Blimp1 regulates development of the posterior forelimb, caudal pharyngeal arches, heart and sensory vibrissae in mice

The zinc-finger transcriptional repressor, Blimp1 (also known as Prdm1),originally identified as a silencer of β-interferon gene expression(Keller and Maniatis, 1991)and a master regulator controlling terminal differentiation of B lymphocytes(Turner et al., 1994), also governs T-lymphocyte homeostasis (Kallies et al., 2006; Martins et al.,2006), primordial germ cell (PGC) specification(Ohinata et al., 2005; Vincent et al., 2005) and stem cell maintenance in the sebaceous gland(Horsley et al., 2006). Blimp1 functions to regulate cell cycle progression and globally redirects patterns of gene expression(Shaffer et al., 2002; Ohinata et al., 2005; Shapiro-Shelef and Calame,2005; Kallies and Nutt,2007). Blimp1 contains an N-terminal PR/SET domain, a proline-rich region, five C₂H₂ zinc fingers and a C-terminal acidic domain (Tunyaplin et al.,2000; Turner et al.,1994). The zinc-finger region mediates nuclear import and DNA binding (Keller and Maniatis,1992; Gyory et al.,2003; Gyory et al.,2004). The evolutionarily conserved PR/SET domain, closely related to histone methyl transferases, recruits additional epigenetic modifiers that cooperatively function to modify chromatin structure and mediate gene silencing (Ren et al., 1999; Yu et al., 2000; Gyory et al., 2004; Ancelin et al., 2006; Hayashi et al., 2007; Surani et al., 2007). Recent experiments suggest that the transcriptional networks controlled by Blimp1 are cell-type specific. For example, Blimp1 represses c-myc (also known as Myc - Mouse Genome Informatics)expression in mature B cells (Lin et al.,1997) and sebaceous gland progenitors(Horsley et al., 2006) but c-myc is not a transcriptional target in activated T-effector cells(Gong and Malek, 2007). In addition to governing exit from the cell cycle, Blimp1 also protects plasma cells from ER stress by maintaining XBP expression, a key transcription factor required for immunoglobulin secretion(Shaffer et al., 2004). Blimp1 has impressive abilities to regulate cell growth, differentiation and survival through its diverse molecular partnerships in highly specialised cell types.

Blimp1-deficient embryos die around embryonic day (E) 10.5(Vincent et al., 2005) because of widespread haemorrhaging and blood pooling within the dorsal aortae. The primitive heart seems to function normally, but placental development is severely compromised. Blimp1 mutants fail to expand the labyrinthine region where the bulk of foetal and maternal exchange takes place leading to placental insufficiency. In the embryo proper Blimp1 is strongly expressed in the key signalling centres that pattern the forebrain, namely the leading anterior mesendoderm and prechordal plate, but development of the head structures proceeds normally in Blimp1 mutants. Strikingly Blimp1 functional loss disrupts PCG specification(Ohinata et al., 2005; Vincent et al., 2005). Blimp1 silences the default somatic pathway and allows a few epiblast cells to become exclusively allocated into the germ cell lineage (reviewed by Hayashi et al., 2007). At later stages, Blimp1 is transiently expressed in subsets of the pharyngeal endoderm and ectoderm, splanchnic mesoderm, endothelial cells and somites (Vincent et al.,2005), and in the papillae of the teeth, hair and taste buds(Chang et al., 2002). Owing to the early lethality, possible Blimp1 functions in these diverse tissues have yet to be elucidated.

To bypass placental defects and examine Blimp1 activities in the embryo, here we used a Sox2-Cre deleter strain(Hayashi et al., 2002) in conjunction with a Blimp1 conditional allele(Shapiro-Shelef et al., 2003). Blimp1 functions normally in the extra-embryonic cell lineages and as a result, mutant embryos survive to late gestation stages. The present experiments demonstrate that Blimp1 is required for development of the posterior forelimb bud, pharyngeal epithelia, the secondary heart field(SHF) and dermal papillae (DP) of the sensory vibrissae. Blimp1maintains expression of key signalling molecules and coordinates cell proliferation and differentiation at these diverse tissue sites. Additionally we characterised Blimp1 mutant embryos carrying an IRES-gfpreporter allele that survive to late gestation(Kallies et al., 2004). In addition to expression of truncated Blimp1 protein lacking the C-terminal zinc fingers (Kallies et al.,2004), Blimp1^gfp/gfp embryos also express full-length Blimp1 protein. These embryos display forelimb, pharyngeal and heart defects and a complete loss of germ cells, but sensory vibrissae develop normally. Interestingly, double heterozygotes carrying the hypomorphic Blimp1^gfp allele and a null allele display more severe defects that closely resemble those described here for Blimp1mutants. Collectively, the present experiments demonstrate that Blimp1 requirements in diverse progenitor cell populations are exquisitely dose dependent.

