Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
Vacuolar H+-ATPases play an important role in maintaining the pH of endomembrane compartments in eukaryotic cells. The functional relevance of this homeostasis for multicellular development has not been studied in plants. Here, we analyze the biological consequences resulting from the lack of subunit E isoform 1 (VHA-E1) encoded by the Arabidopsis TUFF gene. tuff mutant embryos are lethal, displaying variably enlarged cells with multiple nuclei, large vacuoles containing inclusions, abnormal organization of Golgi stacks, and cell wall defects. Rescue of embryo lethality by cell cycle-regulated expression of VHA-E1 results in abnormal seedlings with non-functional meristems and defective cell differentiation. VHA-E1 is the predominant isoform in embryogenesis whereas VHA-E3 is expressed mainly in the endosperm and surrounding maternal tissues during seed development, and VHA-E2 is pollen-specific. VHA-E1 protein accumulates at endomembrane compartments including vacuoles and endosomes, but appears absent from the plasma membrane. Our results suggest an essential role for VHA-E1 in maintaining a functional secretory system during somatic development but not in the haploid gametophytes.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Endomembrane compartments of eukaryotic cells maintain a lower pH value relative to the cytosol, which is required for certain aspects of membrane trafficking as well as for transport of small molecules across membranes (Nishi and Forgac, 2002). The vacuolar H+-ATPase (V-ATPase) carries out ATP-dependent proton transport from the cytosol into the lumen of endomembrane compartments. V-ATPases are composed of a membrane-anchored V0 domain involved in H+ transport and a peripherally associated V1 domain catalyzing ATP hydrolysis. V0 and V1 domains consist of subunits a-e and A-H, respectively, with subunit E contributing to the peripheral stalk that connects the peripheral with the membrane-anchored domain. The structure of V-ATPases appears to be conserved across eukaryotes (Domgall et al., 2002; Nishi and Forgac, 2002). However, plant V-ATPases appear to be divergent in subunit composition, as indicated by the capacity of plant genomes to encode multiple isoforms of several subunits (Kluge et al., 2003). In Arabidopsis, V1 subunits B, E and G are each encoded by three different VHA genes, and multiple isoforms exist for each of the five V0 subunits (Sze et al., 2002). The significance of this genetic diversity is not known but may be related to tissue-specific regulation or response to environmental stress (Kluge et al., 2003). Only one study has addressed potential roles of V-ATPase subunit isoforms in plants genetically. Arabidopsis VHA-c1 and c3 subunit isoforms have been knocked down by RNAi, each resulting in reduced root length and decreased tolerance to moderate salt stress (Padmanaban et al., 2004). This similarity of defects, which suggests functional relatedness of these two VHA-c genes, may reflect their overlapping expression patterns in the root cap, as revealed by promoter-GUS fusions.
Here we analyze the three VHA-E genes of Arabidopsis, which display distinct developmental expression patterns. VHA-E1 encoded by the TUFF (TUF) gene is the major isoform during embryogenesis. tuf mutations impair Golgi organization, vacuole function and cell wall deposition, resulting in embryo lethality, whereas haploid gametophytes are unaffected. Our results demonstrate that multiple genes encoding V-ATPase subunit isoforms can serve divergent roles in plant development.
Two mutations in the Arabidopsis TUF gene, EMS-induced allele UU2573 and T-DNA insertion S08 (SALK 087937; Alonso et al., 2003), caused embryo lethality and displayed very similar embryo phenotypes (Figure 1). In contrast to wild-type embryos, tuf mutant embryos failed to undergo morphogenesis (compare Figure 1B,D with A,C). The mutant cells were variably enlarged, multinucleate and contained large vacuoles that accumulated inclusions staining pink with the pH-sensitive dye toluidine blue (Figure 1B,D). Cell wall deposition was also affected, as evidenced by thin cell walls, local thickening and cell wall stubs (Figure 1E). In addition, Golgi stacks were horseshoe-shaped, with large membrane vesicles accumulating nearby, which indicated that their organization was altered (compare Figure 1G,I with F,H). These observations suggested that the TUF gene plays an essential role in maintaining a functional secretory system required for normal embryogenesis.
