AEG-1 Regulates Retinoid X Receptor and Inhibits Retinoid Signaling

Retinoid X receptors (RXRs α, β, and γ) are the master coordinators of cell growth, metabolism, and development (1, 2). RXRs heterodimerize with one-third of the 48 human nuclear receptor superfamily members, including retinoic acid receptor (RAR; ref. 3). RXR/RAR heterodimer mediates the retinoid-dependent transcription of genes involved in cell proliferation, differentiation, and apoptosis. RXRs are frequently dysregulated because of altered expression and inactivation in various cancers, including hepatocellular carcinoma (HCC), thyroid carcinoma, and prostate cancer (4–6), which might lead to compromised RAR functions in these malignancies (7–9). However, molecular mechanisms or regulators that govern the differential expression and function of RXR still remain poorly understood.

AEG-1, also known as MTDH and LYRIC, is a multifunctional oncogene that is overexpressed in a wide variety of malignancies (10, 11). AEG-1 is significantly upregulated in >90% of human patients with HCC, affecting diverse aspects of HCC pathogenesis, including proliferation, angiogenesis, invasion, and metastasis (12, 13). Although numerous studies have established clinicopatholgic correlation of AEG-1 in cancer, the underlying molecular mechanism by which AEG-1 exerts its oncogenic functions requires further clarification. AEG-1 has been shown to interact with few proteins to drive tumorigenesis (10). To identify crucial AEG-1–interacting partners, we screened a human liver cDNA library by yeast two-hybrid (Y2H) assay (14), and identified RXR as a novel interacting partner of AEG-1.

Using human HCC cell lines with endogenous low and high AEG-1 expression corresponding with low to high aggressive tumorigenic features, HCC cells with stable AEG-1 overexpression or knockdown and hepatocytes isolated from AEG-1 transgenic (Alb/AEG-1; ref. 13) and knockout (AEG-1KO) mice, we have now deciphered how AEG-1 interaction with RXRα and RXRβ drives RXR-mediated functions in HCC. Our studies uncover the mechanism by which AEG-1 upregulation affects RXR inactivation and retinoid response contributing to carcinogenesis. Moreover, AEG-1 knockdown and chemical inhibition of AEG-1–mediated signaling enhances the efficacy of retinoids by sensitizing the cells, leading to inhibited tumor growth in in vitro and xenograft models. AEG-1 inhibition might be an effective strategy to augment effects of retinoids in patients with diverse cancer indications.

Generation and characterization of a hepatocyte-specific AEG-1 transgenic mouse (Alb/AEG-1) in B6/CBA background have been described previously (13). AEG-1KO mouse was generated in C57B6/129Sv background and the procedure is described in detail in the Supplementary Information. All animal studies were approved by the Institutional Animal Care and Use Committee at Virginia Commonwealth University, and were conducted in accordance with the Animal Welfare Act, the PHS Policy on Humane Care and Use of Laboratory Animals, and the U.S. Government Principles for the Utilization and Care of Vertebrate Animals Used in Testing, Research, and Training.

HepG3, QGY-7703, THLE3, Hep3B, HuH7, and HEK-293 cells were cultured as reported earlier (12). Generation of Hep-PC-4 (control clone), Hep-AEG-1–14 (a C-terminal HA-tagged AEG-1–overexpressing clone), Hep-CTRLsi (control siRNA-expressing clone), and Hep-AEG-1si (expressing AEG-1 shRNA) in HepG3 background has been described before (12, 14). A C-terminal HA-tagged AEG-1 construct mutated at the LXXLL motif was stably expressed in HepG3 cells, Hep-AEG1-Lxxmut, and was generated following the same protocol as for Hep-AEG1–14 cells. The clones were selected and maintained in hygromycin containing DMEM.

