Essential roles of a zebrafish prdm1/blimp1 homolog in embryo patterning and organogenesis

Vertebrate development entails a series of cell fate specification and morphogenetic events regulated by networks of evolutionarily conserved transcription factors and secreted molecules. In zebrafish, maternally deposited dorsal determinants are translocated by microtubule-dependent mechanisms to the future dorsal side of the embryo, leading to nuclear accumulation of β-catenin in the dorsalmost extra-embryonic yolk syncytial layer (YSL) and the overlying dorsal blastomeres. At the onset of zygotic transcription, β-catenin activates expression of bozozok(boz) and squint (sqt), encoding a transcriptional repressor and the Nodal-related 1 secreted ligand, respectively, both of which play independent and cooperative roles during formation of the gastrula organizer (Shimizu et al.,2000; Sirotkin et al.,2000). This signaling center, conserved in all vertebrate gastrulae, induces neuroectoderm, patterns the germ layers, and orchestrates gastrulation movements (Spemann and Mangold, 2001).

A wealth of data support the notion that the activity of the gastrula organizer is highly dynamic, but our understanding of the underlying genetic hierarchies remains incomplete (Spemann and Mangold, 2001) (reviewed by Hibi et al., 2002). In zebrafish embryos, many organizer-specific genes, such as goosecoid(gsc), are expressed in the dorsal blastomeres exclusively, whereas others, for example boz, are initially expressed in dorsal blastomeres, somewhat later in both dorsal blastomeres and the dorsal YSL, and finally exclusively in the dorsal YSL(Stachel et al., 1993; Yamanaka et al., 1998; Koos and Ho, 1998). The contribution of the dorsal YSL to the specification and function of the gastrula organizer remains puzzling. Transplantation studies demonstrate that the yolk cell, including the YSL, is capable of inducing organizer gene expression in host blastomeres in a non-cell-autonomous fashion(Mizuno et al., 1999). Furthermore, function of boz only in the YSL is sufficient for normal development (Fekany et al.,1999). However, experimental degradation of RNA in the YSL does not prevent organizer formation (Chen and Kimelman, 2000). Therefore, the YSL and the dorsal blastoderm probably function in a redundant fashion to mediate Spemann-Mangold organizer formation and function.

The vertebrate gastrula organizer contributes to the elaboration of the embryonic pattern by secreting factors, including Chordin, Noggin and Cerberus, that antagonize the ventrally expressed TGF-β superfamily cytokines, the BMPs. The resulting BMP activity gradient is thought to specify a dorsoventral and anteroposterior progression of cell fates, with head and dorsal midline structures developing at the lowest levels of BMP activity,trunk structures at intermediate levels, and ventroposterior structures being specified by the highest BMP activity levels (reviewed by Hammerschmidt and Mullins,2002; De Robertis et al.,2000). Therefore, understanding of the genetic hierarchy that regulates BMP activity during vertebrate gastrulation is of particular importance. In mouse embryos homozygous for mutations that inactivate both chordin and noggin, axis formation proceeds rather normally,except that forebrain and notochord are missing(Bachiller et al., 2000),suggesting that yet additional factors operate to limit BMP activity in murine embryos. In zebrafish, boz and chordin (chd) act in partially overlapping pathways to limit BMP activity during head and trunk development (Gonzalez et al.,2000). Whereas chd mutant embryos exhibit mild ventralization and boz embryos lack dorsal structures such as forebrain and notochord (Schulte-Merker et al., 1997; Fekany et al.,1999), boz chd double-mutant embryos exhibit synergistic loss of head and trunk and have an enlarged tail, largely due to excess BMP activity (Gonzalez et al.,2000). Expression of chd is strongly reduced in young boz gastrulae but, at late gastrulation, returns almost to normal levels (Fekany-Lee et al.,2000). Additionally, chd expression is only mildly affected in ndr1/sqt mutant embryos(Dougan et al., 2003). Thus,it is likely that several genes play redundant roles in regulating chd gene expression in zebrafish gastrulae.

The Krüppel-type zinc-finger gene Prdm1/Blimp1 (PR domain containing 1, with ZNF domain; B lymphocyte-induced maturation protein 1), a transcriptional repressor known to promote terminal differentiation of B lymphocytes and macrophages in the mouse embryo(Turner et al., 1994; Chang et al., 2000), is expressed in multiple tissues during early embryogenesis in Xenopus laevis, mouse and chick (de Souza et al., 1999; Chang et al.,2002; Ha and Riddle,2003). Notably, Prdm1/Blimp1 is expressed in the anterior endomesoderm in X. laevis and the visceral endoderm in mice,tissues hypothesized to possess similar activities as the YSL of the zebrafish embryo supported by expression of orthologous genes such as hematopoietically expressed homeobox (hhex) and LIM homeobox 1a (lhx1a) (reviewed by Sakaguchi et al., 2002). Due to early embryonic lethality of mice with inactivated Prdm1/Blimp1, functional studies have so far focused on its involvement in B lymphocyte development. Conditional targeting of prdm1 shows that mature B cells lacking Prdm1/Blimp1 activity cannot differentiate into immunoglobulin-secreting plasma cells (Turner et al.,1994; Shapiro-Shelef et al.,2003). In X. laevis, Prdm1/Blimp1 is hypothesized to induce anterior endomesoderm and promote head formation by positively regulating expression of cerberus in cooperation with Chordin and Mix.1 (de Souza et al.,1999).

