The absence of Prep1 causes p53-dependent apoptosis of mouse pluripotent epiblast cells

Several crucial events take place during early development. In the E3.5 blastocyst, the inner cell mass (ICM) contains progenitor cells, including epiblast (Epi) precursors (Chazaud et al., 2006), which generate embryonic stem (ES) cells (Evans and Kaufman, 1981; Martin, 1981). The Epi is established during implantation around E4.5, and from E5.5 to E6.5 forms an epithelium [a process known as cavitation (Coucouvanis and Martin, 1995)], maintains its pluripotent state (Niwa, 2007) and proliferates actively [undergoing a major expansion with a cell cycle as short as 2 hours (O'Farrell et al., 2004; Snow, 1977)]. At this time, the Epi is very sensitive to DNA damage (Heyer et al., 2000) and is not protected by specific G1 and G2 check points (O'Farrell et al., 2004). DNA damage at this stage leads to p53-dependent apoptosis (Heyer et al., 2000). Formation of the primitive streak (PS) around E6.5 marks the beginning of gastrulation and requires interactions between the Epi, the extra-embryonic AVE (anterior visceral endoderm) and the ExE (extra-embryonic ectoderm). Gastrulation, during which the Epi gives rise to the three embryonic layers (ectoderm, mesoderm and definitive endoderm), is followed by the neural tube formation from the neural plate (Stern, 2006). After neurulation, the body plan of the embryo is established, with distinct antero-posterior, dorso-ventral and left-right axes (Tam and Behringer, 1997).

Prep1 (Pknox1 – Mouse Genome Informatics) homeodomain transcription factor belongs to the TALE (three amino acids loop extension) superclass of proteins that include Meis1-Meis3, Prep2 and Pbx1-Pbx4. Deletion of Pbx and Meis genes in mouse shows that these genes are essential for organogenesis and differentiation. Homozygous Meis1^–/– embryos die around E14.5 with a severe hematopoietic phenotype (Azcoitia et al., 2005; Hisa et al., 2004) that is similar to that of the hypomorphic Prep1 mutants (Prep1^i/i) (Di Rosa et al., 2007; Ferretti et al., 2006). Prep1^i/i embryos express ∼2% of Prep1 mRNA and 2-10% of the protein, die mostly at E17.5 with liver hypoplasia, anemia, angiogenesis and eye defects (Ferretti et al., 2006). The hematopoietic phenotype is due to malfunction of long-term repopulating hematopoietic stem cells (Di Rosa et al., 2007). No loss of function mutation for Meis2, Meis3 or Prep2 has been described. A compound Pbx1-Pbx2 knockout mouse is lethal around E12.5 and displays pallor, edema, general organ hypoplasia and homeotic features (Capellini et al., 2006; Selleri et al., 2001). While the single Pbx2 knockout mice are viable and fertile (Selleri et al., 2004), Pbx3 null mice are born, but die within a few hours owing to central respiratory failure (Rhee et al., 2004). Importantly, Pbx2 partly compensates for Pbx1 (Capellini et al., 2006; Selleri et al., 2004).

We have analyzed Prep1^–/– embryos in which the expression of the protein was eliminated by targeting the DNA-binding homeodomain. This mutation leads to early post-implantation lethality. Apoptosis reduces the number of pluripotent Epi cells, which, hence, fail to form the AVE, PS and differentiated lineages. This phenotype uncovers a genetic interaction between Prep1 and p53 (Trp53 – Mouse Genome Informatics) as p53 ablation partially rescues the Prep1^–/– phenotype. Therefore, unlike all other TALE proteins, Prep1 is responsible for protecting the embryo very early in development, a unique function within the Meis-Prep families of transcription factors.

Exons 7 and 8 of the Prep1 gene were replaced with pSAβ-geo vector cassette (see Fig. S1A-D in the supplementary material) (Hisa et al., 2004) and homologous recombination events tested by Southern blotting of EcoRI digested DNA with 5′ and 3′ probes (see Fig. S1E,F in the supplementary material). Of several independent isolated ES cell clones carrying the mutation, one was introduced into the germ line using standard techniques. Pure C57/BL6 double KO Prep1-p53 or Prep1-Atm mutant mice were obtained by standard crossing p53^–/– (Jackson Labs, Bar Harbor, ME, USA) or Atm^–/– (Borghesani et al., 2000) mice with Prep1^+/– females. The Prep1-p53-ink4A^p16 mutant line was obtained by crossing an ink4A^p16-null (Krimpenfort et al., 2001) with a Prep1^+/–p53^+/– double heterozygous mouse.

