illustrate the interaction of CARM1 and TFE3. Results: CARM1 was upregulated in clinical GC tissues and cell lines, and higher CARM1 expression predicted worse prognosis. CARM1 enhanced GC cell proliferation, facilitated G1-S transition and inhibited ER stress-induced apoptosis by regulating autophagy. Importantly, the treatment of CARM1 inhibitor rescued the tumor-promoting effects of CARM1 both in vitro and in vivo. Furthermore, we proved CARM1 heightened TFE3 nuclear translocation to induce autophagy via cytoplasmic AMPK-mTOR and nuclear CARM1-TFE3 signaling pathways. Conclusion: CARM1 promoted GC cell proliferation, accelerated G1-S transition and reduced ER stress-induced apoptosis by regulating autophagy. Mechanically, CARM1 triggered autophagy by facilitating TFE3 nuclear translocation via AMPK-mTOR and CARM1-TFE3 signaling pathways. of dissected xenograft showed that the overexpression of CARM1 (top) also accelerated in compared to controls (bottom). J Tumor volumes at indicated formula: tumor volume width 2] K Tumor weights were also measured, that the xenograft tumors derived from CARM1 overexpression cells grew faster than that of control gastric cells. Data are clone C and D BGC823 CARM1-overexpression cells were treated with CARM1i 24h). Cell cycle phase distribution and apoptosis in indicated were assessed by ow cytometry. E, F and G BGC823 xenograft model was performed to investigate the therapeutic effect of CARM1i and HCQ on GC in vivo. Nude mice injected with CARM1 overexpression cells were randomly divided into four groups at day 6 when tumor volumes reached 50 mm3. CARM1i was performed twice daily at 100 mg/kg i.p. HCQ was administered once daily at 50 mg/kg i.p. The combination group was given two treatments and the control group was only intraperitoneally injected with PBS. E The photograph of subcutaneous tumors separated from indicated groups. F The tumor volumes were measured every four days, following this formula: tumor volume (mm3) = [length (mm) × width (mm) 2] ×π/6. G The tumor weights were evaluated. Both CARM1i and HCQ could slow down the tumor growth of GC. Furthermore, CARM1i exhibited a synergistic effect in combination with the treatment of HCQ. Data are demonstrated as mean ± SD. *Represents *P < 0.05, **P < 0.01 and ***P < 0.001. in nucleus protein by western blotting. β- Actin and H3 served as loading controls for cytoplasmic proteins and nuclear proteins respectively. E and F Colony formation and CCK-8 assays revealed TFE3 downregulation reversed the proproliferative effect of CARM1 on BGC823 cells. G TFE3 knockdown rescued increased G1-S transition and decreased apoptosis resulted from CARM1 overexpression in BGC823 cells.


