Mechanism of Endoplasmic Reticulum Retention of Mutant Vasopressin Precursor Caused by a Signal Peptide Truncation Associated with Diabetes Insipidus*

Autosomal dominant neurohypophyseal diabetes insipidus is caused by mutations in the gene encoding the vasopressin precursor protein, prepro-vasopressin-neurophysin II. We analyzed the molecular consequences of a mutation (ΔG227) recently identified in a Swiss kindred that destroys the translation initiation codon. In COS-7 cells transfected with the mutant cDNA, translation was found to initiate at an alternative ATG, producing a truncated signal sequence that was functional for targeting and translocation but was not cleaved by signal peptidase. The mutant precursor was completely retained within the endoplasmic reticulum. The uncleaved signal did not affect folding of the neurophysin portion of the precursor, as determined by its protease resistance. However, formation of disulfide-linked aggregates indicated that it interfered with the formation of the disulfide bond in vasopressin, most likely by blocking its insertion into the hormone binding site of neurophysin. Preventing disulfide formation in the vasopressin nonapeptide by mutation of cysteine 6 to serine was shown to be sufficient to cause aggregation and retention. These results indicate that the ΔG227 mutation induces translation of a truncated signal sequence that cannot be cleaved but prevents correct folding and oxidation of vasopressin, thereby causing precursor aggregation and retention in the endoplasmic reticulum.

Autosomal dominant neurohypophyseal diabetes insipidus is caused by mutations in the gene encoding the vasopressin precursor protein, prepro-vasopressin-neurophysin II. We analyzed the molecular consequences of a mutation (⌬G227) recently identified in a Swiss kindred that destroys the translation initiation codon. In COS-7 cells transfected with the mutant cDNA, translation was found to initiate at an alternative ATG, producing a truncated signal sequence that was functional for targeting and translocation but was not cleaved by signal peptidase. The mutant precursor was completely retained within the endoplasmic reticulum. The uncleaved signal did not affect folding of the neurophysin portion of the precursor, as determined by its protease resistance. However, formation of disulfide-linked aggregates indicated that it interfered with the formation of the disulfide bond in vasopressin, most likely by blocking its insertion into the hormone binding site of neurophysin. Preventing disulfide formation in the vasopressin nonapeptide by mutation of cysteine 6 to serine was shown to be sufficient to cause aggregation and retention. These results indicate that the ⌬G227 mutation induces translation of a truncated signal sequence that cannot be cleaved but prevents correct folding and oxidation of vasopressin, thereby causing precursor aggregation and retention in the endoplasmic reticulum.
Human neurohypophyseal diabetes insipidus is a disease characterized by a severe disturbance of antidiuresis becauseof the lack of the hypothalamic/neurohypophyseal antidiuretic hormone arginine-vasopressin. Patients produce great amounts of dilute urine and accordingly must drink large volumes of fluid to avoid exsiccosis. Neurohypophyseal diabetes insipidus most often results from trauma (including brain operations) or infiltrative or metastatic processes that damage the hypothalamus or the neurohypophysis, or it occurs as a primary idiopathic disorder. More rarely, the disease is hereditary and generally transmitted in an autosomal dominant manner (autosomal dominant neurohypophyseal diabetes insipidus, ADNDI) (1, 2). 1 Various mutations were identified in the coding sequence of the prepro-vasopressin-neurophysin II gene (3).
As illustrated in Fig. 1A, the precursor protein consists of four main segments (4): 1) a 19-amino acid signal sequence for targeting to the endoplasmic reticulum (ER); 2) the nonapeptide hormone vasopressin; 3) neurophysin II (NPII) consisting of 93 residues, which serves as the vasopressin transport protein in the circulation; and 4) a 39-amino acid glycopeptide of unknown function with a single N-glycosylation site. The signal sequence is cleaved off by signal peptidase upon translocation of the precursor into the ER lumen. After folding and disulfide bond formation, pro-vasopressin-NPII passes through the Golgi apparatus into secretory granules where the other segments are separated by proteolytic removal of the linker residues. The mutations identified in different kindred of ADNDI are located in the sequences encoding the NPII moiety, the signal peptide, and in one case the vasopressin sequence (3,(5)(6)(7)(8)(9).
