β-Elemene Inhibits Proliferation of Non-Small-Cell Lung Cancer by Targeting ALDH3A1 to Regulate Metabolic Reprogramming


 Background

Non-small cell lung cancer (NSCLC) is the most common cause of cancer-related death in the world. Although the improvement of treatment has significantly prolonged the survival of NSCLC patients, most patients still face recurrence because of drug resistance. Therefore, it is urgent to find more effective drugs to improve the NSCLC treatment and survival rate of patients.
Methods

Network pharmacology tools were comprehensive applied to analyze β-elemene and the anticancer effects of β-elemene were evaluated both in vitro and in vivo. What’s more, the underlying mechanisms were investigated via metabolomics analysis, 18FDG-micro-PET scan and molecular biology experiment.
Results

β-elemene exhibited potent inhibitory effects on NSCLC cells proliferation by alternating the cell cycle without significantly apoptosis in NSCLC cells. Mechanistically, we found that β-elemene regulating metabolic reprogramming of NSCLC via targeting Aldehyde Dehydrogenase 3 Family Member A1 (ALDH3A1). Subsequently, the metabolic reprogramming of NSCLC cells regulated by β-elemene lead to increased hydroxylation of α-ketoglutarate (α-KG), and degraded Hypoxia Inducible Factor1 Subunit Alpha (HIF1α), which cause accelerated reduction of aerobic glycolysis in NSCLC in vivo and vitro through facilitating HIF1α/LDHA pathway. Moreover, the reduction of aerobic glycolysis effect of β-elemene in vivo was confirmed by the reduced SUV values and the decreased expression of Ki-67 and HIF1α/LDHA pathway proteins in tumor tissues from xenograft mouse models.
Conclusion

β-Elemene exhibits strong metabolic reprogramming effects, glycolysis was suppressed while OXPHOS was increased in NSCLC cells via targeting ALDH3A1 and HIF1α/LDHA pathway, which provides a new theoretical basis for the clinical treatment of patients with NSCLC.


Results
β-elemene exhibited potent inhibitory effects on NSCLC cells proliferation by alternating the cell cycle without signi cantly apoptosis in NSCLC cells. Mechanistically, we found that β-elemene regulating metabolic reprogramming of NSCLC via targeting Aldehyde Dehydrogenase 3 Family Member A1 (ALDH3A1). Subsequently, the metabolic reprogramming of NSCLC cells regulated by β-elemene lead to increased hydroxylation of α-ketoglutarate (α-KG), and degraded Hypoxia Inducible Factor1 Subunit Alpha (HIF1α), which cause accelerated reduction of aerobic glycolysis in NSCLC in vivo and vitro through facilitating HIF1α/LDHA pathway. Moreover, the reduction of aerobic glycolysis effect of β-elemene in vivo was con rmed by the reduced SUV values and the decreased expression of Ki-67 and HIF1α/LDHA pathway proteins in tumor tissues from xenograft mouse models. Conclusion β-Elemene exhibits strong metabolic reprogramming effects, glycolysis was suppressed while OXPHOS was increased in NSCLC cells via targeting ALDH3A1 and HIF1α/LDHA pathway, which provides a new theoretical basis for the clinical treatment of patients with NSCLC.

