Translational Control of Putative Protooncogene Nm23-M2 by Cytokines via Phosphoinositide 3-Kinase Signaling* □ S

The expansion and differentiation of hematopoietic progenitors is regulated by cytokine and growth factor signaling. To examine how signal transduction controls the gene expression program required for progenitor expansion, we screened ATLAS filters with polysome-asso-ciated mRNA derived from erythroid progenitors stimulated with erythropoietin and/or stem cell factor. The putative proto-oncogene nucleoside diphosphate kinase B ( ndpk-B or nm23-M2 ) was identified as an erythropoietin and stem cell factor target gene. Factor-induced expression of nm23-M2 was regulated specifically at the level of polysome association by a phosphoinositide 3-kinase-dependent mechanism. Identification of the transcription initiation site revealed that nm23-M2 mRNA starts with a terminal oligopyrimidine sequence, which is known to render mRNA translation dependent on mito-genic factors. Recently, the nm23-M2 locus was identified as a common leukemia retrovirus integration site,

The expansion, survival, and differentiation of hemopoietic progenitors is largely controlled by growth factors, cytokines, and hormones. Human and murine erythroid progenitors can be expanded in vitro in the presence of erythropoietin (Epo), 1 stem cell factor (SCF), and dexamethasone (Dex). In the absence of SCF and Dex, the same cells undergo Epo-dependent terminal differentiation into enucleated erythrocytes (1,2). The in vitro expansion of erythroid progenitors most likely reflects stress erythropoiesis induced by hypoxia, because glucocorticoids and SCF are dispensable for adult erythropoiesis, whereas both factors are required for spleen erythropoiesis in anemic mice (3,4). The ability of normal cells to be expanded under stress conditions is exploited by genetic alterations in leukemia. For instance, the v-ErbB oncogene, encoded by the avian erythroleukemia virus, appears to mimic the synergistic effect of the Epo plus SCF-induced signal transduction required for renewal divisions (1). To examine mechanisms that control the expansion of committed erythroid progenitors, we screened cDNA arrays using cDNA derived from growth factorstimulated (Epo, SCF, or both) versus growth factor-deprived erythroid progenitors. To identify genes that are differentially regulated either at the transcriptional or the translational level, cDNA prepared from polysome-bound mRNA was used. Here, we demonstrate that expression of the nm23-M2 gene is specifically regulated by the signal-controlled polysome association of its constitutively expressed mRNA.
In a recent screen for retroviral integration sites, 9% of the samples in a panel of primary retrovirally induced leukemias harbored integrations in the nm23-M2 gene (5), suggesting that nm23-M2 may be a proto-oncogene. Nm23-M2 or nucleoside diphosphate kinase-B and its human homologue NM23-H2 are members of a gene family that converts G protein-bound GDP to GTP (6 -8). NM23-H1 and NM23-H2 are described as differentiation inhibitory factors, a function independent of their nucleoside diphosphate kinase activity (9). Furthermore, the enhanced expression of NM23-H1 is associated with hematopoietic malignancies in man (10,11). How NM23 proteins exert their leukemic function is unknown, although numerous functions and interactions have been described (12). NM23-H1 has been shown to negatively correlate with tumor metastasis (13), support endocytosis (14), act as a histidine kinase (15), and nick DNA during cytotoxic T lymphocyte-induced apoptosis (16). Nm23-M2 possesses a DNA binding domain by means of which it is able to regulate c-Myc expression (17).
Here, we report that nm23-M2 mRNA expression is constitutively high in erythroid progenitors but that its translation is strictly dependent on Epo-and SCF-induced phosphoinositide 3-kinase (PI3K) activity. Cloning of the full 5Ј-untranslated region (UTR) of nm23-M2 revealed that it contains both a terminal oligopyrimidine (TOP) sequence and an inverted repeat. Deletion of this 5Ј-UTR caused PI3K-independent, constitutive mRNA translation.