The TUF gene was isolated by map-based cloning (Figure 2). Following the initial mapping to a small genomic interval of chromosome 4, analysis of a larger population narrowed the critical region down to six predicted open reading frames (ORFs; Figure 2A). Sequencing of ORF At4g11150 from the EMS-induced mutant UU2573 revealed a G–A mutation at the splice-acceptor site of exon 4, suggesting that this ORF corresponded to the TUF gene (Figure 2B). This inference was confirmed by the analysis of T-DNA insertion line S08, which failed to complement the embryo lethality of the EMS mutant UU2573. The T-DNA was inserted in intron 1 of At4g11150 (Figure 2B). In addition, a 5.3-kb-long genomic fragment encompassing At4g11150 and adjacent intergenic sequences rescued the lethality of UU2573 (data not shown). The TUF gene encodes VHA-E1, one of three closely related isoforms of subunit E of vacuolar H+-ATPase in Arabidopsis (Figure 2D). RT-PCR analysis of mRNA extracted from UU2573 heterozygous plants yielded two abnormal TUF mRNA variants. About 70% of the mRNAs had a 49-bp deletion, resulting in a predicted premature stop codon. The truncated protein would consist of the N-terminal 136 of 230 amino acid residues of TUF protein plus four additional amino acid residues (Figure 2C, asterisk). About 30% of the mRNAs had a shorter ORF that would translate into a protein with an internal 17 amino acid deletion (Figure 2C, underlined).
To determine whether tuf mutant embryos accumulated detectable levels of mutant protein, wild-type and UU2573 mutant embryos were stained with antiserum raised against VHA-E from barley (Betz and Dietz, 1991). Wild-type embryos displayed endomembrane labeling whereas no specific signal was detected in tuf mutant embryos, indicating that the point mutation interfered with stable accumulation of TUF protein, and both genotypes displayed the control label for the plasma membrane H+-ATPase (Figure 3A–F). This result was confirmed by immunogold labeling of cryosectioned embryos (Figure 3G,H). The small vacuoles of wild-type embryos displayed strong VHA-E labeling whereas no such labeling was detected in tuf mutant embryos. These observations suggest that the tuf mutant phenotype results from the complete absence of VHA-E1.
The majority of TUF protein was detected at vacuolar membranes although some other endomembranes were also labeled (Figure 3G). We used a genomic TUF-GFP fusion that rescued tuf mutant plants to determine the subcellular localization of VHA-E1 in seedling root cells, which lack large central vacuoles (Figure 3I–N). TUF-GFP labeled endomembrane compartments whereas much less signal was detected at the plasma membrane, as shown by double labeling with a specific antibody for plasma membrane-localized H+-ATPase (PM-ATPase; Figure 3I–K). Treatment of seedling roots with the trafficking inhibitor brefeldin A (BFA) resulted in internalization of PM-ATPase, as previously reported (Geldner et al., 2003). These endosomal BFA compartments were also labeled by TUF-GFP, whereas adjacent patches were labeled by TUF-GFP but devoid of PM-ATPase labeling (Figure 3L–N; arrows). Additional large patches that were specifically labeled by TUF-GFP may represent small vacuoles characteristic of young root cells (Figure 3L–N). Thus, some TUF protein accumulated in endosomal BFA compartments and possibly also in surrounding Golgi stacks. However, direct double labeling for VHA-E and gamma-COP, a marker for cis-Golgi, showed no obvious overlap of the signals (data not shown).
The Arabidopsis genome contains two additional genes (At3g08560, At1g64200) encoding isoforms 2 and 3 of VHA-E (Figure 2D; Sze et al., 2002). Genomic GFP fusions of all three isoforms were generated and analyzed for expression in transgenic plants (Figure 4). TUF/E1 was strongly expressed in the developing embryo but not in the endosperm and only at low level in the surrounding maternal tissue (Figure 4A–C). E3 displayed a nearly complementary expression pattern during seed development: no or low-level expression in the developing embryo but strong expression in the endosperm and in the surrounding maternal tissue (Figure 4D–F). During post-embryonic development, E3 was expressed at low level mainly in the epidermis whereas expression of E2 was pollen-specific (J. Dettmer and K. Schumacher, unpublished data). During pollen development, TUF/E1 was initially expressed in the microspore but subsequently its expression was confined to the two sperm cells (Figure 4G). In contrast, E2 was expressed in the vegetative cell (Figure 4H), and E3 was expressed in both the vegetative cell and the sperm cells (Figure 4I). These results indicate that TUF/VHA-E1 is the major isoform in the developing embryo. Its expression in the sperm cells but not in the vegetative cell also provides an explanation for the fact that TUF/VHA-E1 is not required for the development of the haploid gametophyte, but only after fertilization in the developing embryo. The zygotic function of TUF/VHA-E1 contrasts with the requirement of VHA-A during gametophytic development (see accompanying paper by Dettmer et al., 2005).