Typically, symptoms of ADNDI develop gradually over a period of months to years after birth (10 -13). Postmortem studies revealed degeneration of the hypothalamic magnocellular neurons in the nucleus supraventricularis and paraopticus, which synthesize the vasopressin-NPII precursor (14 -19). Based on these findings, it has been postulated that impaired transport and/or processing of the mutant precursor may result in its intracellular accumulation, eventually leading to degeneration of the vasopressinergic neurons and to the gradual manifestation of clinical symptoms. A recent study provided evidence for the toxicity of mutant precursors or their degradation products to cultured neuronal cells (20).
We have investigated the molecular mechanism of an unusual mutation associated with ADNDI recently identified in a Swiss kindred (21). Guanosine 227, the third nucleotide of the translation initiation codon of the vasopressin precursor gene, is deleted on one allele in affected family members. These heterozygous patients are completely vasopressin-deficient and show an abnormal appearance of the neurohypophysis in magnetic resonance imaging. Because loss of expression of one allele is not a plausible cause for the observed dominant phenotype, we hypothesized that a second, in-frame ATG present four codons downstream might serve as an alternative initiation site generating a mutant precursor with a truncated signal sequence (Fig. 1B, Vm). To test this hypothesis and to analyze how a signal sequence truncation might affect the fate of the protein, we transiently transfected mutant or wild-type preprovasopressin-NPII cDNA into COS-7 cells and characterized the effects of the guanosine 227 deletion on expression, membrane translocation, signal cleavage, polypeptide folding, and secretion of the precursor. The protein was found to be synthesized and translocated into the ER. However, the truncated signal sequence was not cleaved and interfered specifically with the folding and disulfide bond formation of the vasopressin nonapeptide portion of the precursor, thus causing the formation of disulfide aggregates and retention within the ER.

EXPERIMENTAL PROCEDURES
DNA Constructs-The full-size cDNA of wild-type prepro-vasopressin-NPII was a gift from Dr. M. Ito (Northwestern University, Chicago, IL). To construct the full-size cDNA of the guanosine 227 deletion mutant, Vm, the 5Ј end of the precursor cDNA up to the SmaI site was amplified by polymerase chain reaction using Vent polymerase, the mutagenic primer CAGGATCCTGACACCATGCTGC (the BamHI site produced by the deletion of guanosine 227 is underlined), and a second primer corresponding to a sequence in the plasmid vector. The product was ligated at the SmaI site to the unamplified 3Ј portion of the cDNA. The cDNA was confirmed by sequencing. In the process, a separate cDNA clone, Vm⌬, was identified with an additional, accidental mutation, deletion of codons 28 -36 of pro-vasopressin-NPII (residues 16 -24 of NPII).
Polymerase chain reaction amplification of the precursor cDNA using the mutagenic primer CGCAGATCTACCATGCTGCCCGCCAGCT-TCCTCGG in combination with a second primer in the flanking vector sequence was performed to generate the plasmid pVm(CϪ11S), which encodes the vasopressin precursor with a truncated signal sequence and with Cys Ϫ11 of the signal peptide mutated to serine.
Cysteine 6 was mutated to serine by polymerase chain reaction amplification of the 5Ј portion of the wild-type precursor cDNA using the mutagenic antisense primer CGGGAGCTCTGTTCTGGAAGTAG-CACGC in combination with a second primer in the flanking vector sequence. Separately the 3Ј portion was amplified using the primer CAGGATCCTGACACCATGCTGC and a vector primer. The mutagenic primers mutated the codon TGC of cysteine 6 to the sequence AGAGCTC, thus inserting four nucleotides to generate a new SacI site (22). The cloned polymerase chain reaction products were ligated at this SacI site. Finally, the four extra nucleotides were deleted by cutting the DNA with SacI, blunting the 3Ј protruding ends with T4 DNA polymerase, and ligation, producing the sequence AGC encoding serine. The final cDNAs were confirmed by sequencing.
