|Home | About | Journals | Submit | Contact Us | Français|
The eukaryotic translation initiation factor 4E (eIF4E) alters gene expression on multiple levels. In the cytoplasm, eIF4E acts in the rate-limiting step of translation initiation. In the nucleus, eIF4E facilitates nuclear export of a subset of mRNAs. Both of these functions contribute to eIF4E's ability to oncogenically transform cells. We report here that the homeodomain protein, HOXA9, is a positive regulator of eIF4E. HOXA9 stimulates eIF4E-dependent export of cyclin D1 and ornithine decarboxylase (ODC) mRNAs in the nucleus, as well as increases the translation efficiency of ODC mRNA in the cytoplasm. These activities depend on direct interactions of HOXA9 with eIF4E and are independent of the role of HOXA9 in transcription. At the biochemical level, HOXA9 mediates these effects by competing with factors that repress eIF4E function, in particular the proline-rich homeodomain PRH/Hex. This competitive mechanism of eIF4E regulation is disrupted in a subset of leukemias, where HOXA9 displaces PRH from eIF4E, thereby contributing to eIF4E's dysregulation. In regard to these results and our previous finding that ~200 homeodomain proteins contain eIF4E binding sites, we propose that homeodomain modulation of eIF4E activity is a novel means through which this family of proteins implements their effects on growth and development.
Dysregulation of the eukaryotic translation initiation factor 4E (eIF4E) is linked to oncogenic transformation in cell culture and in vivo (28). eIF4E levels are upregulated in a subset of acute myelogenous and chronic myelogenous leukemias, as well as in non-Hodgkin B-cell lymphomas, breast cancer, and head and neck squamous cell carcinoma (7, 23, 34, 35). Overexpression of eIF4E leads to dysregulated cellular proliferation and malignant transformation in immortalized cell lines (19-21). In addition, eIF4E overexpression can contribute to leukemogenesis by impeding granulocytic and monocytic differentiation (34). eIF4E plays roles in both the nucleus and the cytoplasm, and both of these functions contribute to its physiological effects on cell growth and oncogenic transformation.
In the cytoplasm, eIF4E functions in the rate-limiting step of cap-dependent translation initiation (28). Here, eIF4E directly binds the methyl-7-guanosine (m7G) cap present on the 5′ end of mRNAs, thereby recruiting transcripts to the ribosome (28). In order for translation to proceed, eIF4E must associate with other factors of the eIF4F complex, including eIF4G, the scaffold of this complex, eIF4A, and an RNA helicase, as well as other factors such as the ribosome-bound eIF3 and the poly(A)-binding protein (28). Surprisingly, overexpression of eIF4E does not lead to increased production of all transcripts (28). For example, overexpression of eIF4E leads to increased translation of ornithine decarboxylase (ODC) and vascular endothelial growth factor but not of actin or GAPDH (glyceraldehyde-3-phosphate dehydrogenase) (4, 9). Transcripts that are more efficiently produced when eIF4E is overexpressed are referred to as eIF4E-sensitive transcripts, and those that are not are referred to as eIF4E insensitive (4, 9). This sensitivity is thought to result from long and highly structured 5′ untranslated region (5′UTRs) (4, 9).
In the nucleus, eIF4E functions in nucleocytoplasmic mRNA transport of a subset of transcripts (5, 27, 29). A substantial fraction (up to 68%) of eIF4E is found in multiprotein nuclear structures referred to as eIF4E nuclear bodies (12, 17, 22, 29). The majority of these bodies colocalize with promyelocytic leukemia protein (5, 17). Here, eIF4E promotes the selective transport of specific mRNAs, such as cyclin D1 and ODC, from the nucleus to the cytoplasm without affecting transport of housekeeping mRNAs such as GAPDH and actin or altering the levels of these transcripts (17, 27, 32, 33). As in the cytoplasm, eIF4E requires its m7G cap binding activity for its mRNA transport function (5). The molecular mechanism for how eIF4E-sensitive transcripts are transported and whether eIF4E directly transports mRNAs or participates in a process required for transport is not yet known. However, for ease of terminology we will refer to this general phenomenon as eIF4E-dependent mRNA transport. The underlying basis for sensitivity to eIF4E at this level of regulation is due to the presence of a 100-nucleotide eIF4E sensitivity element (4ESE) in the 3′UTR of targeted transcripts (6). The mRNA transport function of eIF4E contributes to its ability to transform cells and to impede differentiation (5, 34). Interestingly, eIF4E-dependent mRNA transport is upregulated in a distinct subset of acute myeloid leukemia (AML) and blast crisis chronic myeloid leukemia (bcCML) patient specimens (34). Thus, eIF4E-dependent mRNA transport likely contributes to its oncogenic activities.
Given that eIF4E modulates gene expression at the levels of translation and mRNA transport (27), its activity must be regulated at multiple levels to effectively keep its proliferative properties under control. The best-described family of regulators involve the translation functions of eIF4E and include the eIF4E binding proteins (4EBPs) which contain conserved eIF4E binding sites (28). This site is defined by YXXXXLΦ (where X is any amino acid and Φ is any hydrophobic amino acid [see Fig. Fig.1A1A ]) (4, 9). These proteins use this site to interact with the dorsal surface of eIF4E. In general, regulatory proteins using this site do not substantially alter eIF4E's cap binding activity. Instead, they act by sterically blocking association with eIF4G, which also binds the dorsal surface, and thereby impede translation.
Regulation of the nuclear fraction of eIF4E is only now becoming clear. To date, two negative regulators have been identified: the promyelocytic leukemia protein (PML) and the proline-rich homeodomain protein (PRH), also known as the hematopoietically expressed homeodomain Hex (1, 5, 17, 31-33). In cells that express all three proteins endogenously, the majority of these three proteins colocalize to the same nuclear structures (33). Both PML and PRH directly associate with the dorsal surface of eIF4E, repress its mRNA transport function, and subsequently repress its transformation activity (5, 31-33). At the biochemical level, the RING domain of PML represses eIF4E activity by inducing a conformational change in eIF4E, which in turn reduces its affinity for the m7G cap of mRNA by >100-fold (5, 14). Importantly, PML does not contain a conserved eIF4E binding site, indicating that at least in the nucleus, not all regulators need to contain this motif.
