Man1, an inner nuclear membrane protein, regulates vascular remodeling by modulating transforming growth factor β signaling

The nuclear envelope (NE) is composed of outer and inner nuclear membranes(INM), nuclear lamina and nuclear pore complexes. The outer nuclear membrane is directly continuous with the endoplasmic reticulum membrane, whereas the INM is lined by a nuclear lamina network formed mainly by type V intermediate filaments containing type A and type B lamins. The nuclear lamins bind to several INM proteins, including lamina-associated polypeptides 1 and 2β(LAP1, LAP2β), emerin and Man1 (Lemd3 - Mouse Genome Informatics; the subject of this study), which share a common structural motif of about 40 amino acid residues, called the LEM (LAPs, emerin and Man1) domain(Lin et al., 2000). The nuclear lamins and the LEM domain proteins interact with barrier-to-autointegration factor, through which they are involved in higher-order chromatin organization, nuclear assembly and gene regulation(Gruenbaum et al., 2005). However, the functions of a large number of putative integral NE proteins remain to be elucidated (Schirmer et al.,2003).

Human MAN1 was originally identified as one of three antigens recognized by autoantibodies from an individual with a collagen vascular disease(Lin et al., 2000). Man1 has a N-terminal LEM domain, two putative transmembrane domains, a Man1-SRC1P C-terminal domain of unknown function(Mans et al., 2004) and an RNA recognition motif. An RNA interference-mediated loss-of-function experiment with the Man1 ortholog in C. elegans, Ce-MAN1, provides evidence that Ce-MAN1, in combination with Ce-emerin, plays an essential role in chromosome segregation and cell division (Liu et al.,2003).

The linkage between the INM and growth factor-related signal transduction was initially discovered by analyzing the roles of XMAN1 and SANE (Smad1 antagonistic effector), both of which are Xenopus orthologs of Man1,in early embryogenesis (Osada et al.,2003; Raju et al.,2003). We identified XMAN1 as a novel factor that neuralizes the ectoderm and dorsalizes the ventral mesoderm. In Xenopus, inhibition of the bone morphogenetic protein (Bmp) pathway is a crucial step for neural induction (Harland, 2000). We have demonstrated that XMAN1 antagonizes the Bmp pathway by interacting with the MH2 domain of the receptor-associated Smads (R-Smads) that mediate Bmp signaling (Smad1, Smad5 and Smad8) through its C-terminal region. SANE was isolated as a novel Smad1-interacting protein by the yeast two-hybrid system and shows essentially the same activities as XMAN1.

Abnormal function of the NE is implicated in a wide range of human diseases, collectively termed laminopathies(Burke and Stewart, 2002). Heterozygous loss-of-function mutations in human MAN1 (LEMD3- Human Gene Nomenclature Database) cause osteopoikilosis, Buschke-Ollendorf syndrome and melorheostosis (Hellemans et al., 2004), all of which are characterized by hyperostotic bones. The associated mutations involve the deletion of the C-terminal region of MAN1, which is the interaction domain for R-Smads. Interestingly, expression of wild-type human LEMD3 in mammalian cells suppresses both transforming growth factor (TGF) β1-dependent and BMP-dependent reporter activation,whereas expression of a mutant MAN1 with either of the LEMD3mutations is unable to suppress TGFβ1 signaling. These observations indicate that human MAN1 antagonizes both the BMP- and TGFβ/activin-signaling pathways, a conclusion that was also demonstrated independently by two other groups using mammalian cells(Lin et al., 2005; Pan et al., 2005). Thus,increased bone density and disseminated connective tissue nevi found in individuals with the LEMD3 mutations and elevated elastin production in fibroblasts from patients with Buschke-Ollendorf syndrome(Giro et al., 1992) might be explained by augmented sensitivity to the bone-forming activity of BMPs and the extracellular matrix-producing activity of TGFβ1.

In this study, to better understand the role of Man1 in cellular and developmental processes, we analyzed the consequence of Man1 deficiency in vivo. We used gene trapping to generate Man1-deficient mice, in which the C-terminal region containing the Smad-interaction domain was deleted. We found that the angiogenesis processes required to build a mature capillary were severely perturbed in Man1-deficient embryos. These embryos had an abnormal augmentation of Smad2/3 signaling, resulting in increased extracellular matrix deposition, which is believed to inhibit endothelial cell proliferation and migration (Pepper,1997). These results provide the first evidence that Man1 plays a crucial role in angiogenesis, acting as a `gatekeeper' at the INM to modulate the activity of the Tgfβ signaling pathway through the interaction with receptor-associated Smads.

