Igbp1 is part of a positive feedback loop in stem cell factor–dependent, selective mRNA translation initiation inhibiting erythroid differentiation

The balance between expansion and differentiation of hematopoietic progenitor compartments is controlled by cytokines and growth factors. In erythropoiesis, stem cell factor (SCF), the ligand for cKit, cooperates with glucocorticoids and erythropoietin (Epo) to suppress differentiation and sustain renewal divisions of erythroblasts in vitro^1,2  as well as in vivo.^3,4  Activation of phosphotidylinositol-3 kinase (PI3K) is essential for proliferation of erythroblasts.^2,5-8  Activated PI3K generates PIP3, which serves as an anchor for multiple PH-domain–containing proteins, such as protein kinase B (PKB). Although both Epo and SCF induce activation of PI3K in erythroblasts, the efficiency with which respective downstream signaling pathways are activated shows large differences.^9,10  In cultured erythroblasts, the activation of PKB is much more responsive to SCF compared with Epo. PKB activates mammalian target of rapamycin (mTOR) through phosphorylation of the tumor suppressor complex Tsc1/Tsc2 (tuberous sclerosis protein), which releases RAS homolog enriched in brain (Rheb) to phosphorylate mTOR.^11,12  Activation of mTOR results in phosphorylation and activation of p70S6 kinase (Rps6kb1)¹³  and hierarchical phosphorylation of 4E-binding protein (4EBP),^14,15  resulting in release of the mRNA cap-binding factor eukaryotic initiation factor 4E (eIF4E).¹⁶  Subsequently, eIF4E can bind the scaffold protein eIF4G to the site otherwise occupied by 4EBP, which enables the formation of an eIF4F-scanning complex containing eIF4E, eIF4G, and the RNA helicase eIF4A.¹⁷  eIF4F associates with several other translation factor complexes, including eIF3, eIF2, and the ribosomal subunits.¹⁸  This complex scans the 5′UTR for the first AUG codon in an appropriate sequence context to start protein synthesis.¹⁹  eIF4E is the rate-limiting factor in the scanning process,^20,21  and its release on phosphorylation of 4EBP is a crucial control mechanism in the recruitment of mRNAs to polysomes. Importantly, transcripts with a short and simple 5′UTR show a limited sensitivity to 4EBP phosphorylation, whereas transcripts with a long and structured 5′UTR or with a terminal oligopyrimidine tract (TOP) are highly sensitive to the concentration of eIF4F complexes in the cell.^19,22,23 

The mechanism by which mTOR controls 4EBP phosphorylation involves regulation of 4EBP-phosphorylation but also regulation of the serine/threonine phosphatase Pp2a,^24-27  which can dephosphorylate p70S6K and 4EBP, thereby suppressing translation initiation. Pp2a exists in various complexes that shift target specificity depending on the binding of regulatory components. Binding of the α4 subunit (designated immunoglobulin binding protein 1 [Igbp1]) to the Pp2a catalytic subunit modifies Pp2a phosphatase activity. mTOR stimulates the formation of the Pp2a-α4 complex, which was shown to sequester the phosphatase activity away from its own downstream targets 4EBP and p70S6 kinase.^26,28  The effects of Pp2a-α4 complex formation, however, may be context dependent,²⁹  and Igbp1(α4) has additional functions, for instance, in stabilizing the interaction of the midline complex with microtubules.³⁰ 

We showed that 4EBP is potently phosphorylated by SCF,¹⁰  and described 3 transcripts that require SCF-induced PI3K/mTOR activation to be recruited to polysomes, whereas SCF does not affect their transcript levels.¹⁰  We also demonstrated that overexpression of eIF4E increased the levels of eIF4F complexes and suppressed erythroid differentiation in the absence of SCF. Thus, mechanisms that control eIF4F formation are important to regulate the balance between expansion and differentiation in erythropoiesis,¹⁰  in line with reports stating that overexpression of eIF4E in tumor samples is associated with increased malignancy.^31-34  Apparently, selective mRNA translation plays a major role in erythropoiesis, but knowledge on mRNAs subject to factor-dependent polysome recruitment remained scarce.