Given that PRH is a homeodomain protein, we were surprised to find that it contained a conserved eIF4E binding site, N-terminal to its homeodomain, which it uses for its direct interaction with eIF4E (33). Using a bioinformatics approach, we discovered that ~200 of the ~800 homeodomain-containing proteins in the Swiss-Prot database contain putative eIF4E binding sites (33). Thus, it is possible that eIF4E is regulated in tissue specific manners by association with a wide variety of homeodomain proteins in a variety of contexts.
Loss of homeodomain protein regulation of eIF4E may contribute to disease progression in ~40% of AML specimens (34). Here, eIF4E-dependent cyclin D1 and ODC mRNA transport is substantially upregulated (34). This apparently occurs because of upregulation of eIF4E, enlargement of eIF4E nuclear bodies, and downregulation of PRH, as well as the displacement of PRH from the nucleus. Importantly, restoration of a normal phenotype, including reassociation of PRH with eIF4E and downregulation of eIF4E levels, leads to downregulation of eIF4E-dependent mRNA transport (34).
Here we investigate the possibility that eIF4E can be regulated at multiple levels by the interplay of homeodomain proteins. Further, we examine whether loss of this level of regulation contributes to leukemogenesis. HOXA9 is another homeodomain protein that also contains a putative eIF4E binding site (34) and is implicated in leukemogenesis (30). Because HOXA9 is upregulated in a variety of myeloid leukemias and its overexpression in collaboration with Meis leads to leukemogenesis in animal models (30), we investigated whether HOXA9 does indeed bind eIF4E and thus whether it modulates eIF4E function. We demonstrate that HOXA9, unlike PRH, is a stimulator of eIF4E activity and suggest that this stimulation could be important in its leukemogenic role. Our studies indicate that HOXA9 modulates both the nuclear and the cytoplasmic functions of eIF4E. Importantly, competition between HOXA9 and PRH for eIF4E appears to be important in maintaining normal cell growth control. These results provide an example of a novel mechanism for regulating eIF4E function: competition between inhibitory and stimulatory homeodomain proteins. Further, this subset of homeodomain proteins are positioned to affect gene expression transcriptionally and independently, at the level of mRNA transport and translation, allowing them to act as potent modulators of gene expression.
AML blood cells and normal bone marrow cells were isolated and processed as described previously (11, 13, 34). Primary AML and CML cells were obtained from the peripheral blood of patients at the Markey Cancer Center, University of Kentucky Medical Center. Normal bone marrow was obtained as waste material after pathological analysis or surgical marrow harvest or from the National Disease Research Interchange. All tissues were obtained with the approval of the Institutional Review Board and appropriate informed consent.
Adenovirus (Ad) vectors were constructed to express either green fluorescent protein (GFP) alone or a combination of GFP with the NF-κB inhibitor IκB as previously described (11). Isolated populations were at least 95% pure. Control populations of normal granulocytes and monocytes were obtained by labeling peripheral blood mononuclear cells with CD14-phycoerythrin and CD15-fluorescein isothiocyanate (FITC) (Becton Dickinson). Cells were sorted by using appropriate forward- versus side-scatter gates and CD14+ CD15− (monocytes) and CD14− CD15+ (granulocytes) were isolated.
The bicistonic MSCV-Hoxa9-GFP construct was kindly provided by Guy Sauvageau. In this construct, full-length murine Hoxa9 cDNA lays downstream of the retroviral long terminal repeat, followed by a pgk-GFP reporter cassette. PCR-based site-directed mutagenesis (QuikChange; Stratagene) was used to generate Y11A HOXA9 mutant from the MSCV-Hoxa9-GFP construct. The integrity of both constructs was verified by automated DNA sequencing. Each plasmid was transiently transfected into the Phoenix-Ampho packaging line (kindly provided by Gary Nolan), and retroviral supernatants were used to infect human U937 cells from the American Type Culture Collection. For experiments with Ad infection, purified populations of GFP+ cells were isolated by using a FACSVantage flow cytometer. Dead cells were excluded by using propidium iodide, and sorted populations were at least 95% pure. U937 cells infected with retroviruses were sorted twice to obtain GFP+ populations that were at least 99% pure.
Western blot analysis and coimmunoprecipitation studies were as described previously (32). A total of 20 μg was subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). The following antibodies were used: mouse monoclonal anti-eIF4E antibody (BD Transduction Laboratories), mouse monoclonal anti-cyclin D1 antibody (BD Pharmingen), rabbit polyclonal anti-HOXA9 antibody (Upstate), mouse monoclonal anti-β-actin antibody (Sigma), rabbit polyclonal anti-ODC antibody (Biomol), and affinity-purified rabbit polyclonal anti-PRH antibody (33). All primary antibodies were used at 1:2,000, except rabbit polyclonal anti-HOXA9 antibody that was used at 1:500. Note that experiments for specificity of the HOXA9 antibody indicate that the signal from the antibody is specifically blocked if the HOXA9 antibody is preincubated with purified HOXA9 protein. Horseradish peroxidase-conjugated secondary antibodies were used at 1:20,000, and the signals were detected by chemiluminescence (Super Signal West Pico; Pierce). Coimmunoprecipitations were carried out as described previously (3, 17). Briefly, the appropriate antibody or immunoglobulin G (IgG; Calbiochem) previously cross-linked to protein A-Sepharose beads were added to precleared lysates and incubated overnight at 4°C. Beads were washed five times with immunoprecipitation buffer, collected, and examined by Western blot analysis.