Background
NSCLC is the most frequent cause of cancer-related deaths worldwide [1]. Because the implementation of high-quality lung cancer screening still faces many challenges, 79% of lung cancers still are diagnosed as regional or distant disease [2]. There are a variety of treatments, including target therapy and immunotherapy which signi cantly improve the ve-year survival rate of patients [3], most of them still face treatment failure [4,5]. Therefore, it is urgent to explore and optimize the treatment to gradually improve the poor 5-year survival rate of patients.
Traditional Chinese Medicine has long been indicated can suppress tumor progression [6]. The most recent study indicated that TCMs were more and more widely used in clinical cancer therapy [7]. βelemene is a monomer compound extracted from the traditional Chinese medicine herb Curcuma wenyujin and has shown promising anticancer activity in a broad spectrum of tumors [8]. Mounting studies have reported that β-elemene can promote cell apoptosis in brain tumor and reverse the multidrug resistance in gastric adenocarcinoma cells [9,10]; β-elemene also can inhibit breast cancer metastasis by inhibiting pyruvate kinase M2 (PKM2) mediated glycosysis [11,12]. In addition, β-elemene has also been reported to be able to inhibit the macrophages polarization from M2 to M1 and suppressed the development of lung cancer cells [13]. Therefore, β-elemene plays a crucial role in anti-tumor therapy.
However, the hub target of β-elemene in cancer therapy is still largely unknown. In recent years, there has been an increasing interest in bioinformatics with the increased availability of genomic resources, and network pharmacology has already been widely used in targets discovery, especially in Chinese medicine (CM) formula [14,15]. Therefore, network pharmacology and bioinformatics analysis will help to identify the targets of β-elemene and clarify its anti-tumor mechanism [16].
In this study, we identi ed ALDH3A1 was the hub target of β-elemene for the rst time. We found that βelemene could signi cantly suppress the proliferation of NSCLC cells in vivo and in vitro by inhibiting the expression of ALDH3A1, and the following HIF1α/LDHA pathway mediated glycolysis. As aerobic glycolysis plays a pivotal role in tumor development and drug resistance, while β-elemene has long been a great choice to suppress tumor progression after standard treatment fails. Overall, the study proposes that β-elemene is a promising drug used in combination with chemotherapy or other therapies in the treatment of early lung cancer.

Cell culture
NSCLC cell lines A549 and H1299 were purchased from the Academy of Military Medical Science (Beijing, China). A549 was cultured in DMEM/F-12 medium (Gibco, Carlsbad, CA), and the others were cultured in RPMI-1640 medium (Gibco, Carlsbad, CA) containing with 10% fetal bovine serum (FBS), penicillin (100U/mL) and streptomycin (100 mg/mL) at 37℃, in a humidi ed incubator with 5% CO 2 . Cells were routinely passed every 2-3 days and all cells maintained in culture for a maximum 8 weeks.

Cell proliferation assay
The proliferation assays in selected NSCLC cell lines were measured using a 3-(4,5Dimethylthiazolyl-2yl)-2,5-diphenyltetrazolium bromide (MTT) assay. Firstly, A549 and H1299 cells were seeded at 3×10 3 cells/well in 96-well plates while HCC827 and PC9 cells were seeded at 4×10 3 cells/well in 96-well plates. After incubation for 24 h, the cells were treated with various concentrations of β-elemene for 24 or 48 h, Twenty microliter of MTT solution (5 mg/ml) was added to each well followed by 4 h incubation at 37 °C.
The cell culture medium was removed and the cells were lysed in 200 µl dimethylsulphoxide (DMSO) and the results were measured using a microplate reader (Model 570, Bio-Rad Laboratories, Hercules, CA, USA).
We apply the method described in ref to detect cell proliferation. Brie y, the cells transfected with ALDH3A1-speci c siRNAs (Si1 and Si2) and negative control siRNA (NC) for 4 days. The cell growth was determined using the CellTiter 96 Aqueous One Cell Proliferation Assay (Promega, Madison, WI). Percentage of net growth at day 4 (100%) relative to day 0 (0%) was calculated and the concentration of compound required to inhibit growth by 50% (GI50) determined. The assays were done in triplicate across different plates.

Colony-forming assay
A549 and H1299 cells were plated at 300 cells into each well of 12-well plates in the medium containing 10% FBS. Cells were incubated at 37℃ in 5% CO 2 for 3 days, after cell proliferation reaches the logarithmic growth phase, then cells were treated with multiple concentrations of β-elemene (20ug/ml and 30ug/ml). After 14 days, cells were stained by Wright-Giemsa. Finally, the number of colonies was counted by light microscopy.

The Edu assay
Selected NSCLC cells were seeded into 96-well plates (3000 cells per well). After incubation for 24 h, the cells were pretreated with β-elemene for 24 h or 48 h. Then 50 µM of 5-ethynyl2′-deoxyuridine (EdU, Ribobio, Guangzhou, China) was added to each well and the cells were incubated at 37 °C for 2 h before being xed with 4% formaldehyde for 30 min and incubated with 2 mg/ml glycine for 5 min. After being washed with PBS for 3 times, the cells were reacted with 100µL of 1×Apollo reaction cocktail for 30 min.
Afterwards, the 1×Hoechst 33342 (5 µg/mL) was used to stain the nuclei. Five representative elds of each insert were randomly counted under a microscope (Olympus, Tokyo, Japan) and analyzed statistically.