EXPERIMENTAL PROCEDURES
Cells-I/11 cells were cultured in StemPro medium (Invitrogen) as described (1). For expansion, the medium was supplemented with 0.5 units/ml Epo (a kind gift of Ortho-Biotech, Tilburg, The Netherlands), 100 ng/ml SCF, and 10 Ϫ6 M Dex (Sigma-Aldrich); for differentiation, the medium was supplemented with 10 units/ml Epo, 10 ng/ml insulin, and 0.5 mg/ml iron-loaded transferrin (Intergene). To analyze cellular signaling mechanisms, cells were washed twice with Hepes-buffered balanced salt solution (Invitrogen) and reseeded in Iscove's medium (Invitrogen) without any additives. Factor stimulation was with 5 units/ml Epo and 200 ng/ml SCF. Hemoglobin accumulation was measured as described (18).
RNA Isolation and Northern Blot Analysis-Isolation of total RNA and Northern blot analyses was performed as described by Chomczynski and Sacchi (19) with minor modifications (20). Isolation of polysomal RNA by sucrose gradient fractionation was performed essentially as described (21). Cell extracts were prepared by lysis at 4°C in extraction buffer (10 mM Tris-HCl, pH 8, 140 mM NaCl, 1.5 mM MgCl 2 , 0.5% Nonidet-P40, 20 mM dithiothreitol, 150 g/ml cycloheximide, 1 mM phenylmethylsulfonyl fluoride, and 500 units/ml RNasin), and nuclei were removed by centrifugation (1000 ϫ g for 10 min at 4°C). The supernatant was supplemented with 665 g/ml heparin and centrifuged (12,000 ϫ g for 5 min at 4°C) to eliminate mitochondria. The supernatant was layered on a 10-ml (ATLAS hybridization and Northern blots) or 4-ml (TaqMan analysis) linear sucrose gradient (15-40% sucrose (w/v) supplemented with 10 mM Tris-HCl, pH 7.5, 140 mM NaCl, 1.5 mM MgCl 2 , 10 mM dithiothreitol, 100 g/ml cycloheximide, and 0.5 mg/ml heparin) and centrifuged in a SW41 Ti rotor (Beckman) at 38,000 rpm for 120 min at 4°C or in a SW50 Ti rotor at 42,000 rpm for 60 min at 4°C without brake. Fractions (550 l) were collected and digested with 100 g of proteinase K in 1% SDS and 10 mM EDTA (30 min at 37°C). RNAs were recovered by phenol/chloroform/isoamyl alcohol extraction followed by ethanol precipitation. RNAs were analyzed by electrophoresis on denaturing 1.2% formaldehyde agarose gels and subsequent Northern blotting (using Hybond-N ϩ membranes; Amersham Biosciences). These gels indicated that fractions 1-11 (10-ml gradient) and 1-4 (4-ml gradient) contain nonpolysomal and subpolysomal mRNA, whereas fractions 13-20 (10-ml gradient) and 5-8 (4-ml gradient) consisted of polysome-bound RNA. The latter fractions were pooled for array hybridization and TaqMan analysis, respectively. As probes we used a 612-bp nm23-M2 cDNA fragment or cDNAs containing the entire open reading frame of pim1 or c-jun. After hybridization, filters were scanned on a PhosphorImager (Amersham Biosciences), and signals were quantified by phosphorimaging.
Poly(A) ϩ mRNA Isolation and cDNA Synthesis-RNA was quantified by UV absorbance. Poly(A) ϩ mRNA was purified from isolated total RNA (25-30 g) with the Oligotex mRNA mini kit (Qiagen) according to the manufacturer's recommendations. Equivalent amounts of poly(A) ϩ RNA (1-2 g) were reverse transcribed into cDNA using 2-l cDNA synthesis primers provided by the ATLAS kit. RNA was denatured (for 7 min at 65°C and 7 min at 4°C), 10 units of RNasin, 10 mM dithiothreitol, 100 nM (each) nucleotides, and the buffer supplied with the enzyme plus 100 units of SSII reverse transcriptase (Invitrogen.) were added, and the reaction was put at 42°C for 1.5 h. Subsequently, RNA was degraded in 0.3 M NaOH at 65°C for 1 h, which was stopped by adding Tris, pH 7.5, to 20 mM and HCl to neutralize the NaOH. cDNA was precipitated in the presence of glycogen (Roche Applied Science) and dissolved in 15 l of H 2 O. 1 l of this dissolved cDNA was used in a random primed labeling reaction (Roche Applied Science) using [␣-32 P]dATP. After removal of free nucleotides, the 32 P incorporation was measured in a Cherenkov counter, and the same amount of counts was used in hybridizations of a single experiment (2,22).