Cell fractionation and Northern blot analysis was performed as previously described (17, 32, 33). Total RNA and RNA from cytoplasmic and nuclear fractions were isolated by the TRIzol (Gibco) procedure according to instructions of the manufacturer and as described previously (32). RNA from nuclear fractions was additionally treated with RNase-free DNase I (Promega). A total of 5 μg of total and fractionated RNA was loaded on a 1% formaldehyde-agarose gel and subsequently transferred onto positively charged nylon membrane (Roche). Membranes were prehybridized in ULTRAhyb buffer (Ambion) for 1 h at 45°C and probed with cyclin D1 cDNA probe (10 pM), eIF4E cDNA probe (5 pM), GAPDH cDNA probe (5 pM) (Ambion), tRNALys antisense oligoprobe (30 pM), and U6 antisense oligoprobe 3′ (30 pM) in the same buffer for 16 h at 45°C. cDNA probes were generated as described previously (32), and signals were detected by using CDP Star chemiluminescence (Ambion) as described by the manufacturer.
Cells were fixed and permeabilized as described previously (5, 32, 33) and incubated, as indicated, with mouse monoclonal anti-eIF4E antibody (1:100; BD Transduction Laboratories), mouse monoclonal anti-PML antibody 5E10 (1:10), rabbit polyclonal anti-HOXA9 antibody (1:100), or affinity-purified rabbit polyclonal anti-PRH antibody (1:50) in blocking buffer for 2 h at room temperature. After incubation with primary antibody, cells were washed three times in 1× phosphate-buffered saline (PBS; pH 7.2) and incubated with Texas Red-conjugated donkey anti-rabbit antibody and Cy5-conjugated donkey anti-mouse antibody (Jackson Immunoresearch Laboratories) for 45 min at room temperature. After secondary antibody incubation, cells were washed three times in 1× PBS (pH 7.4) and mounted in Vectashield with DAPI (4′,6′-diamidino-2-phenylindole; Vector Laboratories, Inc.). For triple staining, cells were additionally fixed with 3.7% paraformaldehyde for 10 min at room temperature, washed, and then incubated with 1:20 dilution of FITC-conjugated mouse monoclonal anti-eIF4E antibody (BD Transduction Laboratories) at 4°C overnight. Fluorescence was observed by using a ×100 objective lens on a Leica inverted scanning confocal microscope with excitation at 488, 568, or 351/364 nm. All channels were detected separately, and no cross talk between the channels was detected. Micrographs represent single sections through the plane of cells with a thickness of ~300 nm. Experiments were repeated three times with more than 500 cells in each sample.
Cell pellets (~500 mg) were homogenized in 1 ml of ice-cold lysis buffer (20 mM HEPES [7.5], 10 mM magnesium acetate, 100 mM potassium acetate, 1× EDTA-free complete protease inhibitor cocktail [Roche], 400 U of SUPERasine/ml [Ambion]) and incubated for 30 min on ice with occasional vortexing. Lysates were spun at 3,000 × g for 10 min at 4°C. Supernatants were transferred to clean tubes and spun at 12,000 × g for 20 min at 4°C. Supernatants obtained in the previous step were spun at 100,000 rpm (TLA 100.3 rotor; Optima TLX ultracentrifuge; Beckman) for 1 h at 4°C to pellet the ribosomes. Ribosomal pellets were resuspended in 200 μl of ice-cold lysis buffer, loaded on the top of the 10 to 40% sucrose gradients (buffered with the lysis buffer), and centrifuged at 55,000 rpm for 60 min by using a TLS 55 rotor in Optima TLX ultracentrifuge (Beckman) at 4°C. Then, 200-μl fractions were collected, and 1/10 of each fraction was used for the measurements of the optical density at 254/280 (OD254/280). Two of ten were saved for Western blot analysis. RNA was isolated from the remaining seven of ten of each fraction by the TRIzol procedure according to the manufacturer's instructions. RNA from each fraction was quantified by spectrophotometry, and 40 ng was converted into cDNA by using the Sensiscript Reverse Transcription kit (Qiagen). Real-time PCR was carried out in triplicate with the QuantiTect SYBR Green real-time PCR kit (Qiagen) in an Opticon thermal cycler (MJR) under the following conditions: 95°C for 15 min, followed by 40 cycles of 94°C for 30 s, 57°C for 30 s, and 72°C for 30 s. The following gene-specific primers were used: GAPDH forward (5′-ACCACAGTCCATGCCATCAC-3′), GAPDH reverse (5′-TCCACCACCCTGTTGCTGTA-3′), cyclin D1 forward (5′-CAGCGAGCAGCAGAGTCCGC-3′), cyclin D1 reverse (5′-ACAGGAGCTGGTGTTCCATGGC-3′), ODC forward (5′-GCATCAGCTTTCACGCTTG-3′), and ODC reverse (5′-TCACCCACATGCATTTCAGG-3′). Semiquantitative reverse transcription-PCR was performed by using a One-Step RT-PCR kit (Qiagen) according to the manufacturer's instructions. Reactions were carried out for 27 cycles under the same conditions as for real-time PCR.
