Dominant pro-vasopressin mutants that cause diabetes insipidus form disulfide-linked fibrillar aggregates in the endoplasmic reticulum

Familial neurohypophyseal diabetes insipidus, a disorder of water homeostasis, results from lack of the circulating antidiuretic hormone vasopressin, normally produced in vasopressinergic neurons in the hypothalamic magno- and parvocellular nuclei. As a consequence, vasopressin-mediated reabsorption of water from the renal collecting ducts is deficient, causing patients to lose large amounts of unconcentrated urine and to suffer from increased thirst. The disease is caused by mutations in the gene encoding the vasopressin pre-prohormone. The vasopressin precursor consists of a signal peptide of 19 amino acids, the nine-amino acid hormone, a three-amino acid linker, the 93-amino acid carrier protein neurophysin II (NPII), a single linker amino acid and a C-terminal glycopeptide of 39 amino acids (Fig. 1). The precursor is stabilized by the formation of eight disulfide bridges, one in the hormone and seven in the NPII moiety.

Over 50 pathogenic mutations of the human vasopressin gene have been reported that alter the signal peptide, the hormone, or the NPII moieties (Christensen and Rittig, 2006). So far, no mutations have been found in the glycopeptide. Almost all mutations are dominant. Only the P7L mutant (Pro7 of vasopressin mutated to Leu; Willcutts et al., 1999) causes a recessive form of the disease. Autosomal dominant neurohypophyseal diabetes insipidus (ADNDI) shows a high penetrance, with symptoms beginning weeks to months after birth. Postmortem histological studies of affected individuals have shown degeneration of the vasopressinergic magnocellular neurons (Bergeron et al., 1991; Braverman et al., 1965; Green et al., 1967; Nagai et al., 1984). A knock-in mouse model expressing the human pathogenic mutant C67X (Cys67 of NPII mutated to a stop codon) confirmed the neurotoxic effect of the mutant protein on vasopressinergic cells (Russell et al., 2003), but the mechanism causing cell death remains unknown.

Expression studies have shown that dominant pro-vasopressin mutants are not secreted, but are retained in the endoplasmic reticulum (ER) (Beuret et al., 1999; Christensen et al., 2004; Friberg et al., 2004; Ito and Jameson, 1997; Nijenhuis et al., 1999; Olias et al., 1996; Siggaard et al., 1999). Most of the retained mutant protein is degraded by cytosolic proteasomes after retrotranslocation (Friberg et al., 2004). Here, we show that newly synthesized mutant precursors form disulfide-linked homo-oligomers and large ER-derived accumulations with a fibrillar ultrastructural appearance. Fibril formation was reproduced by the purified precursor in vitro. The results suggest that disulfide-linked oligomers in part escape degradation and gradually aggregate. Autosomal dominant diabetes insipidus thus belongs to the neurodegenerative diseases associated with fibrillar protein aggregation.

To assay for secretion, wild-type pro-vasopressin, the recessive mutant P7L, and the dominant mutants ΔE47, Y2H, C28Y and ΔG227 were expressed in COS-1 cells, pulse-labeled with [³⁵S]methionine/cysteine, chased for 2 hours, isolated from the cells and from the media by immunoprecipitation, and analyzed by SDS-gel electrophoresis and autoradiography (Fig. 2A). As expected, a considerable fraction of the wild type and of P7L was recovered from the media, reflecting their ability to fold and pass ER quality control. By contrast, the dominant mutants ΔE47, Y2H and C28Y were completely retained in the cells. Only the dominant mutant ΔG227, in which the signal sequence is inefficiently cleaved, was also partially secreted (Fig. 2A, lanes 13 and 14). The reduced molecular weight of the secreted protein indicates that the uncleaved pre-pro-vasopressin was retained and only cleaved pro-vasopressin, which is identical to wild type, was allowed to exit.