Genotyping was carried out by Southern blot using standard techniques or by PCR (see Fig. S1 in the supplementary material). The standard protocol was one cycle of 5 minutes at 94°C, 35 cycles of 30 seconds at 95°C, 30 seconds at 55°C and 30 seconds at 72°C. A final elongation step of 5 minutes at 72°C was performed. For E8.5 and E7.5 embryos, DNA polymerase concentration was doubled, and the PCR reaction extended to 40 cycles. The sequences of the primers for genotyping (arrows in Fig. S1C,D in the supplementary material) were: P12, 5′-GAGAGCTCAAGGACAGCCAGGCTA-3′; M7, 5′-CCAGGAGATAATGCCTGCGTGACC-3′; and T3, 5′-ACCGCGAAGAGTTTGTCCTCAACC-3′.

RT-PCR was performed with the Superscript II (Invitrogen, Carlsbad, CA, USA) (see Fig. S1H in the supplementary material). Primers used were: F1, 5′-GACACCGTGTGCTTCTCGCTCAAG-3′; and R1, 5′-AGACAAGCAATGTACCGACTACAG-3′. For the KO mRNA, the primer R2 corresponds to M7 and R3 to T3.

Antibodies used were Pbx1 (Abcam, Cambridge, UK); Pbx2 and β-Actin (Santa Cruz Biotec, CA); β-Gal (Promega, Madison, WI, USA); and Prep1 (Ferretti et al., 2006).

Electrophoretic mobility shift assay was carried out as described previously (Berthelsen et al., 1996) with 20,000 cpm ³²P-labeled oligonucleotides: B2PP2, 5′-GGAGCTGTCAGGGGGCTAAGATTGATCGCCTCA-3′ (Ferretti et al., 2000); and SP1, 5′-AAGACAGGGGAGGGAGCCGGGCGGGAGAGGGAGGGGCGGCGCCGGGGCGGGCCCT-3′ (Ibanez-Tallon et al., 2002).

Embryos were dissected from the maternal deciduas by standard procedures and in situ hybridized (including double in situ) as described previously (Liguori et al., 2003). TUNEL was performed using the ApopTag Kit (Millipore, Billerica, MA, USA).

RT-PCR was performed on total RNA (pools of 100 embryos) (TRIzol reagent; Invitrogen, Carlsbad, CA, USA) as described (Fiorenza et al., 2004). Primers used were Prep1, 5′-ATGATGGCGACACAGACGCTAAGTATA-3′ (sense) and 5′-GGGGTCTGAGACTCGATGGGAGGAGGACTC-3′ (antisense); β-actin, 5′-GGTTCCGATGCCCTGAGGCTC-3′ (sense) and 5′-ACTTGCGGTGCATGGAGG-3′ (antisense).

Embryonic stem (ES) cells were derived from E2.5-3.5 blastocysts from Prep1^+/– crosses using standard procedures. Mouse ES cells E14Tg2a were cultured without feeders in Glasgow-modified Eagle's MEM with 15% ES-screened FBS (Hyclone, Logan, UT, USA), under standard conditions with 1000 U/ml leukemia inhibitory factor (LIF; Chemicon, CA, USA). Embryoid bodies (EBs) were obtained using the standard protocol of hanging drops method. One thousand cells per drop were plated in LIF-free ES medium.

Passage 14-16 ES cells were grown on coverslips, fixed in paraformaldehyde, permeabilized and blocked in PBS/10% calf serum/1% BSA/0.1% Triton X-100, and incubated with anti-Oct3/4 antibody (1/500, Santa Cruz). For whole-mount immunofluorescence, the antibodies were: rabbit anti-cleaved caspase 3 (Cell Signaling, Boston, MA, USA; 1:50); rabbit anti-phospho-histone H3 (Upstate, Billerica, MA, USA; 1:1000); goat anti-Oct4 (Abcam, 1:500); donkey anti-rabbit CY3-conjugated antibody (Jackson Labs, Bar Harbor, ME, USA; 1:400), donkey anti-goat A488-conjugated antibody (Invitrogen, Carlsbad, CA, USA; 1:200) and DAPI (1 μg/ml). Confocal microscopes are TCS SP2 and TCS SP2 AOBS (Leica Microsystem, Germany). For each embryo, a series of optical sections (z stacks) was collected. The mitotic index and cleaved caspase 3 staining were quantified as the ratio of phospho-histone H3 or cleaved caspase 3-positive areas compared with the DAPI stained areas, on five sections per embryo, using ImageJ software. The same procedures were used for immunostaining EBs with cleaved caspase 3 antibody and signal quantification.