The Prdm1^BEH(Vincent et al., 2005), Prdm1^flox(Shapiro-Shelef et al., 2003), Blimp1^gfp (Kallies et al., 2004) targeted alleles and R26R reporter mice(Soriano, 1999) have been described. Mice and embryos were genotyped by PCR as described in the original reports. The Sox2-Cre (Hayashi et al., 2002), Tie2-Cre(Kisanuki et al., 2001) and Blimp1-Cre (Ohinata et al.,2005) transgenic strains were maintained on a mixed outbred background and genotyped using Cre-specific PCR.

Embryos were fixed overnight in 4% paraformaldehyde (PFA). Whole-mount in situ hybridisation on intact embryos was performed according to standard protocols. Standard probes specific for Shh, Fgf8, Fgf4, Tbx2, Tbx3, Ptc1,Gli1, Lmx1, Tbx1, Blimp1, Bmp4, Grem1 and Cre were used. X-gal staining was performed as described (Nagy et al., 2003). Embryos were post fixed, photographed, dehydrated and embedded for histology. Sections (8 μm) were mounted and counterstained with eosin. For standard histology, material was fixed overnight in 4% PFA in PBS and processed for standard haematoxylin and eosin staining. Hearts were manually dissected, photographed and processed as above.

Embryos were fixed overnight in 4% PFA, dehydrated in ethanol, embedded and sectioned at 6 μm. Sections were subject to antigen retrieval by boiling for 20 minutes in either Antigen Retrieval Solution (DAKO), (Ki67 and phosphohistone 3 staining) or Tris-EDTA pH 9.0 (c-myc staining), blocked for 5 minutes in peroxidase block (DAKO K4011), washed in PBS, incubated in primary antibody overnight at 4°C, washed in PBS and developed using the appropriate DAKO peroxidase-labelled polymer kit and DAB, and counterstained with haematoxylin. Antibodies were anti-c-myc (Santa Cruz N-262, sc-764 1:250), Ki67 (NovoCastra NCL-L-MM1, 1:200) and anti-phosphohistone 3 (Upstate 06-570, 1:200). For visualisation of apoptotic cells, embryos were incubated for 30 minutes at 37°C in Lysotracker Red (Molecular Probes L-7528 RED;diluted 1:400 in Hank's balanced salt solution), washed in HBSS, fixed overnight in 4% PFA, dehydrated in methanol and viewed by fluorescence microscopy.

PGCs were visualised at E9.5 by staining for alkaline phosphatase activity,flat mounted under coverslips in 80% glycerol, and photographed under bright-field microscopy as described previously(Lawson et al., 1999).

Intact embryos or isolated limbs were fixed in Bouin's and cartilage and bone were stained using Alcian Blue and Alizarian Red as described(Nagy et al., 2003). Specimens were cleared in benzyl alcohol:benzyl benzoate (1:1) before photography.

Total RNA from STO fibroblasts, individually genotyped E9.5 mouse embryos,or spleen cells cultured with LPS was prepared using Trizol (Invitrogen). The OneStep RT-PCR kit (Qiagen) was used with Blimp1 primer pairs (Exon 6 forward, 5′-CGGAAAGCAACCCAAAGCAATAC-3′ and Exon 6 reverse,5′-CCTCGGAACCATAGGAAACATTC-3′; Exon 6 forward,5′-GGTTACAAGACTCTTCCTTAC-3′ and Exon 8 reverse,5′-GCTCTTGTGACACTGGGCACA-3′; Exon 8 forward,5′-GCAATCTCAAGACCCACCTTC-3′ and Exon 8 reverse,5′-CGAACCTCTCAATTTCTTCATT-3′), and Hprt primer pair(forward, 5′-GCTGGTGAAAAGGACCTCT-3′ and reverse,5′-CACAGGACTAGAACACCTGC-3′) with an annealing temperature of 58°C for 30 cycles.

Spleen cells cultured for 3 days in the presence of LPS (50 μg/ml), were washed and the cell pellets lysed in buffer containing 1% Nonidet P-40, 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 5 mM EDTA, and protease inhibitors (Sigma). Individually genotyped E9.5 embryos were lysed in RIPA buffer plus protease inhibitors (Sigma). Extracts were centrifuged for 30 minutes at 20,800 g. Lysates were mixed with 2 × Laemmli buffer and boiled for 5 minutes before fractionation on an 8% polyacrylamide gel. Proteins were transferred onto a nitrocellulose membrane (catalogue BA83; Schleicher &Schüll) for 2 hours at 500 mA. Membranes rinsed in TBS-T were blocked for 1 hour in TBS-T with 5% dried milk, and incubated at 4°C overnight with rat monoclonal anti-Blimp1 (Kallies et al., 2004) then secondary antibody (goat anti-rat Ig, cat. no. NA935; Amersham Biosciences) and developed by chemiluminescence using ECL Plus(Amersham Biosciences).