In this study, we compared factor-dependent mRNA expression in total and polysome-bound (pb) mRNA and identified 111 transcripts that require PI3K or increased eIF4F levels for polysome recruitment. From these we selected 9 genes suspected to be involved in signal transduction or gene expression and analyzed their expression regulation and biologic function in erythroblasts. Except for one gene that was regulated by both gene transcription and mRNA translation, the selected genes were strictly regulated by polysome recruitment in response to SCF-induced activation of PI3K and eIF4F formation. Strikingly, we identified the Pp2a-associated protein Igbp1 (α4 subunit of Pp2a) as a target of SCF-dependent polysome recruitment. Constitutive expression of Igbp1 in erythroblasts enhanced the mTOR-dependent phosphorylation of S6K and 4EBP. Exogenous Igbp1 strongly impaired differentiation of erythroblasts and enhanced polysome recruitment of other target gene mRNAs identified in this screen.

I/11 cells were cultivated in StemPro-34 (Invitrogen, Breda, The Netherlands).²  For expansion, the medium was supplemented with 0.5 U/mL of Epo, (kind gift from Ortho Biotech Products, Tilburg, The Netherlands), 100 ng/mL of SCF (supernatant of CHO producer cells), and 10⁻⁶ M of dexametasone (Dex; Sigma-Aldrich, St Louis, MO). To induce differentiation, cells were cultivated in StemPro-34 supplemented with 5 U/mL of Epo and 0.5 mg/mL of iron-loaded transferrin (Scipac, Sittingbourne, United Kingdom). Cell numbers and cell size distribution were determined using an electronic cell counter (CASY-1; Schärfe-System, Reutlingen, Germany). LY294002 (10 μM in final volume) and rapamycin (40 ng/mL) were obtained from Alexis (Lausen, Switzerland); actinomycin D (10 μg/mL) and cycloheximide (50 μg/mL) were from Sigma-Aldrich.

Small aliquots of the cultures were removed and analyzed for hemoglobin content by photometry as described.^9,35  Cell morphology was analyzed in cytospins stained with histologic dyes and neutral benzidine,³⁶  using an OlympusBX40 microscope (40× objective, NA 0.65), an OlympusDp50 CCD camera, and Viewfinder Lite 1.0 acquisition software. Images were cropped using Adobe Photoshop version 6.0 (Adobe Systems, Mountain View, CA).

A MIAME-compatible description of sample preparation and hybridization protocols is given in Document S1 (available on the Blood website; see the Supplemental Materials link at the top of the online article). Microarray data were analyzed using the Rosetta Resolver system, as described.³⁷  All microarray data have been deposited at http://www.ebi.ac.uk/microarray-as/ae/browse.html?keywords=E-MEXP-1689.

For acute stimulation with growth factors, proliferating I/11 cells were washed twice with phosphate-buffered saline (PBS) and seeded at 4 × 10⁶ cells/mL in plain Iscove modified Dulbecco medium (IMDM, Invitrogen). After 4 hours of factor deprivation, cells were stimulated at 37°C with SCF (100 ng/mL) or Epo (5 U/mL). Cells were harvested by addition of ice-cold PBS. Cell lysates, sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), immunoprecipitation, and Western blotting were performed as described³⁸ ; 10 μL of protein extract (1 × 10⁶ cells) was loaded onto a 15% polyacrylamide gel. The antibodies used were α-4EBP1 (Cell Signaling Technology, Danvers, MA), α-Uhmk1 (gift from M. Boehm, National Institutes of Health/National Heart, Lung, and Blood Institute), α-Igbp1 (gift from D. L. Brautigan, Center for Cell Signaling, University of Virginia, Charlotesville), S6K-P (Cell Signaling Technology), and α-Fli1 and α-Myc (Santa Cruz Biotechnology, Santa Cruz, CA).

The coding sequences of selected genes were amplified from cDNA derived of I/11 cells using Pfu polymerase (M7741; Promega, Leiden, The Netherlands) and primers designed to insert an EcoRI at the 5′ end and a ClaI site at the 3′ end of the polymerase chain reaction (PCR) product (Table S4). The PCR product was inserted in TA vector (KNM2040-01; Invitrogen); subsequently, the EcoRI/ClaI fragment was transferred to a pBlueScript vector. The PCR primer inserted the ATG of the coding sequence in frame to a 6-myc-tag sequence at the EcoRI site of the pBlueScript vector. Next, a BamHI/SalI fragment containing the N-terminal myc-tag and the coding sequence was inserted into the retroviral expression vector pBabe. Retroviral transduction was performed as described.⁹ 

