The expression level of FH in lung adenocarcinoma tissue compared with normal lung tissue was analyzed using the Oncomine online database (https://www.oncomine.org).

Lung cancer tissues were obtained from patients undergoing surgical treatment at Kaohsiung Medical University Hospital (Kaohsiung, Taiwan) and E-Da Hospital (Kaohsiung, Taiwan) between February 2007 and February 2013. All patients with lung cancer (adenocarcinoma and squamous cell carcinoma) who underwent a lung resection, including 68 men and 36 women, were included in the study. The age distribution was from 29 to 84 years old. Patients with diseases other than lung cancer were excluded. Overall survival (OS) was defined as the interval between the date of diagnosis and death. IRB approval was received from the Institutional Review Board (or Ethics Committee) of Kaohsiung Medical University Hospital [KMUHIRB-E(I)-20180026] and the Institutional Review Board for Human Studies of E-Da Hospital (EMRP-098-132 and EMRP-101-040). Written informed consent was obtained from all the patients.

All tissues used to create the tissue microarray were obtained from formalin-fixed, paraffin-embedded tissue blocks. Histopathological slides prepared from hematoxylin and eosin-stained sections were evaluated by a pathologist who selected representative areas of tumor or normal tissues for scoring. The tissue microarray was constructed using Booster Arrayer & TMA designer software (Alphelys) according to a previously described procedure (21).

The immunohistochemical (IHC) staining of FH was performed using a Bond-Max automated system (Leica Microsystems GmbH) with an anti-FH antibody (GTX110128; 1:500; GeneTex, Inc.). The relative expression of FH in the lung cancer specimens was quantified using a TissueFAXS microscopy system and HistoQuest software 2.0 (TissueGnostics GmbH). The IHC score of the lung cancer tissue was calculated by multiplying the percentage (1–100%) of positively stained cells by the intensity of staining (0, 1+, 2+ or 3+). For further statistical analysis, low and high expression categories were established based on a receiver operating characteristic curve analysis. Patients with lung cancer were classified into two groups using the these scoring categories as follows: Low FH, IHC score <45; and high FH, IHC score ≥45.

The CL1-0, H441, H1299 and CL1-5 human lung cancer cell lines were purchased from the Bioresource Collection and Research Center and were maintained in Roswell Park Memorial Institute (RPMI)-1640 medium (Gibco; Thermo Fisher Scientific, Inc.). A549, H250 and H460 cells were purchased from the American Type Cell Collection and maintained in RPMI-1640 medium. All cell lines were cultured with 5% CO₂ at 37°C in a humdified incubator. All culture media were supplemented with 10% fetal bovine serum (FBS; Biological Industries) and 1% penicillin G, streptomycin and amphotericin B.

For the knockdown of FH in the CL1-0 and H441 lung cancer cell lines, a pLKO.1_puro lentiviral vector (National RNAi Core Facility Platform, Academia Sinica) expressing double-stranded shRNA oligonucleotides targeting human FH (2 clones) was used: Clone 1 (shFH1), IDTRCN0000052466, target sequence, 5′-GTGGTTATGTTCAACAAGTAA-3′ and oligo sequence: 5′-CCGGGTGGTTATGTTCAACAAGTAACTCGAGTTACTTGTTGAACATAACCACTTTTTG-3′; and clone 2 (shFH2), ID TRCN0000310398, target sequence, 5′-CCCAACGATCATGTTAATAAA-3′ and oligo sequence, 5′-CGGCCCAACGATCATGTTAATAAACTCGAGTTTATTAACATGATCGTTGGGTTTTTG-3′ (National RNAi Core Facility, Academia Sinica). A pLKO.1_puro lentiviral vector expressing shRNA targeting firefly luciferase (shluc), which is not associated with the human genome sequence, was used as a negative control (National RNAi Core Facility, Academia Sinica). For the overexpression of FH in H1299 cells, a lentivirus with a pLVX-puro backbone (#GRVL7005G) and empty control (#GRV7006G) were purchased from Topgen Biotechnology Co., Ltd. Cell infection was performed according to the detailed procedures described in a previous study (22).

Cell migration assays were carried out using Transwell (Costar; Corning Inc.) membrane filter inserts (diameter, 6.5 mm; pore size, 8 µm) in 24-well tissue culture plates. Following the knockdown or overexpression of FH, CL1-0, H441 and H1299 lung cancer cells were trypsinized, suspended in serum-free RPMI 1640 medium and seeded (2×10⁴ cells) on the Transwell filter in the upper chamber, while RPMI-1640 medium containing 10% FBS (Biological Industries) was added to the lower chamber. The cells were then incubated for 24 h at 37°C. After incubation, the cells were fixed with 4% formaldehyde for 10 min and stained with crystal violet for 2 h, at room temperature. Non-migrating cells were removed by wiping the upper side of the filter, and the migrated cells were imaged using Olympus SZX10 stereo light microscope (Olympus Corporation) and analyzed using ImageJ software (ij153-win-java8; National Institutes of Health).