Full-length Blimp1 cDNA(Turner et al., 1994) and mutagenised Blimp1 were subcloned into a modified version of pCAGGS(Niwa et al., 1991). For mutagenesis, Val556 of the Blimp1 coding sequence was changed to a stop codon using the QuikChange II Site-Directed Mutagenesis Kit (Stratagene). Primers used for mutagenesis were 5′-GCTCTCCAACCTGAAGTAACACCTGAGAGTGCAC-3′ (sense) and 5′-GTGCACTCTCAGGTGTTACTTCAGGTTGGAGAGC-3′ (antisense). COS cells were grown on coverslips and transfected with 0.8 μg DNA complexed with Lipofectamine 2000 (Invitrogen). Cells were washed with PBS,fixed with 4% formaldehyde, quenched with NH₄Cl, incubated in blocking buffer containing 2.5% goat serum, 2.5% donkey serum, 2% bovine serum albumin (BSA), 2% fish skin gelatin (Sigma) and 2.5% Triton X-100 and incubated overnight at 4°C in blocking buffer containing mouse monoclonal anti-Blimp-1 (1:200, Novus Biologicals). After three rinses, coverslips were incubated for 1 hour at room temperature in blocking buffer containing secondary antibody (1:2000 dilution, Invitrogen A11029), rinsed and placed inverted onto a drop of mounting medium containing DAPI (Vectashield, Vector Laboratories). Samples were visualised with a Zeiss LSM510 META confocal microscope.

It seemed likely that Blimp1 null embryos die by E10.5 due to placental defects, but loss of Blimp1 expression in the developing endothelium (Vincent et al.,2005) might also compromise integrity of the blood vessels. To distinguish these possibilities, mice homozygous for the Prdm1conditional allele (Shapiro-Shelef et al.,2003) (Prdm1^flox/flox) were crossed to Sox2-Cre transgenic mice (Hayashi et al., 2002) also carrying the Prdm1^BEH null allele (Vincent et al., 2005). Sox2-Cre activated at the blastocyst stage is exclusively expressed in epiblast derivatives: that is, only the embryo proper(Hayashi et al., 2002). Sox2-Cre Prdm1^BEH/flox embryos were collected at E10.5 through to E16.5 (Table 1). As expected, homozygous Prdm1^-/- or Prdm1^BEH/BEH null embryos failed to survive beyond E10.5. By contrast, Sox2-Cre Prdm1^BEH/flox embryos (hereafter referred to as Blimp1 mutant embryos) were present at mendelian frequencies at E16.5 (Table 1),and were still viable at E18.5 (data not shown). However, we failed to recover live Blimp1 mutants in newborn litters.

The rescued embryos displayed normal blood vessels. Additionally,conditional inactivation of Blimp1 in endothelial cells using the well described Tie2-Cre transgene(Kisanuki et al., 2001)yielded viable animals. Mendelian numbers of Tie2-Cre Prdm1^BEH/flox-positive weanlings were recovered (22 out of 93), from crosses between Tie2-Cre Prdm1^BEH/+ and Prdm1^flox/flox animals. We therefore conclude that the early lethality of Blimp1 null embryos exclusively reflects Blimp1 requirements in the extra-embryonic cell lineages and is not caused by endothelial cell defects.

At late gestational stages Sox2-Cre-rescued embryos consistently displayed two overt abnormalities. The forelimbs were missing posterior digits and sensory vibrissae were absent. Otherwise mutant embryos were indistinguishable from wild-type and heterozygous littermates. PCR genotyping of a large panel of individual mutant embryos confirmed that these overt defects were associated with efficient and complete Cre recombination (see Fig. S1 in the supplementary material).

Blimp1 mRNA expression in the mesenchyme marks the developing forelimb and hindlimb buds coincident with their emergence(Vincent et al., 2005). During limb bud outgrowth Blimp1 expression was restricted to the posterior mesenchyme encompassing the zone of polarising activity (ZPA)(Fig. 1). By E12.5 Blimp1 expression in the mesenchyme was downregulated, and Blimp1 transcripts were weakly induced in the apical ectodermal ridge(AER) (Fig. 1H, and data not shown). To further characterise forelimb defects, the elements of the appendicular skeleton were visualised at E13.5 and 16.5 using Alcian Blue and Alizarian Red staining. As shown in Fig. 1A-E, the forelimbs attained normal size and the skeletal elements were correctly proportioned. However, the posterior proximal elements, and associated muscles and connective tissue were completely absent (see Fig. S2 in the supplementary material). Thus mutant forelimbs entirely lacked the ulna, and either two (7/15) or three (8/15) of the posterior digits, but were otherwise indistinguishable from the wild type.