Recently, zebrafish prdm1 was shown to be affected by a hypomorphic mutation known as u boot^tp39 (ubo)and to be required for slow muscle fiber differentiation(Roy et al., 2001; Baxendale et al., 2004). Whereas prdm1 expression is dynamic, analyses of the loss-of-function phenotype have focused on a limited set of developmental events(Baxendale et al., 2004). Thus our understanding of zebrafish Prdm1 during embryogenesis remains incomplete.

Here we report that zebrafish prdm1 exhibits a dynamic expression pattern shared by the murine, chick and X. laevis Prdm1/Blimp1 encoding genes. Misexpression of prdm1 inhibits the formation of dorsoanterior structures and reduces expression of the BMP antagonist Chordin. Conversely, interference with Prdm1 translation using morpholino oligonucleotides (MOs) increases chordin expression and dorsalizes the embryo. Our studies propose a role for Prdm1 in limiting the function of the gastrula organizer and show that its activity is essential for many cell fate specification and morphogenetic processes in precise correspondence with the intricate expression pattern of this transcription factor.

Fish (Danio rerio) were maintained as described(Solnica-Krezel et al., 1994). Embryos were staged according to Kimmel et al.(Kimmel et al., 1995). The following mutant alleles were used: boz^m168(Solnica-Krezel et al., 1996; Fekany et al., 1999), din^tt250(Hammerschmidt et al., 1996; Schulte-Merker et al., 1997), oep^tz57 (Gritsman et al., 1999), ogo^tm305(Hammerschmidt et al., 1996), swr^tc300a (Mullins et al., 1996; Kishimoto et al.,1997), smu^b641(Barresi et al., 2000), cas^ta56 (Alexander et al., 1999b).

An expressed sequence tag (EST) with sequence similarity to reported Prdm1/Blimp1 proteins was used to isolate a zebrafish homolog by 3′/5′ SMART RACE (BD Biosciences) from a 6.5-hour postfertilization (hpf) cDNA preparation. Full-length prdm1 was obtained by proof-reading PCR (PfuUltra, Stratagene), cloned into the EcoRV site of the pGEM-T Easy vector (Promega) and used for antisense probe synthesis with SP6 RNA polymerase after ApaI linearization. For misexpression, the full-length cDNA was cloned into the pCS2+ vector(Rupp et al., 1994) and used for capped RNA synthesis with SP6 RNA polymerase after NotI linearization (mMESSAGE mMACHINE, Ambion). The NCBI accession number is AY841759.

Single and double color whole-mount in situ hybridization was performed essentially as described (Thisse et al.,1993); BM Purple (Roche) was used as blue and Magenta phosphate(175 μg/ml in DMF; Fluka) combined with Tetrazolium Red (350 μg/ml in 70% DMF; Sigma) as red alkaline phosphatase substrates. The following molecular markers were used: gsc(Stachel et al., 1993), dlx3b (Akimenko et al.,1994), six3b(Kobayashi et al., 1998; Seo et al., 1998a), pax2a (Krauss et al.,1991), egr2b/krox20(Oxtoby and Jowett, 1993), hgg1/ctslb (Vogel and Gerster, 1997), boz(Koos and Ho, 1998; Yamanaka et al., 1998), chd (Schulte-Merker et al.,1997; Miller-Bertoglio et al.,1997), szl/ogo(Martyn and Schulte-Merker,2003; Yabe et al.,2003), bmp4 (Chin et al., 1997; Martínez-Barberá et al.,1997), ntl(Schulte-Merker et al., 1992), wnt8a (Kelly et al.,1995), mix (Kikuchi et al., 2000), sox17(Alexander and Stainier, 1999), ptc1 (Concordet et al.,1996), vox, vent(Melby et al., 2000; Kawahara et al., 2000a; Kawahara et al., 2000b), six7 (Seo et al.,1998b), dkk1(Hashimoto et al., 2000), nog1 (Fürthauer et al.,1999), hhex (Ho et al., 1999), ndr1(Rebagliati et al., 1998), ndr2 (Erter et al.,1998), dlc (Smithers et al., 2000), gata1(Detrich et al., 1995), eve1 (Joly et al.,1993).

Embryos were microinjected at the 1-cell stage using 50, 100 or 200 pg doses of synthetic prdm1 capped RNA(Marlow et al., 1998) or 1, 2 or 4 ng doses of a prdm1-specific morpholino oligonucleotide(MO^prdm1, 5′-TGTGTGATCTCTCCCCTGAGTGTGT-3′;GeneTools, LLC) (Nasevicius and Ekker,2000). A mutant form of prdm1(prdm1^mut) predicted to be unable to bind MO^prdm1 was constructed by site-directed mutagenesis(5′-AC(A/t)CACT(C/a)AGG(G/t)(G/t)AGAGATCA(C/t) ACA-3′) and used to evaluate the specificity of the MO^prdm1-induced phenotype. In co-injection experiments each reagent was microinjected independently at the 1-cell stage. Injection and photoactivation of anionic dextran DMNB caged fluorescein (Molecular Probes, D-3310) was performed as described(Sepich et al., 2000).

