Overexpressed-eIF3I interacted and activated oncogenic Akt1 is a theranostic target in human hepatocellular carcinoma


  • Potential conflict of interest: Nothing to report.

  • Supported by funding from the National Research Program for Biopharmaceuticals (NSC 101-2325-B-001-011) and the National Science Council (NSC101-2320-B-001-029-MY3) of Taiwan.


Eukaryotic translation initiation factor 3 subunit I (eIF3I) with transforming capability is often overexpressed in human hepatocellular carcinoma (HCC) but its oncogenic mechanisms remain unknown. We demonstrate that eIF3I is overexpressed in various cancers along with activated Akt1 phosphorylation and kinase activity in an eIF3I dose-dependent manner. A novel eIF3I and Akt1 protein interaction was identified in HCC cell lines and tissues and was required for eIF3I-mediated activation of Akt1 signaling. Expression of either antisense eIF3I or dominant negative Akt1 mutant suppressed eIF3I-mediated Akt1 oncogenic signaling and various other tumorigenic effects. Oncogenic domain mapping of the eIF3I and Akt1 interaction suggested that the C-terminal eIF3I interacted with the Akt1 kinase domain and conferred the majority of oncogenic functions. In addition, eIF3I interaction with Akt1 prevented PP2A dephosphorylation of Akt1 and resulted in constitutively active Akt1 oncogenic signaling. Importantly, concordant expression of endogenous eIF3I and phospho-Akt1 was detected in HCC cell lines and tissues. Treatment of eIF3I overexpressing HCC cells with the Akt1 specific inhibitor API-2 suppressed eIF3I-mediated tumorigenesis in vitro and in vivo. Conclusion: We describe a constitutive Akt1 oncogenic mechanism resulting from interaction of overexpressed eIF3I with Akt1 that prevents PP2A-mediated dephosphorylation. Overexpression of eIF3I in HCC is oncogenic and is a surrogate marker and therapeutic target for treatment with Akt1 inhibitors. (HEPATOLOGY 2013;)

Hepatocellular carcinoma (HCC) is the most common liver malignancy and the sixth leading cause of cancer deaths worldwide. The major risk factors for developing HCC are chronic viral hepatitis, alcoholism, nonalcoholic fatty liver disease, and exposure to environmental agents such as aflatoxins.1 In addition to hepatitis B virus (HBV) endemic countries in sub-Saharan Africa and Eastern Asia, the increasing incidence of hepatitis C virus (HCV) infection in Western countries and persistent HBV infection in developing countries resulted in an estimated 748,300 new liver cancers and 695,900 cancer deaths worldwide annually.2 Despite the available treatments, the overall survival of HCC patients remains poor.3 Targeted therapy with a US Food and Drug Administration (FDA)-approved multiple kinase inhibitor, sorafenib, was shown to suppress tumor growth and angiogenesis but displayed a marginal benefit, with an average prolonged survival for 3-4 months in advanced HCC patients.4 Therefore, new biomarkers and therapeutic options are urgently required for improving the survival of HCC patients.

Deregulation of messenger RNA (mRNA) translation, especially at the rate-limiting step of translational initiation, results in a global increase of protein synthesis or enhanced translation of selected oncogenic mRNAs, which may play an important role in cancer development and progression.5, 6 At least nine eukaryotic initiation factors (eIFs) were identified for initiation of protein translation. Among them, eIF4E, the best studied of the eIFs, is frequently overexpressed in human cancers and functionally correlates to cellular transformation, tumorigenesis, and metastasis.7, 8 Overexpression of eIF4E in cancer cells can stimulate phosphorylation of 4EBP through the PI3K/Akt1/mTOR pathway, release eIF4E from 4EBP sequestration to bind to the 5′ cap of mRNAs, and selectively unwind the secondary structures at the 5′ untranslated region (UTR) to enable aberrant protein translation of oncogenic and metastatic genes.9, 10 In the PI3K/Akt1/mTOR pathway, the serine-threonine protein kinase Akt1 is the major regulator of protein synthesis by regulating every stage of the multistep process of mRNA translation from ribosome biogenesis, translation initiation to translation elongation.5, 11 Aberrant activation of Akt1 has been widely reported in many cancers, and resulted in promotion of cancer cell metabolism, growth, proliferation, survival, and angiogenesis.12, 13 Current therapeutic efforts targeting aberrant translational control are mainly focused on inhibition of eIF4E hyperactivation to block initiation of translation and these approaches are currently in phase I and II clinical trials.14

