The zinc finger transcriptional repressor Blimp1/Prdm1 is dispensable for early axis formation but is required for specification of primordial germ cells in the mouse

Current models suggest that programmed cell growth and differentiation during development are controlled by master regulatory genes that coordinate the activities of cell-type-specific transcription factors. However,relatively few examples of such regulators with proven capabilities to govern global patterns of gene expression have been described in the literature. Blimp1 (Prdm1 - Mouse Genome Informatics), a Krüppel-type zinc-finger-containing DNA-binding transcriptional repressor originally cloned by its ability to bind the human β-IFN promoter(Keller and Maniatis, 1991),was also independently identified due to its cell-type-specific expression in murine plasmacytoma cell lines (Turner et al., 1994). Blimp1 expression is dramatically upregulated coincident with plasma cell differentiation, and enforced expression is sufficient to drive B cells to become antibody-secreting cells (reviewed by Calame et al., 2003). Recent studies have shown that Blimp1 plays an essential role in B-cell development (Shapiro-Shelef et al.,2003). Thus conditional inactivation of Blimp1 in B cells disrupts their terminal differentiation to plasma cells, and consequently serum Ig levels are markedly reduced. Blimp1 has been shown to repress expression of cell cycle regulators such as Myc, p18 and p21(Sciammas and Davis, 2004; Shaffer et al., 2002; Yu et al., 2000), and consequently Blimp1-deficient B cells display a hyperproliferative phenotype (Shapiro-Shelef et al.,2003). Blimp1 downregulates the expression of transcription factors such as CIITA and Pax5, which are essential for B-cell antigen presentation and receptor signaling(Lin et al., 2002; Piskurich et al., 2000). Blimp1 has also been implicated as a master regulator of myeloid cell terminal differentiation (Chang et al.,2000).

Considerable evidence suggests that Blimp1 acts to control anteroposterior axis formation and development of the head structures in early vertebrate embryos. In Xenopus and zebrafish, Blimp1transcripts localize to the signaling centers that pattern the forebrain,namely the leading anterior mesendoderm and prechordal plate(Baxendale et al., 2004; de Souza et al., 1999). Overexpression of Blimp1 in Xenopus embryos downregulates Xbra and Myf5 activities, which are required for formation of trunk mesoderm, and consequently causes axis truncations. Blimp1also activates anterior mesendodermal markers such as goosecoid and cerberus in animal cap explant assays(de Souza et al., 1999). Similarly, zebrafish embryos treated with Blimp1 antisense morpholino oligonucleotides show loss of anterior structures and display a foreshortened body axis (Baxendale et al.,2004). The analogous anterior signaling centers in mouse, namely the anterior visceral endoderm (AVE) and anterior definitive endoderm(ADE)/prechordal plate, have been shown to express Blimp1 transcripts(de Souza et al., 1999). Thus it has been widely assumed that Blimp1 also plays a crucial role in anterior patterning during mammalian development.

Blimp1 activity is also required for lineage specification in two additional tissues in zebrafish. A hypomorphic allele of the Blimp1ortholog u-boot (ubo) results in viable animals lacking slow-twitch muscle fibers (Baxendale et al., 2004; Roy et al.,2001). Molecular analysis reveals muscle precursors fail to differentiate into slow-twitch fibers in response to Hh. uboactivates expression of the slow myosin heavy chain, and represses expression of the fast myosin heavy chain. Interestingly, ubo also acts downstream of BMP signaling in specification of the trunk and cranial neural crest precursors (Roy and Ng,2004). Whether Blimp1 similarly controls muscle development and/or formation of neural crest derivatives in the mammalian embryo remains unknown.

