XPACE4 is a localized pro-protein convertase required for mesoderm induction and the cleavage of specific TGFβ proteins in Xenopusdevelopment

In Xenopus, the initial specification of the mesoderm and endoderm germ layers relies on the coordinated activities of stored maternal regulators; the T-box transcriptional activator VegT and the derepressorβ-catenin (Kofron et al.,1999; Xanthos et al.,2002; Xanthos et al.,2001; Zhang et al.,1998). These directly regulate the expression of zygotic transcription factors and signaling molecules, which in turn establish the endoderm and induce the neighboring mesoderm. With the onset of zygotic transcription in Xenopus, the signaling milieu around the embryonic cells rapidly becomes complex. VegT and β-catenin initiate the expression of at least seven TGFβ family members (nodal-related proteins 1-6, and derrière) and three FGFs (Fgf3, Fgf4 and Fgf8). Other TGFβ family members, Vg1 and BMP2, are inherited maternally, and Bmp4 and Bmp7 are expressed zygotically soon after the mid-blastula transition. Here, we examine the role in this process of the pro-protein convertase, XPACE4 (paired basic amino acid converting enzyme 4)/SPC4 (subtilisin-kexin like pro-protein convertase 4).

XPACE4 is a secreted, heparin-binding member of the pro-protein convertase(PC) family (Mains et al.,1997; Tsuji et al.,2003). It recognizes the multibasic consensus motif, RXXR, and cleaves substrates, including TGFβ family members, to release the active mature protein. PCs are important in vertebrate embryogenesis (for a review,see Taylor et al., 2003). In vertebrate embryos, they have been shown to process activin(Roebroek et al., 1993),TGFβ1 (Dubois et al.,1995), BMPs (Cui et al.,1998) and Nodal (Constam and Robertson, 1999; Beck et al.,2002; Constam and Robertson,2000). PCs have also been implicated in remodeling of the extracellular matrix by processing matrix metalloproteases(Leighton and Kadler, 2003; Yana and Weiss, 2000) and regulating cell adhesion by processing integrins(Berthet et al., 2000; Stawowy et al., 2004). Recently glypican 3, a heparan sulfate proteoglycan involved in morphogen gradient formation was found to be processed by PCs(De Cat et al., 2003).

In this study, we have analyzed the role of XPACE4 in mes-endoderm specification in Xenopus. We find that XPACE4 maternal mRNA is localized to the vegetal hemisphere of oocytes, whereas zygotic mRNA is localized to the notochord, the brain and a subset of endodermal precursors. We show that XPACE4 protein is essential for normal development in vivo, and for the production of endogenous mesoderm inducing signals. By comparing the cleavage of overexpressed tagged TGFβ proteins in wild-type and XPACE4-depleted backgrounds, we identify the TGFβ proteins that are specific substrates for cleavage by XPACE4.

Degenerate oligonucleotides were designed based on sequence motifs conserved among multiple members of the PC family. A partial length XPACE4 cDNA was amplified from oocyte cDNA by PCR using the oligonucleotide pair downward [5′-TGG GG(A/T/C) CC(A/T) GA(A/T/C/G)GA(T/C) GA(T/C) GG-3′ (coding for WGP(D/E)DG)] and upward[5′-AC(A/T) TGG (A/C)G(A/G/T) GA(T/C) (A/G)T(G/T) CA(G/A)-3′(coding for TWRC(M/V)QH)]. This cDNA was used to screen a Xenopusoocyte library and two partial-length XPACE4 clones were obtained and fully sequenced. Both of these lacked sequence encoding N- and C-terminal regions of the protein based on alignment with PACE4 from other species. cDNAs encoding a single XPACE4 N terminus and two distinct C termini were obtained using 5′ and 3′ Rapid Extension of cDNA Ends (RACE), respectively(AY836768 and AY836769). A single cDNA containing the entire open reading frame of the longer of the two XPACE4 splice isoforms was amplified from oocyte cDNA using PCR and was used for all subsequent studies.

The antisense oligonucleotide complementary to XPACE4 used in this study (AS-5) was 5′-C^*A^*A^*GGTTCAGGTAGCC^*G^*T^*-3′,where residues with phosphorothioate bonds are indicated by an asterisk(^*) and was used in doses of 4-5 ng. XPACE4 morpholino oligo was 5′-GCATGTTTGAAATGCTCAGAGGGAG-3′ and was used in doses of 30-45 ng.