BioCoat Matrigel invasion chambers (Corning Inc.) were used for the invasion assay. Prior to use, Transwell invasion chambers were rehydrated with serum-free medium for 2 h. The remainder of the protocol was identical to that of the migration assay, with the exception that the lung cancer cells were not transfected. ImageJ software was used to count the number of invaded cells. The invasive ability was calculated based on the percentage of invaded cells and normalized to CL1-5.

Western blot analysis was performed to check the knockdown and overexpression efficiency of lentivirus infection and to evaluate the expression of other proteins in the cells using a previously described procedure (23). Antibodies against FH (#GTX110128; 1:2,000; GeneTex, Inc.), p-AMPK (GTX52341; 1,1,000; GeneTex, Inc.), AMPK (#5831; 1:1,000; Cell Signaling Technology, Inc.), DAB2 (AF8064; 1:500; R&D Systems, Inc.), jagged canonical Notch ligand 1 (JAG1; GTX31607; 1:1,000; GeneTex, Inc.), interferon-related developmental regulator 1 (IFRD1; GTX104578; 1:1,000; GeneTex, Inc.) and actin (#A5441; 1:5,000; MilliporeSigma) were used. The AMPK inhibitor BML-275 was purchased from Santa Cruz Biotechnology, Inc. (sc-200689). H441 cells were treated with BML-275 (15 µM) for 24 h at 37°C in 1% FBS-containing medium before collection for western blot analysis. In brief, after protein transfer, the polyvinylidene fluoride (PVDF) membrane was incubated overnight at 4°C with primary antibodies, followed by incubation with secondary rabbit (HRP conjugate; GTX2131101; 1:5,000; GeneTex, Inc.) or mouse (HRP conjugate; GTX213111; 1:5,000; GeneTex, Inc.) antibodies for 1 h at room temperature. The protein bands on the PVDF membrane were visualized using Western Lightning^R Plus-ECL enhanced chemiluminescence substrate (PerkinElmer, Inc.) and analyzed using Image Lab software 6.0.1 (Bio-Rad Laboratories, Inc.).

A cell proliferation assay was performed using 2,3-bis-(2-methoxy-4-nitro-5-sulfophenyl)-2H-tetrazolium-5-carboxanilide (XTT) as described in previous studies (24,25). In brief, H441, CL1-0 and H1299 cells were seeded in 96-well plates (3,000 cells/well). After 24–72 h, XTT (X4251; Sigma-Aldrich; Merck KGaA) solution containing phenazine methosulfate (P9625; Sigma-Aldrich; Merck KGaS) was added. After 30 min, the absorbance was measured at 475 and 660 nm. The proliferation rate was calculated as 475 nm absorbance minus non-specific reading at 660 nm absorbance.

Following the knockdown of FH, CL1-0 lung cancer cells were seeded (1×10⁶ cells/plate) in a 10-mm dish. After 24 h, the medium was removed and the cells were washed with PBS twice. The PBS was removed and 1 ml ice-cold methanol diluted with water (80:20) was added in accordance with a previously described method (26). The cells were scraped from the dish and transferred to Eppendorf tubes, in which the mixture was vortexed and put on ice for 5 min. The samples were then centrifuged at 18,528 × g at 4°C for 10 min, and the supernatants were collected into clean Eppendorf tubes and subjected to speed vacuum concentration followed by lyophilization to obtain the samples in powdered form. Metabolic data were acquired from the samples using a Q Exactive™ Plus Orbitrap Mass Spectrometer (Thermo Fisher Scientific, Inc.) coupled with a Vanquish™ UPLC (Thermo Fisher Scientific, Inc.) under John Hopkins University (Baltimore, USA) metabolomics analysis service. All metabolomics data were normalized by the protein concentration of each sample. Multiple reaction monitoring for fumarate and malate was assessed at the mass transitions 115 to 100 and 133.01 to 100 m/z, respectively, with a retention time of 4 min. Other parameters were as follows: Negative ionization mode of detection, 350°C nitrogen gas temperature, 35 psi nebuliser pressure and an 11 l/min sheath gas flow rate.

TRIzol^® reagent (Thermo Fisher Scientific, Inc.) was used to extract total RNA from the FH knockdown CL1-0 cells and the respective shluc control cells in accordance with the manufacturer's instructions. An aliquot of RNA (2 g/sample) was processed with DNase (Merck &Co., Inc.) and converted into cDNA using an RT2 First Strand Kit (Qiagen, Inc.). Then, using the human-signal transduction pathway finder RT² profiler PCR array (PAHS-014Z; Qiagen, Inc.), 84 pathway-associated genes and 5 housekeeping genes were screened according to the manufacturer's instructions.