Artificial fertilization, culture of embryonic tissues and embryo staging were performed as previously described(Kay and Peng, 1991; Nieuwkoop and Faber, 1967). Whole-mount in situ hybridization of Xenopus embryos was performed. RT-PCR and luciferase assays using ectodermal explants were carried out as previously described (Osada et al.,2003).

A mouse embryonic stem cell line (cell line XST167, strain 129/Ola)containing an insertional mutation in Man1 was created in a gene-trapping program by BayGenomics(http://baygenomics.ucsf.edu/). The gene-trapping vector, pGT1Mpfs, was designed to create an in-frame fusion between the 5′ exons of the trapped gene and a reporter gene,β geo. The ES cells were injected into C57BL/6J blastocysts to create chimeric mice, which bred to generate heterozygous Man1-deficient (Man1^+/Δ) mice. Mouse embryonic fibroblasts (MEFs) were established from embryonic day (E) 16.5 wild-type and Man1^+/Δ embryos using standard procedures (Nagy et al.,2003).

Genotypes of embryos older than E9.5 were determined by Southern blotting. Genomic DNA isolated from tails, yolk sacs and embryos were digested with PstI and hybridized with a Man1 probe located upstream of the insertion site. The probe was amplified from mouse genomic DNA with primers 5′-GCGCTGGGTTACTTTGTGTGCTG-3′ and 5′-GCTTCCCGTTCACCACACTTCTG-3′. Signals were visualized with the Gene Image Random-Prime Labeling and Detection System (Amersham).

Genotypes of embryos from E7.0-E9.0 and gene expression profiles of wild-type and mutant embryos were analyzed by RT-PCR. Total RNA was isolated from yolk sacs, ectoplacental cones and embryos with the High Pure RNA Tissue Isolation kit (Roche). First-strand cDNA was synthesized with Superscript Reverse Transcriptase (RT) III (Invitrogen). Primer sequences and PCR conditions are presented in Table 1.

Whole-mount in situ hybridization of mouse embryos was performed as described previously (Yamaguchi et al.,1993) with the exception that hybridization was performed at 70°C. Digoxygenin (DIG)-labeled antisense RNA probes were synthesized with the DIG RNA Labeling mix (Roche).

X-gal staining of heterozygous embryos was performed using standard procedures (Nagy et al.,2003). For whole-mount immunohistochemistry, E9.5 yolk sacs and embryos were fixed in 4% paraformaldehyde (PFA) in phosphate-buffered saline followed by dehydration. After blocking, the specimens were incubated with an anti-Pecam-1 antibody (MEC13.3; BD Pharmingen; 1:400) and then treated with a horseradish peroxidase (HRP)-coupled anti-rat IgG antibody (Biosource, 1:500)as a secondary antibody. The signals were detected with 0.3 mg/ml 3,3′-diaminobenzidine (DAB) containing 0.8 mg/ml nickel chloride.

For section immunohistochemistry, dissected embryos were fixed in 4% PFA,dehydrated, embedded in paraffin and sectioned. Antibodies to smooth muscleα-actin (1A4; Dako; 1:500), phospho-histone H3 (Sigma; 1:2000),activated caspase 3 (1:2000), or ssDNA (Dako; 1:400) were used as primary antibodies. Vecta stain Elite peroxidase kit (Vector) and the ImmunoPure Metal Enhanced DAB Substrate kit (Pierce) were used for visualization

For immunofluorescence assays, paraffin-embedded sections were treated with the following primary antibodies: anti-fibronectin (Sigma; 1:300),anti-phospho-Smad1/5/8 (Cell Signaling; 1:100) and anti-phospho-Smad2 (Cell Signaling; 1:500). After washing, the sections were incubated with an Alexa 546-conjugated anti-rabbit IgG antibody (Invitrogen; 1:2000). Nuclei were visualized with 4,6-diamidino-2-phenylindole (DAPI). Images were obtained using a confocal microscope (model LSM 510; Carl Zeiss).

E9.5 embryos were fixed with 3% glutaraldehyde in 0.1 M sodium cacodylate buffer at pH 7.4 and post-fixed with 1% osmium tetroxide in the same buffer. After dehydration, they were substituted by propylene oxide and embedded in epoxy resin. Ultrathin sections were doubly stained with uranyl acetate and lead citrate, and examined with an H-7650 transmission electron microscope(Hitachi).

