The Translation Initiation Factor eIF3i Up-regulates Vascular Endothelial Growth Factor A, Accelerates Cell Proliferation, and Promotes Angiogenesis in Embryonic Development and Tumorigenesis*

Background: Hypoxia regulates VEGFA expression at both the transcriptional and post-transcriptional levels. Results: eIf3i is induced by HIF1A under hypoxia and controls VEGFA protein translation. Conclusion: eIF3i controls VEGFA signaling during normal and tumorigenic angiogenesis. Significance: Our data reveal a new mechanism for VEGFA regulation, and eIF3i can be an oncotarget in anticancer therapies. Vascular endothelial growth factor A (VEGFA) is a critical proangiogenic factor that is activated by hypoxia at both the transcriptional and post-transcriptional levels. In hypoxia conditions, stabilized hypoxia-inducible factor 1α (HIF1A) is the key regulator for transcriptional activation of VEGFA. However, the post-transcriptional control of VEGFA expression remains poorly understood. Here, we report that the eukaryotic translation initiation factor 3i (eIF3i) is required for VEGFA protein expression in both normal embryonic and tumorigenic angiogenesis. eIF3i is dynamically expressed in the early stages of zebrafish embryogenesis and in human hepatocellular carcinoma tissues. eIF3i homozygous mutant zebrafish embryos show severe angiogenesis defects and human hepatocellular cancer cells with depletion of eIF3i to induce less angiogenesis in tumor models. Under hypoxia, the HIF1A protein can interact with its binding sequence in the eIF3i promoter and activate eIF3i transcription. The expression of VEGFA, which should rise in hypoxia, is significantly inhibited by eIF3i siRNA treatment. Moreover, eIF3i knockdown did not cause a general translation repression but specifically reduced the translation efficiency of the VEGFA mRNAs. Taken together, our results suggest that eIF3i is induced by HIF1A under hypoxia and controls normal and tumorigenic angiogenesis through regulating VEGFA protein translation.


Vascular endothelial growth factor A (VEGFA) is a critical
proangiogenic factor that is activated by hypoxia at both the transcriptional and post-transcriptional levels. In hypoxia conditions, stabilized hypoxia-inducible factor 1␣ (HIF1A) is the key regulator for transcriptional activation of VEGFA. However, the post-transcriptional control of VEGFA expression remains poorly understood. Here, we report that the eukaryotic translation initiation factor 3i (eIF3i) is required for VEGFA protein expression in both normal embryonic and tumorigenic angiogenesis. eIF3i is dynamically expressed in the early stages of zebrafish embryogenesis and in human hepatocellular carcinoma tissues. eIF3i homozygous mutant zebrafish embryos show severe angiogenesis defects and human hepatocellular cancer cells with depletion of eIF3i to induce less angiogenesis in tumor models. Under hypoxia, the HIF1A protein can interact with its binding sequence in the eIF3i promoter and activate eIF3i transcription. The expression of VEGFA, which should rise in hypoxia, is significantly inhibited by eIF3i siRNA treatment. Moreover, eIF3i knockdown did not cause a general translation repression but specifically reduced the translation efficiency of the VEGFA mRNAs. Taken together, our results suggest that eIF3i is induced by HIF1A under hypoxia and controls normal and tumorigenic angiogenesis through regulating VEGFA protein translation.
Protein synthesis plays an important role in the control of cell proliferation, and its deregulation is associated with cancer susceptibility (1). The translation initiation of specific mRNAs is one of the key steps for the regulation of protein synthesis (2). eIF3 is the largest and most complicated translation initiation factor with a molecular mass of about 550 -700 kDa, including 13 nonidentical protein subunits in mammalian cells known as eIF3a-eIF3m (3)(4)(5). eIF3 binds to the 40 S ribosome and interacts with other eIFs and mRNA, stimulating the generation of the 43 S pre-initiation complex. The eIF3 complex is critical for both 5Јcap-dependent and internal ribosome entry site (IRES) 3 -dependent translation initiation. Accumulating evidence suggests that altered expression of eIFs is associated with a wide range of human tumors and cancer cell lines (4 -9). eIF3i was initially identified as a WD40 repeat-containing protein that is bound and phosphorylated by the TGF-␤ type receptor II and modulates the TGF-␤ signaling pathway, so it was first named TGF-␤ receptor II interacting protein 1 (Trip1) (10 -12). Later, this gene was also named eIF3i, eIF3S2, or eIF3b, because it is a part of the eIF3 complex. Elevated expression of eIF3i was found in various human cancers, including breast tumor, head and neck squamous cell carcinomas, and HCC tissues (13)(14)(15). Overexpression of eIF3i in NIH3T3 cells induced malignant transformation, and the knockdown of eIF3i reversed cadmium-induced carcinogenesis (13,16). In addition to being a translation initiation factor, eIF3i can also interact with Akt1 and prevents PP2A-mediated Akt1 dephosphorylation, resulting in constitutively active Akt1 oncogenic signaling (17). Thus, these observations suggested that eIF3i is a proto-oncogene. However, how the expression of eIF3i is regulated during carcinogenesis remains unknown.