Here, we describe for the first time the defects caused by Blimp1loss of function in the mouse. Blimp1-deficient embryos die at mid-gestation, but surprisingly early axis formation, anterior patterning and neural crest formation proceed normally. Rather loss of Blimp1expression disrupts morphogenesis of the caudal branchial arches and leads to a failure to correctly elaborate the labyrinthine layer of the placenta. Blimp1 mutant embryos also show widespread blood leakage and tissue apoptosis. Blimp1 transcripts are first detectable in the visceral endoderm overlying the proximal epiblast prior to gastrulation, and within the epiblast are restricted to the primordial germ cells (PGCs) at the time they first appear in the posterior allantois. Blimp1 expression persists as PGCs migrate anteriorly toward the genital ridges(Chang and Calame, 2002). Strikingly, we report here that Blimp1 homozygous mutants entirely lack PGCs. Quantitative analysis of PGC numbers at sequential embryonic stages reveals that this loss occurs at the time of their allocation around 7.25 days post coitum (dpc). Moreover Blimp1 heterozygous embryos exhibit a dose-dependent decrease in PGC numbers. The present experiments thus identify Blimp1 as the first transcriptional regulator required for establishment of the mammalian germ line.

We engineered two independent mutations at the Prdm1/Blimp1 locus. The homology arms were subcloned into the vector described previously that contains LoxP-flanked pgk-hygro, and hsv-tk and pgk-dtanegative-selection cassettes (Vincent et al., 2004). As shown in Fig. 2, the BAH construct contains 3.7-kb (XbaI-ApaI)5′ and 3.7-kb (PstI-MscI) 3′ homology arms, and deletes 3 kb spanning the promoter and exons 1 to 3, whereas the BEH vector,containing 4.8-kb (XbaI-EagI) 5′ and 3.7-kb(PstI-MscI) 3′ homology arms, creates a smaller deletion beginning upstream of the ATG in exon 1 and including exon 3. Linearized vectors were electroporated into CCE ES cells. Drug-resistant colonies were screened by Southern blot analysis using a 5′ external probe (SalI-SpeI). Both vectors gave a targeting frequency of approximately 3%. Candidate targeted clones were expanded and re-screened using a 3′ external probe (MscI-NcoI). For each construct, three correctly targeted ES clones were injected into C57BL/6J blastocysts to generate germ line chimeras. To excise the hygroselection cassette, prdm1^BAH and prdm1^BEH germ-line male chimeras were mated with Sox2Cre transgenic females that express Cre in the germ line(Vincent and Robertson, 2003),yielding the so-termed prdm1^BA and prdm1^BE derivative alleles, respectively. The removal of the selection cassette did not affect the phenotype.

The prdm1^BAH and prdm1^BEH mutant strains were maintained on a C57BL/6×129/Sv//Ev background, whereas prdm1^BA and prdm1^BE mutations were outcrossed to ICR. Primers ApaMUT- (5′-CGA AGT TAT GGA TCA TCA AGG-3′) and ApaCOM+ (5′-TGT ATG CCC CGT GTG TTT AG-3′) that amplify a 259-bp band were used to detect the prdm1^BA and prdm1^BAH alleles. The prdm1^BE and prdm1^BEH alleles were detected using primers EagMUT-(5′-ATT ATA CGA AGT TAT GGA TCA TC-3′) and EagCOM+ (5′-TTG GTA AAG AAG TCT GCG GC-3′) that amplify a 188-bp band. Primers EagCOM+and EagWT- (5′-TCG TAC CCA CAC GTT TTT CC-3′) were used to detect a 307-bp band corresponding to the wild-type allele. The prdm1^CD strain was maintained on an albino outbred ICR background and genotyped as described(Shapiro-Shelef et al., 2003). Embryos were individually genotyped prior to histological analysis,whole-mount in situ hybridization and primordial germ-cell staining.

Whole-mount in situ hybridization was performed according to standard protocols (Nagy et al., 2003). Two Blimp1/Prdm1 probes were used. The first was derived from the EST mB3 (IMAGE clone 1165721) specific for the 3′UTR, and the second, generated by PCR, was specific for exons 4 and 5. No expression pattern differences were observed between these two probes. Fgf8, Crabp1,Shh, Hoxb1 and Evx1 probes have been described previously(Dolle et al., 1989; Vincent et al., 2003). India ink was injected into the outflow tract of freshly isolated 9.5 dpc embryos to visualize the major arteries. For histological analysis, embryos fixed in 4%paraformaldehyde, were dehydrated through an ethanol series, and embedded in wax before sectioning. Hematoxylin and Eosin staining was performed according to standard protocols.