The eIF3 protein complex is the largest eIF and consists of 13 subunits named eIF3a to eIF3m.15 Various eIF3 subunits were reported to be abnormally expressed in malignant tumors and demonstrated to play important roles in tumorigenesis.16, 17 eIF3I (also known as eIF3S2) was initially identified as a type II transforming growth factor beta (TGF-β) receptor interacting protein (TRIP-1) and functions as a negative regulator for the TGF-β signaling pathway.18 eIF3I overexpression is commonly observed in several cancer types.19-21 Overexpressing eIF3I (or TIF3, a mouse ortholog with 99% amino acid similarity) in NIH3T3 cells enhanced colony forming capability in vitro and tumor xenografts in nude mice.17, 22 Moreover, eIF3I displayed strong anchorage-independent growth (AIG) in HCC cells and was up-regulated in 62% (28/45) of HCC tumors as compared with adjacent nonmalignant tissues, suggesting that eIF3I plays an important role in HCC progression.19


AIG, anchorage-independent growth; eIF, eukaryotic translation initiation factor; HCC, hepatocellular carcinoma; IHC, immunohistochemistry; IP, immunoprecipitation; KM-Akt1, dominant negative mutant Akt1; Myr-Akt1, constitutively active Akt1; PP2A, protein phosphatase 2A.

Materials and Methods

Cell Lines, Constructs, Antibodies, Gene Expression Analysis, and Tumorigenic Assays.

Details and experimental protocols are provided in the Supporting Information.

In Vitro Akt1 Kinase Assay.

In vitro Akt1 kinase activity was measured by Akt1 kinase assay kit (Cell Signaling) and described in the Supporting Information.

In Vitro Akt1 Dephosphorylation Assay.

HCC Cells were transiently transfected with HA-Akt1, Flag-wild-type eIF3I, Flag-mutant eIF3I (amino acids 1-174), or vector control. After serum starvation, these transfected cells were treated with IGF-1 for 45 minutes and lysed in isotonic buffer (142.5 mM KCl, 5 mM MgCl2, 10 mM HEPES, 0.1% Triton X-100, 1 mM PMSF, and 1× protease inhibitor). Equal amounts of cell lysates were treated with anti-HA antibody overnight at 4°C for IP assay. After washes, the pellets were reacted with PP2A proteins in buffer B (20 mM imidazole, 150 mM NaCl, 14.4 mM β-mercaptoehanol, 1 mM PMSF and 1× protease inhibitor) at 30°C for 1 hour. This reaction was terminated by 6× sample buffer and western blot analysis for phospho-Akt1 expression.

PP2A Immunoprecipitation Phosphatase Assay.

PP2A phosphatase activity was measured by PP2A immunoprecipitation phosphatase assay kit (Millipore) and described in the Supporting Information.

In Vivo Xenograft Tumor Formation Assay.

The animal protocol was approved by the Animal Safety Committee of Academia Sinica. A total of 1 × 106 HepG2 cells were mixed with matrigel (356237; BD) and subcutaneously injected into BALB/c NU mice. Tumor volumes were measured weekly (volume = length × width 2 × 0.52). When tumor volume reached >20 mm3, mice were randomly divided into two groups and intravenously injected with Akt1 inhibitor API-2 (API-2/DMSO at 1 mg/kg/day) and placebo (dimethyl sulfoxide, DMSO) daily.


Overexpression of eIF3I-Activated Akt1 Signaling in Cancers.