Flow cytometry analysis
The selected cells were cultured in 6-well plates and treated with or without different doses of β-elemene and then further incubated for 12 h or 24 h. Before the test, cells were incubated with 10µL phosphatidylinositol and 5µL Annexin V for 20 min protected from light for cell apoptosis. For cell cycle analysis, the cells were suspended in 70% ethanol, incubated at -20 °C for 2 h, and collected by centrifugation at 1500 rpm for 3 min. After removal of the ethanol, the cells were re-suspended in 300 µl PI staining solution and incubated for 30 min in dark. Then, the samples were evaluated by ow cytometry, and nally analyzed by Cell Quest software (USA) and mod t software.

Network pharmacology analysis
BATMAN-TCM (http://bionet.ncpsb.org/BATMAN-TCM) is the rst tool for bioinformatics analysis for studying the molecular mechanism of TCM's by predicting potential targets for ingredients. The PubChem_CID / InChl of β-elemene (6918391) were entered as search queries. The following parameter settings were used: Target Prediction's Score cutoff was 20; Target Analysis's Adjusted P-value is less than 0.05. The PPI pairs of all the targets of β-elemene were analyzed using the online database STRING version 10.5 (http://string-db.org/). The pairs with combined scores > 0.4 were used for the PPI network construction, then the Cytoscape v.3.6.0 software was used to construct the network and analyze the interaction relationship of the proteins. Pathway enrichment analysis was performed using the clusterPro ler package [17].

Intracellular glucose uptake assay
First of all, A549 and H1299 cells were seeded into 96-well plates (3000 cells per well), after 24 h, treated with β-elemene, then incubate for another 24 h. Lastly, cultured in glucose-free medium overnight. After removal of the culture medium, the levels of intracellular glucose were assessed using a uorescencebased glucose assay kit (BioVision) according to manufacturer's instructions.

Lactate Production Assay
Selected NSCLC cells were plated at 1×10 5 cells into each well of 6-well plates in the medium containing 10% FBS. Treat with different doses of β-elemene after 24 h and incubate for 4 hours. Then, the levels of lactate in the medium were quantitated with a lactate assay kit that is based on uorescence activity (BioVision) according to manufacturer's instructions.

NADP+/NADPH, NAD + /NADH Ratio and α-KG Assay
Selected NSCLC cells were plated at 1×10 5 cells into each well of 6-well plates in the medium containing 10% FBS. Treat with different doses of β-elemene after 24 h and incubate for 4 hours. Then, NADP+/NADPH, NAD+/NADH Ratio or α-KG levels in the cell lysates were determined using a uorescence-based lactate assay kit (AAT Bioques for NADP+/NADPH, AAT BioquesforNAD + /NADH Ratio, BioVision for α-KG) according to the manufacturers' instructions.

Intracellular PH assay
Dissolve Nigericin in 345 µL anhydrous DMSO to make a 20 mM solution. At the same time, dissolve Valinomycin in 225 µL anhydrous DMSO to make a 20 mM solution. Then, combine 100 µL of Nigericin and 100 µL of Valinomycin solutions to make a 1000× stock solution (10 mM each). Dilute 10 µL of this solution into 10 mL of the desired Intracellular pH Calibration Buffer to make a cell loading solution. Perform desired cellular experiment with pH-sensitive pHrodo™ dyes or conjugates. Washing cells 2× with Live Cell Imaging Solution (LCIS) (Cat. no. A14291DJ), replacing LCIS with the cell loading solution, and incubate at 37 °C for at least 5 minutes. Analyze the cells using the appropriate Ex/Em maxima.
2.12 ALDH3A1 enzymatic assay ALDH3A1 was measured as previously described using 5 µg/ml of recombinant protein. Brie y, enzymatic activity was measured spectrophotometrically by the reduction of NAD + to NADH at A340 over 5 min in the presence of increasing concentrations of β-elemene or DMSO vehicle control. Assays were conducted in 50 mM sodium pyrophosphate buffer (pH 7.4) in the presence of 2.5 mM NAD + and 10 mM substrate and measured at 25 °C. Dose-response curve ts and EC50's were calculated using GraphPad Prism 7 software.