We have previously observed the formation of disulfide-linked products of retained ΔG227 in COS cells (Beuret et al., 1999). Here, we investigated the formation of disulfide complexes for the wild-type and all five pro-vasopressin mutants in COS-1 and Neuro2a mouse neuroblastoma cells. Transiently transfected cells were radiolabeled with [³⁵S]methionine/cysteine for 1 hour and then incubated with iodoacetamide to alkylate free SH-groups and prevent post-lysis oxidation. Vasopressin precursors were immunoprecipitated and separated by gel electrophoresis either under non-reducing conditions or after reduction of disulfide bonds with β-mercaptoethanol. Under reducing conditions, the precursors expressed in Neuro2a cells were found as a ∼21 kDa species representing glycosylated pro-vasopressin (Fig. 2B). By contrast, non-reducing conditions (Fig. 2C) revealed a ladder of bands corresponding in size to prohormone oligomers. Whereas wild-type pro-vasopressin and the recessive mutant P7L produced less than 20% of high molecular weight forms, most of the dominant mutant proteins (50-90%) were found in covalent oligomers (Fig. 2D). Qualitatively the same result was obtained with expression in COS-1 cells (not shown), indicating that disulfide-linked oligomerization of mutant pro-vasopressin is not a phenomenon specific to neuronal cells.

The strikingly regular spacing of the disulfide-linked products suggested homo-oligomerization of the precursor protein. To test this hypothesis, a lane containing the non-reduced ΔE47 products was cut out, placed horizontally onto a second gel, and overlaid with reducing sample buffer before electrophoresis. All disulfide-linked oligomers collapsed to bands of the same mobility as the monomeric protein (Fig. 2E). The same result was observed for the ΔG227 mutant (not shown). The disulfide-linked products generated predominantly by the dominant mutants thus represent homo-oligomers of glycosylated vasopressin precursor.

To characterize the intracellular localization of the vasopressin precursors, transfected COS-1 cells were analyzed by immunofluorescence staining 48 hours after transfection (Fig. 3). COS cells are particularly suitable for visualizing intracellular organelles and aggregations due to their flat cell body and large cytoplasm. As previously shown (Beuret et al., 2004), expression of regulated secretory proteins such as granins and prohormones, including pro-vasopressin, is sufficient to generate granule-like structures in a variety of nonendocrine cell lines due to aggregation in the trans-Golgi. Accordingly, wild-type precursor was concentrated in granular accumulations in the cell periphery of about half the expressing cells (Fig. 3A). The recessive mutant P7L showed the same immunofluorescence pattern as the wild type (Fig. 3B), which is in agreement with the observation that P7L was not retained in the ER, but was secreted into the culture medium as was the wild-type prohormone (Fig. 2A).

In many cells, the dominant mutants showed strong reticular staining typical for proteins retained in the ER. However, in up to 60% of the transfected cells, they also produced condensed accumulations dispersed throughout the cell (Fig. 3C-F), with morphologies distinct from that of the granule-like structures formed by the wild-type protein or the recessive P7L mutant. These accumulations had a coarse, short tubular (Fig. 3C-E) or irregular, laminar appearance (Fig. 3F). The different morphologies of the aggregates were not specific to particular mutants, but were found for all of them. The accumulations were distinct from mitochondria as there was no co-staining with the mitochondrial marker cytochrome c (not shown).

To unambiguously distinguish between granule-like structures and accumulations formed in the ER, co-staining experiments were performed. When wild-type pro-vasopressin and other regulated secretory proteins (e.g. secretogranin II) were co-expressed in COS-1 cells, they colocalized in granule-like structures (Beuret et al., 2004). This was also the case for the P7L mutant (Fig. 4A). By contrast, the secretogranin structures and the accumulations of the dominant mutants were clearly distinct (shown in Fig. 4C for C28Y). When COS-1 cells expressing pro-vasopressin were co-stained with antibodies against NPII and against BAP31 (B-cell receptor-associated protein 31; an endogenous ER membrane protein), the granule-like structures of P7L [like those of the wild type (Beuret et al., 2004)] were not positive for BAP31 (Fig. 4B). By contrast, there was extensive co-staining of the C28Y accumulations with BAP31 (Fig. 4D), indicating that the aggregations produced by the dominant mutant are located within the ER. The same was observed with all dominant mutants.