Embryos processed for whole-mount in situ hybridization were mounted in 80%glycerol/PBT and photographed using a Zeiss Axiophot compound microscope and an Axiocam digital camera. Live embryos were anesthetized if needed and mounted in 1.5% or 2.5% methylcellulose.

In order to investigate further the role of Prdm1/Blimp1 during early vertebrate development we searched the NCBI-database and identified a single zebrafish EST with sequence similarity to vertebrate Prdm1/Blimp1genes (de Souza et al., 1999; Turner et al., 1994; Ha and Riddle, 2003; Keller and Maniatis, 1991). The full-length cDNA sequence was obtained from a gastrula stage cDNA pool and designated prdm1. The transcript spans a total of 2522 bp and encodes a polypeptide of 776 amino acids (∼87.7 kDa). Comparative sequence analysis revealed high similarity to human, mouse, chick and X. laevis Prdm1/Blimp1 proteins in the N-terminal SET domain predicted to harbor histone methyltransferase activity(Tschiersch et al., 1994), as well as in the C-terminal DNA-binding domain, which encompasses five Krüppel-type (C₂H₂) zinc-fingers(Fig. 1). The overall sequence identity to vertebrate Prdm1/Blimp1 proteins is 55-56% (67-68% similarity),whereas within the zinc-finger motif sequence identity is 88-89% (91-94%similarity).

Expression of prdm1 was first detected by whole-mount in situ hybridization in the external YSL at the end of the blastula period (5.5 hpf; Fig. 2A). At the onset of gastrulation (6 hpf), prdm1 expression in the external YSL became more prominent, and shortly thereafter prdm1 transcripts appeared at low levels in the anteriormost part of the prechordal mesoderm(Fig. 2B). By contrast, gsc is expressed in the entire prechordal mesoderm and part of the chordamesoderm throughout gastrulation(Stachel et al., 1993)(Fig. 2F,G). The abundance of prdm1 transcripts in the prechordal mesoderm increased during gastrulation and expression extended more posteriorly compared with the hgg1 expression domain (Fig. 2C,D,H,I). At mid-gastrulation, prdm1 transcripts were identified in the non-neural ventral ectoderm (not shown) and gradually confined to a line of cells marking the boundary between neural and non-neural ectoderm during late gastrula stages (Fig. 2C,D) following the dynamic pattern described for dlx3b(Fig. 2H,I). Furthermore, two stripes of slow muscle precursor cells adjacent to the notochord began to express prdm1 (Fig. 2D). Throughout segmentation, slow muscle expression of prdm1 decreased gradually from anterior to posterior somites but was maintained in the posterior and thus youngest somites(Fig. 2E,J,K). In addition, prdm1 was expressed in a variety of tissue precursors, including the otic vesicle, the branchial arches and unidentified cells or cell groups in the central nervous system (Fig. 2E,J). At 24 hpf, prdm1 transcripts were maintained in the developing ears, branchial arches and hatching gland, and newly detected in the pectoral fin buds, ectodermal cells of the fin folds and the cloaca(Fig. 2K).

A new prdm1 expression domain in the photoreceptor layer of the retina was detected at 2 days postfertilization (dpf) and maintained until 3 dpf (Fig. 2L,M). Expression of prdm1 in the cloaca, the pharyngeal pouches and the pectoral fins was maintained at least until 5 dpf, whereas prdm1 expression in the ear disappeared by 3 dpf (Fig. 2L,M,N). Similarly, prdm1 transcripts disappeared gradually in the fin folds and were maintained only in the tip of the tail by 5 dpf (Fig. 2N). Starting at 3 dpf, prdm1 expression was detected in the developing neuromasts of the lateral line organ (Fig. 2N).

To place prdm1 in genetic hierarchies that regulate early development, we examined its expression in several pattern formation mutants. In boz^m168 embryos lacking dorsoanterior tissues such as forebrain, chorda and prechordal mesoderm(Solnica-Krezel et al., 1996; Fekany et al., 1999), the two lines of prdm1-expressing slow muscle precursor cells were either absent or fused, whereas the prechordal plate expression domain of prdm1 was missing (Fig. 3A,H,B,I). By contrast, injection of synthetic boz RNA induced ectopic prdm1 expression at early gastrulation (not shown). Therefore, both loss-of-function and gain-of-function data place mesendodermal prdm1 expression downstream of boz. Similarly, mesodermal and endodermal prdm1 expression depends on Nodal signaling, as the prechordal plate prdm1 expression domain was not detected in maternal-zygotic one-eyed pinhead^tz57 (MZoep)embryos lacking activity of the EGF-CFC co-factor essential for Nodal activity(Gritsman et al., 1999). By contrast, the ectodermal expression of prdm1 was not affected in MZoep mutants (Fig. 3A,H,C,J). prdm1 expression is also dependent on BMP signaling. Embryos homozygous for a null mutation in the bmp2b locus swirl^tc300a (swr) lack most ventral cell fates(Mullins et al., 1996). We observed that expression of prdm1 in the non-neural ectoderm was reduced and confined to a small ventral region of swr embryos(Fig. 3A,H,D,K). Conversely, dino^tt250 (din) mutant gastrulae lacking activity of the negative BMP regulator Chordin displayed reduced dorsal structures with a concomitant expansion of the ventral prdm1 expression domain(Fig. 3E,F)(Hammerschmidt et al.,1996).