Because eIF3I overexpression was shown in some cancers including HCC, we first performed in silico microarray analysis to detect aberrant eIF3I expression in cancers. We found that overexpression of eIF3I existed in HCC, cervical, and colon cancers versus the normal tissues and also in metastatic versus primary melanoma (Fig. 1A). To elucidate the molecular mechanism of eIF3I overexpression-mediated tumorigenesis, we examined activation of crucial HCC signal mediators including STAT3, ERK, and Akt1.23-25 Only phospho-Akt1 (Ser473) was significantly increased in an eIF3I dose-dependent manner in 293T cells (Fig. 1B). An eIF3I-induced increase in phospho-Akt1 was observed in two HCC cells, HepG2 and SK-Hep1, stably expressing eIF3I and reversed by silencing endogenous eIF3I with antisense eIF3I (eIF3I-AS) (Fig. 1C). To determine whether the eIF3I-enhanced phospho-Akt1 could further activate its kinase activity, we examined phosphorylation of direct downstream substrates of Akt1 in HCC cells including GSK-3α/β and FoxO1/FoxO3a. In these cells, increased phosphorylation forms of GSK-3α and FoxO1/FoxO3a were observed (Fig. 1D). Conversely, knockdown of endogenous eIF3I by antisense or cotransfection of eIF3I with dominant negative Akt1 (KM-Akt1, K179M) resulted in decreased expression of phospho-GSK-3α and phospho-FoxO1/FoxO3a (Fig. 1D). The eIF3I-induced Akt1 kinase activity was further confirmed by in vitro Akt1 kinase assay. Coexpression of eIF3I with Akt1 significantly activated Akt1 kinase activity, as evidenced by the increased intensity of phospho-GST-GSK-3α/β (Fig. 1E). Collectively, our results indicated that overexpression of eIF3I activates phospho-Akt1 and enhances its kinase activity in HCC cells.

Figure 1.

eIF3I overexpression enhances Akt1 phosphorylation and kinase activity. (A) eIF3I overexpression was revealed in HCC (T for tumor, n = 265; N for normal, n = 23), cervical cancer (T, n = 70; N, n = 12), colon cancer (T, n = 90; N, n = 19), and metastatic melanoma (M for metastatic tumor, n = 96; P for primary tumor, n = 60) by in silico microarray analysis. T test **P < 0.01 and ***P < 0.0001, respectively. (B) Overexpression of eIF3I increases endogenous Akt1 phosphorylation in 293T cells. (C) eIF3I-enhanced Akt1 phosphorylation could be reversed by antisense-eIF3I (eIF3I-AS) expression in HCC cells. Arrow and star symbols stand for transfected and endogenous eIF3I, respectively. (D) eIF3I overexpression increased endogenous Akt1 kinase activity and downstream signaling including phospho-GSK-3α and phospho-FoxO1/FoxO3α. Overexpression is countered by antisense-eIF3I or a dominant negative mutant of Akt1 (KM-Akt1) expression in HCC cells. Arrowhead and double stars represent HA-tagged KM-Akt1 and eIF3I, respectively. (E) In vitro Akt1 kinase assay. Expression of tubulin served as internal control.

Akt1 Kinase Activity Is Required for eIF3I-Mediated Tumorigenic Properties.

Given that overexpression of eIF3I can enhance Akt1 kinase activity, we next asked whether Akt1 plays a role in eIF3I-induced oncogenesis. Indeed, overexpression of eIF3I in HepG2 cells enhanced cell proliferation, increased colony numbers in AIG, and enlarged tumor volume in xenografts (Fig. 2A-C). In contrast, knockdown of endogenous eIF3I by eIF3I-antisense diminished the above-mentioned oncogenic properties. To further determine the involvement of Akt1 signaling pathway, coexpression of KM-Akt1 with eIF3I not only decreased the level of eIF3I-enhanced expression of phospho-GSK-3α and phospho-FoxO1/FoxO3a but also mitigated eIF3I-mediated tumorigenic effects (Figs. 1D, 2A-C). Since eIF3I was originally identified as a type II TGF-β receptor-interacting protein-1 (TRIP-1) and since TGF-β-induced apoptosis has been shown to be involved in regulation of early-stage HCC, we tested the effect of overexpressing eIF3I on TGF-β-induced apoptosis. Hep3B cells displayed apoptotic effects after serum starvation or the combination of serum starvation with TGF-β (Fig. 2D). These effects were diminished when Hep3B cells overexpressed eIF3I. Conversely, knockdown of endogenous eIF3I by eIF3I-antisense further increased cellular apoptosis. In addition, coexpression of KM-Akt1 with eIF3I reversed the eIF3I-induced antiapoptotic effect. Together, our results suggested that overexpression of eIF3I attenuated TGF-β-induced apoptosis and promoted HCC tumorigenic effects through activation of the Akt1 signaling pathway.