Isolation of polysomal RNA by sucrose gradient fractionation was performed as described.^10,39  Total RNA was isolated by the same protocol, omitting the density centrifugation. Cell extracts were layered on a 4-mL linear sucrose gradient (15%-40% sucrose wt/vol), and 8 fractions were collected. Northern blotting indicated that fractions 1 to 4 contain nonpolysomal and subpolysomal mRNA, whereas fractions 5 to 8 consisted of pb RNA. These fractions were pooled to generate subpolysomal and polysomal mRNA of each sample. RNA was quantified by UV-absorbance. Poly(A)⁺ mRNA was purified and cDNA was generated as described.³⁹ 

Real-time PCR was performed using SYBR green and a Taqman PCR machine (model 7700 sequence detector, Applied Biosystems, Foster City, CA).⁴⁰  The amplification program started with 2 minutes at 50°C (AmpErase UNG incubation), 10 minutes at 95°C (AmpliTaq Gold Activation), followed by 40 cycles of 15 seconds' denaturation at 95°C, 30 seconds of annealing at 62°C, and 30 seconds' of extension at 72°C. All primer pairs had similar PCR annealing temperatures. To confirm amplification specificity, the dissociation curve was checked at the end of each run, and PCR products from each primer pair were checked by gel electrophoresis. Gene-specific primers are listed in Table S5 and were obtained from Invitrogen or Sigma-Genosys (The Woodlands, TX).

To identify mRNA transcripts that are selectively recruited to polysomes on growth factor stimulation of erythroblasts, we compared factor-induced gene expression at the level of total and pb RNA using mRNA profiling. We used immortalized p53^−/− erythroblasts having a lifespan that allows analysis on large numbers of cells while closely resembling primary progenitors.¹  The lack of p53 will make a difference, for instance, in response to DNA damage, but not in response to mitogenic factors. Erythroblasts were factor-deprived (4 hours) and restimulated (2 hours) with Epo plus SCF, or left unstimulated. Four independently generated replicates were prepared from 2 established p53^−/− erythroblast cultures with different genetic background (I/11 and R10).^2,41  Total and pb RNA were isolated twice from each culture, cRNA was generated and hybridized to Affymetrix oligonucleotide arrays. Rosetta Resolver software was used to normalize and analyze the intensity data as described³⁷ ; samples are described in Table S1. The ratio of gene expression in Epo/SCF stimulated samples (ES) versus factor-deprived cells (NF) was calculated for total and pb mRNA (Figure 1A). Next, we applied a 2-step selection to identify genes differentially regulated in total versus pb RNA. First, genes with sufficient variance between the ES/NF ratios for total and pb RNA were selected by analysis of variance (P = .01). Second, selected genes had to be differentially expressed in presence or absence of Epo/SCF in total or pb mRNA in either both I/11 or in both R10 hybridizations (P = .001; for details on this strategy, see Document S1). This approach selected 115 probe sets representing 111 different transcripts subject to Epo/SCF-controlled polysome recruitment (Table S5). To relate differential polysome recruitment to gene expression during differentiation, we derived pb mRNA from steady-state expanding and differentiating erythroblasts (48 and 60 hours after differentiation induction; again, 4 independent replicates). After hybridization of oligonucleotide arrays, the gene expression ratio of differentiation over steady-state renewal was calculated. Subsequent cluster analysis of gene expression ratios of the 115 selected probe sets resulted in a matrix that groups the selected probe sets into 5 separate clusters based on Epo/SCF-controlled gene expression detected in total and pb RNA and on expression during differentiation (Figure 1B; Table S5).

Genes represented in clusters 1, 3, and 5 (83 genes) were subject to Epo/SCF-induced polysome recruitment; clusters 1 and 3 were also characterized by decreased expression during differentiation. To gain further insight into regulation of gene expression in the various clusters, we compared normalized intensity data obtained from the array hybridization with total and pb RNA from factor-deprived and Epo/SCF-restimulated cells (Figure S1). Cluster 1 represents genes with constitutive expression in total RNA and Epo/SCF-induced polysome recruitment. Cluster 3 represents genes of which transcription is increased by growth factor stimulation, but polysome recruitment shows a more pronounced increase in response to Epo/SCF. Cluster 5 represents genes characterized by Epo/SCF-induced repression of expression in total mRNA, but with maintained levels in pb RNA, which implies increased Epo/SCF-induced polysome recruitment from a smaller transcript pool. Genes present in clusters 2 and 4 are subject to enhanced polysome recruitment after factor deprivation, which is abrogated by Epo/SCF restimulation (Figure S1). These genes may play a role in the initiation of stress responses. Only genes represented by cluster 4 were up-regulated during differentiation.