The Prep1 homeobox was deleted by replacing exons 7 and 8 with a pSAβ-geo recombination cassette (Friedrich and Soriano, 1991) (see Fig. S1A-D in the supplementary material). The vector contains a splice acceptor site with stop codons in all reading frames, and β-geo, a fusion lacZ-Neo^R gene, with an internal ATG and a bovine growth hormone polyadenylation sequence (see Fig. S1D in the supplementary material). One single cell line was injected to create chimeric animals and heterozygous germ-line males were backcrossed to C57/BL6-NCr (B6) more than 10 times and are therefore essentially C57/BL6 congenic. The insertion was tested by Southern blotting at the 5′ (see Fig. S1E in the supplementary material) and 3′ (see Fig. S1F in the supplementary material) ends (the expected fragments are shown as double arrows in Fig. S1A-B in the supplementary material), and the results confirmed by PCR (data not shown). We have subsequently used a PCR assay to genotype the embryos. Fig. S1G in the supplementary material shows a typical assay on Prep1^+/+, Prep1^+/– and Prep1^–/– embryos. Primers are indicated by small arrows in Fig. S1C-D in the supplementary material.

Heterozygous intercrosses did not yield homozygous mutant mice (Table 1). Timed pregnancy analysis placed the death of Prep1^–/– embryos around E7.5. No Prep1^–/– homozygous embryo was recovered after E10.5, whereas at E9.5 and E8.5 homozygous structures mostly looked like empty yolk sacs (Table 1). At E7.5, homozygous embryos appeared mostly small and developmentally delayed (Table 1 and see below). Prep1 is therefore essential for early post-implantation mouse development.

The approach used to knock out Prep1 might in principle lead to transcription of a di-cistronic mRNA producing two proteins, the N-terminal part of Prep1 with its own ATG and the C-terminal β-geo from an internal ATG (see Fig. S1H in the supplementary material). The fusion di-cistronic mRNA was observed by amplification from Prep1^–/– E8.5 cDNA (see Fig. S1H-I in the supplementary material). Several combinations of primers were used to test for anomalous Prep1 mRNAs, but no additional mRNA was found (data not shown). We also sequenced the fusion region of Prep1-β-Geo cDNA (not shown) and found the expected STOP codons in the three frames, which should prevent the production of a fusion Prep1-β-Geo protein. The fusion mRNA was also detected in E11.5 heterozygous mouse embryonic fibroblasts (MEFs), liver, kidney, lung and spleen (not shown).

At the protein level, we never found any truncated form of Prep1 (see Fig. S2A in the supplementary material). We also did not find any difference in DNA binding in EMSA performed with Prep1^+/+ and Prep1^+/– extracts (see Fig. S2D-E in the supplementary material). lacZ was detected by immunoblotting only in heterozygous testis extract (in the cytoplasm, see Fig. S2B in the supplementary material) and by β-gal staining in the E13.5 retina of Prep1^+/– embryos (see Fig. S2C in the supplementary material) where Prep1 has been shown to be strongly expressed in wild-type embryos (Ferretti et al., 2006). Finally, western blotting of wild-type, heterozygous and homozygous nuclear extracts of Prep1 ES cell lines (see below) (two different clones per genotype) showed no Prep1 in homozygous (see Fig. S1J in the supplementary material) and reduced Prep1 levels in heterozygous nuclear extracts (see Fig. S1J in the supplementary material). We obtained the same result with two different monoclonal antibodies (not shown). No C-terminally truncated Prep1 was observed using antibodies specifically recognizing either the N terminus or the C terminus of Prep1 (see Fig. S3A-C in the supplementary material). The western blot analysis of the ES cell nuclear extracts using the N terminus-specific antibody allowed us to observe only background staining at the low molecular weight range of the gel in Prep1^–/– ES cells (see Fig. S3D in the supplementary material); this was the case even after cultivating the cells with MG132 proteasome inhibitor (see Fig. S3E in the supplementary material). We conclude that Prep1 is totally absent and hence that Prep1^–/– are null embryos.

