The Carboxyl-terminal 90 Residues of Porcine Submaxillary Mucin Are Sufficient for Forming Disulfide-bonded Dimers*

COS-7 cells transfected with expression vectors encoding 90 and 154 amino acid residues, respectively, from the carboxyl terminus of the disulfide-rich domain (240 residues) of porcine submaxillary mucin were shown to form disulfide-bonded dimers. Cells with expression vectors that encoded the disulfide-rich domain lacking the last 90 and 150 carboxyl-terminal residues, respectively, from the carboxyl terminus of the disul-fide-rich domain were unable to secrete truncated domains. These results indicate that the information required to form disulfide-bonded dimers resides in only 90 residues, including 11 half-cystines. Site-specific mutagenesis was employed to change, one at a time, each codon for the 11 half-cystines to serine. Eight of the 11 mutants formed disulfide-bonded dimers indistinguishable from those produced by unmutated vector, although 6 of the 8 mutants also produced aggregates thought to be misfolded protein with scrambled disulfide bonds. Two additional mutant vectors encoding serine instead of half-cystine at residues 13244 and 13246 in submaxillary mucin expressed both monomers and dimers of the disulfide-rich domain but no aggregates. The final mutant vector, C13223S, expressed protein aggregates that were poorly secreted from transfected cells. A mutant vector with two codon changes, C13244A/ C13246A, expressed both monomers and dimers, just like the single mutants at these half-cystines. These results suggest that three half-cystine residues (Cys 13223 , Cys 13244 , and Cys 13246 ) may be involved in forming interchain disulfide

Porcine submaxillary mucin contains a polypeptide chain of 13,288 residues, 1 which can be divided into domains typical of those found in several other mucins secreted by various human tissues (1). Most of the polypeptide chain is comprised of 81 residue tandem repeats that are rich in serine, threonine, gly-cine, and alanine, and together account for 75% of the residues in this domain. There are at least three different genes encoding porcine submaxillary mucin that differ from one another in the number (90, 125, and 135) of repeats they encode (1). The tandem repeat domain is flanked on either end by unique sequences that are similar in composition to the repeat domains but exhibit no sequence identity to one another or to the tandem repeat domains. The hydroxyl groups of serine and threonine in the tandem repeat domain (2), and presumably the unique sequences flanking this domain, are in O-glycosidic linkage with oligosaccharides, which account for about 75% of the weight of native mucin. Three disulfide-rich domains are at the amino terminus of submaxillary mucin (1). These domains show considerable sequence identity to the D-domains found in human prepro-von Willebrand factor 2 (3), frog integumentary mucin FIMB.1 (4), and a human intestinal mucin (MUC2) 3 (5). The carboxyl terminus of submaxillary mucin is formed by a 240-residue disulfide-rich domain, which has significant sequence identity with similar domains at the carboxyl terminus of human prepro-von Willebrand factor (3), frog integumentary mucin FIMB.1 (6), and human mucins secreted by different tissues, including MUC2, MUC5AC, 4 MUC5B, 5 and MUC6 6 (5,(7)(8)(9)(10). Moreover, the 11 half-cystines in a 133-residue protein (norrin) 7 associated with Norrie disease in humans (11,12) are conserved in the carboxyl-terminal disulfide-rich domains of human prepro-von Willebrand factor, porcine submaxillary mucin, and several human mucins (13). It has been shown that the disulfide-rich domain of prepro-von Willebrand factor (14), porcine submaxillary mucin (15), and norrin (16) forms interchain disulfide bonds between two polypeptide chains in these molecules.
We describe here studies designed to determine whether the entire 240-residue carboxyl-terminal disulfide-rich domain is required for the formation of disulfide-linked dimers of porcine submaxillary mucin. The approach taken was to mutate plasmids encoding the domain by deletion of segments of DNA from either the 5Ј or the 3Ј end and to examine the ability of the truncated plasmids when transfected into COS-7 cells to form disulfide-bonded dimers. Further studies by site-specific mutagenesis permitted determination of those half-cystine residues that appear to form the interchain disulfide bonds in dimers.