Embryos were dissected from the decidua and Reichert's membrane. Pieces of the extraembryonic tissue (7.75-8.5 dpc embryos) or yolk sac (9.5 dpc embryos)were kept for genotyping. Alkaline phosphatase staining was performed as described previously (Labosky and Hogan,1999; Lawson et al.,1999). SSEA1 immunostaining was performed as previously described(Gomperts et al., 1994).

Regression analysis and comparison of regression lines were performed using the F test to compare variances (Snedecor and Cochran, 1967).

Considerable evidence suggests that Blimp1 plays a crucial role as a master regulator during axis formation and anterior patterning in vertebrate embryos (de Souza et al.,1999). Consistent with this suggestion, in mouse embryos Blimp1 transcripts are strongly expressed in the anterior region in the prechordal plate, as well as in the anterior definitive endoderm(de Souza et al., 1999). Immunohistochemical experiments have also described expression at later stages in the primordial germ cells and gut endoderm(Chang and Calame, 2002). To assess Blimp1 expression from pre-gastrulation stages onwards, we performed WISH analysis. Blimp1 transcripts are initially detected at day 6.0 post coitum (dpc) in the visceral endoderm overlying the proximal epiblast (Fig. 1A), and by 7.5 dpc are confined to the anterior mesendoderm(Fig. 1D,E)(de Souza et al., 1999) and the PGCs (Fig. 1B-F). We have previously shown that during gastrulation Nodal signals are required to specify derivatives of the anterior primitive streak, including the anterior mesendoderm (Vincent et al.,2003), raising the interesting possibility that Blimp1might potentially function as a transcriptional regulator downstream of the Nodal signaling pathway.

To learn more about possibly conserved Blimp1 functions in axis formation and anterior patterning, we generated two independent mutations designated prdm1^BEH and prdm1^BAH. Both alleles delete 5′ regions of the Blimp1 locus, including the first three coding exons (see Materials and methods; Fig. 2). Heterozygous mice carrying these mutations are healthy and fertile. However, intercross matings produced no live born homozygous offspring (see Table 1). Mutant embryos were recovered at the expected Mendelian ratio up to 9.5 dpc, but became hemorrhagic and were dying one day later. Homozygous mutant embryos failed to survive beyond 10.5 dpc.

The prdm1^CD mutation eliminates exons 6-8, encoding the five zinc fingers crucial for DNA binding, and disrupts Blimp1 protein expression as assessed by western blot analysis(Shapiro-Shelef et al., 2003). To compare the mutant phenotypes, prdm1^CD heterozygotes were intercrossed, and embryos collected between 9.5 and 11.5 dpc. We observed exactly the same range of tissue defects as described in detail below. Finally, when mice carrying the prdm1^CD and prdm1^BEH alleles were intercrossed, we found that the trans-heterozygotes were indistinguishable from prdm1^BEH, prdm1^BAH and prdm1^CD homozygous embryos. These results strongly suggest that our prmd1^BEHand prdm1^BAH targeted alleles create loss-of-function mutations.

Blimp1 is expressed in the anterior definitive endoderm, with higher levels detected in the lateral region that subsequently gives rise to the branchial arches (Fig. 3A,B). In mammalian embryos, the six pairs of branchial arches develop in a cranial to caudal sequence. In the mouse, the first branchial arch, the precursor of the maxilla and mandible, becomes morphologically apparent at 8.5 dpc, with the second and third arches emerging over the course of the next 24 hours. At 9.5 dpc, Blimp1 transcripts are restricted to the endodermal layer of the proximal-most region, or cleft, of the first branchial arch (Fig. 3C,D),whereas, by contrast, expression expands into the endoderm, ectoderm and mesenchyme of the more caudally situated second and third arches(Fig. 3C,E). Interestingly Blimp1 expression is transient, and from 10.5 dpc onwards is barely detectable in pharyngeal tissues (Fig. 3F,G).