eIF4E was purified for pull-downs as described earlier (14). PRH was purified as described earlier (33). eIF4E for cap binding and fluorescence studies was purified as a fusion protein of the B1 domain of protein G (eIF4E-GB [kindly provided by Gerhard Wagner]). Expression of HOXA9-glutathione S-transferase (GST) was induced in BL21(RIL) codon plus cells at an OD600 of 0.8 with 0.8 mM IPTG for 16 h at 18°C. Sedimented cells were suspended in 0.5 M NaCl-50 mM Na-Tris (pH 7.5)-1 μM tris(2-carboxyethyl)phosphine supplemented with protease inhibitors and 50 U of micrococcal nuclease, 5 mg of RNase A, and 100 U of DNase I (per pellet of 1-liter culture) and then lysed by 20 rounds of sonication on ice by using a duty cycle. Lysates were cleared by centrifugation and incubated with FastFlow glutathione-Sepharose at room temperature for 20 min. Bound beads were exhaustively washed with 0.3 M NaCl-10 mM Na-Tris (pH 7.5), including a high-stringency 0.7 M NaCl-10 mM Na-Tris (pH 7.5) wash, all of which were done at 4°C. Bound protein was eluted by incubation with 0.3 M NaCl-50 mM Na-Tris (pH 7.5)-50 mM reduced glutathione at room temperature for 20 min and dialyzed against 0.3 M NaCl-10 mM Na-Tris (pH 7.5)-1 μM ZnCl2 at 4°C. Purity was assessed by SDS-PAGE, and the concentration was calculated as 280 = 77,520 M−1 cm−1. Measurement of binding affinity using difference circular dichroism (CD) was performed as described previously (14). Briefly, far-UV CD spectra were recorded by using Jasco-810 spectropolarimeter with a 0.847-cm tandem cuvette (Hellma). Binding partners were diluted in 0.3 M NaCl-50 mM Tris (pH 7.5)-5 mM glutathione at 25°C. Partners in two chambers of the cuvette were mixed by inversion and allowed to equilibrate for 30 min. Five spectra for each condition before and after binding were collected by using a 1-nm bandwidth and a 1-nm resolution and then averaged. Relative ellipticity was converted to molar ellipticity (14). The effects due to twofold dilution during mixing of the two tandem cuvette sections were monitored by omitting one of the binding partners. Differences in molar ellipticity at 222 nm for variable binding partner ratios were normalized, where a value of 1.0 corresponds to the maximal CD change at saturating partner concentrations. Binding isotherms were fit to a heuristic expression, assuming stoichiometry of a single binding site as described previously (14). Cap affinity methods are given elsewhere (10).
Our previous studies indicated that ~200 homeodomain proteins contain potential eIF4E binding sites (34). One such homeodomain protein, HOXA9, contains a putative eIF4E binding site (Fig. (Fig.1A)1A) and is upregulated in many myeloid leukemias (2, 8, 16, 18). Similar to PRH, this site is found N-terminal to the homeodomain (Fig. (Fig.1A).1A). Importantly, this binding site is found in HOXA9 from a wide variety of species ranging from fish to humans (Fig. (Fig.1A).1A). In order to determine whether HOXA9 directly binds eIF4E, we carried out GST pull-down experiments with proteins purified to homogeneity. Murine HOXA9 directly binds to eIF4E (Fig. (Fig.1B)1B) with a similar affinity as observed for other eIF4E interacting proteins PML, Bicoid, and PRH (Fig. 1B and C) (5, 14, 24, 25). Mutation of the conserved tyrosine to alanine (Y11A) in the eIF4E binding site of HOXA9 results in a significant reduction in binding (Fig. (Fig.1D).1D). Proteins that contain conserved eIF4E binding sites typically associate with the dorsal surface of eIF4E (28). Consistently, mutation of W73 on the dorsal surface of eIF4E inhibits association with HOXA9 (Fig. (Fig.1E).1E). Similarly, PRH, PML, and Bicoid also require W73 (Fig. (Fig.1E)1E) (5, 14, 24, 25). Mutation of the cap-binding site (W56A) does not impede association with HOXA9, PML, Bicoid, or PRH (Fig. (Fig.1F)1F) (34). Thus, HOXA9 utilizes its conserved eIF4E binding site to directly interact with the dorsal surface of eIF4E.
To establish the physiological importance of the HOXA9 eIF4E interaction, we examined whether these proteins colocalized in primary human blood cells (where they are endogenously expressed) using immunofluorescence and confocal microscopy and separately using immunoprecipitation studies (Fig. (Fig.2).2). Incubation of the polyclonal anti-HOXA9 antibody with recombinant HOXA9 protein, which served as a specific antigen, resulted in the disappearance of the signal in both immunofluorescence and Western studies, indicating that this antibody specifically recognizes HOXA9 protein (http://www.iric.ca/portail_labs.asp?titre_HTML=Supp%2E+Data&labo=16§ion=8 [referred to below as “our website”]). Specimens were taken from normal donors or from a variety of leukemic subtypes. Our previous studies indicated that the PRH eIF4E interaction was disrupted in a distinct subset of myeloid leukemias (34). In particular, normal bone marrow, French-American-British Classification (FAB) subtype AML specimens M1, M2, and ALL specimens all had normal levels of eIF4E and PRH proteins and normal cyclin D1 mRNA transport (34). However, in M4/M5 AML and bcCML specimens, which account for ~40% of myelogenous leukemias, eIF4E levels were substantially upregulated, PRH levels were downregulated and PRH was displaced from the nucleus. This was correlated with dysregulation of eIF4E and upregulation of eIF4E-dependent cyclin D1 mRNA transport (34).
Thus, we examined the distribution of HOXA9 in these specimens to determine whether its interaction could contribute to a subset of leukemias. In normal and M1/M2 AML specimens, PML, PRH, and eIF4E colocalize as expected (34; data not shown). In these same M1 and M2 subtype specimens, HOXA9 also formed nuclear dot structures that were indistinguishable from normal specimens, i.e., diffuse cytoplasmic and nucleoplasmic staining (Fig. (Fig.2A,2A, subpanels A to E and K to O, and Fig. Fig.3A,3A, subpanels A to E). These bodies (red dots in panels E and O) did not have any apparent spatial relationship to those containing PML, PRH, and eIF4E. Occasionally, HOXA9 partially overlapped with an eIF4E nuclear body, but this event appeared to be random.