Microarray Hybridization-Hybridization of Atlas cDNA arrays (588 mouse cDNA probes; Clontech) was performed basically as recommended by the manufacturer with the modifications described below. Filters were prehybridized for 8 h at 68°C in 10 ml of prewarmed ExpressHyb plus denatured sheared salmon sperm DNA, both provided by the kit. Subsequently, 6 ϫ 10 7 dpm of denatured radioactive cDNA (kept for 5 min at 95°C without denaturing solution and thereafter chilled on ice) were added and hybridized for 20 h. Filters were washed three times at 68°C in 200 ml of 2ϫ SSC 1% SDS for 40 min. Thereafter, the filters were washed again three times in 0.2ϫ SSC 0.5% SDS for 20 min and subjected to phosphorimaging analysis (Amersham Biosciences).
Western Blotting-I/11 cells were factor-deprived and stimulated as described above. Samples were taken at regular intervals and processed as described previously (18). Following separation of proteins on 10% polyacrylamide gels and Western blotting, filters were incubated with antibodies against Nm23-M2 (Seikagaku, Falmouth, MA), phosphoserine 473-PKB and total PKB (Cell Signaling Technology, Beverly, MA), and ERK1/2 (Santa Cruz Biotechnology). Immune complexes were detected with horseradish peroxidase-conjugated goat anti-rat IgG antiserum (Santa Cruz Biotechnology) followed by an enhanced chemoluminescence reaction (PerkinElmer Life Sciences).
Isolation and Cloning of the 5Ј-Untranslated Region of nm23-M2-A nested PCR was performed on an oligo(dT)-primed cDNA library of primary erythroblasts (a gift from Walbert Bakker, Department of Hematology, Erasmus MC, Rotterdam, The Netherlands). The library was ligated into the lambda ZAP express vector (Stratagene, Amsterdam, The Netherlands). The first PCR (45 s at 94°C, 1 min at 57°C, and 1 min at 72°C for 25 cycles) was performed using the M13 reverse primer (5Ј-ACAGGAAACAGCTATGACCTTG-3Ј) in the vector in combination with pN6 (5Ј-TCGCCCACCAGGCCGCGC-3Ј) in nm23-M2 on 50 ng of cDNA. 1 l of PCR product was transferred to the nested PCR (45 s at 94°C, 1 min at 56°C, and 2 min at 72°C for 30 cycles). The T3 primer (5Ј-AATTAACCCTCACTAAAGGG-3Ј) in the vector and pN8 (5Ј-TGCACGCCATCTGGCTTG-3Ј) in nm23-M2 were used for this reaction. Subsequently, the final products were cloned directly into pCR2.1 (Invitrogen) according to the instructions of the manufacturer. Nucleotide sequencing was carried out using a binding domain sequencing kit according to instructions from the provider (PE Biosystems) Sequencing was carried out on an ABI 310 automatic sequencer (PE Biosystems) using the M13 forward primer (5Ј-GTAAAACGACGGCCAGT-3Ј). All primers were obtained from Invitrogen.
Nm23-M2 Expression Constructs-The cDNA of nm23-M2 (NCBI accession number X68193) was cloned in the pBluescript KS vector. A triple HA tag was added 5Ј of the start codon of the introduced sequence. Artificially introduced NcoI and BspHI restriction sites were used to isolate HA-nm23. Fragments were blunted by adding 1 unit of alkaline phosphatase (Roche Applied Science) for 15 min at 37°C. Subsequently, the fragments were gel-purified and inserted into the HpaI site of the eukaryotic retroviral expression vector pBabe.
Preparation of Retrovirus and Transduction of I/11 Cells-Retroviral transduction was performed as described (18). In short, ecotropic Phoenix cells were transfected with the use of calcium phosphate. After 24 h, cells were treated with mitomycin C (10 g/ml) for 1 h, washed three times, and washed three more times 2 h later. I/11 cells were added and co-cultured for 20 -24 h in StemPro medium supplemented with Epo, SCF, and Dex as described above. Subsequently, I/11 cells were removed carefully from the Phoenix cells and cultured in supplemented StemPro medium. To select for stable transfectants, puromycin (2 g/ ml) was added 48 h later.
