Suppression subtractive hybridization identified differentially expressed genes in lung adenocarcinoma: ERGIC3as a novel lung cancer-related gene

The primary tumors and adjacent nonmalignant lung tissues were obtained at the time of surgery and quickly frozen in liquid nitrogen. No patients were treated before undergoing surgical resection. The adjacent nonmalignant lung tissues which were away from the cancer tissues at least 5 cm, did not contain cancer cells but usually appeared inflammatory response and fibrosis. Pathological diagnosis was based on light microscopy according to the World Health Organization classification [12]. Tumors were staged according to TNM criteria published by the International Union Against Cancer in 1997 [13]. Tumor regions selected for RNA and protein isolation contained a tumor cellularity greater than 60%. The use of all of the human tissue samples and the experimental procedures for this study were reviewed and approved by the Tumor Hospital of Yunnan Province and Kunming Institute of Zoology. All researches were carried out according to the Helsinki Declaration.

Six lung cancer cell lines and an immortalized human bronchial epithelial cell line (BEAS-2B) were used. A549 (AC), 801-D (large cell carcinoma), NCI-H446 (SCLC), and BEAS-2B were obtained from Cell Bank of Kunming Institute of Zoology, Chinese Academy of Sciences (CAS, Kunming, China); SPCA-1 (AC) was purchased from the Cell Bank of Type Culture Collection, CAS (Shanghai, China); EPLC-32M1 (SCC) and GLC-82 (AC) were obtained from German Cancer Research Center (Heidelberg, Germany) and Chinese National Cancer Institute, Chinese Academy of Medical Sciences (Beijing, China), respectively. The BEAS-2B cell line was fed with DMEM, while the other cell lines were cultured with RPMI 1640 containing 10% fetal bovine serum (FBS) and maintained in a humidified incubator with 5% CO₂ at 37°C.

To construct the SSH library, we used a tumor tissue and its adjacent nonmalignant lung tissue derived from one patient with well-differentiated lung adenocarcinoma. The tumor tissue contained a tumor cellularity greater than 80%. The adjacent nonmalignant lung tissue was away from the cancer tissues at least 10 cm and appeared completely normal histological structure. Total RNA was isolated using TRIzol (Invitrogen Corp., Carlsbad, CA, USA). Poly(A)⁺ RNA was isolated using the PolyATtract® mRNA isolation systems (Promega Corp., Madison, WI, USA). The cDNA synthesis and subtraction were performed using the PCR-select™ cDNA subtraction kit (Clontech, Palo Alto, CA, USA). Using the cDNA of lung adenocarcinoma tissue as tester and its adjacent nonmalignant lung tissue as driver, we generated the FSL which represented up-regulated transcripts. Using the cDNA of lung adenocarcinoma tissue as driver and its adjacent nonmalignant lung tissue as tester, the RSL which represented down-regulated transcripts, was constructed. The subtracted cDNA fragments obtained in each experiment were cloned into the pGEM-T Easy vector (Promega Corp.) and used to transform E.coli strain DH5α cells. All the transformants were isolated from white colonies on X-gal/isopropyl-beta-D-thio-galatopyranoside agar plates.

Total RNA was isolated from lung cancer tissues and adjacent nonmalignant tissues using TRIzol (Invitrogen Corp.). RNA was used as a template to synthesize the first strand cDNA with M-MLV reverse transcriptase (Promega Corp.). The q-RT-PCR reactions were performed on an ABI stepone™ real-time RT-PCR system using the SYBR Green dye method. The q-RT-PCR reactions were performed in 25 μl volumes that included 0.5 μl of SYBR green, 8 ng of cDNA template and 1.0 μl each of the forward and reverse primers (10 μM) (see Additional file 1). The PCR was done under the conditions at 94°C for 30 seconds, then 40 cycles of amplification at 94°C for 30 seconds, 60°C for 30 seconds. Each gene was normalized to the internal β-actin levels. Each sample was run in triplicate to ensure quantitative accuracy, and the threshold cycle numbers (Ct) were averaged. The results were reported as tumor tissues: normal tissues (T:N) ratios and calculated using the 2^−ΔΔCt method [14].