RNA and Lentivirus interference
ALDH3A1-speci c siRNAs (Si1 and Si2) and negative control siRNA (NC) were prepared by ViewSolidBiotech (Beijing, China). The coding strand of human ALDH3A1Si1 was 5′-GGAACUCAGUGGUCCUCAATT-3′, and that of ALDH3A1Si2 was 5′-UUGAGGACCACUGAGUUCCTT-3′. siRNAs were transfected into cells using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol.Lentivirus construction of ALDH3A1 knockdown carrying green uorescent protein (GFP) was designed and provided by Obio Technology Corp., Ltd. (Shanghai, China). A549 cells were transfected with the lentiviral vector according to the manufacturer's instructions. Control cells were transfected with empty vector carrying GFP. The ALDH3A1 knockdown cells were named Sh-ALDH3A1, and empty vector cells, named as Sh-NC, were used as control.

Western blot analysis
For western-blotting in A549 and H1299 cell lines, the detailed method was described in ref. The cells were collected after treatment for indicated time and washed twice with cold PBS. Then cells were lysed in 1% Triton lysis buffer (50 mmol/L Tris-HCl, pH 7.4, 10 mmol/L EDTA, 100 mmol/L NaF, 150 mmol/L NaCl, 1% Triton X-100, 1 mmol/L PMSF 1 mmol/L Na3VO4 and 2 µg/mL aprotinin). After sonication and centrifugation, samples were quanti ed with the Lowry method. Next, all the protein samples were boiled at 95 °C for 5 min with 3× sampling buffer. After that, all the lysate samples were resolved by SDS polyacrylamide gel electrophoresis (SDS-PAGE) and followed by western blotting which means electronically transferred to polyvinylidene uoride (PVDF) membranes. After blocking with 5% skim milk in TBST buffer (10 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.1% Tween 20) at room temperature for 1 h, the blots were incubated with indicated antibodies shaking for 2 h, then overnight at 4 °C. Then the blots were washed four times with TBST buffer, and then incubated with secondary antibodies for 30 min at room temperature. After washing for another four times, the protein bands were detected with enhanced chemiluminescence reagent (SuperSignal Western Pico Chemiluminescent Substrate; Pierce, Rockford, IL, USA) and visualised with the Electrophoresis Gel Imaging Analysis System (DNR Bio-Imaging Systems, Jerusalem, Israel).

Micro-PET scan
Firstly, the model mice were anesthetized by abdominal cavity. The volume of 18 FDG was not more than 0.1 ml. The activity, activity and measurement time were measured. 18 FDG was injected into tail vein. The injection time was recorded. The residual activity of syringe was measured. The residual activity and measurement time were recorded. After 30 minutes, the 18 FDG intake by mice was full, the model animals were transferred to the scanning bed and adjusted. The whole posture stretches its limbs, con gures the micro-PET scanning (Metis-6) and related control software (Metis Console) and scanning.

13 C metabolite labeling and quanti cation of energy metabolites
To extracting metabolites, rst of all, add 200µL of pre-cooled ultrapure water to 60 mg of each mouse tumor tissue perform MP homogenization, followed by 800µL of pre-cooled methanol/acetonitrile(1:1, v/v).Samples were separated using an Agilent 1290 In nity LC Ultra High Performance Liquid Chromatography System. Mass spectrometry in negative ion mode using a 5500 QTRAP mass spectrometer (AB SCIEX) analysis. Finally, the peak area and retention time were extracted using Multi quant software. The quality standard corrects the retention time and performs metabolite identi cation.

Immunohistochemistry
Tumors were formalin-xed, para n-embedded and prepared as described in our previous study for staining with haematoxylin and eosin. The immunohistochemical antibodies ALDH3A1, HIF-1α, GLUT1, LDHA, Actin and Ki67 have been described already. The staining was evaluated by scanning the entire tissue specimen under low magni cation (×10) and con rmed under high magni cation (×20 and ×40). The protein expression was visualised and classi ed based on the percentage of positive cells and the intensity of staining. From each section, ve visual elds were randomly selected. The degree of protein expression was based on the percentage of positive cells and the intensity of staining. Staining intensity was scored as 0 (no staining), 1 (low staining), 2 (intermediate staining), and 3 (high staining). For the staining area, ≤ 5%, 5-25%, 26-50, 51-75% >75% were recorded as 0, 1, 2, 3 and 4 points, respectively. Histological score = staining intensity× staining area. A score 0 was classi ed as negative (−), 1-4 points as weakly positive (+), 6-12 points as a strong positive (++). Final scores were assigned by two independent pathologists.

