Several online bioinformatics tools were used to predict the subcellular distribution of eIF3g, including WoLF PSORT (http://wolfpsort.org), PROST II Prediction (http://psort.hgc.jp/form2.html), Subnuclear Compartments Prediction System (http://array.bioengr.uic.edu/subnuclear.htm) and PredictProtein (https://www.predictprotein.org). The sequence-based protein function was predicted using the online tool, PredictProtein (https://www.predictprotein.org). Protein sequence data were obtained from the National Center for Biotechnology Information Protein database (http://www.ncbi.nlm.nih.gov/protein). Gene Ontology terms are organized in two different ontolo-gies, Molecular Function Ontology and Biological Process Ontology (http://geneontology.org/).

The human breast cancer cell line, Bcap37 (The Cell Bank of Type Culture Collection of the Chinese Academy of Sciences, Shanghai, China), with doxycycline-inducible overexpression of eIF3g (Bcap37/Tet-on-eIF3g) was used as previously described (13). Briefly, the human breast cancer Bcap37 cells were cultured in RPMI 1640 medium (Boster Biological Technology) supplemented with 10% fetal bovine serum (FBS; Gbico; Thermo Fisher Scientific, Inc., Waltham, MA, USA) or 10% Tet-approved FBS (Clontech Laboratories, Mountain View, CA, USA) in a humidified 37°C incubator. The full length eIF3g cDNA in the pLVX-Tight-Puro vector (Clontech Laboratories) was used to produce the lentivirus, following packaging with a Lenti-X™ Tet-On Avanced Inducible Expression System (Clontech Laboratories), according to the manufacturer's protocol. HEK293T cells (8×10⁶ per dish; American Type Culture Collection, Manassas, VA, USA) were cultured in Dulbecco's modified Eagle's medium (DMEM) with high glucose (Boster Biological Technology), supplemented with 10% FBS, and used for transfection. The lentivirus-containing supernatant was collected. The Bcap37 cells (6×10⁵ per dish) were infected with the lentivirus in a 6-well plate. Geneticin (250 µg/ml; Amresco LLC, Solon, OH, USA) and puromycin (0.25 µg/ml; Sigma-Aldrich, St. Louis, MO, USA) were used to select and maintain stable transfectants. Doxycycline (1 µg/ml; Sigma-Aldrich) was used to induce the overexpression of eIF3g.

Fractionation was performed to separate the cytosolic fraction and nuclear fraction of the cells (14). Briefly, the trypsinized cells were washed once for 1 min with ice-cold phosphate-buffered saline (PBS) and lysed with 1 µl ice-cold 0.1% NP-40 in PBS. A small volume (300 µl) of the lysate was saved as 'whole cell lysate'. The remaining lysate was centrifuged at 15,000 × g (4°C) for 5 min, and the supernatant was collected as the 'cytosolic fraction'. The pelleted 'nuclear fraction' was further lysed in radioimmunoprecipitation assay (RIPA) buffer [50 mM Tris-HCl (pH 7.4), 1% NP-40, 0.5% sodium deoxycholate, 150 mM NaCl and 0.1% SDS], containing protease inhibitor cocktail tablets (Roche Diagnostics, Basel, Switzerland). For crosslinking experiments, the modified RIPA buffer (1X PBS, 1% NP-40, 0.5% sodium deoxycholate, 150 mM NaCl and 0.1% SDS) was used. The lysate was centrifuged at 15,000 × g (4°C) for 15 min and the supernatant was collected as the 'nuclear fraction'.

The cells were lysed in RIPA buffer and the lysate was centrifuged at 15,000 × g for 15 min at 4°C. The protein supernatant was quantified using a DC Protein Assay kit (Bio-Rad Laboratories, Inc., Hercules, CA, USA). The cell lysates and protein samples, prepared in reducing sample buffer [50 mM Tris-HCl (pH 6.8), 2% SDS, 0.01% Bromophenol Blue, 10% glycerol and 50 mM DTT] were subjected to SDS-PAGE and transferred onto polyvinylidene fluoride (0.2 µm PVDF) membranes. Following blocking with 10% nonfat milk for 1 h, the membranes were incubated with the following antibodies: Polyclonal rabbit anti-human eIF3g (1:10,000; Bethyl Laboratories, Inc., Montgomery, TX, USA; cat. no. A301-757A), polyclonal rabbit anti-human hnRNP U (1:1,000; Abcam Cambridge, UK; cat. no. ab20666), polyclonal goat anti-human HSZFP36 (1:1,000; Santa Cruz Biotechnology, Inc., Santa Cruz, CA, USA; cat. no. sc-247189), polyclonal goat anti-human β-actin (1:1,000; Santa Cruz Biotechnology, Inc.; cat. no. sc-1615), polyclonal rabbit anti-human Histone H4 (cat. no. 07-108; 1:1,000; EMD Millipore, Billerica, MA, USA) overnight at 4°C. This was followed by incubation with secondary peroxidase-conjugated goat anti-rabbit IgG (1:5,000; ZB-2301) and rabbit anti-goat IgG (1:5,000; ZB-2306) (Zhongshan Golden Bridge Biotechnology, Co, Ltd, Beijing, China) or horseradish peroxidase-conjugated polyclonal rabbit anti-human GAPDH (cat. no. sc-25778; 1:1,000; Santa Cruz Biotechnology, Inc.) for 2 h at room temperature. Enhanced chemiluminescence substrates (EMD Millipore) were then applied and signals were detected using a chemiluminescence imaging system (ChemiDoc™ MP Imaging System; Bio-Rad Laboratories, Inc.).