Cell Culture, Transfection, and Immunoprecipitation-COS-7 cells were grown in Eagle's minimal essential medium with 10% fetal calf serum at 37°C with 7.5% CO 2 . The media were supplemented with 2 mM L-glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. Transfection of COS-7 cells using DEAE-dextran and Me 2 SO was performed according to Cullen (23) in 6-well plates. Transfected cells were processed 48 h after transfection. For in vivo labeling, transfected cells were incubated for 30 min in medium lacking methionine and labeled for 30 min at 37°C with 100 Ci/ml [ 35 S]methionine in starvation medium. Where indicated, cells were chased in complete medium containing excess methionine. Lysed cells and media were subjected to immunoprecipitation using a polyclonal rabbit anti-NPII antiserum (ICN) and protein A-Sepharose (Amersham Pharmacia Biotech). For analysis with endo-␤-N-acetylglucosaminidase H (Endo H; Roche Molecular Biochemicals), immunoprecipitates were boiled for 2 min in 50 l of 50 mM sodium citrate, pH 6, 1% SDS. Aliquots were incubated with 1 milliunit of Endo H for 2 h at 37°C. Finally, samples were boiled in SDS sample buffer and analyzed by SDS-polyacrylamide gel electrophoresis. Gels were fixed, soaked in 1 M sodium salicylate containing 1% glycerol, dried, and fluorographed on Kodak BioMax MR films. For quantification, fluorographs were densitometrically scanned.
In Vitro Translation and Radiosequencing-The cDNAs of wild-type and mutant prepro-vasopressin-NPII were subcloned into the plasmid pGEM3 (Promega), in vitro transcribed with SP6 RNA polymerase, and translated using rabbit reticulocyte lysate with dog pancrease microsomes as described (24). Translation products were labeled with [ 35 S]methionine for fluorography and with [ 3 H]leucine for sequencing. After translation, the microsomes were pelleted through a sucrose cushion (24), separated by SDS gel electrophoresis, and transferred to polyvinylidene difluoride membrane. The region containing the labeled protein was cut out and subjected to automated Edman degradation using an Applied Biosystems 477A sequencer. The eluants at each cycle were analyzed by scintillation counting. Inhibition of signal peptidase was accomplished by addition of N-methoxysuccinyl-Ala-Ala-Pro-Valchloromethyl ketone (Sigma) in Me 2 SO at a final concentration of 5 mM. 2 Membrane Integration, Disulfide Aggregation, and Protease Sensitivity-To assess membrane integration of wild-type and mutant vasopressin precursor, labeled cells were scraped and subjected to alkaline extraction as described by Gilmore and Blobel (25). Alternatively, cells were extracted with 0.1% saponin in PBS at 4°C for 20 min (24). Extracted and membrane-associated material was separately taken up in lysis buffer and subjected to immunoprecipitation. To analyze disulfide formation, labeled transfected cells were treated with 100 mM iodoacetamide for 60 min at 4°C before lysis and immunoprecipitation. Samples were split in two aliquots and boiled with SDS sample buffer with or without 100 mM ␤-mercaptoethanol before SDS gel electrophoresis and fluorography. To assess the folding state of wild-type and mutant vasopressin precursor, labeled protein was immunoprecipitated with rabbit anti-NPII or anti-vasopressin antiserum (ICN), and protein A-Sepharose. The immunoprecipitates were incubated for 10 min at 4°C with 0 -5 mg/ml trypsin. The protease was then stopped by the addition of 10 mM phenylmethylsulfonyl fluoride, the resin was washed, and the bound protein was analyzed by SDS gel electrophoresis and fluorography.
Immunofluorescence-For indirect immunofluorescence staining, cells were grown on 14-mm glass coverslips, fixed with 3% paraformaldehyde for 20 min at room temperature, washed in PBS, and quenched with 50 mM NH 4 Cl in PBS. Cells were permeabilized with 0.1% Triton X-100 for 10 min. Nonspecific antibody binding was blocked with PBS containing 1% bovine serum albumin (PBSB). The fixed cells were incubated with anti-neurophysin II primary antibody diluted 1:400 in PBSB for 1 h, washed with PBSB, and incubated with fluorescein isothiocyanate-labeled anti-rabbit immunoglobulin secondary antibody diluted 1:200 in PBSB for 30 min. After additional washes with PBSB, PBS, and water, the coverslips were mounted in Mowiol 4 -88 (Hoechst) containing 2.5% 1,4-diazobicyclo-(2, 2, 2)-octane and analyzed using a Zeiss Axiophot microscope. For double immunofluorescence, the fixed cells were incubated simultaneously with anti-neurophysin II and a mouse monoclonal anti-p63 antibody (26) followed by incubation with fluorescein isothiocyanate-labeled anti-rabbit immunoglobulin and then Cy3-labeled rabbit anti-mouse immunoglobulin antibodies.