Primary mouse hepatocytes were isolated from wild-type (WT; B6/CBA), Alb/AEG-1, WT (C57B6/129Sv), and AEG-1KO mice in the Cell and Molecular Biology Core in VCU as described previously (13) and were plated on collagen-coated dishes (BD BioCoat collagen type I; BD Biosciences) and cultured in Williams E. medium (SIGMA) containing NaHCO₃, l-glutamine, insulin (1.5 μmol/L) and dexamethasone (0.1 μmol/L). For MTT assays, a total of 1.0 to 1.5 × 10⁴ mouse hepatocytes were plated in each well of a 96-well plate and treated with retinoids and rexinoids for respective time points as mentioned in the figure legends of Fig. 2 and Supplementary Fig. S3. Cell viability was determined by standard MTT assay as described previously (12, 15, 16).

Transfections and luciferase assays were done according to the manufacturer's protocol for human HCC cells as described elsewhere (14, 17, 18) and primary hepatocytes (Supplementary Information). Each experiment was performed in triplicates and repeated three times to calculate means and SE.

Total RNA was extracted from human HepG3 cells, livers and hepatocytes of WT (B6/CBA and C57B6/129Sv), Alb/AEG-1, and AEG-1KO mouse using the QIAGEN miRNAeasy Mini Kit (QIAGEN). cDNA preparation was done using the ABI cDNA Synthesis Kit. Real-time polymerase chain reaction (RT-PCR) was performed using an ABI ViiA7 fast real-time PCR system and TaqMan gene-expression assays according to the manufacturer's protocol (Applied Biosystems). The best available TaqMan primers–probes spanning two exons for RARB, CYP26A1, NROB2, CRABP2, FOXA1, TLL1, RXRA, and RXRB for human as well as mouse were purchased from ABI.

Sheared chromatin was prepared following the manufacturer's instructions and was immunoprecipitated using RXRα (Santa Cruz Biotechnology), AHH3 (Acetyl Histone H3) and SRC-1 (steroid receptor coactivator-1; Cell Signaling Technology) antibodies. The eluted DNA and inputs were subjected to PCR for RARB and HOXA1 genes. For RARB2 (sense: 5′AGCTCTGTGAGAATCCTGGGAG3′, antisense: 5′TAGACCCTCCT GCCTCTGAACA3′) and HOXA1 (sense: 5′CTGGGG CAATCAGATTCAAACC3′, antisense: 5′CTCAGATAAACTGCTGGGACTC3′) primers were used for PCR amplification using the Taq PCRx Polymerase Kit (Invitrogen), following the manufacturer's instructions. These PCRs were performed without enhancers and repeated at least three times.

Subcutaneous xenografts were established in flanks of athymic nude mice using QGY-7703 cells (5 × 10⁵). After 1 week, these mice were injected with all-trans retinoic acid (ATRA; 10 mg/kg) or DMSO i.p., a total of seven injections, once every alternative day. One week after the final treatment, mice were sacrificed. Tumor volume was measured twice weekly with a caliper and calculated using the formula |$\pi /6\! \times\! {\rm larger\, diameter}\!\, \times\! ({\rm smaller\,\, diameter})^2$|⁠. Tumor samples were immunostained using antibodies against AEG-1, PCNA (Cell Signaling Technology), and cleaved caspase-3 (Cell Signaling Technology). All experiments were performed with 6 to 8 mice in each group and were repeated at least two times.

Data were represented as mean ± SEM and analyzed for statistical significance using one-way ANOVA followed by the Newman–Keuls test as a post hoc test. A P value of <0.05 was considered as significant.

AEG-1 is a 582 amino acid (a.a.) protein with a transmembrane domain (51–72 a.a.) and multiple nuclear localization signals. A number of AEG-1–interacting proteins have been identified that interact with its large C-terminal region (14, 17). However, a “LXXLL” motif, by which coactivators and corepressors interact with nuclear receptors (19), is located in the NH₂-terminal region of AEG-1, suggesting that this region might be important for AEG-1 functions. On the basis of this, a human liver cDNA library was screened by Y2H assay using the NH₂-terminal 1–57 a.a. of AEG-1 as bait, which lead to identification of RXRβ as an AEG-1–interacting protein (Supplementary Table S1). Coimmunoprecipitation (Co-IP) assays using lysates from the human HCC cell line QGY-7703 demonstrated pull-down of RXRα and RXRβ by anti–AEG-1 antibody and vice versa, confirming the interaction (Fig. 1A).