The embryo lethality of tuf mutants precluded the analysis of post-embryonic defects resulting from the loss of VHA-E1 activity. To circumvent this problem, Myc-tagged TUF was expressed under the cis-regulatory control of the cell cycle-regulated KNOLLE (KN) gene in the tuf mutant background (Lukowitz et al., 1996; Müller et al., 2003). In six independent transgenic lines analyzed, the KN::Myc-TUF transgene rescued the lethality of tuf embryos (data not shown). However, in all six lines, the resulting seedlings were defective and did not develop into adult plants, although Myc-TUF protein was expressed and accumulated in endomembranes (Figure 5). All seedling organs were stunted. The cotyledons were roundish and yellowish-green, the hypocotyl was short and inflated, and the root was shortened and almost lacked root hairs except for a few at the hypocotyl–root junction (Figure 5A–E). Primordia of primary leaves were present but did not develop further (Figure 5B). The root meristem was reduced and disorganized as was the root cap (Figure 5E). Vascular strands were incompletely differentiated in the cotyledons and in the hypocotyl (Figure 5A,C). The hypocotyl appeared flattened, consisting of fewer internal cell layers, which moreover did not display the well-formed tissue organization characteristic of the wild-type hypocotyl (Figure 5F,G). The outer cell layers consisted of drastically shortened and enlarged cells, when compared with wild type. To determine whether cell elongation was affected, seedlings were grown in the dark. Hypocotyl elongation was observed in the transgenic seedlings, although to a much lesser extent than in wild-type seedlings (Figure 5H). These observations suggest that reduction in VHA-E1 activity impairs cell differentiation, which is comparable to the det3 phenotype resulting from reduced VHA-C activity (Schumacher et al., 1999).
Our results demonstrate that V-ATPase plays an essential role in plant development and also provide evidence for functional divergence among the three Arabidopsis genes encoding different isoforms of the same subunit, VHA-E. tuf mutant embryos are lethal and lack VHA-E1, which is the major isoform during embryogenesis. By contrast, VHA-E3 mainly accumulates in the endosperm and surrounding maternal tissues of the developing seed, and VHA-E2 is pollen-specific. This differential expression of VHA-E genes also suggests why tuf mutations do not cause gametophyte lethality, although a mutation in the single-copy gene VHA-A indicates that V-ATPase is essential in pollen development (see accompanying paper by Dettmer et al., 2005). Presumably the other isoforms, VHA-E2 and VHA-E3, provide the necessary V-ATPase subunit E activity in pollen and embryo sac.
Isoforms of V-ATPase subunit E have been detected in several plant species (Kawamura et al., 2000; Kluge et al., 2003). In pea, immunologically distinguishable isoforms differ in their tissue specificity. E2 is present in all organs whereas E1 is absent from cotyledons of etiolated seedlings and barely detectable in green leaves of mature plants (Kawamura et al., 2000). In the salt-tolerant plant Mesembryanthemum crystallinum, VHA-E expression is complex and appears adapted to the export of salt from the root to the shoot for storage in the vacuole. In response to salt, transcript levels increase in the leaves and decrease in root cells (Golldack and Dietz, 2001). Although it has not been reported which of the three subunit E isoforms was studied, these observations support the notion that different V-ATPase complexes may be involved in different physiological or developmental conditions (Kluge et al., 2003).
Several lines of evidence indicate that the lethality of tuf embryos is caused by the absence of VHA-E1 protein, as visualized by immunofluorescence staining. Two different tuf alleles disrupting the VHA-E1 gene displayed essentially the same phenotype. In addition, another embryo-lethal mutant, emb2448, has recently been reported to carry a T-DNA insertion in the VHA-E1 gene but no phenotypic analysis was provided (Tzafrir et al., 2004). Furthermore, a genomic fragment of TUF/VHA-E1 fully restored normal development of tuf mutants. Finally, cell cycle-regulated expression of VHA-E1 from the KN::Myc-TUF transgene also rescued the embryo lethality. It is currently not known why the transgenic embryos gave rise to abnormal seedlings, which did not develop further. Nonetheless, the partial rescue activity of the transgene confirms that the lethality of tuf embryos is exclusively caused by the loss of VHA-E1 gene function.