Deletion of Guanosine 227 Results in a Truncated Signal
Sequence That Is Functional but Not Cleaved-Deletion of guanosine 227 in the prepro-vasopressin-NPII gene of the Swiss ADNDI kindred eliminates the original translation initiation codon. Expression of a mutant protein would depend upon initiation at an alternative in-frame ATG. There are only two internal ATG sequences in the entire prepro-vasopressin-NPII cDNA, encoding Met 14 (the second codon of NPII) and Met Ϫ15 (the fifth codon of the signal sequence; Fig. 1B). Translation initiation at Met 14 would generate a cytosolic polypeptide, whereas a protein beginning with Met Ϫ15 would still contain most of the signal peptide truncated by four residues (Fig.  1B). The wild-type signal sequence is unusual in that it contains a negatively charged aspartate and no positively charged residues in its N-domain (27,28). In comparison, the truncated signal appears rather more typical, because the negative residue is deleted.
To test whether translation of the mutant cDNA initiates at one of the downstream ATGs and whether insertion into the ER occurs, wild-type and mutant cDNAs were transiently expressed in transfected COS-7 cells, labeled with [ 35 S]methionine for 30 min, and immunoprecipitated with an antiserum against the NPII portion of the precursor protein ( Fig. 2A). As expected, the wild-type vasopressin precursor was synthesized as a major species of ϳ22 kDa ( Fig. 2A, lane 1). Sensitivity to digestion with Endo H showed it to be glycosylated and thus translocated into the ER lumen (lane 3). Often, a small amount of a ϳ19-kDa species was also produced (lane 1). Its insensitivity to Endo H digestion and the electrophoretic mobility, which was identical to that of the deglycosylated major product, suggested it to be an unglycosylated form of the precursor protein. The mutant cDNA with the guanosine 227 deletion was also efficiently expressed (Vm, lanes 2 and 4). The major product, however, migrated with an increased apparent molec-ular mass of ϳ23 kDa that was reduced to ϳ20 kDa by deglycosylation. Glycosylation indicates that even in the absence of the original initiator codon a mutant protein with a functional signal sequence was translated. Similar to the wild-type protein, a minor fraction was produced in unglycosylated form. Because there is no other ATG upstream of the coding sequence, initiation must have occurred at codon 5 of the segment encoding the signal sequence. The increased molecular mass suggested that the signal was not cleaved.
To experimentally determine the N-terminal sequence of Vm, its cDNA was translated in vitro using rabbit reticulocyte lysate and dog pancreas microsomes in the presence of radioactive amino acids for subsequent detection after Edman degradation. The mutant precursor labeled with [ 35 S]methionine was efficiently glycosylated in the presence of microsomes (Fig.  3A, lanes 1 and 2). As expected, it migrated slightly more slowly than the wild-type precursor (lane 4) but with the same mobility as the wild-type precursor translated in the presence of a signal peptidase inhibitor (lane 3). For sequencing, Vm was synthesized with [ 3 H]leucine, and the glycosylated products were subjected to automated Edman degradation. [ 3 H]Leucine was released in cycle 2 and in cycles 7-10 ( Fig. 3B). Within the entire prepro-vasopressin-NPII sequence, this pattern corresponds only with the sequence starting with Met Ϫ15 , which has leucines at positions 2, 7, 9, and 10 from the N terminus (positions Ϫ14, Ϫ9, Ϫ7, and Ϫ6 of the precursor protein). This confirms that translation initiated at Met Ϫ15 and that the truncated signal was not cleaved.