The nuclear protein (200 µg) was incubated with the primary antibodies mentioned above (5 µg anti-eIF3g, 5 µg anti-hnRNPU, 5 µg anti-HSZFP36 or 5 µg anti-β-actin) and 50 µl protein A/G PLUS-Agarose (Santa Cruz Biotechnology, Inc.) on a rotating device (Tube Tumbler Rotating Mixer SBS550-2; Select BioProducts, Edison, NJ, USA) at 4°C overnight. Normal rabbit serum (2 µl) was used as a control group. Following centrifugation at 5,000 × g for 10 min at 4°C, the co-immunoprecipitated proteins were washed with PBS four times (5 min each time), treated with reducing sample buffer and subjected to 10% SDS-PAGE. Following electrophoresis of the proteins, the gel was stained using 50 ml Bio-safe Coomassie blue stain (Bio-Rad Laboratories, Inc.) for 1 h at room temperature with gentle agitation, followed by destaining in distilled water for >2 h until the protein bands were clearly visible. The protein bands were excised and placed into microtubes, and were sent for mass spectrometry characterization of proteins by service provider (Institute of Biomedical Sciences, Fudan University, Shanghai, China).

The separated nuclear fraction was lysed with modified RIPA buffer. DSS (Thermo Fisher Scientific, Inc.; spacer arm, 11.4 Å) was added to the nuclear fraction to a final concentration of 5 mM and incubated at room temperature for 30 min. To quench the reaction, quenching buffer (1 M Tris-Cl; pH 7.5) was added to a final concentration of 20–50 mM Tris (Bio-Rad Laboratories, Inc.) at room temperature for 15 min. The mixtures were then prepared in SDS-PAGE loading buffer for western blot analysis.

The full length coding sequence of eIF3g was inserted into the pGEX-5X-2 vector (Pharmacia Biotech, Tokyo, Japan), in-frame with the GST open reading frame, which led to the production of soluble GST-eIF3g fusion protein when transformed into Escherichia coli DH5α (Takara Bio Inc., Otsu, Shiga, Japan) and induced by 0.1 mM isopropyl-β-D-1 -thiogalactopyranoside (Amresco LLC) for 4 h at 37°C. For the GST pull-down assays, the GST or GST-eIF3g proteins expressed in E. coli were purified using magnetic glutathione-beads (Promega Corp., Madison, WI, USA), according to the manufacturer's protocol. Briefly, bacteria were collected following culture by centrifugation at 5,000 × g for 15 min at 4°C, resuspension in 3 ml ice-cold PBS and lysis by sonication on ice. The lysate was centrifuged at 5,000 × g, for 10 min at 4°C to remove the cell debris, and 100 µl magnetic glutathione beads were added to the supernatant for binding of the GST/GST-eIF3g protein. The beads were washed three times (5 min each time) with PBS and then incubated with the nuclear lysates from the Bcap37 cells with rotation for 30 min at room temperature. The beads were then washed at least three times with PBS (5 min each time); bound proteins were resolved by 8% SDS-PAGE and detected by western blotting.

The Bcap37/Tet-on-eIF3g and Bcap37 cells were seeded separately into a 6-well plate with a glass-bottomed dish at a density of 2×10⁴ cells/dish and treated with doxycycline for 3 days at 37°C. The cells were then washed with PBS three times (10 min each time), fixed with 4% paraformaldehyde (Shanghai Ling Feng Chemicals Co., Ltd., Shanghai, China) for 10 min, and permeabilized with 0.1% Triton X-100 (Bio Basic Inc., Markham, ON, Canada) for 10 min. Following washing with PBS three times, the cells were blocked with 5% goat serum (Zhongshan Golden Bridge Biotechnology Co., Ltd.) and incubated with anti-eIF3g (1:400), anti-hnRNP U (1:200), anti-HSZFP36 (1:200) or anti-β-actin (1:200) primary antibodies overnight at 4°C. The cells were then washed with PBS three times (10 min each time) and further incubated with following secondary antibodies: Fluorescein isothiocyanate-conjugated donkey anti-rabbit IgG (1:500; 711-095-152) and/or tetramethylrhodamine-conjugated donkey anti-goat IgG (1:500; 705-025-003) (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA). The nuclei were stained with 200 µg/ml DAPI (Sigma-Aldrich). The stained cells were observed under a confocal microscope (LSM710; Carl Zeiss, Jena, Germany).