Table S1 in the supplementary material lists the specific peri-gastrulation markers used and the number of embryos analyzed in each case. PS formation is evidence of the onset of gastrulation around E6.5 (Tam and Behringer, 1997). The T gene (Brachyury) marks the PS and axial mesoderm (Wilkinson et al., 1990). Prep1^+/+ and Prep1^+/– embryos were T positive but Prep1^–/– embryos were not (Fig. 1A,B), suggesting that Prep1-deficient embryos lack the PS. To exclude the possibility of a simple delay in T expression, we tested its expression at E7.5 and E8.5; at neither stage did Prep1^–/– embryos express T (not shown). As T is expressed in the distal ExE before PS formation (Perea-Gomez et al., 2004; Rivera-Perez and Magnuson, 2005) and as we did not observe T in Prep1^–/– embryos, we repeated the staining for T and increased the color reaction. We also tested for Fgf8 as a second PS marker (Crossley and Martin, 1995). Again, Prep1^–/– embryos were negative for T (see Fig. S4A-D in the supplementary material) and Fgf8 (Fig. 1C-D). We tested Lefty2 as a nascent mesoderm marker (Meno et al., 1999) at E6.75-E7.5 and, in accordance with the absence of PS, Prep1^–/– embryos were also negative for Lefty2 (Fig. 1E,F).

PS induction depends on the crosstalk between the Epi and two extra-embryonic tissues: AVE and ExE (Ang and Constam, 2004; Tam et al., 2006). In this process, important signaling molecules include Nodal, Wnt3 and Bmp4. While Nodal and Wnt3 are expressed in the posterior embryonic/extra-embryonic junction, next to where the PS arises (Brennan et al., 2001; Liu et al., 1999), Bmp4 is expressed in the distal ExE (Coucouvanis and Martin, 1999; Lawson et al., 1999). In Prep1^–/– embryos at E6.5-E6.75, Nodal was indeed expressed in the Epi (Fig. 1G,H), whereas Wnt3 was strongly decreased or absent (Fig. 1I,J); however, Bmp4 was normally expressed in the ExE as in wild-type embryos (Fig. 1K-L). The areas of expression of Nodal and Bmp4 in the Epi and the ExE, respectively, were reduced in size (Fig. 1G,H,K,L). Finally, we examined the formation of the AVE, a signaling center implicated in PS formation (Bertocchini and Stern, 2002; Perea-Gomez et al., 2002) and neural induction (Albazerchi and Stern, 2007; Kimura et al., 2000; Perea-Gomez et al., 2001) because it expresses antagonists of Nodal (Meno et al., 1999; Takaoka et al., 2006), Bmp (Belo et al., 2000) and Wnt (Kimura-Yoshida et al., 2005) genes. The AVE marker Cerberus-like (Cer1) (Belo et al., 1997; Biben et al., 1998; Shawlot et al., 1998) was not detectable in Prep1^–/– embryos (Fig. 1M,N), suggesting the absence of AVE. Thus, although the Epi and the ExE of Prep1 mutant embryos do express important signaling molecules such as Nodal and Bmp4, Prep1^–/– embryos at the onset of gastrulation are reduced in size, do not express Wnt3 and form no AVE, PS or mesoderm.