Lethal embryos lacking VHA-E1 display a range of defects, such as variably enlarged, multinucleate cells, cell wall stubs and abnormal division planes, which have also been observed in cytokinesis-defective mutants knolle and hinkel (Lukowitz et al., 1996; Strompen et al., 2002). Similar defects occur in cyt1 mutants, which are defective in N-glycosylation required for cellulose biosynthesis (Lukowitz et al., 2001; Nickle and Meinke, 1998). tuf mutant embryos are barely distinguishable from wild-type embryos during early stages, suggesting a rather general primary defect. It is conceivable though that early development of tuf embryos is sustained by VHA-E protein carried over from the female gametophyte such that development becomes progressively abnormal as this supply is depleted.
What do the morphological aberrations of tuf embryos suggest about the role of V-ATPase in embryo development? As shown here, tuf mutant cells display specific subcellular defects, such as vacuolar inclusions, abnormal Golgi morphology and vacuolation, and defective cell wall deposition, suggesting that the secretory system is compromised. Consistent with this, VHA-E1 accumulates not only in the vacuolar membrane but also in other endomembranes including endosomes and possibly Golgi stacks. V-ATPase has been previously localized to Golgi stacks and shown to be displaced to Golgi-derived swollen vesicles upon treatment of sycamore cells with the ionophore monensin (Zhang et al., 1996). Golgi vacuolation has also been observed in tobacco BY-2 cells treated with V-ATPase inhibitors concanamycin A and bafilomycin A (Robinson et al., 2004). In addition, concanamycin A-treated tobacco cells display missorting of soluble vacuolar protein precursors that are normally transported from the trans-Golgi compartment (Matsuoka et al., 1997). Comparable morphological Golgi aberrations have also been observed in pollen tubes treated with concanamycin A as well as in pollen of the Arabidopsis vha-A mutant, which is defective in the single gene encoding subunit A of V-ATPase (see accompanying paper by Dettmer et al., 2005). Thus, essentially the same Golgi defect is caused by mutations in two different V-ATPase subunit genes of Arabidopsis and by chemical inhibition of V-ATPase activity. Taken together, the available evidence suggests that elimination of V-ATPase activity in embryogenesis compromises the secretory system, which gradually impairs a variety of cellular processes, including cell division, and eventually results in grossly abnormal lethal embryos.
Plant material and growth conditions
The tuf allele UU2573 was induced by EMS mutagenesis in the Arabidopsis thaliana ecotype Landsberg erecta (Ler). The T-DNA insertion allele S08 (SALK 087937; http://signal.salk.edu/cgi-bin/tdnaexpress) was obtained from the Arabidopsis Biological Resource Center. Plants were grown as previously described (Mayer et al., 1993). For BFA treatment, seedlings were incubated in 50 μg ml−1 BFA for 1.5 h. Hypocotyl elongation was determined on seedlings grown on 1% agar at 23°C for 10 days, with agar plates exposed to light or wrapped in aluminum foil.
Isolation and molecular identification of the TUF gene
For molecular mapping, 500 F2 plants from a cross of the tuf allele UU2573 to Columbia (Col) wild-type plants were examined. Genomic DNA for PCR amplification was prepared as described (Lukowitz et al., 1996). The markers nga8 and nga12 were used to localize the TUFF gene to chromosome 4 (http://www.arabidopsis.org). Additional markers for fine-mapping (deposited with TAIR; http://www.arabidopsis.org) were identified by sequencing PCR-amplified Ler genomic DNA from the region depicted in Figure 2(A). The TUF candidate gene, At4g11150, was sequenced from Ler and the mutant allele UU2573, using the following primers: TUFF/F1, ACGAAAATCTCTGAAACCTCC; TUFF/F2, CTCTAATTCAATCGTAAGCGG; TUFF/F3, GTTATCGTGTATTACTTTGCC; TUFF/F4, GGAGATTATGTAATCGTGTCG; TUFF/F5, CATTATCAGTGTTTAGATCCGG; TUFF/B1, GTAAATAACCAAACGCAAACCG; TUFF/B2, GAGTTTCATACGGAATGCCAC; TUFF/B3, GACATCAGCAGAACGGATC; TUFF/B4, CATTGACAATATCGTCTTGCG; TUFF/B5, GATATTGTAATCCAGAACAAGC. The T-DNA insertion S08 was detected by PCR amplification with the primers TL2, CAATCAGCTGTTGCCCGTCTCAC, and TUFF/B5, GATATTGTAATCCAGAACAAGC, which yielded a PCR product of 600 bp.