Figure 2.

Akt1 kinase activity is essential for eIF3I-induced oncogenic effects. HepG2 transfectants expressing eIF3I, eIF3I-antisense, or the combination of eIF3I with KM-Akt1 were used to determine their tumorigenic effects including (A) cell proliferation, (B) AIG, and (C) in vivo xenograft tumor assay (five mice per group). (D) Hep3B transfectants were treated with serum starvation or in combination with TGF-β for assays of cell apoptosis at 405 nm. Protein expression and eIF3I knockdown efficiency by antisense are shown in the left panel and tubulin served as an internal control. Arrowhead, double stars, arrow, and single star symbols represent HA-KM-Akt1, HA-eIF3I, transfected-eIF3I, and endogenous eIF3I, respectively.

eIF3I Interacts with Akt1 in an Akt1 Kinase Activity-Dependent Manner.

With the structural feature of four WD40 domains and its ability to activate Akt1 kinase activity, we examined the possibility of eIF3I modulated Akt1 activity through a protein-protein interaction. Akt1 was detected in the endogenous eIF3I immunocomplex by using the endogenous coimmunoprecipitation assay. Reciprocally, pull-down of endogenous Akt1 also coprecipitated eIF3I (Fig. 3A). The endogenous interaction of eIF3I and Akt1 was further confirmed in two human HCC tumor and corresponding adjacent nonmalignant tissue paired samples (Fig. 3B). When compared with their corresponding adjacent nonmalignant tissue, stronger expression of eIF3I, Akt1, and phospho-Akt1 were detected in tumor tissues and resulted in higher immunoprecipitated interacting proteins. Our results suggested that overexpressed eIF3I forms a complex with Akt1.

Figure 3.

Interaction of eIF3I with Akt1 in an Akt1 kinase activity-dependent manner. Endogenous interaction of eIF3I and Akt1 in (A) HpeG2, (B) two human HCC tumors (T), and their corresponding adjacent nonmalignant (N) tissues with reciprocal immunoprecipitation shown by western blot analysis. PP2A-C in Akt1-immunocomplex was to examine the influence of eIF3I overexpression to the PP2A and Akt1 interaction. (C) Treatment of CIP or insulin-like growth factor 1 (IGF-1) reduced or increased interaction of overexpressed eIF3I and Akt1 in HCC cells, respectively. (D) Transfection of kinase-dead Akt1 mutant (KM-Akt1) reduced its interaction with eIF3I in 293T cells by western blot analysis. Phospho-Akt1 and phospho-GSK3 expression served as functional controls of WT-Akt1 and KM-Akt1. Tubulin expression served as internal control.

Because Akt1 is the downstream modulator of several kinases, we next asked whether the phosphorylation level of phospho-Akt1 could further influence the interaction between Akt1 and eIF3I. First, We treated protein lysates of eIF3I and Akt1 transfected Hep3B, HepG2 and 293T cells with calf intestinal alkaline phosphatase (CIP) to dephosphorylate Akt1 and examined the protein interactions in the eIF3I immunocomplex (Fig. 3C; Supporting Fig. 1A). We found that decrease of Akt1 phosphorylation reduced eIF3I interaction with Akt1. By contrast, we treated eIF3I and Akt1 transfected HCC and 293T cells with IGF-1 (insulin-like growth factor-1) to activate Akt1 kinase activity. We found that IGF-1 treatment resulted in increased phspho-Akt1 expression, leading to increased precipitation of Akt1 in the eIF3I-immunocomplex (Fig. 3C; Supporting Fig. 1B). Finally, we asked whether Akt1 kinase activity is required for eIF3I and Akt1 interaction by comparing the eIF3I-binding ability to the wild-type Akt1 and the kinase-deficient mutant KM-Akt1. After validation of phospho-Akt1 expression and its kinase activity by measuring the downstream phospho-GSK-3α expression of wild-type Akt1 and KM-Akt1 (lower panel, Fig. 3D), we found that the Akt1 kinase-deficient mutant (KM-Akt1) displayed a significant reduction when compared with wild-type Akt1, and of eIF3I-binding capability in both the eIF3I-immunocomplex and the Akt1-immunoprecipitated complex (upper panel, Fig. 3D). Taken together, our results strongly indicated that eIF3I interacts with Akt1 through an Akt1 kinase activity-dependent manner.