Introduction
Gastric cancer(GC) is the fth commonly diagnosed malignant tumor and the fourth leading cause of cancer death globally(1), causing severe economic burden. Multidisciplinary treatment has become the mainstream treatment for GC patients, especially with the development of immunotherapy and targeted therapy (2). However, the number of patients bene ting from the new treatment is very limited due to individual heterogeneity. Therefore, it is very important to explore the fundamental molecular pathogenesis of GC and nd new diagnostic biomarkers and therapeutic targets.
Autophagy is a self-eating process to recycle wastes of cells and maintain homeostasis through clearing longevity proteins or old organelles. These cellular contents and organelles are sequestered in twomembrane structures called autophagosomes, which are then fused with lysosomes to degrade the cargo (3). Autophagy is activated when cells encounter environmental stress, such as malnutrition, hypoxia, pathogen infection, oxidative stress, leading to adaptation or cell death depending on the intensity of the stimulus and the host (4). Autophagy is reported to exhibit crucial con icting functions in tumor progression, including pancreatic adenocarcinoma, myeloid leukemia, gastric carcinoma, squamous cell carcinoma, sarcoma, multiple myeloma, etc(5-8). More importantly, there have been several clinical trials targeting autophagy in malignant glioma, non-small cell lung cancer, colon cancer, melanoma, pancreatic cancer and melanoma, implicating the promising prospects for targeting autophagy therapy (9).
Arginine methylation regulated by the protein arginine methyltransferase (PRMT) family is one of the most vital epigenetic modi cations in autophagy (10). Of which coactivator-associated arginine methyltransferase-1 (CARM1), is special because it is the only enzyme capable of methylating arginine with proline sequences (11). CARM1 has two main domains, chromatin remodeling proteins and proteins possessing RNA binding activity, which play essential roles in chromatin remodeling, gene transcription, DNA packing, pre-mRNA splicing, and mRNA stability (12,13). The role of CARM1 in cancers is paradoxical. It acts as an oncogene in lung, prostate, colorectal, ovarian, breast cancer, Osteosarcoma, myeloid leukemogenesis and multiple myeloma (14)(15)(16)(17)(18)(19)(20)(21)(22)(23). However, CARM1 inhibits the progression of liver and pancreatic cancers (19,20). Previous studies mainly focused on the role of CARM1 as a coactivator in regulating tumor-related gene (24,25). Recent studies have shown that CARM1 is an important regulator of autophagy mainly through two dependent pathways. In the nucleus, AMPK is activated under glucose starvation and subsequently phosphorylates and activates FOXO3a, leading to a decrease in SKP2 expression and resulting in the stabilization of CARM1 in the nucleus, which acts as a coactivator of TFEB to promote autophagy-related gene transcription (26). While in the cytoplasm, C9orf72 mediates CARM1 lysosomal degradation by interacting with the PH-like domain and modulates lipid metabolism (27). However, the role of CARM1 in GC has not been reported previously, and it remains unclear whether CARM1 could affect the progression of GC by regulating autophagy.
In this study, we illustrate the effect of CARM1 on GC and the mechanism of CARM1 regulating autophagy. CARM1 and autophagy marker were upregulated in GC tissues and higher CARM1 expression predicted worse prognosis. Knockdown CARM1 resulted in autophagy ux blockade and succeeding ER stress-induced apoptosis, reduced cell proliferation via increased G1 phase cell cycle arrest, which could be partially reversed by autophagy agonists rapamycin. While overexpression of CARM1 exhibited the opposite results. More importantly, the treatment of CARM1 inhibitor could rescue the tumor-promoting effects of CARM1 both in vitro and in vivo. Furthermore, we proved CARM1 facilitated TFE3 nuclear translocation to induce autophagy via cytoplasmic AMPK-mTOR and nuclear CARM1-TFE3 signaling pathways. Thus, our ndings suggest CARM1 alone or in combination with agents represent novel therapeutic strategies targeting GC.

Tissue microarray and clinical samples
Tissue microarray (G6038-3) including 48-paired GC tissues and relevant adjacent non-neoplastic tissues were obtained from Servicebio Biotech (Wuhan, China) containing clinical information such as gender, age, tumor location and size, metastasis, vessel carcinoma embolus and the overall survival of patients.
After immunohistochemistry assay, the immunostaining intensity scores were evaluated by two experienced pathologists, performed double-blind reading. The percentage of positive cells and the staining intensity were analyzed by semi-quantitative results according to the following criterion: On the one hand, the percentage of positive cells was divided into 5 grades: the number of positive cells < 5% was 0, 5% ~ 25% was 1, 26% ~ 50% was 2, 51% ~ 75% was 3, 76% ~100% was 4 points. On the other hand, the staining intensity was recognized as 4 grades: negative was 0, weak was 1, moderate was 2, and strong was 3 points. The total scores were calculated by adding up the product on a scale range of 0-12. Tissues of total score < 7 were classi ed into low expression while score ≥ 7 were recognized as high expression. 33 pairs of clinical tissues were collected from the Second A liated Hospital of Xi 'an Jiaotong University with informed consent from patients.

Immunohistochemistry (IHC)
Then tissue microarray and para n sections were soaked in xylene twice for depara nating. Rehydrated the slides with series ethanol (100%, 95%, 85% and 75%). Slides were evenly spaced in citrate buffer for microwaving treatment to perform antigen retrieval. Afterwards, added 3% hydrogen peroxide to remove endogenous peroxidase and then the slides were blocked with goat serum for 30 minutes at room temperature (RT). Then, added appropriate concentration of primary antibodies and incubated at 4°C overnight. The next day, added secondary biotinylated antibodies and incubated at RT for 1 hour. Then, fresh DAB staining solution was added and the staining degree was observed under a microscope.
RNA extraction and qRT-PCR Total RNA was extracted by TRIzol (Invitrogen, Thermo, Waltham, MA, USA). Afterwards, the messenger RNA was transcribed reversely by the Transcriptor First-Strand cDNA Synthesis Kit (Roche, Basel, Switzerland). Subsequently, we used the FastStart Universal SYBR Green Master (Roche) to conduct qRT-PCR analysis following the manufacturer's instructions. All of the primer sequences were listed in Additional le 4: Table S2.
Cell culture Human GC cell lines HGC27, BGC823, MKN45, SNU-1 and the normal gastric epithelial cell line GES1 were purchased from Genechem (Shanghai, China), identi ed by STR pro ling and tested free of mycoplasma contamination. All cell lines were cultured in high glucose DMEM (HyClone, Logan, Utah, USA) supplemented with 10% fetal bovine serum (Gemini, Calabasas, CA, USA) and 1% penicillin-streptomycin (Gibco, Grand Island, NY, USA) in a humidi ed incubator in 5% CO2 at 37°C.