VEGFA is the major pro-angiogenesis factor in tumor angiogenesis and is enhanced in tumor cells by hypoxia (18). Rapid tumor growth leads to localized hypoxia, which induces VEGFA expression at both transcriptional and post-transcriptional levels (19). The accumulation of HIF1A is one of the major factors that trigger VEGFA transcription through binding to the hypoxia-responsive element (HRE) in the 5Ј-upstream sequence of VEGFA gene (20). At the post-transcriptional level, hypoxia increases VEGFA mRNA stability (21) and increases the IRES-directed translation of VEGFA but impairs the 5Јcap-dependent translation (22). Recently, the DEAD-box RNA helicase 6 (DDX6) has been identified as an IRES transacting factor that can particularly interact with the 5Ј-untranslated region (5ЈUTR) of VEGFA mRNA and inhibit VEGFA IRES-mediated translation under normoxic conditions. While under hypoxia, the level of DDX6 declines, and its interaction with VEGFA mRNA is diminished, which enhances VEGFA translation and promotes angiogenesis (23).
Here, we report that eIF3i is required for VEGFA protein expression in both normal embryonic and tumorigenic angiogenesis. In zebrafish, eIF3i is dynamically expressed at the early stage of embryonic development. eIF3i Ϫ/Ϫ mutant zebrafish embryos show severe angiogenesis defects, and human hepatocellular cancer cells with depletion of eIF3i induce less angiogenesis in tumor models. Besides, overexpression of eIF3i promotes colony formation, and knockdown of eIF3i inhibits cancer cell proliferation. In hepatocellular cells, hypoxic conditions can enhance eIF3i expression, and HIF1A directly regulates eIF3i transcription by binding to the HRE in the eIF3i promoter. Taken together, these findings demonstrate that eIF3i is critical for VEGFA protein expression in embryonic and tumorigenic development, offering a regulatory mechanism for eIF3i expression in cancer cells.
Cell Culture-Human embryonic kidney cell line 293T, human liver cell line Lo2, and human hepatocellular carcinoma cell line HepG2 were obtained from the American Type Culture Collection (ATCC, Manassas, VA). These cells were cultured in DMEM supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 units/ml streptomycin. All of the cells were maintained in a 37°C incubator with a humidified 5% CO 2 atmosphere, and the medium was changed every other day. CoCl 2 (Sigma) was added at a final concentration of 200 M for 5 h to mimic hypoxia. In parallel, 5% CO 2 and 95% N 2 in the cell culture incubator were also designed to mimic hypoxia. Lipopolysaccharide (LPS) (Sigma) and/or BAY 11-7082 (Alexis Biochemicals, Lausen, Switzerland) were added in cell culture medium to regulate the expression of NF-B in HepG2 cell line.
Establishment of Stable Cell Line-The eIF3i coding region was amplified by PCR with primers 5Ј-AAGAATTCGCCAC-CATGAAGCCGATCC-3Ј and 5Ј-GCTCTAGATCAAGCCT-CAAACTCAAATTCG-3Ј and was then cloned into pLVX-IRES-ZsGreen1 to get recombinant plasmid, which was transformed into Escherichia coli TOP10. After positive screening, the recombinant plasmids were transfected into 293T cells with the packaging and envelop plasmids to make lentivirus. To obtain the eIF3i overexpression cells, the lentivirus infection was conducted into Lo2 cell line, and the fluorescent cells were sorted by flow cytometry. The eIF3i knockdown cell was constructed by the infection of lentivirus-mediated eIF3i RNAi. Cells were selected by puromycin (Sigma) at 2 g/ml for 2 weeks and maintained in growth medium supplemented with puromycin (1 g/ml).
Colony Formation Assay-eIF3i overexpression cells were digested with trypsin to obtain a single cell suspension, plated onto a 10-cm dish, and incubated for 2-3 weeks. When the colony was visible to the naked eye, cell medium was removed, and the colonies were carefully washed twice with PBS. Then the colonies were fixed with 4% paraformaldehyde for 15 min and stained using 1% methylene blue. The residual methylene blue was slowly removed with flowing water. The colony formation ability was measured by comparing with the control Lo2 cells.
Histological Analysis-The human hepatocellular carcinoma tissue microarrays (TMA) were stained with immunohistochemistry for testing eIF3i expression. The TMA slides were dewaxed and rehydrated. Then high pressure cooking was conducted for antigen retrieval. After blocking with 3% hydrogen peroxide and 1% goat serum, the slides were incubated with rabbit polyclonal antibodies against eIF3i at 4°C overnight, and immunoreactivity was visualized using peroxidase/3,3Ј-diaminobenzidine. The hematoxylin was used as a counterstain. For Ki67 immunofluorescence assay, the HepG2 cells were plated in 24 wells at a density of 20,000 cells/well in a total volume of 0.5 ml of complete medium. After interfering with eIF3i siRNAs for 48 h, cell medium was removed, and the cells were fixed, blocked, incubated with Ki67 (Abcam) at 4°C overnight, and then stained with goat anti-rabbit IgG/FITC. DAPI was used to stain cell nuclei (blue). The TMAs and HepG2 cells were viewed under a fluorescence microscope (Leica Mikrosysteme Vertrieb GmbH, Bensheim, Germany).
Cell Counting-For cell counting, the embryos of wild type or eIF3i Ϫ/Ϫ zebrafish were anesthetized with Tricaine and washed using sterilized fish water. Then the embryos were exposed to proteinase K and incubated at 37°C until complete embryonic digestion. After dilution with L-15 medium, the embryo cells were counted using a hemocytometer.
Body Size Calculation-The heterozygous eIF3i zebrafish were in-crossed, and the embryos were collected. Then the chorion envelopes of embryos were removed manually at 2 dpf, and the sizes of embryos were measured with Leica fluorescence microscopy.