Domain Mapping Reveals eIF3I C-Terminal End Possesses Akt1 Interaction, Activation, and Oncogenic Functions.

Next, we dissected the interacting domains of eIF3I and Akt1 to ask whether the interaction is essential for eIF3I-mediated tumorigenic properties (Fig. 4A,B). To dissect the interacting domains on eIF3I (Fig. 4A), a full-length and four truncated mutants of eIF3I including a.a. 1-174 (amino acid 1∼174 fragment containing clone), a.a. 1-216, a.a. 40-327, and a.a. 175-327 were constructed to examine their binding ability with Akt1. Most of eIF3I constructs were able to bind to Akt1 except a.a. 1-174 mutant, indicating that the C-terminal region (a.a. 175-327) of eIF3I is responsible for Akt1-binding (Fig. 4A). Reciprocally, a series of Akt1 truncated mutants including a.a. 1-148, a.a. 1-412, a.a. 149-412, and a.a. 149-480 were constructed and examined for the eIF3I-binding ability (Fig. 4B). The full-length and most Akt1 truncated mutants maintained eIF3I-binding ability except a.a. 1-148 construct failed to interact with eIF3I. Our results suggested that the kinase domain of Akt1 (a.a. 149-412) is required for eIF3I interaction (Fig. 4B).

Figure 4.

Akt1 interacting domain of eIF3I is essential for eIF3I-mediated tumorigenesis. Domain mapping of (A) eIF3I interacted with Akt1 and (B) Akt1 interacted with eIF3I in 293T cells. The interaction results are summarized in the schematic diagrams. In comparison to the full-length eIF3I in HepG2 cells, the C-terminal Akt1 interacting domain of eIF3I (a.a. 175-327) but not the N-terminal end of eIF3I (a.a. 1-147) retained its major capability to interact with Akt1 (C), promote cell proliferation (D), enhance colony formation (E), and finally to increase xenograft tumor volume (F). Star symbol: the expression of designated wild-type or mutant eIF3I and Akt1.

To determine the tumorigenicity of eIF3I domains, we expressed two eIF3I truncated mutants, the N-terminal Akt1-binding deficient mutant (a.a. 1-174) and C-terminal Akt1-binding mutant (a.a. 175-327) in HepG2 cells. Wild-type and C-terminal eIF3I activated phospho-Akt1 as compared to the mock control, whereas N-terminal eIF3I lost this effect (Fig. 4C). Consistently, although C-terminal eIF3I with some decreased tumorigenic effects, both wild-type and C-terminal eIF3I maintained the oncogenic capabilities to promote cell proliferation, enhance AIG ability, and increase tumor volume in nude mice but the N-terminal eIF3I lacked these properties (Fig. 4D-F). Taken together, our results demonstrate that the ability of eIF3I to interact with Akt1 determines its activation of Akt1 signaling pathway and oncogenic effects.

eIF3I Enhances Phospho-Akt1 Activation by Preventing PP2A-Mediated Akt1 Dephosphorylation.