Western blotting
Total protein was extracted using cold RIPA cracker buffer (Beyotime, Shanghai, China). Cytoplasmic and nuclear extracts were separated by applying NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Scienti c™, MA, USA). Determine the protein concentration of each cell lysate using BCA Protein Assay Kit (TIANGEN, Beijing, China). Next, the same amount of protein samples was applied to SDS-PAGE gel wells and then transferred to the PVDF membranes. The membrane was incubated overnight at 4 ℃ in a primary antibody solution. The involved antibodies were demonstrated in Additional le 5: Table S3. The next day, the membrane was incubated with horseradish peroxidase bound secondary antibody and visualized with enhanced chemiluminescence reagents.

Lentivirus and siRNA transfection
The overexpression lentivirus and control lentivirus were obtained from Hanbio Biotrchnology (Shanghai, China). The component order of overexpression lentivirus was HBLV-h-CARM1-3x ag-ZsGreen-PURO, while the control component was HBLV-ZsGreen-PURO. The transduction was performed according to the manufacturer's protocol. Then culture the transfected cells in medium containing 2.5 µg/mL puromycin to obtain stable overexpression cells. HGC27 and BGC823 CARM1-overexpression cells were transfected with speci c siRNAs (GenePharma, Shanghai, China), targeting CARM1 and TFE3 respectively. Lipofectamine 2000 reagent (Invitrogen, Calrsbad, CA) was added to facilitate transfection in accordance with the manufacturer's instructions. The interference sequences involved were shown in Additional le 4: Table S2.

Immuno uorescence
Cells were cultured on small cover glass. 4% paraformaldehyde was added into the hole for cell xation for 30 minutes at RT. Then treat cells with 0.1% Triton 100 to permeate cells for 15 minutes. Block cells with 5% goat serum without washing. Then su cient amount of the appropriate concentration of primary antibodies were added overnight in a moist container at 4℃. In the following day, cells were incubated with uorescence labeled secondary antibody and for 1 hour and DAPI for 1 minute in dark. In the end, sealed the coverslip with anti-fading buffer (BD Biosciences, NJ, USA). Photographs of ve random elds were captured with confocal microscopy (Nikon C2, Tokyo, Japan).

CCK8 assay
Cell suspensions were added in 96-well plates with 3 × 10 3 cells per well in 100 µL DMEM and incubated for 1-4 days. Each well was added 10ul CCK-8 regent (7 Sea Pharmatech Co., Ltd, Shanghai, China) and incubated in the cell incubator for 1 hour. Then read the absorbance value at 450nm determined by a microplate spectrophotometer (Thermo, Waltham, MA, USA).

Colony formation assay
Add cells in the exponential growth phase at a density of 1 × 10 3 cells each well in 6-well plates. Then cultivate the cells for 10-14 days until visible clones were observed. In the process, replace fresh culture medium timely according to the pH change. Afterwards, xed cells with 4% paraformaldehyde for 20 minutes and stained with crystal violet dye for 10 minutes. Finally, counted the number of colonies containing more than 50 cells.
Flow cytometry 2 × 10 5 cells were plated in 6-well plates and grew in serum-free medium for 24 hours to keep the same cell cycle. Then culture cells for another 24 hours. The adherent cells were digested and xed by 70% ice ethanol at 4℃ overnight. The next day, the xed cells were washed and resuspended with 500µL PI/RNase Staining Buffer (BD Biosciences, Franklin Lakes, NJ, USA). After culturing for 15 minutes in dark at RT, ow cytometry (BD Biosciences, Franklin Lakes, NJ, USA) was used to detect different cell cycle phases of cells. Seed 1.5 × 10 5 cells in 6-well plates each well and then continue to cultivate cells to a con uence of 80%-90%. Afterwards, the proportion of apoptosis cells was determined by PE Annexin V/7-aminoactinomycin (7-AAD) Detection Kit (BD Biosciences, NJ, USA) following the manufacturer's instructions and analyzed by ow cytometry.
In vivo xenograft tumor model A subcutaneous xenograft model was established using 5-6 weeks old male BALB/c nude mice (Xi'an Jiaotong University Medical Laboratory Animal Center, Xi'an, China). The experiments were approved by the animal ethics committee of Xi'an Jiaotong University. We subcutaneously injected 1.0 × 10 6 tumor cells into the left groin of nude mice where blood ow was abundant. The volume of subcutaneous tumor was evaluated every 3 days. For the drug research, nude mice injected with CARM1 overexpression cells were randomly divided into four groups at day 6 when tumor volumes reached 50 mm3. CARM1 inhibitor EZM2302 was performed twice a day at 100 mg/kg i.p. HCQ was administered once a day at 50 mg/kg i.p. The combination group was given two treatments and the control group was only intraperitoneally injected with PBS. When the subcutaneous tumor showed ulceration and necrosis, the mice were sacri ced, the transplanted tumors were removed, and the weight and volume were measured.

