Cytoplasmic Prep1 Interacts with 4EHP Inhibiting Hoxb4 Translation

Background Homeobox genes are essential for embryonic patterning and cell fate determination. They are regulated mostly at the transcriptional level. In particular, Prep1 regulates Hox transcription in association with Pbx proteins. Despite its nuclear role as a transcription factor, Prep1 is located in the cytosol of mouse oocytes from primary to antral follicles. The homeodomain factor Bicoid (Bcd) has been shown to interact with 4EHP (eukaryotic translation initiation factor 4E homolog protein) to repress translation of Caudal mRNA and to drive Drosophila embryo development. Interestingly, Prep1 contains a putative binding motif for 4EHP, which may reflect a novel unknown function. Methodology/Principal Findings In this paper we show by confocal microscopy and deconvolution analysis that Prep1 and 4EHP co-localize in the cytosol of growing mouse oocytes, demonstrating their interaction by co-immunoprecipitation and pull-down experiments. A functional 4EHP-binding motif present in Prep1 has been also identified by mutagenesis analysis. Moreover, Prep1 inhibits (>95%) the in vitro translation of a luciferase reporter mRNA fused to the Hoxb4 3′UTR, in the presence of 4EHP. RNA electrophoretic mobility shift assay was used to demonstrate that Prep1 binds the Hoxb4 3′UTR. Furthermore, conventional histology and immunohistochemistry has shown a dramatic oocyte growth failure in hypomorphic mouse Prep1i/i females, accompanied by an increased production of Hoxb4. Finally, Hoxb4 overexpression in mouse zygotes showed a slow in vitro development effect. Conclusions Prep1 has a novel cytoplasmic, 4EHP-dependent, function in the regulation of translation. Mechanistically, the Prep1-4EHP interaction might bridge the 3′UTR of Hoxb4 mRNA to the 5′ cap structure. This is the first demonstration that a mammalian homeodomain transcription factor regulates translation, and that this function can be possibly essential for the development of female germ cells and involved in mammalian zygote development.

Prep1 and Pbx1 form stable complexes that regulate the transcription of some Hox genes [1,[3][4][5][6]. Expression of Hox genes is regulated not only at the transcriptional but also at the posttranscriptional level. Indeed, Hoxb4 expression in mouse embryos is restricted by selective translation and/or degradation of its mRNA [7]. Transcriptional and translational regulation of homeobox genes also occurs in Drosophila embryos, where nuclear Bcd regulates the transcription of Hunchback or Even-skipped in the nucleus, while in the cytosol Bcd regulates the translation of Caudal (cad) mRNA [8][9][10]. This cytosolic effect is due to the interaction with Drosophila 4EHP (d4EHP) through a YxxxxxxL motif [11] distinct from the consensus binding site for the eukaryotic translation initiation factor 4E (eIF4E), YxxxxLW (where x is any amino acid and W any hydrophobic residue) [12]. d4EHP binds the 59 cap of cad mRNA, while Bcd binds the 39UTR, preventing the coordinate assembly of the translational machinery [13] .
In most animal species, female gametes contain a pool of stable stored but not translated transcripts in the cytoplasm, including Hox mRNAs [14][15][16][17]. Translation of these mRNAs occurs at meiosis, upon fertilization, and during early embryo development [16], but little information is available about Hox translational regulation and its importance during oocyte development.
Prep1 and Pbx1 are present in the cytosol of mouse oocytes from primary to antral follicles [18]. In early zebrafish embryos, Prep1 and Pbx1 proteins are located in the cytoplasm and they translocate to the nucleus only around gastrulation [6]. So far, no information is available about any specific developmental function of cytosolic Prep1.
Since Prep1 contains a putative 4EHP-binding motif, we have studied a possible cytoplasmic function of Prep1, discovering that Prep1 is involved in a 4EHP-dependent translational regulation of at least Hoxb4 mRNA, and concluding that this function is possibly essential for mammalian female germ cell development.