Real Time PCR-The real time PCR assay involved TaqMan technology (PE Applied Biosystems model 7700 sequence detector) using the double-stranded DNA-specific fluorescence dye SYBR green I to detect the PCR product as described previously (23). The amplification program consisted of one cycle of 50°C with a 2-min hold (AmpErase UNG incubation) and one cycle of 95°C with a 10-min hold (AmpliTaq Gold Activation) followed by 40 cycles of denaturation at 95°C for 15 s, annealing at 62°C for 30 s, and extension at 62°C for 30 s. All of the different primer pairs had identical optimal PCR annealing temperatures. Acquisition of the fluorescence signal from the samples was carried out at the end of the elongation step. To confirm amplification specificity, the PCR products from each primer pair were subjected to agarose gel electrophoresis, and the dissociation curve was checked at the end of each run.

Nm23-M2 Expression Is Regulated by Translational Con-
trol-Similarly to primary erythroid progenitors, the erythroid cell line I/11 proliferates in the presence of Epo, SCF, and Dex, although the cells undergo terminal differentiation into enucleated erythrocytes in medium supplemented with Epo and insulin (1,2). We aimed to identify genes whose expression is controlled by Epo plus SCF to serve as endogenous targets in the elucidation of Epo-and SCF-specific signaling. As signal transduction may control the activation of both transcription and translation, we used polysomebound mRNA, which was derived from I/11 erythroid cells that were factor-deprived and restimulated with Epo, SCF, or Epo plus SCF, to screen ATLAS filters containing 588 cDNA probes (2) (see also the supplementary data available in the on-line version of this article). Among other genes, we found nm23-M2 to be up-regulated by Epo, SCF, or Epo plus SCF ( Fig. 1A; for complete data see the on-line supplement). The highest expression was observed when cells were stimulated with Epo plus SCF. To validate these results, we tested total RNA derived from I/11 cells that were factor-deprived and restimulated by Epo, SCF, or Epo plus SCF for nm23-M2 expression on a Northern blot. Surprisingly, we did not detect altered expression, although the Epo target gene pim1 was up-regulated by Epo, and the SCF target gene c-jun was up-regulated by SCF (Fig. 1B). Subsequently, we hybridized Northern blots containing fractions of subpolysomal and polysome-bound RNA with the nm23-M2 probe. Quantitative analysis showed that, in the absence of factor, Ͻ20% of all nm23-M2 mRNA was present in the polysomal fractions. Upon stimulation with Epo, 53% of nm23-M2 mRNA shifted into the polysomal fractions. In the presence of SCF 64% of the nm23-M2 mRNA was found in the polysomal fractions, and in the presence of Epo plus SCF the percentage was 78% (Fig. 1C). These data suggest that expression of nm23-M2 is regulated at the level of translation initiation rather than by transcriptional control.
Cytokine Regulated Polysome Association of nm23-M2 mRNA Is PI3K-dependent-The signaling routes that may be involved in growth factor-stimulated translational control are the MEK/ERK pathway, resulting in activation of p90RSK and MNK-1/2, and the PI3K/PKB pathway, activating mTOR and S6 kinase (24). To test whether these pathways are involved in growth factor-induced polysome association of nm23-M2 mRNA, I/11 cells were factor-depleted and subsequently SCFstimulated in the absence or presence of the PI3K-inhibitor LY294002, the mTOR inhibitor rapamycin, or the MEK1 inhibitor PD98059. The PI3K inhibitor LY294002 completely abrogated polysome association of nm23-M2 mRNA, whereas the MEK inhibitor PD98059 did not alter the distribution of nm23-M2 transcripts as compared with untreated SCF control samples ( Fig. 2A). Rapamycin only weakly reduced polysome association of nm23-M2 mRNA following SCF treatment. However, rapamycin treatment abrogated SCF-induced phosphorylation of the eIF-4E-binding protein (4E-BP), and PD98059 inhibited SCF-induced phosphorylation of ERK1/2 in parallel experiments, indicating that the quality and concentration used were effective (Fig. 2B).