Acridine Orange Staining-Acridine orange staining was performed on zebrafish embryos at 3 dpf to quantify apoptosis in zebrafish embryos as described previously (29). Briefly, living embryos were incubated in a solution of 5 g/ml acridine orange (Sigma) in fish water for 30 min and washed twice with fish water. For imaging, fish were embedded in 2% methylcellulose supplemented with Tricaine (Sigma). The green nuclei were counted in the whole spinal cord of the embryos under a Leica fluorescence microscopy.
In Situ Hybridization-Antisense probes for eIF3i, cp, foxa3, and fli were synthesized with a digoxigenin labeling kit in the presence of digoxigenin (Roche Applied Science). Whole mount in situ hybridizations were performed as described previously (30). Hybridized transcripts were recognized in situ by antidigoxigenin antibodies linked to alkaline phosphatase and visualized with nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate solutions. Then the embryos were photographed using a microscope (Leica Mikrosysteme Vertrieb GmbH) with an attached camera.
Luciferase Reporter Assay-Dual-luciferase reporter assays were performed as described previously (31). Briefly, a proximal human eIF3i promoter, referred to as wild type eIF3i promoter that flanks the Ϫ593 to Ϫ24 of eIF3i gene, was cloned into pGL3 luciferase vector for reporter assays. The HIF1A-binding site in this promoter was mutated by overlap extension PCR using the primers described below. The reporter constructs were co-transfected with the pRL-CMV plasmid that contains the CMV-driven Renilla luciferase gene as internal control. 24 h after transfection, 200 M CoCl 2 was added to mimic hypoxia for 5 h. Then cells were harvested for dual-luciferase assay. Primers used to construct the eIF3i promoter were as follows: peIF3i F-593, 5Ј-GGTACCCCTTTCACAGAGACGCACCT-3Ј; peIF3i R1-24-bam, 5Ј-GGATCCGCAACGTGAGTAAGA-CCGGA-3Ј; peIF3i-MT F, 5Ј-AGCGGCTGATCGCCACCCC-ACAGAGCGG-3Ј; and peIF3i-MT R, TGGGGTGGCGA TCAGCCGCTGCTGAAGCGG.
Chromatin Immunoprecipitation (CHIP)-CHIP assays were performed as described previously with modifications (32). Briefly, the treated hepatocytes were cross-linked with 1% formaldehyde for 10 min and were stopped by the addition of glycine to a final concentration of 0.125 M. The cross-linked hepatocytes were rinsed in cold PBS and scraped. Then cells were homogenized, and nuclei were pelleted. The extract was sonicated to generate 300 -500-bp fragments. Fragmented chromatin was immunoprecipitated overnight at 4°C with 3 g of HIF1A antibody or the control isotype IgG2b (Lab Vision Corp., Fremont, CA) and then pulled down by Mabselect beads (GE Healthcare). Precipitates were washed and extracted with 1% SDS and 0.1 M NaHCO 3 . Eluates were pooled and heated. DNA fragments were purified and used as template for PCR. The primers used for the amplification of eIF3i promoter containing HRE were 5Ј-CCTTTCACAGAGACGCACCT-3Ј and 5Ј-TCAGGGTGACTCACTCGTCT-3Ј; the primers flanking the junction of the 2nd intron and 3rd exon were 5Ј-GCCTA-GAGAACGATACCTCC-3Ј and 5Ј-CCAGTGAGGACATG-CTTGGT-3Ј.
Cell Microinjection-The wild or eIF3i Ϫ/Ϫ green fluorescence-labeled HepG2 cells with stable expression of either control shRNA or eIF3i shRNA were harvested at a concentration of 10 7 cells/ml. The cell mixture was then loaded into a borosilicate glass needle pulled by a Flaming/Brown micropipette puller (Nari-shige, Japan, PN-30). About 50 -100 cells were implanted into the perivitelline space of Flk:mcherry transgenic zebrafish embryos by using an electronically regulated air-pressure micro-injector (Harvard Apparatus, NY, PL1-90). After injection, the zebrafish embryos were maintained under normal fish husbandry conditions, examined for the presence of fluorescent cells, and subsequently documented photographically.
Polysome Analysis-Polysome analysis was performed as described previously (33). Briefly, control and eIF3i shRNA HepG2 cells were treated with cycloheximide at a final concen-eIF3i Controls VEGFA Signaling during Angiogenesis tration of 100 g/ml in growth media for 15 min before being harvested. After that, the cells were washed twice with cold PBS, scraped, and collected in a 50-ml tube. Then the cells were lysed with lysis buffer containing 10 mM HEPES (pH 7.8), 150 mM NaCl, 10 mM KCl, 1 mM EDTA (pH 8.0), 2 mM MgCl 2 , 10% glycerol, 0.5% Triton X-100, 1 mM DTT, 1 IU/l RNase inhibitor. The lysates were spun, and the supernatants were transferred to premade 10 -50% sucrose gradients and centrifuged in a Beckman SW40Ti rotor at 35,000 rpm for 4 h at 4°C with an Optima L80XP ultracentrifuge (Beckman Coulter). The gradients were then manually fractioned into 80 aliquots at a volume of 150 l. The absorbance at 260 nm of each aliquot was measure by a spectrophotometer to obtain the ribosome profile. Each six aliquots were combined into one fraction to extract RNAs for electrophoresis and reverse transcription-PCR (RT-PCR) analysis. Primers used for RT-PCR analysis were as follows: VEGFA RT-F, 5Ј-CTAC-CTCCACCATGCCAAGT-3Ј, and VEGFA RT-R, 5Ј-GCA-GTAGCTGCGCTGATAGA-3Ј; GAPDH-F, 5Ј-AGAAGGCTG-GGGCTCATTTG-3Ј, and GAPDH-R, AGGGGCCATCCAC-AGTCTTC; ␤-actin-F, 5Ј-AGCGAGCATCCCCCAAAGTT-3Ј, and ␤-actin-R, 5Ј-GGGCACGAAGGCTCATCATT-3Ј. The PCR products were analyzed by DNA electrophoresis and then quantified with ImageJ software.