slow muscle omitted^b641 (smu) mutant embryos,which lack activity of the Shh signal transducer Smoothened (Smo), fail to specify slow muscle (Barresi et al.,2000; Varga et al.,2001). This expression domain of prdm1 was absent in smu mutant embryos (Fig. 3L,M), suggesting that Prdm1 functions downstream of Shh signaling in the slow muscle precursors. Cloaca expression of prdm1 was missing in casanova^ta56 (cas) mutant embryos carrying a mutation in the sox32 locus and lacking endodermal structures(Fig. 3G,N) (Alexander et al.,1999b; Kikuchi et al., 2001; Dickmeis et al., 2001),underscoring dynamic expression of Prdm1 in all germ layers.

In order to investigate the role of Prdm1 during embryogenesis, we microinjected synthetic prdm1 RNA at the one-cell stage and studied the effects on embryo morphology. Injection of high doses (400 pg) resulted in 100% embryonic lethality by 24 hpf, whereas most embryos injected with 100 or 200 pg doses completed embryogenesis. Starting at 4 hpf and throughout gastrulation (6-9.5 hpf), these embryos exhibited irregular epibolic movements that normally thin and spread the blastoderm around the yolk cell (not shown). However, the completion of epiboly appeared unaffected. We used 100 pg doses in all the gain-of-function experiments described below if not indicated otherwise.

At 1 dpf embryos misexpressing Prdm1 exhibited reduction (45%) or loss(42%) of forebrain, including eyes, anterior midbrain and notochord, while the remainder of the body axis appeared largely unaffected (n=286; Fig. 4A,B,C). A low number of these embryos appeared to be affected much more severely (9%; Fig. 4D) and 4% died by 24 hpf. In support, forebrain expression of six3b was reduced (52%) or missing (42%) and pax2a expression at the mid/hindbrain boundary was reduced (11%), whereas hindbrain expression of krox20 was not affected (n=189; Fig. 4E,F).

The defects observed after Prdm1 misexpression phenocopied boz^m168 mutants(Solnica-Krezel et al., 1996). To gain further insight into the molecular nature of these patterning defects,we analysed expression of the organizer genes boz and gscand of the dorsoventral patterning genes chd, szl and bmp4in prdm1 RNA-injected embryos. Expression of boz, normally detected from midblastula transition until early gastrulation in the dorsal aspect of the blastoderm and the YSL (Koos and Ho, 1998; Yamanaka et al.,1998), was missing in 39% of injected embryos at 4.5 hpf(n=167; Fig. 4G,H). Accordingly, expression of the early anterior axial mesoderm marker gsc, which depends on boz function(Fekany et al., 1999), was strongly reduced or completely absent at 5.5 hpf (45%; n=143; Fig. 4I,J). These observations indicated that the gastrula organizer was not properly specified in prdm1-misexpressing embryos. In support of this notion, expression of chd, broadly distributed in dorsal mesoderm and ectoderm(Miller-Bertoglio et al.,1997), was reduced in 58% of early gastrulae (n=176; Fig. 4K,L). Moreover, ventral expression of szl encoding another BMP antagonist positively regulated by BMP signaling (Martyn and Schulte-Merker, 2003; Yabe et al., 2003) and of bmp4 was expanded (69%, n=188 and 65%, n=200, respectively; Fig. 4M-P). Furthermore, expression of hgg1 in the anterior prechordal plate mesoderm was either reduced and positioned more posteriorly(43%), or completely absent at 10.5 hpf (44%, n=238; Fig. 4Q,R). These effects are consistent with the head deficiencies observed at 24 hpf.

The similarity of dorsoanterior deficiencies observed in embryos misexpressing Prdm1 and boz mutants prompted us to test whether prdm1 RNA-injection could modify the boz phenotype, which exhibits both variable penetrance and expressivity and decreases with the age of the female parent (Fekany et al.,1999). At sub-threshold doses (50 pg), prdm1 RNA injection affected AP neural patterning mildly in 5.9% of wild-type embryos(Fig. 4U), indicated by slightly reduced forebrain six3b expression and unaltered pax2a and krox20 expression. By contrast, Prdm1 misexpression at the same dose in boz embryos caused increased penetrance of forebrain defects (66, 67 and 85%, respectively) compared with untreated siblings (0, 0 and 12%, respectively; Fig. 4U)(Fekany et al., 1999). The expressivity of the boz phenotype was also increased, as revealed by further reduction of six3b expression(Fig. 4S,T)(Fekany et al., 1999). Thus,given that boz^m168 is a strong/null allele(Fekany et al., 1999; Koos and Ho, 1999), Prdm1 misexpression at a sub-threshold dose can impact organizer formation by regulating expression of other genes in addition and/or parallel to boz.