Interestingly, examination of FAB M4/M5 specimens, which have dysregulated eIF4E, revealed that eIF4E and HOXA9 proteins almost completely colocalized in the abnormally enlarged eIF4E nuclear bodies characteristic of these leukemias (Fig. (Fig.2A,2A, subpanels F to J and P to Y). In addition, both proteins are present throughout the cytoplasm. Because of the diffuse localization in the cytoplasm, confocal microscopy is not best suited for establishing interaction, and thus we carried out immunoprecipitation studies. Consistently, immunoprecipitation studies of specimens from these leukemias reveal that eIF4E and HOXA9 interact, whereas in normal specimens only a small fraction of these proteins interact consistent with the confocal data (Fig. (Fig.2B).2B). In M1/M2 specimens with overexpressed HOXA9 but normal levels of eIF4E, an interaction was observed between HOXA9 and eIF4E; however, a much smaller fraction of eIF4E immunoprecipitated with HOXA9 than in the M4/M5 AML specimens (Fig. 2B and C). Since the confocal data indicate that HOXA9 nuclear bodies are not colocalizing with eIF4E nuclear bodies in the M1/M2 or normal specimens (Fig. (Fig.2A),2A), the interactions detected by immunoprecipitation likely reflect interactions within the populations of eIF4E and HOXA9 which are found diffusely throughout the cytoplasm and/or nucleus. Importantly, unlike the M4/M5 specimens, PRH is still present at the eIF4E nuclear bodies in the M1/M2 or normal specimens, a finding consistent with the observation that eIF4E-dependent cyclin D1 transport is not upregulated in these specimens (34). Note that the cyclin D1 levels are upregulated transcriptionally in M1/M2 specimens (34).
In the M4/M5 specimens, PRH is, in general, absent from eIF4E nuclear bodies (34), whereas enlarged HOXA9 nuclear bodies colocalize with the enlarged eIF4E nuclear bodies (Fig. (Fig.2A,2A, subpanels F to J and U to Y). This is due to two factors, the substantial downregulation of PRH in these leukemias and the fact that PRH is found almost completely in the cytoplasm of these specimens (34). We propose that HOXA9 overexpression in conjunction with low PRH and high eIF4E levels contributes to the upregulation of eIF4E-dependent cyclin D1 mRNA transport previously observed in the M4/M5 AML and bcCML specimens (34).
The subcellular localization of eIF4E and PRH in M4/M5 AML leukemia specimens returns to the pattern observed in normal specimens when the cells express a dominant-negative inhibitor of NFκB, IκB-super repressor (IκB-SR) (34). Specifically, expression of this inhibitor leads to downregulation of eIF4E, upregulation of PRH and return of PRH to the eIF4E nuclear bodies. Importantly, the return of PRH to eIF4E nuclear bodies correlates with downregulation of eIF4E-dependent cyclin D1 mRNA transport (34). Thus, we utilized the same system to determine whether return of normal cyclin D1 mRNA transport was correlated with displacement of HOXA9 from the eIF4E body in the M4/M5 AML and bcCML specimens (Fig. (Fig.3).3). We examined the effects of NF-κB activity by using an Ad vector encoding IκB-SR that mediates strong repression of NF-κB activity within 6 to 12 h (11). AML or bcCML cells were transduced with bicistronic vectors coding for Ad-GFP or Ad-IκB-SR-GFP, and CD34+/GFP+ cells were isolated by fluorescence-activated cell sorting (Fig. (Fig.3)3) (our website). In AML or CML CD34+ cells overexpressing GFP from the control virus, the subcellular distribution of eIF4E, PML, PRH, and HOXA9 proteins (as analyzed by confocal microscopy) is indistinguishable from their respective distributions in the untransduced patient cells (Fig. (Fig.3)3) (34).
In cells expressing IκB-SR, we observe a substantial alteration in eIF4E nuclear body architecture, as expected (Fig. (Fig.3A,3A, subpanels K to O) (our website) where eIF4E nuclear bodies are smaller and there is less eIF4E in the nucleus. Importantly, IκB-SR expression leads to downregulation of HOXA9 protein levels (Fig. (Fig.3B)3B) (our website) and to its reorganization into smaller bodies that are distinct from eIF4E nuclear bodies (Fig. (Fig.3A)3A) (our website). Thus, HoxA9 nuclear bodies (seen as red dots) no longer colocalize with eIF4E nuclear bodies (green dots). This leads to a reduction in cyclin D1 mRNA transport and thus a reduction in levels of cyclin D1 protein in the identical specimens (see Fig. Fig.33 in reference 34). These HOXA9 bodies are similar to those we observed in normal bone marrow controls (Fig. (Fig.3A,3A, subpanels A to E). Identical results are observed whether experiments were done in an M4/M5 AML or bcCML background. Concurrently, we observed the return of PRH to the eIF4E nuclear body in these identical specimens (34).
The direct interaction with eIF4E and the observation that HOXA9 associates with eIF4E in a subset of leukemia specimens led us to investigate whether HOXA9 was positioned to modulate eIF4E functions by modulating its interactions with PRH (Fig. (Fig.4).4). First, we determined whether HOXA9 overexpression alone was sufficient to alter the eIF4E and PRH association observed in control cells. U937 cells were transfected with a bicistronic MSCV vector encoding GFP, along with HOXA9 or the Y11A HOXA9 mutant deficient in eIF4E binding. GFP-positive cells were isolated by using fluorescence-activated cell sorting. In vector control cells, the majority of PML, PRH, and eIF4E colocalized as observed previously (33, 34). Endogenous HOXA9 bodies do not associate with eIF4E nuclear bodies, similar to the studies above in the normal primary specimens (our website). HOXA9 overexpression resulted in two major changes (Fig. (Fig.4).4). First, it leads to a reduction in the association of PRH with eIF4E nuclear structures (our website). Immunoprecipitation studies indicate that when HOXA9 is overexpressed, PRH no longer physically associates with eIF4E (Fig. (Fig.4D).4D). Second, immunoprecipitation studies indicate that overexpressed HOXA9 interacts with eIF4E, in both the nuclear and cytoplasmic compartments (Fig. (Fig.4).4). Consistently, HOXA9 now colocalized with eIF4E nuclear bodies (our website). The subcellular distribution of PML and eIF4E did not appear to be altered by HOXA9 overexpression (data not shown); thus, HOXA9 only alters association of PRH with eIF4E. Importantly, the Y11A HOXA9 mutant, which cannot bind eIF4E, does not associate with eIF4E nuclear bodies and thus does not alter the distribution of PRH, as shown by immunoprecipitation experiments (Fig. 4A and D) and independently by confocal microscopy experiments (our website).