The signalling centres controlling outgrowth and patterning of the early limb bud are well characterised (reviewed by Tickle, 2006; McGlinn and Tabin, 2006). Anterior-posterior (A-P) polarity is patterned by Shh signalling from the ZPA, proximal-distal (P-D) outgrowth is controlled by Fgfsignalling from the AER, whereas the dorsal-ventral (D-V) axis requires Wnt signals from the dorsal ectoderm. The AER and ZPA are maintained by reciprocal inductive interactions between Fgf and Shhsignalling pathways. We previously found at E9.5 that limb bud formation initiates normally, and Shh and Fgf8, markers of the ZPA and AER, respectively, are correctly induced in Blimp1 mutant embryos(Vincent et al., 2005).

To further evaluate the forelimb patterning defects observed here, we analysed a panel of molecular markers between E10.5 and 12.5(Fig. 2). Shhexpression in the ZPA was induced with the appropriate temporal and spatial kinetics in both forelimbs and hindlimbs(Fig. 2A,B). However, Shh⁺ ZPA progenitor cells in the forelimb are short-lived. Thus Shh transcripts are prematurely downregulated, and except for a small patch of expression corresponding to the distal-most edge of the ZPA(Fig. 2C,D), are largely absent by E11.5. Analysis of Fgf expression(Fig. 2G-J; see Fig. S3 in the supplementary material), together with the observation that mutant forelimbs attain wild-type length, demonstrated that the AER is correctly induced and at early stages seems to function normally. However, premature cessation of Shh signalling resulted in a failure to maintain localised Fgf8 expression within the AER and as a result, the posterior region of the AER was selectively lost (Fig. 2J; see Fig. S3 in the supplementary material). Nonetheless, digit 2, which depends on paracrine Shh signalling(Harfe et al., 2004), was correctly patterned (Fig. 1). The Shh target genes Ptch1 and Gli1 were activated(Fig. 2P, data not shown). Thus the ZPA functions appropriately to impart initial A-P pattern. However, at later stages, Blimp1 mutant forelimbs displayed truncated expression of posterior markers genes, such as Tbx2(Fig. 2M,N).

The positive feedback loop between the ZPA and the AER is controlled by the BMP antagonist Grem1 (Khokha et al., 2003; Panman et al.,2006). ZPA and AER signals are prematurely lost in Grem1mutants causing limb patterning defects similar to those described here(Khokha et al., 2003). However, the forelimb defects cannot be attributed to loss of Grem1function as Grem1 was correctly induced, albeit at lower levels, in mutant forelimb buds (Fig. 2Q,R). Wnt7a signalling from the dorsal ectoderm is also required for normal A-P patterning. Loss of Wnt7a(Parr and McMahon, 1995) or its target Lmx1a similarly results in loss of the ulna and posterior digits (Chen and Johnson,2002). However, as shown in Fig. 2L, Lmx1a was expressed normally in the dorsal mesenchyme and histological analysis at late stages confirmed that mutant forelimbs develop normal D-V tissue polarity (see Fig. S2 in the supplementary material). Collectively, these experiments demonstrate that A-P, P-D and D-V patterning is correctly initiated at E10.5,but defects in the developing fore but not the hindlimb are caused by premature loss of the ZPA.

To generate a fate map of Blimp1-expressing cells in the early limb bud used a Blimp1-Cre transgenic strain(Ohinata et al., 2005). Our results shown in Fig. 1,together with the previous work (Ohinata et al., 2005) demonstrated that Cre activity closely resembles the endogenous Blimp1 expression pattern(Fig. 1I). Embryos obtained from crossing Blimp1-Cre males to carrying females the R26R lacZ reporter cassette (Soriano,1999), were stained for β-galactosidase activity at different gestational stages. As shown in Fig. 1, Blimp1⁺ cells gave rise to the posterior elements of the proximal limb. Blimp1⁺ cells in the forelimb selectively formed the ulna, digits 4 and 5, and a subset of cells in digit 3(Fig. 1J-L). Similarly in the hindlimb, Blimp1⁺ cells represented the progenitors of the fibular and the posterior three digits (data not shown). Blimp⁺ cells contributed to the muscles and connective tissues in the posterior fore and hindlimbs, but not the surface ectoderm(Fig. 1L). Remarkably, this fate map of Blimp1⁺ cells could be precisely superimposed on that recently described for Shh⁺ cells(Harfe et al., 2004). We conclude that Blimp1⁺ Shh⁺ progenitor cells in the ZPA give rise to much of the proximal posterior region of the limb and this discrete cell lineage is entirely lost in Blimp1 mutant forelimbs.

Careful histological examination, together with Lysotracker Red staining,failed to reveal any evidence for selective cell death in Blimp1mutant forelimbs (data not shown). Moreover staining with antibodies specific for phospho-histone3 and Ki67 at E10.5, revealed a uniform high labelling index throughout the mesenchyme in both wild-type and mutant forelimb buds(see Fig. S4 in the supplementary material).