Statistical analysis
All the presented data were veri ed by three separate experiments, and are expressed as the means ± standard deviation (SD). Differences between groups were calculated by Student's two-tailed t-test. All analyses were calculated using GraphPad Prism software or R. P < 0.01 and P < 0.05 were considered statistically signi cant.

β-Elemene inhibited NSCLC proliferation by alternating the cell cycle
We identi ed the inhibition effect of β-elemene in NSCLC cells proliferation, and the results revealed that β-elemene inhibited the proliferation of the A549 and H1299 NSCLC cells signi cantly, the IC 50 values of β-elemene in A549 cells was 27.4 ± 0.60 ug/ml and 22.42 ± 0.89 ug/ml, and in H1299 cells was 27.47 ± 1.71 ug/ml and 23.07 ± 1.345 ug/ml; the effect was dose depended while the time dependence was not so obviously (Fig. 1A).The colony-forming assay showed 25 ug/ml and a higher dose 40ug/ml of βelemene inhibited the proliferation of the A549 and H1299 NSCLC cells considerably (Fig. 1B). We further examined DNA replication in early phase using EdU assays, And the results showed β-elemene could affect DNA replication signi cantly after cells were incubation in with β-elemene for 12 hours, and this effect was enhanced by increasing the dose (Fig. 1C-D). β-Elemene treatment increased the G0/G1 cell cycle subpopulation cells and reduced the subpopulation of G2/M and S phase cells. This effect was increased when extending incubation time to 24 hours (Fig. 1E). but there were no obvious apoptotic cells ( Figure S1A). Consistently, β-elemene treatment down regulated Cyclin A, Cyclin D1, Cyclin E in both A549 and H1299 cells (Fig. 1F). β-Elemene can also signi cantly inhibit tumor proliferation in the xenograft model (Fig. 1G). We did not observe treatment-related deaths or body weight loss in mice treated with βelemene and the tumor-inhibiting capability is 61.6% (Fig. 1H).

β-Elemene regulated metabolic reprogramming
The network analyses via BATMAN-TCM showed the main KEGG pathway β-elemene work is metabolic pathway ( Fig. 2A). There were only 31 proteins with a score higher than 50 (Table 1). So, we used these proteins to construct a protein-protein interaction (PPI) network with STRING ( Figure S2A), and pathway enrichment analyses were performed via KEGG. The top 4 core pathways are simultaneously related to metabolic including Tyrosine metabolism, Metabolic pathways, Drug metabolism and Glycolysis/Gluconeogenesis (Fig. 2B). Furthermore, to identify whether β-elemene could regulate metabolism reprogramming, we established A549 xenograft model in nude mice (Fig. 2C). The 18 FDG micro-PET scan revealed that after treating with β-elemene, the SUV values of the tumors in vivo were signi cantly down regulated (Fig. 2D).
LC-MS analysis was performed on whole-tumor extracts, in order to dissect metabolic alterations, 13 C metabolite labeling and quanti cation showed there were some heterogeneity between the treatment and control group, but we could still see some signi cant differences (Fig. 3A). In the treatment group, glucose, pyruvate, lactate, ATP and NAD/NADH ratio, which were produced during glycolysis, were obviously decreased, while the production of TCA cycle such as citrate and α-KG were increased signi cantly. The NADP/NADPH ratio and fumarate had an increase trend but P > 0.05 (Fig. 3B-C). We further used NSCLC cells to identify some of the metabolites in vitro. The in vitro and in vivo results were consistent, and showed β-elemene signi cantly decreased the glucose, lactate and NAD + /NADH ratio in cells, while increased α-KG and NADP/NADPH ratio. We also observed that β-elemene can signi cantly decrease the PH intracellular because of the NAD/NADH ratio changed in cells (Fig. 3D).