E9.5 embryos were dissected out and yolk sacs were removed for genotyping. Dissected embryos and MEFs were lysed in modified RIPA buffer (50 mM Tris-HCl at pH 7.4, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 1 mM EDTA supplemented with phosphatase and protease inhibitors). Lysates were cleared by centrifugation and quantitated with the BCA Protein Assay Kit (PIERCE). The lysates (20 μg) were separated on a 5-10% gradient gel (BioRad). The following primary antibodies were used: anti-β-galactosidase (Promega;1:1000), anti-N-terminal human MAN1 (Santa Cruz Biotechnology; 1:200),anti-fibronectin (Sigma; 1:1000) and anti-β-tubulin (Sigma; 1:5000).

Although recent cell culture data indicate that Man1 inhibits Tgfβ1 signaling in mammalian cells by interacting with the Tgfβ1/activin-responsive Smad proteins Smad2 and Smad3(Hellemans et al., 2004; Lin et al., 2005; Pan et al., 2005), it is not clear whether this is also the case in vivo. To address this, we first examined whether a Xenopus ortholog of MAN1, XMAN1, inhibits activin/nodal signaling during Xenopus development. Chordin is expressed in the Spemann organizer at the early gastrula stages and is a downstream target for activin/nodal signaling(Sasai et al., 1994). When XMAN1 mRNA was injected into the future organizer region, the expression of chordin was almost completely suppressed(Fig. 1A). Conversely, delivery of the antisense morpholino oligonucleotides (MO), which specifically inhibit XMAN1 functions in vivo (Osada et al.,2003), caused a large expansion of the expression domain for chordin, whereas it was unchanged in embryos injected with four-base-mismatched MO.

The above results were supported by animal cap (ectodermal explant) assays. Xenopus nodal-related factor 1 (Xnr1) induced mesodermal markers(Xbra, chordin and goosecoid) in animal caps as previously described (Jones et al.,1995). Wild-type XMAN1 and the C-terminal fragment of XMAN1(XMAN1-CT) containing the Smad-interacting domain, suppressed the induction of these markers by Xnr1, but a mutant lacking the CT (XMAN1-ΔCT) did not(Fig. 1B).

Xnr1 expression is autoregulated through an activin/nodal response element composed of three Foxh1 (FAST-1)-binding sites present in the first intron(Osada et al., 2000). We performed a reporter assay with a luciferase construct (Int1) containing this element to study whether XMAN1 inhibits Xnr1-dependent transcription. As shown in Fig. 1C, the activation of Xnr1-dependent Int1 was suppressed by XMAN1 and XMAN1-CT, but not by XMAN1-ΔCT. We also found that XMAN1 could directly bind to Smad2 and Smad3 via its C-terminal region (Osada et al., 2003) (data not shown) in the Xenopus embryo,consistent with the works using cultured cells(Hellemans et al., 2004; Lin et al., 2005; Pan et al., 2005). These data indicate that XMAN1 negatively regulates activin/nodal signaling in addition to Bmp signaling in vivo.

Although Man1 loss-of-function analyses have been performed in C. elegans (Liu et al.,2003), Xenopus (Osada et al., 2003) and cultured mammalian cells(Hellemans et al., 2004; Pan et al., 2005), the consequence of the complete loss of functional Man1 in whole organisms is unknown, even in human disease states. To obtain insights into the requirement for Man1 in mammalian development and cellular functions, we generated Man1-deficient mice using an embryonic stem cell line in which Man1 had been disrupted by gene trapping. A DNA sequence tag generated by 5′ rapid amplification of cDNA ends analysis revealed that the insertional mutation in Man1 occurred in intron 4(Fig. 2A). Embryos were genotyped by RT-PCR at E7.0-E9.0 with Man1 and βgeoprimer pairs (Fig. 2B) and by Southern blots at E9.5 and later with a probe located upstream of the insertion site (Fig. 2C). We amplified a fusion transcript composed of the fourth exon of Man1, an Engrailed-2 splice acceptor and full-length βgeo from Man1-deficient embryos (Fig. 2D). The deduced gene product of this transcript indicates that a Man1-βgeo fusion protein lacking the C-terminal Smad-interacting domain was generated in Man1-deficient embryos. We refer to the genetrap allele as Man1Δ hereafter.

We determined whether any wild-type Man1 protein was made from the genetrap allele by western blotting. An antibody to Man1 recognized a ∼100 kDa band, which was comparable with exogenously overexpressed human MAN1 (not shown), in lysates prepared from wild-type and Man1^+/Δ embryos and MEFs, but not in those from Man1^Δ/Δ embryos(Fig. 2E). Conversely, a band of ∼230 kDa corresponding to the predicted Man1-βgeo fusion protein was detected by an antibody to β-galactosidase in the Man1^+/Δ and Man1^Δ/Δspecimens, but not in the wild-type ones. Thus, the Man1-βgeo fusion protein in Man1^Δ/Δ embryos is unlikely to retain regulatory activity on Smad signaling.

