Estrogen receptor α promotes protein synthesis by fine-tuning the expression of the eukaryotic translation initiation factor 3 subunit f (eIF3f)

Approximately two thirds of all breast cancer cases are estrogen receptor (ER)–positive. The treatment of this breast cancer subtype with endocrine therapies is effective in the adjuvant and recurrent settings. However, their effectiveness is compromised by the emergence of intrinsic or acquired resistance. Thus, identification of new molecular targets can significantly contribute to the development of novel therapeutic strategies. In recent years, many studies have implicated aberrant levels of translation initiation factors in cancer etiology and provided evidence that identifies these factors as promising therapeutic targets. Accordingly, we observed reduced levels of the eIF3 subunit eIF3f in ER-positive breast cancer cells compared with ER-negative cells, and determined that low eIF3f levels are required for proper proliferation and survival of ER-positive MCF7 cells. The expression of eIF3f is tightly controlled by ERα at the transcriptional (genomic pathway) and translational (nongenomic pathway) level. Specifically, estrogen-bound ERα represses transcription of the EIF3F gene, while promoting eIF3f mRNA translation. To regulate translation, estrogen activates the mTORC1 pathway, which enhances the binding of eIF3 to the eIF4F complex and, consequently, the assembly of the 48S preinitiation complexes and protein synthesis. We observed preferential translation of mRNAs with highly structured 5′-UTRs that usually encode factors involved in cell proliferation and survival (e.g. cyclin D1 and survivin). Our results underscore the importance of estrogen-ERα–mediated control of eIF3f expression for the proliferation and survival of ER-positive breast cancer cells. These findings may provide rationale for the development of new therapies to treat ER-positive breast cancer.

Breast cancer is the leading cause of cancer-related deaths among females worldwide. Approximately 70% of breast cancers are estrogen receptor (ER) 2 positive, which underscores the dependence of cancer cells on estrogen for growth and survival (1). ER-positive breast cancers are usually treated with endocrine therapies that inhibit ER function either by antagonizing the binding of estrogen to ER (selective estrogen receptor modulators, e.g. tamoxifen), promoting ER degradation (selective estrogen receptor degraders, e.g. fulvestrant), or blocking estrogen biosynthesis (aromatase inhibitors, e. g. letrozole, anastrozole, and exemestane) (2). However, their effectiveness is compromised by the emergence of intrinsic or acquired resistance in treated patients (2)(3)(4). Therefore, better understanding of ER-positive breast cancer biology is critical to development of more effective therapeutic strategies that minimize resistance and cancer recurrence.
ER␣ is a nuclear receptor whose activity is primarily regulated by the binding of its ligand estrogen (17␤-estradiol). Estrogen-ER␣ complex acts as a transcription factor that activates or represses the expression of multiple target genes (genomic pathway) (5,6). Alternatively, extranuclear ligandbound ER elicits rapid, stimulatory effects on cytoplasmic signal transduction pathways mediated by the mitogen-activated protein kinase (MAPK)/ERK or the phosphatidylinositol 3-kinase (PI3K)/AKT/mTOR complex 1 (mTORC1), also termed the nongenomic pathway (7). By acting through these signaling pathways, increased levels of estrogen-ER␣ complex promote cell proliferation, cell cycle progression, survival, angiogenesis, invasion, and migration in cancer cells.
Regulation of mRNA translation is critical to define the proteome, maintain homeostasis, and control cell proliferation, growth, and development. Protein synthesis occurs in four steps: initiation, elongation, termination, and ribosome recycling, with initiation being the rate-limiting phase (8). The translation initiation comprises: (a) the assembly of the eukaryotic translation initiation factor 4F complex (eIF4F) composed of the cap-binding protein eIF4E, the RNA helicase eIF4A, and the scaffolding protein eIF4G on the 5Ј cap-structure of cellular mRNAs; (b) the formation of the 43S preinitiation complex (PIC) that consists of the 40S ribosomal subunit, translation initiation factors eIF1, eIF1A, eIF3, eIF5, and the ternary complex that includes eIF2, initiator Met-tRNA i Met , and GTP; (c) . The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This article contains supporting "Experimental procedures," Figs. S1 and S2, and Table S1. 1 To whom correspondence should be addressed. Tel.: 914-594-4110; E-mail: mholz@nymc.edu. 2 The abbreviations used are: ER, estrogen receptor, MAPK, mitogen-activated protein kinase; ERK, extracellular signal-regulated kinase; PIC, preinitiation complex; YLC, yeast-like core; ERE, estrogen response element; MTT, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide; mTOR, mechanistic target of rapamycin; RT-qPCR, quantitative reverse transcription PCR.
cro ARTICLE the recruitment of the 43S PIC to the mRNA via the interaction of eIF3 with eIF4G to form the 48S PIC; (d) the scanning of the 5Ј-UTR; and (e) the assembly of the 80S ribosome-initiation complex at the AUG start codon (9). Translation initiation is tightly regulated by the oncogenic MAPK/ERK and PI3K/AKT/ mTORC1 signaling pathways (10). The mechanisms by which mTORC1 controls this phase have been well-studied. They involve (a) phosphorylation of translational inhibitor 4E-binding protein 1 (4E-BP1), which prevents its interaction with eIF4E and facilitates eIF4E binding to eIF4G to form the eIF4F cap-binding complex, and (b) eIF3-mediated activation of mTOR effector, the 40S ribosomal kinase 1 (S6K1), which results in enhanced assembly of the 48S PIC and, therefore, increased protein synthesis (10,11).
Dysregulation of mRNA translation is observed in cancer. Altered levels of translation initiation factors and/or the activation of the upstream oncogenic pathways increase global protein synthesis and/or the translation of specific mRNAs. These translational changes promote malignant transformation and tumor development (10,(12)(13)(14)(15). Many studies have demonstrated the contribution of altered eIF4F levels and activity to cancer etiology. Targeting eIF4F components or the upstream signaling pathways (e.g. PI3K and mTOR inhibitors) has shown promising results in preclinical studies and clinical trials (16 -19). However, ineffectiveness or resistance in monotherapy setting and/or toxicity in combination limited their clinical utility.
Here we focus on expanding our understanding of the role of the translation initiation factor eIF3 in tumor biology. eIF3 is a large complex composed of 13 nonidentical subunits (eIF3a-m) in human cells. Current model proposes that the assembly of eIF3 starts with the interaction of eIF3a and eIF3b to form the eIF3 nucleation core. The association of eIF3g and eIF3i to eIF3b gives rise to the subcomplex known as yeast-like core (YLC). Then, the sequential interaction of the seven subunits eIF3c, eIF3e, eIF3f, eIF3h, eIF3k, eIF3l, and eIF3m with eIF3a forms the eIF3 octamer. The nonoctameric eIF3d subunit joins eIF3 complex through its binding to eIF3e (20). Overexpression of eIF3a, eIF3b, eIF3c, eIF3h, eIF3i, and eIF3m, or underexpression of eIF3e and eIF3f have been reported in several cancers, including breast tumors (12,15,21). First evidence supporting a role for eIF3 in cancer biology was obtained by ectopic overexpression of individual subunits in NIH3T3 cells. Ectopic expression of eIF3a, eIF3b, eIF3c, eIF3h, and eIF3i stimulates global protein synthesis and translation of mRNAs that encode growth-regulating proteins, and leads to malignant transformation. In contrast, ectopic expression of eIF3e and eIF3f inhibits protein synthesis and decreases cell growth and proliferation (22). Recent studies have demonstrated that changes in the levels of a single eIF3 subunit can affect the expression of other subunits and result in the formation of eIF3 subcomplexes responsible for the translation of specific set of mRNAs (23). Additionally, Cate and colleagues (24,25) have shown that eIF3 binds to a specific group of mRNAs involved in cell growth control and promotes their cap-dependent translation independently of eIF4E activity. These results indicate that altered levels of individual eIF3 subunits and, therefore, the formation of the eIF3 complex and/or subcomplexes significantly define the translational landscape of cancer cells.
To investigate the role of eIF3 in breast cancer biology, we have determined the levels of eIF3a, eIF3b, and eIF3f in a panel of breast cancer cells. Interestingly, we observed lower levels of eIF3f in ER-positive cells compared with ER-negative cells. Our studies demonstrate that genomic and nongenomic estrogen-ER␣ pathways coordinately fine-tune the expression of eIF3f in ER-positive MCF7 cells. This tight regulation buffers the levels of eIF3f to ensure the synthesis of factors required for estrogendependent cell proliferation and survival.