Immunoprecipitation assay
The cells were harvested and added appropriate amount of IP lysis buffer (including protease inhibitor) for 30 min at 4°. After centrifugation, 6 µg CARM1 antibody or control IgG and 40 µ L Protein A/G Mix Magnetic Beads (Merck Millipore, Germany) were added into the supernatant and incubated at 4°C overnight. After immunoprecipitation reaction, the protein was carefully eluted from the protein A/Gbeads and denatured for western blotting analysis.

Statistical analysis
We used GraphPad Prism 6 (San Diego, CA, USA) and SPSS 20.0 (Chicago, IL, USA) to perform statistical analysis. When comparing the differences in measurement data between the two groups, Student's t test was used. When comparing the differences between more than two groups, ANOVA was applied, in which p value were adjusted in face of multiple comparisons. Before Student's t test and ANOVA, a variance homogeneity test and normality analysis were carried out. Overall survival curves were analyzed by the Kaplan-Meier method and log-rank tests. Chi-squared test was utilized to estimate the difference of CARM1 level between gastric cancer tissues and adjacent non-tumor tissues. The hazard ratio in clinical samples was determined by cox proportional hazards model for univariate and multivariate analyses. All values demonstrate means ± SD. All statistical tests were two-sided, and P < 0.05 was considered statistically signi cant.

Results
CARM1 and autophagy marker are upregulated in human GC tissues and higher CARM1 expression demonstrates poor prognosis Through analysis of CARM1 expression in TCGA database, we found CARM1 mRNA expression was upregulated in GC compared to normal tissue samples (80 cases) in Fig. 1B. CARM1 expression also increased in GEPIA database (Additional le 1: Fig. S1A). To further investigate the expression of CARM1 in GC tissues, we used GC tissue microarrays containing 48 pairs of patients. The immunohistochemistry (IHC) staining indicated that CARM1 expression increased substantially in GC tissues in comparison with that in adjacent non-tumor tissues (Fig. 1A, C). Furthermore, as shown in Fig. 1D, the overall survival was shorter in patients with higher CARM1 expression, consistent with the KM Plotter database (Additional le 1: Fig. S1B), suggesting that CARM1 was an important biomarker of GC. Moreover, univariate and multivariate cox regression analyses demonstrated that CARM1 was an independent risk factor for predicting poor survival (Table 1, 2). The association of CARM1 level with clinicopathological parameters of GC patients was shown in Additional le 3: Table S1. Due to the important role of autophagy in tumor progression and the regulatory role of CARM1 on autophagy, we analyzed the expression of autophagy markers in human GC tissues. As detailed in Fig. 1E, ATG5, LC3B and beclin1 protein expression increased in cancer compared to normal tissues, meanwhile, the mRNA levels were also higher in cancer analyzed by 33 paired clinical samples (Fig. 1F).

CARM1 promotes autophagy in GC cells
To explore the role of CARM1 in GC cells, we evaluated the expression of CARM1 in 4 GC cell lines and normal gastric mucosal epithelial cell line GES1. The results revealed that both CARM1 protein and mRNA expression were increased in GC cell lines compared to GES1, CARM1 levels increased slightly in BGC823 while grew dramatically in HGC27 (Fig. 1G, H). Then we established CARM1-knockdown HGC27 cells using speci c siRNA and stable CARM1-overexpression BGC823 cells by transfection with lentivirus. The expression of CARM1 in treated HGC27 and BGC823 cells was con rmed by western blotting and qRT-PCR, presented in Fig. 2A, B.
Previous studies proved CARM1 was related to the regulation of autophagy(11), so we identi ed autophagosomes and autolysosomes to assess autophagy in GC cells. As can be seen in Fig. 2C, HGC27 cells with lower CARM1 expression presented fewer autophagosomes and autolysosomes in comparison with control cells, while the number in BGC823 cells transfected overexpression lentivirus increased signi cantly compared to normal control cells. Consistently, enhancive LC3 puncta and increased protein expression of LC3 II /I, ATG5, beclin1and reduced p62 expression in CARM1 overexpression cells further proved that CARM1 could induce autophagy (Fig. 2D, E).