Prep1 interacts with 4EHP
The 59 YRHPLFPLL 67 amino acid motif of Prep1 (Fig. 1A) is similar to the 66 YNYIRPYL 73 sequence of Bcd, that binds the translation inhibitor 4EHP [11]. This motif is present in all members of the MEIS subfamily of TALE proteins (Fig. 1B), and is conserved among orthologs (Fig. 1C). Interestingly, a glutamic acid (depicted in blue) present in Bcd is conserved in all the MEIS subfamily members. The presence of this motif and the Prep1 cytoplasmic localization in mouse oocytes [18] led us to study the interaction between Prep1 and 4EHP.
Prep1 and 4EHP co-localize in the cytoplasm of mouse primary oocytes (oo), as shown by confocal immunofluorescence analysis ( Fig. 2A-D). From secondary to antral follicles, Prep1 is located in the nucleus of granulosa cells (gc), where 4EHP is mainly cytosolic, and no co-localization is observed (Fig. 2E-L). In contrast, Prep1 and 4EHP still co-localize in the cytosol of oocytes from secondary to antral follicles. The co-localization between Prep1 and 4EHP in the cytosol of antral oocytes is confirmed ( Fig. 2M-P) by deconvolution analysis, which increases image resolution and decreases false positives [19]. As it is shown in Fig. S1A, the 4EHP antibody specifically detects 4EHP but not its close homolog eIF4E. In the case of Prep1 antibody, its specificity has been described previously [2,3].
Prep1-4EHP interaction in ovarian cytosolic extracts was confirmed by co-immunoprecipitation of endogenous 4EHP by an anti-Prep1 antibody (Fig. 3A). However, the 4EHP antibody did not co-immunoprecipitate endogenous Prep1 due to the very low cytosolic Prep1 concentration and/or to the scarce immunoprecipitation capacity of the anti-4EHP antibody (data not shown). We have also investigated if Prep1-4EHP interaction is RNA mediated, but we did not observe any difference with or without RNase A treatment (data not shown). This result suggests that Prep1-4EHP interaction is not RNA mediated.
We further investigated Prep1-4EHP interaction by pulling down in vitro synthesized 35 S-Met-labeled proteins. Prep1-GST and 4EHP-GST beads pulled down 35 S-Met-4EHP and, respectively, 35 S-Met-Prep1 (Fig. 3B). Moreover, 4EHP-GST or Prep1-GST beads were able to pull down endogenous Prep1 or 4EHP from ovarian cytosolic extracts, respectively (Fig. 3C). We observed a doublet for 4EHP when it was produced in vitro, probably due to premature translation terminations. In contrast, a single band was observed for endogenous 4EHP (Fig 3A-C).
We exploited the above technique to identify the 4EHP-binding sequence in Prep1. Mutational analysis of Prep1 showed that the substitution of the conserved tyrosine 59 and leucine 66 residues with alanine (Y59A and L66A, GST-Prep1 Y-L mutant, Fig. 1D) slightly reduced the interaction between mutant Prep1 and 4EHP (data not shown). In contrast, alanine-substitution of Y59 and both L66 and L67 in Prep1 (GST-Prep1 Y-LL mutant, Fig. 1D) strongly reduced the interaction, even if it was not completely abolished (Fig. 3C).
Overall, the results show that Prep1 and 4EHP interact in vivo and in vitro and that the 59 YRHPLFPLL 67 amino acid motif of Prep1 is functional and required for 4EHP-binding. Prep1 protein is associated with a ribosome-free fraction of mouse ovarian cytosol RNA-binding proteins and mRNAs are fractionated in polysomes, ribosomes and ribosome-free fractions by continuos (15-45%) sucrose gradient centrifugation [20]. In mouse ovarian post-nuclear supernatants, Prep1 and 4EHP were found in the first fractions, which do not contain ribosomes or polysomes [20], as assessed by immunoblotting ( Fig. 3D-E). Then, we conclude that Prep1 and 4EHP are not associated with polysomal fractions.