Total protein was extracted from cultured cells and tissues, separated by electrophoresis, and then transferred into PVDF membrane according to the routine protocol. The blotted membrane was incubated with the rabbit anti-ERGIC3 polyclonal antibody (Sigma-Aldrich, St. Louis, MO, USA) followed by horseradish peroxidase-conjugated anti-rabbit secondary antibody (AbMART, Shanghai, China), and then proteins were detected with chemiluminescence reagents (Thermo Fisher Scientific Inc., Waltham, MA, USA). The membrane was reprobed with an antibody against β-actin (Abcam, Cambridge, UK) as a control for equivalent protein loading.

Immunohistochemical staining was done on 4-μm-thick paraffin sections cut from the formalin fixed tissues. Heat-induced epitope retrieval was performed in Tris-EDTA Buffer (10 mM Tris Base, 1 mM EDTA Solution, 0.05% Tween 20, pH 9.0). The sections were then incubated in 3% H₂O₂ for 10 minutes to eliminate endogenous peroxidase activity. The primary rabbit anti-ERGIC3 polyclonal antibody (Abcam) and horseradish peroxidase-conjugated anti-rabbit secondary antibody were used. Color development was accomplished with 3,3'-Diaminobenzidin. The nuclei were then counterstained with hematoxylin. Only manifest cytoplasmic staining was defined as a positive reaction. Negative controls were incubated with normal rabbit serum instead of the polyclonal antibody.

Immunofluorescence analysis was done on the cultured cells. Briefly, the cells grown on coverslips were fixed in −20°C acetone for 10 minutes and then incubated with the rabbit anti-ERGIC3 polyclonal antibody (Abcam), and anti-MUC1 mouse monoclonal antibody (mAb) A76-A/C7 (Glycotope, Berlin, Germany), anti-ST mAb [15], anti-calreticulin mAb (Abcam), and anti-58K Golgi protein mAb (Abcam) at 4°C overnight. Following the washes, the cells were incubated with FITC-coupled anti-rabbit antibody (BD sciences, Franklin Lakes, NJ, USA) or with Cy3-coupled anti-mouse antibody (Millipore, Billerica, MA, USA) for 1 hour. Subsequently, the nuclei were finally counterstained with diamidinophenylindole. The slides were analyzed under confocal laser scanning microscopy.

To generate the plasmid expressing ERGIC3, we cloned the open reading frame of ERGIC3 and used the following specific primers (upper case, restriction enzyme sequences): forward, 5'-GGTGGTGAATTCatgaggcgctggggaagct-3'; reverse,5'-GGTGGTGGATCCaccgaggagggtgactacgttgtctt-3'. After running PCR, the product was cut by EcoR I and BamH I. The purified DNA was ligated into the pLXSN expression vector (Clontech Corp.). The empty vector served as a control. The vectors were transfected into BEAS-2B cell by lipofectamine LTX and PLUS (Invitrogen Corp.) according to the instructions. The over-expression of ERGIC3 was validated by q-RT-PCR and western blot.

For gene silencing of ERGIC3, pGPU6/GFP/Neo-ERGIC3-shRNA and its control vector, pGPU6/GFP/Neo-NC-shRNA, were used (Jima Corp., Shanghai, China). Briefly, the oligonucleotide GGAGGACTATCCAGGCATTGT was designed to interfere specifically with ERGIC3 gene expression. As a negative control, we used the oligonucleotide GTTCTCCGAACGTGTCACGT, which has no significant match in a BLASTn search (human NCBI nr database). The vectors were transfected into GLC-82 cells by lipofectamine LTX and PLUS. The gene silencing of ERGIC3 was validated by q-RT-PCR and western blot.