Blocked autophagy ux induces endoplasmic reticulum stress
It has been reported that autophagy and ER stress have complex interactions, and ER stress can induce autophagy through unfolded protein reactions(UPR)(28). Classical UPR includes three stress transducers, ATF6, IRE1α and PERK (29). So we evaluated ER stress related proteins and discovered that IRE1α and ATF6 expression showed no difference in GC cells treated with CARM1 siRNA or overexpression virus (Additional le 1: Fig. S1C). However, the expression of p-PERK, p-eIF2α and ATF4 accelerated in CARM1 downregulation HGC27 cells while retarded when CARM1 was upregulated, consistent with the change of GRP78, a molecular chaperones of ER to help proteins fold properly(30) (Fig. 3A), suggesting knockdown of CARM1 triggered ER stress.
On the other hand, lack of autophagy can in turn promote ER stress, the inhibition of autophagy causes excess p62 accumulation, which could impair the delivery of polyubiquitinated proteins to induce UPR(31) , (32). Therefore, we utilized the autophagy agonist rapamycin to investigate whether the increased ER stress was caused by the loss of autophagy ow. As assumed, enhancive autophagy partially reverted ER stress in CARM1 downregulation cells (Fig. 3B). We further used the autophagy inhibitor HCQ to study the changes of ER stress in BGC823 cells, when autophagy was inhibited, the level of ER stress heightened (Fig. 3C). To sum up, these results showed that impaired autophagy resulted in the deposit of p62 and thereby induced ER stress.

CARM1 aggravates the proliferation of GC cells through regulation of autophagy
Since autophagy was recognized as a double-edged in the progression of cancer, we then explored the effect of CARM1 and the induced autophagy on GC cells proliferation. CCK-8 analysis revealed downregulation of CARM1 signi cantly inhibited cell growth (Fig. 4A), however, when HGC27 cells were treated with autophagy stimulant rapamycin, the decreased cell viability partially restored (Fig. 4B). On the contrary, BGC823 cells handled with autophagy suppressant HCQ attenuated cell viability induced by CARM1 (Fig. 4C, D). Then we performed colony formation assay to verify the role of autophagy induced by CARM1 in cell proliferation. Similarly, upregulation of CARM1 increased colony numbers of clone, which could be reversed by HCQ (Fig. 4E, F), while downregulation of CARM1 showed the opposite effect and could be rescued by rapamycin in part (Fig. 4G, H). To further explore the role of CARM1 in tumorigenesis in vivo, we carried out a subcutaneous xenograft model in nude mice. The result showed that overexpression of CARM1 substantially accentuated tumor growth compared to mice injected with CARM1-EV cells (Fig. 4I, J). In keeping with the xenograft tumor volumes, the nude mice bearing CARM1-OE cells revealed heavier weights too (Fig. 4K).

CARM1 facilitates G1-S transition in cell cycle and restrains apoptosis of GC cells by inducing autophagy
To investigate how CARM1 affect cell proliferation through regulating autophagy, we then performed cell cycle assay through ow cytometry analysis. Fig. 5A, B exhibited downregulation of CARM1 increased while upregulation decreased the proportion of cells in G0/G1 phase, accompanied by a relevant decline and promotion in S phase. While treatment with rapamycin partially recovered the decrease of S phase in CARM1 knockdown HGC27 cells, and HCQ reversed the increase of S phase in CARM1 overexpression cells to some extent. These results proved CARM1 accentuated G1-S transition via regulating autophagy.
Furthermore, apoptosis also played important role in the growth of tumor cells. More importantly, as shown in Fig. 3A, B, blockade of autophagy due to silencing CARM1 provoked ER stress and subsequently promoted the expression of CHOP and cleaved-caspase3, which were the vital proapoptotic effector (33). So we detected the apoptosis of GC cells under different treatments. As expected, the apoptosis rate enhanced in CARM1 silencing HGC27 cells, which could be counteracted by rapamycin (Fig. 5C) and in CARM1 overexpression cells presented the contrary results (Fig. 5D). Therefore, we con rmed that the inhibition of CARM1 lead to the de cient of autophagy ow, resulting in the increase of ER stress-related apoptosis.