To generate HA-tagged TGFβs, one HA epitope (YPYDVPDA) was inserted at the C terminus immediately after the coding sequence, and before the stop codon by high fidelity PCR. The inserts were subcloned into pCS2+ vector. Each construct was then sequenced for confirmation and tested for activity to ensure that the activity of the protein was not disrupted by the presence of the tag. The linearized plasmids were purified and mRNA was transcribed with SP6 polymerase using the Megascript kit (Ambion). The doses and sites of injection are described in the text.

Total RNA and cDNA were prepared according to Zhang et al.(Zhang et al., 1998). cDNA was synthesized using oligo dT primers, or random hexamer (R6) primers where indicated. Real-time RT-PCR was carried out using the Light Cycler System(Roche) as described by Kofron et al.(Kofron et al., 2001) using the primers and cycling conditions as listed in Table 1. All samples were normalized to levels of ornithine decarboxylase (ODC), which was used as the loading control. Every experiment was repeated at least twice with different oocyte and embryo batches to show that the results were reproducible.

Full-grown oocytes were isolated and cultured as described previously(Xanthos et al., 2001), before being injected with antisense oligos and fertilized using the host transfer technique as described previously (Zuck,1998). Staging was according to Nieuwkoop and Faber(Nieuwkoop and Faber,1967).

Oocytes at all stages were removed from follicle cells by incubating in 2 mg/ml collagenase in Ca²⁺ and Mg²⁺-free 1×MMR and fixed in MEMFA for 1-2 hours. To increase probe penetration into the yolky vegetal hemisphere, fixed oocytes and embryos were bisected into halves in the animal-vegetal axis using a clean scalpel blade, refixed in MEMFA, washed in PBS and transferred into 100% methanol. Whole-mount in situ protocol was modified from Harland (Harland,1991). XPACE4 antisense probe was transcribed using T7 polymerase after pCS2+XPACE4 vector was digested with EcoRI. For the sense probe, pCS2+ XPACE4 vector was digested with XhoI and transcribed using SP6 polymerase. In situs were developed using BM Purple as substrate (Roche).

Vegetal masses were dissected from control and XPACE4-depleted embryos at stage 9. Wild-type animal caps from stage 8-9 embryos were co-cultured with vegetal masses for 1-2 hours. They were then separated carefully using tungsten needles and contaminating vegetal cells, recognized by their vital dye coloring, were removed. The caps were incubated until they reached the desired stage and were frozen down for analysis.

The paracrine assay was carried out as described previously(Lustig and Kirschner, 1995). Oocytes were injected with an antisense oligo, cultured for 48 hours at 18°C, and then injected vegetally with 150-200 pg of Xnr1 mRNA,before culturing for an additional 12-24 hours to allow time for protein synthesis and processing. The oocytes were transferred into wells in agar plates with indentations, and stage 9 animal caps were placed on their animal poles, bringing the inner cell layer of caps into contact with the surface of the oocytes. Caps were co-cultured with oocytes in OCM for 1 hour at room temperature, and then carefully separated from the oocytes. The caps were cultured separately until sibling embryos reached the desired stage when they were frozen down for analysis.

Control and XPACE4-depleted embryos were injected vegetally at the two-cell stage with 50 pg of the firefly luciferase reporter construct pGL3-ARE-luciferase described previously(Huang et al., 1995) together with 20 pg of control HSTK Renilla luciferase plasmid. Batches of four embryos were collected in triplicate for each injection mixture at stage 10.5. Luciferase assays were performed using the Dual Luciferase Reporter Assay system (Promega) as described(Kofron et al., 2004). The assay was repeated twice, and a representative experiment is shown.

Needles were prepared in the size range of 10-20 nl/second, and rinsed with the protease inhibitor PIC (Roche). The needle was inserted through the blastocoel roof into the blastocoel cavity of early gastrula stage embryos,and the blastocoel fluid was withdrawn gradually, using a Medical Systems picoinjector. Care was taken not to contaminate the fluid with cellular debris. Depending on the batch of embryos, 0.1-0.2 μl of fluid/embryo was collected. After blastocoel fluid from 10 embryos was collected, the fluid was snap frozen. Similar volumes of fluid were withdrawn from control and experimental embryos.