Deletion of a.a. 1 to 70 from N-terminus (N1) of AEG-1 and mutations in its LXXLL motif prevented the interaction of AEG-1 with RXRα and RXRβ (Fig. 1B). However, this interaction was maintained in full-length AEG-1 or upon deletion of C-terminal region (1–513), validating that RXRs indeed interact with AEG-1 at its LXXLL motif. Further to check at which domain of RXR AEG-1 interacts with, we made domain-specific deletion constructs of RXRα (Supplementary Fig. S1). Deletion of the C-terminus ligand-binding AF-2 domain of RXRα resulted in the loss of interaction, whereas deletion of its NH₂-terminus AF-1 and DNA-binding domain (DBD) did not affect the interaction of AEG-1 with RXRα (Fig. 1B). These findings demonstrated that AEG-1 binds RXRα at its ligand-binding domain (LBD), the same domain in which coactivators and corepressors also bind.

Co-IP assays using stable clones of human HCC HepG3 cells overexpressing AEG-1 (Hep-AEG-1–14), AEG-1 knockdown (Hep-AEG-1si), and control clone Hep-PC-4 cells (12) confirmed that the interaction of AEG-1 with RXRα and RXRβ is indeed AEG-1–dependent (Fig. 1C). Double immunofluorescence analysis in HepG3 cells detected colocalization of AEG-1 with RXRα and RXRβ predominantly in the nuclear and perinuclear regions, which was not observed upon AEG-1 knockdown (Fig. 1D).

RAR/RXR binds to retinoic acid response elements (RARE) on the target promoters. Basal luciferase reporter activity of RARE-containing plasmid, pGL3–RARE–luc was decreased in Hep-AEG-1–14 clone and increased in Hep-AEG-1si cells compared with control Hep-PC-4 and Hep-CTRL-si clone, respectively (Fig. 1E). RARE activity remained suppressed even upon treatment with different retinoic acid/retinoids (RA) and rexinoids (RXA) in Hep-AEG-1–14 cells, whereas it was significantly augmented in Hep-AEG-1si cells (Fig. 1E, Supplementary Fig. S2A). As a corollary, both basal and ligand-dependent RARE activity was significantly inhibited in Alb/AEG-1 hepatocytes (Fig. 1F) and amplified in AEG-1KO hepatocytes (Fig. 1G).

HepG3 cells express low levels of AEG-1 and are nontumorigenic in nude mice, whereas QGY-7703 cells express high levels of AEG-1 that generate very aggressive tumors (12). Consequently, both basal and ligand-dependent RARE activities were markedly less in QGY-7703 cells as compared with HepG3 cells (Fig. 1H). Transient overexpression of AEG-1 in HepG3 and HEK-293 cells inhibited, whereas transient knockdown of AEG-1 by siRNA augmented ligand-dependent RARE activity (Supplementary Fig. S2B).

Transient expression of full-length AEG-1 that interacts with RXRs decreased RARE activity, whereas AEG-1 deletion constructs N1 and N2 (lacking NH₂-terminal 70 and 100 a.a., respectively), and LXXLL-mutant expression construct failed to inhibit RARE activity in HepG3 (Fig. 1I) or HEK-293 cells (Supplementary Fig. S2C). AEG-1 expression level in the cells used in this study is shown in Fig. 1J. Taken together, AEG-1–RXR interaction affected RXRs negatively and AEG-1 upregulation led to a decrease in RXR-dependent RARE reporter activity. These activities were increased upon inhibiting this interaction, suggesting that AEG-1 provides a homeostatic balance to RXR functions.