Decreased nm23-M2 Protein Levels during Erythroid Differentiation Due to the Loss of Polysome Association of nm23-M2 mRNA-Previous studies reported down-regulation of Nm23-M2 protein expression in myeloid and erythroid differentiation (25,26). Our finding that nm23-M2 mRNA associates with polysomes under erythroid "renewal" conditions is in agreement with these findings. To investigate whether nm23-M2 down-regulation was due to a change in translation, we studied nm23-M2 mRNA polysome binding in I/11 cells during terminal differentiation. I/11 cells were shifted from medium supplemented with Epo, SCF, and Dex to medium supplemented with Epo plus insulin. I/11 cells differentiated synchronously as shown previously (1,27); cell volume is gradually reduced, and hemoglobin accumulation is accelerated between 36 and 48 h after differentiation induction (Fig. 3A). Western blotting showed that the Nm23-M2 protein is downmodulated between 36 and 48 h upon differentiation induction (Fig. 3B). Northern blots containing fractions of subpolysomal and polysomal mRNA derived from cells at different stages of differentiation showed that nm23-M2 mRNA is present in large polysomes 6 and 20 h after the induction of differentiation but that polysome association is largely lost 48 h after differentiation induction (Fig. 3C). To examine whether polysome recruitment is associated with PI3K/PKB activation, we differentiated I/11 cells for 0, 24, and 48 h, factor-depleted the cells for 4 h, and restimulated with 5 units/ml Epo and 10 units/ml insulin for 10 min. Both Epo and insulin induced comparable phosphorylation of PKB in expanding progenitors. Following 24 h differentiation, insulin-induced phosphorylation of PKB was increased, which may explain the increased polysome recruitment of nm23-M2 at this stage of differentiation (Fig. 3, C and  D). As observed previously (18), PKB phosphorylation is downregulated 48 h following differentiation, as expression is strongly decreased (Fig. 3D). The loss of polysome association 48 h after differentiation induction coincides with major changes in the gene expression program of the cell (2).
Nm23-M2 mRNA Starts with a TOP Tract and Contains an Inverted Repeat-Translational regulation may occur through specific sequences and structures present in the 5Ј-and/or 3Ј-UTR of the mRNA. We obtained 5Ј-fragments of nm23-M2 cDNA from a cDNA library by PCR. Fragments of ϳ170 bp, primed from the ATG in nm23-M2 to the 5Ј-cloning site of the vector, were isolated and cloned. Three independent clones were sequenced, which started at their 5Ј-site with a pyrimidine stretch of 8, 11, and 13 nucleotides, respectively. These clones were identical to a cDNA submitted by the RIKEN Genomic Sciences Center (AKO12447), 2 which had a 12-nucleotide TOP sequence (Fig. 4). TOP tracts are structures known to be essential in the control of growth factor-dependent translation of several mRNAs, including those for ribosomal proteins (28). In addition, we observed a 9-bp inverted repeat in the 5Ј-UTR of nm23-M2 mRNA (indicated as a stem structure in 2 Y. Hayashizaki, unpublished data, submission date July 10, 2000.  Fig. 4). Mapping of the novel 5Ј-cDNA sequence to genomic DNA revealed that the 90 most 5Ј nucleotides of the mRNA are located on a separate exon that is separated from the second exon from which the nm23-M2 open reading frame starts by an intron of 241 nucleotides. The TOP tract maps to a genomic sequence of 14 pyrimidines.
Translation of nm23-M2 mRNA Expressed from a Retroviral Vector Is Cytokine-independent-To examine whether the TOP sequence and the inverted repeat in the 5Ј-UTR of nm23-M2 determine the efficiency of polysome association, I/11 cells were transduced with a pBabe HA-Nm23-M2 retroviral expression construct in which the 5Ј-UTR is replaced by the HA tag sequence. HA-Nm23-M2-expressing clones were factor-deprived and restimulated with Epo and SCF, and RNA was separated into subpolysomal and polysomal fractions that were assayed for endogenous and exogenously expressed nm23-M2. The polysome-associated fraction of HA-nm23-M2 mRNA was refractory to factor deprivation, in contrast to the endogenous mRNA that was consistently only present in the polysome-associated fractions when cells were restimulated with Epo/SCF (Fig. 5). Upon restimulation of the cells with Epo and SCF in the presence of LY294002 or rapamycin, polysome association of endogenous nm23-M2 mRNA was lost, but polysome association of HA-nm23-M2 was not altered. This shows that the 5Ј-UTR sequences are responsible for PI3K-dependent regulation of the nm23-M2 mRNA translation.