CARM1 inhibitor attenuated the tumor-promoting effect of CARM1
Given that CARM1 could promote the GC tumor growth both in vitro and in vivo, we sought to investigate whether CARM1 inhibitor (CARM1i) EZM2302 could exert a therapeutic e cacy. As revealed in Fig. 6A, B, CARM1i suppressed the increased cell viability and clone numbers caused by CARM1 overexpression.
Concordantly, CARM1i also triggered G1 phase cell cycle arrest and signi cantly reversed the decreased apoptosis rate in CARM1 elevated cells (Fig. 6C, D). Then we tested the treatment effect of CARM1i utilizing mice xenograft models in vivo and we identi ed CARM1i had a bene cial therapeutic effect, evidenced by the slowed growth rate and eventual reduction in tumor volume, as well as the decreased tumor weight compared to the untreated group injected with CARM1 overexpression cells (Fig. 6E-G).
Interestingly, CARM1i exhibited a synergistic effect in combination with the treatment of HCQ, suggesting a potential therapeutic strategy in GC (Fig. 6E-G).

CARM1 activates autophagy by promoting TFE3 nuclear translocation
Previous studies showed that the MiTF/TFE family, including TFE3, TFEB, MITF and TFEC, played important role in various physiological processes especially in lysosomal homeostasis and autophagy regulation (34,35). As TFEC was mainly expressed in bone marrow-derived cells(36), we detected the MITF, TFEB and TFE3 expression to explore whether MiTF/TFE family was involved in the regulation of CARM1 on autophagy. As shown in Fig. 7A, MITF exhibited no difference while TFEB and TFE3 attenuated in CARM1 silencing cells and increased in CARM1 overexpression cells. CARM1 has been reported to interact with TFEB as its coactivator to promote transcription of autophagy-related genes(26), and TFE3 and TFEB could share regulatory networks (37). Therefore, this study mainly focused on whether CARM1 could promote autophagy via regulating TFE3. Western blotting of nuclear proteins and immuno uorescence staining demonstrated CARM1 could promote nuclear translocation of TFE3 (Fig.   7B, C). To illustrate the effect of TFE3 nuclear translocation on autophagy, we knocked down TFE3 expression using siRNA. TFE3 level was successfully silenced, bringing about the de ciency of autophagy and subsequent increased ER stress-mediated apoptotic protein expression (Fig. 7D).
Functional experiments further veri ed TFE3 knockdown could reverse increased cell growth caused by CARM1-induced autophagy (Fig. 7E-G).
The TFE3 activity is activated via cytoplasmic AMPK-mTOR and nuclear CARM1-TFE3 signaling pathways Then we tried to explore the mechanisms by which increased CARM1 promoted nuclear translocation of TFE3. Previous studies showed AMPK-mTOR was an important pathway involved in the regulation of MiTF/TFE family (34). Our results demonstrated in cytoplasm, p-AMPK/AMPK expression enhanced while p-mTOR/ mTOR reduced, accompanied by increased TFE3 nuclear expression in CARM1 overexpression cells (Fig. 8A), suggesting TFE3 activity was regulated by cytoplasm AMPK-mTOR signaling pathway partially. It has been reported CARM1could bind to the promoter region of TFEB to enhance transcription(26). Since TFE3 always shares regulatory mechanism with TFEB, we conducted an immunoprecipitation assay to explore whether CARM1 could also interact with TFE3, as exhibited in Fig.   8B, the overexpression of CARM1 group increased the TFE3 binding level, without the in uence of different CARM1 expression levels (Fig. 8B). To further investigate the effect of AMPK on TFE3, we applied Compound C, an effective reversible inhibitor of AMPK(38). As shown in Fig. 8A, B, Compound C suppressed cytoplasmic AMPK-mTOR pathway and reduced both TFE3 nuclear translocation and binding activity to CARM1.