Next, 9 genes were selected that (1) require Epo/SCF for polysome recruitment, (2) are down-regulated in differentiation, (3) are suggested to function in signal transduction or control of gene expression, and (4) were hitherto not known to be translationally regulated (Table 1; for details on these genes, see Table S2). Quantitative reverse-transcribed PCR (Q-PCR) was used to analyze whether polysome recruitment was dependent on PI3K and mTOR, and on eIF4E expression.

First, we tested how signal-dependent expression in total and pb RNA correlates between array data and Q-PCR. We determined transcript levels in subpolysomal and pb RNA by Q-PCR and calculated the percentage polysome recruitment. In addition to the 9 selected genes, Nm23-M2 (Nme2) and Ybx1 were tested as positive controls.¹⁰  A control gene, Fli-1, that is not regulated at the level of RNA-specific polysome recruitment, showed at most a 2-fold difference in polysome recruitment in response to Epo or SCF (Table 1; Figure 2C,F). In contrast, Nme2, Ybx1, and 8 of 9 of the selected genes showed a 10-fold increase or more in polysome recruitment in response to Epo plus SCF; only Grwd1 failed to reproduce the regulation detected on the arrays (Table 1; Figure 2A,B,D,E). The SCF-induced increase in polysome recruitment exceeded Epo-induced polysome recruitment, whereas Epo plus SCF showed an additive or even synergistic effect (Table 1, Figure 2D,E). Cluster 1 genes Igbp1, mEd2, Rnf138, Nap1l1, and Cnih, cluster 3 gene Nubp1, and cluster 5 gene Uhmk1 were almost exclusively up-regulated in pb but hardly in total mRNA (Table 1); cluster 5 gene Hnrpa1 was down-regulated in total mRNA in accordance with the array data (Tables 1 and S5). Inhibition of PI3K (LY294002) or mTOR (rapamycin) decreased polysome recruitment of the control gene Fli-1 approximately 2-fold but completely dissociated all other genes from polysomes (Table 1; Figure 2D-F).

Overexpression of eIF4E is expected to render the selected transcripts less sensitive to Epo/SCF-induced PI3K activity.¹⁰  Indeed, polysome recruitment of all transcripts became partially factor independent in cells overexpressing eIF4E (Table 1 and Figure 2G,H; for cells, see Blazquez-Domingo et al¹⁰ ). Effects in the Fli-1 controls were again maximally 2-fold (Figure 2I).

We previously showed that PI3K activity is required for the expansion of erythroblast cultures. Overexpression of eIF4E rendered cells independent of PI3K activation and impaired differentiation.^2,10  Therefore, proteins whose expression is controlled by PI3K- and eIF4E-dependent mRNA polysome recruitment are expected to be required during erythroblast proliferation. pb mRNA was isolated from I/11 cells induced to differentiate, and Q-PCR showed that the selected genes were all down-regulated during differentiation (Table 1; Figure 2J,K). The erythroid transcription factor Nfe2 is known to be up-regulated during differentiation and is shown as a control (Figure 2L).

In conclusion, translation of 10 of 11 transcripts selected from 83 genes subject to Epo/SCF-induced polysome recruitment (Igbp1, mEd2, Rnf138, Nap1l1, Cnih, Nubp1, Uhmk1, Hnrpa1, Nm23-M2, and Ybx1) was strictly controlled by the PI3K-mTOR-eIF4E pathway.

The high-density complexes with which the transcripts associated in response to PI3K activation and eIF4E expression are assumed to be polysomes. Potentially, however, these could represent high-density hnRNP complexes. To examine whether the association of transcripts in high-density complexes reflects protein expression, we analyzed protein expression of Igbp1 and Uhmk1 on Western blot and compared it with factor-dependent protein expression of Fli-1. In contrast to Fli-1, expression of Igbp1 and Uhmk1 was more sensitive to inhibition of translation compared with inhibition of transcription (Figure S2). Restimulation of factor-deprived erythroblasts with Epo and SCF rapidly increased Igbp1 and Uhmk1 protein expression, which was inhibited by rapamycin and CHX. Fli-1 expression, however, was inhibited by CHX, but not by rapamycin (Figure S2). Igbp1 protein expression was also enhanced in cells constitutively expressing eIF4E (Figure S5). Thus, factor-dependent Igbp1 and Uhmk1 protein expression was appropriately reflected by RNA present in high-density polysomes.