The severely affected embryos misexpressing Prdm1(Fig. 4A,D) suggested that additional cell types were impaired. Expression of the pan-mesodermal marker ntl at the blastoderm margin was reduced in 73% of prdm1misexpressing embryos at late blastula stages (n=188; Fig. 5A,B). Similarly,expression of wnt8a, also involved in mesoderm development(Lekven et al., 2001; Erter et al., 2001), was reduced as well (92%, n=160; Fig. 5C,D). Likewise, expression of two endodermal markers, mix and sox17, was reduced in prdm1 misexpressing embryos at late blastula (21%, n=100; Fig. 5E,F) and early gastrula stages (91%, n=190; Fig. 5G,H), respectively. In conclusion, Prdm1 can also affect endodermal and mesodermal cell fate specification when ectopically expressed.

To address the requirement for Prdm1 activity during development, we aimed to eliminate prdm1 gene function by using morpholino antisense oligonucleotides (MO^prdm1) designed to interfere with translation of the targeted transcripts(Nasevicius and Ekker, 2000). Embryos injected with 1 ng of MO^prdm1 typically developed until 7 dpf (100%, n=376; see Fig. S1A,B in the supplementary material), while higher doses of 2 or 4 ng caused widespread necrosis in all germ layers and developmental arrest by 24 hpf (see Fig. S1A,C in the supplementary material). Therefore, in all loss-of-function experiments,MO^prdm1 doses of 2 ng/embryo were used for phenotypic analysis during gastrulation and early segmentation and doses of 1 ng/embryo for analysis during late segmentation and beyond 24 hpf. We evaluated effectiveness and specificity of the MO^prdm1-induced phenotype in a series of co-injection experiments using synthetic capped RNA of wild type and a mutated form of prdm1(Table 1A,B). First we assessed the effectiveness of MO^prdm1 by testing its ability to suppress the prdm1 gain-of-function phenotype. Whereas most embryos injected with 100 pg of prdm1 mRNA exhibited irregular epibolic movements at late blastula stages and throughout gastrulation, this phenotype was suppressed when co-injected with 1 ng of MO^prdm1(Table 1A). No such suppression of ectopic Prdm1 activity was observed when synthetic prdm1^mut mRNA, harboring mutations predicted to prevent binding of MO^prdm1, was used instead(Table 1A). These experiments provide support for the effective and specific interference with prdm1 RNA activity.

The earliest morphological abnormalities of MO^prdm1-injected embryos were observed at the conclusion of gastrulation (10 hpf), featuring an elongated animal-vegetal (AV) axis and mediolateral expansion in a dose-dependent fashion(Fig. 7N-S). At 2 ng doses all injected embryos (n=358) were affected, 93% of which strongly,whereas at 1 ng doses 85% (n=376) of injected embryos were affected less strongly, and the remainder appeared unaffected (data not shown). To investigate whether the defects observed in MO^prdm1-injected embryos resulted from the reduction of Prdm1 function alone, we carried out rescue experiments using a sub-threshold dose of synthetic prdm1^mut RNA (50 pg) causing mildly impaired epiboly in small numbers of injected embryos(Table 1B). When co-injected with 1 ng of MO^prdm1, the number of injected embryos exhibiting animal-vegetal elongation at 10 hpf was significantly reduced to 31% (Table 1B). This result demonstrates suppression of the MO^prdm1 phenotype by ectopic Prdm1 activity, supporting the notion that MO^prdm1specifically impairs prdm1 function.

To investigate whether the prdm1 morphant phenotype exhibiting an elongated AV axis is associated with dorsalization, as described for swr^tc300a or somitabun^dtc24(sbn; lacking Smad5 activity) mutant embryos(Mullins et al., 1996), we studied expression of bmp4, szl and chd. Consistently,ventroposterior expression of bmp4 was reduced in 60% of prdm1 morphant embryos at 10 hpf, while the anterior expression domain in prechordal mesoderm was unchanged (n=180; Fig. 6A,B). Correspondingly,mediolateral expansion of chd expression (68%, n=163) and reduction of ventral szl expression (70%, n=174) was observed at midgastrulation (6.5 hpf; Fig. 6C-F). However, these alterations of expression domains of dorsoventral (DV) patterning genes were milder than those reported for bmp2b/swr and smad5/sbn mutants(Kishimoto et al., 1997; Nguyen et al., 1998; Hild et al., 1999; Kramer et al., 2002). Furthermore, in agreement with the endogenous expression profile of prdm1, BMP signaling was not altered before the onset of gastrulation(not shown).