The Mutant Precursor Is Efficiently Retained in the ER-To analyze their secretion properties, wild-type and mutant prepro-vasopressin-NPII precursors were expressed in COS-7 cells, pulse-labeled for 30 min with [ 35 S]methionine, chased for up to 4 h, and then immunoprecipitated from the medium and from the lysed cells (Fig. 4). The wild-type precursor was rapidly secreted into the medium with a half-time of less than 1 h (lanes 1-8). Because COS cells lack a regulated secretory pathway and the corresponding processing proteases, pro-vasopressin-NPII was secreted intact. The secreted protein was resistant to Endo H digestion (lanes 9 and 10), indicative of conversion of the N-linked glycan from high mannose to complex type in the Golgi apparatus. In contrast, the mutant precursor was almost completely retained inside the cells in an Endo H-sensitive form (lanes [13][14][15][16][17][18][19][20][21][22][23][24][25][26]. Localization of the mutant protein by indirect immunofluorescence showed strong staining with the typical reticular pattern of the ER extending through the entire cell (Fig. 5A). Double immunofluorescence staining for the ER resident protein p63 (Fig. 5D) (29) and for mutant vasopressin precursor (Fig. 5C) produced identical patterns in expressing cells. For comparison, the weak staining pattern of cells expressing the wild-type protein reflects the steady state distribution of the precursor in the secretory pathway, i.e. the network of the ER and a perinuclear spot typical of the Golgi apparatus (Fig. 5B).
An uncleaved, hydrophobic signal peptide could cause ER retention by anchoring the protein in the ER membrane. Membrane integration was tested by alkaline extraction (Fig. 2B). Microsomes of labeled cells were incubated at pH 11.5 to disrupt organelles and release peripherally associated and soluble proteins. As expected, wild-type vasopressin precursor was recovered in the soluble extract (lanes 1 and 2), whereas subunit H1 of the asialoglycoprotein receptor, a control integral membrane protein, was pelleted with the membranes (lanes 5 and  6). The mutant vasopressin precursor was as almost as efficiently extracted as the wild-type protein (lanes 3 and 4), suggesting that it was not anchored in the bilayer of the ER.
The Uncleaved Signal Causes Disulfide Aggregation but Does Not Prevent NPII Folding-The uncleaved hydrophobic signal peptide might cause ER retention by interfering with correct ]methionine, lysed, and subjected to immunoprecipitation using an antiserum against NPII. Immunoprecipitates were incubated with or without Endo H and analyzed by SDS gel electrophoresis and fluorography. The positions of marker proteins are indicated with their molecular masses in kDa. (Expression of the mutant cDNA reproducibly yielded better incorporation of radioactivity into immunoprecipitable protein, probably because of the additional methionine remaining with the uncleaved signal.) B, alkaline extraction. Microsomes of labeled COS-7 cells were incubated at pH 11.5 and centrifuged through a sucrose cushion. Vasopressin precursor was then separately immunoprecipitated from the membrane pellet (P) and the supernatant (S) and analyzed by gel electrophoresis. As a control for a membrane-anchored protein, subunit H1 of the asialoglycoprotein receptor was expressed and analyzed in parallel. FIG. 1. Sequence of wild-type and mutant prepro-vasopressin-NPII. A, the general organization of the precursor protein is schematically shown with the disulfide-linked cysteines connected by lines and the site for N-glycosylation indicated with a diamond. The protein segments produced by proteolytic cleavage of the precursor are visualized by different shades of gray; the linker sequences are cross-hatched. B, the N-terminal sequence of the wild-type protein (V) and of the predicted mutant precursor protein (Vm) is shown.
folding and disulfide bonding of the precursor. Pro-vasopressin-NPII contains eight intramolecular disulfide bridges. Formation of disulfide bonds is often reflected in the electrophoretic mobility of a protein under reducing versus nonreducing conditions. To address this, transfected COS cells were metabolically labeled and then incubated at 4°C with iodoacetamide to alkylate free sulfhydryl groups to prevent post-lysis oxidation. Vasopressin precursor was immunoprecipitated and split into two samples that were boiled in SDS sample buffer with or without mercaptoethanol. Upon gel electrophoresis, the wildtype precursor migrated slightly faster in its nonreduced, more compact folded state than after reduction of the disulfide bonds (Fig. 6, lanes 1 and 3). In contrast, the nonreduced mutant precursor was hardly detectable at the corresponding position in the gel but appeared as aggregates that migrated more slowly or did not enter the gel at all (lane 4). Such aggregates are typically observed for misfolded proteins (28).
