A172 glioma cell line was purchased from the Cell Bank of Chinese Academy of Sciences (Shanghai, China). A172 cells were cultured in DMEM medium (Gibco, New York, NY, USA) supplemented with 10% fatal bovine serun (FBS) (Invitrogen, Carlsbad, CA, USA), 100 U/mL penicillin and 100 µg/mL streptomycin (Gibco) at 37°C in a humidified incubator (Thermo-Fisher, Waltham, MA, USA) with 5% CO₂. All experiments were conducted with cells passage less than eight times. DNA fingerprinting and profiling were performed every 5 months to verify the cell line phenotype. Human astrocyte culture was performed according to previous studies [11,21].

A total of five patients with glioma were recruited in present study (male, World Health Organization grade III, age 54–68) at Minhang Hospital. Right after surgery, tumor tissues and surrounding normal brain tissues were collected, and separated carefully under microscope. Glioma was defined by pathological tests with both fresh frozen section tissue and formalin-fixed paraffin-embedded (anaplastic astrocytoma). Tissues were carefully washed with pre-cold PBS, and then minced for homogenization with tissue lysis buffer (Invitrogen). Written informed consent was obtained from each participant.

Methyl thiazolyl tetrazolium (MTT) assays were performed for cell viability assessment. After different treatments, 0.5 mg/mL MTT reagent was added into the culture medium and then incubated for 4 hours at 37°C. Purple formazan salt crystals were dissolved by adding the solubilization solution with 10% sodium dodecyl sulfate (SDS) and 0.01 M HCl. The absorption at 490 nm was measured using a multi-well plate reader.

For cell death assessment, trypan blue staining approach was applied. Attached cells were digested with trypsin/ethylene diamine tetraacetic acid (EDTA), suspended in 1X PBS, and mixed with 0.4% trypan blue dye (Sigma, St. Louis, MO, USA). Dead cells would take up trypan blue due to compromised cell membranes, while viable cells would not due to the membrane integrity. The percentages of dead cell were recorded for further assessments. Cell Apoptosis DNA ELISA Detection Kit (Roche, San Francisco, CA, USA) was used for apoptosis assessment according to manufacturer’s instructions.

To evaluate cell proliferation, colony formation assays were performed. A172 cells were harvested with trypsin/EDTA and re-suspended in pre-cold PBS before being seeded in a 6-well plate at a density of 500 cells per well with completed DMEM medium (10% FBS). The cells were cultured for 21 days before the cell colonies were stained with crystal violet solution (Sigma). Number of viable colonies was recorded.

Control lentivirus pLKO.1 vector and pLKO.1 vectors containing shRNA for EID3 (RHS4531-EG493861) were purchased from Thermo Fisher Scientific (Waltham, MA, USA). Glioma cells were infected with viral particles twice for the establishment of stable EID3 knocking down clones. Stable transfected cells were selected with puromycin (2 µg/mL). Knocking down efficiency were confirmed by real-time polymerase chain reaction (RT-PCR) and Western blot (WB) assays.

Two lentiviral AMPKα1 shRNAs were purchased from Santa Cruz Biotech (shAMPKα1-1; Santa Cruz, CA, USA) and GeneChem (shAMPKα1-2; Montreal, Canada). Each shRNA was added into A172 cells separately for 24 hours. The cell culture medium was then replaced by fresh medium for an additional 24 hours. Stable clones expressing AMPKα shRNA were selected by puromycin (0.5 µg/mL; Sigma) for a period of 10 days. Control cells were infected with lentiviral scramble shRNA (Santa Cruz Biotech). Knocking down efficiencies were evaluated by RT-PCR and WB assays.

The dominant negative AMPKα (DN-AMPKα, T172A) construct was designed as mentioned before. Glioma cells were cultured in a medium without antibiotics when reached 60–70% confluence. Then DN-AMPKα cDNA (0.10 µg/mL) was transfected into the glioma cells followed by Lipofectamine 2000 protocol. Afterwards, stable cells were selected via neomycin (1.0 µg/mL; Sigma). Transfection efficiency was verified via WB.