COS-1 cells transfected with the four dominant mutants were analyzed by immunofluorescence at different times after transfection (Fig. 4E). All mutants produced similar results. At 24 hours after transfection, anti-NPII immunostaining showed the reticular ER pattern, identical to that produced by anti-BAP31 antibody staining, in a majority of the cells, whereas less than 30% of transfected cells showed distinct aggregates. After 48 and 72 hours, the percentage of aggregate-forming cells increased to ∼55% and ∼75%, respectively, and the aggregations generally increased in size. As a control, we also analyzed the immunofluorescence staining of COS-1 cells expressing an ER-retained mutant of cathepsin Z, a secretory protein with five disulfide bonds. In these experiments, the frequency of apparent aggregations did not increase with time above the control levels of ∼20%. Large aggregates and disulfide oligomers were also produced by dominant vasopressin mutants expressed in CV-1 cells (the parental cell line of COS-1 cells) lacking the large T-antigen needed for high-level expression (not shown). Thus, the formation of aggregates was not an artifact of excessive expression levels due to plasmid amplification.

COS-1 cells expressing dominant mutants were analyzed by immunogold electron microscopy 48 hours after transfection (Fig. 5). As expected, gold staining was found in many cells in ER tubules cut longitudinally or across, some of which had a dilated appearance (Fig. 5A,A′). Other cells contained multiple large structures of ∼1 μm in diameter decorated with 15-nm gold for pro-vasopressin (Fig. 5B,B′,C), corresponding to the bright accumulations seen by immunofluorescence (Fig. 3C,D,E). In some cells, large regions were densely filled with labeled structures (Fig. 5D,D′) that might appear almost continuous in immunofluorescence (as in Fig. 3F), adjacent to regions with normal-sized ER tubules of ∼60-90 nm diameter. With double-labeling, the pro-vasopressin aggregations stained for NPII with 15-nm gold were also found to be decorated with 10-nm gold for the ER chaperone calreticulin (Fig. 5C), confirming that these are ER-derived structures. In some micrographs, the core of the aggregates appeared to be structured, and gold particles were aligned in continuous rows (Fig. 5B′), suggesting a filamentous arrangement of the antigen.

To also test their behavior in a neuronal cell type representing the physiological situation more closely, dominant pro-vasopressin mutants were expressed in Neuro2a cells and analyzed by immunofluorescence and electron microscopy. By confocal microscopy, dispersed accumulations of pro-vasopressin could again be observed that were strongly co-stained for the ER chaperone Hsp47 (heat shock protein 47; Fig. 6). At the ultrastructural level, immunogold staining showed pro-vasopressin in compact accumulations of similar size to those in COS-1 cells (Fig. 7A and enlarged areas in Fig. 7B,C). The aggregations are structured and appear to be composed of tangled filaments of approximately 10-15 nm diameter (Fig. 7B-F). Using double-labeling, both mutant pro-vasopressin and calreticulin were again found to decorate these structures (Fig. 7D,F), demonstrating that they are derived from the ER. Clearly, aggregations containing mutant pro-vasopressin are formed not only in fibroblasts, but also in neuronally derived cells, which better mimic the pathological situation in vivo.

To test the ability of pro-vasopressin to form fibrils in isolation, the ΔE47 mutant was expressed with a C-terminal His6 tag replacing the signal sequence in Escherichia coli, recovered in inclusion bodies, solubilized with 8 M urea, and purified under reducing conditions by metal-chelate affinity chromatography. After removal of urea and reducing agent by dialysis, and pelleting of amorphous precipitates by centrifugation, fibrillar structures of defined diameter were detected in the supernatant by electron microscopy (Fig. 8). No fibrils were observed when reducing conditions were maintained throughout. These results show that pro-vasopressin can form fibrils in vitro in the absence of any other cellular proteins and that oxidation is necessary to stabilize them.

To investigate whether the formation of disulfide-linked oligomers and of large aggregations is dependent on specific cysteine residues, we generated artificial mutants in which the cysteines in any of four segments of the precursor protein were replaced by serines, or in which all 16 cysteines were mutated (Fig. 9A). Mutation of cysteines almost certainly prevents native folding of the precursor (10 of the 16 have been found mutated in diabetes insipidus alleles). These mutants were tested for the formation of disulfide-linked oligomers in Neuro2a cells (Fig. 9B,C) and of aggregates in COS-1 cells (Fig. 9D,E). Mutation of any one group of cysteines still allowed efficient formation of disulfide oligomers and the appearance of large aggregations. None of the cysteines is thus essential for disulfide oligomerization or aggregation. As expected, the mutant without any cysteines did not produce covalent homo-oligomers (Fig. 9B, abcd). Also, this cysteine-less mutant did not induce the formation of aggregates detectable by immunofluorescence (Fig. 9D,E). When these mutants were tested for ER retention, two constructs, aBCD and abcd, were surprisingly found to be partially secreted into the medium (Fig. 9G, lanes 10 and 14).