Blimp1-deficient embryos are readily identified at 9.5 dpc by their relatively small size and the presence of a unique first branchial arch rather than the three arches observed in wild-type littermates (Fig. 3H, compare with 3K). One of the main components of branchial arch mesenchyme is cranial neural crest. In zebrafish, ubo is expressed at the boundary of the neural plate and the non-neural ectoderm at the site where neural crest cells later originate, and loss of ubo activity impairs specification of the neural crest cells(Roy and Ng, 2004). In the mouse embryo, Blimp1 transcripts are not detected in the neural plate nor in migrating neural crest cells (Fig. 3, data not shown), and to evaluate the possibility that the branchial arch defects are caused by aberrant neural crest formation, we assessed expression of the neural crest marker Crabp1(Dolle et al., 1989). At 9.0 dpc, mutant and control embryos show no differences in Crabp1expression (data not shown). Thus we conclude that loss of Blimp1fails to perturb specification and migration of the neural crest cells. Rather, histological analysis reveals that the defect is caused by selective cell death of mesenchyme cells specifically at the site where the second and third arches normally emerge within the strongest Blimp1 expression domain (Fig. 3K, part 2). The presence of apoptotic cells was confirmed by TUNEL analysis (data not shown). In addition, Fgf8 transcripts, normally co-expressed with Blimp1 in the endoderm and ectoderm of the emerging caudal second and third arches, are significantly reduced(Fig. 3I-M). By contrast, Fgf8 expression in the cleft of the first arch is robustly maintained(Fig. 3J,M).

Blimp1 shows a dynamic pattern of expression from 9.5 dpc onwards. Interestingly, Blimp1 is expressed at the site where myogenesis is initiated in the myotome of the most rostral somites (reviewed by Tajbakhsh and Buckingham,2000) (Fig. 4A). As the more caudal somites mature, Blimp1 expression is detected in the myotomal fibers along the rostrocaudal axis(Fig. 4B,C), and by 11.5 dpc becomes restricted to the myotome of the most caudal somites(Fig. 4D). Expression is confined to the intercalated mytome, located between the most epaxial and hypaxial domains, where muscular differentiation is initiated(Cheng et al., 2004; Spörle, 2001). Blimp1-deficient embryos die at 10.5 dpc, precluding our ability to investigate its functional role in the myogenic lineage. However, the expanded Fgf8 domain in the epaxial dermomyotomal lip of Blimp1-deficient embryos suggests a potential role for Blimp1 in regulation of the myogenic program in mouse(Fig. 3J,M).

Blimp1 transcripts are present throughout the limb bud as it emerges from the flank (Fig. 4A). By E10.5, expression shifts to demarcate the posterior region of the bud (Fig. 4D),encompassing the zone of polarizing activity (ZPA). Interestingly at 10.5 dpc, Blimp1 expression is also restricted to the distal part of the bud,in a domain that includes the apical ectodermal ridge AER(Fig. 4B,D, and data not shown). These signaling centers control anteroposterior and proximodistal patterning of the growing limb bud, respectively (reviewed by Tickle, 2003). However AER and ZPA patterning proceeds normally in Blimp1 mutant embryos. Moreover Fgf8 expression in the AER is unaffected(Fig. 3I-M), strongly suggesting that Blimp1 does not lie upstream of Fgf8. Similarly, the ZPA marker Shh shows delayed onset of expression,which is likely to be due to growth retardation of the mutant embryos, but by 10.5 dpc anteroposterior polarity is correctly established, as judged by induction of Shh expression (data not shown). The otic vesicle, a tissue showing very strong Blimp1 expression(Fig. 3C,F) also forms normally(Fig. 3K, part 2).