PI3K-dependent Polysome Association of RpS4, EF1B2, and nm23-M2 mRNA-We next compared the translation of nm23-m2 to other mRNAs that meet the following criteria: (i) are known to posses a TOP-tract i.e. ribosomal protein S4 (rpS4) and elongation factor 1␤2 (EF1␤2); (ii) are known to be regulated at the level of translation by a very different mechanism in other cell systems, i.e. FLI-1 (29); and (iii) are known to be constitutively translated i.e. ␣ tubulin. The cells were treated as described above, and the subpolysomal and polysomal fractions were isolated, pooled into two fractions, and transcribed into cDNA. Real time PCR (TaqMan; PE Applied Biosystems) was performed to quantitatively determine mRNA expression (Fig. 6). RpS4, EF1B2, and nm23-M2 mRNAs showed a significant shift from the subpolysomal to the poly-somal RNA pool in response to Epo/SCF restimulation. The increase in polysome-associated mRNA was 10-fold for nm23-m2 and 4 -5-fold for rpS4 and EF1B2. Furthermore, polysome association of nm23-M2, rpS4, and EF1B2 was almost completely abrogated by the PI3K inhibitor LY294002 and clearly reduced by rapamycin. In contrast, Epo/SCF induced only a minor increase in polysome association of FLI-1 or ␣-tubulin mRNAs (Ͻ2-fold), and the presence of LY294002 or rapamycin did not affect polysome association of these transcripts more than what would be expected from general metabolic activation via the eIF-4E pathway. In conclusion, the TOP sequences present in the mRNAs of various translationally regulated genes (rpS4, EF1B2, and nm23-M2) renders the mRNAs sensitive to PI3K-dependent mRNA translational control. As polysome recruitment of nm23-M2 is regulated more tightly than rpS4 and EF1B2, the inverted repeat in the 5Ј-UTR of nm23-M2 likely contributes to factor-dependent mRNA translation.

DISCUSSION
Expression of the putative oncogene nm23-M2 in erythroid progenitors is controlled by Epo and SCF, factors that induce renewal in erythroid progenitors. Here, we demonstrate that mRNA translation is strictly controlled by pathways dependent on PI3K and mTOR, whereas nm23-M2 gene transcription is constitutively high. The nm23-M2 transcript starts with an oligopyrimidine tract and contains an inverted repeat, structures known to invoke control of translation by growth factors (28,30). This TOP hairpin structure was shown to be essential for translational control of the nm23-M2 gene Regulation of Nm23-M2 mRNA Translation-Control of mRNA translation is increasingly recognized as an important additional regulatory mechanism in gene expression regulation. Internal ribosomal entry sites were found in eukaryotic mRNA and are (for instance) frequent present in mRNAs of genes involved in apoptosis to specifically assure the translation of these mRNAs under conditions that impair cap-dependent translational initiation of mRNA. Because the synthesis of ribosomes is extremely energy consuming, translation of ribosomal proteins and a number of translation factors is restricted FIG. 4. The 5-UTR of nm23-M2 mRNA contains a TOP tract and an inverted repeat. The 5Ј-UTR of nm23-M2 mRNA was cloned by reverse transcription PCR from a cDNA library of I/11 cells. Three independent clones were sequenced, and their respective 5Ј-ends are indicated by vertical arrows above the sequence. The open arrow below the underlined sequence is the start of cDNA AK012447, which has been deposited to the data base of the RIKEN Genomic Sciences Center. The cDNA of nm23-M2 started with a stretch of pyrimidines (italics and underlining). The 5Ј-UTR contains a 9-nucleotide inverted repeat creating a potential stem-loop structure as indicated with a predicted free energy of Ϫ15.4 kcal/mol. Within the gene, the TOP and potential stem-loop sequence are separated from the translation start site by an intron of 214 nucleotides (arrow below the sequence). The protein-coding region is marked in bold, and the encoded amino acids are shown. nt, nucleotide.