Cell proliferation analysis was based on the capacity of mitochondrial enzymes to transform MTT to MTT formazan. More succinctly, after being transfected for 48 hours, GLC-82 and BEAS-2B were collected and transferred into 96-well plates (4*10³ cells/well), then all cells were treated with MTT (5 mg/ml) every 24 hours. 10 μl of MTT was added to each well and cells were incubated at 37°C for 2 hours. Then, the culture medium with dye was removed and dimethylsulfoxide was added at 100 μl per well for formazan solubilization. The absorbance of converted dye was measured at a wavelength of 490 nm using a 96-well microplate reader (model 680, Bio-Rad Laboratories).

The transfected GLC-82 and BEAS-2B cells were used for the cell migration assay. The transwell migration assay was conducted in 24 well plates with membrane inserts (8 mm pore size; Millipore). The cells were then seeded in the upper chamber of transwells (10⁵ cells/well) without FBS. The lower chambers were loaded with 500 μl of medium with 10% FBS. After incubation for 24 hours, filters were washed and the cells on the upper surface were gently removed with a cotton swab. The cells were fixed with 95% ethanol, and were stained by Giemsa I (Jiancheng Corp. Nanjing, China) and Giemsa II. The cells were then counted under microscope. The experiments were repeated three times.

Measurement data (levels of mRNA and protein, the data of cellular migration and proliferation) and enumeration data (the data of immunohistochemical staining) were respectively analyzed using the paired t-test and the Fisher’s exact test (two-tailed). All of the values were evaluated using IBM SPSS 19 (SPSS inc., Chicago, Illinois). Differences were considered significant if P < 0.05.

Using SSH, two cDNA libraries were constructed with the primary lung AC tissue derived from a single patient. From the FSL library, 485 subtractive cDNA clones were obtained, representing genes up-regulated in tumor tissues. Meanwhile, from the RSL library, 172 subtractive cDNA clones were obtained, representing genes down-regulated in tumor tissues. Using the BLAST algorithm at NCBI RefSeq, GenBank, and dbEST, we detected 177 genes and 44 unknown expressed sequence tags (ESTs) in the FSL as well as 59 genes and 10 unknown ESTs in the RSL. Further bioinformatic analysis demonstrated the major functions of these genes were related molecular transport, cellular signaling and interaction, cellular polarization, cell cycle, cell apoptosis, cellular growth and proliferation, cellular movement, DNA replication, recombination and repair (see Additional file 2 and 3).

Eight earlier constructed SSH libraries of lung cancer (five FSLs and three RSLs) were available for us to obtain gene expression profiles [6‐10]. From the previous five FSLs, 152 genes from our FSL were not reported. Similarly, 54 genes in our RSL were not present in the previous three RSLs. In the six aggregated FSLs from our study and earlier publications, 42 genes appeared twice, while four genes (EEF1A1, FTH1, GSTP1, STAT1) appeared three times (see Additional file 4). Additionally, in the four RSLs from our study and the three previous, only six genes (ANXA8, CAV1, CEBPD, GPX3, TPM3, NACA) appeared twice. Among those six, CAV1, CEBPD, GPX3, TPM3 and NACA were present in our RSL (see Additional file 5).

As a first stage analysis, we selected 16 genes among the differentially expressed genes from our FSL and RSL libraries for further study based on two criteria: 1) The 10 genes were previously reported in lung cancer while six were not reported; 2) These genes belong to importantly functional genes. We examined mRNA expression of 16 genes selected from the SSH libraries using q-RT-PCR. Among the 16 genes, two were selected from the RSL, one from both the RSL and FSL, and 13 from the FSL. The percentage of the altered expression in the tumor tissues compared to their adjacent nonmalignant lung tissues is shown in Table 1. The mRNA levels of the 16 genes in the lung cancer tissues were compared with those in the adjacent nonmalignant lung tissues using the paired t-test. The two genes selected from the RSL were significantly down-regulated in the tumor tissues as compared with the adjacent nonmalignant lung tissues (GPX3, P = 0.000 and TIMP3, P = 0.000), and six genes selected from the FSL were significantly up-regulated in the lung cancer tissues (DDR1, P = 0.009; HSP90B1, P = 0.000; SDC1, P = 0.007; RPSA, P = 0.013; ERGIC3, P = 0.000; and LPCAT1 P = 0.036). However, seven genes selected from FSL (FOXA2, C4BPA, SCGB3A1, DDX58, CCNDBP, TMSB4X, and CXCL17), and one gene selected from both FSL as well as RSL (CD9), were up-regulated in the lung cancer tissues, but not significantly (P > 0.05). The tendency of the seven genes expressions was consistent with that of the FSL which represented genes up-regulated in cancer tissues. We noted that CD9, present in both the FSL and RSL, was increased in 41.2% of lung cancer cases. Perhaps the presence of CD9 in the RSL was a false signal. Over-expression of two genes (ERGIC3 and LPCAT1) had not been previously linked to lung cancer. We opted to focus on ERGIC3 in this study.