Because SCF-induced activation of the PI3K-mTOR-eIF4E pathway sustains expansion and delays differentiation of erythroblast cultures, we investigated the functional role of the proteins requiring this pathway for their expression. From the 8 selected genes that showed PI3K-dependent polysome recruitment, 7 were expressed from retroviral expression vectors downstream of a myc-epitope tag (introducing the ORFs without the UTRs that control translation; for unclear reasons, Nubp1 resisted cloning in pBabe). Expression of selected genes in Phoenix cells was checked on Western blots, and correct cellular distribution of these proteins was assessed by immunofluorescence (Figure S3). The expression vectors were transduced into I/11 cells, and single cell-derived clones were established, generating empty vector control clones in every respective experiment.

With 4 of the 7 genes (Uhmk1, Cnih, Rnf138, and Nap1l1), we repeatedly failed to establish I/11 clones stably expressing these genes despite proper transient expression in Phoenix cells. Hnrpa1, Igbp1, and mEd2 (2010315L10Rik) were expressed in Phoenix and I/11 cells, and several single cell-derived I/11 clones were established (Figure S4 and data not shown). To analyze if and to what extent these 3 proteins mimic SCF-induced suppression of differentiation, clones positive for exogenous protein expression were subjected to differentiation conditions (Epo plus iron-loaded transferrin). Differentiation parameters, including cell number, mean cell volume, hemoglobin per cell volume, and cell morphology, were analyzed at various time points (Figure 3; and data not shown). On induction of differentiation, empty vector-transduced control cells showed the expected transient proliferation (∼3 divisions within 48 hours) and hemoglobin accumulation (Figure 3A), accompanied by size decrease and a predominantly mature normoblast/erythrocyte morphology (Figure 3D). Constitutive expression of Hnrpa1 yielded similar results as empty vector (data not shown), whereas expression of Igbp1 and mEd2 impaired differentiation. Erythroblasts constitutively expressing Igbp1 proliferated exponentially for more than 72 hours, remained large, hardly accumulated hemoglobin, and maintained a blast-like morphology (Figure 3B,D). Erythroblasts constitutively expressing mEd2 were partially impaired in differentiation; the cells became smaller but did not accumulate hemoglobin and showed a partially differentiated morphology (Figure 3C,D). In these experiments, we used the mEd2 ORF published to enhance an ERK-dependent reporter construct (PM26).⁴²  We also expressed a longer cDNA, recently described to encode the novel Q-SNARE protein D12 (Use1) in Phoenix cells.⁴³  The intracellular localization of D12 was similar to mEd2, but we failed to express the D12 protein in I/11 cells, suggesting that the truncated protein inhibits differentiation, whereas the full-length protein seems incompatible with erythroblast proliferation.

Transforming growth factor-β (TGF-β) counteracts the effect of SCF and enhances differentiation of erythroblasts in presence of Epo plus SCF (H.B., M.v.L., unpublished data, June 2000). Therefore, we examined whether TGF-β could induce differentiation of cells constitutively expressing Igbp1. In the presence of Epo plus SCF, the constitutive expression of Igbp1 in I/11 cells enhanced proliferation (Figure 4A,B), completely prevented the slow size decrease seen in control cells, and fully inhibited hemoglobin accumulation and morphologic maturation (Figure 4C-G). As expected, addition of TGF-β inhibited proliferation and accelerated differentiation of control cells (Fig 4A,C,E). TGF-β also induced differentiation in Igbp1 expressing erythroblasts but to a much lesser extent. Compared with control cells, the addition of TGF-β to the Igbp1 expressing cells showed delayed onset of inhibition of proliferation (day 5 instead of day 2), size decrease (day 4 instead of day 1), and hemoglobinization (maximal at day 7-8 instead of day 3-4; Figure 4B,D,F). In line with this, TGF-β–treated Igbp1 expressing erythroblasts were still largely immature at the time (t96), whereas most control cells were in an advanced state of maturation (small and hemoglobinized, Figure 4G). In conclusion, TGF-β induced only a limited differentiation in Igbp1-expressing cells, which suggests that Igbp1 may interfere with aspects of TGF-β signaling.