During gastrulation, between E7.5 and E7.75, Chordin (Chrd) is expressed in the node and axial mesoderm (Bachiller et al., 2000), Otx2 in the anterior neuroectoderm (Ang et al., 1994; Simeone et al., 1993), Gbx2 in the posterior neural tube (Wassarman et al., 1997), Cer1 in the definitive endoderm (Belo et al., 1997; Biben et al., 1998; Shawlot et al., 1998) and Sox1 in the neural plate (Wood and Episkopou, 1999). Unlike wild-type and heterozygous embryos, Prep1^–/– embryos did not express Chrd at E7.5 (Fig. 1O,P), Otx2 at E7.75 (Fig. 1Q,R), Gbx2 at E7.75 (Fig. 1S,T), Cer1 at E7.5 (Fig. 1U,V) or Sox1 at E7.75 (see Fig. S4E-F in the supplementary material). Therefore, in agreement with the absence of PS, Prep1^–/– embryos lack mesoderm, endoderm and ectoderm patterning, i.e. establish no embryonic lineage. Next, we looked at Rhox5 (Lin et al., 1994; Maclean et al., 2005), Bmp4 (Lawson et al., 1999; Winnier et al., 1995) and Flk1 (Yamaguchi et al., 1993) as extra-embryonic markers. Prep1^–/– embryos expressed Rhox5 at E7.5 and E8.5 (Fig. 1W,X and data not shown), most probably marking the extra-embryonic part of the visceral endoderm. Prep1^–/– embryos expressed Bmp4 (Fig. 1Y,Z) that most likely represents the residual expression in the distal ExE. In agreement, Prep1^–/– embryos also expressed Cdx2 (not shown), a second ExE marker (Beck et al., 1995). Finally, Flk1 was almost undetectable (see Fig. S4G,H in the supplementary material) arguing that there is no formation of endothelial cells in the Prep1 KO. Hence, at E7.5-E7.75 Prep1^–/– embryos have no embryonic derivatives and the extra-embryonic compartments are restricted to residual cells of the ExE and derivatives of the visceral endoderm.

Early post-implantation Epi cells are pluripotent. Descendants of single early PS Epi cells are able to contribute to more than one tissue type (Lawson et al., 1991; Tam and Behringer, 1997). Moreover, it is possible to extract Epi-stem cells from egg cylinder stage embryos (Brons et al., 2007; Tesar et al., 2007). Hence, the absence of PS in Prep1^–/– embryos might be due to premature differentiation of pluripotent Epi cells. To assess this hypothesis, we examined the expression of Oct4 (Pou5f1), which is widely accepted as a marker for pluripotency in post-implantation embryos (Ding et al., 1998; Liguori et al., 2003). Although reduced in number, Oct4-positive cells were observed in Prep1^–/– embryos at E6.75 (Fig. 2A,B), E7.5 (Fig. 2C,D) and E8.5 (Fig. 2E,F). The expression of Oct4 in Prep1^–/– embryos argues against the above hypothesis and shows that in Prep1^–/– mutants the number of epiblast cells is reduced.

Altogether, our marker analysis shows that by E7.5-E7.75, Prep1^–/– embryos have a reduced number of Epi cells and a residual ExE all surrounded by derivatives of the visceral endoderm. We have confirmed this point with a double in situ against Oct4 and Bmp4 (Fig. 2G,H).

The above results suggest that the phenotype may be due to a basic, general cellular failure, i.e. excessive apoptosis or a block of proliferation. Epi cells at the egg cylinder stage are highly sensitive to DNA damage, which induces apoptosis without arresting the cell cycle (Heyer et al., 2000). Therefore, we tested whether apoptosis might account for the decreased number of Epi cells in Prep1^–/– embryos. Whole-mount confocal immunofluorescence of cleaved caspase 3 showed that, in the absence of Prep1, pluripotent Epi cells underwent apoptosis at E6.5 (Fig. 2I,J) and E7.5 (data not shown). A quantification of cleaved caspase 3 staining is shown in Fig. 2K. Similar results were obtained in a TUNEL staining at E6.5 and E7.5 (see Fig. S4K-O in the supplementary material). Cleaved caspase 3 and TUNEL staining of Prep1^–/– embryos concentrate in the Epi region (Fig. 2J, see Fig. S4J-L in the supplementary material), arguing that the apoptotic cells are indeed the Oct4-positive Epi cells.

The high sensitivity of Epi cells to irradiation during the egg cylinder stage is p53 dependent (Heyer et al., 2000). To test whether the apoptosis we observed in Prep1^–/– embryos depends on p53, we crossed Prep1^+/– and p53-null mice. Mice with mutations in Prep1 or p53 were both created in a full C57BL6 background. Embryos from double-heterozygous Prep1^+/–p53^+/– intercrosses were analyzed with specific markers.