Statistical Analysis-Data were assayed by analysis of variance and unpaired Student's t test using SPSS statistics 18.0 analysis software. Differences between means or ranks as appropriate were considered significant when p value was Ͻ 0.05.

RESULTS
eIF3i Is Dynamically Expressed in Early Embryonic Development-Molecular mechanisms of embryonic development often overlap with those in cancer, and various signaling pathways are shared between the two processes (34,35). Although overexpression of eIF3i has been reported in several cancer types (15,17), no data are available so far on its expression in embryonic development of any model animals. We used zebrafish to analyze the expression pattern of eIF3i in embryonic development by whole mount in situ hybridization (WISH). We found that eIF3i mRNA was maternally expressed as it can be detected at the one-cell stage (Fig. 1A). The zygotic transcripts were expressed dynamically (Fig. 1, B-D). After 24 h post-fertilization, eIF3i mRNAs were expressed in the anterior neural tissue and the somites at high levels ( Fig. 1, E and F). After 2 dpf, the somite expression domain of eIF3i was dramatically reduced, and anterior neural tissue expression domain  eIF3i Controls VEGFA Signaling during Angiogenesis was limited to the retina and ependymal layer. Interestingly, eIF3i began to be expressed in endodermal organs such as liver, pancreas, and intestine at 3 dpf, and these expression domains were maintained at 4 dpf but weakened at 5 dpf (Fig. 1, G-M).
eIF3i Mutation Affects Zebrafish Embryonic Development-To assess eIF3i function in vivo, we used an eIF3i mutant line identified in a large scale zebrafish retrovirus-mediated insertional mutagenesis. This line harbors a retroviral insertion in the 2nd exon of the eIF3i gene causing a premature stop codon in the eIF3i open reading frame ( Fig. 2A). We crossed heterozygous eIF3i fishes and observed no obvious morphogenetic defects in the embryos before 24 hpf. At 2 dpf, 25% embryos (n Ͼ500) showed reduced body size and less pigment as compared with the wild type embryos (p Ͻ 0.01) (Fig. 2B). Genotyping of the incrossed embryos revealed that all the embryos with morphological defects were homozygous for eIF3i insertion (Fig. 2E). In parallel, we determined eIF3i protein levels in homozygous eIF3i zebrafish at 2 dpf and did not observe any detectable eIF3i proteins, suggesting that this line was likely an eIF3i null mutant (Fig. 2C). Because eIF3i mutation caused a small body size in zebrafish, we next sought to examine whether cell proliferation and apoptosis are abnormal in these mutant embryos. We found that eIF3i mutant embryos had decreased cell numbers as compared with wild type siblings (Fig. 2D). We also conducted the acridine orange staining and found that a higher level of apoptosis was induced in eIF3i Ϫ/Ϫ mutant embryos (p Ͻ 0.01) (Fig. 2F). The expression pattern of eIF3i during embryonic development also suggested a role of eIF3i in liver organogenesis. To address this issue, we examined the expression of two liver-specific marker cp and hepatocyte nuclear factor 3␥ (foxa3) by WISH. We found that the mRNA expression of both genes in the liver was present but greatly reduced, indicating that eIF3i mutation mainly affected the proliferation of hepatocytes and the outgrowth of the liver bud, but it did not affect the initial hepatocyte specification (Fig. 2G).
eIF3i Mutation Affects Zebrafish Angiogenesis-In comparison with normal zebrafish embryos (supplemental movie 1), eIF3i Ϫ/Ϫ mutant embryos nearly lacked blood circulation in their intersegmental regions at 2 dpf (supplemental movie 2). To assess whether mutation of eIF3i affects angiogenesis, expression of fli1, a vascular endothelial marker gene, was analyzed by WISH at 28 hpf. At this stage, no general morphological defects were observed in embryos derived from eIF3i heterozygous parents. Fli expression in the dorsal aorta and pectoral vein was not changed in all embryos examined, but its expression in the inter-segmental vessels showed defects in about 25% embryos (n ϭ 126), consistent with the Mendelian ratio (Fig. 3A). These defects ranged from morphological defects of the tip cells to lacking tip cells or even lacking both stalk and tip cells, suggesting that specification and function of tip cells were disrupted. After WISH and photography, individual embryos (n ϭ 32) were genotyped by PCR. The results showed that all the embryos with defects in fli expression were eIF3i homozygous mutants, and the rest were either eIF3i heterozygous or wild type, confirming that the phenotype was specific to eIF3i homozygous mutants. To further confirm the data, we crossed the eIF3i ϩ/Ϫ zebrafish with the Flk:GFP transgenic line expressing GFP in the vascular endothelial cells to obtain eIF3i ϩ/Ϫ ;flk:GFP ϩ/Ϫ double transgenic fishes. Then we inter-crossed the double transgenic fishes and collected the embryos for time-lapse observation using a fluorescence microscopy. At 24 hpf, we found that 25% GFP-positive embryos exhibited similar angiogenesis defects (n Ͼ100, Fig.  3B). Again, the embryos with angiogenesis phenotypes were genotyped by PCR, and the results showed that embryos with angiogenesis defects were all eIF3i homozygous mutant. We separated live 24 hpf mutant embryos from their siblings for cell number and apoptosis analysis. However, we did not find obvious differences in cell number and apoptosis levels between homozygous mutant embryos and their wild type or heterozygous siblings (data not shown). Together, these data suggested that eIF3i has specific roles in the sprouting and specification of endothelial tip cells during zebrafish embryonic angiogenesis.