Levels of eIF3f are significantly reduced in ER-positive compared with ER-negative breast cancer cells
Altered amounts of individual eIF3 subunits have been detected in breast tumors (15,21,26,27). Particularly, elevated levels of eIF3a and eIF3b and reduced levels of eIF3f are observed in breast cancers (21). eIF3a and eIF3b subunits serve as the nucleation core around which other subunits assemble to form the eIF3 complex, and eIF3f plays a critical role in stabilizing the complex (23). To investigate the role of these proteins in breast cancer biology, we first evaluated their expression in a panel of ER-positive (MCF7, T47D, ZR75.1, and MDA-MB-361) and ER-negative (BT-474, MDA-MB-231, and MDA-MB-436) breast cancer cells (Fig. 1). Interestingly, we observed a notably reduced expression of eIF3f in ER-positive cells compared with ER-negative cells. However, the expression levels of eIF3a were quite similar among cell lines, and eIF3b levels were significantly higher in BT-474 cells compared with the ER-positive and MDA-MB-231 cell lines (p Ͻ 0.05) (Fig. 1B). Our results suggested that eIF3f expression in ER-positive breast cancer cells may be a function of the estrogen-ER pathway.

Estrogen represses the transcription of the EIF3F gene
We first asked whether EIF3F gene is a direct target of the genomic estrogen-ER pathway. We measured the levels of the eIF3f mRNA in response to vehicle or estrogen stimulation in MCF7 cells by RT-qPCR ( Fig. 2A, right plot). We found a significant reduction in the amount of the eIF3f mRNA starting at 12 h of estrogen treatment. As expected, the levels of the TFF1 mRNA, a well-characterized ER target gene, significantly increased at earlier time points (6 h) ( Fig. 2A, left plot). In contrast, the amount of the eIF3a mRNA did not change and the levels of the eIF3b mRNA only showed a significant increase at the 6-h time point (Fig. S1).
To rule out the possibility that estrogen affects the stability of the eIF3f mRNA, we determined its half-life. In vehicle-and estrogen-stimulated MCF7 cells, the half-life of the eIF3f mRNA was ϳ12 h (Fig. 2B, right plot). Interestingly, we found that estrogen treatment significantly stabilized the TFF1 mRNA, in addition to positively regulating TFF1 transcription (Fig. 2B, left plot). Next, we investigated the effect of tamoxifen on EIF3F transcription. As expected, tamoxifen antagonized estrogen-mediated activation of TFF1 transcription, but, like estrogen, it repressed EIF3F transcription (Fig. 2C). Additionally, we confirmed that neither estrogen nor tamoxifen affects the amount of the eIF3f mRNA in ER-negative MDA-MB-231 cells (Fig. 2C). Thus, estrogen induces transcriptional repression of the EIF3F gene.

ER␣ regulates expression of eIF3f ER␣ mediates estrogen-induced transcriptional repression of EIF3F
To confirm that ER␣ directly affects the observed response to estrogen, we performed specific silencing of ER␣ and determined the levels of the eIF3f mRNA in control or ER␣-silenced MCF7 cells treated with vehicle or estrogen for 24 h. As shown in Fig. 3A, suppression of ER␣ expression prevented ligandinduced repression of EIF3F transcription. As expected, we detected reduced levels of the TFF1 mRNA in estrogen-stimulated ER␣-silenced cells compared with control cells. In addition, we confirmed ER␣ silencing by immunoblotting (Fig. 3B). Interestingly, the amount of the eIF3f protein did not significantly change under any experimental condition, although mRNA levels did ( Fig. 3 and Fig. S2). These results suggest that the decrease in eIF3f mRNA levels by the estrogen-ER␣ complex is buffered by an increase in estrogen-stimulated mRNA translation, as demonstrated below. In contrast, we did not observe any effect of estrogen on the translation of the eIF3a and eIF3b mRNAs ( Fig. 3B and Figs. S1 and S2).
The mechanisms by which estrogen-bound ER␣ represses transcription are poorly understood. It has been proposed that early gene inhibition involves transient binding of ER␣ to the promoter and sequestration of limiting factors away from the repressed gene, a mechanism known as physiologic squelching. However, late gene inhibition (after 6 h of treatment) requires the binding of ER␣ and repressors or corepressors at sites adjacent to the repressed gene. Some reports indicate that the expression of specific repressors is induced by estrogen, which explains the lag (5,28). The results shown in Fig. 2A suggested that estrogen-mediated repression of EIF3F transcription might occur according to the late inhibition model. However, we did not observe any change in the amount of eIF3f transcripts when estrogen-stimulated cells were pretreated with the protein synthesis inhibitor cycloheximide (Fig. 4A). These results indicated that estrogen-induced expression of a repressor was not required for EIF3F repression, and suggested regulation by physiologic squelching. To test this hypothesis, we performed chromatin immunoprecipitation (ChIP) assays using nuclear extracts from MCF7 cells treated with vehicle or estrogen for 30 min (Fig. 4B). For these studies, we used three pairs of primers to amplify different regions of the EIF3F promoter, and another pair that amplified an upstream region containing an experimentally identified ER-binding site (ERE). Distal EREs are involved in late gene inhibition (5). As expected, we did not detect increased binding of ER␣ to this distal ERE upon estrogen stimulation. In contrast, we observed elevated binding of ER␣ to the EIF3F promoter in response to estrogen treatment, particularly to the region proximal to the transcription start site, which correlated with reduced loading of RNA polymerase II (Fig. 4B). However, these changes were much less pronounced compared with the differential binding observed at the TFF1 promoter, which contains a canonical ERE (Fig. 4C). These results indicated a transient and labile interaction of ER␣ with the EIF3F promoter and supported a physiologic squelching mechanism for estrogen-ER␣ repression of EIF3F transcription. Although transcription inhibition may occur as early as 30 min upon estrogen stimulation, changes in eIF3f mRNA levels were detectable at later times because of its ϳ12-h halflife (Fig. 2B).

Estrogen-ER␣ pathway promotes the binding of eIF3 to eIF4F and increases cap-dependent translation
We observed that estrogen-induced reduction of the eIF3f mRNA levels was not associated with a decrease in the eIF3f protein amount ( Figs. 2A and 3, A and B). This result suggested that eIF3f expression is buffered by up-regulation of eIF3f mRNA translation in response to estrogen. This effect may be mediated by the mTORC1 pathway, which is activated in estrogen-treated MCF7 cells as indicated by the increased phosphorylation of its downstream targets S6K1 and 4E-BP1 (Fig. 3B). mTORC1 is activated by the nongenomic estrogen-ER pathway and, as previously described, is a key regulator of cap-dependent translation (7,10). To test this hypothesis, we evaluated protein synthesis using a Dual-Luciferase reporter system that monitors the ratio between cap-dependent and cap-independent translation initiation (Fig. 5A). We observed that estrogen