To more directly assess the folding state of the protein, wild-type and mutant precursors were subjected to a protease sensitivity assay. Labeled protein was immunoprecipitated using anti-NPII antiserum and protein A-Sepharose and was incubated with increasing concentrations of trypsin. The Sepharose beads were washed, and bound material was analyzed by gel electrophoresis and fluorography (Fig. 7). From the wild-type precursor (lanes 1-5), a fragment of ϳ12 kDa (fragment 3) was generated that was completely resistant to trypsin at up to 5000 g/ml. Its size corresponds to that of the NPII domain, indicating that the flanking segments (vasopressin and glycopeptide) were cleaved off at the connecting peptides normally hydrolyzed by processing enzymes in secretory granules. Intermediates of ϳ21 and ϳ13 kDa (fragments 1 and 2) were weakly detectable at the lowest trypsin concentration (lane 2). Their sizes suggest that these intermediates correspond to NPII-glycopeptide and vasopressin-NPII fragments, respectively. This was confirmed by performing the same experiment using an antiserum directed against the vasopressin nonapeptide (data not shown). Only the full-size precursor and the 13-kDa intermediate (fragment 2) were recovered, indicating that they contain the vasopressin sequence.
Upon incubation of the mutant precursor with trypsin ( Fig.  7, lanes 6 -10), the 12-kDa NPII fragment was generated as well, although it was slightly less resistant to the highest protease concentrations. Intermediate fragments of ϳ21 kDa (fragment 1) lacking the uncleaved signal and vasopressin, and of ϳ14 kDa (fragment 2*) lacking the glycopeptide portion were also produced. The latter intermediate was the only one to be recovered using anti-vasopressin antibody (not shown), supporting the notion that it consists of the uncleaved signal, vasopressin, and NPII. It is more abundant than the corresponding fragment of the wild-type precursor (even considering that it contains an additional methionine in the signal sequence). The uncleaved signal thus appears to affect the acces- sibility of the peptide connecting vasopressin to NPII. Most importantly, it did not interfere with the folding of NPII itself.
In contrast, the precursor Vm⌬, a cloning artifact of Vm with an additional small deletion of codons 28 -36 of pro-vasopressin-NPII (residues 16 -24 of NPII), did not generate specific trypsin-resistant fragments (Fig. 7, lanes 11-15). Vm⌬ thus serves as a control showing that protease resistance depends on NPII folding.
The Uncleaved Signal Interferes with Disulfide Formation in Vasopressin-In the wild-type pro-vasopressin-NPII, all cysteines are engaged in disulfide bonds, seven in NPII and one in  vasopressin. The mutant precursor, however, contains an additional cysteine, Cys Ϫ11 , in the uncleaved signal peptide (Fig.  1B). To test whether this cysteine is responsible for the formation of disulfide aggregates and retention, we constructed the mutant Vm(CϪ11S), in which Cys Ϫ11 in the truncated signal sequence of Vm was mutated to serine. As shown in Fig. 8, mutation of this cysteine did not significantly affect the properties of the protein. It was synthesized by COS cells as a 23-kDa glycosylated, Endo H-sensitive protein (Fig. 8A, lanes 1  and 2), indicating that the truncated signal without cysteine was functional and uncleaved. The majority of the protein was retained in an Endo H-sensitive form within the cell (lanes 5-8). After 2 h of chase, ϳ7% was recovered from the medium as Endo H-resistant protein, which is slightly more than for Vm. Trypsin resistance of the 12-kDa NPII fragment indicates correct folding of NPII (Fig. 8B). Most notably, elimination of Cys Ϫ11 did not prevent most of the protein from forming disulfide-linked aggregates (Fig. 8C), indicating that covalent aggregation is not, or not solely, caused by Cys Ϫ11 . The extraordinary protease resistance of the NPII portion in Vm and Vm(CϪ11S) argues that its seven disulfide bonds were correctly formed. Thus, the disulfide bond in the vasopressin segment may not be correctly formed in the mutant precursors and cause aggregation.
The first residue of vasopressin, Cys 1 , forms a disulfide bridge with Cys 6 . It is conceivable that lack of signal cleavage in Vm and Vm(CϪ11S) interferes with the formation of this disulfide bond, leaving its thiol groups potentially available for oxidation to form covalent aggregates. To test whether unpaired cysteines of vasopressin might cause ER retention, we constructed the mutant V(C6S). Cys 6 was mutated to serine in the context of the wild-type precursor sequence, leaving Cys 1 without its normal disulfide partner, even if the signal is removed. Upon expression in COS cells, V(C6S) showed an electrophoretic mobility of 22 kDa like the wild-type precursor (Fig.  9A, lanes 1 and 2), indicative of signal cleavage. Nevertheless, it was not secreted from the cells and remained sensitive to Endo H digestion (lanes 6 -9). The mutation did not interfere with folding of NPII, because the trypsin-resistant NPII fragment of ϳ12 kDa was generated in the protease sensitivity assay (Fig. 9B). However, V(C6S) was also found largely as aggregates when analyzed under nonreducing conditions (Fig.  9C). These results indicate that failure to form the disulfide bond in vasopressin is sufficient to cause aggregation and ER retention of the precursor protein.