Sequence data were analyzed with MacVectorTM (Kodak Scientific Imaging Systems, New Haven, CT, USA). blast searches were carried out using the TAIR database (http://www.arabidopsis.org) and the databases at the NCBI (http://www.ncbi.nlm.nih.gov). Protein alignments were generated with MegAlign version 3.1.7 (DNA Star Inc., Madison, WI, USA).
For analysis of splicing, mRNA was extracted from UU2573 heterozygous plants and amplified by RT-PCR, using the OneStep RT-PCR kit (Qiagen, Venlo, the Netherlands) according to the manufacturer's instructions. PCR amplification with the primers TUFF/F3neu, GATCTTCTGAACGTTAGTCGCG, and TUFF/B2, GAGTTTCATACGGAATGCCAC yielded a 360-bp wild-type cDNA fragment and two mutant fragments that were shorter by 51 and 49 bp, respectively. The same primers were also used for sequencing the PCR products.
For complete rescue of the UU2573 mutant, a 5.3-kb genomic TUFXhoI fragment from BAC clone T22B4 was ligated into the SalI site of the pJH212 vector (Hajdukiewicz et al., 1994) and transformed into UU2573 heterozygous plants. Transgenic lines were selected by kanamycin resistance. Rescued UU2573 homozygous mutant plants were identified by PCR amplification with primers TUFF/F0, CCGCTCGCACCGTTAGTCC, and TUFF/B2.2, CAGTGGAGACCATGAGGATC, yielding a 1.7-kb product, which was cleaved by BseMI into 1.5 kb and 200 bp fragments in UU2573 but not in wild-type DNA (see Figure 2B). The TUF transgene was detected by PCR amplification with primers B0neu, CATTGCATAGTCTCTCGCG, and pJH212.rev, GTTGTAAAACGACGGCCAGTGCC, yielding a 300-bp product.
The construct KN::Myc-TUF for transformation of UU2573 heterozygous plants was generated as follows. The TUF coding sequence was amplified from an embryo-enriched cDNA library (Grebe et al., 2000), using the primers TUFF cDNA/F, ATATCCCGGGATGAACGACGGAGATGTATCG, and TUFF cDNA/B, ACCTGAATTCTCAGGCAGTAACTTGGCCGAAC. The 693 bp PCR product was digested with XmaI and BamHI, cloned into the KN90MYC cassette in the pGREEN vector, transformed into Escherichia coli XL1 (selected on 50 μg ml−1 kanamycin) and sequenced. The KN90MYC cassette contains 2.5 kb KN promoter, the coding sequence for the Myc epitope, restriction sites for directional in-frame cloning, and 1 kb KN 3′UTR (Müller et al., 2003). Plants were transformed by floral dipping with Agrobacterium tumefaciens GV3101 selected on 50 μg ml−1 rifampicin and 50 μg ml−1 kanamycin (Clough and Bent, 1998). Transgenic T1 plants were selected by spraying BASTA (Völker et al., 2001). UU2573 homozygous seedlings were identified by PCR amplification with the primers TUFF/F1, ACGAAAATCTCTGAAACCTCC, and TUFF/B2.2, CAGTGGAGACCATGAGGATC, which yielded a 1.3-kb product that was cleaved by BseMI in UU2573 mutant but not in wild-type DNA (see Figure 2B). The transgene was detected by PCR amplification, using primers KN90MYC, GCTTATTTCTGAGGAGGATCTTCTTTCTAG, and KN90off, ATTGGATACAAGACACGAAAGG.