ER␣ regulates expression of eIF3f
stimulation increased cap-dependent translation by ϳ 2-fold, which was prevented by pretreatment of the cells with either mTORC1 (rapamycin) or ER␣ (fulvestrant) inhibitors (Fig. 5A). Additionally, we found that, unlike estrogen, tamoxifen failed to activate mRNA translation (Fig. 5A). These results support a role of estrogen-ER␣ signaling in the control of mRNA translation through the activation of mTORC1. mTORC1 regulates the formation of the cap-binding complex eIF4F and its interaction with eIF3 to form the 48S preinitiation complex (10,11,29). Accordingly, we postulated that estrogen may enhance translation by promoting the assembly of the 48S PIC. To test this hypothesis, we isolated translation initiation complexes using m 7 GTP-agarose beads from MCF7 cells grown in DMEM containing 5% charcoal-stripped FBS and supplemented with vehicle or estrogen. As shown in Fig.  5B, the binding of eIF4G and the eIF3 subunits eIF3a, eIF3b, and eIF3f to eIF4E increased in response to estrogen stimulation, whereas 4E-BP1 interaction with eIF4E decreased. As expected, inhibition of mTORC1 with rapamycin promoted the binding of dephosphorylated 4E-BP1 to eIF4E and, consequently, the dissociation of eIF4G and the eIF3 subunits from eIF4E. Additionally, we observed that inhibition of ER activity with fulvestrant or tamoxifen prevented estrogen-induced interaction of the eIF3 subunits with the cap-binding complex eIF4F (Fig. 5B   Figure 2. Estrogen reduces transcription of the EIF3F gene. A, ER-positive MCF7 cells were grown in phenol red-free DMEM supplemented with 10% charcoal-stripped FBS (low-estrogen medium) for 3 days before stimulation with vehicle (EtOH) or estradiol (10 nM) for the indicated times. Total RNA was purified and the levels of TFF1, eIF3f, and GAPDH mRNAs were determined by RT-qPCR. TFF1 and eIF3f values were normalized to GAPDH and mean Ϯ S.E. of three independent experiments were expressed relative to vehicle-treated sample at time 0 (set to 1) (*, p Յ 0.05; **, p Յ 0.01). B, MCF7 cells were grown in low-estrogen medium for 3 days and then stimulated with vehicle (EtOH) or estradiol (10 nM) for 12 h before adding actinomycin D (5 g/ml) and continuing the incubation for indicated times. TFF1 and eIF3f mRNAs were analyzed as described in A. Data represented as mean Ϯ S.E. of three independent experiments (*, p Յ 0.05). C, MCF7 and MDA-MB-231 cells were grown in low-estrogen medium for 3 days before stimulation with vehicle (EtOH), estradiol (10 nM), or tamoxifen (100 nM) for 24 h. TFF1 and eIF3f mRNAs were purified and analyzed as described in A. Mean Ϯ S.E. of three independent experiments were plotted (*, p Յ 0.05; **, p Յ 0.01).

ER␣ regulates expression of eIF3f
and Table S1). However, we found no effect of estrogen on the assembly of the eIF3a, eIF3b, and eIF3f subunits (Fig. 5C). These results indicated that estrogen-mediated activation of mTORC1 facilitates the binding of eIF3 and, therefore, the 43S PIC to the cap-binding complex eIF4F. This increase in the levels of the preinitiation complexes specifically promotes translation of mRNAs harboring long and structured 5Ј-UTRs that encode factors required for cell proliferation, survival, or angiogenesis (12). Accordingly, we observed enhanced synthesis of cyclin D1 and survivin in estrogen-stimulated cells, which is reversed by the treatment with the inhibitors of mTORC1 or estrogen receptor (Fig. 5D). However, the amount of the antiapoptotic factor Mcl1, whose translation required high levels of phosphorylated eIF4E, did not change under any condition ( Fig. 5D) (30). As predicted, the levels of eIF3f protein were similar in vehicle-and estrogen-stimulated cells, but they decreased after 36 h of treatment with rapamycin or tamoxifen (Fig. 5E).
Strikingly, we detected the interaction between ER␣ and the eIF3 complex in unstimulated MCF7 cells, which was dissociated by the treatment with estrogen. Using immunoprecipitation assays, we observed an increased binding of ER␣ to eIF3b and eIF3f in the absence of estrogen (3.5-and 1.6-fold over background, respectively), which was prevented by estrogen stimulation (Fig. 6, A-C). Ligand binding to ER␣ induces a conformational change that promotes ER␣ dimerization and activation and determines its association with transcriptional coactivators and its dissociation from inhibitory chaperones (31,32). Consistently, this structural change may also prevent the binding of ER␣ to eIF3. In addition, estrogen-bound ER␣ translocates to the nucleus, where it activates or represses transcription of target genes. Accordingly, we observed estrogeninduced accumulation of ER␣ in nuclear extracts, but cellular distribution of eIF3b and eIF3f did not change (Fig. 6D). These results suggested that estrogen-liganded ER␣ may also facilitate the interaction of eIF3 with other translation initiation factors.