Total RNA was extracted from cells using RNeasy Mini Kit (Qiagen, Dusseldorf, Germany), and reversed transcript using Quantitect Reverse Transcription Kit (Qiagen) according to the manufacturer’s instructions. The RT-PCR mixtures contained 2× QuantiTect SYBR Green PCR Master Mix (Qiagen), 10× QuantiTect primer assay mix, and synthesized cDNA. Amplification was performed in triplicate for each sample on a Lightcyle 480 platform (Roche, Mannheim, Germany). EID3 protein expression levels were normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) levels. Primer sequences were shown as follows : GADPH, forward (F), TGCACCACCAACTGCTTAGC; reverse (R), GGCATGGACTGTGGTCATGAG; EID3, F, GGAGGCACCTCAGTCATGATA; R, GTCCACAAACAGTTCTCTGAC.

Cells were lysed in RIPA lysis buffer (Beyotime, Shanghai, China) with PMSF (Roche, Mannheim, Germany). Supernatant was collected after centrifugation at 13000 ×g at 4°C for 10 minutes. Protein concentration was measured with a BCA protein kit (Beyotime) and the whole lysates were mixed with 4× SDS loading buffer (125 mmol/Tris-HCl, 4% SDS, 20% glycerol, 100 mmol/L DTT, and 0.2% bromophenol blue). Protein samples were heated to 100°C for 5 minutes. For the WB, 30 µg total protein samples were loaded to a SDS-polyacrylamide gel. After electrophoresis, the protein was transferred to a polyvinylidene fluoride membrane blocked in 5% fat-free milk and incubated with antibodies against EID3 (1 : 800; Abcam, Burlingame, CA, USA), AMPKα1 (1 : 1000; CST, Chicago, IL, USA), pS6K1 (1 : 1000; Abcam), S6K1 (1 : 750; Abcam), LC3B-I (1 : 1000; Abcam), LC3B-II (1 : 800; Abcam), Becllin (1 : 750; Pharmingen, San Diego, CA, USA), ATG-5 (1 : 800; Abcam), ATG-7 (1 : 1000; Abcam), p62 (1 : 1000; Abcam), pAMPKα1 (1 : 1000; Abcam), ACC (1 : 1000; Abcam), pACC (1 : 1000; Abcam), and Tubulin (1 : 1000, Proteintech, Wuhan, China) overnight at 4°C, followed by another 2 hours incubation with horseradish peroxidase-conjugated anti-rabbit secondary antibody (1 : 5000; Jackson ImmunoResearch Labs, West Grove, PA, USA). GAPDH was served as a loading control. NIH Image J software (Bethesda, MD, USA) was used for the densitometric analysis of the WB. The protein levels were normalized with GAPDH.

Specific pathogen-free BALB/c nude mice (4–6 weeks old) purchased from the Chinese Academy of Medical Science were injected with 5×10⁶ glioma cells subcutaneously into the left upper flank regions. The subcutaneous tumor size was calculated and recorded every week with vernier caliper as follows : tumor volume (mm³)=(L×W²)/2, where L=long axis and W=short axis, the measurements were repeated three times.

Glioma patient-derived xenograft (PDX) models were generated according to a previous study [17]. Briefly, 4 weeks old male non-obese diabetic severe combined immune deficiency (NOD/SCID) mice were used as transplant recipient models and raised in the aseptic environment. Fresh glioma samples were cut into 1mm pieces within 1 hour after removal from patients. Tissue fragments were incubated in DMEM medium mixed with 50% Matrigel™ (BD, Franklin Lakes, NJ, USA), 10 ng/mL epidermal growth factor (Gibco), 10 ng/mL basic fibroblast growth factor (Gibco), 100 U/mL penicillin, and 100 U/mL streptomycin for 30 minutes. The whole tumor tissue mixture was then transplanted into the right flanks of mice (n=3; 4–5 weeks old; Shanghai Institute of Material Medicine, Chinese Academy of Science) subcutaneously with a #20 trocar. Animal welfare and experimental protocols were approved by the Shanghai Medical Experimental Animal Care Commission. Primary EID3 expression levels were detected by RT-PCR and WB prior to tissue transplantation. Single glioma cell suspension was made according to a previous study [6]. All the samples were mechanically disaggregated and digested with type IV collagenase (Gibco) within one hour, and resuspended in DMEM medium. The treated tissue was filtrated through a 40 µm filter to obtain the single-cell suspension. Red blood cells were lysed with ACK buffer (Invitrogen). The number of viable cells was counted and analyzed by trypan blue analysis (Procell, Wuhan, China).