By late gestation, five prominent rows of sensory vibrissae were normally visible on either side of the snout. Additionally, single or clusters of discrete whiskers were induced around the eye and on the underside of the lower jaw (Fig. 3A,C). Blimp1 mutant embryos lacked all sensory vibrissae(Fig. 3B,D), but the other head structures that express Blimp1, including the papillae of the teeth and taste buds, and the neural retina(Chang et al., 2002), all developed normally (Fig. 3C,Dand data not shown).

Hair follicles are induced at two discrete stages in the developing mouse embryo. Induction of the sensory vibrissae initiates at E12.5, whereas the hair follicles of the coat, or pelage, appear several days later. Reciprocal inductive interactions between the mesenchyme of the dermis and the overlying surface ectoderm induce formation of localised thickenings or placodes that then invaginate to form the hair follicle. The underlying mesodermal cells coalesce to form the dermal papilla (reviewed by Fuchs et al., 2001; Millar, 2002; Fuchs, 2007). As shown in Fig. 3E, at E12.5 Blimp1 localised to the mesenchymal condensates at sites of vibrissae induction, and closely overlapped with the Bmp4 and Shhexpression domains (Fig. 3G-I). Blimp1 expression thus identifies the DP progenitors located under characteristic surface elevations that preconfigure the sites of ectodermal placode induction (Fig. 3G). Fate-mapping studies showed that these Blimp1⁺ cells give rise to the mature DP and also expand to form a mesenchymal layer immediately surrounding the hair follicles (Fig. 3J-L, and data not shown). Blimp1 expression was maintained within the mature DP (Chang et al., 2002) (data not shown) but Blimp1⁺ progenitors not incorporated into the DP subsequently downregulated Blimp1 expression and migrated around the sides of the hair follicle. Interestingly Blimp1 was also expressed in the dermal condensates and prospective DP of the pelage hair follicles (Fig. 3M) but in striking contrast to sensory vibrissae, pelage hair follicles developed normally (Fig. 3N,O).

Shh, which is activated in response to Wnt/Lef1 signals from the ectoderm, is required to organise the dermal cells into focal condensates(reviewed by Fuchs, 2007). In Blimp1 mutants Shh, Bmp4 and Gli1 expression was weakly induced, but not maintained, in the dermal mesenchyme(Fig. 3F, see Fig. S5 in the supplementary material). The surface ectoderm elevations that preconfigure the sites of vibrissae induction were visible but mutant embryos lacked localised dermal condensates and placodal thickenings(Fig. 3S). Thus Blimp1is required for maintenance but not initial specification of DP progenitors.

The absence of DP is not due to apoptosis of the dermal condensate because there was no evidence for foci of Lysotracker-Red-positive cells in E12.5 or 13.5 mutant embryos (data not shown). Next, we examined cell proliferation by staining with a Ki67 antibody. The prospective DP of the vibrissa was conspicuously quiescent compared with cells in the adjacent mesenchyme and overlying surface ectoderm (Fig. 3P-R). These localised foci of Ki67 negative cells were entirely absent in Blimp1 mutant embryos. Rather the mesenchyme underlying the surface ectoderm elevations showed a uniform mitotic index(Fig. 3S).

Finally, as has been shown in B cells and the sebaceous gland(Horsley et al., 2006), we wondered whether Blimp1 functions in DP progenitors to repress c-myc expression. The Ki67-negative DP cells normally failed to express c-myc (Fig. 3T) but we did not detect patches of c-myc-negative cells in the mesenchyme underneath the ectodermal elevations of Blimp1 mutants(Fig. 3U). Rather, in Blimp1 mutants, DP progenitors continued to divide and disperse and failed to act as localised signalling centres. Similarly, the Shh⁺ cells transiently observed in E12.5 Blimp1mutants are also present in diffuse patches and not discrete clusters(Fig. 3F).

The pharyngeal arches develop as a series of reiterated bulges formed in a rostral to caudal fashion. Reciprocal signalling between the surface ectoderm and endoderm, and the underlying mesoderm and neural crest (NC) cells control tissue morphogenesis. The surface epithelia expand to form discrete pouches into which the mesoderm and NC migrate (reviewed by Graham, 2001; Graham et al., 2005). We previously found that only the first arch forms and the caudal arches are completely lost in Blimp1 null embryos(Vincent et al., 2005).

As shown in Fig. 4, Blimp1 was strongly expressed in the anterior epithelia of the individual pharyngeal arches (Vincent et al., 2005) (Fig. 4A). These Blimp1⁺ cells expand rapidly giving rise to the bulk of the pharyngeal arch surface ectoderm and endoderm at E10.5(Fig. 4B-D). Histological examination of Sox2-Cre-rescued Blimp1 mutant embryos from E10.5 onwards showed that the jaws - which derive from the first pharyngeal arch - develop normally, but none of the tissue structures formed from the caudal arches are present. Blimp1 mutants lack the thymus, which normally forms from an outpocketing of the pharyngeal endoderm of the third arch (Fig. 4G,H). Additionally,hypoplasia of the pharyngeal epithelium secondarily resulted in failure to form the pharyngeal arch arteries. The pharyngeal arch arteries normally give rise to a complex vascular network connecting the dorsal aorta to the outflow tract vessels of the heart. Loss of the pharyngeal arch artery system led to massive disturbances of the thoracic vasculature(Fig. 4E,F).