Because the RAR/RXR pathway mediates antiproliferative effects of retinoids and rexinoids, we determined the extent to which AEG-1 might provide protection. Using eight different RA and RXA and their synthetic analogs at multiple doses (Supplementary Table S2), we documented that Hep-AEG1–14 cells displayed 2- to 3-fold resistance and Hep-AEG-1si cells showed enhanced sensitivity to different RA/RXA treatment at different doses up to 96 hours, when compared with control Hep-PC-4 cells (Fig. 2A; Supplementary Fig. S3A–S3H). Colony formation assays also confirmed marked protection from RA- and RXA-mediated inhibition of cell growth in Hep-AEG-1–14 cells and potentiation of cell growth inhibition in Hep-AEG-1si cells (Fig. 2B). Similarly, QGY-7703 cells with basal high AEG-1 expression displayed resistance toward RA/RXA as compared with HepG3 (Supplementary Fig. S3Q).

WT hepatocytes showed <25% cell growth inhibition at 48 hours, whereas only <5% in Alb/AEG-1 hepatocytes, even at high doses of different RA/RXA (data not shown). At 96 hours, WT hepatocytes showed variable cell viability with different doses of retinoids and rexinoids. However, Alb/AEG-1 hepatocytes were noticeably protected from inhibition of cell growth with >80% viability (Fig. 2C). Conversely, AEG-1KO hepatocytes demonstrated marked decrease in cell growth with retinoids and rexinoids at 48 to 96 hours in a dose-dependent manner (Fig. 2D; Supplementary Fig. S3I–S3P). We next tested whether AEG-1 had any effect on retinoid-mediated killing in other cancer cells such as human acute myelogenous leukemia (AML) HL-60 cells. Notably, profound resistance was obtained with AEG-1 transient expression, and decreased cell viability was observed with AEG-1 knockdown upon treatment of different RA/RXA (Fig. 2E).

Relative mRNA levels were measured for representative RAR/RXR target genes, RAR beta (RARB), Cytochrome P450 26A1 (CYP26A1), nuclear receptor subfamily 0, group B, member 2 (NR0B2), Homeobox A1 (HOXA1), cellular retinoic acid–binding protein 2 (CRABP2), and Forkhead box protein A1 (FOXA1). Hep-AEG-1–14 cells showed significant downregulation of all these genes at basal level and post 9-cis reinoic acid (9CRA) or ATRA treatment, when compared with control Hep-PC-4 (Fig. 3A and Supplementary Fig. S4A). Similar decrease and corresponding increase in the above-mentioned genes were also observed in Alb/AEG-1 and AEG-1KO hepatocytes, respectively (Fig. 3B and C and Supplementary Fig. S4B–S4C). The RA negatively regulated gene, Tll1 showed significant upregulation in Alb/AEG-1 hepatocytes with or without ligand, and a reduction in AEG-1KO hepatocytes at basal level (Supplementary Fig. S4B–S4C). Downregulation of RA-responsive gene, RARB, which determines cell growth and apoptosis, was confirmed in additional Alb/AEG-1 mice (Supplementary Fig. S4D).

Chromatin immunoprecipitation (ChIP) assays were performed using the RA target genes RARB and HOXA1. In the absence of ligand, significantly decreased and increased recruitment of RXRα was observed in both target gene promoters in Hep-AEG-1–14 and Hep-AEG-1si cells, respectively, as compared with Hep-PC-4 cells (Fig. 3D and Supplementary Fig. S4E). Similar effects were also observed for (AHH3) and SRC-1. Upon treatment with 9CRA or ATRA, RXRα recruitment did not change substantially, whereas both AHH3 and SRC-1 recruitment was significantly augmented in all three cell lines. However, ligand-induced recruitment of AHH3 and SRC-1 was notably less in Hep-AEG-1–14 cells and more in Hep-AEG-1si cells (Fig. 3D and Supplementary Fig. S4E). To check whether AEG-1 interferes with DNA-binding property of RXR, EMSA was performed in a cell-free system using in vitro translated RARβ, RXRα, and AEG-1 using a DR5 RARE response element as probe. RARβ/RXR heterodimer efficiently bound to DNA in the absence or presence of ATRA, which was not affected by addition of AEG-1 (data not shown). These findings suggest that AEG-1 does not interfere with DNA-binding or heterodimerization of RXR with its partners rather it might modulate coactivator/corepressor recruitment. The decreased recruitment of RXRα to RARB or HOXA1 promoter in Hep-AEG-1–14 cells might be due to reduced levels of nuclear RXRs in these cells, a hypothesis we will address in subsequent sections.