Expression of the ERGIC3 protein was analyzed using western blot. Expression of ERGIC3 was increased in 67% (10/15) of the tumor cases (Figure 1A). Similarly, expression of ERGIC3 protein was increased in all three lung cancer cell lines by comparison with the BEAS-2B (Figure 1B).

In cultured cells, the subcellular localization of ERGIC3 was examined by immunofluorescence double-staining using markers of the Golgi apparatus and endoplasmic reticulum (ER). ERGIC3 was mainly located at the Golgi apparatus and ER in the lung cancer cell lines. Interestingly, ERGIC3 was distributed at the side of nucleus in EPLC-32M1, 801D, and NCI-H446 cells, but uniformly present around nucleus in SPCA-1, GLC-82 and A549 cells (Figure 2).

We also found that ERGIC3 was co-localized with the epithelia mucin MUC1 and β-Galactoside α2,6 Sialyltransferase (ST), which were principally located at the ER and Golgi apparatus in the cultured cells (Figure 2).

Expression and localization of the ERGIC3 protein was further investigated by immunohistochemistry in 35 cases of NSCLC. ERGIC3 was diffusely distributed in the cytoplasm of tumor cells but not expressed in normal bronchial epithelial cells and alveolar cells (Figure 3). The antibody of ERGIC3 used in the study was the polyclonal anti-ERGIC3 serum, this may be the reason that the non-specific staining appeared in nucleus.

ERGIC3 was positive in 89% of (31/35) NSCLCs, and strongly stained in 63% (22/35). Interestingly, the positive rate of AC (100%, 22/22) was higher than that of SCC (69%, 9/13). In poorly differentiated NSCLCs, 36% (4/11) cases were not stained. We noticed that all of negative specimens were poorly differentiated tumor cells, and the staining was decreased to the largest extent and even disappeared in poorly differentiated NSCLCs, compared with well and moderately NSCLCs (P <0.05). No correlations were observed between expression of ERGIC3 and the gender as well as smoking histories of the patients and TNM stage (Table 2).

In order to study whether altered expression of ERGIC3 affects the cell proliferation, GLC-82 (with the high level of endogenous ERGIC3) and BEAS-2B cells (with the low level of endogenous ERGIC3) were used in the study. In GLC-82 cells, compared with the cells transfected with the control vector, the expression of ERGIC3 was reduced by RNA interference. While the expression of ERGIC3 in BEAS-2B cells was increased after the transfection with the expression vector. The altered expression of ERGIC3 was indicated by q-RT-PCR (Figure 4A,B) and western blotting (Figure 4C,D). We found the reduced expression of ERGIC3 in GLC-82 significantly reduced the rate of cell proliferation (Figure 4E), and the over-expression of ERGIC3 in BEAS-2B cells significantly increased the rate of cell proliferation (Figure 4F).

We also analyzed the impact of ERGIC3 on cellular migration. The expression of ERGIC3 in GLC-82 and BEAS-2B cells was successfully altered through gene silencing and over-expression, as described earlier. The treated cells were used for the transwell migration assay based on Boyden Chamber. We found that the lower-expression of ERGIC3 could inhibit the cellular migration in GLC-82 (Figure 5A), and the over-expression of ERGIC3 could promote the cellular migration in BEAS-2B (Figure 5B).