We hypothesized that HOXA9 modulates the mRNA transport activity of eIF4E. Western analysis of cyclin D1 protein levels indicate that overexpression of HOXA9 leads to increased cyclin D1 protein levels, whereas cyclin D1 levels are only modestly increased in cells expressing the Y11A mutant (Fig. (Fig.4B).4B). Importantly, overexpression of HOXA9 did not alter PML or eIF4E protein levels, and PRH appeared to be slightly reduced in both wild-type and mutant-overexpressing cells (Fig. (Fig.4B4B and data not shown). To determine whether upregulation of cyclin D1 is related to upregulation of eIF4E's mRNA transport function, we examined cyclin D1 mRNA levels (Fig. (Fig.4C).4C). Consistent with previous studies, HOXA9 upregulates cyclin D1 mRNA levels. Note that the two bands present in the cyclin D1 Northern blots are both cyclin D1, with different poly(A) tail lengths. Further, there is no alteration in the production of GAPDH mRNA, an eIF4E-insensitive mRNA.
We examined whether eIF4E-dependent cyclin D1 mRNA transport was modulated by HOXA9 (Fig. (Fig.4C,4C, right panel). Here, GFP+ cells expressing either HOXA9, the Y11A mutant, or vector were fractionated into nuclear and cytoplasmic compartments, and the subcellular distribution of cyclin D1 mRNA was monitored. U6snRNA and tRNALys serve as controls for the quality of the fractionation. Comparison of HOXA9 with controls indicates that HOXA9 overexpression was correlated with substantially increased cyclin D1 mRNA levels in the cytoplasmic fraction, indicating increased cyclin D1 mRNA transport. In contrast, the subcellular distribution of an eIF4E-insensitive mRNA, GAPDH, was not altered. Thus, the specificity of the mRNA transport effects of HOXA9 is similar to those observed for eIF4E. Taken together, these findings indicate that HOXA9 increases cyclin D1 expression both through upregulation of its mRNA levels and by promoting its mRNA transport.
To ensure that the HOXA9-dependent increase in cyclin D1 mRNA transport was mediated through its interaction with eIF4E, parallel studies were carried out with the Y11A HOXA9 mutant, which binds eIF4E with substantially reduced affinity (Fig. (Fig.1).1). Analysis of total mRNA indicates that cyclin D1 mRNA levels are increased to the same extent as wild-type HOXA9, a finding consistent with a transcriptional upregulation of cyclin D1 (Fig. (Fig.4C).4C). However, the Y11A mutant did not alter the subcellular localization of cyclin D1 mRNA relative to vector controls (Fig. (Fig.4C).4C). Thus, the ability of HOXA9 to displace PRH from nuclear bodies and to increase cyclin D1 mRNA transport requires its interaction with eIF4E.
To establish whether HOXA9 was a general inhibitor of eIF4E-dependent mRNA transport or specifically inhibited cyclin D1, we monitored the levels and subcellular distribution of ODC, another transcript sensitive to eIF4E at the mRNA transport level (Fig. (Fig.4C,4C, right panel). HOXA9 overexpression, but not expression of the Y11A mutant, led to increased transport of ODC transcripts correlated with increased ODC protein levels. Importantly, neither HOXA9 nor the mutant altered levels of ODC mRNA (Fig. (Fig.4C,4C, left panel). Thus, HOXA9 stimulates eIF4E-dependent mRNA transport of both ODC and cyclin D1 mRNAs. This activity is independent of the transcriptional functions of HOXA9.
Since HOXA9 directly binds eIF4E and modulates the nuclear function of eIF4E, we extended our studies to assess whether the cytoplasmic fraction of HOXA9 modulates translation of eIF4E-sensitive transcripts (Fig. (Fig.5).5). Previous studies indicated that overexpression of eIF4E leads to increased polysomal loading of ODC but not of cyclin D1 or GAPDH transcripts (27). Thus, we monitored polysomal loading of these transcripts in HOXA9- or Y11A mutant-overexpressing cells or vector controls. Appropriately transduced and GFP+ sorted cells were separated into different ribosomal fractions by centrifugation on sucrose gradients, and ribosomal profiles were determined for ODC, cyclin D1, and GAPDH mRNA by performing quantitative real-time PCR. The results are represented as average cycle threshold values (CT) ± the standard deviations for each ribosomal fraction. Parallel experiments using semiquantitative PCR methods confirmed these results (our website). Importantly, for any given set of cells, we used the same ribosomal fractions to monitor all three transcripts so differences in polysomal loadings are not the result of differences in fractionations between experiments.
Clearly, overexpression of HOXA9 leads to a substantial change in the polysomal profile of ODC transcripts (Fig. (Fig.5).5). Here, more transcripts were found in the heavier polysomal fractions (fractions 10 and 11) compared to vector controls and cells expressing the Y11A mutant, which were found with small polysomes (fractions 7 and 9). Loading of transcripts onto heavier polysomes indicates increased translational efficiency. This correlates with increased translation of ODC (Fig. (Fig.4B).4B). Importantly, the polysomal loading profile of GAPDH was not affected by expression of HOXA9 or the Y11A mutant compared to vector controls. For the case of cyclin D1 mRNA, ribosomal distribution was not qualitatively affected by enforced expression of either HOXA9 or the Y11A mutant. Note that the total amount of polysomal cyclin D1 (as reflected by lower CT values) was elevated in the cells overexpressing wild-type form of HOXA9 compared to control cells and cells overexpressing HOXA9Y11A, a finding consistent with the upregulation of transport of cyclin D1 transcripts to the cytoplasm, which was detected in the HOXA9-overexpressing cells.
We further analyzed the polysomal fractions of these cells (our website). Analysis of the polysomal profiles by monitoring OD254 or by Western analysis of ribosomal proteins indicated the expected distributions of eIF4E and ribosomal proteins in these fractions (our website). Overexpression of HOXA9 or the Y11A mutant did not alter the distribution of these ribosomal proteins, indicating that increased polysomal loading is not due to differences in the overall polysomal profile in HOXA9-overexpressing cells. Importantly, endogenous HOXA9, overexpressed HOXA9, and Y11A mutants were not found in the polysomal fractions (our website). Note that eIF4E was found both in fraction 1, with HOXA9, and independently on the polysomes (our website). Thus, eIF4E is available for association with eIF4G consistent with the observed increased protein production (our website).