eIF3f overexpression prevents estrogen-induced up-regulation of protein synthesis, reduces cell proliferation, and induces apoptosis
Down-regulation of eIF3f is also observed in melanoma and pancreatic tumors. In cell derived from these tumors, ectopic expression of eIF3f results in the inhibition of protein synthesis, reduced cell proliferation, and increased apoptosis (27). Similar effects are detected in immortalized NIH3T3 fibroblasts in response to eIF3f overexpression (22). These results underscore the importance of a tight control of eIF3f levels for proper cellular function. To confirm the role of estrogen-ER␣-mediated regulation of eIF3f in ER-positive MCF7 cell biology, we evaluated the rate of protein synthesis, proliferation, and apoptosis in cells expressing the HA tag or HA-tagged eIF3f. mRNA translation was tested using the Dual-Luciferase reporter system, as described in Fig. 5A. As expected, estrogen stimulation enhanced cap-dependent translation in HA-expressing cells, but it did not have any effect on translation in cells expressing HA-eIF3f (Fig. 7A). Consistently, ectopic expression of eIF3f also prevented estrogen-induced synthesis of cyclin D1 and survivin (Fig. 7B). Estrogen stimulates the transcription of the CCND1 gene, which could explain the lower effect of eIF3f overexpression on cyclin D1 expression compared with survivin (33). These results corroborated the inhibitory effect of high eIF3f levels on protein synthesis in MCF7 cells. Then, we investigated the mechanism by which eIF3f overexpression affects protein synthesis. Because estrogen promoted the binding of eIF3 to eIF4F, we first examined the interaction of eIF3b with eIF3a and eIF3f, and their association with the cap-binding complex eIF4F in HA-or HA-eIF3f-expressing cells. As shown in Fig. 7, C and D, ectopic expression of eIF3f did not interfere with the assembly of the translation initiation complex. We did not observe any effect on mTORC1 activity either, because phosphorylation of the ribosomal protein S6 and 4E-BP1 was not affected. Alternatively, eIF3f overexpression might alter the expression of other eIF3 subunits, the assembly of eIF3 complexes, the formation of the 43S PIC, or other translation-related functions of eIF3f as discussed below. Further studies need to be performed to evaluate these options.
Next, we determined the effect of ectopic expression of eIF3f on cell proliferation using MTT cell viability assays. MCF7 cells transiently expressing HA tag (control) or HA-eIF3f (eIF3f) were grown in phenol red-free DMEM supplemented with 5% charcoal-stripped FBS and estrogen (100 nM) for 0, 2, and 4 days. As shown in Fig. 7E, we observed proliferation of control cells; however, cell viability decreased in eIF3f-overexpressing

ER␣ regulates expression of eIF3f
cells, particularly in the first 2 days. This effect correlated with a marked reduction in the levels of the HA-eIF3f protein (Fig.  7E). These results indicated that the overexpression of eIF3f induced apoptosis. Accordingly, we observed a significantly higher fraction of early and late apoptotic cells in eIF3f-expressing cells than in HA-expressing cells after double staining with Annexin V-PE and 7-AAD (Fig. 7F). As control, cells transfected with the empty vector were treated with the vehicle or etoposide to induce apoptosis. Surprisingly, we observed that eIF3f overexpression was more efficient in inducing apoptosis than etoposide in MCF7 cells.
All together, these results indicate that the levels of eIF3f are critical to ensure proper synthesis of proteins required for cell proliferation and survival, such as cyclin D1 and survivin in estrogen-stimulated MCF7 cells. Therefore, we conclude that a tight and coordinated control of eIF3f expression by estrogen-ER␣ genomic and nongenomic pathways is essential for adequate proliferation and survival of ER-positive cells.

Discussion
ER-positive breast tumors rely on estrogen-ER activity for their development and progression. Through its genomic and  Figure 6. Estrogen induces the dissociation of ER␣ from eIF3 and the nuclear localization of ER␣. A, HA-transfected or HA-eIF3f-transfected MCF7 cells were incubated in low-estrogen DMEM for 3 days, followed by stimulation with estrogen (10 nM) for 30 min. Cell extracts were prepared, precleared with protein G-agarose beads for 1 h at 4°C, and incubated with anti-HA antibody overnight at 4°C. Isolated immunocomplexes were resolved by SDS-PAGE and indicated proteins were detected by immunoblotting. B and C, MCF7 cells were treated as in A, and cell extracts were precleared with protein A/G-agarose for 1 h at 4°C followed by incubation with anti-ER␣ (B) or anti-eIF3b (C) antibodies overnight at 4°C. Isolated immunocomplexes were resolved by SDS-PAGE and indicated proteins were analyzed by immunoblotting. D, MCF7 cells were estrogen-deprived for 3 days, followed by stimulation with estrogen (10 nM) for 1 h. Nuclear and cytoplasmic extracts were generated as described in "Experimental procedures" and separated by SDS-PAGE. Indicated proteins were analyzed by immunoblotting.