Thirty-five E7.5 embryos were extracted and hybridized with an Oct4 probe (Table 2). Among them, the two Prep1^–/–p53^–/– double homozygous embryos showed a strong increase in the size of the Oct4 expression domain (arrow in Fig. 2L-N). Prep1^–/– embryos heterozygous for p53 (asterisk in Fig. 2M) were indistinguishable from Prep1^–/– embryos (see Fig. 2D). Thus, the absence of p53 rescues the number of Epi cells. Indeed, the increase in Oct4 staining was accompanied by a decrease in TUNEL staining in Prep1^–/–p53^–/– double KO embryos (see Fig. S4K-M in the supplementary material). Cer1 was re-expressed in Prep1^–/–p53^–/– embryos (arrow in Fig. 2O-Q), but in a pattern more similar to that in the AVE than to the pattern in the definitive endoderm. A Prep1^–/–p53^+/+ embryo was negative for Cer1 (asterisk in Fig. 2P), as expected (see Fig. 1V). Moreover, T and Chrd were expressed at E7.5 in double KO embryos, arguing that, unlike Prep1^–/– embryos (Fig. 1A-B,O-P), Prep1^–/–p53^–/– embryos formed the PS (Fig. 2R-T), which reached the most distal part of the embryo (Fig. 2U-W). Finally, Prep1^–/–p53^–/– embryos were Bmp4-positive and possessed a better organized ExE than did Prep1^–/– embryos (data not shown).

Interestingly, in spite of the recovery in the number of Oct4-positive cells at E7.5 in the two embryos, and the re-expression of Cer1, T and Chr, this recovery was still strongly delayed at E8.5 in Prep1^–/–p53^–/– embryos. The expression of T in E8.5 Prep1^–/–p53^–/– embryos is still evident (Fig. 3A-C) but the staining does not reach the anterior part of the embryo (Fig. 3C). We did not precisely determine the latest stage at which double KO mutants could be recovered. However, E9.5 and E10.5, Prep1^–/–p53^–/– embryos are almost reabsorbed (Fig. 3D-F and data not shown).

We conclude that Prep1 is required to prevent p53-dependent apoptosis in the embryonic Epi at peri-implantation stages and that the absence of p53 allows the increase in number of epiblast Oct4-positive cells inducing AVE and PS formation, thus only a partial recovery of the phenotype. This indicates that Prep1 is also required during gastrulation.

During peri-gastrulation stages, the Epi is hyper-proliferative (O'Farrell et al., 2004). First, we tested the proliferation of E6.75 Prep1^–/– embryos with Ki67 immunostaining, a marker of non-quiescent cells, and found no difference between Prep1^+/+, Prep1^+/– and Prep1^–/– embryos (data not shown). Moreover, at E6.75, whole-mount immunofluorescence for Oct4 and for the mitotic cell marker phospho-histone 3 (P-H3) (Gurley et al., 1978), analyzed by confocal microscopy (Fig. 3G-H), showed that embryos were reduced in size, but had a large P-H3-positive signal (Fig. 3H). The mitotic index quantification is shown in Fig. 3I. To complete our proliferation analysis of the Prep1 KO, we produced a Prep1-p53-p16 triple KO. Ink4A^p16 is an inhibitor of pRb and its absence confers a proliferative advantage (Sharpless, 2005). We crossed ink4A^p16-null (Krimpenfort et al., 2001) with Prep1^+/–p53^+/– double heterozygous mice. The Ink4A^p16 mice were in a mixed C57BL6-SV129 genetic background. Therefore, in order to exclude that a change in the genetic background would affect the phenotype we verified that the phenotype was conserved in the new background (data not shown). Although we did not yet find a triple homozygous Prep1^–/–p53^–/–p16^–/– embryo, the Prep1^–/– phenotype was not rescued in p16 Ko (see expression of Oct4 in the p16^–/– double KO embryos) (see Fig. S4N,O in the supplementary material). Thus, despite the reduced size, Prep1^–/– embryos retained their proliferation capacity.

From E8.5 to birth, Prep1 is expressed ubiquitously and weakly (Ferretti et al., 1999). We examined the expression pattern of Prep1 in wild-type embryos during gastrulation and ES cell differentiation. Prep1 whole-mount in situ hybridization shows a weak ubiquitous signal from E6.5 to E8.5, in some cases difficult to distinguish from the sense strand control signal, especially at E7.5 (see Fig. S4P-U in the supplementary material). In whole-mount immunofluorescence, however, Prep1 was expressed in the whole embryo at E6.5 (Fig. 4A,B) and at E7.5 (Fig. 4D,E), and was localized to the nucleus (Fig. 4C,F). The specificity of the antibody was verified using Prep1^–/– embryoid bodies (EBs) (data not shown).