Cloning of genomic VHA-E1, E2, and E3 constructs for expression of C-terminal translational GFP fusions
VHA-E1 (TUF, At4g11150): A 3624-bp fragment including 2000 bp E1 promoter sequence was PCR-amplified from BAC clone T22B4, using primers E1gfor, ATACCCGGGGGTAATGATGAGCCTTCTTCATGACAC, and E1grev, ATGGGATCCAGGCAGTAACTTGGCCGAACAAC, digested with XmaI/BamHI and ligated into the pGTkan vector, which is a derivative of pPZP212 (Hajdukiewicz et al., 1994), containing the GFP5(S65T) coding sequence and the rbcs terminator. VHA-E2 (At3g08560): A 2403-bp fragment including 605 bp E2 promoter sequence was PCR-amplified from BAC clone F17O14, using primers E2.GFPBamHI.f, TATAGGATCCGCACGGAGCTGATGCATTCGC, and E2.GFPBamHI.r, ATATGGATCCTTGCTCTGGAGGTTTCAGGAG, digested with BamHI and ligated into the pGTkan vector. VHA-E3 (At1g64200): A 2497-bp fragment including 1011 bp E3 promoter sequence was PCR-amplified from BAC clone F22C12, using primers E3.GFPBamHI.f, ATATGGATCCGTCTCTAACTGTAAAACAGAG, and E3.GFPBamHI.r, ATATGGATCCTAGCTGCACCAACCTTGCCGAAG, digested with BamHI and ligated into the pGTkan vector. Wild-type plants were transformed by floral dipping and transgenic plants were selected on plates for kanamycin resistance as above. For analysis of tuf rescue activity, E1-GFP transgenic plants were crossed with UU2573 heterozygous plants, F1UU2573 heterozygotes were selected and selfed, and their rescued homozygous mutant progeny were identified by PCR, using the BseMI polymorphism (see above). The E1-GFP transgene was detected by PCR amplification with the primers TUFF/F5, CATTATCAGTGTTTAGATCCGG, and GFPantisense, CCACTAGTTTTGTATAGTTCATCCATGCC, yielding a 900-bp product.
Microscopy and immunolocalization
Whole-mount preparations of ovules and plastic sections for light microscopy were prepared and analyzed as described (Mayer et al., 1993). Staging of mutant embryos was approximate and based on stages of wild-type siblings from the same silique. Immunofluorescence localization of TUF, Myc-TUF, TUF-GFP, gamma-COP, and plasma membrane H+-ATPase with rabbit anti-barley VHA-E antiserum (diluted 1:500; Betz and Dietz, 1991; kind gift of K.-J. Dietz, University of Bielefeld, Germany), mouse anti-Myc monoclonal antibody 9E10 (1:600), rabbit anti-GFP antiserum (1:800), rabbit anti-gamma-COP antiserum (Movafeghi et al., 1999; kind gift of D.G. Robinson, University of Heidelberg, Germany) and mouse anti-PM-ATPase (1:1000) monoclonal antibody, respectively, was carried out as described (Lauber et al., 1997). Signals were detected with Cy3-conjugated goat anti-mouse (1:600) and FITC-conjugated goat anti-rabbit (1:400) secondary antibodies (Dianova, Hamburg, Germany). DNA was stained with DAPI (1 μg ml−1). After mounting in Citifluor (Amersham, Freiburg, Germany), specimens were analyzed with a Leica confocal laser-scanning microscope (CLSM) and Leica TCS-NT software (Leica Microsystems, Bensheim, Germany). The CLSM standard objective was 63× (water immersion), scanning was carried out with electronic magnification. GFP fluorescence signals were detected by confocal laser scanning microscopy. Ultrastructural analysis of embryos was carried out by electron microscopy as described previously (Lauber et al., 1997). Immunogold labeling with anti-VHA-E antiserum (diluted 1:400) and silver-enhanced nanogold was carried out on cryosections and analyzed as described previously (Stierhof et al., 1991; Völker et al., 2001). Images were processed with Adobe PhotoshopTM software.
We are grateful to K.-J. Dietz for kindly providing antiserum. We thank H. Schwarz for help with the EM analysis, U. Hiller and B. Sailer for technical assistance, and Claudia Oecking for helpful comments on the manuscript. This work was supported by the EU Framework Programme 5 through contract QLG2-CT-1999-00454 (ECCO) and by the Deutsche Forschungsgemeinschaft through grant Ju 179/7-1.