The vector pMCab was made by introducing into pMC a PstI site downstream to the Gly 13213 codon with the antisense mutagenesis oligonucleotide 5Ј-CAACTTTAATCCTGCAGACCATTGTACC-3Ј. The new plasmid encodes a transin signal peptide of 25 residues, 3 residues (GRE) that precede a truncated mucin domain (residues 13045-13213), and 3 additional residues (LQD) at the carboxyl terminus. To construct plasmid pMCa, a frameshift in the mucin coding region of pMC was created downstream from the Glu 13128 codon with the following antisense mutagenic oligonucleotide, 5Ј-TTTTTCACAGTGTCTCGAGTTC-ACAGCAGG-3Ј. This plasmid expressed a mutant protein containing a transin signal peptide of 25 residues, followed by 3 amino acids (GRE) and a truncated mucin domain (residues 13045-13128) and the tetrapeptide LETL at the carboxyl terminus.
The expression vector pMCcH was constructed by replacing the protein A coding sequences of plasmid pPA-MC1 (15) with a new multicloning site as described earlier (15). The resulting plasmid was then digested with XhoI and SpeI and ligated to a XhoI/SpeI mucin DNA fragment encoding residues 12572-13288 (1). A new XbaI site was made by site-specific mutagenesis upstream from the codon for serine residue 12572 with the following antisense mutagenic primer; 5Ј-GG-TTTGCAACTACTTCTAGATGTATAGCAACA-3Ј. The following oligonucleotide was used as selection primer: 5Ј-CTAGAAAGAATTCGAC-GACCTTCTTCAC-3Ј, which destroys a unique SalI site in the multicloning site of the template plasmid. A poly-His-encoding cassette was inserted upstream to the stop codon by published procedures (15). The resulting plasmid was digested with XbaI and religated to give pMCcH, which encodes a fusion protein containing a 25-residue transin signal peptide, 7 residues of the cloning site (GRRILSR), followed by 90 residues at the carboxyl terminus of mucin (residues 13198 -13288) and 6 histidine residues. The Cys to Ala mutants of pMCcH were constructed with the same antisense mutagenic oligonucleotides used with pMC, and the antisense selection primer 5Ј-GCAACTACTTCTGGAAAGAAT-TCG-3Ј, which destroys a unique XbaI site. To facilitate the screening process, the mutagenic oligonucleotides were designed to produce silent mutations that destroy or create restriction sites in the mucin sequences along with the desired codon changes. All mutations in the expression vectors were confirmed by DNA sequencing with the Sequenase 2.0 system from U. S. Biochemical Corp. The entire polynucleotide encoding the mutant C13223S was completely sequenced to establish that no unintended mutation had been made.
Transfection of Mammalian Cells-COS-7 cells were grown and transfected with Lipofectin (Life Technologies, Inc.) as described (15) except that chloroquine, at a final concentration of 50 mM, was included during the 6-h transfection. NIH-3T3 cells expressing mutant or normal mucin domains were obtained by co-transfection of cells with plasmid pHBAPr1neo (19) and the indicated construct as reported earlier (15).
Purification and Analysis of Recombinant Proteins-Antisera 3814 against the carboxyl-terminal disulfide-rich domain of mucin was described earlier (15). Metabolic labeling of cells with Tran 35 S-label or 35 S-cysteine (both from ICN Pharmaceutical, Inc.) and immunoprecipitation with antisera 3814 were performed as described earlier (15) except that immunoprecipitates were washed further with 50 mM Tris-HCl, pH 8.0, 0.25 M NaCl containing 0.01% each of SDS, sodium deoxycholate, and Nonidet P-40. To block free thiol groups that could permit formation of incorrectly paired half-cystines during cell lysis and immunoprecipitation, intracellular precursors were reacted in vivo with iodoacetamide as reported earlier (20), and media were reacted with iodoacetamide prior to immunoprecipitation. SDS-gel electrophoresis and autoradiography were done as described earlier (15). 14 C-labeled proteins (Amersham Pharmacia Biotech) were used as molecular weight standards, including myosin (220,000), phosphorylase b (97, 400 -100,000), bovine serum albumin (66,000), ovalbumin (46,000 -50,000), carbonic anhydrase (30,000), and lysozyme (14,300).