Prep1 co-immunoprecipitates Hoxb4 mRNA
Bicoid homologs have been identified only in close relatives of the schizophoran fly Drosophila. Stauber et al. have shown that Bcd gene originated from a recent duplication of the direct homolog of the vertebrate gene Hox3, termed zerknüllt [21]. Prep1 is not a Hox protein, but belongs to the TALE family of homeodomain proteins, regulating Hox expression at the transcriptional level. For this reason, we decided to investigate if Prep1 could also regulate Hox genes during translation.
RT-PCR analysis with specific primers shows that Hoxb4, 5, 6, 7 and 8 are expressed in the oocyte and associated ganulosa cells (OGC, Fig. 4A). To test whether Prep1 binds mRNAs coding for Hox genes, we immunoprecipitated crosslinked RNA from OGC using a Prep1 antibody (see Material and methods). Degenerated primers (HoxA and HoxB) based on an early nucleotide consensus for vertebrate Antennapedia class homeodomains [18,22,23] (see Materials section) were used to amplify homeobox sequences in the co-immunoprecipitated RNA from OGC. As shown in the top line of Fig. 4B, Hox amplicons were detected by PCR, meaning that Hox RNAs were co-immunoprecipitated by Prep1. After cloning and sequencing those amplicons, we found Hoxb4 and Hoxb8 sequences highly represented among the different clones. Knowing that Hoxb4 and Hoxb8 mRNAs can be co-immunoprecipitated by Prep1, we used specific primers to confirm this result. In fact, we were able to amplify Hoxb4 and Hoxb8 from the co-immunoprecipitated OGC RNA (Fig. 4B, second line, and data not shown for Hoxb8). In contrast, we could not amplify other Hox members from the co-immunoprecipitated RNA, such as Hoxb5 (third line, Fig. 4B), which was present in OGC extracts (Fig. 4A). Prep1, therefore, associates at least to Hoxb4 and Hoxb8 mRNA in oocyte-associated granulosa cells.  antibodies. The RNA was extracted and subjected to RT-PCR with degenerated Antennapedia primers (upper part), which amplified Hox messengers (HoxA and B clusters). After cloning and sequencing of the amplicons, Hoxb4 was highly represented among the amplicons. Then, specific Hoxb4 primers were used to confirm the previous result (middle part), amplifying Hoxb4 mRNA from the OGC co-immunoprecipitated RNA. Notice that specific primers for Hoxb5, which is expressed in OGC but was not identified among the Hox amplicons, is not amplified from the OGC co-immunoprecipitated RNA (lower part). doi:10.1371/journal.pone.0005213.g004 Prep1 and 4EHP co-regulate Luc-39UTR Hoxb4 translation in vitro Since Prep1 associates with 4EHP and at least two mRNAs, these two interactions might be functionally linked. We decided to focus our work in a single mRNA, and we selected Hoxb4 for our studies.
It has been already described that Drosophila and human 4EHP are able to bind cap analogs using an m 7 GTP-Sepharose approach [11,24,25]. For this reason, we investigated if mouse 4EHP had the same capacity. Pull-down of cytosolic extracts with m 7 GTP-Sepharose suggests that 4EHP can bind the m 7 GpppN (where N is the first template-encoded nucleotide of the transcript) cap structure of mRNAs. Both in vitro-translated and endogenous cytoplasmic 4EHP interact with m 7 GTP-Sepharose, but not with GTP-Sepharose ( Fig. S1B-C). However, Prep1 does not bind m 7 GTP-Sepharose directly, as expected (not shown).