To further understand the molecular mechanism of overexpressed eIF3I increased phospho-Akt1 expression and interaction in a kinase activity-dependent manner, we examined the potential involvement of a reported Akt1 negative modulator, a serine/threonine protein phosphatase 2A (PP2A).26 Our results showed that the incubation of immunoprecipitated-Akt1 with PP2A indeed resulted in a considerable reduction in phospho-Akt1 (Fig. 5A). However, coexpression of wild-type eIF3I disrupted PP2A-mediated decrease of phospho-Akt1 but not the Akt1-binding deficient N-terminal eIF3I (a.a. 1-174) mutant (Fig. 5A). Since PP2A has been demonstrated to form a complex with Akt1 to inactivate Akt1 kinase activity,27 we hypothesized that eIF3I inhibits PP2A-mediated Akt1 dephosphorylation through interference of Akt1-PP2A binding. As expected, eIF3I overexpression markedly blocked Akt1-PP2A binding, whereas no significant effect was found on the Akt1-binding deficient N-terminal eIF3I mutant (a.a. 1-174) (Fig. 5B). To rule out the possibility that eIF3I directly regulates PP2A phosphatase activity, we measured PP2A phosphatase activity in cells overexpressing mutant and wild-type eIF3Is. No significant difference of PP2A activity was observed between mock control, wild-type, and N-terminal (a.a. 1-174) mutant of eIF3I (Fig. 5C). PP2A-B and PP2A-C retained their capability to interact with and modulate Akt1 (Supporting Fig. 2). To further verify these results in HCC cells, we found that the eIF3I-AS knockdown eIF3I decreased Akt1-immunoprecipitated eIF3I and phospho-Akt1 but increased the pull-down of Akt1-interacted PP2A (Fig. 5D). Similarly, overexpressed eIF3I in two human HCC tumor samples decreased Akt1-immunoprecipitated PP2A when compared to that of corresponding adjacent nonmalignant liver samples (Fig. 3B, bottom panel). These results suggest that eIF3I associates with phospho-Akt1 to prevent PP2A binding and thereby abrogates PP2A-mediated Akt1 dephosphorylation (Fig. 5E).

Figure 5.

eIF3I interacted with Akt1 prevented PP2A binding and dephosphorylation of Akt1. Full-length eIF3I (WT), but not N-terminal eIF3I (a.a. 1-174), form a complex in immune-precipitated-Akt1 (A) preventing PP2A dephosphorylation of Akt1; (B) preventing association of PP2A-C to Akt1 in western blot analysis; (C) Expression of eIF3I proteins has no effect on the phosphatase activity in upper panel. Western blot analysis of the above-mentioned proteins and tubulin in 293T cell lysates served as expression and internal controls in lower panels. (D) Knockdown of eIF3I by eIF3I-antisense reduced Akt1 pull-down of eIF3I and phospho-Akt1 but increased Akt1-precipitated PP2A-C in HCC cells; (E) The model of eIF3I overexpression mediated increase of phspho-Akt1 through preventing PP2A interaction and dephosphorylation of oncogenic Akt1 and downstream signaling to promote HCC tumorigenicity. In normal cells, the level of Akt1 phosphorylation is kept in balance by its upstream kinases including PDK1 and mTORC2 and its direct phosphatase PP2A. In eIF3I-overexpressing HCC cells, eIF3I could interact with Akt1 in an Akt1 kinase activity-dependent manner. The interaction between Akt1 and eIF3I disrupted PP2A binding to Akt1. Without PP2A-mediated dephosphorylation, accumulated phospho-Akt1 will lead to constitutive activation of Akt1 and its downstream signaling to promote tumor progression.

Endogenous eIF3I and Phospho-Akt1 Are Concordantly Expressed in HCC Cell Lines and Tissues.

With the strong functional relationship between eIF3I and phospho-Akt1 established, we next explored the concordance of expression of these two proteins in HCC cell lines and tumor specimens. Of the 10 HCC cell lines that showed disconcordant expression, except SK-Hep1 and HCC36, eight demonstrated concordance of eIF3I and phosphor-Akt1 expression. Four dual high expressions (PLC5, Mahlavu, HA22T, and Hep3B) showed strong colony-forming capability and four dual low expressions (Tong, HepG2, Huh6, and Huh7) displayed fewer numbers of colonies in soft agar assays (Fig. 6A).

Figure 6.

Concordant expression of endogenous phospho-Akt1 and eIF3I in HCC cell lines and tumor tissues. (A) Protein expression of endogenous eIF3I and phospho-Akt1 in 10 HCC cell lines correlated with cellular transforming capability assayed by colony formation; (B) Expression of phospho-Akt1 and eIF3I in HCC tissues using IHC assay on tissue arrays. IHC intensity (mean ± SD) and expression correlation are presented in the table with Pearson's correlation and representative IHC images from the same HCC patients.