Because VEGFA is the major factor that induces angiogenic sprouting and tip cell specification (36), we next analyzed whether Vegfa expression was affected in eIF3i mutant embryos. As shown in Fig. 3C, compared with wild type, eIF3i Ϫ/Ϫ mutant embryos expressed much less Vegfa proteins. We then tested whether ectopic forced expression of VEGFA could rescue angiogenesis caused by eIF3i mutation. We injected 5 pg of human VEGFA plasmids or empty pcDNA3.1 plasmids into one-cell stage embryos that were produced by in-crossing eIF3i ϩ/Ϫ ;flk:GFP ϩ/Ϫ double transgenic fishes. The presence of proper sprouting tip cells was scored to evaluate rescue. In the pcDNA3.1-injected control group at 24 hpf, angiogenesis defects were observed in 25% embryos (n ϭ 64), whereas only 6.4% embryos (n ϭ 78) showed obvious angiogenesis defects in the VEGFA-injected group (p Ͻ 0.01) (Fig. 3D), suggesting a partial rescue by VEGFA ectopic overexpression. These data demonstrated that eIF3i controls embryonic angiogenesis by regulating VEGFA expression.
eIF3i Is Highly Expressed in Human Hepatocellular Carcinoma-According to the expression pattern of eIF3i and the phenotype caused by eIF3i mutation, we reasoned that eIF3i might be associated with liver carcinogenesis. Consistent with this hypothesis, eIF3i overexpression has been observed in several cancer types, including HCC (15,17). To confirm the data, we performed immunohistochemistry analysis of eIF3i expres-FIGURE 2. eIF3i mutation affects zebrafish embryo development. A, position of retroviral insertion in the eIF3i gene was revealed by searching the ensemble database. The dark dashed ellipse indicates the insertion. B, eIF3i Ϫ/Ϫ homozygous embryos showed reduced body size (**, p Ͻ 0.01) and less pigment as compared with their siblings and wild type embryos at 2 dpf. Scale bar, 500 m. C, no obvious eIF3i protein was detected in eIF3i Ϫ/Ϫ embryos. Western blot analyses of eIF3i Ϫ/Ϫ homozygous zebrafish at 2 dpf were done with equivalent amounts of protein. D, eIF3i Ϫ/Ϫ zebrafish embryos showed significantly lower cell proliferation than wild type ones (**, p Ͻ 0.01). The wild type or eIF3i Ϫ/Ϫ zebrafish embryos were anesthetized and digested, and embryonic cells were counted using a hemocytometer. E, embryos with mutant or wild type phenotypes were subjected to genotyping. All the embryos with mutant phenotypes were eIF3i homozygous mutant, and embryos with normal phenotypes were either eIF3i heterozygous or wild type. F, eIF3i Ϫ/Ϫ mutation caused increased apoptosis in zebrafish embryos. Acridine orange staining was performed on zebrafish embryos at 3 dpf, and the green nuclei were counted in the whole spinal cord of the embryos. Obvious higher levels of apoptosis were observed in eIF3i Ϫ/Ϫ mutant embryos (**, p Ͻ 0.01). G, expression of liver-specific marker cp and hepatocyte nuclear factor 3␥ (foxa3) were examined by WISH. Both of cp and foxa3 expression domains in the liver were greatly reduced (arrows). eIF3i Controls VEGFA Signaling during Angiogenesis OCTOBER 10, 2014 • VOLUME 289 • NUMBER 41 sion in a TMA consisting of 100 human hepatocellular carcinoma tissues. We found that eIF3i proteins were very low in noncancerous tissues but increased in HCC tissues (Fig. 4, A  and B). Of 100 TMA, 66% were eIF3i protein-positive (17% weakly positive, 25% moderately positive, and 24% strongly positive, Fig. 4C). These data suggested that aberrant expres-sion of eIF3i is associated with human hepatocellular carcinoma.
eIF3i Promotes Cell Proliferation and Colony Formation-To investigate the function of eIF3i in tumorigenesis, we analyzed loss-of-function of eIF3i by transfection of two different eIF3ispecific siRNAs into HepG2 cells. Both siRNAs displayed com-  eIF3i mutation affects zebrafish angiogenesis. A and B, eIF3i Ϫ/Ϫ mutant embryos showed inter-segmental vessels defects. A, expression of endothelial marker fli of zebrafish embryos at 28 hpf was analyzed by WISH. fli expression in the dorsal aorta and pectoral vein was unchanged, but its expression in the inter-segmental vessels showed defects in eIF3i Ϫ/Ϫ embryos (arrows). B, living images showed similar angiogenesis defects to that described in A. Asterisks indicate the abnormal inter-segmental vessels. C, eIF3i Ϫ/Ϫ mutant embryos express much less Vegfa proteins than wild type embryos. D, forced overexpression of human VEGFA (hVEGFA) rescued angiogenesis defects in eIF3i Ϫ/Ϫ mutant embryos. Embryos from in-crossed eIF3i ϩ/Ϫ ;flk:GFP ϩ/Ϫ double transgenic fishes were injected with pcDNA3.1-hVEGFA or empty pcDNA3.1 plasmids. Angiogenesis defects were observed in 25% of embryos injected with pcDNA3.1, although only 6.4% embryos injected with pcDNA3.1-hVEGFA showed angiogenesis defects (**, p Ͻ 0.01).