Since Prep1 can bind both some mRNAs and the 4EHP translation inhibitor, we studied the effect of the Prep1-4EHP complex on Hoxb4 mRNA translation in vitro using a rabbit reticulocytes lysate translation system. We cloned the 39UTR of Hoxb4 at the 39 end of a luciferase reporter gene, expressed under the SP6 promoter (Luc-39Hoxb4). As shown in Fig. 5A (nvalues = 5), addition of in vitro-translated Prep1 (previously synthesized under the T7 promoter) inhibited Luc-39Hoxb4 translation by more than 90% (column 1 versus 6). In contrast, the Prep1 mutant (Prep1 YLL) inhibited only around 40% (column 2). This result completely agrees with the capacity of Prep1-YLL to bind 4EHP, which is low but not completely abolished (Fig. 3C). Addition of exogenous 4EHP to the reaction apparently had no major effect on Luc-39Hoxb4 mRNA translation (see columns 1, 4, Fig. 5A). However, we suspected that 4EHP may already be present in excess in the rabbit reticulocyte lysate. In fact, western blot analysis identified 4EHP in rabbit reticulocyte lysates (data not shown). Moreover, RT-PCR identified 4EHP mRNA in the micrococcal nuclease-untreated rabbit reticulocyte lysate (Fig. S1E) . To verify that the inhibitory effect of Prep1 was not due to a difference in the amount of RNA produced in the reaction, we also extracted total RNA from the samples shown in Fig. 5A and analysed the amount of Luc-39Hoxb4 mRNA by RT-PCR (at 25 and 30 cycles). The amount of Luc-39Hoxb4 mRNA produced in each reaction was comparable in all cases in non saturated PCR cycles (Fig. 5B), suggesting that the strong differences observed in Fig. 5A cannot be explained by a differential RNA production between reactions. Moreover, the amount of Prep1 or Prep1-YLL protein added to the reactions was comparable (Fig. S1D). We also verified that the inhibitory effect of Prep1 was specific for Hoxb4 39-UTR. Indeed, translation of a Luc-39Cdx2 mRNA, containing the 39UTR of the mammalian ortholog of Caudal Cdx2 [26], was only marginally affected by Prep1 (Fig. 5C, n = 3). Finally, we also show that the inhibition of Luc-39Hoxb4 mRNA by Prep1 is dose-dependent (compare columns 1, 2 and 3 with column 4 on Fig. 5D, n = 4).
In order to address if 4EHP was required for the inhibition of Luc-39Hoxb4 mRNA translation by Prep1, we used a 4EHP antibody in the reaction. Addition of 2 mg of 4EHP antibody prevented the inhibition of Luc-39Hoxb4 mRNA translation by Prep1 (from over 95% to 20%, column 1 versus 3, Fig. 5E, n = 3). In contrast, the addition of 2 mg of an unrelated antibody (resuspended in the same buffer and at the same concentration) had no effect (column 4, Fig. 5E, n = 3).
We conclude, therefore, that Prep1 and 4EHP inhibit in vitro translation of mRNAs that specifically contain Hoxb4 39UTR.
To identify the region of Hoxb4 39UTR required for the inhibition mediated by Prep1, we subcloned Hoxb4 39UTR in 3 parts (R1, R2, and R3, Fig. S2A) into a luciferase vector and in vitro translated them individually. Translation of none of the three luciferase mRNA constructs was inhibited by Prep1, suggesting that the entire 39UTR or regions across R1-R2 or R2-R3 were necessary for Prep1 inhibition (Fig. S2B, n = 3).
Finally, we analyzed the Prep1-Hoxb4 mRNA interaction by RNA-electrophoretic mobility shift assays using recombinant Prep1 and the Hoxb4 39UTR. As a control, we used an antisense probe (Fig. 5F). Prep1 induced a specific mobility shift (Fig. 5F, lane 2, arrow), which was supershifted by anti-Prep1 antibody (Fig. 5F, lane 3, arrowhead). In contrast, an antibody against other transcription factors such as Pbx proteins (which recognize Pbx1, Pbx2, Pbx3 and Pbx4 members) had no effect (Fig. 5F, lane  4). In contrast, no binding was detected with the antisense probe. This confirms that Prep1 specifically binds Hoxb4 39UTR mRNA. Whether 4EHP is required to increase Prep1 affinity for the 39UTR has not been investigated.