As Prdm1 functions in slow muscle development downstream of Shh signaling(Fig. 3L,M)(Roy et al., 2001) and shh is expressed in the gastrula organizer and axial mesoderm throughout gastrulation (Krauss et al.,1993), we investigated whether the effects on DV patterning observed in prdm1 morphant embryos might depend on Shh signaling. After injection of MO^prdm1, most smu/smomutant embryos, identified by missing ptc1 slow muscle expression,were elongated animal-vegetally and showed reduction of bmp4expression compared with untreated siblings (n=44; Fig. 6G,H). Thus, Prdm1 function in DV patterning does not depend on Shh signaling. Furthermore,interfering with ectodermal Prdm1 activity resulted in increased chdexpression in 61% of MO^prdm1-injected MZoepmutant embryos at the end of gastrulation (10 hpf; n=174; Fig. 6I,J). Hence, Prdm1 can regulate chd expression independently of Nodal signaling.

Further supporting the mild dorsalization phenotype, 85% of MO^prdm1-injected embryos displayed a mediolaterally enlarged neuroectoderm (krox20) and somitic mesoderm (dlc)at the onset of somitogenesis (10.5 hpf), whereas position of the anterior prechordal mesoderm (hgg1) was unchanged (n=142; Fig. 6K-N). In addition, AP neural patterning (six3b, pax2a krox20) was not affected in prdm1 morphant embryos (n=167; Fig. 6O,P). At mid-somitogenesis (16 hpf), expression of the ventral tissue markers gata1 in blood precursors and eve1 in ventroposterior cells in the tail were reduced in 94% (n=139) and 98% (n=135) of prdm1 morphant embryos, respectively(Fig. 6Q-T), suggesting a class C3 dorsalization phenotype according to Mullins et al.(Mullins et al., 1996). Nevertheless, at 24 hpf, judged by overall morphology, morphant embryos rather exhibited class 1 dorsalization characteristics (see Fig. S1A,B in the supplementary material).

The increase of chd and corresponding decrease of szlexpression in prdm1 morphant embryos(Fig. 6A-F) prompted us to investigate whether interference with prdm1 function can suppress the ventralized phenotypes of din/chd and ogon^tm305 (ogo; lacking Szl activity) mutant embryos. Whereas misexpression of Chd can fully suppress the ventralization phenotype of ogo/szl mutants (Miller-Bertoglio et al.,1999), misexpression of Szl does not rescue din/chd mutants,suggesting that Szl requires Chd function for its dorsalizing activity(Yabe et al., 2003). At 24 hpf, the ventral fin fold of the tail is characteristically duplicated and enlarged in both ogo/szl and din/chdembryos (Fig. 7A,B,D). Interestingly, when injected with MO^prdm1, we observed partial suppression of this phenotype in ogo/szl (92%, n=53) but not din/chd (0%, n=69) embryos(Fig. 7A-E). Accordingly, bmp4 expression was reduced in 72% of ogo/szlembryos (n=36; Fig. 7F,G) but not in din/chd embryos (0%, n=80; Fig. 7H,I)injected with MO^prdm1. Furthermore, the neuroectoderm was mediolaterally enlarged in ogo/szl but not in din/chd embryos with impaired prdm1 function(Fig. 7J-M). However, other characteristics of the ventralized phenotypes were not affected. The observation that the dorsalization of prdm1 morphants is associated with increased chd expression together with the above gain- and loss-of-function experiments suggest that Prdm1 activity promotes BMP signaling during zebrafish gastrulation, probably through limiting expression of chd in the organizer region.

Given that boz negatively regulates bmp2b expression(Fekany-Lee et al., 2000; Koos and Ho, 1999; Leung et al., 2003) and ectopic Prdm1 activity can suppress boz expression(Fig. 4H), we tested whether dorsalization of prdm1 morphants is caused by excessive or prolonged expression of boz and/or its downstream targets. We found that the boz expression domain was not affected in prdm1 morphant embryos at late blastula stages, and was correctly downregulated at the onset of gastrulation (not shown). Similarly, expression of vox and vent, negatively regulated by Boz(Kawahara et al., 2000a; Kawahara et al., 2000b; Melby et al., 2000; Imai et al., 2001), gsc,six3b and six7 (Kobayashi et al., 1998; Seo et al., 1998a,b),the Wnt8 antagonist dkk1 and the BMP antagonist nog1,confined to the presumptive anterior prechordal mesoderm(Hashimoto et al., 2000; Fürthauer et al., 1999),and hhex, expressed in the dorsal YSL(Ho et al., 1999), was normal in prdm1 morphant embryos at early gastrula stages (not shown). Furthermore, mesodermal ntl and wnt8a expression and endodermal mix and sox17 expression were unchanged in prdm1 morphant embryos (not shown). Hence, although Prdm1 can suppress boz expression in misexpression experiments, its function does not appear to be essential for the regulation of boz and its downstream effectors. We conclude that Prdm1 appears to function in limiting chd expression via boz- and sqt-independent mechanisms.