eIF3i Controls VEGFA Signaling during Angiogenesis
parable interference efficiency as determined by Western blot. In contrast, the standard control siRNA had no effect on eIF3i expression (Fig. 5A). After transfection of siRNAs, the cell viability of HepG2 was evaluated by MTT assay at different time intervals (24,48,72, and 96 h). Transfection of either eIF3i siRNA led to reduced viable cells when compared with negative control siRNA transfection (Fig. 5B). In addition, we also evaluated cell proliferation ability by evaluating the expression level of a proliferation marker Ki-67 and found that its expression was significantly inhibited by eIF3i siRNA (Fig. 5C). These results indicated that eIF3i silencing effectively inhibited HepG2 cell proliferation.
To further explore the effect of overexpressed eIF3i in tumorigenesis, forced overexpression of eIF3i in normal liver cell line, Lo2 cell, was achieved by infection of lentivirus-expressing eIF3i-IRES-zsgreen. Empty lentivirus-expressing IRES ZsGreen was used as a negative control. Stable cell lines were sorted by fluorescence-activated cell sorting (FACS), and cells with strong green fluorescence were pooled for further experiments. The expression level of eIF3i protein in stably transfected cells was determined by Western blot (Fig. 5D) and the stable cells were then used to conduct colony formation assay. As shown in Fig. 5E, the Lo2 cells overexpressing eIF3i-IRESzsgreen formed more colonies than those expressing ZsGreen. These results suggested that overexpression of eIF3i would likely promote malignant transformation.
Expression Regulation of eIF3i by Hypoxia-Next, we sought to explore whether and how eIF3i expression is regulated in human hepatocellular carcinomas. An analysis of the proximal promoter region of the human eIF3i gene revealed that it harbors multiple cis-elements, including DNA binding sites of c-MYC, NF-B1, and HIF1A (Fig. 6A). Notably, the HIF1Abinding site in the human eIF3i promoter was accompanied by an HIF1A ancillary sequence that is necessary for a functional HRE, and this region is very conserved across mammalian spe-cies (Fig. 6E). To examine whether these factors are indeed required for eIF3i expression regulation, we determined eIF3i expression by Western blot in cells with either over-or underexpression of these factors. As expected, hypoxia conditions (95% N 2 ϩ 5% CO 2 ) or 200 M CoCl 2 treatment significantly induced eIF3i expression (Fig. 6B), and knockdown of c-Myc with two published siRNAs mildly reduced eIF3i expression (Fig. 6C). However, adjustment of NF-B signaling, either by LPS and/or BAY 11-7082, had no effect on eIF3i expression (Fig. 6D). Because eIF3i expression is more sensitive to hypoxia conditions and the eIF3i promoter region that contains HRE is very conserved during evolution, we focused our study on elucidating whether eIF3i is a direct target of HIF1A. We observed that eIF3i mRNA was up-regulated upon CoCl 2 treatment (Fig.  6F), suggesting that eIF3i transcription was regulated by hypoxia. To test whether HIF1A can bind to the eIF3i promoter through these conserved sequences, we performed ChIP analysis of the eIF3i promoter in HepG2 exposed to CoCl 2 . Using the chromatin pulled down by anti-HIF1A or normal IgG antibody as templates, a DNA fragment that flanks Ϫ593 to Ϫ395 of the eIF3i promoter containing HRE was analyzed. A region flanking ϩ1773 to ϩ1993 in the eIF3i gene locus that contains no HRE was used as a negative control. HIF1A antibody efficiently immunoprecipitated the HRE-containing region of the eIF3i proximal promoter when compared with control IgG antibody, although it did not show any specific affinity for the region that does not contain HRE (Fig. 6G). To further confirm the direct interaction of HIF1A with HRE in the proximal eIF3i promoter, we constructed luciferase reporter vectors that either contain the wild type (WT) eIF3i promoter or the mutant promoter with a 3-bp substitution in the HRE core sequence (named pGL3-peIF3i-WT and pGL3-peIF3i-MT, respectively, Fig. 6H). The reporter constructs were co-transfected with an internal control vector, pRL-CMV, into HepG2 cells, which were then subjected to normoxia or hypoxia for 5 h. The lucif- eIF3i Controls VEGFA Signaling during Angiogenesis OCTOBER 10, 2014 • VOLUME 289 • NUMBER 41 erase activity of the constructs containing wild type HRE was significantly higher than that of HRE mutant promoter in the HepG2 cell under normoxia. Under hypoxia, the wild type eIF3i promoter showed increased activity, although the mutant promoter did not (Fig. 6I). The ChIP and luciferase reporter assays demonstrated that HIF1A directly interacts with the HIF1Abinding sequence in eIF3i promoter and activates eIF3i transcription during hypoxia.