Prep1 hypomorphic mice show drastic defects in ovary and oocyte development
To test for an in vivo role of Prep1 in oocytes and ovary development, we analyzed some of the very few Prep1 i/i females that reach adulthood [3]. Because of the low number (n = 5) of available mice, we cannot claim that homozygous Prep1 i/i females are sterile, but we have never observed pregnancies in mouse Prep1 i/i females. However, Prep1 i/i ovaries had a drastic phenotype: they were smaller and underdeveloped (10/10), presented no oocytes (5/10) or developed cysts (4/10) (Fig. 6A-F).

Hoxb4 expression is increased in Prep1 i/i oocytes
If the Prep1-4EHP interaction negatively regulates Hoxb4 mRNA translation in mouse oocytes, one would expect an increased Hoxb4 production in Prep1 i/i oocytes. Indeed, Hoxb4 was increased in Prep1 i/i oocytes in about 40% of the secondary to antral oocytes (10 Prep1 i/i ovaries analysed, with 16 secondary to antral oocytes in total, Fig. 6G-I). No differences were observed in primary follicles, where Hoxb4 was almost undetectable by immunohistochemistry (data not shown). These data suggest a translation-inhibition function of cytosolic Prep1 in vivo, and indicate that Prep1 could repress Hoxb4 mRNA translation in oocytes. Interestingly, Hoxb4 was localised in the cytosol in antral oocytes.

Injection of Hoxb4 in mouse zygotes delays embryo development in vitro
In order to test whether the oocyte phenotype of Prep1 i/i mice correlates with the increased Hoxb4 mRNA translation, we microinjected fertilized oocytes from super-ovulated females with either CMV-IRES-GFP or CMV-Hoxb4-IRES-GFP vector and examined their development in culture. The overall death rate due to micro-injection was not significantly different between GFP and Hoxb4 injected zygotes (not shown). Those zygotes lysed within the first 24 hours were not included in the calculations. We performed three series of injections for each vector, using 140 fertilized oocytes with the control and 240 with the Hoxb4 vector. Fluorescence microscopy showed that the GFP was expressed at very low levels in several (although not all) injected zygotes, at the various stages (Fig. S2C). Figure 7 shows the (averaged) results of the three experiments in which at 24 hour intervals the percentage of embryos at each developmental stage (1-2 cells and 3-8 cells) was scored and expressed as percent of the total ''live'' embryos. Overall, the development was slowed down at all stages in the Hoxb4-microinjected zygotes. The results were statistically signif- icant at the very early (1-2 and 3-8 cells) stages. Overexpression of Hoxb4 showed the same trend also at the morula/blastocyst stage, where it did not reach statistical significance (not shown). We conclude, therefore, that the overexpression of Hoxb4 in mouse zygotes slows down embryo development.
In Drosophila embryos, the homeodomain protein Bcd interacts with 4EHP to regulate the translation of Cad mRNA through a YxxxxxxL motif [11]. Although Bcd homologs have been identified only in close relatives of Drosophila, we show in this paper that the ability to act in both transcriptional and translational levels is conserved in some mammalian homeodomain proteins, and that at least the TALE class protein Prep1 specifically represses translation of Hoxb4 mRNA.
Bcd and Prep1 mechanisms are different. First, Bcd is related to Hox [21], not to TALE proteins. Second, cytosolic Bcd regulates embryonic patterning while cytosolic Prep1 in mammals likely regulates Hoxb4 expression in female germ cells. Moreover, Bcd represses Cad mRNA during embryo development, but we were . Lane 4 shows that the effect of the antibody is specific since an anti-Pbx antibody has no effect. Same experiment using antisense probe is shown in lanes 5-8. doi:10.1371/journal.pone.0005213.g005 not able to find any apparent effect of Prep1 on Cdx2 39UTR mRNA (the mammalian ortholog of Cad). Another difference lies at the level of the 4EHP-interacting sequence which is 66 YNYIR-PYL 73 in Bcd and 59 YRHPLFPLL 67 in Prep1 (Fig. 1A), i.e. with an additional important leucine in the case of Prep1. The 59 YRHPLFPLL 67 sequence present in Prep1 is highly conserved in proteins of the same family, suggesting that the translation inhibition function might be shared with other members of the family (Fig. 1B). If this prediction is verified, it is possible that members of the same family are able to bind different mRNAs. Interestingly, the Prep1 4EHP-binding sequence overlaps with the Pbx1-binding sequence [29], suggesting that the binding of Prep1 to 4EHP or to Pbx1 is mutually exclusive. This agrees with the ability of Prep1 to bind the 39UTR of Hoxb4 in the absence of any Pbx proteins. In fact, this could explain why Prep1 is located in the cytosol of mouse oocytes. The formation of a Prep1-Pbx complex is necessary to transport Prep1 to the nucleus [30].
Translation inhibition by Prep1-4EHP is most likely due to the inability of 4EHP to bind eIF4G [31]. The interaction with 4EHP-Prep1 would sequester the target mRNA preventing its association with the translation initiation machinery. Unlike Hoxa9 [12], we were not able to find an interaction between Prep1 and the translation initiation factor eIF4E (not shown).
In this paper, we have focused our study on Hoxb4. However, the target of Prep1 may be not only Hoxb4 mRNA, since Hoxb4 mRNA was not the only one co-immunoprecipitated in our experiments. Moreover, although we have not demonstrated the formation of a Prep1-mRNA-4EHP complex, Prep1 might bind simultaneously to 4EHP and to Hoxb4 39UTR mRNA. In turn, mRNA would be bound by 4EHP at the cap site. The Prep1binding sequence in Hoxb4 mRNA is located in the 39UTR, since Prep1 inhibition was specific for this 39UTR, but Prep1 was unable to repress translation when only part of the 39UTR was present (Fig. S2B). However, we cannot exclude that the Hoxb4 39UTR binding region is located in a sequence bridging R1 to R2 or R2 to R3.
Translational control is an important mechanism regulating the earliest stages of embryogenesis [16,32,33]. In mammals, maternal mRNA translation is tightly controlled delaying translation of specific maternal mRNAs during the mammalian oocyte-embryo transition [34][35][36]. The novel oocyte/ovary phenotype of the Prep1 i/i mice correlates with the increased production of Hoxb4. The increased synthesis of Hoxb4 protein in Prep1 i/i oocytes agrees with the hypothesis that the absence of Prep1 relieves a block of Hoxb4 mRNA translation leading to an oocyte growth failure and cyst formation. However, Hoxb4 null mutant females are viable and fertile [37], possibly due to compensation by another Hox gene. On the other hand, overexpression of Hoxb4 in mouse developing oocytes leads to developmental delay at the transition between one to eight cells, and the same trend is also observed at morula/blastocyst stages. In fact, Prep1 is the first homeodomain protein whose translational repression activity may be functionally relevant in vivo in mammals.
In summary, we conclude that Prep1 is involved in translational regulation of Hoxb4 mRNA in mouse oocytes, in cooperation with 4EHP. This function may be essential for mammalian female germ cell development and also involved during the first stages of embryo development.
Figures in this paper were prepared using the Adobe Phostoshop CS4 version 11.0.