Statistical analyses were performed with SPSS ver. 20.0 for Mac (IBM, Armonk, NY, USA). Quantitative data presented as mean±standard error are from at least three independent experiments. Continuous data were analyzed by one-way analysis of variance and Student’s t-test, and categorical data were analyzed by Fisher’s exact test or chi-square test. A p-value <0.05 was considered statistically significant.

EID3 expression level in glioma cells was evaluated by quantitative PCR and WB. Results of RT-PCR assays showed that EID3 only expressed in glioma cells, and undetectable in human astrocyte and neuronal cells (HCN-1a cell line) (Fig. 1A). WB assays further validated the findings of RT-PCR (Fig. 1B). Since AMPK was identified as a key suppressor in glioma, we next investigate the correlation between EID3 and AMPKα1, an active form of AMPK. The WB results showed that EID3 expression was correlated with AMPKα1 downregulation (Fig. 1B). AMPK inhibition could result in mTORC1 activation process, during which phosphorylation of S6K1 was considered as a hallmark. Hence, we further explore the S6K1 activation status. Our data showed that the S6K1 phosphorylation was increased in AMPKα1-low A172 cells. However, the pS6K1 level was low in EID3-null astrocytes and HCN-1a cells which with a high AMPKα1 expression (Fig. 1B).

EID3 and AMPKα1 expression status were also evaluated in human glioma tissues. As shown in Fig. 1C, EID3 expression could be detected in four of five patients (80%) enrolled, however, it could not be detected in surrounding normal brain tissues. Also, EID3 expression was correlated with downregulation of AMPKα1, which consistent with the results in A172 cells.

EID3 expression was knocking down with two shRNAs targeting different sites of EID3 cDNA, the efficiencies of knocking down were validated through RT-PCR and WB (Fig. 2A). Remarkably, EID3 knocking down could restore AMPKα1 expression, resulting in inhibition of S6K1 (Fig. 2A). Colony formation assays demonstrated that EID3 knocking down significantly inhibited A172 cell proliferation (p<0.05, Fig. 2B). Meanwhile, Transwell assays indicated that the number of migrated and invaded cells was significantly decreased when EID3 expression is knocked down (Fig. 2C and Supplementary Fig. 1). Based on the MTT results, we found the number of viable cells was significantly decreased after EID3 knocking down. We also found that EID3 shRNA could induce cell death according to trypan blue staining. Additionally, Histone DNA apoptosis ELISA assay revealed EID3 shRNA could induce apoptosis in glioma cells (Fig. 2D), while the scramble non-sense control shRNA exhibited no effects on glioma cell proliferation, death, and apoptosis.

According to a previous report [22], AMPK activation would trigger autophagy by activating Ulk1 or inhibiting mTOR signaling. Since our results indicated that EID3 could restore mTOR activation by downregulating AMPKα1, we further investigated the effects of EID3 on autophagy in glioma cells. Our data indicated that EID3 shRNA could induce increased Beclin-1, ATG-5, and ATG-7 expression, LC3B-I to LC3B-II switch, and p62 degradation, which confirmed that EID3 knocking down glioma cells exhibited autophagy-like changes (Fig. 2E).

We also validated in vivo proliferation inhibition of EID3. A172 cells that stable express EID3 shRNAs and parental control A172 cells were injected into the flanks of SCID mice, and tumor volumes and weights were recorded. As Fig. 2F demonstrated, tumors generated from EID3 knocking down A172 cells grew significantly slower than control tumors (sh-C). Moreover, weights of tumor generated from EID3 knocking down A172 cells were significantly lighter than those of control tumors.