Protein extracts were prepared from batches of 3-5 embryos using 10 μl homogenization buffer/embryo. Homogenization buffer for anti-HA blots is 2.5%IGEPAL CA-630 Sigma I3021, 50 mM Tris pH 7.5, 150 mM NaCl, 1 mM EDTA, 1%Triton X-100, 0.1% SDS; containing 1:100 PIC (Sigma P-8340). For anti-phospho-SMAD2 blots, homogenization buffer is 20 mM Tris pH 8.0, 2 mM EDTA, 5 mM EGTA, 0.5% NP-40 (Sigma), 25 mM sodium β-glycerophosphate, 100 mM NaF, 20 nM Calyculin A, 10 mM sodium pyrophosphate, containing 1:100 PIC and PMSF. Homogenates were spun at 13,000 g for 10 minutes and the supernatant was collected and mixed with equal volume of 2× reducing sample buffer. Blastocoel fluid samples were also spun down and 10 μl of sample buffer was added. Samples were boiled and 0.5-1 embryo equivalents of protein extracts or the entire blastocoel fluid samples were loaded on 10-12%SDS-PAGE gels and transferred on to nitrocellulose membranes. The membranes were blocked with 5% NFDM in PBS-Tween overnight at 4°C for anti-phospho-Smad2 blots and 1-2 hours at room temperature for others. Anti-phospho-Smad2 antibody (Cell Signaling Technology #3101) and anti-Smad2 antibody (BD Transduction Laboratories) were used at a dilution of 1:500. Anti-HA high affinity rat monoclonal antibody 3F10 (Roche 1-867-423) was used at a dilution of 1:1000. Signal detection was carried out using the Amersham ECL detection system. No signal was detected with this HA antibody in the uninjected control samples. The relative mobility of each protein was calculated by comparison with a standard curve generated by migration of the prestained kaleidoscope standards (BioRad). One HA tag is approximately 1.1 kDa.

As a loading control, membranes were stripped after signal detection, and incubated with α-tubulin antibody (DM1A, Neomarkers) at 1:10,000. Noα-tubulin was detected in the blastocoel fluid samples. The experiments were repeated at least twice.

The predicted amino acid sequence of one splice isoform of XPACE4 is shown in Fig. 1. It is most closely related to rat PACE4 (rPACE4) and human PACE4A-II isoform (hPACE4A-II). To determine the temporal expression profile of XPACE4 mRNA, real-time RT-PCR was performed using XPACE4-specific primers on a staged series of embryos. Fig. 2A shows that XPACE4 mRNA is present in oocytes and continues to be expressed throughout blastula and gastrula stages with maximum expression at the early blastula stage. As mRNA levels increase before mid-blastula transition (MBT),when transcription is considered to be absent, we asked whether this was due to polyadenylation. We compared oligo dT primed versus random hexamer primed cDNA for XPACE4 levels. Fig. 2B shows that XPACE4 mRNA levels remain constant at early blastula stage compared with oocyte levels in random hexamer primed cDNA,indicating that the increase in XPACE4 levels is due to polyadenylation of the maternal mRNA. Northern blot analysis reveals the presence of three distinct XPACE4 RNA species that range in size from 2.5 to 7 kb and show differing patterns of temporal regulation (data not shown).

To determine the spatial localization of XPACE4 mRNA during oogenesis, real-time RT-PCR was performed using oocytes dissected into animal-vegetal halves, and embryos dissected into dorsal and ventral halves or equatorial and vegetal explants. XPACE4 mRNA is enriched in vegetal halves of oocytes and vegetal masses of embryos, with no significant asymmetrical distribution between dorsal and ventral halves (data not shown). To confirm the localized expression pattern of XPACE4, we performed whole-mount in situ hybridization using XPACE4 antisense probe at different stages of oogenesis and embryogenesis. Fig. 2C shows that in stage 1 oocytes, XPACE4 localizes to the mitochondrial cloud. As the mitochondrial cloud disperses at stage 2(Heasman et al., 1984), XPACE4 mRNA co-localizes with the vegetal cortex where it remains for the rest of oogenesis (Fig. 2C;inset). Hemi-sections of full-grown oocytes reveals that XPACE4 mRNA is not exclusively associated with the cortex of the vegetal hemisphere but is also diffusely distributed in the vegetal cytoplasm and excluded from the animal hemisphere (Fig. 2D). Fig. 2E,F shows that at the gastrula stages, XPACE4 mRNA persists vegetally and becomes localized to the presumptive endoderm. After gastrulation, XPACE4 expression declines dramatically and Fig. 2G shows that it is restricted to a small number of cells in the endoderm. Fig. 2H shows that at tailbud stages, the signal is detected specifically in the olfactory bulb, the brain and in the notochord (inset shows staining in the notochord in a transverse section with sense control on the right).