We also analyzed the expression of Prep1 at pre-implantation stages by RT-PCR (Fig. 4I) and during ES cell differentiation by qRT-PCR (Fig. 4G). Prep1 was expressed from the one-cell to the blastocyst stage (Fig. 4I). During ES cell differentiation, Prep1 was not up- or downregulated, unlike the pluripotent genes Oct4 and Nanog (the expression of which decreased during differentiation), and the neural marker Nestin [the expression of which increased (Fig. 4G)].

We have also studied the expression of Prep2, Meis1 and Meis2 during ES cell differentiation. Unlike Prep1, the levels of expression of Prep2, Meis1 and Meis2 increased during differentiation (Fig. 4H). Accordingly, in the E8.5 embryo, when organogenesis starts, Prep2 (Fig. 4J), Meis1 (Fig. 4K) and Meis2 (Fig. 4L) were not ubiquitously expressed, but had more restricted expression patterns.

We conclude that Prep1 is expressed from the earliest stages of embryogenesis, ubiquitously at gastrulation, and is not regulated during ES cells differentiation. These features are unique within the MEIS class of transcription factors.

We have derived ES cells from E3.5 blastocysts of Prep1^+/– intercrosses. The derivation yield of Prep1^–/– ES cell lines was low (2/26) (Fig. 5A,B; data not shown). However, these lines expressed Oct4 and Nanog (Fig. 5C; data not shown).

Prep1^–/– ES cells were able to differentiate using the hanging drop protocol; however, this gave rise to smaller EBs than in wild-type ES cells, particularly for one of the two Prep1^–/– cell lines (Fig. 5D). Similarly to our whole-mount results, the cleaved caspase 3 staining was more intense in Prep1^–/– than in wild-type EBs (Fig. 5E,F), especially in the smaller EBs. The differences were not as pronounced as in embryos (Fig. 2I,J). Analysis of specific markers (Oct4 and Nanog for pluripotency, T for mesoderm, Fgf5 and Sox1 for neuroectoderm, Gata4 for endoderm and Hand1 for trophoectoderm) showed that, although there was a slight delay, Prep1^–/– ES cells were overall capable of differentiation (data not shown).

Altogether, these results confirm the role of Prep1 in homeostasis in early embryogenesis; Prep1^–/– ES cells are extracted in lower ratio than expected (8% instead of the expected 25%). The two viable Prep1^–/– ES cell lines phenocopy Prep1^–/– embryos, although in a milder fashion, giving rise to small EBs with increased apoptosis.

E6.5 Epi cells do not tolerate double-stranded DNA damage and undergo p53 dependent apoptosis (Heyer et al., 2000). In Prep1^i/i hypomorphic MEFs, we observed an increase of basal apoptosis (Micali et al., 2009) and increased double strand break response (G.I. and F.B., unpublished). To test whether DNA damage was at the basis of Prep1^–/– p53-dependent apoptosis of Epi cells, we crossed Prep1^+/– mice with Atm^+/– mice (Borghesani et al., 2000), both in a full C57BL6 background. As Atm is important in DNA repair, one would expect its absence or reduction to worsen the Prep1 phenotype. The absence of Atm induced a phenotype in two out of four heterozygous Prep1^+/– embryos (Fig. 5G-J), which were delayed in their development, whereas in the other two embryos the effects were less evident. Moreover, in the single recovered Prep1^–/–Atm^–/– double knockout embryo, the phenotype was stronger than in the Prep1^–/–Atm^+/+, because, by E7.5, the Oct4-positive cells were almost undetectable (Fig. 5G-J). In agreement with the above results, downregulation of Atm with an sh-RNA in Prep1^–/– ES EBs induced a further decrease of their already deficient size and increased cleaved caspase 3 staining, i.e. apoptosis (data not shown).

The absence of Prep1 causes p53-dependent apoptosis of Epi cells, which prevents gastrulation and differentiation. This is probably due to the accumulation of DNA damage. The absence of Prep1 affects all cells but apoptosis was mostly observed in the Epi, probably owing to the tremendous proliferative expansion of the Epi at this stage. Thus, the role of Prep1 in early embryogenesis is to protect epiblast cells from accumulating damage that induces apoptosis.