We analyzed the temporal and spatial expression domains for Man1 during early embryogenesis by whole-mount X-gal staining in Man1^+/Δ embryos. lacZ expression was detected in the epiblast, the allantois and the ectoplacental cone at E7.5(Fig. 2L), and in the heart,the branchial arches, the foregut, the paraxial mesoderm, the sinus venosus and the tail mesenchyme at E8.5 (Fig. 2N). No signals were detected in a wild-type littermate embryo(Fig. 2K). As for the vascular system, lacZ was detected in visceral endodermal, mesodermal and endothelial cells in the yolk sac (Fig. 2Q), and endothelial and vascular smooth muscle cells (VSMCs) in the dorsal aorta (Fig. 2R). lacZ was also observed in a wide range of tissue/cell populations in the adult (see Fig. S1 in the supplementary material). These lacZ-positive tissues correspond well to the localization of Man1 transcripts (Fig. 2F-J,M,O), suggesting that the βgeo gene was under the control of endogenous Man1 promoter. Man1 is also strongly expressed in the head mesenchyme at E9.0(Fig. 2P). An antisense probe specifically hybridizes to the sequence covering the C-terminal region (P2)gave no signals in Man1^Δ/Δ embryos(Fig. 2J), supporting further that the C-terminal transcript was not expressed in Man1^Δ/Δ embryos.

Man1^+/Δ mice are normal in terms of growth,fertility and lifespan. Although humans with heterozygous LEMD3mutations exhibit prominent bone and connective tissue phenotypes(Hellemans et al., 2004), we did not detect any abnormalities in bone density by magnetic resonance imaging or connective tissue nevi by skin histology in Man1^+/Δ mice (data not shown). No homozygous mice(Man1^Δ/Δ) were recovered at weaning. To elucidate further the apparent embryonic lethal phenotype, we analyzed the offspring of heterozygote intercrosses at various stages of development. Although Man1^Δ/Δ embryos at E7.5, E8.5, E9.5 and E10.5 were recovered at the predicted mendelian frequencies (see Table S1 in the supplementary material), the number of Man1^Δ/Δ embryos at E11.5 was lower than predicted. No homozygous mutants were recovered at E12.5 or later; thus, Man1^Δ/Δ embryos died at about E10.5-E11.5.

Members of the TGFβ superfamily play a crucial role in mesoderm formation and patterning of the embryo(Chang et al., 2002; Whitman, 1998). We first examined the possibility that Man1^Δ/Δ embryos die from impaired mesoderm formation by performing whole-mount in situ hybridization. Because Man1^Δ/Δ embryos showed a 0.5- to 1.0-day delay in development, we compared the mutant embryos with stage-matched wild-type embryos. We found no obvious differences in the expression domains for the mesodermal markers Nodal(Fig. 3A; ^+/+, n=6; ^Δ/Δ, n=4), chordin(Fig. 3B; ^+/+, n=7; ^Δ/Δ, n=4), Bmp4(Fig. 3C; ^+/+, n=5; ^Δ/Δ, n=5) and brachyury(Fig. 3E; ^+/+, n=2; ^Δ/Δ, n=3), between the wild-type and mutant embryos, although we observed consistently stronger intensity of expression of Nodal and chordin in the mutants. Nodalnormally shows an asymmetric expression in the left lateral plate mesoderm(LPM) at E8.25 (Lowe et al.,1996). However, in the mutant embryos, Nodal was ectopically expressed in the right LPM(Fig. 3A′; ^+/+, n=2; ^Δ/Δ, n=5),suggesting that Nodal expression is deregulated in Man1^Δ/Δ embryos.

Man1 deficiency did not grossly affect the anteroposterior patterning of the central nervous system or axial midline, as assessed by examining the expression of cerberus-related 1(Fig. 3D; ^+/+, n=2; ^Δ/Δ, n=2), Otx2 and Krox20 (Fig. 3F; ^+/+, n=3; ^Δ/Δ, n=5), Emx2 (Fig. 3G; ^+/+, n=4; ^Δ/Δ, n=3), and sonic hedgehog (Fig. 3H; ^+/+, n=5; ^Δ/Δ, n=5).