Igbp1 associates with the serine/threonine phosphatase Pp2a to modulate Pp2a phosphatase activity toward its different targets.^26,28  Among these targets are 4EBP and p70S6k.^44,45  We analyzed Epo- and SCF-induced phosphorylation of 4EBP and p70S6k in Igbp1-expressing I/11 cells and respective control cells. As previously shown, SCF but not Epo induced full phosphorylation of 4EBP.¹⁰  On constitutive expression of Igbp1, however, stimulation of erythroblasts by Epo was sufficient to induce hyperphosphorylation of 4EBP (Figure 5A). Phosphorylation of p70S6k required the presence of both Epo and SCF in control cells. Interestingly, constitutive expression of Igbp1 increased Epo-induced phosphorylation of p70S6k to levels obtained only with Epo plus SCF but did not enhance SCF-induced phosphorylation of p70S6k (Figure 5B). Apparently, the activation of p70S6K requires at least 2 cooperating pathways: SCF-induced activation of Igbp1 plus an Igbp1 independent pathway activated by Epo. The enhanced phosphorylation of 4E-BP and p70S6k in Igbp1-expressing cells in the presence of Epo is in line with enhanced proliferation and impaired differentiation of Igbp1-expressing erythroblasts in the presence of Epo only.

Next we tested whether constitutive expression of Igbp1 also enhances translation of transcripts previously found to depend on increased levels of free eIF4E. Expression levels of various genes were measured by Q-PCR in subpolysomal and pb mRNA fractions isolated from empty vector control erythroblasts and erythroblasts constitutively expressing Igbp1. Polysome recruitment of Igbp1 itself was increased on factor deprivation and Epo restimulation (Figure 5D), which is expected as the expressed construct lacks the regulatory sequences responsible for factor-dependent translation. Polysome recruitment of Fli-1, which is not subject to factor-dependent translation, is not affected by Igbp1 expression (Figure 5C). However, 2 transcripts with a terminal oligopyrimidine tract, eIF1β and Rps4, as well as 2 transcripts with a highly structured 5′UTR, mEd2 and Nm23-M2, show increased polysome recruitment, both during steady-state expansion and after Epo stimulation (Figure 5E-H). Together, these data indicate that translational control of Igbp1 is an important positive feedback signal to enhance polysome recruitment of eIF4E-sensitive mRNAs.

Activation of the PI3K-mTOR-eIF4E pathway is an important mechanism by which SCF delays differentiation and sustains proliferation of erythroblasts.¹⁰  This suggests that selective polysome recruitment of eIF4E-sensitive mRNAs is crucial to erythropoiesis. In this study, we identified a unique set of genes that are hardly regulated at the level of gene transcription but depend on SCF-induced, PI3K-dependent increase in eIF4F complexes to be recruited to polysomes. Functional analysis revealed that constitutive expression of one of these genes, Immunoglobulin binding protein 1 (Igbp1, also known as the α4 subunit of Pp2a), strongly impaired terminal differentiation of erythroblasts. Constitutive expression of Igbp1 maintained phosphorylation of 4EBP and p70S6K in the presence of Epo to levels otherwise only obtained in presence of SCF. Igbp1 appeared to function in a positive feedback loop of mTOR-regulated polysome recruitment of eIF4E-sensitive mRNAs (Figure 6). Our data emphasize the important role of selective polysome recruitment in control of gene expression and cell fate determination.

We show that polysome-bound mRNA profiling, integrating gene expression control at the level of transcription, mRNA nuclear export, and polysome recruitment,^46,47  allowed the identification of multiple genes that were regulated specifically at the level of polysome recruitment via Epo/SCF signaling and would not be detected by standard mRNA profiling. One of these genes, Igbp1, was critically involved in regulating erythroid progenitor renewal versus differentiation. This indicates that polysome recruitment is an important level at which signaling-dependent gene expression regulates erythropoiesis. Rajasekhar et al demonstrated PI3K- and Mek1-dependent, selective polysome recruitment of mRNA in v-Ras/v-Akt transformed glioblastoma cells.⁴⁸  Others identified transcripts that are specifically recruited to polysomes on overexpression of eIF4E in mouse embryo fibroblasts⁴⁹  or in human epithelial cells.⁵⁰  We found little overlap between genes identified in our studies because of differences in cell type and experimental approach. Notably, overexpression of eIF4E improved polysome recruitment of, eg, Nm23-M2 and Igbp1 in absence of PI3K activity, but not to levels obtained by full stimulation with Epo/SCF. This suggests that polysome recruitment of the targets identified in our screen may depend on additional mechanisms that the other screens did not select for. Together, however, screens using polysome-bound bound mRNA indicate that regulation of gene expression at the translational level is an important mechanism in development and cancer.^51,52 