Next, we measured RARE reporter activity in control and AEG-1–overexpressing cells with coexpression of a corepressor Silencing mediator for retinoid/thyroid hormone receptors (SMRT) or a coactivator, SRC-1. SMRT overexpression led to a minimal but significant reduction in RARE activity (Fig. 3E). Interestingly, SRC-1 overexpression profoundly elevated RARE activity in Hep-AEG-1–14 cells similar to Hep-PC-4 cells without SRC-1 overexpression (Fig. 3E). Histone deacetylase inhibitor, Trichostatin A (TSA), markedly induced basal RARE activity in a dose-dependent manner (Fig. 3F). Collectively, these findings suggest that overexpressed AEG-1 competes with the coactivators to bind RXR by interacting through the LXXLL motif maintaining histones in a deacetylated state.

Co-IP analysis using nuclear and cytoplasmic fractions demonstrated that AEG-1 interaction with RXRs mainly occurs in the cytoplasm of Hep-AEG-1–14 cells and in the nucleus in control cells (Fig. 4A). Immunofluorescence studies revealed that AEG-1 resides mainly in the nucleus of control Hep-PC-4 cells and WT hepatocytes, which show low AEG-1 expression (Fig. 4B and C and Supplementary Fig. S5). However, in Hep-AEG-1–14 cells and Alb/AEG-1 hepatocytes, AEG-1 displays predominantly cytoplasmic localization (Fig. 4B and C and Supplementary Fig. S5), which has been attributed to the enhanced monoubiquitination and stabilization (13, 20). Interestingly, AEG-1 overexpression and cytoplasmic translocation also resulted in the coordinated translocation of RXRα and RXRβ from nucleus to the cytoplasm in Hep-AEG-1–14 cells. Although in Hep-PC-4 cells, AEG-1/RXR interaction took place predominantly in the nucleus (Fig. 4B and Supplementary Fig. S5). As nuclear receptors shuttle between the cytoplasm and the nucleus, we determined whether cytoplasmic retention of RXRs might depend on ligand. Treatment with 9CRA or ATRA at various doses up to 24 hours, failed to induce nuclear translocation of RXRα and RXRβ in AEG-1–overexpressing cells, suggesting that AEG-1 entraps RXRs into cytoplasm obstructing their nuclear import (Fig. 4B and Supplementary Fig. S5).

In WT hepatocytes, RXRα is distributed throughout the cell but more in the nucleus, and RXRβ showed universal staining. 9CRA treatment induced the import of RXRα and RXRβ into the nucleus (Fig. 4C). In contrast, both RXRα and RXRβ were mostly cytoplasmic in Alb/AEG-1 hepatocytes and 9CRA treatment did not alter this profile significantly (Fig. 4C). Dominant to complete cytoplasmic localization of AEG-1 and RXRs was dependent upon AEG-1 expression level (Supplementary Fig. S6A–S6E). Cytoplasmic translocation of AEG-1 and RXR was not affected upon culturing them on collagen–fibronectin matrix and synchronization (Supplementary Fig. S6F–S6G).

Double immunofluorescence analysis of AEG-1 with RXRα and RXRβ was performed using HepG3 stable clone overexpressing AEG-1 mutated at the LXXLL motif (Hep-AEG1-LXXmut). Interestingly, both RXRα and RXRβ remain nuclear and do not translocate to cytoplasm in these cells (Fig. 4D), confirming that AEG-1 indeed brings RXRs to the cytoplasm, which is prevented upon losing the interaction.