to conditions under which both nutrients and mitogens are abundant. This type of regulation is dependent on a TOP tract at the 5Ј-end of the respective mRNAs (28) that causes inhibition of translation in the absence of growth factors and nutrients. It is crucial that the very first nucleotide following the cap structure is a pyrimidine, and a stretch of five pyrimidines is sufficient to execute its function (28). A TOP tract is usually followed by a GC-rich sequence. The nm23-M2 TOP sequence consisting of 14 pyrimidine residues meets these requirements. The mechanism by which a TOP sequence inhibits translation is not known. It has been suggested that specific proteins bind TOP sequences to inhibit translation (31).
Other structural motives present in UTRs are also known to be critical regulators of growth factor-dependent translational control of gene expression. Nm23-M2 contains an inverted repeat able to form a stem-loop structure. However, the free energy of this structure (Ϫ15.4 kcal/mol energy as calculated by RNAstructure, version 3.6; Ref. 32) is too small to impair scanning by the initiation complex and can easily be resolved by eIF-4A. The structure may, however, be stabilized by an interaction with RNA-binding proteins. For example, the stemloop structure in ferritin mRNAs, the "iron responsive element," is stabilized by a protein whose association with the mRNA is controlled by iron (30). Similarly, the potential stemloop structure in nm23-M2 may be stabilized by a protein whose association may be dependent on its phosphorylation status. Several proteins have been shown to be able to interact with the 5Ј-UTR of mRNAs, for instance P56 (33,34), the La autoantigen (35), and the cellular nucleic acid-binding protein (CNBP) (36). Ro60 is involved in the mutually exclusive binding of La or CNBP to 5Ј-UTRs (37). We do not yet know whether the TOP sequence and the inverted repeat in nm23-M2 mRNA both contribute to its translational control. A comparison with known TOP mRNAs that contain no other regulatory domains show that nm23-M2 mRNA translation is more strictly controlled by growth factor signaling than is the translation of ribosomal protein S4 and elongation factor-1␤ (EF1␤) This finding suggests that both types of regulatory mechanisms may contribute to the translational control of nm23-M2.
PI3K Regulates nm23-M2 mRNA Polysome Association-Translation of most TOP mRNAs is under the control of mTOR and PI3K-dependent pathways (38). Consistent with this fact, the translation of nm23-M2 mRNA is fully repressed upon the inhibition of PI3K. During differentiation, nm23-M2 mRNA is released from polysomes 48 h after differentiation induction concurrent with the loss of PKB activity (18). However, regulation via the inhibition of mTOR by rapamycin appears to be less stringent, similar to that of EF1␤ and rpS4, whose translation is also less efficiently repressed by rapamycin. Therefore, the effectiveness of LY294002 compared with that of rapamycin may be specific for erythroid progenitors and/or the conditions used rather than being a property of nm23-M2 mRNA. PI3K activates PKB and mTOR, which results in phosphorylation of eIF-4E-binding proteins (4E-BP) and the release of eIF-4E. Increased levels of available eIF-4E are thought to facilitate scanning of the translation initiation complex past structural obstacles like TOP sequences and hairpin structures. In addition, the PI3K pathway may result in the phosphorylation and release of RNA-binding proteins that stabilize the mRNA secondary structure, although such mechanisms have not yet been identified. In erythroid progenitors, the cooperation of Epo with SCF induces renewal and delays differentiation (39). The inhibition of PI3K abrogates renewal and induces differentiation (1), whereas inhibition of the MEK/ERK pathway does not affect the balance between expansion and differentiation. This suggests that PI3K targets like nm23-M2 may be involved in renewal induction. Accordingly, the enhanced activity of PI3K and its downstream targets is also crucial in tumors induced by a mutated PI3K (v-p3k) functioning as an oncogene in avian sarcoma virus 16 (ASV16)-induced tumors (40) and in Friend spleen focus-forming virus-induced erythroleukemia (41). However, limited attempts to show the effects of Nm23 overexpression in murine I/11 erythroblasts failed to detect clear effects of Nm23 on renewal, survival, or factor dependence. Thus, Nm23 may function in myeloid or multipotent cells rather than in erythroid cells and/or may require cooperation with other, to date unknown, oncogenic events to cause leukemia.