β-Elemene targeted ALDH3A1 to inhibit NSCLC cell proliferation
We intersected the genes in the top four enriched KEGG pathways to obtain 9 key role proteins, including ALDH1A3, ADH1A, ADH1B, ALDH3B2, ADH7, ALDH3A1, ADH5, ADH4, ADH6 (Fig. 4A). Firstly, we analyze the correlation between these molecules ( Figure S3A-B). Then we evaluated the expression of these proteins in NSCLC patients and normal patients with the data from TCGA and GTEx. ALDH3A1 was the only one to have signi cant difference between cancer and normal tissue, while ALDH3B2 only had signi cant difference between cancer and normal tissue in LUSC (Fig. 4B). On the other hand, the correlation between ALDH3A1 and glycolysis related molecules SLC2A1 was particularly evident with R = 0.63, while the correlation between ALDH3B2 and SLC2A1 was not considerably noticeable with R = 0.36 (Fig. 4C).
The effect of ALDH3A1 on cell proliferation was assessed using NSCLC cells treated with β-elemene, siALDH3A1 or ALDH3A1 inhibitor DEAB individually or together. After treatment with different doses of βelemene for 24 h, both cell lines exhibited suppressed ALDH3A1 mRNA and protein expression (Fig. 4D-E). The enzyme activation of ALDH3A1 was also dose-dependent down regulated by β-elemene in both cell lines (Fig. 4F). After comparing the knockdown effect of Si-1, Si-2 and Si-3, Si-1 and Si-2 displayed higher effect both at the mRNA and protein levels ( Figure S3C-D). We next applied lentivirus Sh-ALDH3A1 to knockdown ALDH3A1, and the knockdown effect of Sh-ALDH3A1 both at the mRNA and protein levels were higher than 50% ( Figure S3E-F). Si-1, Si-2, Sh-ALDH3A1 and DEAB successfully suppressed ALDH3A1 and inhibited NSCLC cells proliferation (Fig. 4G). The Sh-ALDH3A1 was effective alone identi ed via colony forming assays, and its effect was enhanced upon co-administration of DEAB (Fig. 4H). β-elemene could suppressed DNA replication obviously, and this effect was enhanced by the addition of β-elemene in both NSCLC cells (Fig. 4I-J).

ALDH3A1 control HIF1α-Mediated NSCLC glycolysis in vivo and in vitro
We could see the metabolites in TCA cycle increased while the production of ATP was obviously decreased after β-elemene treatment (Fig. 3B-C), which was the main energy source. We next examined whether the loss of ALDH3A1 in uenced tumor metabolism and possible mechanism. 18 FDG micro-PET imaging of mice with Sh-NC tumors in the left and Sh-ALDH3A1 tumors in the right, which had a similar metabolic index (Fig. 5A), showed considerably lower SUV values in the right tumors indicating a potential decrease in glucose uptake in Sh-ALDH3A1 tumors, and this effect was signi cantly enhanced upon β-elemene treatment (Fig. 5B-C).The metabolites analysis showed glucose uptake, production of lactate and NAD+/NADH ratio were signi cantly decreased, the cytosolic levels of NAD + and the NAD+/NADH ratio were major role to sustain glycolytic ux [18],while the α-KG was considerably up regulated and these effects were simultaneously enhanced upon β-elemene treatment (Fig. 5D).These reminded us that α-KG is used as the electron donor of PHDs for prolyl hydroxylation and degradation of HIF1α [19]. Indeed, silencing of ALDH3A1 suppressed HIF1α,and also decreased the expression of a number of transcriptional targets of HIF1α, which including GLUT1, PDK1, LDHA, and PFKL in A549 NSCLC cells. Consistently, overexpression of ALDH3A1 increased HIF1α and also up-regulated the expression of GLUT1, PDK1, LDHA, and PFKL in H1299 NSCLC cells (Fig. 5E-F). This impact suggested that the regulatory effect of ALDH3A1 on glycolysis are mediated by HIF1α. Furthermore, treatment with Octyl-α-KG, the cell-permeable α-KG, similar to knockdown ALDH3A1 in A549 and reversed the effect of overexpression of ALDH3A1 in H1299 (Fig. 5G-H). Taken together, these results indicate that ALDH3A1 suppress NSCLC cell glycolysis via downregulating HIF1α and downstream signaling pathways.