Figure 5. Estrogen-ER␣ pathway facilitates the binding of eIF3 to eIF4F and promotes cap-dependent translation. A, MCF7 cells transfected with
R-HCV-L bicistronic plasmid were grown in phenol red-free DMEM containing 5% charcoal-treated FBS for 3 days before being treated with vehicle, estradiol (100 nM), estradiol and rapamycin (20 nM), estradiol and fulvestrant (100 nM), or tamoxifen (100 nM) for 24 h. Cell extracts were obtained, and Renilla and Firefly Luciferase activities were determined. Renilla/Firefly ratios were calculated and mean Ϯ S.E. of three independent experiments were plotted relative to vehicle-treated cells (set to 1) (*, p Յ 0.05). B, MCF7 cells were estrogen-deprived for 3 days. Cells were treated with vehicle, rapamycin (20 nM), or fulvestrant (100 nM) for 30 min before being stimulated with estradiol (100 nM) or tamoxifen (100 nM) for 2 h. Cell extracts were obtained and translation initiation complexes isolated using m 7 GTP-agarose beads. Indicated proteins were analyzed by immunoblotting. C, cell extracts obtained as in B were incubated with anti-eIF3b antibody overnight at 4°C. Purified immunocomplexes were resolved by SDS-PAGE and indicated proteins were analyzed by immunoblotting. D, MCF7 cells were cultured and treated for 24 h as in B. Cell extracts were prepared and resolved by SDS-PAGE, and indicated proteins were detected by immunoblotting. E, MCF7 cells were grown in phenol red-free media supplemented with 5% charcoal-stripped FBS for 3 days before stimulation with vehicle (EtOH), estradiol (100 nM), estradiol and mTOR inhibitor pp242 (2.5 M), or tamoxifen (100 nM) for 24 or 36 h. Cell lysates were prepared and indicated proteins were analyzed by immunoblotting.