eIF3i Controls Tumorigenic Angiogenesis and Regulates Expression of VEGFA in Hypoxia-To determine whether eIF3i also helps to induce tumorigenic angiogenesis, we constructed stable HepG2 cell lines that expressed either negative control shRNA or eIF3i shRNA. The stable cell lines were selected by puromycin and FACS taking advantage of co-expression of GFP and the puromycin-resistant gene with shRNA. To evaluate the effects of eIF3i knockdown on the tumorigenic angiogenesis, the control or eIF3i shRNA green fluorescence-labeled HepG2 cells were implanted into the perivitelline space of 48-hpf flk:mcherry transgenic zebrafish embryos. The transplanted embryos bearing hepatocellular carcinoma xenografts were allowed to grow at 33°C for 1 week. As shown in Fig. 7A, the embryos bearing HepG2 tumors expressing negative control shRNA had obvious tumor angiogenesis. However, the embryos bearing HepG2 tumors expressing eIF3i shRNA had much less or eIF3i Controls VEGFA Signaling during Angiogenesis even no obvious angiogenesis induction. These results suggested that eIF3i also control tumorigenic angiogenesis.
As eIF3i controls embryonic angiogenesis by regulating Vegfa expression in zebrafish, we next sought to determine whether this mechanism also functions in human cells. We first checked the expression levels of VEGFA protein in eIF3i shRNA and control HepG2 cells. As expected, VEGFA protein expression was dramatically reduced in the eIF3i shRNA cell (Fig. 7B). In addition, because VEGFA is also a direct target of HIF1A (37), we also determined whether eIF3i depletion also reduced the VEGFA expression in the hypoxia environment. We first used eIF3i siRNAs to interfere with the eIF3i expres- had no effect on eIF3i protein expression. E, HRE in the eIF3i promoter is conserved across mammalian species. F, CoCl 2 treatment induced eIF3i mRNA level as determined by quantitative PCR analysis (**, p Ͻ 0.01). G, CHIP analysis of the eIF3i promoter revealed the direct binding of HIF1A to the HRE region in the eIF3i promoter. HIF1A antibody efficiently immunoprecipitated the HRE-containing region of the eIF3i proximal promoter when compared with control IgG antibody, but did not show any specific affinity for the region that does not contain HRE. H, schematic diagram represents luciferase reporter vectors that either contain the wild type eIF3i promoter or the one with HRE mutation. I, luciferase activity of constructs containing wild type HRE or the HRE mutant was tested under normoxia or hypoxia. The wild type eIF3i promoter showed significantly higher activity than the HRE mutant eIF3i promoter under normoxia (**, p Ͻ 0.01). Wild type eIF3i promoter showed increased activity under hypoxia, although activity of mutation eIF3i promoter was not affected (**, p Ͻ 0.01).
eIF3i Controls VEGFA Signaling during Angiogenesis OCTOBER 10, 2014 • VOLUME 289 • NUMBER 41 sion in HegG2 cells for 48 h, and these cells were then exposed to hypoxia. After 5 h, these cells were collected to analyze the expression of HIF1A, eIF3i, and VEGFA through Western blot. As shown in Fig. 7D, hypoxia induced the expression of HIF1A, eIF3i, and VEGFA. The treatment of eIF3i siRNAs effectively interferes with the expression of eIF3i even in hypoxia. Also, the expression of VEGFA, which should rise in hypoxia, is significantly inhibited by eIF3i siRNAs. According to these results, we revealed a new mechanism for VEGFA expression regulation involving eIF3i, which itself is also a target of hypoxia.
eIF3i Controls VEGFA Signaling during Angiogenesis eIF3i Silencing Reduces Translation of VEGFA mRNA-To study the detailed mechanism that eIF3i controls VEGFA expression, we examined whether eIF3i silencing decreased the transcription or the translation of VEGFA. Total RNAs of HepG2 cells expressing control or eIF3i shRNA were extracted and analyzed with quantitative PCR. As shown in Fig. 7C, the eIF3i mRNA was markedly down-regulated by eIF3i shRNA treatment (p Ͻ 0.01), although no obvious change in the VEGFA mRNA level was detected, suggesting that knockdown of EIF3i did not alter VEGFA transcription. Then we examined whether eIF3i silencing reduced VEGFA translation by using polysome analysis (Fig. 7E). Cell extracts from control and eIF3i shRNA HepG2 cells were fractionated using 10 -50% sucrose gradient to isolate polysome-bound mRNAs from ribosomefree mRNAs. Then the levels of VEGFA, ␤-actin, and GAPDH mRNAs were analyzed by RT-PCR and quantified with ImageJ software. As shown in Fig. 7, F and G, in eIF3i shRNA, treatment led to a significant shift of VEGFA mRNAs from the heavy polysome fractions toward light polysome and monosome fractions. However, no significant redistribution of ␤-actin or GAPDH mRNAs was detected in eIF3i shRNA cells. Taken together, these analyses suggested that eIF3i knockdown did not cause a general translation repression but specifically reduced the translation efficiency of the VEGFA mRNAs.

DISCUSSION
Molecular mechanisms of embryonic development are demonstrated to correlate with those in cancer, and many embryonic genes that are activated during embryonic organogenesis and silenced in adult tissues are re-expressed in malignant tumors. In this work, we studied the expression and function of eIF3i in embryonic development. We found that the expression of eIF3i is dynamic during zebrafish embryogenesis and is accumulated in endodermal organs during organogenesis. We isolated a mutant line with retrovirus inserting to the second exon of eIF3i and found that deletion of eIF3i induces angiogenesis defects characterized by the lack of or disrupted tip cells, which we demonstrated is the result of insufficient Vegfa production. WISH experiments revealed that depletion of eIF3i also results in defected liver organogenesis. Thus, our results indicated that eIF3i expression is critical for embryogenesis especially for angiogenesis and liver organogenesis.