Discussion
Gastric cancer remains one of the deadliest malignancies, lack of effective early diagnosis and prognostic molecular markers. CARM1 is elevated in various tumors and exhibits a tumor-promoting role mainly as transcriptional coactivators through methylating histones and non-histones (25,39).
Contradictorily, CARM1 shows tumor-inhibition effect on liver and pancreatic cancers (19,20). However, the role of CARM1 in GC and involved pathomechanism have not been studied before. Here we revealed CARM1 expression increased in databases and clinical samples of GC and proved that CARM1 was an independent risk factor for predicting poor survival for the rst time (Fig. 1A-D), indicating the potential detective and therapeutic strategy for GC.
Recent studies showed the vital role of CARM1 in autophagy (11). Autophagy maintains homeostasis through the degradation of long-life proteins and damaged organelles, which is considered to be a selfdefense pathway against certain stressful conditions (40). The role of autophagy in tumors is paradoxical depending on the speci c environment (41,42). On the one hand, autophagy markers are often located in cancer-related regions with frequent mutations or deletions as protective mechanisms to inhibit tumor initiation (43), and it has been proved that 5-FU can inhibit GC cells by inducing autophagy related death (44).On the other hand, autophagy can help tumor cells resist nutritional de ciencies and other malignant conditions to promote tumor progression and chemotherapy resistance (45). In this study, we demonstrated that CARM1 promoted GC cells proliferation through autophagy regulated G1-S transition ( Fig. 4A-H, 5A-B).
Furthermore, our results showed de cient autophagy ux caused by CARM1 inhibition contributed to ER stress related apoptosis, which affecting tumor growth too. There are three main UPR signaling pathways initiated by IRE1α, PERK, and ATF6 to lessen protein load and enhance protein-folding capacity(46), and we found the ER stress induced by CARM1 mediated autophagy was regulated mainly by PERK pathway (Fig. 3B,C). PERK undergoes dimerization and autophosphorylation, leading to activation of its kinase domain, which phosphorylates eIF2α, and p-eIF2α inhibits most protein translation to relieve the unfolded protein burden of ER(47) but selectively enhances ATF4 translation, leading to the increase of downstream protein CHOP expression(48). CHOP is responsible for the apoptosis of cells with dysfunctional ER (49). In our study, increased ER stress mediated by impaired autophagy promoted CHOP expression and eventually potentiated apoptosis (Fig. 5C).
Our results elucidated CARM1 exerted signi cant tumor-promoting role in GC, so we sought to explore the therapeutic effect of targeting CARM1 through utilizing small molecule compound EZM2302, which was a speci c inhibitor of CARM1 and had shown signi cant inhibition effect on Multiple Myeloma both in vitro and in vivo (21). In AML and diffuse large B-cell lymphoma mice model, CARM1i also exhibited effective inhibitory effect (50,51). We then wanted to know whether CARM1i could also inhibit the progression of GC, a solid tumor. As shown in Fig. 6A-D, CARM1i retarded GC cell viability, induced G1 cell cycle arrest and trigged apoptosis in vitro, and the implanted tumors in vivo were also restrained ( Fig. 6E-G). Furthermore, the autophagy inhibitor HCQ, which had been proved to possess appreciable antitumor activity in clinical trials (52,53), demonstrated synergistic antitumor effects with CARM1i, suggesting the combination of CARM1i and HCQ as a potential therapy target.
Lastly, we aimed to clarify how CARM1 regulated autophagy. The MiT/TFE family was reported to exert essential effect on autophagy regulation(54), we found TFEB and TFE3 both increased in CARM1 overexpression cells (Fig. 7A), as TFEB had been clearly demonstrated to interact with CARM1 to promote autophagy(26), we focused our study on whether CARM1 could induce autophagy by regulating TFE3.
TFE3 helps maintain lysosomal homeostasis and promotes autolysosome formation through binding CLEAR elements of lysosome and autophagy relevant genes (55). The results showed CARM1 promoted TFE3 nuclear translocation to activate autophagy, and silencing TFE3 could reverse increased cell proliferation and decreased apoptosis induced by CARM1-mediated autophagy. As reported, MTOR complex 1 (MTORC1) was the most important regulator of TFE3 by directly phosphorylating TFE3 serine residue 321, causing TFE3 cytoplasmic retention(56). The MTORC1 activity is negatively regulated by AMPK, an energy sensor, through direct and indirect phosphorylation under energy de ciency (57,58). Moreover, activated AMPK could directly phosphorylate TFE3 serine residues, resulting in TFE3 transcriptional activity (59). So we detected cytoplasmic AMPK-mTOR pathway and found activated AMPK-mTOR pathway was responsible for the nucleus translocation of TFE3 (Fig. 8A). Furthermore, CARM1 was reported to coactive TFEB in the nucleus, while TFEB and TFE3 always shared regulatory network, we then explored the interaction of CARM1 and TFE3 in the nucleus and proved CARM1 could also regulate autophagy though nucleus CARM1-TFE3 signaling pathway (Fig. 8B). Interesting, a recent study reported CARM1 activated autophagy as an enhancer through CARM1-Pontin-FOXO3a signaling axis, different from the direct interaction with TFEB or TFE3 (59). CARM1 appears to regulate autophagy through several functionally redundant transcription factors, suggesting the key role of CARM1 in autophagy regulation.
In conclusion, the present study rst revealed CARM1 was upregulated in clinical GC tissues and cell lines, and higher CARM1 expression was related to an unfavorable prognosis. More importantly, our study   5μm) and the images below indicated higher magni cation (scale bar: 1μm). D Representative photographs of immuno uorescence staining presented LC3B positive cells (scale bar: 50μm). E Protein levels of p62, LC3BII/I, ATG5, Beclin1 were detected by western blotting using β-Actin as an internal control. (n = 3; error bar, SD). *Represents *P < 0.05, **P < 0.01 and ***P < 0.001.