To exclude the possibility that the absence of aggregates with construct abcd is due to reduced ER concentrations, the ER retention signal KDEL was fused to the C-terminus. This effectively prevented secretion of both aBCD^KDEL and abcd^KDEL (lanes 12 and 16). The cysteine-less construct, even when retained in the ER as abcd^KDEL, did not increase aggregate formation above backgound levels (Fig. 9D,E). Even at the ultrastructural level, no aggregations could be observed for abcd but there was dispersed staining in the ER (mostly tubular elements in the cytosol and the nuclear envelope; Fig. 9F).

To test the ability of the cysteine-free construct to form fibrils in vitro, abcd without the signal sequence was expressed in bacteria. Unlike ΔE47 pro-vasopressin, which accumulated in inclusion bodies, abcd remained soluble in the bacterial cytoplasm. After purification in 8 M urea, dialysis to remove the denaturant and dithiothreitol (DTT), and analysis by electron microscopy, hardly any structures resembling the fibrillar aggregates found for ΔE47 in a parallel experiment could be detected (not shown). Together with the in vivo results this suggests that, whereas no individual cysteine in the precursor is necessary for oligomerization or aggregation, the formation of fibrils and aggregates is dependent on the presence of cysteine residues, irrespective of their positions within the precursor protein.

Many different mutations in the vasopressin precursor have been shown to cause diabetes insipidus. Only two of these are recessive. The human P7L mutation in the vasopressin nonapeptide (Willcutts et al., 1999) greatly reduces hormone binding to its renal receptor, but does not affect precursor folding and secretion. In addition, the Brattleboro mutation in rat, a frameshift mutation in the gene segment encoding NPII, shows a recessive phenotype (Schmale and Richter, 1984). Here, the complete absence of a stop codon in the new reading frame results in highly inefficient translation of the mutant protein, resulting essentially in a null allele. All other known mutations in the vasopressin precursor are dominant, eventually causing death of the vasopressin-producing neurons.

The mutations are diverse and not localized to a specific region in the precursor protein, except that no pathogenic mutations have been identified within the C-terminal glycopeptide. Cytotoxicity appears therefore not to be caused by a specific effect of each individual mutation within the prohormone, but to result from consequences common to all the mutations. This notion is supported by two mutations in the signal sequence, ΔG227 (resulting in a signal truncation; Beuret et al., 1999) and A(–1)T (mutation of the signal cleavage site; Ito et al., 1993). In both cases, removal of the signal peptide is impaired, but the pro-vasopressin sequence itself remains unchanged. A common effect of all dominant mutants tested is their retention in the ER, most probably due to misfolding. For frameshift and point mutations affecting cysteine residues normally involved in disulfide bonds, misfolding is an obvious consequence. Failure to cleave the signal sequence has a more subtle effect, as it prevents the N-terminus of the vasopressin sequence from folding into NPII, a requirement for the stabilization of the disulfide bond within the hormone segment (Beuret et al., 1999; Chen et al., 1991). How can misfolding and ER retention cause cytotoxicity?

At least half of the newly synthesized precursor proteins of all four dominant mutants were found to be part of disulfide-linked homo-oligomers. This indicates a specific interaction between unfolded pro-vasopressin polypeptides, bringing together cysteines to form intermolecular crosslinks. Using in vitro mutagenesis, we tested which cysteines might be essential for this process. Whereas the presence of some cysteines was required for the formation of disulfide-linked oligomers as well as aggregations visible by immunofluorescence, no single cysteine was specifically necessary.