ER␣ regulates expression of eIF3f
nongenomic pathways, ER enhances the transcription and the activity of multiple factors that promote cell proliferation, cell cycle progression, angiogenesis, and survival (7). In this report, we demonstrated that estrogen-bound ER␣ also regulates mRNA translation and, therefore, controls the expression of target genes such as CCND1 at translational level. This regulation occurs by two mechanisms in MCF7 cells: First, by maintaining proper levels of the eIF3f subunit (Figs. 2 and 3) and second, by activating the mTORC1 pathway to facilitate the assembly of the 48S preinitiation complex through the interac-tion of eIF3 and eIF4G (Fig. 5). Additionally, we found that elevated levels of eIF3f reduce the rate of protein synthesis, decrease proliferation, and increase apoptosis (Fig. 7). These results suggest that eIF3f may be a negative regulator of cancer cell growth. Therefore, any therapeutic strategy aiming to induce eIF3f expression may be effective for treatment of ERpositive breast cancers, particularly those with low eIF3f levels.
We detected reduced levels of the eIF3 subunit eIF3f in ERpositive breast cancer cells compared with ER-negative cells (Fig. 1). In agreement with this observation, overexpression of A, MCF7 cells transfected with empty vector or an eIF3f-expressing plasmid and R-HCV-L bicistronic plasmid were grown in phenol red-free DMEM containing 5% charcoal-treated FBS for 3 days before being treated with vehicle or estradiol (100 nM) for 24 h. Cells extracts were obtained and Renilla and Firefly Luciferase activities were determined. Renilla/Firefly ratios were calculated and values plotted relative to untreated control cells (*, p Յ 0.05). B, MCF7 cells were transfected with plasmids expressing HA tag or HA-eIF3f and grown in phenol red-free DMEM containing 5% charcoal-treated FBS for 2 days. Then, cells were treated with vehicle (EtOH) or estrogen (100 nM) for 24 h. Whole-cell extracts were obtained and resolved by SDS-PAGE, and indicated proteins analyzed by immunoblotting. C, MCF7 cells were cotransfected with pcDNA3 and empty vector or an eIF3f-expressing plasmid (pcDNA3 provides neomycin resistance to transfected cells). After 24 h, cells were selected in low-estrogen medium containing G410 (0.6 mg/ml) for 48 h and then stimulated with estrogen (100 nM) for 2 h. Whole-cell extracts were prepared and cap-binding complexes were isolated and resolved by SDS-PAGE; indicated proteins were analyzed by immunoblotting. D, cell extracts prepared as in C were used to pull down eIF3b protein. Immunoprecipitated proteins were separated by SDS-PAGE and indicated proteins analyzed by immunoblotting. (Band corresponding to heavy IgG chain is marked with an asterisk). E, MCF7 cells were transfected as in B. After 24 h, cells were seeded into 96-well plates and incubated in complete medium for additional 24 h. Then, transfected cells were grown in phenol red-free DMEM containing 5% charcoal-treated FBS and estrogen (100 nM) for indicated days. Viable cells were estimated using MTT cell viability assays, and values represented as mean Ϯ S.E. relative to day 0 cells (set to 1) determined from three independent assays (*, p Յ 0.05; **, p Յ 0.01). The blots show the levels of the endogenous and ectopic eIF3f at the different time points (C, Ctrl; 3f, HA-eIF3f). F, MCF7 cells were transfected as in C and selected in DMEM containing 10% FBS and G418 for 48 h, as indicated. Cells were treated with vehicle (DMSO) or etoposide (50 M) for 24 h, as indicated, and apoptotic cells were labeled with Guava Nexin Reagent and quantified using the Guava easyCyte Flow Cytometer. Percentage of apoptotic cells was represented as mean Ϯ S.E. from three independent experiments (*, p Յ 0.05). eIF3a, eIF3b, eIF3c, eIF3d, eIF3g, eIF3h, eIF3i, and eIF3m, and down-regulation of eIF3e and eIF3f have been documented in several types of tumors, including breast cancer (15,21). Supporting a role of the aberrant expression of eIF3 subunits in cancer biology, Hershey and colleagues demonstrated that overexpression of eIF3a, eIF3b, eIF3c, eIF3h, and eIF3i in NIH3T3 cells promoted malignant transformation. In contrast, ectopic expression of eIF3e and eIF3f inhibited cell growth and proliferation, which suggested a tumor-suppressing activity of these two subunits (22). Further studies have shown that siRNA knockdown of the eIF3 subunits reduces the malignant phenotypes of cancer cells derived from tumors in which they are overexpressed. According to its predicted tumor-suppressor activity, overexpression of eIF3f inhibits proliferation and induces apoptosis in cells with low endogenous eIF3f levels, such as melanoma and pancreatic and breast cancer cells (15,20,21) (Fig. 7, E and F). The role of eIF3e in tumor development is more ambiguous, because both low and high eIF3e levels as well as the expression of truncated eIF3e proteins have been linked to human cancers. All these results associate altered levels of eIF3 subunits to malignant transformation. However, the mechanisms by which these changes contribute to cancer etiology are still being elucidated.
Some evidence indicates that imbalanced expression of the eIF3 subunits leads to the formation of partial eIF3 subcomplexes, which may control translation of a specific set of mRNAs (20,21). Thus, ectopic expression or siRNA knockdown of eIF3a raises or reduces the protein levels of most of the other eIF3 subunits, respectively, which in turn affect the formation and stability of the entire eIF3 complex (22,34). These changes in eIF3a levels also alter the translation of specific mRNAs. Consequently, the synthesis of tyrosinated ␣-tubulin, the ribonucleotide reductase regulator M2, and the N-myc downstream regulated gene-1 (NDRG1) is stimulated in cell overexpressing eIF3a, whereas levels of p27kip1 decrease. Similar effects are observed on other mRNAs that may be responsible for the regulation of cisplatin sensitivity and DNA repair activities. Low levels of eIF3a reverse these changes (21). Another example is eIF3h, whose knockdown induces downregulation of the eIF3k and eIF3l subunits. High levels of eIF3h are observed in prostate and breast cancers, and eIF3h overexpression in NIH3T3 cells enhances the synthesis of cyclin D1, ODC, and FGF2 (23,35). The imbalanced expression of eIF3 subunits and the formation of eIF3 subcomplexes may alter the canonical function of eIF3 in the assembly of the 48S PIC through its interaction with eIF4G. Other translational steps in which eIF3 plays a critical role (e.g. translation reinitiation, termination, and ribosome recycling) may be also affected. Remarkably, Cate and colleagues identified a group of mRNAs whose 5Ј-UTRs specifically cross-linked to eIF3a, eIF3b, eIF3d, and eIF3g. Using c-JUN mRNA, they found that the binding of eIF3 subunits to specific secondary structures at its 5Ј-UTRs promotes the interaction of eIF3d to the 5Ј cap and eIF4F-independent recruitment of the mRNA to the 43S PICs for translation initiation (24,25). These results indicate that the levels of entire eIF3 complex and/or partial eIF3 subcomplexes may be critical to ensure the proper synthesis of factors required for the proliferation, growth, and survival of cancer cells.