These results show that maternal XPACE4 mRNA is localized to the vegetal hemisphere of the oocyte and is regulated by polyadenylation, while during embryogenesis, it is localized in endodermal precursors and later in specific areas of all three germ layers.

Antisense oligos designed against the XPACE4 coding sequence were tested for efficiency by injection into the vegetal poles of full-grown oocytes, which were incubated for 48 hours before real-time RT PCR analysis for XPACE4 mRNA depletion. Fig. 3A shows that antisense oligo 5 (AS-5) is the most efficient, and depletes XPACE4 mRNA to 5-10% of control levels. To increase the stability of oligo, phosphorothioate modified AS-5 (AS-5MP) was used. Fig. 3B shows that the oligo does not affect mRNA levels of a related maternal convertase, furin(XFurA).

To determine when zygotic XPACE4 mRNA is first expressed in maternal XPACE4-depleted embryos, we assayed a series of developmental stages comparing XPACE4-depleted embryos with controls. Fig. 3C shows that XPACE4 mRNA levels remain depleted throughout blastula and gastrula stages. By in situ hybridization, no RNA is detectable in oligo-depleted embryos at the late blastula stage, although a low level of nuclear transcript is visible by mid-gastrulation, indicating that zygotic transcription begins at this time (data not shown).

As mouse Nodal is a known substrate of mammalian PACE4(Beck et al., 2002), we assessed the loss of activity of XPACE4 in a functional assay. For paracrine assays, control and XPACE4-depleted oocytes were injected with Xnr1 mRNA and co-cultured with wild-type animal caps as described in the Materials and methods. Fig. 3D shows that the depletion of XPACE4 decreases the inducing activity of Xnr1. Although the organizer genes, chordin(Chd) and goosecoid (Gsc), and the mesodermal gene Xbra and endodermal gene XSox17α are induced in animal caps co-cultured with control oocytes overexpressing Xnr1 mRNA, caps cultured with XPACE4-depleted oocytes show little inducing activity. This indicates that the Xnr1-processing activity of endogenous XPACE4 protein is reduced by the antisense oligo-mediated depletion of the maternal XPACE4 mRNA.

To determine the function of maternal XPACE4 during embryogenesis, we used the host transfer technique to fertilize XPACE4-depleted oocytes. XPACE4-depleted embryos develop normally through the cleavage and blastula stages. The phenotype caused by maternal XPACE4 depletion is obvious at gastrulation, when the formation of the blastopore is delayed compared with controls (Table 2). This phenotype is rescued by the reintroduction of XPACE4 mRNA into sibling XPACE4-depleted embryos(Fig. 4A,B). After 1-2 hours delay, the blastopore ring forms and closes and the embryos continue to develop, with varying degrees of abnormality of anterior structures(Fig. 4C; Table 3). Similar phenotypes are observed when 30-45 ng of XPACE4 morpholino oligo is used (data not shown).

As XPACE4 depletion in oocytes abrogated the ability of injected Xnr1 mRNA to induce mesoderm in animal caps(Fig. 3D), we reasoned that XPACE4-depleted embryos may be deficient in Xnr signaling and therefore deficient in mesoderm induction. Four assays were used to test this.

First, equatorial regions, which are specified at the late blastula stage to form mesodermal tissue, were dissected from control and XPACE4-depleted embryos, cultured until the neurula stage and analyzed for the expression of mesodermal (MyoD and cardiac actin) and neural (NCAM) markers. Fig. 4D shows that control equatorial explants undergo the typical convergent extension movements indicative of mesoderm induction. By contrast, XPACE4-depleted equatorial explants are unable to undergo convergent extension movements, a deficiency that is specific to XPACE4-depletion as it is rescued by the reintroduction of XPACE4 mRNA. The expression of both cardiac actin and MyoD expression is reduced by XPACE4-depletion and is rescued by the reintroduction of XPACE4 mRNA (Fig. 4E). Neural tissue is normally secondarily induced in control equatorial explants. Neural induction assayed by the expression levels of NCAM is reduced by XPACE4-depletion and rescued by XPACE4 mRNA. Marker analysis of gastrulae using real-time RT-PCR confirms that the expression of mes-endodermal genes is reduced in XPACE4-depleted embryos. In particular, the general mesodermal gene Xbra, and the organizer gene chordin (Chd) are reproducibly reduced in expression, while Xnr1, which is known to be positively autoregulated, and Xnr3 are also affected. The expression of Chd, Xnr1 and Xnr3 is partially rescued by the reintroduction of XPACE4mRNA (Fig. 4F). The expression of the mesodermal markers Fgf8 and derrière in whole embryos is little affected by XPACE4-depletion (data not shown).