We previously found at E8.5 that Blimp1 expression identifies a discrete sub-region of the splanchnic mesoderm(Vincent et al., 2005). By E9.5, this expression domain was positioned anteriorly and dorsal to the developing heart tube (Fig. 4A)marking the anterior or SHF progenitors of the inflow and outflow poles of the heart tube (reviewed by Buckingham et al.,2005; Srivastava,2006). As shown in Fig. 5, Blimp1⁺ SHF progenitors contributed almost exclusively to the myocardium of the right ventricle (RV), the ventricular septum and cells surrounding the flow tract (OFT). Relatively few marked cells were present in the left ventricle (LV) and only occasionally in the atria. At later stages (E16.5), Blimp1⁺ descendents contributed to the walls of the aorta and pulmonary artery as well as the leaflets of the ventricular outflow valves (Fig. 5D). Blimp1 expression thus identifies multipotent myocardial progenitor cells within the SHF (Verzi et al.,2005).

Early heart morphogenesis and formation of the right ventricle proceeded normally in Sox2-Cre-rescued embryos; however, at E16.5, mutant hearts (n=6) had a more pronounced apex, lacked the aortic arch and displayed persistent truncus arteriosus (PTA) and a pronounced ventricular septal defect (VSD) (Fig. 5). Thus, Blimp1 loss disturbs alignment, septation and rotation of the arteries. Fgf8 expression in surface epithelium, primarily the pharyngeal endoderm, promoted migration of the mesenchyme and neural crest into the forming arches. Dose-dependent Fgf8 signalling in the SHF is required for morphogenesis of the OFT(Brown et al., 2004; Macatee et al., 2003; Park et al., 2006). Loss of the T-box transcription factor Tbx1 also results in cardiovascular and pharyngeal defects similar to those described here(Jerome and Papaioannou, 2001; Lindsay et al., 2001; Merscher et al., 2001). Tbx1 expression in the mesenchyme is required to support OFT morphogenesis and pharyngeal arch development, whereas expression in the endoderm is required for formation of the thymus(Zhang et al., 2006). Interestingly, at E9.5, Fgf8 and Tbx1 expression was severely compromised but not lost in Blimp1-deficient embryos(Fig. 5I-L). Thus, Blimp1 is essential to maintain Tbx1 and Fgf8expression in pharyngeal arch epithelia progenitors. These results suggest that the complex pharyngeal and heart abnormalities in Blimp1 mutant embryos are probably caused by a failure to expand and maintain signalling capabilities within the surface epithelium.

A Blimp1^gfp reporter allele previously generated via insertion of a STOP-IRESgfp-pgk neo cassette 3′ to exon 6 eliminates expression of the C-terminal zinc fingers and effectively acts as a null mutation in B cells (Kallies et al.,2004). However, in contrast to the early lethality seen in Prdm1^BEH/BEH and Prdm1^-/- null embryos at E10.5, homozygous Blimp1^gfp/gfp embryos survive to late gestational stages at almost mendelian ratios(Kallies et al., 2004)(Table 2). Because downstream coding exons remain intact, we wondered whether enhanced survival of Blimp1^gfp/gfp embryos was associated with leaky expression of wild-type transcripts. To evaluate this possibility, we initially performed RT-PCR using appropriate primer combinations spanning exons 6, 7 and 8. As shown in Fig. 6A, in addition to transcripts that prematurely terminate within the IRESgfp-pgk neo cassette, Blimp1^gfp/gfp embryos expressed correctly spliced wild-type transcripts. Similar conclusions were reached in RPA experiments(data not shown).

Western blot experiments demonstrate that LPS-activated heterozygous Blimp1^gfp/+ spleen cells and Blimp1^gfp/gfp embryos produce truncated Blimp1 protein lacking the C-terminal zinc-finger domains(Kallies et al., 2004). However, we also clearly detect full-length Blimp1 protein in E9.5 Blimp1^gfp/gfp embryos(Fig. 6B). We also engineered an expression vector via introduction of a stop codon at amino acid 556. In contrast to full-length Blimp1, truncated Blimp1 protein lacking the zinc fingers failed to enter the nucleus (Fig. 6C). Collectively, these results strongly suggest that the truncated Blimp1 protein encoded by the gfp reporter allele has no functional activity. Rather the enhanced survival of Blimp1^gfp/gfpembryos can best be explained by low levels of full-length mRNA and protein expression.