SCF signaling stimulates cap-dependent translation and is expected to identify transcripts that require increased levels of the eIF4F complex. It is broadly accepted that these transcripts are characterized by structured 5′UTRs.^22,31  Our list of translationally regulated genes contained ribosomal proteins and some translation factors assumed to contain a TOP sequence⁵³  (Rps10, Rpl10a, Rpl18, Rpl26, Rpl36, and eIF3s12 in cluster 1; Rps5, Rps16, Rpl22l1 and Rpl27a in cluster 5; Table S5). Many other ribosomal proteins, however, failed to pass the thresholds we set for signaling-dependent polysome recruitment, suggesting that a TOP sequence alone fails to impose strong, signal-regulated translational control on a transcript. The genes selected for functional analysis lacked a TOP sequence, but other structural RNA motifs that contribute to control of translation initiation are difficult to recognize. Most of such structured 5′ UTRs are incompletely represented in the databases because structures hamper reverse transcription. Even knowledge of the correct 5′ UTR does not always allow for recognition of the structural configuration that controls polysome recruitment, examples being the stem-loop iron response element⁵⁴  or the consensus pseudoknot structure bound by Fragile-X mental retardation protein (Frm1).⁵⁵  Besides such structures, upstream AUGs (uAUG) may affect polysome recruitment because translation of upstream open reading frames (uORF) can modulate translation of the proper ORF.^56,57  We have been able to elongate the 5′UTR of Igbp1, mEd2, and Nm23-M2 beyond the reported cDNA start, using RACE experiments at increased temperatures to facilitate melting of secondary structure, and detected potential stem-loop structures and uAUGs in these genes (A. Nieradka, G.G., M.v.L., unpublished data, December 2006).

The activity of the central regulator of protein synthesis, mTOR, is modulated by a variety of signals.¹⁸  Polysome recruitment of transcripts that require activation of the PI3K/mTOR/eIF4E pathway are sensitive not only to SCF but also to amino acid starvation and lack of cAMP.^58,59  The serine/threonine phosphatase Pp2a is one of the potential antagonist of mTOR.⁶⁰  Functional Pp2a consists of a catalytic subunit (Pp2a_C), a structural subunit (Pp2a_A), and a variable regulatory subunit (Pp2a_B).^61,62  One of these regulatory subunits is Igbp1, also known as α4. The interaction of Pp2a with Igbp1 was shown to inhibit its activity toward 4EBP and p70S6K,^44,45  but Pp2a activity on other targets is enhanced.³⁰  Pp2a has been implied in mTOR-regulated polysome recruitment as addition of rapamycin disrupts the Pp2a/Igbp1 complex,²⁶  changing the constitution of the Pp2a trimeric protein.⁶³  However, most of these studies used ectopic expression studies, factor-independent cell lines, or otherwise nonphysiologic conditions. We showed here that constitutive expression of Igbp1, similar to addition of SCF, potently attenuated and delayed differentiation of erythroblasts in the presence of Epo. Igbp1 binds and regulates Pp2a and may prevent dephosphorylation of both 4EBP and p70S6K as described,^44,45  which stimulates polysome recruitment of respective signal regulated transcripts important for erythropoiesis. This is supported by the following observations: (1) Constitutive Igbp1 expression enhances phosphorylation of 4EBP and p70S6K in presence of Epo to levels normally reached by Epo plus SCF (Figure 5A,B). (2) TGF-β is not able to induce differentiation of Igbp1 expressing I/11 cells. TGF-β induces association of the catalytic subunit of Pp2a (Pp2a_C) with a different regulatory subunit (Ppp2r5a or Bα) that stabilizes a complex containing Pp2a and p70S6K, in which p70S6K is dephosphorylated and inhibited.⁶⁴  Activation of both the Smad pathway and Pp2a is required for TGF-β–induced inhibition of proliferation.⁶⁴  Increased expression of Igbp1 is expected to counteract the TGF-β-mediated activation of Pp2a by sequestering Pp2a_C, maintaining p70S6K activity in the presence of TGF-β. This could explain why Igbp1 expression counteracts TGF-β–induced differentiation of SCF-treated erythroblasts. (3) Polysome recruitment of previously identified genes is enhanced by constitutive Igbp1 expression (similar to eIF4E overexpression; Figure S5). This suggests that SCF-induced expression of Igbp1 causes positive feedback in polysome recruitment of multiple eIF4E-sensitive mRNAs, some of which could contribute to attenuation of erythroid differentiation. In line with this, another inhibitor of Pp2a, the putative oncogene SET,⁶⁵  is induced in CML,⁶⁶  perhaps contributing to differentiation arrest of the leukemic cells.