The major vascular structures of the early embryo, including the dorsal aorta and the yolk sac vasculature form normally in Blimp1 mutant embryos (Fig. 4). By 9.5 dpc,the heart tube displays correct patterning, and has undergone normal looping morphogenesis (Fig. 4G, compare with 4F) and contracts rhythmically. Histological analysis shows appropriate formation of the myocardial and endocardial cell populations. However, one day later, mutant embryos show widespread apoptosis and become severely hemorrhagic, with blood pooling in the heart, dorsal aorta and under the surface ectoderm(Fig. 5G). We wondered whether the reduced growth rate and mid-gestation lethality might be caused by a placental defect.

Formation of the placenta is a complex process involving sequential interactions of the allantois and the chorionic plate with the trophoblast derivatives (reviewed by Rossant and Cross, 2002). Around 10.5 dpc, embryonic blood vessels arising from the allantois undergo extensive villus branching within the placenta to create a densely packed network of blood-filled spaces termed the labyrinth,which functions as an exchange zone between maternal and embryonic circulation. Histological analysis shows that these specialized cell types initially develop normally (Fig. 5B,D). However, the labyrinthine region fails to expand correctly,thereby greatly reducing the volume of the blood spaces (Fig. 5D,H; compare with 5B,F, respectively). Whole-mount in situ hybridization experiments failed to demonstrate Blimp1 transcripts in the extra-embryonic structures that contribute to the placenta, excluding a cell-autonomous defect.

The major blood vessels did not display any obvious abnormalities prior to 10.5 dpc (Fig. 4). At 10.5 and 11.5 dpc, Blimp1 transcripts are detectable in the developing intersegmentary arteries and the endothelial cells of the superficial capillary plexus (Fig. 4C,E). PECAM staining of endothelial cells throughout the circulatory network of mutant embryos was unperturbed as assessed by whole mount immunohistochemistry(data not shown). Thus it appears that endothelial cells initially form in Blimp1 mutants, but that the rapid expansion of this cell population necessary to support embryonic growth is compromised. Alternatively, the loss of the caudal branchial arches, which prevents formation of the second and third pharyngeal arteries (Fig. 5C), may secondarily disrupt blood circulation. Conditional deletion of Blimp1 within the endothelial cell lineages is required to distinguish these possibilities.

Around 7.0 dpc, Blimp1 expression identifies committed PGCs as they first appear at the base of the incipient allantois(Fig. 1B). Expression persists in PGCs as they proliferate and migrate anteriorly along the hindgut toward the genital ridges (Chang and Calame,2002). To evaluate whether Blimp1 is required for specification and/or maintenance of the mammalian germ line, we compared PGC numbers in wild-type, heterozygous and homozygous mutant embryos. PGCs are easily identified due to their high endogenous levels of alkaline phosphatase(AP) activity (Ginsburg et al.,1990; Lawson and Hage,1994). Intercross embryos were collected around 9.0-9.5 dpc. In Blimp1 null mutant embryos (n=12), we failed to observe AP-positive PGCs, and the number of PGCs in similarly staged heterozygous embryos (n=58) was significantly reduced, when compared with wild type (n=34) (Fig. 6B). These results were confirmed by immunohistochemical staining with the SSEA1 monoclonal antibody (Gomperts et al.,1994). We failed to detect any migrating PGCs within the hindgut epithelium of Blimp1 mutant embryos, and there were reduced numbers of SSEA1-positive PGCs in heterozygotes, when compared with wild type(Fig. 6C).

To determine the onset of the PGC phenotype, embryos were collected and stained for AP activity between 8.0 and 9.5 dpc. The number of PGCs was plotted against somite number and the regression line of the log PGC number calculated. Extrapolation of the regression lines gives a founding population of 40 PGCs in wild type (Fig. 7), consistent with previous estimates of approximately 45 founder cells (Lawson and Hage, 1994). By contrast, a founding population of nine was observed for Blimp1heterozygotes (a reduction of 78%), revealing strong gene-dosage sensitivity. The slopes of the regression lines are not significantly different, indicating that both wild type and Blimp1 heterozygous PGCs expand at similar rates. Furthermore, we observed no differences in the regional distribution of PGCs along the hindgut epithelium (Fig. 5B). Thus Blimp1 is required for initial PGC specification.