For a long time, eIFs have been thought to be constitutively expressed housekeeping genes, and some of them, such as eIF1a, eIF2a, and eIF4a, are even considered as internal controls in RT-PCR experiments (38 -40). However, the dysregulation of eIFs expression in cancer tissues suggests that at least a part of eIFs are not constitutively expressed. Actually, the aberrant expression of eIF3i has been identified in several diseases, including a broad spectrum of tumors, such as breast, esophagus, cervix, lung, stomach, and liver cancers, suggesting expression of eIF3i is under specific control (3,17,(41)(42)(43)(44)(45). We reported that 66% (66/100) of human hepatocellular carcinoma tissues is eIF3i-positive staining. Then, we analyzed the proximal promoter of eIF3i and found several cis-elements recognized by oncogenic transcriptional factors, including c-MYC, NF-B, and HIF1A. We also showed that HIF1A and c-MYC, but not NF-B, activate eIF3i expression, which provides solid evidence demonstrating expression of eIF3i requires specific signals in cancer cells. In line with this notion, eIF3i is also expressed in a tissue-specific manner during embryogenesis.
Angiogenesis is essential for proper organ growth and repair. The imbalance in the regulation of this process correlates with inflammatory, ischemic, and malignant disorders (46). In our study, we noticed that the blood circulation is quite abnormal in eIF3i Ϫ/Ϫ mutant zebrafish embryos as some somites of these embryos have no blood flow. We found the expression of endothelial marker fli is defective in the homozygous mutant, and the specification and function of tip cells are disrupted in these embryos. Consistent with these observations, the mutant offspring of the double transgenic line (eIF3i ϩ/Ϫ ;Flk:GFP ϩ/Ϫ ) showed similar angiogenesis defects. These angiogenesis defects are quite similar to that caused by insufficient Vegf signaling. As expected, we detected that eIF3i Ϫ/Ϫ mutant embryos express much less Vegfa proteins than wild type embryos. Besides, the injection of hVEGFA plasmids effectively rescued angiogenesis defects in mutant embryos, suggesting Vegfa is a downstream target of eIF3i in angiogenesis control. In addition, tumor angiogenesis primarily relies on VEGFA-driven responses, which result in a complex tumor vasculature to coordinate tumor progression and metastasis (47). Then we tested whether eIF3i also promotes tumorigenic angiogenesis in human cancer and found that the stable knockdown deletion of eIF3i greatly reduced the expression of VEGFA in HepG2 cells. The zebrafish embryos bearing eIF3i shRNA-HepG2 tumors have much less angiogenesis induction in comparison with embryos bearing control HepG2 tumors. Together, these data demonstrated that eIF3i controls embryonic and tumorigenic angiogenesis by regulating VEGFA expression.
The expression of VEGFA is modulated by a variety of signaling, including a number of cytokines, growth factors, and steroid hormones (22). However, in solid tumors, the up-regulation of VEGFA by hypoxia is very significant (48). Hypoxia The treatment of eIF3i siRNAs effectively interfered with the expression of eIF3i in hypoxia. Also, the expression of VEGFA was significantly inhibited by eIF3i siRNAs in hypoxia. E-G, eIF3i silencing reduced translation of VEGFA mRNA as determined by polysome analysis. E, polysomes from control and eIF3i shRNA HepG2 cells were fractionated on sucrose density. Upper panels are the schematic diagram showing the 10 -50% sucrose gradients. Bottom panels are electrophoresis of RNAs extracted from the 12 fractions. F and G, VEGFA, but not ␤-actin or GAPDH mRNA, showed significant shift from the heavy polysome fractions toward light polysome and monosome fractions. VEGFA, ␤-actin, and GAPDH mRNA levels were determined by RT-PCR analysis (F) and quantified with ImageJ software (G). eIF3i Controls VEGFA Signaling during Angiogenesis OCTOBER 10, 2014 • VOLUME 289 • NUMBER 41 regulates VEGFA expression at both the transcriptional and post-transcriptional levels (19). The transcriptional activation of VEGFA in hypoxia is mainly mediated by HIF1A, which accumulates during hypoxia and activates VEGFA transcription through binding to HRE within the VEGFA promoter (37,49). Hypoxia also increases VEGFA mRNA stability and promotes VEGFA IRES-mediated translation (50,51). In this study, we revealed a new mechanism for VEGFA expression regulation involving eIF3i, which itself is also a target of hypoxia. Because eIF3i belongs to the eIF3 complex that is critical for IRES-mediated translation initiation, it is reasonable to hypothesize that elevated expression of eIF3i in hypoxia helps to promote the IRES-mediated translation initiation of VEGFA. Very recently, a study reported that eIF3i can bind to and activate Akt1 signaling. The binding of eIF3i to Akt1 prevents PP2Amediated Akt1 dephosphorylation and results in constitutively active Akt1 oncogenic signaling (17). Previous studies have revealed important roles of Akt1 in the regulation of VEGFA expression, and therefore it is also reasonable that eIF3i promotes VEGFA production through activating Akt1 signaling (52,53). In addition, it has been shown that hypoxia induces Akt1 phosphorylation, and activated Akt1 can also induce HIF1A expression (54). Consistent with this, we also observed knockdown of eIF3i reduces HIF1A expression (data not shown). Thus, this evidence places eIF3i in a positive feedback loop that involves multiple oncogenic molecules and suggests eIF3i can be an oncotarget in anticancer therapies.