Next, we examined the concordant expression of eIF3I and phospho-Akt1 protein levels by immunohistochemistry (IHC) staining on 59 HCC specimens (Supporting Table 1). Strong expression of eIF3I and phospho-Akt1 were found in 76% (45/59) and 54% (32/59) HCC tumor tissues, respectively (Fig. 6B, representative IHC images of HCC tumor tissues). The high proportion of eIF3I up-regulation at the protein level is consistent with a previous study that showed higher RNA expression in 60%∼70% HCC cases with no statistical association with clinicopathological features of HCC samples.19 Interestingly, we found a significant concordant expression of eIF3I and phospho-Akt1 in the same HCC tumor tissues (Fig. 6B, r = 0.7172, P < 0.01). Together, our results suggest that phospho-Akt1 expression was significantly correlated with eIF3I protein expression in HCC cell lines and tissues.

Overexpressed eIF3I Is a Surrogate Marker for Treating HCC Patients with Akt1 Kinase Inhibitor API-2.

With strong concordance of eIF3I and phospho-Akt1 expression in HCC tissues, we tested whether overexpressed eIF3I could serve as a theranostic biomarker for suppressing eIF3I-induced oncogenic properties with Akt1 specific inhibitor API-2.28 Our results showed that the decrease of phospho-Akt1 and its interaction with eIF3I were observed in an API-2 dose-dependent manner (Fig. 7A,B), suggesting that API-2 treatment effectively inhibited eIF3I-induced activation of Akt1. Furthermore, API-2 treatment of eIF3I overexpressing HepG2 cells significantly attenuated cell proliferation in comparison to the vector control (Fig. 7C,D) and decreased colony numbers in soft agar (Fig. 7E). Moreover, treatment of API-2 significantly suppressed eIF3I overexpressing HepG2 xenograft tumor volume in nude mice (Fig. 7F). Altogether, our results indicated that Akt1 kinase activity is essential for eIF3I-exerted oncogenic effects. eIF3I is a potential theranostic biomarker for treatment of HCC patients with Akt1 inhibitor API-2 for suppression of eIF3I-induced HCC tumorigenesis.

Figure 7.

eIF3I overexpression-mediated tumorigenic effects are suppressed by the Akt1-specific inhibitor API-2. The dosage effects of the Akt1 specific inhibitor API-2 on (A) phospho-Akt1 expression and (B) eIF3I/Akt1 interaction in eIF3I-overexpressed HepG2 transfectant by western blot analysis. API-2 suppressed cell growth with (C) mock and (D) vector only controls, (E) AIG of eIF3I transfectant of HepG2, and (F) eIF3I-overexpressing HepG2 xenografted tumors in nude mice (five mice per group).


Aberrant activation of Akt1 has been reported in various cancers through studies on members of the PI3K/Akt1/PTEN pathway. In other studies, loss or down-regulation of PTEN,29 amplification or up-regulation of PIK3CA, mutation of PIK3CA or PIK3R1,30 or amplification or mutation of Akt131 were shown to activate the oncogenic Akt1 signaling pathway. In biochemical studies, novel protein interactions to the pathway members were commonly reported to efficiently regulate Akt1 kinase activity in cancers.32 For instance, HPV type 16 E7 and Caveolin-1 were demonstrated to independently interact with PP2A, inhibit PP2A binding to Akt1 for disrupting its dephosphorylation, and lead to activation of the Akt1 pathway.27 Similarly, hsp90 formed a complex with Akt1 resulting in maintenance of Akt1 kinase activity by preventing PP2A-mediated Akt1 dephosphorylation.33 In a transgenic model, hydrodynamic injection of constitutive active Akt1 into mouse liver resulted in lipogenesis activation and hepatocarcinogenesis, providing compelling evidence of aberrant Akt1 activation during liver cancer pathogenesis.34