To strengthen this interpretation, we evaluated PGC numbers soon after they first appear (Downs and Davies,1993; Lawson and Hage,1994). At the headfold stage (7.75 dpc), PGCs begin to disperse from the base of the allantois and intercalate into the forming hindgut. We observed reduced numbers of PGCs in Blimp1 mutant heterozygotes, when compared with wild type (Fig. 6A), whereas in homozygous mutants (n=6) fewer than two to four AP-positive cells were present, or PGCs were undetectable(Fig. 6A). AP-positive cells thus fail to survive (Fig. 6B,C). We conclude that Blimp1 functions to promote the formation of nascent PGCs.

Here we demonstrate, contrary to expectations from studies in Xenopus and zebrafish, that Blimp1 is not required for establishment of the anteroposterior axis in early mouse embryos. These results are especially surprising because Blimp1 is expressed at the right time and place to influence formation of head structures, namely in axial mesendoderm, the tissue known to provide anterior patterning signals during gastrulation. Blimp1 is not expressed specifically in the AVE but rather is induced in a ring of visceral endoderm at the embryonic/extra-embryonic junction. Loss of highly transient Blimp1expression in the visceral endoderm has no effect on early post-implantation development. Blimp1 transcripts are broadly and dynamically expressed at later stages, but surprisingly mutant embryos display quite restricted tissue defects.

The identical phenotype results from two independent targeting strategies that deleted different regions of the locus. Heart development per se proceeds normally, but at around 10.5 dpc, we observe blood leakage from the superficial capillaries, and accumulation of blood within the heart and dorsal aorta. Mutations causing hemorrhaging phenotypes have been extensively described in the various signaling pathways that control endothelial cell function and formation (reviewed by Sato and Loughna, 2002). Our experiments demonstrate that Blimp1 is expressed in the endothelium and intersegmental arteries. Specification of endothelial cells proceeds normally, as assessed by histological examination and PECAM expression, suggesting that Blimp1is not required during formation of the vasculature but probably acts later to regulate angiogenesis and/or vascular remodeling. It will be interesting to test this possibility in conditional mutants selectively lacking Blimp1 in the endothelium.

The growth retardation and rapid death of the Blimp1-deficient embryos is most likely due to placental insufficiency. All of the specialized placental cell types are correctly specified in Blimp1 mutant embryos. However the labyrinth region, the site of major maternal/fetal exchange, is poorly developed at 9.5 dpc, and by 10.5 dpc, mutant placentas display striking morphogenetic defects coincident with rapid growth of the embryo. At this stage the essential nutrient and oxygen exchange functions provided by the visceral yolk sac normally shift to the mature placenta. Blimp1 transcripts are undetectable in the placenta, suggesting the labyrinth defects arise as a secondary consequence of abnormalities restricted to the embryonic vasculature. The importance of blood flow in shaping the vasculature is well documented. Reduced blood flow due to hemorrhage thus potentially disrupts the mechanical signals necessary for villous branching in the placental labyrinth.

Organogenesis and tissue patterning are largely unaffected. However, Blimp1-deficient embryos display striking branchial arch defects. The branchial arches form from four distinct cell subpopulations. The core,comprising lateral mesenchymal cells and neural crest cells, interacts with the adjacent endoderm and surface ectoderm tissues. Blimp1 is expressed in the dorsal anterior regions of the emerging branchial arches,specifically in the pharyngeal endoderm. In the first arch, Blimp1transcripts are only present in the endoderm, whereas in the second and third arch expression is more widespread, extending into the mesenchyme. Formation of the second and third arches correctly initiates around 9.5 dpc, but in Blimp1 mutant embryos these structures deteriorate shortly thereafter. Reduced Blimp1 activity in the hypomorphic zebrafish mutant ubo similarly leads to pectoral fin defects, implicating an evolutionarily conserved role for Blimp1 in patterning branchial arch derivatives (van Eeden et al.,1996).