Disulfide-linked homo-oligomers were also observed for wild-type pro-vasopressin and the recessive mutant P7L, but to a much lower extent. Not-yet folded protein appears to be transiently crosslinked into disulfide-linked homo-oligomers, but to be subsequently rescued when the free molecules successfully fold into the native structure. Indeed, even the mutant proteins are not irreversibly trapped in disulfide aggregates because we have previously found them to be quite efficiently degraded by the cytosolic proteasome (Friberg et al., 2004). Only a small fraction thus escapes ER-associated degradation to accumulate in large aggregates visible by fluorescence and electron microscopy. Consistent with this, it takes 2-3 days for a majority of expressing cultured cells to develop visible aggregates.

Electron microscopic analysis showed that the pro-vasopressin accumulations have a fibrillar ultrastructure. The ability of pro-vasopressin to form ordered, linear polymers by self-assembly was confirmed in vitro with bacterially expressed and purified protein. The vasopressin precursor thus resembles a whole family of proteins that generate amyloid fibers and are associated with neurodegenerative disorders.

Taken together, our results suggest a model for aggregate formation in ADNDI as illustrated in Fig. 10. After translocation into the ER lumen, vasopressin precursor might acquire a conformation that leads to homo-oligomers, which are stabilized by intermolecular disulfide-links between exposed cysteine residues. This process is reversible and thus allows wild-type proteins to fold into their native structure and exit the ER, and allows folding-deficient mutants to be retrotranslocated into the cytosol for proteasomal degradation. A significant fraction of mutant precursors accumulate as growing oligomers and polymers, forming fibrils and eventually large visible aggregates associated with ER chaperones.

The formation of fibrillar aggregations and inclusion bodies are frequent pathological features of neurodegenerative disorders (Ross and Poirier, 2005). In contrast to the extensively studied aggregations associated with Alzheimer's, Parkinson's and Huntington's disease that are extracellular, cytoplasmic or intranuclear, the aggregations formed by dominant pro-vasopressin mutants are produced within the lumen of the ER.

The role of aggregation in degenerative disease processes is controversial because there is often no direct correlation between the amount of inclusions and cell degeneration (Chun et al., 2002; Ross and Poirier, 2005). The formation of large aggregates and inclusion bodies might represent a cellular protective response, and early protofibrils might be the neuropathogenic species triggering disease (Lansbury and Lashuel, 2006; Ross and Poirier, 2005). In a diabetes insipidus knock-in mouse expressing the cytotoxic pro-vasopressin mutant C67X, induction of the ER chaperone BiP was detected, but no obvious aggregations were apparent (Russell et al., 2003). No upregulation of the ER stress-induced pro-apoptotic transcription factor C/EBP homologous protein (CHOP) or other evidence for apoptosis was observed. However, aggregate-producing cells might have been rapidly eliminated. By contrast, vasopressinergic neurons expressing the same C67X mutant in transgenic rats were resistant to a cytotoxic effect, but produced distended ER structures containing the mutant protein and markers of autophagy (Davies and Murphy, 2002; Si-Hoe et al., 2000). This lysosomal degradation mechanism is thought to protect against the toxic effects of altered or aggregated intracellular proteins (Martinez-Vicente and Cuervo, 2007). Inhibition of autophagy was found to compromise the viability of C67X-expressing rat neurons and Neuro2a cells (Castino et al., 2005). Interestingly, we could not detect significant colocalization of the dominant mutant C28Y with Lamp1 or LC3 (markers for lysosomes and autophagosomes, respectively) neither in COS-1 nor Neuro2a cells (J.B., unpublished). The extent to which the disulfide-linked homo-oligomers of pro-vasopressin mutants detected biochemically and the large aggregations observed by light and electron microscopy contribute to cytotoxicity remains to be analyzed in detail.

The cDNAs of the human wild-type vasopressin precursor and the mutants ΔG227, ΔE47 and C67X, as well as of myc-tagged secretogranin II, have been described before (Beuret et al., 1999; Friberg et al., 2004). The cDNAs encoding the vasopressin mutants Y2H and P7L were a gift from Jane Christensen (Århus University Hospital, Aarhus, Denmark). The C28Y mutant was generated by polymerase chain reaction. The expression plasmid of human pre-pro-cathepsin Z mutant (N184Q, N224Q) (Appenzeller-Herzog et al., 2005) was a gift from Hans-Peter Hauri and Beat Nyfeler (Biozentrum, University of Basel, Basel, Switzerland). All human cDNAs were subcloned into the pRc/RSV expression plasmid (Invitrogen).