Dorsalized mutant embryos exhibit an elongated AV and narrowed mediolateral(ML) axis, reflected by an increased AV/ML ratio(Myers et al., 2002) and an increase of the AP embryonic length at the conclusion of gastrulation(Sepich and Solnica-Krezel,2004). Given that prdm1 morphant embryos also showed altered morphology (Fig. 7N-Q),we carried out morphometric analyses by capturing micrographs of untreated and MO^prdm1-injected sibling embryos at the end of gastrulation (Sepich et al.,2000). The AV/ML ratio of prdm1 morphant embryos was significantly increased (1.27±0.05) compared with untreated sibling embryos (1.1±0.04; Table 2) (Sepich et al.,2000). This increase of the AV/ML ratio of prdm1 morphant embryos prompted us to investigate the length of the nascent embryonic body(Fig. 7N)(Sepich et al., 2000), which is enlarged in dorsalized mutant embryos at the conclusion of gastrulation(Myers et al., 2002; Sepich and Solnica-Krezel,2004). Embryonic length of prdm1 morphant embryos was significantly increased to 1551.7±50.7 μm, compared with 1415.2±52.3 μm in untreated sibling embryos(Fig. 7N,O; Table 2). However, unlike strongly dorsalized swr/bmp2b or sbn/smad5embryos, which cease extension of the AP body axis by the end of gastrulation and do not survive early segmentation(Myers et al., 2002), prdm1 morphant embryos still exhibited increased embryonic length at early segmentation (Fig. 7P,Q; Table 2). Furthermore, ML width of the developing somites was enlarged in prdm1 morphant embryos at 12.5 hpf (Fig. 7R,S).

In order to test whether impaired gastrulation movements are responsible for the morphological defects observed in prdm1 morphant embryos, we tracked the AP extension of the dorsal mesendoderm between onset and conclusion of gastrulation in MO^prdm1-injected and untreated sibling embryos (Topczewski et al., 2001). A co-injected caged dextran-fluorescein conjugate was photoactivated at 6 hpf in a small cell population residing at the blastoderm margin dorsally in the embryonic shield or ventrally opposite to the embryonic shield (Fig. 7T). At the end of gastrulation the AP extension of the labeled tissue was measured. In morphant embryos, the labeled mesoderm formed a significantly longer cell array(1042.9±57.1 μm) than in untreated sibling embryos(989.1±49.2 μm; Fig. 7U,V; Table 2),revealing increased extension movements of dorsal tissues. Furthermore,consistent with mild dorsalization in MO^prdm1-injected embryos, labeled ventral cell populations underwent almost normal cell movements toward the tailbud exhibiting mild ectopic AP extension(Fig. 7W,X). By contrast,strongly dorsalized sbn/smad5 mutants exhibit a dramatic AP extension of ventral cell populations(Myers et al., 2002).

Late loss-of-function phenotypes in slow muscle, photoreceptor cell layer,branchial arches, pectoral fin and cloaca development are presented in Figs S1 and S2 in the supplementary material.

We report the functional analysis of zebrafish prdm1 expressed in a dynamic and complex fashion throughout embryonic development. Previously, Prdm1/Blimp1 had been described in human, mouse, chick and frog sharing many expression domains with zebrafish prdm1(Keller and Maniatis, 1991; Turner et al., 1994; Ha and Riddle, 2003; de Souza,1999). The strong similarities of sequence and expression pattern suggest functional conservation between all vertebrate Prdm1/Blimp1 genes. Conditional targeting of murine Prdm1 had revealed its essential role in B lymphocytes for plasmacytic differentiation, including secretion of immunoglobulins (Turner et al.,1994; Shapiro-Shelef et al.,2003). Misexpression of wild-type and dominant-negative forms of Prdm1/Blimp1 in X. laevis embryos suggested a role for Prdm1/Blimp1 in head formation (de Souza et al., 1999). A recent report by Baxendale and colleagues(Baxendale et al., 2004) showed that the function of zebrafish Prdm1 is partially inactivated in ubo^tp39 mutants, revealing an essential role of this gene in the differentiation of slow muscle downstream of Shh signaling(Roy et al., 2001; Baxendale et al., 2004). The role of Prdm1 in early development was not investigated by that work.

In early mouse and frog gastrulae Prdm1/Blimp1 is expressed in equivalent structures, the anterior visceral endoderm and the anterior endomesoderm, respectively, as well as in the presumptive prechordal plate mesoderm, both of which are implicated in forebrain specification and patterning (reviewed by de Souza and Niehrs, 2000; Kiecker and Niehrs, 2001). Similarly, prdm1 is expressed in the entire YSL underlying the blastoderm margin shortly before gastrulation starts and after onset of gastrulation, in the anterior portion of the prechordal mesendoderm. Interestingly, in mouse gastrulae, Prdm1/Blimp1expression was also detected in the posterior visceral endoderm(de Souza et al., 1999),reminiscent of prdm1 expression in ventral and lateral YSL(Fig. 2A,B).