Both RXRα and RXRβ levels were significantly higher in HCC cells compared with normal immortal human hepatocytes (THLE-3) cells (Fig. 5A). Moreover, truncated form of RXRα (tRXRα, arrow in Fig. 5A) accumulated more in different HCC cell lines. RXRα and RXRβ were also upregulated in AEG-1–overexpressing systems: Hep-AEG-1–14 cells and Alb/AEG-1 hepatocytes and downregulated in AEG-1 knockdown cells (HepAEG-1si; Fig. 5B and Supplementary Fig. S7A). However, their mRNA levels were reduced in Hep-AEG-1–14 cells and Alb/AEG-1 hepatocytes (Fig. 5C and Supplementary Fig. S7B–S7C). These findings suggested potential posttranslational modification of RXRs by AEG-1. Interestingly, phosphorylation of RXRα and RXRβ at serine residues was conspicuously induced in different AEG-1 overexpression systems: Alb/AEG-1 hepatocytes, QGY-7703, and Hep-AEG-1–14 cells (Fig. 5D and E). Phosphorylation was inhibited in Hep-AEG-1si cells in comparison with control cells, which might reflect the reduction of total RXR (Fig. 5D). More phosphorylated RXRα was detected in both the nucleus and cytoplasm of Hep-AEG-1–14 cells although phospho forms of tRXRα and RXRβ remained prevalent in the nucleus (Supplementary Fig. S7D). Inhibition of nuclear phosphorylated RXRα and RXRβ in Hep-AEG-1si cells might explain increased RARE reporter activity, expression of RAR/RXR target genes, and apoptosis (Fig. 1 and 2).

Different kinases have been reported to phosphorylate RXRα in HCC (21–23). We tested which kinase might be responsible for AEG-1–dependent RXR phosphorylation in HCC cells by using specific inhibitors (Supplementary Table S2). Inhibition of PKA, PKC, JNK, ERK, and p38MAPK led to downregulation of pRXRα and p-tRXRα in AEG-1–specific manner (Fig. 5F and G and Supplementary Fig. S7E–S7G). pRXRβ was also reduced to a greater extent in Hep-AEG1–14 cells upon inhibiting PI3K, JNK, ERK, p38MAPK, and TK (Fig. 5F and G and Supplementary Fig. S7E–S7G). RXRΔN antibody detecting all isotypes of RXR showed none to minimal effect of the inhibitors on the full-length RXR. However, RXRα-specific antibody detected inhibition of both pRXRα and p-tRXRα (Fig. 5F and G). Moreover, combined inhibition of PI3K and p38MAPK synergistically decreased phosphorylation of tRXRα.

RARE activity was significantly augmented in AEG-1 stable and transient overexpressing cells upon inhibition of ERK, PKA, and p38MAPK (Fig. 5H and Supplementary Fig. S8A–S8B). ERK has been shown to phosphorylate RXRα at the serine 260 residue in HCC, and inhibition of ERK-mediated RXR phosphorylation led to most increase in RARE activity in a dose- and AEG-1–dependent manner (Supplementary Fig. S8C). We also checked the expression of representative RXR/RAR target genes; RARB, HOXA1, CYP26A1, and NR0B2 upon inhibition of various kinases (Fig. 5I and Supplementary Fig. S8D). ERK inhibition retrieved the expression of all the genes at various extents in AEG-1–overexpressing cells. PKA and p38MAPK inhibitors specifically increased the expression of RARB and HOXA1/FOXA1, respectively. Notably, siRNA mediated knockdown of AEG-1 in HepG3 (∼1.3-fold) and QGY-7703 cells (∼2-fold) and RXRα overexpression in only QGY-7703 cells augmented ligand-dependent RARE activity (Supplementary Fig. S8E), confirming the AEG-1 caused inactivation of RXR in tumorigenic HCC cells, which was significantly retrieved by AEG-1 knockdown or RXRα overexpression.