The 16 cysteine residues in pro-vasopressin were mutated to serines by PCR mutagenesis. For this purpose, the signal sequence was replaced by that of pre-pro-enkephalin as described (Cescato et al., 2000; Friberg et al., 2004), providing a BsrGI restriction site near the start of pro-vasopressin. Silent mutations were introduced to insert a KpnI site at codons 49/50 of the pro-vasopressin coding sequence and a BamHI site at codons 64/65. Together with a natural SmaI site at codons 27/28 and external SalI site, these restriction sites served to clone and recombine PCR-amplified mutant segments. All constructs were verified by DNA sequencing.

COS-1 cells were grown in Dulbecco's modified Eagle's medium (DMEM; Sigma) supplemented with 10% fetal calf serum (FCS), 100 units/ml penicillin, 100 μg/ml streptomycin and 2 mM L-glutamine at 37°C in 7.5% CO₂. COS-1 cells were transiently transfected in six-well plates using Lipofectine (Invitrogen) or FuGENE HD (Roche) and analyzed 2-3 days after transfection or as indicated. Neuro2a cells were grown in DMEM containing 4500 mg/l glucose, supplemented as above. They were transfected using Metafectene (Biontex Laboratories). To induce differentiation, cells were cultured from 1 day before transfection in 1 mM valproic acid in DMEM with 3% FCS, 100 units/ml penicillin, 100 μg/ml streptomycin and 2 mM L-glutamine.

For labeling experiments, transfected cells were starved for 30 minutes in DMEM without methionine and cysteine (Sigma), supplemented with 2 mM L-glutamine. Cells were labeled with 100 μCi/ml [³⁵S]protein-labeling mix (PerkinElmer) and chased with excess methionine and cysteine. Cells were washed in phosphate-buffered saline (PBS), lysed in 500 μl of lysis buffer (PBS, 1% TritonX-100, 0.5% deoxycholate, 2 mM phenylmethylsulfonylfluoride), and scraped. Cell lysates or media were incubated with rabbit polyclonal anti-NPII (Friberg et al., 2004) in lysis buffer containing 1 mM phenylmethylsulfonylfluoride, 0.5% SDS and 1 mg/ml BSA. Bound proteins were immunoprecipitated with protein-A–Sepharose (Zymed) and analyzed by electrophoresis on 10% polyacrylamide Tris/tricine SDS-gels and autoradiography.

Transfected and labeled cells were washed with cold PBS containing 100 mM iodoacetamide; incubated at 4°C for 1 hour in the dark with 200 μl 200 mM Tris pH 8.0, 100 mM iodoacetamide, 2 mM phenylmethylsulfonyl fluoride; lysed in lysis buffer containing 100 mM iodoacetamide; and immunoprecipitated. One half of each sample was boiled in SDS-sample buffer with β-mercaptoethanol and the other half without β-mercaptoethanol before gel electrophoresis and autoradiography. The amount of protein present as disulfide-linked aggregates (signal of high molecular weight oligomers as a percentage of total signal for each construct) was quantified by phosphorimager. For analysis in the second-dimension, a lane containing a non-reduced sample was cut out and soaked for 1 minute in sample buffer containing 200 mM β-mercaptoethanol. The gel strip was then loaded horizontally onto a second Tris/tricine SDS-gel, whose stacking gel was supplemented with 50 mM β-mercaptoethanol, and overlaid with reducing sample buffer for electrophoresis.

Transfected cells were grown for 24-72 hours on glass coverslips; fixed with 3% paraformaldehyde for 30 minutes at room temperature; washed in PBS; permeabilized and blocked with 0.1% saponin, 20 mM glycine and 1% BSA in PBS for 20 minutes; incubated at room temperature with primary antibodies for 2 hours in PBS containing 0.1% saponin, 20 mM glycine and 1% BSA; washed; and stained with fluorescent secondary antibodies in PBS containing 0.1% saponin for 30 minutes After several washes with PBS containing 0.1% saponin, the coverslips were mounted in Mowiol 4-88 (Hoechst). As primary antibodies, we used polyclonal rabbit anti-NPII antiserum, mouse monoclonal anti-myc antibody 9E10, and a mouse monoclonal antibody against the endogenous ER membrane protein BAP31 (a gift from Hans-Peter Hauri; Klumperman et al., 1998). As secondary antibodies, non-cross-reacting Cy3-labeled goat anti-mouse and Cy2-labeled goat anti-rabbit immunoglobulin antibodies (Jackson ImmunoResearch and Amersham Biosciences, respectively) were used according to the manufacturers' recommendations. Staining patterns were analyzed using a Zeiss Axioplan 2 microscope with a Leica DFC420C imaging system or, for Neuro2a cells, with a Zeiss Confocal LSM510 Meta microscope.