Sucrose gradient
We followed the protocol described previously [20].

RNA immunoprecipitation
We modified a protocol described previously [38]. Ovaries were dissected under the microscope, and oocytes with surrounding granulosa cells were isolated. Cells were washed twice with 5ml PBS, and resuspended in 2ml PBS. Formaldehyde was added to a final concentration of 1% and incubate at RT for 10 min with slow mixing. Reaction was quenched by the addition of glycine (pH 7.0) to a final concentration of 0.25M, followed by incubation at RT for 5min. Cells were harvested by centrifugation using a clinical centrifuge at 3000rpm for 5min. Cells were washed twice with icecold PBS. Fixed cells were resuspended in 2ml of IP buffer (20mM HEPES-KOH, pH 7.6, 200mM KCl, 0.5mM EDTA, 10% glycerol, 0.5% Triton X-100 and Protease Inhibitor Cocktail Complete; Roche). Cells were lysed routinely by three rounds of sonication, 30s each. Between each cycle, the samples were kept in an ice-water bath for 2min. Insoluble material wass removed by microcentrifugation at 14.000rpm for 10min at 4uC. Immunoprecipitation was performed by adding the relevant antibody to the supernatant extracts and incubating at 4uC overnight. Reactions were incubated with 20 ml protein A slurry beads (equilibrated in IP buffer containing 1mg/ml BSA, competitor tRNA at 100 mg/ml) and the mix was incubated for 2h at 4uC. Beads were collected using a minicentrifuge at 6.000rpm for 45s and the supernatant was saved for RNA extraction. Beads were washed five times with 1ml of IP buffer by 15min rotation at 4uC. Beads were collected and resuspended in 100 ml of 50mM Tris-Cl pH 7.0; 5mM EDTA; 10mM DTT and 1% SDS. Beads were incubated at 70uC for 3h to reverse crosslinks. RNA extraction was performed with Quiagen RNasin kit. After RNA extraction, a DNA digestion was performed, and RNA was cleared by the same kit.