The guanosine 227 deletion in Vm thus acts indirectly. By eliminating the normal translation initiation site, it induces translation of a truncated signal sequence that, because it cannot be cleaved, prevents the correct folding and oxidation of vasopressin. This is ultimately sufficient to cause aggregation and ER retention.

DISCUSSION
In this study, we have analyzed the mechanism by which a mutation in the translation initiation codon of prepro-vasopressin-NPII causes ADNDI in a Swiss kindred. The mutant precursor was expressed and glycosylated, indicating that translation initiated at an alternative ATG codon and produced a functional signal for ER targeting and translocation (Fig. 2). The only possible initiation site is codon Ϫ15 generating a signal sequence truncated by four residues. This shortened signal is in better agreement with the typical characteristics of ER targeting signals, because an unusual N-terminal negative charge, Asp Ϫ17 , is deleted. The immediate consequence of the truncation was the complete loss of signal cleavage, which is evident from a molecular mass increase of ϳ2 kDa.
It is not clear by which mechanism signal peptidase cleavage was affected. The available algorithms to predict signal cleavage sites (30,31), which are based on statistical evaluation of known sequences, yield the same result for the wild-type and the mutant sequence. The main criteria are the residues close to the cleavage site (particularly residues Ϫ1 and Ϫ3, which preferentially have small and uncharged side chains) and the vicinity to the apolar segment of the signal. These features are not affected by the mutation. However, it is conceivable that the mutant signal is positioned differently within the translocation machinery, because the hydrophilic N-terminal segment with Asp Ϫ17 is lacking. As a result, accessibility of the cleavage site for signal peptidase might be reduced. Nilsson et al. (32) previously reported evidence for a subtle position effect on signal cleavage depending on the length of the hydrophobic signal domain. An effect of the N-terminal hydrophilic segment on signal cleavage has been observed for major histocompatibility complex class II invariant chain and the asialoglycoprotein receptor H1. Both proteins are type II membrane proteins with uncleaved signal anchor sequences. Deletion of the Nterminal hydrophilic portions resulted in partial signal cleavage at cryptic sites that normally are not accessible to signal peptidase (33,34). How alterations in the N-terminal domain of a signal sequence can affect the situation of the signal in the insertion process is not clear, even more so because it is not known precisely in what environment the signal is situated at the time of cleavage (i.e. inside the translocation pore or within the lipid bilayer).
The uncleaved mutant vasopressin precursor was efficiently retained in the ER in a high mannose glycosylated form (Figs. 4 and 5) and could be recovered largely in disulfide-linked aggregates (Fig. 6) typical of unfolded or misfolded proteins (28). All the cysteines in the precursor are normally involved in intramolecular disulfide bonds, except for one in the signal sequence. However, folding of the NPII domain, which contains 7 of the 8 disulfide bonds in the precursor, was not significantly affected by the uncleaved signal, because the NPII domain was found to be almost as resistant to trypsin digestion as in the wild-type protein (Fig. 7). Based on the positions of the N and the C termini of NPII in the crystal structure (35), the small, C-terminal glycopeptide domain is not in contact with the signal vasopressin segments and is thus unlikely to be disturbed during folding. In addition, it does not contain any cysteines and thus cannot be responsible for disulfide aggregation. Hence, retention and aggregation is caused by the uncleaved signal and/or by misfolded vasopressin. Cys Ϫ11 of the signal sequence is not solely responsible, because its mutation to serine did not eliminate aggregation and did not allow secretion (Fig. 8). Our results therefore suggest that the disulfide bond in vasopressin is not formed in the uncleaved mutant protein, leaving the free thiols available for incorrect intermolecular disulfide formation. Consistent with this model, we found that preventing disulfide formation in vasopressin by mutating one of its two cysteines is sufficient to cause ER retention and aggregation, even if the signal is removed correctly (Fig. 9).