At E10.0, Man1^Δ/Δ embryos were recognizable by their smaller and pale yolk sacs (Fig. 4A,B). Histological analysis of wild-type and mutant yolk sacs revealed that the yolk sac vasculature was formed in the mutants but appeared highly dilated (Fig. 4C,D). Some mutant embryos died in the middle of turning(Fig. 4F), and others showed clumps of red blood cells and an enlarged pericardium(Fig. 4G), suggesting that vascular permeability and intracardiac pressure may have been elevated in the mutants.

Vascular defects in Man1^Δ/Δ embryos were confirmed by whole-mount immunostaining with an antibody to platelet endothelial cell adhesion molecule 1 (Pecam1), which specifically detects endothelial cells. In the yolk sac of the wild-type embryo at E10.0, major blood vessels and a well-branched capillary network were present(Fig. 5A). By contrast, the yolk sac of the mutant embryo at E10.0 displayed a primary capillary plexus but no branching from pre-existing vessels(Fig. 5B). No significant difference in endothelial cell proliferation was observed between wild-type and mutant yolk sacs (data not shown). Moreover, in the wild-type embryo proper, well-developed and defined capillary networks were apparent in the head and intersomitic regions (Fig. 5C,E,G), whereas the fine capillary network in the cranial,branchial, cardiac and intersomitic regions of mutant embryos was absent(Fig. 5D,F,H). These data imply that endothelial cell differentiation occurred normally in Man1^Δ/Δ embryos, but that vascular remodeling to build a mature vascular network was severely perturbed.

Interactions between endothelial cells and mural cells (pericytes and VSMCs) are essential for vascular development(Armulik et al., 2005; Carmeliet, 2000). To determine whether differentiation or recruitment of VSMCs is affected in mutant embryos,we examined the expression of smooth muscle α-actin, a marker for VSMC differentiation, by immunohistochemistry. In E10.0 wild-type embryos, the dorsal aorta was surrounded by VSMCs (Fig. 5I). However, in E10.0 mutant embryos, smooth muscle α-actin expression was barely detected around the dorsal aorta(Fig. 5J), although its expression was detected in other tissues, including the heart (data not shown). Electron microscopic analysis supported the results of immunohistochemistry (Fig. 5K-P); in wild-type embryos, endothelial cells in the dorsal aorta were surrounded by mural cells and directly in contact with them(Fig. 5K,M), whereas those in mutant embryos were not supported by mural cells, resulting in a collapsed dorsal aorta (Fig. 5L,N). Tight junctions between endothelial cells were formed in both wild-type and Man1^Δ/Δ embryos(Fig. 5O,P). These results indicate that mural cells were differentiated normally, but their recruitment to the vascular wall was severely impaired in Man1^Δ/Δ embryos.

Null mutants of central components of the Tgfβ pathway demonstrate early embryonic death with growth retardation at about E10.0, defects in vascular development, and abnormal VMSC recruitment(Chang et al., 2002; Goumans et al., 2003). As such, there is a remarkable correlation with the defects observed in Man1^Δ/Δ embryos, consistent with the idea that Man1 is required for manifestation of the Tgfβ signal transduction effect. Thus, we examined the activation status of the Smad1/5 and Smad2/3 pathways using a cellular resolution assay by immunohistochemistry with antibodies to phospho-Smad1/5/8 and phospho-Smad2/3. Compared with wild-type embryos, the intensity of nuclear phospho-Smad2 was significantly elevated in mesenchymal, endothelial, and neural cells of Man1^Δ/Δ embryos(Fig. 6I-P, Fig. S3). Quantitative analysis revealed that accumulation of phospho-Smad2 was significantly increased in these cells in the mutants(Fig. 6R). By contrast, the difference in the intensity of phospho-Smad1 between wild-type and mutant embryos was not significant in all cell types examined(Fig. 6A-H,Q, see Fig. S2 in the supplementary material).

Augmentation of the Smad2/3 pathway was confirmed by analyzing the expression levels of its downstream targets by RT-PCR(Fig. 7A). In Man1^Δ/Δ embryos, the expression level of Id1, a specific downstream target for the Smad1/5 pathway in endothelial cells (Goumans et al.,2002; Ota et al.,2002), was comparable with that in wild-type and heterozygous embryos. By contrast, the expression levels of plasminogen activator inhibitor 1 (PAI-1; Serpine1 - Mouse Genome Informatics) and fibronectin 1, both of which are specific downstream targets for the Smad2/3 pathway in endothelial cells, were significantly upregulated. The expression of Smad7, which is a downstream target for both the Smad1/5 and Smad2/3 pathways, was unchanged in Man1^Δ/Δ embryos. We obtained similar results in embryos from another cross (see Fig. S3 in the supplementary material). Notably, the expression level of Tgfb1 was also elevated in the mutants.