We sought to understand the biochemical underpinnings of the ability of HOXA9 to promote eIF4E's activity. HOXA9, through its direct interaction with eIF4E, is positioned to modulate eIF4E's activity through a number of mechanisms. First, we examined the possibility that HOXA9 altered the affinity of eIF4E for the m7G cap (our website). The affinity of m7GpppG cap was assessed by monitoring fluorescence quenching of two tryptophans in the cap binding site (W56 and W102) as a function of m7GpppG cap analogue concentration as described previously (14, 15). HOXA9 and, for comparison, PRH and PML RING were purified to homogeneity from bacteria. For these studies, we used the eIF4E-GB fusion protein in which GB acts as a solubility enhancement tag (10). The eIF4E-GB construct bound the cap analogue with a Kd of 130 ± 21 nM, a finding consistent with previous reports (see reference 36 and references therein). The addition of HOXA9 led to a slight increase in cap affinity (Kd = 87 ± 34 nM). For comparison, the addition of a PML RING led to a 100-fold reduction in m7GpppG binding (Kd = 12 ± 3 μM), as expected (14, 15). Importantly, PRH, like HOXA9, had little effect on cap binding relative to eIF4E alone (Kd = 89 ± 14 nM). It appears that neither homeodomain protein greatly alters the affinity of eIF4E for the m7GpppG cap, and thus their respective effects on eIF4E must be mediated through another mechanism.
Another possible mechanism by which HOXA9 promotes eIF4E functions is through displacement of an inhibitor. We examined the possibility that eIF4E activity could be modulated via competition between HOXA9 and PRH (Fig. (Fig.6).6). Note that, like HOXA9, PRH binds eIF4E in both subcellular compartments (33). If these proteins compete for binding in vivo, this should be reflected in their binding affinities for eIF4E in vitro. Since PRH and HOXA9 both bind the dorsal surface of eIF4E, requiring W73, one would expect that any given eIF4E molecule can only bind either HOXA9 or PRH, but not both, at any given time. To determine the dissociation constant of HOXA9, and separately PRH, for eIF4E, we utilized CD spectroscopy as we did previously (14, 15). When either HOXA9 or PRH bind to eIF4E, the proteins undergo a conformational change, as observed by CD spectroscopy. The extent of this change as a function of eIF4E concentration is shown (Fig. (Fig.6).6). The apparent Kds are 0.81 ± 0.17 μM for eIF4E-HOXA9 and 0.34 ± 0.035 μM for eIF4E-PRH, indicating that PRH binds to eIF4E slightly more tightly. However, given the closeness of these Kd values, one would expect that binding of eIF4E would be ultimately dominated by whichever protein, HOXA9 or PRH, was in excess or at higher local concentrations. Given these Kd values and ignoring other unknown factors, cells expressing equimolar amounts of HOXA9 and PRH would have a ratio of ~0.4:1 of HOXA9-eIF4E to PRH-eIF4E complexes. Under conditions in which HOXA9 levels are 10- or 100-fold increased, this ratio would change to ~4:1 or ~40:1, respectively. These ratios would be dramatically affected by co-overexpression of eIF4E and HOXA9, as observed in M4/M5 AML and bcCML, and even more affected because PRH is downregulated while other inhibitor discussed, PML, does not change its levels or localization (33). These results strongly suggest that at the levels of overexpression seen in HOXA9-overexpressing cells or in those of primary patient specimens, HOXA9 effectively outcompetes PRH for eIF4E binding and thereby relieves PRH-mediated repression and modulates eIF4E activity.
We describe here a novel function for HOXA9 that depends on its association with eIF4E. HOXA9 upregulates the transport of cyclin D1 mRNA, and the levels of cyclin D1 transcripts. In contrast, HOXA9 upregulates ODC mRNA transport and translation but does not upregulate the levels of ODC mRNA. Together, these findings suggest that HOXA9 has two classes of functions, its role in transcription and a distinct role in promoting both the nuclear and the cytoplasmic functions of eIF4E. Our studies with the Y11A HOXA9 mutant demonstrate that these functions depend on whether HOXA9 interacts directly with eIF4E. Importantly, the transcriptional and eIF4E-dependent properties of HOXA9 are independent of each other, e.g., the Y11A mutant does not bind eIF4E but still upregulates the levels of cyclin D1 mRNA. The parallel activities of HOXA9 and eIF4E strongly suggest that eIF4E is what determines which transcripts are being regulated by HOXA9 and whether regulation occurs at the level of mRNA transport, translation, or both. However, this does not rule out the possibility that HOXA9 directly interacts with these mRNAs (through possible interactions with its homeodomain) and that HOXA9 may only effect a subset of eIF4E-sensitive transcripts. If true, this would infer that HOXA9 preferentially interacts with a specific subset of eIF4E-sensitive transcripts.
In the nucleus, eIF4E promotes the transport of ODC, cyclin D1, and other mRNAs. Many of these mRNAs contain 4ESE in their 3′UTRs which, together with the m7G cap, are required for eIF4E to promote their export to the cytoplasm (6). Negative regulators of this process were identified previously, e.g., PML and PRH (5, 14, 33). PRH is similar to HOXA9 in that it is a homeodomain containing protein with an N-terminal eIF4E binding site (33). Unlike HOXA9, PRH and PML impede eIF4E-dependent mRNA transport (5, 33). PML, PRH, and HOXA9 bind the dorsal surface of eIF4E with similar submicromolar affinity, suggesting that HOXA9 competes with these negative regulators of eIF4E in order to alleviate repression. It is also possible that, through other regions of the HOXA9 protein, HOXA9 recruits factors that enhance eIF4E-dependent transport, perhaps by stabilizing active transport complexes in addition to removing inhibitors.