Cap-Affinity Assay
For Cap-affinity assay we followed the protocol described previously [11].

Deconvolution analysis
Confocal microscopy stacks were deconvolved with 20 iterations using theoretical point spread function (PSF) and maximum likelihood estimation (MLE) algorithms of Huygens software (SVI, Hilversum, the Netherlands). 3D colocalization analyses of 4EHP and Prep1 were performed using the automatic threshold algorithm by Costes and Locket [39] implemented in Bitplane Imaris suite (Bitplane AG, Zurich, Switzerland). 3D colocalization is shown as the white channel.

Co-immunoprecipitations and GST Pull-Down
For co-immunoprecipitation, cytosolic ovarian cell extract was brought up to 0,5 ml with the IP buffer (20 mM HEPES-KOH, pH 7.6, 200 mM KCl, 0.5 mM EDTA, 10% glycerol, 0.5% Triton X-100 and Protease Inhibitor Cocktail Complete; Roche) and precleared for 1h at 4uC with 25 ml of Protein A Sepharose. The supernatant was immunoprecipitated for 1h at 4uC with 25 ml of anti-Prep (Santa Cruz Biotech). The resin was washed three times with lysis buffer. Immunoprecipitates were eluted in 26 sample buffer.

Luciferase Assay
Prep1, mutant-Prep1 and 4EHP proteins were generated using the TNT T7 Coupled Reticulosyte Lysate Transcription/Translation System (Promega), under the T7 promoter following the manufacturer's instructions. As a control-reaction, a T7-reaction with an empty pCDNA3.1 vector was used. T7-reactions were stopped on ice after 1h of incubation at 30uC. 1 ml of the corresponding T7-reactions (containing Prep1, mutant-Prep1, 4EHP protein or control-reaction) was added to the SP6-reactions composed by 20 ml of master-mix, methionine, Luc-Hoxb4 39UTR plasmid (or SP6 Luciferase vector with the R regions of Hoxb4 39UTR), and SP6 enzyme, in a total volume of 25 ml following the manufacturer's instructions. SP6-reactions were incubated at 30uC for 1h and 30min. After the 1 st hour of incubation, the reaction was shacked vigorously for 5 seconds. Reactions were stopped on ice. Then, 2.5 ml were used to analyze luciferase production. SP6-Reactions were peformed everytime in triplicate, and each condition was performed at least three independent times (see nvalues in the text). mRNA extraction and RT-PCR mRNA extraction from luciferase samples were extracted, and retrotranscribed as previously described [18].
Degenerate primers for amplification of HoxA and HoxB cDNA were used as described previously [18,22,23]. Amplified products were cloned with TA-Cloning kit (Invitrogen), sequenced, and screened for homology to known sequences using the NCBI-BLAST software.

Cloning, expression and purification of human Prep1
Human Prep1 protein was used just only for REMSA experiments. Expression in the Escherichia coli strain BL21(DE3) was induced with 0.3mM IPTG. Expression was continued for ,16h at 20uC. Cells were harvested by centrifugation and resuspended in 30ml lysis buffer (20mM Tris, pH 7.4, 0.3mM NaCl, 1mM DTT supplemented with Protease inhibitor cocktail from Calbiochem) per litre of culture. After sonication, the lysate was cleared by centrifugation. The GST-fusion protein was purified using Glutathione-agarose beads (GE Healthcare) equilibrated in 20mM Tris, pH 7.4, 0.3mM NaCl, 1mM DTT and the protein was subsequently cleaved from GST with 10u of PreScission protease (GE Healthcare) per milligram of substrate for 16h at 4uC.