ALDH3A1 down-regulation by β-Elemene induces HIF1α/LDHA pathway inhibited
Combinatorial treatment with β-elemene and Sh-ALDH3A1 enhanced the effect of Sh-ALDH3A1 signi cantly to inhibiting the expression of HIF1α and substantially downregulated GLUT1, LDHA, and PFKL in A549 and H1299 cells (Fig. 6A-B, Figure S4a). The down-regulation of Cyclin A, Cyclin D1, Cyclin E in both A549 and H1299 cells was signi cantly enhanced upon β-elemene treatment (Fig. 6C). We next identi ed the HIF1α/LDHA pathway and cell cycle related proteins in whole-tumor extracted proteins, and the result was consistent with that in A549 and H1299 cells (Fig. 6D-E).
Finally, we identi ed the functional signi cance of ALDH3A1 and β-elemene in vivo. ALDH3A1 knockdown and β-elemene inhibited the growth of A549 cells transfected with lentivirus xenografts in nude mice and we did not observe other side effects, body weight loss or treatment-related mortality in the mice treated with β-elemene and lentiviral (Fig. 7A-E). Immunohistochemistry assay showed a signi cant decrease in Ki-67, ALDH3A1, HIF1α, GLUT1, LDHA, and PFKL in the β-elemene, Sh-ALDH3A1 group and the effect was enhanced upon combinatorial treatment (Fig. 7F). Therefore, these results indicated that the tumor suppressive effect of β-elemene is achieved by targeting ALDH3A1 to inhibit the HIF1α/LDHA pathway.

Discussion
In this study, we found that β-elemene inhibited NSCLC proliferation by targeting ALDH3A1 to control the NAD/NADH ratio, reprogramming glycolysis and OXPHOS balance via suppressing HIF1α/LDHA pathway in vivo and in vitro. Our report proved the great potential of β-elemene in future tumor treatment which had not been deeply studied before.
Although multiple studies have explored the anti-tumor mechanism of β-elemene before, the direct target of β-elemene has not yet been determined. Recently, Y.Pan et al found that β-elemene can inhibit the metastasis of breast cancer via suppressing glycolysis [11], which analyzed by PharmMapper [20,21], the network pharmacology tool which based on pharmacophore model to analyze the action effects. In addition to the pharmacophore there are many other mechanisms. Unlike previous studies, we identi ed ALDH3A1 as the key target of β-elemene via BATMAN-TCM and bioinformatics methods [15,22].
Compared with PharmMapper, BATMAN-TCM integrates the known interactions in DrugBank, TTD and KEGG databases based on similarity algorithms, and sorts potential drug-target interactions, which is more comprehensive and persuasive for nding effective hub targets. ALDH3A1 was reported to be able to regulate cell proliferation and stem cell-like characteristics [23]. As a metabolic enzyme, ALDH3A1 can promote lactic acid production by regulating NAD+/NADH, promoting glycolysis, the main way of tumor energy supply, but weakening the role of OXPHOS, a relatively ine cient way in tumor cells, NAD + metabolism links energy status with adaptive cellular and organismal responses [24][25][26][27], and NAD+/NADH ratio control the balance of OXPHOS and glycolysis [28,29], In this study, we found that β-elemene targeting ALDH3A1 by not only down-regulating its expression at the mRNA and protein levels, but also inhibiting its enzyme activity. Consequently, the suppression of ALDH3A1 by β-elemene leading to decreased the ratio of NAD+/NADH resulting in a large amount of NAD + dissociated in the cytoplasm, and the oxidized form of cellular NAD+, NADH, is critical for mitochondrial function,which was injured in cancer cells [30]. Accumulated NAD + produced negative feedback inhibition on the transformation of pyruvate to lactic acid during glycolysis, and accumulated NADH further inhibited the upstream reaction (the process catalyzed by GAPDH), resulting in downregulation of glycolysis and decrease of NAD+/NADH ratio. More important, our in vivo metabolomics results further veri ed the inhibition of aerobic glycolysis and the activation of oxidative phosphorylation.
The relative amount of ATP in the tumor was signi cantly down-regulated compared with the control group, further certifying that β-elemene can compensate for the slow OXPHOS of energy supply by inhibiting rapid aerobic glycolysis. Reduce the energy supply e ciency of tumor growth, and then inhibit the biosynthesis of DNA, inhibit tumor proliferation.
When the decreased glycolysis under β-elemene lead to energy depletion, mitochondrial function played a secondary compensatory role, resulting in up-regulation of α-KG, succinic acid, malic acid and other intermediate metabolites suppressed in tumors. The up-regulation of α-KG, the product of TCA cycle, induced the degradation of HIF-1α by hydroxylation under normoxic condition [19,[31][32][33]. Moreover, βelemene down-regulates HIF-1α, which was an important transcription factor of Warburg Effect, and the process of aerobic glycolysis by down-regulating its transcribed protein GLUT1, LDHA, and PFKL, which suggests that β-elemene regulates the balance of glycolysis and OXPHOS by targeting ALDH3A1 via HIF1 α / LDHA pathway inhibition. Certainly, metabolic reprogramming also occurs in the tumor microenvironment [34][35][36], and the effects of drugs are often a comprehensive systemic effect. Therefore, whether β-elemene has a similar effect on the tumor microenvironment will be the focus of our future study.
In order to further verify the anti-tumor mechanism of β-elemene. In addition to applying metabolomics, we rst tried to apply 18 FDG-micro-PET to verify the regulation of β-elemene on glucose metabolism in vivo. 18 FDG-PET is one of the main tools for clinical diagnosis and evaluation of drug responsibility.
Taking advantage of the signi cant difference in glucose absorption capacity between tumor tissues and normal tissues, 18 F-labeled glucose ( 18 FDG) is used to track tissues with signi cantly increased metabolic ability in vivo. We used this tool to evaluate the mechanism of β-elemene, which more objectively and accurately re ected the inhibitory effect of β-elemene on glucose metabolism in tumor.
Simultaneously, our results also proved that in the future the e cacy of β-elemene can be effectively evaluated by PET in clinic.