The in vitro growth-suppressing function of retinoids upon AEG-1 knockdown was extended by in vivo assays. Because, QGY-7703 cells develop aggressive tumors in nude mice, we transduced them with none, control lentivirus–expressing scrambled shRNA (shControl) and AEG-1 knockdown lentivirus (shAEG-1; ref. 24). These cells were then subcutaneously xenografted into the flanks of male athymic nude mice. One week after implantation, when tumor is well established, the mice were treated with i.p. injections of DMSO (vehicle) or ATRA (10-mg/kg body weight). In DMSO-treated-only and DMSO-treated shControl groups, tumor gradually increased in size as reflected by tumor volume measured twice a week, and tumor weight at the final day of experiment (Fig. 6A–C). In ATRA-treated groups, only two of the 8 mice showed reduction in tumor growth, which was not significant. In the shAEG-1 DMSO–treated animals, the tumor growth was significantly suppressed as compared with controls and the ATRA-treated group. Interestingly, the shAEG-1 ATRA-treated animals grew none to minimal tumors in size and weight (Fig. 6A–C). Three of 8 shAEG-1 ATRA–treated mice showed no tumor or smaller tumor than 0 day of treatment. Their combination effect was calculated as 0.78 (<1 = synergistic), indicating synergism. Histologic analysis of the tumors revealed significantly increased necrosis/apoptosis in the shAEG-1 ATRA–treated group compared with others (Fig. 6D top). Immunohistochemical studies confirmed knockdown of AEG-1 in shAEG-1 groups (Fig. 6D and E). In addition, there was marked downregulation of the cell proliferating marker PCNA and upregulation of apoptosis marker cleaved caspase-3 in tumors derived from shAEG-1 ATRA, shAEG-1, and ATRA, respectively, as compared with the DMSO control and shControl DMSO groups (Fig. 6D and E).

Molecular mechanism(s) by which RXR controls diverse cellular functions, including cell proliferation and metabolism, has been thoroughly investigated. However, only a few reports have addressed the molecular regulators of RXR itself, which might be indispensable in evaluating and improving retinoid- and rexinoid-based chemotherapy. Our current observations suggest that AEG-1 negatively regulates functions of RXRs by interacting through the LXXLL motif at the AF-2 ligand-binding domain of RXR, thereby interfering with coactivator recruitment and subsequent histone acetylation. AEG-1 is a highly evolutionary conserved gene present only in vertebrates. In rodents, the “LXXLL” motif is present as “LRELL” whereas in primates, it is present as “LREML.” This change from leucine to methionine in primates might affect binding affinity between AEG-1 and RXRs and determine the strength, degree, and duration of inhibition.

We document that AEG-1 not only inhibits RXR-mediated transcriptional regulation, but also induces RXR phosphorylation via ERK and p38MAPK pathways, which are known to be activated by AEG-1 (12). Phosphorylated RXRs are resistant to ubiquitin proteasome–mediated degradation and act dominant negatively for normal RXR functions by abrogating heterodimerization and coactivator(s) recruitment (25, 26), which is indispensable to drive the oncogenic functions. Therefore, inhibition of these molecules suppressed AEG-1–mediated phosphorylation of RXR and rescued transcriptional activation of genes regulating cell proliferation in HCC. RXRα is proteolytically degraded generating N-terminally truncated 44-kDa receptor (tRXRα) that interacts with the p85α subunit of PI3K and activates protumorigenic PI3K/AKT signaling (27). Activation of the PI3K/Akt pathway regulates AEG-1–mediated resistance to apoptosis (28), and AEG-1–induced tRXRα generation might lead to PI3K/Akt activation.

Our findings reveal that overexpressed cytoplasmic AEG-1 in tumorigenic cells leads to dominant to exclusive cytoplasmic translocation of RXRα and RXRβ. RXRs shuttle between the cytoplasmic and nuclear compartments during certain stages of development, in response to ligand, differentiation, apoptosis, and inflammation (29–32). Furthermore, tRXRα has been shown to be exclusively cytoplasmic (27). Overall, RXRs might be regulated by AEG-1 through histone (de)acetylation in normal cells and via cytoplasmic entrapping and phosphorylation in human cancer cells. It should be noted that treatment with ERK inhibitor increased RA-dependent gene expression but did not increase nuclear translocation of RXR (data not shown). We observed phosphorylation of both nuclear and cytoplasmic RXR in AEG-1–overexpressing cells (Supplementary Fig. S7D). ERK inhibition might dephosphorylate nuclear RXR, leading to its increased activity without resulting in nuclear translocation of cytoplasmic RXR, which remains entrapped in the cytoplasm because of physical interaction with AEG-1.