The nm23 Gene Family and Malignant Transformation-Enhanced expression of the nm23 family of genes has been detected in a variety of malignancies (42)(43)(44)(45)(46) and frequently correlated with a poor differentiation stage of certain tumors (42). Although this observation was shown for both mRNA and protein levels, the relation between mRNA and protein levels was rarely addressed. Therefore, it is not clear whether aberrant control of nm23-M2 mRNA translation is involved FIG. 5. Deletion of the 5-UTR abrogates translational control of nm23-M2 mRNA by SCF. The 5Ј-UTR of nm23-M2 was substituted with a short artificial sequence followed by the sequence encoding an HA tag. HA-Nm23 expressing I/11 cells were factor-deprived (NF) and restimulated with Epo and SCF in the absence (Epo/SCF) or presence of 15 M LY294002 (Epo/SCF/LY) or 20 ng/ml rapamycin (Epo/SCF/Rapa). RNA isolated from fractions of the sucrose gradient was tested for the presence of nm23-M2 by Northern blot analysis using a nm23-M2specific probe. HA-nm23-M2 (top four sections) can be discriminated from endogenous nm23-M2 (middle four sections) by its slower electrophoretic mobility. Percentages of polysome-bound nm23 M2 mRNA are indicated at the right of the sections. Lowest section, representative staining for ribosomal RNA by ethidium bromide (SCF sample). LY, LY294002; Rapa, rapamycin.
in altered nm23 protein expression in malignancy. Most likely, transcriptional as well as translational control mechanisms regulate nm23-M2 expression. In retrovirally induced mouse leukemias, integrations occur both upstream and downstream of the nm23-m2 ORF (14). Many integrations, however, occurred downstream of the transcription initiation site and upstream of the ATG start codon. In these cases, the retroviral long terminal repeat could act as a promoter element to give rise to a fusion transcript in which the translational control elements of nm23-M2 (TOP sequence and repeat) are replaced with viral sequences that ensure rapid scanning of the translation initiation complex. As a result, viral integration may abolish regulation of nm23-M2 expression by growth factors/cytokines and, thus, promote leukemogenesis. Unfortunately, it appears to be impossible to isolate polysome-associated mRNA from stored primary mouse leukemia samples. Therefore, the effect of virus integration on the translational regulation of nm23-M2 in the leukemias containing long terminal repeat-nm23-M2 fusion transcripts will have to await future analysis.
Oncogenesis and Translational Control-Control of translation initiation via mRNA-specific mechanisms is increasingly recognized as a potentially important control level in lineage determination, cell survival, proliferation, differentiation, and disease (47,48). The oncogenic role of Ras and PKB in glioblastoma was shown to involve differential recruitment of existing mRNAs to polysomes (49). Overexpression of eIF-4E, the limiting factor in translational initiation, is tumorigenic on its own (50 -52). In addition, the expression of many proto-oncogenes appears to be controlled at the level of translation, because it allows rapid changes in protein levels. For example, the expression of AML1 is regulated through usage of alternative promoters coupled with internal ribosomal entry site-mediated translation control (53). An internal ribosomal entry site is also present in the 5Ј-UTR of c-myc. In several bone marrow samples from patients with multiple myeloma there is aberrant translational regulation of c-Myc, and this correlates with a C 3 T mutation in the c-Myc-internal ribosomal entry site (54). The proto-oncogene Fli-1 is expressed as two protein isoforms generated by alternative translation initiation from two highly conserved in-frame initiation codons (29). Overexpression of Fli-1 is associated with multiple, virally induced leukemias in mouse, and the human counterpart is translocated in Ewing tumors (55).
Several mechanisms for retrovirally induced malignant transformation have, to date, been identified, ranging from the activation of transcription to mRNA stabilization or gene inactivation. Data presented in this paper suggest that an important mechanism may so far have been underestimated, namely the replacement of untranslated sequences important for translational control in mRNAs encoding potential oncogenes such as nm23, leading to loss of such translational control by growth factors/cytokines.