The present study hypothesized that eIF3g may link other biological processes to the initiation of translation. To confirm this, in silico analysis and subcellular fractionation was performed to examine the functional domains and subcellular distribution of eIF3g. As shown in Tables I and II, bioinformatics analyses using state-of-art online tools predicted that eIF3g has a nuclear distribution for potential functions.

To verify this prediction, the present study examined the expression of eIF3g in Bcap37 breast cancer cells by western blotting in a preliminary experiment, and the results showed a weak, but distinct band of eIF3g in the nuclear protein fraction (data not shown). The Bcap37/Tet-on-eIF3g cells with inducible overexpression of eIF3g were then used in subsequent investigations. The cells were harvested through subcellular fractionation of the cytosolic and the nuclear proteins of breast cancer cells expressing eIF3g (Bcap37/Tet-on-eIF3g). As shown in Fig. 1A, the use of this method successfully separated the cytosolic and the nuclear fractions, confirmed by detection of nuclear protein, Histone H4, in the whole cell lysate and nuclear fraction, but not in the cytoplasmic fraction. By contrast, the cytoplasmic protein, GAPDH, was detected in the whole cell lysate and cyto-solic fraction, but not in the nuclear fraction. Although eIF3g was abundant in the cytosolic fraction, it was also detected in the nuclear fraction. Immunofluorescent staining confirmed the nuclear distribution of eIF3g (Fig. 1B). This was a novel finding, considering eIF3g was originally identified as a cytoplasmic protein and as a component of the translation initiation machinery (2).

Functional nuclear proteins exert their functions via binding directly to DNA or RNA to regulate gene expression or RNA processing, or by forming a protein complex to cooperate with other nuclear proteins (15). The present study focused primarily on identifying the possible nuclear proteins, which may interact with eIF3g. The nuclear fraction of the eIF3g-expressing cells was isolated, and co-IP was formed with anti-eIF3g antibody to pull down proteins that physically interacted with eIF3g. The immunoprecipitated proteins were separated by SDS-PAGE and visualized by Coomassie blue staining (Fig. 2), and four bands were selected for subsequent mass spectrometry for protein characterization, excluding non-specific bands. As shown in Table III, the proteins were heterogeneous nuclear ribonucleoprotein U/scaffold attachment factor A (hnRNP U/SAF-A), HSZFP36/zinc finger protein 823 (ZFP823), β-actin and eIF3a, respectively.

The present study confirmed the physical interaction between eIF3g and its potential partners, hnRNP U, HSZFP36 and β-actin, by performing co-IP and in vitro crosslinking in combination with western blotting. As shown in Fig. 3A, eIF3g was co-immunoprecipitated from the nuclear fractions using antibodies against hnRNP U, HSZFP36 and β-actin, respectively. These proteins were also co-immunoprecipitated using antibody against eIF3g (Fig. 3B–D).

The results of the western blotting for the crosslinked proteins are shown in Fig. 3E, in which two specific bands were detected by the anti-eIF3g antibody; these bands were >170 kDa in molecular weight. The larger band was also detected by the anti-hnRNP U antibody, whereas the smaller band was detected by the anti-HSZFP36 or anti-β-actin antibodies. The results of the GST-pulldown assay showed that hnRNP U, HSZFP36 and β-actin were pulled down by the GST-eIF3g proteins (Fig. 3F). These results supported the interaction between eIF3g and hnRNP U, HSZFP36 or β-actin, and suggested that eIF3g, HSZFP36 and β-actin may form a protein complex.

The subcellular distributions of eIF3g, hnRNP U, HSZFP36 and β-actin were examined by immunofluorescent staining and confocal microscopy. The cells incubated with anti-eIF3g antibody showed strong fluorescent signals in the cytoplasm and weak fluorescent signals in the nucleus. Similar patterns of fluorescence were observed in the cells incubated with anti-HSZFP36 and anti-β-actin antibodies. In parallel, fluorescence was predominantly observed in the nuclei of cells incubated with anti-hnRNP U antibody (Fig. 4A). The subcellular distributions of these four proteins were also confirmed by western blotting with the fractionated cytosolic and nuclear protein samples (Fig. 4B).

As shown in Fig. 4C, confocal microscopy revealed the co-localization of eIF3g with HSZFP36 and β-actin in the nucleus, although the intensities of the signals varied due to the different abundance of the proteins in the nucleus.