Tgfβ1 strongly stimulates synthesis of extracellular matrix (ECM) and deposition of ECM inhibits vascular remodeling(Ignotz et al., 1987; Pepper, 1997). In line with the elevation of Tgfb1 expression and the augmentation of the Smad2/3 signaling, the expression of fibronectin 1 was increased in Man1^Δ/Δ embryos(Fig. 7A). This was further supported by western blots, where fibronectin synthesis was highly elevated(Fig. 7B), and by immunofluorescence analyses, where fibronectin was strongly deposited in Man1^Δ/Δ embryos and yolk sacs(Fig. 7C-H). Occasionally,fibronectin deposition was extremely prominent around endothelial cells(Fig. 7D). These results suggest that abnormal activation of Smad2/3 signaling leads to aberrant deposition of ECM to inhibit migration of endothelial cells.

Because NE proteins are implicated in cell proliferation and apoptosis(Cohen et al., 2001; Gruenbaum et al., 2005), we examined these processes in Man1^Δ/Δ embryos. We first analyzed cell proliferation in Man1^Δ/Δ embryos using immunohistochemistry with an antibody against phospho-histone H3, a marker for cells in mitotic prophase. As shown in Fig. 8A,B, the number of phospho-histone H3-positive nuclei was not significantly different in wild-type and mutant embryos at the 17-somite stage, indicating that cell proliferation was normal in the mutants. Consistent with this, the expression of cyclin-dependent kinase inhibitor 1A(Cdkn1a or p21), a cyclin-dependent kinase inhibitor and mediator of the cellular growth arrest program of the Tgfβ pathway, was unchanged in Man1^Δ/Δ embryos(Fig. 7A). By contrast,immunohistochemistry with antibodies against activated caspase 3 and single-strand DNA (ssDNA, a marker for DNA fragmentation during programmed cell death) revealed massive apoptosis in Man1^Δ/Δ embryos(Fig. 8D,F), especially in mesenchymal tissues where Man1 is strongly expressed(Fig. 2P). Electron microscopic analysis clarified that the mesenchymal cells in the mutants shrunk and their nuclei were condensed and fragmented, which are the typical characteristics of apoptotic cells (Fig. 8J). Wild-type embryos had few apoptotic cells(Fig. 7C,E). Notably, apoptotic cells were seldom detected in other tissues even in Man1^Δ/Δ embryos. These results suggest that Man1 is essential for cell survival, but the degree of cell death is dependent on the different sensitivity to the loss of Man1 between cells.

The INM proteins bind to nuclear lamins, which are essential for maintaining nuclear shape (Gruenbaum et al., 2005); emerin, an INM protein, is required for normal nuclear shape (Lammerding et al.,2005). We next investigated whether the loss of Man1 affects the nuclear morphology by electron microscopy(Fig. 8K-O). The outlines of wild-type mesenchymal nuclei were tense and smooth, whereas the most of mutant mesenchymal nuclei were irregularly shaped, although continuity of their NE was apparently intact. At higher magnification, we found mesenchymal cells with herniated nuclei (Fig. 8M,N) and those with an expanded perinuclear space(Fig. 8O) at low frequency(∼1%) in the Man1^Δ/Δ embryos.

In our previous study of Xenopus embryos(Osada et al., 2003), we have proposed that XMAN1 is a novel type of negative regulator for Smad signaling. A similar hypothesis has also been proposed by four other groups(Hellemans et al., 2004; Lin et al., 2005; Pan et al., 2005; Raju et al., 2003). We have verified this hypothesis in vivo by analyzing Man1-deficient mice lacking the Smad-interacting domain. We have found that Man1^Δ/Δ embryos died during embryonic development because of defects in vascular remodeling. We have demonstrated that abnormal deposition of ECM caused by the augmented Smad2/3 pathway and the disrupted intercellular communication between endothelial and mural cells underlie the perturbed vascular remodeling in Man1^Δ/Δ embryos. Our results have shed light on the importance of Man1 in fine-tuning the activity of Tgfβ signaling during angiogenesis.

The Man1-βgeo fusion generated in Man1^Δ/Δ embryo contained the N-terminal LEM domain and the first transmembrane domain, both of which are necessary and sufficient for targeting Man1 to the INM in mammalian cells(Wu et al., 2002), but lacked the C-terminal Smad-interacting domain. Thus, it is likely that the phenotypes observed in the mutants are caused by abnormal Smad regulation at the INM.