Second, we asked whether endogenous mesoderm induction by vegetal masses was reduced in XPACE4-depleted embryos, using Nieuwkoop assays(Fig. 5A). Animal caps were dissected from wild type mid-blastula stage embryos and co-cultured for 1-2 hours with vegetal masses dissected from control or XPACE4-depleted embryos at the mid-blastula stage. After the induction period, the caps were separated and cultured until the mid-gastrula stage and analyzed for induced gene expression. Fig. 5A shows that the mes-endodermal genes chordin (Chd), goosecoid (Gsc), Fgf8, Xbra and XSox17α, which are targets of TGFβ signaling, are significantly induced in animal caps by wild-type vegetal masses, but not XPACE4-depleted vegetal masses. This experiment was repeated twice with similar results.

Third, we assayed the activity of ARE-luciferase reporter in XPACE4-depleted embryos. Control and XPACE4-depleted embryos were injected vegetally with the ARE-luciferase reporter at the two-cell stage and assayed for luciferase activity at stage 10.5. Fig. 5B shows average ARE-luciferase activity is significantly reduced in XPACE4-depleted embryos.

Fourth, we determined the phospho-Smad2 levels in XPACE4-depleted embryos by western blots using a phospho-Smad2 specific antibody. Fig. 5C shows that XPACE4-depleted embryos have decreased phospho-Smad2 levels most significantly at the late blastula.

These results indicate that XPACE4 regulates endogenous mesoderm induction.

To determine which TGFβ family members expressed during early Xenopus embryogenesis are substrates for XPACE4, we assayed control and XPACE4-depleted embryos for processing of different TGFβproteins. We injected mRNA coding for HA-tagged proteins into the vegetal cytoplasm of wild-type and XPACE4-depleted embryos at the two-cell stage, and assayed for precursor and mature forms of proteins by western blotting using a high-affinity HA antibody at the early gastrula stage. The HA tag was inserted into the C terminus of each construct, allowing the visualization of the mature protein, precursor protein and any stable C-terminal-containing intermediaries of Xnr2, Xnr1, Xnr3, Xnr5, Vg1, ActivinB and Derrière. In many cases, the analysis of whole embryo homogenates showed that only a small percentage of the overexpressed proteins were cleaved, making comparison between control and XPACE4-depleted embryos difficult. Therefore, we took advantage of the fact that both precursor and cleaved forms of the HA-tagged TGFβ proteins are secreted into the blastocoel cavity (Williams et al., 2004). As this concentration of pro-protein and mature protein is uncontaminated with cellular debris, it provides a more sensitive assay for the determination of changes due to XPACE4-depletion. The experiments were repeated at least three times and a representative experiment is presented for each of TGFβ protein(Fig. 6A-H).

As Nodal processing is defective in mice lacking both furin and PACE4(Beck et al., 2002), we first tested the effect of XPACE4-depletion on the cleavage of nodal related proteins Xnr1, Xnr2, Xnr3 and Xnr5. Fig. 6 shows the cleavage profile of overexpressed TGFβ proteins in embryo lysates and blastocoel fluid and Table 4 summarizes the effects of XPACE4 depletion on processing of different TGFβ proteins. In XPACE4-depleted whole embryo lysates, the mature Xnr2-HA is reduced compared with controls. Interestingly, the blastocoel fluid analysis shows the reduction in mature Xnr2-HA, and a relative increase in both the uncleaved and intermediate forms as a result of XPACE4 depletion. Thus, XPACE4 is required for the production of mature Xnr2-HA but not for the formation of the intermediate form. The reduction in cleavage of Xnr2-HA is specifically due to XPACE4 depletion, as Fig. 6Bshows that the amount of mature protein increases and the uncleaved and the intermediate forms decrease when XPACE4 mRNA is injected at the two-to four-cell stage, along with Xnr2-HA mRNA into XPACE4-depleted embryos. Processing of Xnr1 (Fig. 6C) and Xnr3 (Fig. 6D) is also reduced by XPACE4 depletion. In contrast to Xnr1, Xnr2 and Xnr3, cleavage of Xnr5-HA (Fig. 6E), ActivinB (Fig. 6F) or Derrière (Fig. 6G) is not affected by XPACE4 depletion.