Our data suggest that Prep1^–/– embryos arrest around E5.0 or E5.25. Indeed, they die before E5.5, which is when AVE is established in the distal tip of the embryo (Tam and Loebel, 2007). However, death must occur after the time of expression of Oct4 (which is present at implantation in the Epi) (Pfister et al., 2007). This timing coincides with the loss of Wnt3 expression in the posterior Epi at E5.75 (Rivera-Perez and Magnuson, 2005).

The lack of gastrulation is due to the reduced size of the epiblast. Prep1 knockout embryos show a decrease of the pluripotent cells expressing Oct4 at E6.5, E7.5 and E8.5 (Fig. 2A-F). The number of Epi cells is a limiting factor for the initiation of gastrulation (Tam and Behringer, 1997). Elimination of one blastomere from a two- or four-cell embryo delays gastrulation; gastrulation starts only when the embryo has accumulated a sufficient number of Epi cells (Power and Tam, 1993; Rands, 1986). The size recovery is achieved by increasing cell proliferation prior to and during gastrulation (Power and Tam, 1993). It is likely that the low number of Epi cells explains the absence of the AVE and PS in Prep1^–/– embryos, and, therefore, the absence of embryonic lineages. Nodal signaling from the Epi is essential for specifying AVE in the most distal part of the embryo (Brennan et al., 2001; Mesnard et al., 2006). The lack of expansion of the Prep1^–/– Epi may not keep the distal cells far enough from the negative signals of the ExE (Mesnard et al., 2006; Richardson et al., 2006; Rodriguez et al., 2005), hence maintaining them under the inhibitory control of the ExE.

The pattern of Prep1 expression during ES cell differentiation and the ability to harvest Prep1^–/– ES cells, which still express Oct4 and can, on the whole, differentiate (data not shown), suggests that Prep1 is not a reprogramming or pluripotency-establishing gene (Avilion et al., 2003; Chambers et al., 2003; Mitsui et al., 2003; Nichols et al., 1998).

The decreased number of Oct4-expressing cells, together with the increased rate of p53-dependent apoptosis, suggest that Prep1 is, however, essential for Epi homeostasis. This is also confirmed by the in vitro analysis of Prep1^–/– ES cells. Our current explanation for the p53-dependent apoptosis of the Prep1^–/– epiblast is DNA damage.

The early post-implantation embryo is very sensitive to UV irradiation, and Epi cells irradiated with doses that do not induce any response at other stages or in somatic cells undergo p53-dependent apoptosis (Heyer et al., 2000). We hypothesize that in the absence of Prep1, DNA damage accumulates in the embryonic cells. The DNA damage hypothesis is suggested by the observation that a decrease in Atm induces a phenotype in the heterozygous, and worsens the phenotype of the homozygous, Prep1 mutant embryos. Likewise, downregulation of Atm in Prep1^–/– but not in wild-type ES cells induces apoptosis and inhibits EB growth. This argues that the inability to repair DNA synergizes with the absence of Prep1 and drives Epi cells into apoptosis. Accordingly, in Prep1^i/i hypomorphic MEFs, we observed an increase in double-stranded breaks and DNA repair signaling (G.I. and F.B., unpublished). Why the absence of Prep1 could induce DNA damage is, at the moment, matter for speculation.

Interestingly, several cancer genes are involved in controlling DNA damage response and DNA repair. Indeed, several pieces of evidence demonstrate that Prep1 is a novel tumor suppressor (Longobardi et al., 2010).

We are very grateful to Dr Claudio Stern for his interest and advice; to the late Graziella Persico and to Giovanna Liguori for much advice and reagents; to Luisa Lanfrancone for reagents and helpful discussions; and to Anton Berns for Ink4A^p16 mice. We also thank Antonello Mallamaci, Shankar Srinivas, Silvia Brunelli, Miguel Torres, Janet Rossant, Juan Pedro Martinez Barberà, Tristan Rodriguez, Michael Shen, Daniel Constam, Vania Broccoli, Hans Schoeler, Isao Matsuo and Robin Lovell-Badge for probes, and Michael Hemann for the shATM vector. This work was funded by grants from TELETHON Onlus (Italy) and Ministero della Salute to F.B., and from FIRC to G.I.