In fish, BMP signals induce Blimp1 expression at the boundary between neural and non-neural ectoderm, which in turn promotes formation of neural crest and sensory neuron progenitors(Roy and Ng, 2004). However,the caudal branchial arch defects described here cannot be explained as a deficiency in neural crest because Crabp1 expression in Blimp1 mutant embryos is unperturbed. Thus, neural crest cells are correctly specified and efficiently migrate into the forming arches. Rather,branchial arch defects closely resemble those caused by reduced Fgf signaling. Fgf8 is expressed in both the branchial arch ectoderm and endoderm,and is required for survival of the neural crest. Localized cell death has been observed in the branchial arches of Fgf8 hypomorphs and conditional mutants (Abu-Issa et al.,2002; Frank et al.,2002; Macatee et al.,2003; Trumpp et al.,1999). Blimp1 mutants similarly display numerous apoptotic cells with pyknotic nuclei in caudal branchial arches. These results suggest that Blimp1 positively regulates Fgf8 in the pharyngeal endoderm. The pharyngeal endoderm is thought to play a crucial role in coordinating branchial arch outgrowth (reviewed by Graham, 2003; Graham et al., 2004). Cell death in the arches probably compromises continued arch outgrowth, and disrupts the survival or mitogenic cues necessary to create a permissive environment for incoming neural crest and paraxial mesenchyme.

Interestingly these defects are restricted to branchial arches 2 and 3,whereas the first arch develops normally, a phenotype reminiscent of the Raldh2 mutants and those rescued by maternal retinoic acid (RA)supplementation (Niederreither et al.,1999; Niederreither et al.,2003). Expression of the RA-responsive genes Hoxa1 and Hoxb1, normally present in the foregut endoderm, is entirely absent in Raldh2 mutants, and these transcripts are expressed at reduced levels in the rescued embryos. Interestingly, the first arch does not express either of these Hox genes. By contrast, formation of the posterior arches is sensitive to fluctuating Hox levels (Rijli et al., 1993; Rossel and Capecchi, 1999). Here, we observe that Blimp1 is co-expressed with Hoxa1/Hoxb1 in the foregut endoderm. An attractive idea is that the focal defects in arch formation reflect selective Blimp1 requirements upstream of Hox gene expression. It will be interesting to learn more about Blimp1 targets in the branchial arches and the extent to which these pathways overlap with those governing terminal B cell differentiation and germ cell specification.

Here, we demonstrate for the first time that the zinc-finger transcriptional repressor Blimp1 is required for specification of the mammalian germ line. Blimp1 transcripts appear to tightly localize to PGCs, coincident with their emergence in the extra-embryonic mesoderm. Blimp1 is initially expressed in a subpopulation of VE overlying the proximal region of the epiblast adjacent to the distal extra-embryonic ectoderm. It is well known that Bmp4 plays an essential role during germ-cell specification. Bmp4 signals from the extra-embryonic ectoderm(Lawson et al., 1999),together with Alk2, which is expressed in the VE(de Sousa Lopes et al., 2004),and the intracellular effectors Smad1, Smad5 and Smad4(Chang and Matzuk, 2001; Chu et al., 2004; Tremblay et al., 2001) are required to promote both germ cell and posterior mesoderm/allantoic fates in a dose-dependent fashion. As for neural crest progenitors in zebrafish(Roy and Ng, 2004), we suggest that the Bmp4 pathway also governs Blimp1 induction during germ-cell specification. Relatively little is known about transcriptional control at the Blimp1 locus, but we speculate that Blimp1 expression in PGCs in part depends on Bmp4 signals acting upstream of Smad effector proteins. Consistent with this idea Smad1, Smad5 and Smad4 mutant embryos display germ-cell defects (Chang and Matzuk, 2001; Chu et al.,2004; Tremblay et al.,2001).