Interestingly, Blimp1^gfp/gfp mutant embryos(n=5) displayed the identical pharyngeal and cardiovascular defects as described above for Blimp1 mutant embryos, including PTA and VSD(Fig. 7A,B). However, forelimb development was less severely compromised. In most cases the ulna formed correctly and only a single posterior digit was absent(Fig. 7G). Moreover, induction of the sensory vibrissae occurred normally(Fig. 7J). Interestingly reduced numbers of PGCs were detectable in heterozygous Blimp1^gfp/+ littermates whereas none of 12 E9.5 Blimp1^gfp/gfp embryos contained alkaline-phosphatase-positive PGCs (Fig. 7C,D). Homozygous Blimp1^gfp/gfp embryos thus completely lacked germ cells and in this respect phenocopy the null mutants(Fig. 7E).

Next, to examine the consequences of further reduced Blimp1expression levels heterozygotes carrying the hypomorphic Blimp1^gfp/+ allele and the Blimp1^BEH/+null allele (Vincent et al.,2005) were intercrossed to generate transheterozygous Blimp1^gfp/- embryos. Only a small proportion survive to E13.5 (Table 2). As for Blimp1 mutant embryos, both the ulna and posterior digits were missing (Fig. 7H), and interestingly the dermal condensates of the vibrissae were induced to form;however, only a small fraction of these proceeded to the ectodermal placode stage (Fig. 7K). We conclude that weak Blimp1 expression is sufficient to promote development of the extra-embryonic tissues, sensory vibrissae and the posterior forelimb,whereas increased Blimp1 activities are required for pharyngeal and heart morphogenesis, and PGC specification requires highest levels of Blimp1 expression.

Blimp1 null embryos arrest at around E9.5-10.5 because of an under-developed labyrinth: the region of the placenta responsible for maternal foetal exchange (Vincent et al.,2005). Bypassing Blimp1 requirements in extra-embryonic tissues allowed us to rescue the early lethality and demonstrate essential roles in the posterior forelimb, caudal pharyngeal arches, OFT of the heart,and dermal papillae of the sensory whiskers. Blimp1 expression identifies multipotent progenitor cell populations allocated during tissue morphogenesis and is essential for maintenance of key signalling centres guiding cell proliferation and differentiation within these complex structures.

Blimp1 functional loss disrupts development of the highly specialised sensory vibrissae. Blimp1 expression in the dermal condensates marks islands of quiescent cells within the expanding population of head mesenchyme. Blimp1 induced in response to instructive signals from the surface ectoderm, represses c-myc expression, arrests cell division,and causes cells to coalesce and form a signalling centre. Neighbouring cells continue to rapidly divide and disperse into the surrounding mesenchyme.

Time-dependent Shh signalling in the ZPA controls specification of the tissues of the posterior limb (reviewed by Tickle, 2006). Likewise, Blimp1 is required in multipotent progenitors giving rise to the skeletal elements, muscles and connective tissue of the posterior forelimb. Our results demonstrate that Blimp1 maintains this short-lived progenitor cell population. However, Blimp1 is not essential for Shh activation because the onset of Blimp1 expression in the prospective ZPA of the forelimb occurs around day 9 of development, well in advance of Shh. Moreover, early forelimb patterning proceeds normally in Blimp1 mutant embryos.

Similarly fate-mapping experiments show that transient Blimp1expression in the forming pharyngeal arches is essential for expansion of ectodermal and endodermal progenitors giving rise to the entire surface epithelia of the pharynx. Loss of Blimp1 function results in downregulated Tbx1 and Fgf8 expression in the surface epithelia leading to selective loss of the mesenchyme and NC that normally migrate into the caudal pharyngeal arches. The failure to elaborate the caudal pharyngeal arches disrupts development of the PA arteries. NC that invade the OFT are required for separation of the arteries and formation of the outflow valves (reviewed by Jiang et al.,2000). Abnormal morphogenesis of the OFT and ventricular septum could thus potentially reflect hypoplasia of the pharyngeal epithelia and concomitant loss of invading NC that normally migrate through the caudal pharyngeal arches and surround the outflow region(Jiang et al., 2000). Shh expression in the ventral pharyngeal endoderm is known to be required for OFT morphogenesis (Goddeeris et al., 2007), partly because of its ability to promote NC survival. However, loss of Shh in the ventral endoderm fails to disrupt pharyngeal arch development, and Fgf8 and Tbx1expression is unperturbed (Goddeeris et al., 2007). PTA has also been observed in Pitx2δc mutants in a context where NC migration occurs normally(Liu et al., 2002). To clarify the underlying cause of these OFT and PTA defects, future studies aim to selectively inactivate Blimp1 in either the pharyngeal epithelia or the SHF.

This cluster of pharyngeal and heart defects closely resembles those associated with DiGeorge syndrome (DGS), highly prevalent in the human population (reviewed by Lindsay,2001). In mice, Tbx1 haploinsufficiency causes aortic arch defects (Jerome and Papaioannou,2001; Lindsay et al.,2001; Merscher et al.,2001), and Tbx1 maps to the DGS interval del22q11 in humans. Manipulating Fgf8 and Tbx1 expression also leads to DGS phenotypes in mice (Brown et al., 2004; Macatee et al., 2003; Park et al., 2006; Zhang et al., 2006). It will be interesting to learn whether Blimp1 mutations are causally associated with DGS and congenital heart defects in the human population.