In contrast to the zygotically expressed Xnr proteins and derrière,Vg1 is a maternal mRNA and protein, which is, like XPACE4, localized to the vegetal cortex of Xenopus oocytes(Dale et al., 1989; Melton, 1987; Tannahill and Melton, 1989; Weeks and Melton, 1987). Next,we asked whether Vg1 processing would be affected by XPACE4depletion. HA-tagged Xenopus Vg1 was cloned from oocyte cDNA as described in the Materials and methods, and found to have a serine instead of a proline residue in position 20 (AY838794) compared with the published sequence (Weeks and Melton,1987). Detailed analysis of the ESTs available for Xenopus laevis revealed that this serine residue is conserved in 14 ESTs compared with 13 ESTs with proline residues, suggesting that this represents a pseudoallelic form of Vg1. Western blot analysis of overexpressed Vg1-HA in Xenopus embryos shows the processing of Vg1-HA is affected in XPACE4-depleted embryos, as the mature protein levels decrease and there is a corresponding increase in the levels of unprocessed forms(Fig. 6H). The intermediate form of Vg1 may be a nonspecific degradation product or product of cleavage at an upstream site. Native Vg1 has been shown to have no mesoderm inducing activity. As our sequence is different from the published one, we tested the mesoderm inducing activity of Vg1 and found that this Vg1 has weak general mesoderm inducing activity (data not shown).

These results demonstrate that endogenous XPACE4 cleaves exogenous Xnr1,Xnr2, Xnr3 and Vg1, but not ActivinB, Derrière and Xnr5. Taken together, this work provides the first evidence that XPACE4 is an essential vegetally localized regulator of mesoderm induction in Xenopusembryos, with specificity towards the TGFβs Xnr1, Xnr2, Xnr3 and Vg1.

Several studies have shown that PCs when overexpressed have the ability to cleave a wide spectrum of substrates. However, the specificity of PACE4 towards endogenous substrates during development has not been investigated. We have studied this, using loss-of-function assays, in early Xenopusembryos. Three novel observations are made. First, maternal XPACE4mRNA is localized to the mitochondrial clouds and vegetal hemispheres of oocytes. Second, it is required for the endogenous mesoderm inducing activity of vegetal cells before gastrulation, and third, it has substrate-specific activity, cleaving Xnr1, Xnr2, Xnr3 and Vg1, but not Xnr5, Derrière and ActivinB pro-proteins.

At the late blastula and early gastrula stages, the expression domains of XPACE4 and several TGFβ mRNAs, including Vg1, Xnr1 and Xnr2,coincide. In mouse, processing depends on the ability of the secreted pro-protein precursor and the PACE4/SPC4 enzyme to contact each other in the extracellular space (Beck et al.,2002; Constam and Robertson,1999). We have no direct evidence that endogenous XPACE4 is a secreted protein, as specific antibodies are not available. However, we find that uncleaved as well as cleaved forms of exogenous Vg1, Xnrs, ActivinB and Derrière accumulate in a stable fashion in blastocoel fluid, and that,for Xnr1, Xnr2 and Xnr3, the uncleaved pro-proteins accumulate more when XPACE4 is depleted. The simplest explanation for this is that XPACE4 is secreted. Thus, the pro-proteins are normally cleaved extracellularly by secreted XPACE4, and accumulate more in the blastocoel in XPACE-depleted embryos, because the enzyme is not there to process it. However, this remains to be confirmed directly. Another possibility is that cleavage of endogenous TGFβs occurs intracellularly, but that excess uncleaved and intermediate forms are produced and secreted in blastocoel fluid as a result of the excessive loading of the cells with these mRNAs (100-600 pg of mRNA for HA-tagged TGFβs was injected). It will be important to determine whether endogenous XPACE4 is present in blastocoel fluid when antibodies become available.

Later in development, XPACE4 mRNA is found in the notochord, the brain and the olfactory bulb, where its position correlates with expression of rodent PACE4 in the developing CNS(Akamatsu et al., 1997; Zheng et al., 1997), and in a subset of cells in the endoderm correlating with the position of primordial germ cells. Possible roles of XPACE4 in these locations remain to be determined.