We suspect that deregulation of the Smad2/3 pathway is fundamental to the defects in vascular remodeling in Man1^Δ/Δembryos. We have demonstrated abnormal activation of the Smad2/3 pathway in Man1^Δ/Δ embryos by immunofluorescence microscopy (Fig. 6, see Fig. S2 in the supplementary material), where nuclear accumulation of phospho-Smad2 is increased, and by RT-PCR (Fig. 7A), where the expression of its downstream targets is upregulated. Importantly, the expression of Tgfb1 is elevated in Man1^Δ/Δ embryos. The increased accumulation of nuclear Smad2 in the mutants may lead to the activation of Tgfb1promoter, resulting in Tgfβ1 autoinduction(Kim et al., 1989). Tgfβ1 stimulates the synthesis of fibronectin(Ignotz et al., 1987), which inhibits proliferation and migration of endothelial cells, impairing vascular remodeling (Pepper, 1997). We have shown that the expression of fibronectin is upregulated in Man1^Δ/Δ embryos at the mRNA and protein levels(Fig. 7). In addition, the expression level of Serpine1 is also elevated in mutant embryos. Serpine1 is a potent inhibitor of endothelial cell migration in vitro(Stefansson and Lawrence,1996) and angiogenesis in vivo(Stefansson et al., 2001). Therefore, abnormal deposition of ECM caused by elevated Smad2/3 signaling could underlie the defects in vascular remodeling in Man1^Δ/Δ embryos.

Our quantitative analyses of phospho-Smads have revealed that Smad2/3 signaling is preferentially activated in endothelial cells of Man1^Δ/Δ embryos(Fig. 6R). Tgfβ1 can stimulate two Smad pathways in endothelial cells - the canonical Alk5-Smad2/3 pathway and the Alk1-Smad1/5 pathway(Goumans et al., 2002; Oh et al., 2000). It has been proposed that Alk1-Smad1/5 signaling promotes endothelial cell proliferation and migration and Alk5-Smad2/3 signaling inhibits them(Goumans et al., 2002). Tgfβ and activin also inhibit proliferation and sheet formation of ES cell-derived endothelial cells, probably by activating the Smad2/3 pathway(Watabe et al., 2003). Although an opposite model on the roles of the Alk1 and Alk5 pathways during angiogenesis has been proposed (Lamouille et al., 2002; Oh et al.,2000; Seki et al.,2003), the balance between the two pathways in endothelial cells appears to be crucial in determining the activation state of the endothelium. In Man1^Δ/Δ embryos at E9.5, the expression of specific targets of the Smad2/3 pathway (Serpine1 and Fn1)was significantly elevated (Fig. 7A), whereas that of Id1, a specific target for the Smad1/5 pathway, was unchanged, suggesting that the balance between these two pathways was disrupted in mutant endothelial cells.

It has been proposed that interactions between endothelial and mural cells in the blood vessel wall play an important role in the regulation of vascular remodeling (Armulik et al.,2005). In Fig. 5I-N, we have clearly shown that the recruitment of mural cells to the vascular wall was severely impaired in Man1^Δ/Δ embryos. Because mural cells are generally thought to be mesenchymal origin, we speculated that the mesenchymal apoptosis observed in E9.5 Man1^Δ/Δ embryos(Fig. 8D,E) led to the decreased numbers of VSMCs. However, we detected the elevated expression of trangelin (Tagln) and Acta2(Fig. 7A) and accumulation of SMA-positive cells around the heart in the mutants (data not shown),suggesting that proliferation and differentiation of VSMCs was normal. The activated Tgfβ1-Smad2/3 signaling, which plays an important role in VSMC development, may result in the elevation of these markers. Thus, VSMC migration rather than proliferation and differentiation may be affected in the mutants. The abnormal deposition of fibronectin around blood vessels observed in Man1^Δ/Δ embryos would be also associated with the inhibition of VMSC migration.

Man1 is highly expressed in the head mesenchyme(Fig. 2P), where massive apoptosis was observed in Man1^Δ/Δ embryos. Loss of Ce-MAN1 in C. elegans and siRNA interference of human MAN1 in cultured cells also induce apoptosis (Liu et al., 2003; Pan et al.,2005). Although these results suggest that the mesenchymal apoptosis in Man1^Δ/Δ embryos is probably a direct consequence of the loss of Man1, the elevated Tgfb1 expression may be implicated in initiation or deterioration of the apoptotic processes. We have clarified that most of the apoptotic cells in the mesenchyme are caspase 3-dependent (Fig. 7D),consistent with the observation that caspase 3 is activated during Tgfβ1-induced apoptosis in many types of cultured cells(Schuster and Krieglstein,2002). In addition, several lines of evidence have shown that Smad3 is an important mediator of Tgfβ1-induced apoptosis(Kim et al., 2002; Yamamura et al., 2000; Yanagisawa et al., 1998). Thus, the elevated Smad2/3 pathway in the mutants may also be involved in the Tgfβ1-induced caspase 3-dependent apoptosis.