Figure 3
Blocked autophagy ux induces endoplasmic reticulum(ER) stress. A ER stress related proteins GRP78, p-PERK, PERK, p-eif2α, eif2α, ATF4 and downstream effector proteins associated with apoptosis were detected by western blotting in CARM1 downregulation HGC27 cells and CARM1 upregulation BGC823 cells. B HGC27 CARM1-knockdown and control cells were treated with rapamycin (50nM, 24h) to detect the effect of autophagy induction on ER stress. C BGC823 overexpression and control cells were treated with HCQ (25μM, 24h) to evaluate the role of autophagy ux blockage in ER stress. β-Actin was applied as loading controls. Each experiment was repeated three times. Data represent mean values ± SD.
*Symbolizes *P < 0.05, **P < 0.01 and ***P < 0.001. Figure 4 CARM1 promotes the proliferation of GC cells through regulation of autophagy. A CCK-8 assay revealed that downregulation of CARM1 signi cantly suppressed the growth rate in HGC27 cells. B HGC27 CARM1-knockdown and control cells were treated with rapamycin (50nM, 24h). Then cell viability was detected by CCK-8 assay. C CCK-8 assay proved that upregulation of CARM1 considerably enhanced cell viability in BGC823 cells. D BGC823 overexpression and control cells were treated with HCQ (25μM, 24h). faster than that injected of control gastric cells. Data are presented as mean ± SD. *Represents *P < 0.05, **P < 0.01 and ***P < 0.001. HGC27 cells transfected with si-CARM1 were treated with rapamycin (50nM, 24h), BGC823 cells transduced with overexpression and control lentivirus were treated with HCQ (25μM, 24h) before the detection of ow cytometry. A and B Distribution of different cell cycle phases was presented in indicated cells by ow cytometry. CARM1 overexpression promoted while CARM1 downregulation attenuated G1-to-S transition of GC cells. However, the effects could be reversed by autophagy activator rapamycin and autophagy inhibitor HCQ respectively. C and D Representative images showed the percentage of cells undergoing apoptosis from three dependent experiments. Treatment with rapamycin partially reversed the increase of apoptosis in CARM1 knockdown HGC27 cells (left panel), and HCQ recovered the decrease of apoptosis in CARM1 overexpression cells to some extent (right panel). Bars represent mean ± SD from three independent experiments. *Represents *P < 0.05, **P < 0.01 and ***P < 0.001. Figure 6 CARM1 inhibitor rescued the tumor-promoting role of CARM1 both in vitro and in vivo. A BGC823 control and upregulation cells were treated with CARM1i (8μM, 24h), the cell viability was determined by CCK-8 assay. B BGC823 control and upregulation cells were cultivated in medium containing 8μM CARM1i for 10-14 days, CARM1i reduced the increase of clone numbers caused by CARM1 overexpression. C and D BGC823 control and CARM1-overexpression cells were treated with CARM1i (8μM, 24h). Cell cycle phase distribution and apoptosis in indicated were assessed by ow cytometry. E, F and G BGC823 xenograft model was performed to investigate the therapeutic effect of CARM1i and HCQ on GC in vivo. Nude mice injected with CARM1 overexpression cells were randomly divided into four groups at day 6 when tumor volumes reached 50 mm3. CARM1i was performed twice daily at 100 mg/kg i.p. HCQ was administered once daily at 50 mg/kg i.p. The combination group was given two treatments and the control group was  expression in BGC823 overexpression cells blocked autophagy ux and subsequently induced ER stress. Autophagy and ER stress related protein expression was detected in the cytoplasm and TFE3 expression was proved in nucleus protein by western blotting. β-Actin and H3 served as loading controls for cytoplasmic proteins and nuclear proteins respectively. E and F Colony formation and CCK-8 assays revealed TFE3 downregulation reversed the proproliferative effect of CARM1 on BGC823 cells. G TFE3 knockdown rescued increased G1-S transition and decreased apoptosis resulted from CARM1 overexpression in BGC823 cells. Bars represent mean ± SD from three independent experiments.

Figure 8
The TFE3 activity is activated via cytoplasmic AMPK-mTOR and nuclear CARM1-TFE3 signaling pathways. A Western blotting analysis of AMPK-mTOR signal pathway in the cytoplasm and TFE3 expression in the nucleus in CARM1-EV, CARM1-OE, CARM1-OE+DMSO, CARM1-OE+Compound C (10μM, 6h) groups. B BGC823 overexpression cells were treated with Compound C (10μM, 6h) or solvent comparison DMSO. Then immunoprecipitation assay using CARM1 antibody was performed and western