The regulation of TGFβ signaling in vertebrate embryogenesis is crucial for correct tissue formation, and this study together with work in mouse embryos shows the importance of subtilisin-family proteases in this process (Beck et al., 2002; Constam and Robertson, 1999; Constam and Robertson, 2000). Although the targeted loss of function of PACE4 in mouse embryos showed relatively mild effects [causing only 25% of animals to die prenatally with cardiac malformations, laterality defects and craniofacial abnormalities, and 75% to survive to birth (Constam and Robertson, 2000)], PACE4/furin double mutants showed severe anterior visceral endoderm and axis formation defects, indicative of disruption of Nodal cleavage and function(Beck et al., 2002). In the studies presented here, we find that the predominant contribution of XPACE4 mRNA is from maternal stores accumulated during oogenesis,with relatively small and localized areas of synthesis of zygotic XPACE4 transcript. Two observations argue that the maternal store of protein accumulated specifically in vegetal, endomesoderm precursors is stable and long lived. First, the XPACE4 phenotype, delayed gastrulation and headless development, was more pronounced the longer the oligo-injected oocytes were incubated before fertilization (48 hours versus 24 hours). Longer incubation after oligo injection would be expected to produce greater turnover of the stable XPACE4 protein pool synthesized during oogenesis. Second, we found that a morpholino oligo against XPACE4 was effective in causing the same phenotype as AS-5MP oligo if the morpholino oligo was injected 2 days before fertilization, but was not effective when injected into fertilized eggs (data not shown). It will be interesting to determine if there is also a maternal contribution of PACE4/SPC4 protein in mouse development, which allows the relatively normal early development of mPACE4^-/- embryos.

Several lines of evidence presented show the importance of XPACE4 in mesoderm induction. First, loss of XPACE4 causes a loss of mesodermal markers in whole embryos. Second, Niewkoop assays show that XPACE-depleted vegetal masses are unable to induce wild-type animal caps to form mesoderm. Third,loss of XPACE4 causes loss of ARE-luciferase activity concomitant with loss of mesodermal gene expression. Last, phospho-Smad2, the immediate downstream target of activin-type TGFβ signaling, is also reduced in XPACE4-depleted embryos. All approaches show the reduction of Nodal signaling activity due to the depletion of maternal XPACE4. Since, of the TGFβs tested, we found that Xnr1, Xnr2, Xnr3 and Vg1 are substrates for XPACE4 cleavage, it is likely that the defect in mesodermal induction due to XPACE4 depletion is caused by the reduction in mature forms of these proteins. Previously, we have shown that Xnr3 is required for convergence extension movements and head formation(Yokota et al., 2003), and that Xnr1 and Xnr2 are able to rescue axis formation and mesoderm induction in embryos depleted of TGFβ signaling by antisense ablation of maternal VegT(Kofron et al., 1999; Xanthos et al., 2001; Zhang et al., 1998). We find here that the Vg1-like protein Derrière is not a substrate for XPACE4. Derrière is known to be important in establishment of tail and posterior axial structures, and rescues VegT-depleted embryos to the extent of causing late gastrulation and the formation of tails and dorsal axes but not heads (Kofron et al., 1999). Thus, one explanation for the late gastrulation and posterior axial development of XPACE4-depleted embryos is that normal Derrière function is maintained. This hypothesis is supported by the fact that phospho-Smad2 levels approach control levels by mid-gastrulation when Derrière is expressed (Fig. 5C). It will be interesting to determine whether Derrière, ActivinB and Xnr5 are substrates for the other maternally supplied PCs.

To date, this work provides the most comprehensive comparison of the cleavage spectra of seven TGFβ proteins in embryo lysates and blastocoel fluids, and the effect of depletion of one of the regulating enzymes on those patterns. As the HA tags were placed at the C termini of the proteins, N terminal pro-domain fragments were not visualized. Table 4 correlates these findings with the known biological activity of each of the proteins as measured by mesoderm induction. Several conclusions can be drawn from the comparisons.

First, XPACE4 regulates Xnr1, Xnr2, Xnr 3 and Vg1 but not Xnr5,Derrière and ActivinB. As all of these proteins (with the exception of ActivinB) share an RXXR consensus cleavage motif, the substrate specificity of XPACE4 may be due to different residues flanking this consensus site. Specificity and optimum activity of convertases have been documented to be dependent on these flanking sequences(Apletalina et al., 1998; Henrich et al., 2003), and studies on Vg1 cleavage indicate that the presentation of the cleavage site in a particular context and/or conformation is crucial for regulating cleavage(Dohrmann et al., 1996; Thomsen and Melton, 1993).