In the cytoplasm, eIF4E functions in the rate-limiting step of translation initiation. Interestingly, eIF4E overexpression increases translational efficiency of a subset of eIF4E sensitive transcripts including ODC but not cyclin D1 (27). The sequence features in the cytoplasm that impart sensitivity are not yet known but are thought to involve the structural complexity of the 5′UTR (28) and are distinct from the 4ESE, which plays a role in the nuclear mRNA export functions of eIF4E (6). Our data demonstrate that HOXA9 overexpression actually promotes translational efficiency of ODC. It seems likely that HOXA9 acts prior to the assembly of the eIF4F complex. In particular, HOXA9 binds the same surface of eIF4E that is bound by eIF4G. The eIF4E-eIF4G interaction is required for formation of a productive translation initiation complex. HOXA9 has a Kd for eIF4E of ~1 μM (Fig. (Fig.6),6), whereas the eIF4E-eIF4G Kd is ~1 nM (10). Thus, eIF4G should easily displace HOXA9, allowing for translation of ODC and other eIF4E-sensitive transcripts. Importantly, HOXA9 is absent from the polysomes, and thus there is no HOXA9-eIF4E interaction in the polysomal fractions (our website). Also, both the Y11A mutant (which is deficient in this activity) and the wild-type proteins have the same subcytoplasmic distribution (our website), strongly suggesting that the HOXA9-eIF4G exchange occurs prior to loading of ODC onto polysomes, a finding consistent with the increased translation we observed. Elucidating further details of this mechanism will be an area of active future work.
Previous studies indicated that another homeodomain protein, Bicoid, suppresses translation of caudal mRNA through its N-terminal eIF4E binding site (25). The mechanism of repression is quite different from that of stimulation of translation by HOXA9. The homeodomain of Bicoid binds an element in the 3′UTR of caudal mRNA, known as a Bicoid response element (BRE), and the eIF4E binding site of Bicoid binds the dorsal surface of eIF4E at the same time (24). In this case, mRNA specificity is determined by Bicoid and not eIF4E. Translational repression of caudal mRNA presumably occurs through blocking the eIF4E-eIF4G interaction (Kd = ~1 nM) (24). The data in Fig. Fig.11 and and66 would suggest that the Kd for Bicoid is similar to that of PRH and HOXA9. Thus, the association of Bicoid with both the BRE in caudal mRNA and the eIF4E appears to be important for forming an effective repression complex. Thus, this mechanism is different from the stimulatory mechanism of HOXA9, for which the choice of mRNAs is determined by eIF4E, and the weaker eIF4E-HOXA9 (Kd = ~1 μM) association readily allows exchange with eIF4G and thus translation. In this way, homeodomain proteins modulate eIF4E function in a variety of ways.
HOXA9 only affects eIF4E in cells that endogenously express the HOXA9 protein. For instance, HOXA9 overexpression in NIH 3T3 cells does not lead to upregulation of cyclin D1 (data not shown). Neither does HOXA9 transcriptionally upregulate cyclin D1 in NIH 3T3 cells (data not shown). These findings suggest that HOXA9 requires tissue-specific cofactors. Similarly, the findings that HOXA9 overexpression does not alter eIF4E function in M1/M2 leukemia specimens and that overexpression of HOXA9 alone in U937s is sufficient for its stimulation of eIF4E activity suggest that HOXA9 acts in a context-dependent manner, for which the precise context is defined not only by cell type but by other multifactorial factors that result in differing gene expression programs and thus different complements of eIF4E regulatory factors.
Our finding that 200 homeodomain proteins contain eIF4E binding sites suggest that there could be a plethora of eIF4E regulators (33). Here, we suggest that such regulators could work by competing for eIF4E binding and thereby modulate development and differentiation, as our in vitro work strongly suggests for HOXA9 and PRH (Fig. (Fig.6).6). For instance, the presence of HOXA9 in the eIF4E nuclear body in HOXA9-overexpressing cells is correlated with concomitant displacement of PRH. This mechanism seems to be physiologically relevant in M4/M5 AML and bcCML primary patient specimens, for which the HOXA9 and eIF4E levels are high (the present study) and PRH is displaced to the cytoplasm (34). In this case, eIF4E functions are upregulated and correlate with increased proliferation and impeded differentiation in human cell lines (34). Importantly, treatment with the NF-κB repressor IκB-SR restores PRH to eIF4E nuclear bodies, results in the relocalization of HOXA9 to a distinct part of the nucleus, and downregulates eIF4E-mediated growth and developmental arrest (34; the present study).
In summary, our findings demonstrate that the activity of eIF4E is regulated through interactions with certain homeodomain proteins. These results are particularly interesting in light of new findings suggesting that homeodomain proteins transit between cells to effect gene expression (26). Given that so many of these proteins contain the eIF4E binding motif, it seems likely that eIF4E is regulated through competition of a variety of homeodomain proteins, some inhibitory and some stimulatory, and that these effect outcomes critical to normal cellular proliferation, differentiation, and development. Furthermore, we propose that changes in homeodomain protein expression and/or localization may orchestrate alterations in eIF4E-mediated gene expression. Such changes are independent of the normal transcriptional activities commonly associated with homeodomain proteins and allow modulation of gene expression at multiple levels, positioning these proteins as potent regulators of cellular proliferation and differentiation.
We are grateful for the gifts of antibodies, constructs, and cell lines kindly provided by Guy Sauvageau, L. de Jong, Gerhard Wagner, and Nahum Sonenberg. We thank Jonathan Licht for use of the Opticon thermal cycler. We are grateful for technical assistance from Melanie McConnell, Vladimir Jankovic, and Biljana Culjkovic and for critical reading of the manuscript and helpful discussions with Martin Wiedmann and Guy Sauvageau.
Confocal laser scanning microscopy was performed at the MSSM-LCSM core facility, supported by funding from the NIH (1 S10 RR0 9145-01) and the NSF (DBI-9724504). K.L.B.B. and C.T.J are scholars of the Leukemia and Lymphoma Society. K.L.B.B. holds a Canada Research Chair. Financial support was provided by the NIH (CA 98571 and CA90446).