The crystal structures of bovine NPII complexed with a dipeptide analog of vasopressin (35) and of the closely homologous bovine neurophysin I-oxytocin complex (36) have been determined. They show that the ␣-amino group of the hormone is hidden in the binding pocket, forming three hydrogen bonds and a salt bridge, and that the disulfide bond of the hormone is facing into the binding site. Binding of the N-terminal vasopressin segment of pro-vasopressin-NPII into the hormone binding site may be essential to stabilize the intramolecular disulfide bond between Cys 1 and Cys 6 by protecting it from further disulfide isomerization. If the signal sequence is not removed, this is impossible, and the signal and vasopressin will remain exposed. The observation that the NPII domain of the wild-type precursor is somewhat more stable at high trypsin concentrations than that of the mutants Vm or V(C6S) (Figs. 7 and 9B) may reflect stabilization of NPII when the correctly oxidized hormone is bound to its binding site. A stabilizing effect of the ligand on NPII has previously been documented in in vitro folding studies (37).
In the prepro-vasopressin-NPII precursor sequence, close to 30 different mutations causing ADNDI have been identified (3,(5)(6)(7)(8)(9). Besides the mutation studied here, three localized to the signal sequence. Ala Ϫ1 was mutated to threonine in six kindred (3,8,13,38,39), representing the most frequent ADNDI mutation known so far. In addition, substitution of Ala Ϫ1 by valine (3,9) and of Ser Ϫ3 by phenylalanine (3) have been identified. These signal mutations involve the residues that are most critical for signal peptidase activity. Reduced signal cleavage was experimentally demonstrated in vitro for the alanine-tothreonine mutation (38). These cleavage site mutants are very likely to cause ER retention by the same mechanism as the guanosine 227 deletion mutant analyzed in this study. There is only one known case of ADNDI with a mutation in vasopressin itself (Tyr 2 to His) (6). Because Tyr 2 is essential for vasopressin binding to NPII, Cys 1 and Cys 6 of the mutant protein are likely to remain exposed to the ER lumen, with the same consequences as observed for the signal mutant analyzed here. All other mutations were distributed throughout the NPII se-  Fig. 7, and disulfide aggregation (C) was analyzed as in Fig. 6. CϪ, cells without Endo H digestion; Cϩ, cells with Endo H digestion; MϪ, medium without Endo H digestion; Mϩ, medium with Endo H digestion; red., reduced; non-red., nonreduced.
quence and include missense, deletion, and nonsense mutations that most likely cause ER retention by disturbing the correct folding of the NPII domain.
The gradual development of the dominant phenotype of ADNDI and postmortem studies suggest that the disease is caused by the degeneration of the hypothalamic cells producing mutant hormone precursor. In a recent study, reduced viability of neuro2A neuroblastoma cell lines expressing different vasopressin precursor mutants (a signal cleavage site mutant and several NPII mutants) was observed (20), indicating that the retained precursor (or its degradation products) has a cytotoxic effect. ER retention, first shown for the glycine 17-to-valine mutation of NPII (40), and the resulting cytotoxicity might be a common feature of many, if not all, prepro-vasopressin-NPII mutations. This is reminiscent of ␣ 1 -antitrypsin deficiency, which causes liver cirrhosis in susceptible individuals. Mutant antitrypsin was found in hepatocytes in characteristic insoluble inclusions that were associated with hepatocellular damage (41). Several mutant proteins, including the relatively frequent Z variant, were shown to be retained in the ER in a misfolded form producing large noncovalent homopolymers of a specific structure (loop-sheet polymerization) (42,43). In contrast to this mechanism, mutant vasopressin precursors were not found in inclusions but rather evenly distributed throughout the ER (Fig. 4 and Ref. 20). The disulfide aggregates appear to be heterogeneous, suggesting that disulfide linkage occurs to other still unfolded proteins in the ER rather than by oxidative homooligomerization. In addition, the mutations that cause ADNDI are quite different. For example, truncation of the C-terminal third of NPII is very likely to cause gross misfolding of the rest of the precursor, whereas lack of signal cleavage in the guanosine 227 deletion mutant did not impede NPII folding (Fig. 7). Elucidating the mechanism of cytotoxicity by retained vasopressin precursors is the main challenge in understanding the molecular pathogenesis of ADNDI and may provide important insights into the mechanisms underlying neurodegenerative diseases.