A recent screen identified two candidate genes potentially involved in the establishment of germ-cell competence in the proximal epiblast(Saitou et al., 2002). Fragilis/Ifitm3, an interferon-inducible transmembrane protein, is broadly induced in proximal epiblast cells adjacent to the extra-embryonic ectoderm(Tanaka and Matsui, 2002; Tanaka et al., 2004), and is activated in response to Bmp4 signaling in a dose-dependent fashion. Expression rapidly resolves to the proximal posterior side of the embryo, and becomes confined to the PGCs, but its functional role in germ-cell specification has yet to be characterized. The gene stella(Dppa3 - Mouse Genome Informatics), encodes a novel intracellular protein, activated later within the PGCs, that is nonessential for germ-cell specification (Payer et al.,2003; Bortvin et al.,2004). It has been proposed that genes such as fragilis and stella act to turn off genes that promote somatic cell fates, including transcription factors such as Hoxb1 and Evx1, and thus create a niche for PGC specification (Saitou et al.,2002).

The present data suggests that Blimp1 turns off the default pathway that allows epiblast cells to adopt a somatic cell fate and shifts the transcriptional profile so that Blimp1-expressing cells become exclusively allocated into the germ-cell lineage. Germ-cell specification involves the repression of region-specific homeobox genes such as Hoxa1/Hoxb1, Lim1 and Evx1 in adjacent somatic cells. Evx1 and Hoxb1 expression domains were not expanded in Blimp1-deficient embryos (data not shown), but this might reflect our inability to assess expression at single-cell resolution via WISH. Interestingly, we observe pronounced dose-dependent Blimp1requirements, which suggests that a relatively high threshold of Blimp1 activity is required for germ-cell differentiation and survival. An important future goal is the identification of Blimp1targets downstream in the molecular program responsible for specification of germ cells.

In Xenopus, Blimp1 activates Cerberus expression and antagonizes TGFβ and Wnt signals (de Souza et al., 1999). However TGFβ antagonists play more a subtle role in axis patterning in mouse embryos. Indeed, mutants lacking cerberus (Cer1 - Mouse Genome Informatics) are viable, most likely as a result of redundancy in these signaling pathways. Cerberus activities in mice were only uncovered in double-mutant embryos also lacking Lefty1, a second TGFβ family antagonist(Perea-Gomez et al., 2002). Cerberus is thus also a candidate Blimp1 target in early mouse embryos.

Blimp1 contains five consecutive Krüppel-type zinc fingers in the C-terminal region, and a positive-regulatory (PR) domain at the N terminus that is necessary for chromatin remodeling. The middle proline-rich region recruits members of the Groucho family of co-repressors and members of the histone deacetylase families (Ren et al.,1999; Yu et al.,2000). These various functional modules coordinate the ability of Blimp1 to function as a master regulator of gene expression. Considerable evidence suggests that Blimp1 acts as a DNA-binding scaffold protein capable of directly recruiting multiple chromatin-modifying enzymes to targeted promoters. The molecular mechanisms underlying Blimp1 activities during plasma cell differentiation have been extensively characterized. As a master regulator of the B-cell lineage, Blimp1 functions upstream of transcription factors such as CTIIA and Pax5, which govern antigen presentation and receptor signaling(Lin et al., 2002; Piskurich et al., 2000). Blimp1 also represses genes, such as Myc, which are involved in cell cycle progression (Lin et al.,1997). In contrast, Blimp1 activates expression of XBP, a transcription factor in the unfolded-protein-response pathway that is required for plasma cell differentiation (Lin et al., 2002; Shaffer et al.,2002; Shapiro-Shelef et al.,2003). The present study demonstrates that Blimp1 is an essential regulator of branchial arch formation and germ cell specification. Future experiments aim to characterize how Blimp1 promotes programmed differentiation and/or survival of these distinct cell types in the developing mammalian embryo.

We thank Debbie Pelusi for help with genotyping. S.D.V. was supported by a Long Term Fellowship from the Human Frontier Science Program Organization(HFSPO). N.R.D. was supported by a postdoctoral fellowship from the NICHD. This work was supported by grants from the NIH and the Wellcome Trust to E.J.R.