It is important to note, however, that Pp2A is a very general phosphatase and that Igbp1 may control multiple cellular processes in addition to phosphorylation of 4EBP and p70 S6K,³⁰  whereas, on the other hand, Igbp1 only modulates a part of all cellular Pp2a (70 regulatory subunits are known to date^61,62 ). Lack of Igbp1 results in very early embryonal lethality,²⁸  and knockdown of Igbp1 would not give any insight into its importance in control of mRNA translation.

Whereas Igbp1 and mEd2 could only be expressed at restricted levels in erythroblasts, and 4 other selected genes could not be expressed at all, Hnrpa1 could be constitutively expressed at abundant levels without altering the erythroblast phenotype. Hnrpa1 is involved in the generation of correct splice variants of the erythrocyte membrane protein Band4.1,^67,68  and incorrect splicing has major consequences in vivo that may not become apparent in vitro.⁶⁹  With respect to the proteins resisting expression, there may be a need to express them at precisely regulated levels or only during a specific phase of the cell cycle. Notably, these genes included Uhmk1/Kis and Cnih, which associate with and control stability of tubulins during mitosis.^70,71  Rnf138 contains ubiquitin ligase and nuclear localization domains. Its constitutive expression may result in degradation of essential nuclear proteins. mEd2 is also known as MAPK-activating protein PM26,⁴²  and its human homolog is uncharacterised hematopoietic stem/progenitor cells protein MDS032. Recently, mEd2 was shown to be a Q-SNARE protein, termed D12 or Use1, involved in endoplasmic reticulum (ER) trafficking.^43,72  Interestingly, reduced expression of D12 resulted in increased phosphorylation of eIF2a. This opens the interesting possibility that mEd2-dependent surveillance of protein quality in the ER may also cause feedback from eIF4E activation to polysome recruitment of mEd2 and subsequent protection of eIF2a from inactivation by phosphorylation.

Constitutive activation of PI3K has frequently been found in both solid tumors and leukemia.^73-78  The strong inhibition of erythroid differentiation by overexpression of eIF4E or by constitutive expression of Igbp1, and the high levels of eIF4E found in several cancer types³¹  indicate that regulation of mRNA translation is a critical event in carcinogenesis downstream of PI3K. Currently, rapamycin homologs are tested as anticancer drugs in a large variety of tumors, yielding promising results.^79-81  Although it is generally assumed that the anticancer effect of rapamycin and its analogs is the result of a general inhibition of protein synthesis in proliferating cells, it is more likely that inhibition of mTOR specifically targets structured mRNAs. Control of translation not only regulates cell growth to reach “start” in the cell cycle but also is an important and selective mechanism to regulate gene expression. However, many translationally controlled genes await further characterization with respect to regulation and function.

The authors thank Dr Victor de Jager for assistance with the Rosetta Resolver software; Dr Ivo Touw for many fruitful discussions and critical reading of the manuscript; Liu Wing for technical assistance; Drs Peter Seither, Andreas Weith (Boehringer Ingelheim, Biberach, Germany), Helmuth Dolznig, Thomas Waerner, and Sandra Pilat (IMP, Vienna, Austria) for mRNA profiling of erythroblasts, of which the complete data will be published elsewhere; Dr Bart Aarts (Erasmus MC, Rotterdam, The Netherlands)for assistance in confocal scanning microscopy; Dr David Brautigan (University of Virginia, Charlottesville) for anti-Igbp1 antibodies; Dr Manfred Boehm (National Institutes of Health/National Heart, Lung, and Blood Institute, Bethesda, MD) for anti-Uhmk1 antibodies; and Ortho-Biotech (Tilburg, The Netherlands) for their kind gift of Eprex (erythropoietin).

This work was supported by grants from the European Union (Brussels, Belgium; HPRN-CT-2000-00 083; MRTN-CT-2004-005499), The Netherlands Organisation for Scientific Research (Den Haag, The Netherlands; 050-10-051), the Dutch Cancer Society/KWF (Amsterdam, The Netherlands; EUR 2000-2230), and the Austrian Science Foundation, FWF (Vienna, Austria; SFB-F28; H.B., E.W.M.).

National Institutes of Health

Contribution: G.G. and M.v.L. designed and performed most of the research and drafted the paper. M.B.-D., A.K., W.J.B., E.W.M., and H.B. designed and/or performed the expression profiling.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Marieke von Lindern, Department Hematology, Erasmus MC, PO Box 1738, 3000 DR Rotterdam, The Netherlands; e-mail: m.vonlindern@erasmusmc.nl.