Our finding that eIF3I was a newly discovered Akt1 binding partner raised several interesting questions on the role of overexpression of eIF3I on the function of the Akt1/PP2A axis. First, given that eIF3I could interact with Akt1, we analyzed the possibility whether eIF3I is an Akt1 substrate. The results of an in vitro Akt1 kinase assay demonstrated that GSK-3β, but not eIF3I, was phosphorylated by myr-Akt1 (constitutively active Akt1) (Supporting Fig. 3A). Consistently, the GSK-3β-immunocomplex could be recognized by phospho-Akt1 substrate antibody, whereas no signal was detected in eIF3I-immunoprecipitated complex (Supporting Fig. 3B). These results indicated that eIF3I interacts with Akt1 to constitutively activate its downstream oncogenic signaling, but eIF3I is not an Akt1 substrate. Second, since the interaction between phosphatidylinositol (3,4,5)-triphosphate (PtdIns(3,4,5)P3 or PIP3) with the PH domain of Akt1 has been shown to recruit Akt1 to the plasma membrane where Akt1 is phosphorylated by it upstream kinase PDK1, we therefore determined the subcellular localization of phospho-Akt1 in eIF3I-overexpressing cells. Akt1 protein was mainly localized in cytoplasmic fraction and was also detected in plasma membrane and nuclear fractions (Supporting Fig. 4). Coexpression of Akt1 with eIF3I showed a similar pattern of Akt1 subcellular fraction to that of Akt1 expression. Our results suggest that eIF3I may be one of the cellular factors preventing PP2A dephosphorylation of cytoplasmic phosphor-Akt1 in order to maintain and deliver Akt1 oncogenic signaling. Finally, because eIF3I/Akt1 interaction disrupted the association of PP2A to Akt1 for dephosphorylation, we also examined the possible inhibitory effects of eIF3I to PP2A. We validated the known interaction of PP2A and Akt1 (Fig. 5B) and demonstrated no influence of phosphatase activity under eIF3I overexpression (Fig. 5C). Furthermore, we showed that endogenous PP2A was precipitated in the Akt1-immunocomplex, but not in eIF3I-immunoprecipitates (Supporting Fig. 2). Therefore, our results ruled out the possibility that eIF3I-mediated activation of Akt1 signaling is through the inhibition of PP2A functions.

Although we revealed a new oncogenic mechanism of eIF3I overexpression in HCC and its potential application to other eIF3I-overexpressed cancers, further efforts will be needed to explore potential participation of other oncogenic mechanisms found in other aberrant eIFs. For example, selective translation of “weak” oncogenic mRNAs with potential secondary structures at 5′-UTR might be involving in eIF3I overexpression-mediated tumorigenic effects. Moreover, eIF3I overexpressed in HCC cells might function as a scaffold protein to recruit other Akt1 pathway kinase and effectors to enhance Akt1 kinase activity or as a shuttling protein to change Akt1 subcellular localization for promoting tumor progression.35

Since there is no drug currently available to target oncogenic eIF3I, an improved mechanistic understanding and the finding of concordant expression of eIF3I and phospho-Akt1 in HCC provide opportunities to apply eIF3I as a theranostic biomarker for treating HCC patients with the Akt1-specific inhibitor API-2. Since sorafenib is an FDA-approved multiple kinase inhibitor with therapeutic benefit to late-stage HCC patients, we examined the possible inhibitory effects of sorafenib to the eIF3I-overexpressed HepG2 cells. Our results showed no significant growth difference after sorafenib treatment between eIF3I overexpressing and vector control HepG2 cells (Supporting Fig. 5). Nevertheless, our results open a new possibility to target aberrant mRNA translation control on eIF3I and eIF4E for individual and combination therapies in cancers. Furthermore, since Akt1 has two other closed related isoforms, Akt2 and Akt3, with over 73% identity in entire amino acid sequence and over 88% identity in kinase domain (eIF3I interacting domain), we hypothesize a possibility that Akt2 and Akt3 could potentially associate with eIF3I. Since Akt2 has been reported to be involved in insulin signaling and since Akt3 is mainly expressed in brain and testis,36 we speculate that additional efforts may be warranted to study the role of eIF3I and its modulation of Akt signaling pathway in cardiovascular and neurological conditions.


We thank Dr. Adi Gazdar, MD, professor of pathology at the Hamon Center for Therapeutic Oncology Research, UT-Southwestern Medical Center, Dallas, TX, for careful and critical reading of the article.

Author contributions: Performed experiments: Y.W.W. and S.C.C.; Data analysis and interpretation: Y.W.W., K.T.L., D.L.G., C.F.C., P.H.T., and Y.S.J.; Technical support: K.T.L. and P.H.T.; Drafting of the article and study supervision: Y.W.W. and Y.S.J.