Silencing of eIF3f significantly reduces the levels of eIF3h, eIF3m, eIF3k, and eIF3l subunits, which impairs the formation of the eIF3 octamer and the entire complex. However, the assembly of YLC complex (eIF3a, eIF3b, eIF3g and eIF3i) is not affected. This complex represents a minimal functional unit of eIF3 able to promote the binding of ternary complex and mRNA to the 40S ribosome and the scanning of the 5Ј-UTR of mRNAs (23,34). These results suggest that the levels of entire eIF3 complex may be reduced, whereas YLC-like subcomplexes may be accumulated in cells containing low amounts of eIF3f such as MCF7 cells. We observed the association of eIF3f to eIF3a and eIF3b and their binding to eIF4F, which is mediated by eIF3c, eIF3e and eIF3d (Fig. 5, B and C) (36). Our results confirm the presence of the entire functional eIF3 complex in MCF7 cells, whose estrogen-stimulated binding to eIF4F enhances cap-dependent mRNA translation, including the synthesis of cell cycle and survival regulators such as cyclin D1 and survivin (Fig. 5, A and D). Further studies will be performed to investigate the existence of eIF3 subcomplexes, as well as their implication in the translation of a specific group of mRNAs. Altered eIF3f levels affect the synthesis of proteins involved in cell proliferation and survival such as cyclin D1 and survivin, and result in reduced proliferation and increased apoptosis in MCF7 cells (Fig. 7, B, E, and F). Consequently, we expect to identify other tumor-promoting factors, whose expression is sensitive to eIF3f levels. Similar effects are observed in other cancer cells containing low levels of eIF3f such as melanoma and pancreatic cancer cells (27,37). The mechanisms by which increased levels of eIF3f inhibit protein synthesis are not wellknown. Some studies indicate that ectopic expression of eIF3f promotes degradation of the 28S and 18S rRNAs, resulting in reduced number of ribosomes and, therefore, decreased rate of protein synthesis (27,37). Consequently, cells undergo apoptosis, which in turn induces the activation of CDK11p46, a kinase that phosphorylates eIF3f and stabilizes its binding to other eIF3 subunits into an inactive complex (38). Our studies did not show changes in the levels of 18S rRNA by RT-qPCR (data not shown), altered binding of eIF3f to eIF3a and eIF3b, or impaired association of eIF3 with eIF4F in MCF7 cells overexpressing eIF3f, compared with HA-expressing cells (Fig. 7, C and D); however, we cannot rule out this mechanism. In fact, estrogen-ER pathway has been implicated in the regulation of ribosome biogenesis (39).
In addition to its function as a stabilizer of the eIF3 complex, eIF3f plays a critical role in the regulation of translation initiation by the mTORC1 pathway in certain cell lines. Thus, eIF3f works as a scaffolding protein for the activation of S6K by mTORC1, which results in the phosphorylation of components of the translational machinery and, consequently, increased assembly of the 48S PIC, as described above (11,40). Therefore, eIF3f levels may modulate mTORC1 activity in translation initiation. However, we did not observe changes in mTORC1 activity or 48S PIC assembly in MCF7 cells overexpressing eIF3f (Fig. 7, B-D). Interestingly, cell lines with low eIF3f levels (MCF7, T47D, and ZR75.1) harbor activating alterations in components of the PI3K/AKT/mTORC1 pathway, and increased eIF3f levels may not have a significant impact on downstream targets. In contrast, high levels of eIF3f in cells
As discussed, low levels of eIF3f are essential to keep the proper expression of proteins involved in cell proliferation and survival in melanoma, pancreatic, and breast cancer cells. Although allelic loss of the EIF3F gene accounts for reduced expression in melanoma and pancreatic cells, gene amplification or deletion or mutations of this gene are not frequently detected (Ͻ 0.9%) in the breast tumor samples available in The Cancer Genome Atlas (TCGA). These data highlight the relevant role of the estrogen-ER␣ genomic and nongenomic pathways in the regulation of eIF3f expression in MCF7 cells. By describing this regulatory mechanism, we contribute to a better understanding of ER-positive MCF7 breast cancer cell biology and provide rationale for the investigation of eIF3 and/or eIF3f as druggable targets. Our results may also provide insight into the response of ER-positive cells to endocrine therapies. In agreement with this notion, we observed that tamoxifen reduces eIF3f mRNA and protein levels, which was associated with protection from drug-induced apoptosis in melanoma cells (Figs. 2C and 5E) (27). These results suggest that tamoxifen may not be as effective as other endocrine therapies for the treatment of ER-positive tumors with low eIF3f levels.
MCF7 cells were transfected using FuGENE HD (Promega) according to the manufacturer's protocol. Lipofectamine RNAiMAX (Invitrogen) was used for transfection of siRNAs following the manufacturer's protocol.
For immunoprecipitation assays, equal amounts of protein were precleared with 20 l of 50% protein A-or protein G-agarose bead slurries for 1 h at 4°C followed by incubation with anti-eIF3b (10 l), anti-ER␣ (10 l), or anti-HA (2 l) overnight at 4°C. Protein A-or protein G-agarose beads, previously blocked with BSA, were added to the samples and incubation continued for 1 h at 4°C. Beads were washed five times with lysis buffer for 5 min at 4°C and collected by centrifugation at 3,000 rpm for 3 min. Immunocomplexes were resolved and detected as described above.
For cap-binding assays, equal amounts of protein were incubated with m 7 GTP-agarose (Creative BioMart; m7GTP-001A) overnight at 4°C. Beads were washed five times with lysis buffer, and complexes were resolved and detected as described above.