Second, we show here that the mature form of Xenopus Vg1 can be detected by western blots using a high-affinity HA antibody, that this form has weak mesoderm-inducing activity and that the processing of Vg1 is reduced in XPACE4-depleted embryos. Previously, although native Vg1 protein has been shown to be present vegetally in oocytes and early Xenopusembryos (Dale et al., 1989),the mature form was neither detected nor active in mesoderm induction(Tannahill and Melton, 1989). In addition, while Vg1 chimeras with prodomains of either Activin or BMP proteins were processed efficiently and were potent mesoderm inducers(Dale et al., 1993; Thomsen and Melton, 1993),mature exogenous Vg1 could be detected but had no mesoderm-inducing activity(Kramer and Yost, 2002). By contrast, the zebrafish homolog of Vg1, zDVR-1(Dohrmann et al., 1996) and chick Vg1 (Shah et al., 1997)were detectable as mature proteins with weak mesoderm-inducing activity in Xenopus oocytes.

The Vg1 used here has some sequence differences compared with the published Vg1 sequence. One difference is a C-to-T transition resulting in a serine residue (TCN codon) instead of a proline (CCN codon) at position 20. The analysis of available EST sequences has suggested that both of these alleles are present in Xenopus oocytes. The proline residue in the helix rich N terminal region of Vg1 may be responsible for disruption of the structure,and affect processing and render the protein functionless. This may explain the previous observation that N-terminal signal peptide of Vg1 does not get cleaved and Vg1 protein fails to dimerize(Dale et al., 1993), whereas the Vg1 with a serine residue at position 20 has weak mesoderm-inducing activity (data not shown). It will be interesting to determine the function of this Vg1 allele in detail and to understand the structural distinctions between the different forms.

Third, the Xnr2 cleavage spectrum is more complex than that of the other six proteins. Although the intermediate form of 33 kDa may be a degradation product or caused by nonspecific cleavage at an upstream cryptic site, the fact that it accumulates in an XPACE-depleted background suggests that it may be formed by the specific pro-protein digestion activity of another endoprotease, generating a processing intermediate that is subsequently cleaved at the known downstream site. We have shown previously that BMP4 processing involves sequential cleavage that is important in regulating the stability and activity of the mature protein(Cui et al., 2001). Analysis of Xnr2 protein sequence reveals the presence of a consensus RXXR motif located∼70 amino acids upstream of the known cleavage site that would be predicted to yield a protein form of ∼33 kDa upon endoproteolysis. Xnr2 generated from a precursor in which this upstream site is mutated, however,maintains reduced activity, suggesting that processing at the downstream site occurs independent of the upstream site and is sufficient for Xbra induction activity. A double cleavage mutant of Xnr2 that fails to generate any mature protein also induces Xbra, suggesting that unprocessed Xnr2 has some signaling activity (Eimon and Harland,2002).

Fourth, as has been shown previously for Activin and Vg1, there is no simple correlation between the amount of mature form of each of the TGFβs and their biological activity. Here, Xnr5 and Activin are the most active proteins. They work at very low doses of RNA in mesoderm induction assays and yet accumulate less well than Xnr1 and 2 in whole embryo lysates. They do,however, accumulate substantially in blastocoel fluid. Cleaved forms of Xnr3 and Vg1 are only very weakly detected in either the lysates or the blastocoel fluid. Undoubtedly, the regulation of each TGFβ is complex and individual, involving the interplay of stabilizing or competing intermediary cleavage forms, other competing co-expressed TGFβ family members and the presence or absence of localized co-receptors. Specific loss-of-function and mutation analysis for each individual TGFβ and detailed functional analysis of the other endoproteolytic enzymes are required to fully understand their function.

This work was supported by NIH RO1 to J.H. (HD038272) and by NIH to J.L.C.(HD37976 and HD042598). Excellent technical support was provided by Matt Kofron, Douglas W. Houston, Chika Yokota and Nambi Sundaram. ARE-luciferase reporter was provided by Malcom Whitman (Harvard Medical School, USA). Thanks to Karel Dorey and Caroline Hill (Developmental Signalling Laboratory, Cancer Research UK) for the phospho-Smad2 western blot protocol.