Conclusions
In summary, we provided new evidence that β-elemene targets ALDH3A1 in NSCLC in vivo and in vitro, and found that high expression of ALDH3A1 promote the proliferation of NSCLC via interfering with the NAD+/NADH ratio, which nally causing glycolysis considerably enhanced while OXPHOS capacity signi cantly reduced in cancer cells. Moreover, β-elemene targeted ALDH3A1 to regulate NAD+/NADH ratio and the increased the product of TCA cycle, α-KG, to further inhibit HIF1 α / LDHA signal pathway, thus inhibiting NSCLC progression. This study demonstrated the potential of β-elemene as a front-line treatment.   Figure 1 β-Elemene inhibited NSCLC cells proliferation by alternating the cell cycle. a. MTT assays indicating the cytotoxicity of β-elemene in NSCLC cells including A549 and H1299. b. NSCLC cells treated with high or low dose of β-elemene, followed by a colony-forming assay, and the clones were numerated. c-d. An EdU assay of two cell lines under identical conditions for 24 h, followed by sequential EdU and Hoechst staining. Hoechst 33342 (blue) represents cell nuclei and EdU (red) represents nuclei of proliferative cells.e.Effect of the β-elemene on cell cycle distribution of A549 and H1299 cells. Data are presented as mean ± SD (n= 3, *p < 0.05). Student's t-tests were used for statistical analyses. f. Effect of β-elemene on the expression of cell cycle markers Cyclin E, Cyclin A and Cyclin D1.g. Photographs of tumors four weeks after injection with 1×107 A549 cells in 150 μL PBS. Each trial included six mice treated with PBS (150 μL by oral gavage twice daily); β-elemene (50 mg/kg by oral gavage twice daily) h. The weight and volume of the photographed tumors and mice. Figure 2 β-Elemene regulated metabolic reprogramming. a. Network pharmacology analysis by BATMAN-TCM. b. KEGG pathway analysis of all the target proteins in PPI. c. The diagrammatic sketch of the steps taken to build the mice model to identify the e cacy of β-elemene. d. The 18FDG micro-PET scan before and after treating with β-elemene and the SUV values of the tumors in vivo.

Figure 3
Page 22/27 β-Elemene regulated the balance of glycolysis and oxidative phosphorylation.a.13C metabolite labeling and quanti cation of energy metabolites in the mice models. b-c. The metabolites in glycolysis and TCA cycle in tumors. d. The metabolites in glycolysis and TCA cycle in NSCLC cells. Figure 4 β-Elemene targeted ALDH3A1 to inhibit NSCLC cell proliferation a. Venn diagram of the genes in the top 4 KEGG pathways that were affected by β-elemene. b. The expression of ALDH3A1 and ALDH3B2 in NSCLC mice tumors transduced with Sh-NC and Sh-ALDH3A1with or without treatment with β-elemene (50ug/kg) were subjected to Western blot analysis.  β-elemene (50ug/kg by oral gavage twice daily); b. Changes in mouse weight during treatment. c-d.