Cultured cells were fixed in 3% formaldehyde and 0.2% glutaraldehyde for 2 hours at room temperature. After addition of 1% picric acid overnight at 4°C, cells were scraped, pelleted, resuspended and washed three times in PBS, treated with 50 mM NH₄Cl in PBS for 30 minutes to quench free aldehydes, and then washed three times in PBS. The final pellet was resuspended in 2% warm agarose and left on ice to solidify. Agarose pieces were dehydrated and infiltrated with LR-gold resin (London Resin, London, UK) according to the manufacturer's instructions and allowed to polymerize for 1 day at 4°C. For immunogold labeling, sections of about 60 nm were collected on carbon-coated Formvar-Ni-grids, incubated with rabbit anti-NPII in PBS containing 2% BSA and 0.1% Tween-20 for 2 hours, washed with PBS, and incubated with 15-nm colloidal gold-conjugated goat anti-rabbit immunoglobulin antibodies (BioCell, Cardiff) in PBS containing 2% BSA and 0.1% Tween-20 for 90 minutes Grids were washed five times for 5 minutes in PBS and then five times in water before staining for 10 minutes in 2% uranylacetate followed by 1 minute in Reynolds lead citrate solution. Sections were viewed in a Phillips CM100 electron microscope.

For double labeling, the first labeling was performed using rabbit anti-human calreticulin antibody (Stressgen) and 10-nm colloidal gold-conjugated goat anti-rabbit immunoglobulin antibodies as described above. After washing six times for 2 minutes with PBS, the sections were fixed again for 2 hours with 1% formaldehyde and 1% glutaraldehyde in PBS, washed with PBS, and quenched for 10 minutes with 50 mM NH₄Cl. The sections were incubated with PBS containing 2% BSA and 0.1% Tween-20 for 10 minutes, followed by incubation for 90 minutes with rabbit anti-NPII immunoglobulins that had been purified from serum by a protein-A–Sepharose column (Amersham Pharmacia Biotech), pre-coupled to 15-nm gold–protein-A (BioCell) at a molar ratio of 2:1, and incubated with a twofold excess of unrelated mouse IgG. Grids were washed and stained as above. As negative controls to ensure specificity of labeling, cells were stained as above, except that 15-nm gold–protein-A was used that had not been loaded with anti-NPII immunoglobulins. This produced very low background labeling (not shown).

For bacterial expression, the coding sequence of ΔE47 and cysteine-free (abcd) pro-vasopressin (without the signal sequence) were cloned into the plasmid pET21b (Novagen) and thus provided with a C-terminal His6 tag. The proteins were expressed in E. coli BL21(DE3) after induction with 1 mM isopropyl-d-thiogalactoside for 2-3 hours, solubilized from inclusion bodies (E47) or the cytosol fraction (abcd) in 8 M urea, 10 mM Tris-HCl pH 8, 1 mM DTT, and protease inhibitors (10 μg/ml each of leupeptin, pepstatin and aprotinin, 0.5 mM orthovanadate, 1 mM PMSF), and purified by Ni-NTA chromatography. Protein at 1 mg/ml was dialyzed stepwise for 8-16 hours each to 6, 5, 4, 2, 1 and 0.5 M urea, and finally to 10 mM Tris-HCl pH 7.4, 100 mM NaCl without DTT overnight. Large precipitates were removed by centrifugation in a tabletop centrifuge for 5 minutes. The supernatant was analyzed by electron microscopy. An aliquot of 10 μl was adsorbed to Formvar carbon-coated Ni grids, washed, dried, stained with 0.5% uranyl acetate for 1 minute, dried, and imaged using a Phillips CM100 electron microscope.