In zebrafish Blimp1 (also known as prdm1 - ZFIN) is expressed in the mesenchyme as well as the ectodermal cells of the early fin bud and later in the cells of the AER. In ubo mutants, loss of Blimp1 activity severely impairs pectoral fin outgrowth and patterning (Lee and Roy,2006), whereas Blimp1 morphant embryos entirely lack fin buds (Mercader et al., 2006). Blimp1 acts downstream of Shh signalling for specification of the slow-twitch muscle lineage(Baxendale et al., 2004), and in the neural crest, Blimp1 functions downstream of Bmpsignalling to promote formation of sensory neuron progenitors(Hernandez-Lagunas et al.,2005; Roy and Ng,2004). However, experiments to date strongly suggest that Blimp1 is not required in either muscle or neural crest lineages in mice. In light of these differences, we conclude that Blimp1 functions in the fin and limb are not conserved between fish and mammals.

Blimp1 is essential for specification of primordial germ cells(Ohinata et al., 2005; Surani et al., 2007). In the developing mouse embryo, Blimp1 induced in the proximal epiblast in response to BMP/Smad signals (Lawson et al., 1999; Tremblay et al.,2001; Chu et al.,2004; Arnold et al.,2006) represses the default pathway that would normally give rise to posterior mesodermal derivatives such as the allantois or the primitive streak. However, Blimp1 contributions to the germ cell lineage appear to be mammalian specific because Blimp1 is not expressed in zebrafish PGCs (Ng et al., 2006). Similarly, the present experiments establish that Blimp1 is required in trophoblast cells of the labyrinthine placenta, a uniquely mammalian cell lineage. Collectively these observations strongly suggest that diverse tissue-specific transcriptional networks controlled by Blimp1 as a master regulator have evolved independently in different organisms.

The tumour suppressor p53 regulates the expression of numerous target genes to protect cells against genotoxic stress. Recent studies have identified Blimp1 as a p53 target (Yan et al., 2007). The NC and mesenchyme of the prospective caudal arches in Blimp1 null embryos undergoes apoptosis(Vincent et al., 2005). However, the dying cells are not Blimp1⁺ and cell death most likely occurs as a secondary consequence of loss of Fgf8 expression in the pharyngeal epithelia (Macatee et al.,2003). Moreover, we found no evidence that apoptosis contributes to any of the tissue abnormalities described here. Interestingly, p63is strongly expressed in embryonic epithelia at sites undergoing reciprocal signalling. Strikingly, as for Blimp1 mutants, p63-deficient embryos display limb defects, pharyngeal arch hypoplasia, and lack epithelial structures including the vibrissae and hair(Mills et al., 1999; Yang et al., 1999). It will be interesting to learn whether p63 and Blimp1 cooperatively govern self-renewal of diverse progenitor cell populations in the pharyngeal surface epithelium.

Blimp1 selectively maintains the ZPA in the forelimb but not the hindlimb. Similarly, Blimp1 is required for induction of sensory vibrissae but not pelage hair follicles. One possibility is that yet-to-be-discovered, divergent signalling pathways may selectively control development of these morphologically similar structures. Alternatively,closely related zinc-finger PR/SET domain family members may selectively compensate for the loss of Blimp1 activity in the developing hindlimb and pelage hair follicles. More than 15 PRDM genes have been identified and several have been shown to control embryonic development(Hoyt et al., 1997) and/or tumour growth (Steele-Perkins et al.,2001). An important future goal is to test whether additional members of the PRDM family are coexpressed with Blimp1 in these unaffected tissues.

The present experiments demonstrate that homozygous Blimp1^gfp/gfp embryos carrying an IRESgfp reporter cassette, only weakly express Blimp1 but entirely lack PGCs and display fully penetrant pharyngeal and heart defects, whereas extra-embryonic tissues and sensory vibrissae develop normally. More pronounced defects were observed in transheterozygous Blimp1^gfp/- embryos, with further reduced expression levels. These results demonstrate graded Blimp1 activities in the developing mouse embryo and strongly suggest that moderate changes in Blimp1 expression levels may have a significant impact on target gene selection and/or Blimp1 functions as a scaffolding protein for recruitment of co-repressors.

We thank Kathryn Calame for the Blimp1 conditional mice, Azim Surani and Michel Nussenzweig for the Blimp1-Cre mice, Lynn Corcoran for Blimp1 antibody, John Fallon, Malcolm Logan, Peter Scambler and Rolf Zeller for probes, and Andreas Trumpp and Albert Basson for c-myc and Lysotracker Red staining protocols, respectively. This work was supported by a programme grant from the Wellcome Trust.