Our results indicated that EID3 knocking down could restore AMPKα1 expression and inhibit tumor growth and invasion. We further explored whether AMPKα1 served as the key inhibitor that mediated the inhibition effects following EID3 knocking down. For this purpose, AMPKα1 was knocked down in A172 cells expressing shEID3 (Fig. 3A). We found that AMPKα1 inhibition could restore the activation of S6K1 in EID3 knocking down cells (Fig. 3A). However, AMPKα1 knocking down did not affect expression status of EID3. Moreover, cell viability reduction and cell apoptosis caused by EID3 knocking down could be largely attenuated by AMPKα1 knocking down inhibition (Fig. 3B).

To further study the role of AMPKα1, a dominant negative AMPKα1 mutant (DN-AMPKα1, T172A) were introduced into A172 cells, aiming to block AMPKα1 activation following EID3 knocking down. WB assays confirmed exogenous expression of the DN-AMPKα1 in A172 cells (Fig. 3C). As expected, EID3 knocking down inhibition greatly activated AMPKα1, which can be proven by increased expression levels of p-AMPKα1 and p-ACC (Fig. 3D). However, such effects were significantly hindered by DN-AMPKα1 (Fig. 3D). Consistently, forced expression of DN-AMPKα1 did not rescue the inhibition effects of EID3 knocking down on cell survival (Fig. 3E) and death (Fig. 3F). Meanwhile, all control shRNAs showed no significant effects on A172 cells. Through RT-PCR assays, we found that neither EID3 shRNA could decrease AMPKα1 mRNA expression (Fig. 3G), which suggested EID3 regulates AMPKα1 expression through a non-transcriptional way. Therefore, MG132, a proteasome inhibitor, was introduced to test our hypothesis. WB assays showed that MG132 restored AMPKα1 expression in A172 cells with high EID3 expression (Fig. 3H). Also, MG132 reactivated the mTOR signaling, supported by increased pS6K1 level after treatment (Fig. 3H). However, MG132 showed no significant influence on the mRNA expression of AMPKα1 (Fig. 3I). To exclude the possibility that EID3 inhibits the protein synthesis process of AMPKα1, cycloheeximide (CHX), an inhibitor of protein synthesis was used. Results showed that CHX treatment could not affect the expression level of AMPKα1 in EID3 knocking down A172 cells (Fig. 3J), which indicated that increased expression of AMPKα1 in EID3 knocking down cells might not due to promoted protein synthesis. Collectively, these data implied that EID3 might induce ubiquitination and proteasomal degradation of AMPKα1.

To validate the function of EID3 in primary glioma, we generated three patient-derived xenograft model with descending EID3 expression (T1, high; T2, moderate; T3, low). The EID3 expression status were confirmed by WB assays (Fig. 4A). Accordingly, primary glioma with high EID3 expression showed lower level of AMPKα1 but high pS6K1 expression (T1), while tumor with undetectable EID3 exhibited high level of AMPKα1 and low level pS6K1 (T3, Fig. 4A). We observed that T1 (high EID3) grew significantly faster than T2 and T3, based on bigger tumor volumes (Fig. 4B). Moreover, T2 also showed significantly faster growth than T3 did (Fig. 4B).

We then separated single glioma cells from T1 tumor tissue and transfected it with EID3 shRNA to explore the effects of EID3 knocking down in primary glioma. WB assays confirmed the efficiencies of EID3 knocking down and demonstrated that EID3 knocking down resulted in upregulation of AMPKα1 and inactivation of S6K1 (Fig. 4C). Furthermore, we also observed an autophagy-like phenotype in primary glioma cells with EID3 knocking down, which were similar to those findings in A172 cells (Fig. 4D). To further validate the role of AMPKα1, primary T1 glioma cells were transfected with AMPKα1 shRNA. Notably, AMPKα1 knocking down could effectively restore the pS6K1 level in primary glioma cells transfected with EID3 shRNAs (Fig. 4E). Moreover, AMPKα1 knocking down largely attenuated the viability reduction and cell death caused by EID3 inhibition (Fig. 4F). These results in primary glioma cells further confirmed that AMPK mediated viability reduction and cell death by EID3 knocking down.