One important consequence of RXRs inhibition by AEG-1 is abrogation of RA-induced gene transcription. Inactivation or downregulation of RARB and other RAR/RXR downstream genes have been reported in various malignancies with increased cell survival and augmentation of disease (8). A transgenic mouse expressing dominant negative RARα develops spontaneous HCC (9). By negatively regulating the retinoid-inducible genes, such as RARB and CRABP2, important regulators of normal cell growth, AEG-1 provided significant protection from cell growth inhibition by different retinoids/rexinoids whereas AEG-1 knockdown potentiated these effects. Similar level of AEG-1–mediated resistance in HL-60 cells provided the importance of AEG-1 in determining response to retinoid, a frontline therapy for leukemia.

Retinoids and rexinoids have been evaluated as candidates for cancer chemoprevention from past two decades (33, 34); however, there have been multiple drawbacks and limitations. One important factor is that retinoid signaling is often lost or compromised in carcinogenesis, leading to retinoid ineffectiveness and resistance (35–37). An important reason is the inactivation/inhibition of RXR and/or RAR pathways in various cancers, including breast cancer, HCC, and AML (36–38). Combination of AEG-1 inhibition with ATRA treatment resulted in tremendous synergistic inhibition of tumor growth in HCC xenograft assays, and demonstrated reduced proliferation and increased apoptosis. Targeting AEG-1 could enhance the efficacy of retinoids- and rexinoids-based therapeutics overcoming drawbacks in HCC and other malignancies, including leukemia. This hypothesis needs to be confirmed using orthotopic xenograft models in nude mice and endogenous tumor models.

In summary, a schematic representation is presented to describe AEG-1–driven regulation of RXRα and β (Fig. 7). Generally, RXR heterodimerizes with RAR and binds to target gene promoters recruiting the corepressor complex in the absence of ligand, and coactivators in the presence of ligand, thus regulating transcription of target genes for normal cell proliferation and apoptosis. AEG-1 in the nucleus of normal nontumorigenic cell balances this phenomenon by interfering with coactivator recruitment. When AEG-1 increases and accumulates in the cytoplasm of tumorigenic cells, this equilibrium is perturbed. First, AEG-1 translocates and entraps RXRα and β into cytoplasm obstructing RXR binding and transcriptional activation of target genes. Second, AEG-1 induces phosphorylation of RXRs that act dominant negatively on normal RXR. Finally, RXRα truncation is also elevated by AEG-1. All together, AEG-1 upregulation causes inactivation of RXRs, thereby negatively affecting downstream signaling, thus favoring unrestrained cancer cell proliferation leading to HCC and other cancers.

No potential conflicts of interest were disclosed.

Conception and design: J. Srivastava, P.B. Hylemon, D. Sarkar

Development of methodology: N.D. Mukhopadhyay, D. Sarkar

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.L. Robertson, D. Rajasekaran, R. Gredler, S. Ghosh, K. Shah, D. Bhere, M.A. Subler, J.J. Windle, D. Sarkar

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Srivastava, D. Rajasekaran, A. Siddiq, N.D. Mukhopadhyay, P.B. Hylemon, M.A. Subler, D. Sarkar

Writing, review, and/or revision of the manuscript: J. Srivastava, N.D. Mukhopadhyay, P.B. Hylemon, M.A. Subler, J.J. Windle, P.B. Fisher, D. Sarkar

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Gredler, L. Emdad, N.D. Mukhopadhyay, D. Sarkar

Study supervision: P.B. Fisher, D. Sarkar

Other (performed and analyzed experiments): J. Srivastava

Other (supplied reagents and carried out discussion): G. Gil

This study was supported in part by grants from The James S. McDonnell Foundation and National Cancer Institute Grant R01 CA138540 (D. Sarkar) and NIH grants R01 CA134721 (P.B. Fisher). D. Sarkar is the Harrison Endowed Scholar in Cancer Research and Blick scholar. P.B. Fisher holds the Thelma Newmeyer Corman Chair in Cancer Research.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.