We have addressed the role of Prdm1 during early development in gain-of-function and loss-of-function experiments. Misexpression of Prdm1 primarily caused reduction of dorsoanterior structures similar to defects observed in boz mutants(Solnica-Krezel et al., 1996). Furthermore, expression of the dorsal genes boz, gsc and chdwas downregulated. The notion of an important role of Prdm1 in limiting the gastrula organizer function is further supported by the loss-of-function experiments, which revealed that prdm1 morphant embryos were mildly dorsalized at the conclusion of gastrulation. These embryos exhibited an elongation of the animal-vegetal axis typical of dorsalized zebrafish mutants(Mullins et al., 1996; Myers et al., 2002). Accordingly, molecular analyses revealed mediolateral expansion of chd expression dorsally and reduction of ventral expression of bmp4 and szl, which encodes another BMP antagonist positively regulated by BMP signaling(Yabe et al., 2003; Martyn et al., 2003). Furthermore, at midsegmentation ventroposterior gata1 and eve1 expression was reduced. Additionally, depletion of Prdm1 activity partially suppressed the ventralized ogo mutant phenotype,but had no such effect on chd mutant embryos. This is also in agreement with recent work by Yabe et al.(Yabe et al., 2003), which showed that Chd activity is required for the ogo-dependent dorsalization. We hypothesize that Prdm1 acts directly or indirectly on chd in limiting gastrula organizer function.

In contrast to gain-of-function experiments, expression of boz and many other dorsal genes downstream of boz, such as gsc, hhex,six3b, six7, dkk1 and nog1, as well as expression of ventral genes such as vox and vent and of the Nodal-related genes ndr1 and ndr2, was not altered in prdm1 morphant embryos. This and the fact that many of the gain-of-function experiments did not reflect endogenous Prdm1 function, led us to propose that although ectopic Prdm1 activity is able to suppress boz expression, its function may not be essential for the regulation of boz and many of its downstream effectors during normal development. We conclude that Prdm1 limits the gastrula organizer function in zebrafish by negatively regulating chdexpression, largely via boz independent mechanisms. We hypothesize that Prdm1 regulates boz expression during normal fish development in a functionally redundant manner with other unknown genes.

Our analyses of morphogenetic defects in prdm1-depleted gastrulae are in support of the molecular data presented above. In embryos deficient in BMP signaling, the characteristic abnormalities in embryonic shape are a consequence of specific alteration of convergence and extension gastrulation movements (Myers et al.,2002). During normal gastrulation strong extension and moderate convergence movements are restricted to the dorsal hemisphere, generating embryos of slightly ovoid shape at the conclusion of gastrulation. Dorsalized sbn mutant embryos exhibit increased elongation of the animal-vegetal dimension and of the AP axis, in part due to slightly increased extension movements of dorsal cell populations(Myers et al., 2002; Sepich and Solnica-Krezel et al., 2004). Our morphometric analyses of prdm1morphant embryos are consistent with the increased extension movements of dorsal cell populations causing the increased length of the animal-vegetal and AP embryonic axes at the end of the gastrula period, typical of dorsalized mutants.

However, in strongly dorsalized smad5/sbn mutants at early segmentation, AP axis elongation becomes reduced compared with wild type and premature tail eversion takes place(Myers et al., 2002). Both defects have been attributed to altered cell movements of ventral cell populations. Whereas cells residing in the ventral regions of wild-type gastrulae, and experiencing the highest levels of BMP signaling, do not engage in convergence and extension movements and migrate toward the vegetal pole, in embryos with strongly reduced BMP signaling these cell populations undergo strong extension typical of more dorsal cell populations(Myers et al., 2002). By contrast, prdm1 morphant embryos manifest AP axes longer than wild type at 12.5 hpf, and develop a normal tail. Labeled ventral cell populations of embryos injected with MO^prdm1 underwent relatively normal movements toward the tailbud (Fig. 7W,X), consistent with a mild dorsalization of the morphant embryos.

Together our gain- and loss-of-function experiments establish a role for Prdm1 in limiting organizer function, uncovering noteworthy similarities and differences to the proposed function of X. laevis Prdm1/Blimp1. In zebrafish embryos prdm1 gain-of-function impaired primarily dorsoanterior structures, whereas in X. laevis it affected the entire anteroposterior axis, without exerting stronger effects on head versus trunk and tail (de Souza et al.,1999). While in agreement with our observations, expression of the BMP antagonist chordin was downregulated in X. laevisgastrulae overexpressing Prdm1/Blimp1, expression of other organizer genes, such as gsc and cerberus (cer),was either unchanged in dorsal marginal zone explants or ectopically induced in ventral explants (de Souza et al.,1999). These data lead to the conclusion that in X. laevis Prdm1/Blimp1 activity is required for head formation via positive regulation of cer expression (de Souza et al., 1999). Thus specific roles of Prdm1/Blimp1 in the organizer gene regulatory networks might differ between the two systems.

We thank L.S.-K.'s laboratory members for discussions, and B. Appel, J. R. Jessen, T. Van Raay and A. Inbal for critical comments. We thank P. Ingham, K. Artinger and W. Driever for sharing their results before publication. We acknowledge excellent fish care by J. Clanton and A. Bradshaw and thank C. Yin for assistance with confocal microscopy, D. S. Sepich for assistance with photoactivation experiments, and A. Flynt for help with the misexpression experiments. We thank A. Schier and M. Mullins for mutant lines, M. Westerfield, M. Halpern, T. Hirano, P. Ingham, M. Ekker, B. Thisse, C. Thisse,T. Schilling and T. Jowett for probes, and J. Malicki and F. E. Stockdale for a generous gift of antibodies and probes. Work in L.S.-K.'s laboratory is supported by NIH grants GM55101 and GM62283.