Nuclear lamins and the INM proteins play a key role in maintaining nuclear morphology (Burke and Stewart,2002). We observed nuclear anomalies, especially in mesenchymal cells of Man1^Δ/Δ mutants(Fig. 8L-O), although we seldom found the cells with discontinuous NE. It remains to be elucidated whether the nuclear abnormalities in the mutants were caused by the abnormal regulation of Smad signaling at the INM, or by the presence of structurally aberrant Man1 with the long C-terminal tail of βgeo, and they are associated with the increased mesenchymal apoptosis. Emerin-deficient MEFs show abnormal nuclear shape and increased apoptosis, but impaired mechanotransduction rather than nuclear fragility induces their apoptosis(Lammerding et al., 2005).

Nodal expression in the node and left LPM during early embryogenesis plays a crucial role in determining the left-right axis of the embryo (Hamada et al., 2002; Lowe et al., 1996). In Man1^Δ/Δ embryos, initiation of Nodalexpression and gastrulation occurred normally. However, asymmetric Nodal expression was impaired to become expressed bilaterally in mutant embryos (Fig. 3A′). It would be intriguing to examine whether Man1 is involved in establishing left-right asymmetry through the regulation of Nodal expression.

In Man1^Δ/Δ embryos, the overall anteroposterior patterning of the embryo proper and the initiation of the expression of neural markers are normal(Fig. 3), which is consistent with previous data in XMAN1 morphants(Osada et al., 2003). However,in contrast to the XMAN1 morphants that show eyeless phenotypes with anterior truncations, the anterior structures are relatively normal in Man1mutants. Early embryonic lethality at about E10.5 in Man1^Δ/Δ embryos prevented us from examining the role of Man1 in neural development. Generation of conditional Man1 mutant mice would be useful for analyzing the cell- and tissue-specific roles of Man1 more precisely.

In cultured cells, depletion of human MAN1 leads to elevated responsiveness to both TGFβ and BMP (Hellemans et al., 2004; Lin et al.,2005; Pan et al.,2005). Accordingly, the sensitivity to both activin/nodal and Bmp stimulation was expected to be elevated in the setting of Man1deficiency. The mechanism by which the Smad2/3 pathway was preferentially augmented in Man1^Δ/Δ embryos is currently unknown. Intracellular antagonism between Bmp4 and activin signaling through competition for a limited pool of Smad4 has been proposed(Candia et al., 1997). Thus, it is possible that the Smad2/3 pathway activated by continuous Tgfβ1 stimulation consumed more Smad4 than the Smad1/5 pathway, leading to the dominance of the Smad2/3 pathway in Man1 mutants.

We did not detect overt morphological defects caused by the abnormal activation of the Bmp-Smad1/5 pathway in Man1^Δ/Δ embryos, aside from the fact that the allantois was enlarged in the mutant embryos(Fig. 3E). Bmp signaling plays a crucial role in the development of the allantois(Fujiwara et al., 2001). It would be important to examine the development of tissues/cell population where the involvement of the Bmp-Smad1/5 pathway is implicated, such as the formation of the heart and primordial germ cells, and the dorsoventral patterning of the neural tube (Hogan,1996).

We thank Dr N. Osumi for instructions on mouse dissection; Drs Y. Suda, H. Sasaki, A. Shimono and R. Behringer for mouse in situ probes; Dr H. Worman for the MAN1 expression plasmid; Drs K. Hamada and M. Nakaya for the experimental protocols; Drs N. Inagaki, K. Hasegawa and N. Ban for the reagents; Drs M. Shibuya and Y. Sakurai for the PCR primer sequences; Dr T. Momoi for the antibody to activated caspase 3; Drs C. Wright, T. Watabe and K. Kawamura for helpful suggestions and discussions. We also thank S. Hatazawa, H. Watanabe and S. Chida for technical assistance; and A. Takahashi for assistance in performing the in situ hybridization. This work was supported by Grants-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports,Science and Technology of Japan; by a grant from the Kanae Foundation for Life and Socio-Medical Science to S.O.; and by an NHLBI-funded Program in Genomics Applications [`BayGenomics' (HL66621, HL66600 and HL66590)].