Nuclear-cytoplasmic fractionation
Nuclear and cytoplasmic extracts were prepared using NE-PER TM Nuclear and Cytoplasmic Extraction Reagents (Thermo Scientific) according to the manufacturer's instructions.

ER␣ regulates expression of eIF3f
Luciferase reporter assays Cells expressing Renilla-HCV IRES-Firefly reporter mRNA were treated as indicated in figure legends. Cells were lysed with 1ϫ Passive Buffer, and Renilla and Firefly Luciferase activities were determined using Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer's protocol. Assays were performed in triplicates and results were analyzed as described in figure legends.

Cell proliferation assays
Cell viability was determined using MTT Cell Viability Assay Kit from Biotium. Briefly, MCF7 cells were seeded in duplicate into 96-well plates at 5,000 cells/well. Cells were cultured in phenol red-free DMEM supplemented with 5% charcoalstripped FBS and estrogen (100 nM) for 0, 2, and 4 days. Cell viability was determined according to the manufacturer's protocol. Absorbance was measured at 570 nm.

Apoptosis assays
HA-expressing MCF7 cells were treated with vehicle (DMSO) or etoposide (50 M) for 24 h. Selected HA-and HA-eIF3f-expressing cells were treated with vehicle for 24 h. Cells were labeled with Guava Nexin Reagent (EMD Millipore) according to the manufacturer's protocol. Cells were sorted using a Guava easyCyte Flow Cytometer (EMD Millipore) and analyzed with GuavaSoft software (EMD Millipore).

Statistical analysis
Statistical analysis was performed using the Prism GraphPad 7.0 software. Significance was determined by paired two-tailed Student's t test. p values Ͻ 0.05 were considered significant.