REMSA
The probe for REMSA was prepared and labeled by the in vitro transcription of the cloned DNA fragment of Hoxb4 39UTR using [alpha-32P]rUTP and RiboprobeH Combination System (Promega, Madison, WI). After treatment with DNase, it was described by RNeasy kit (Qiagen, Germany). REMSA was carried out as previously published with minor modifications [40]. Briefly, the reaction was performed in the CEB buffer (10mM HEPES, pH 7.5, 3mM MgCl2, 14mM KCl, 5% glycerol, 1mM DTT) using 0.3 mg of human Prep1 recombinant protein. After 20min incubation on ice with or without 5 mg of anti-Prep1 [30] or anti-Pbx1 (sc-889X, Santa-Cruz Biotechnology, Santa Cruz, CA), the probe (50,000 cpm) was added and the mixture incubated at room temperature, followed by 10min incubation with RNase T1 (0.5u) and 10min incubation with heparin (6mg/ml). The RNA-protein complexes were resolved in 5% polyacrlylamide mini-gels (acrylamide:bis acrylamide of 36:1) and vacuum-dried. RNAprotein interactions were visualized by use of PhosphoImager 445 SI (Molecular Dynamics Sunnyvale, CA).

Microinjection of Hoxb4 into mouse oocytes
Fully grown, germinal vesicle-intact (GV) fertilized mouse oocytes were obtained from 4-week-old female mice and freed of attached cumulus cells as previously described [41,42]. The collection medium was bicarbonate-free minimal essential medium (Earle's salts) supplemented with polyvinylpyrrolidone (3mg/ml) and 25mM Hepes, pH 7.3. The denuded oocytes were matured in CZB medium [43] in an atmosphere of 5% CO2 in air at 37uC. Images were captured by Zeiss Discovery V12 stereo microscope, and fluorescence with Nikon SMZ 1500 Microscope.

Statistical analysis
All the experiments were performed at least three times. For statistical analysis of data, Student's t test was used. Values are expressed as mean6standard error of the mean. Data were considered statistically different at a p value of ,0.04. Figure S1 (A) This control shows the specificity of the anti-4EHP antibody that does not recognize the close homolog eIF4E. (B) Cytosolic extracts from wild type mouse ovaries were pulled down using m7-GTP or GTP (control) beads and eluted as described in the Material and Methods section. The presence of 4EHP in the eluate was monitored by immunoblotting. (C) Same experiment as in (B), but performed with in vitro translated 35S-4EHP. (D) This control shows that the amounts of Prep1 and mutant Prep1 added to the reactions (Fig. 5A) were equivalent, as shown by the radiographic evaluation of in vitro translated 35S-Met-labeled proteins. (E) 4EHP messenger RNA is detected in the crude untreated rabbit retyculosyte lysate, suggesting that there is at least endogenous 4EHP mRNA in the reaction. Found at: doi:10.1371/journal.pone.0005213.s001 (2.76 MB TIF) Figure S2 (A) Hoxb4 mRNA sequence, from the stop codon TAG (black box) to the poly-A signal. The Hoxb4 39UTR was divided in 3 regions (R1, R2, and R3) and cloned using specific primers (sequences underlined) in a luciferase vector, in order to study the effect of Prep1 protein. (B) Prep1 does not inhibit the translation of luciferase-Hoxb4 R1, R2 or R3 39UTR mRNA, suggesting that the whole 39UTR is required for the inhibition. (C) Expression of fluorescent GFP in mouse embryos micro-injected with a CMV-Hoxb4-IRES-GFP construct (mouse embryos, left; GFP merge, right). This representative picture was taken at an early developmental stage, after 1.5 days in culture. Found at: doi:10.1371/journal.pone.0005213.s002 (5.48 MB TIF)