All statistical analyses were performed using the SPSS 14.0 statistical package for PC (SPSS, Inc.). Comparisons between two groups were performed using unpaired Student's t-test while comparisons among multiple groups were performed using one-way ANOVA followed by Tukey's post hoc test. The relationship between the expression of FH and invasive ability was evaluated by linear regression. Associations of the expression of FH with age, sex, stage, tumor size, lymph node metastasis, distant metastasis, histologic type, tumor recurrence and smoking status were investigated by χ² or Fisher's exact tests. Survival curves were generated using Kaplan-Meier estimates and the significance of difference between curves was evaluated by log-rank test. Furthermore, univariate and multivariate Cox regression models were used to investigate the associations between clinicopathological characteristics and OS. P<0.05 was considered to indicate a statistically significant result.

Assessment of the FH expression level in patients with lung cancer using the Oncomine database revealed that FH mRNA expression was significantly lower in lung adenocardinoma tissues compared with normal lung tissues (Fig. 1A). Furthermore, the protein expression levels of FH in lung cancer tissues collected from patients were analyzed by IHC staining (Fig. 1B). The results showed that lung cancer tissue had lower expression of FH compared with normal lung alveoli. Survival analysis of the patients with lung cancer revealed that the OS of the high FH expression group was significantly prolonged compared with that of the low FH expression group (P=0.029; Fig. 1C).

The associations between FH expression levels and the clinicopathological characteristics of the patients with lung cancer were also examined. The results revealed that low FH expression was significantly associated with lymph node metastasis (P=0.028) and disease recurrence (P=0.05) (Table I). Although the multivariate analysis did not show that FH expression level was a significant factor for OS (P=0.093), the univariate analysis indicated that FH expression level was a significant predictor of OS for patients with lung cancer (P=0.013; Table II), and high FH expression was associated with lower hazard ratio.

The analysis of the clinicopathological data indicated that FH might play an important role in lung cancer metastasis (Table I). Therefore, the effect of FH on lung cancer cell invasion and migration was investigated in vitro. Western blotting showed that the level of endogenous FH was downregulated in highly invasive CL1-5, H1299 and A549 cells, and FH expression was negatively associated with invasion ability in seven human lung cancer cell lines (Figs. 2A, B and S1). Knockdown of FH expression in the poorly invasive CL1-0 and H441 lung cancer cell lines significantly increased the migration ability of the cells compared with the respective shluc-transfected control cells (Fig. 2C). To ensure that the knockdown of FH had an effect on the TCA cycle and reduced FH activity, the fumarate level in FH knockdown CL1-0 cells was also measured by metabolomics analysis and the results revealed that fumarate level was significantly higher while the malate level was significantly lower in the FH knockdown cells compared with the shluc control (Fig. S2). Conversely, the overexpression of FH in the highly invasive H1299 cell line significantly reduced the migration ability of the cells compared with the control cells (Fig. 2D). This supported the patient data which showed that low FH expression was significantly associated with lymph node metastasis in lung cancer. The XTT assay showed that the neither the knockdown nor the overexpression of FH in lung cancer cells affected the proliferative potential of the cells (Fig. 2E).

The human-signal transduction pathway finder RT² profiler PCR array was used to investigate the downstream and upstream genes in FH knockdown CL1-0 cells. The results showed that three mRNAs, namely IFRD1 (fold change 2.03), JAG1 (fold change 2.35) and DAB2 (fold change-2.37) showed >2-fold changes in the FH knockdown group compared with the control group (Fig. 3A and B). These findings were evaluated at the protein level by western blotting and the results confirmed that DAB2 protein expression in FH knockdown cells CL1-0 and H441 was downregulated by 0.3 and 0.4 fold, respectively, compared with the shluc control group. By contrast, DAB2 protein level in FH overexpressed cells was upregulated by 1-fold, compared with the vector control (Fig. 3C). However, IFRD1 and JAG1 protein expression did not change following FH knockdown in the same manner as the mRNA expression observed in the RT² array analysis. (Fig. S3). Therefore, DAB2 was focused upon for further experiments.

Previous studies have shown that fumarate accumulation affects AMPK protein expression. For example, in kidney cancer, FH deficiency has been reported to downregulate AMPK expression (20). Also, in renal cancer cells, it has been reported that fumarate activates AMPK, which protects FH-deficient cancer cells from apoptosis (27). Therefore, AMPK and p-AMPK protein levels were evaluated by western blotting. The results demonstrated that p-AMPK protein levels were upregulated in FH knockdown CL1-0 and H441 cells by 2- and 1.5-fold, respectively, compared with the control shluc group (Fig. 3C) while p-AMPK protein levels were downregulated in FH overexpressing H1299 cells by 0.7-fold (Fig. 3C). Furthermore, to investigate whether DAB2 was downstream or upstream of p-AMPK, FH knockdown H441 cells were treated with the p-AMPK inhibitor BML-275. Western blotting demonstrated that the inhibition of p-AMPK upregulated DAB2 protein expression by 1.5-fold in FH knockdown cells (Fig. 3D). These results suggest that the downregulation of FH leads to an increase in AMPK phosphorylation, which decreases DAB2 protein expression, resulting in increased cancer cell motility (Fig. 3E).