Detection of Free Thiol Groups-His-tagged proteins were purified from the medium of transfected cells on TALON-IMAC beads (CLON-TECH) (15), eluted from the beads with 0.5 ml of Tris, pH 8, containing 10 mM EDTA, and 6 M urea (binding buffer), and then applied to a column of activated thiolpropyl-Sepharose 6B (Sigma) equilibrated with binding buffer. After 2 h at 25°C, the column was sequentially eluted with binding buffer, the same buffer containing 1 M NaCl, and then with binding buffer-NaCl containing 10 mM dithiothreitol. Fractions (0.15 ml) were collected and counted for radioactivity. In some experiments, fractions were desalted on Sephadex G-25 (Pharmacia-LKB) in 50 mM Tris, pH 8, containing 0.15 M NaCl and 0.1% Triton X-100, and the eluted domain was isolated by immunoprecipitation with antiserum 3814. Immunoprecipitates were analyzed as described earlier (15) by SDS-gel electrophoresis and autoradiography. (15) is a plasmid that encodes a fusion protein containing a transin signal peptide of 25 residues followed by 3 residues (GRE) derived from a multicloning site and then 244 residues that encode the disulfide-rich domain at the carboxyl terminus of mucin. As a first step to identify which of the 30 half-cystine residues are involved in formation of interchain disulfide bonds between two of the disulfide-rich domains, pMC was truncated at its amino terminus by site-specific mutagenesis. Two truncated plasmids were produced. One, designated pMCbc, is devoid of the first 86 residues of the disulfide-rich domain and encodes 158 residues (residues 13131 to 13288 in mucin 1 ) including 20 half-cystine residues. The other, designated pMCcH is devoid of the first 154 residues of the disulfiderich domain and encodes the last 90 residues (residues 13198 to 13288) at the carboxyl terminus of the domain, including 11 half-cystine residues, followed by 6 histidine residues, which were introduced to aid in purification of the protein expressed by the plasmid. Fig. 1A shows the electrophoretic pattern on SDS gels in 2-mercaptoethanol of the proteins immunoprecipitated from the medium of COS-7 cells that had been transfected with the truncated plasmids and metabolically labeled with Tran 35 S-label. The protein produced by pMCbc (lane 2) migrated as a broad band with M r ϭ 22,000 -28,000, whereas pMCcH (lane 3) was expressed as two proteins with M r ϭ 17,000 and 19,000, respectively. However, if the proteins produced by these plasmids were digested with N-glycanase, only one protein species was observed on gels (Fig. 1B). pMCbc gave a protein with M r ϭ 22,000, and pMCcH gave a protein with M r ϭ 17,000. Treatment of transfected cells with tunicamycin also showed single protein species of these sizes on gel electrophoresis (data not shown). These results are similar to those obtained earlier (15) on transfection of COS-7 cells with pMC and show that the different sizes of proteins expressed by the plasmids result from different extents of N-glycosylation, with pMCbc containing two glycosylation sites and pMCcH containing one. Fig. 1A also shows the proteins produced by COS-7 cells transfected with these plasmids and analyzed on nonreducing gels. pMCbc (lane 5) produced a protein with M r ϭ 48,000, about twice the size of the species observed in the presence of reducing agent. pMCcH (lane 6) produced three protein bands with M r ϭ 33,000, 35,000, and 37,000, respectively. These results indicate that the proteins produced by the plasmids formed disulfide-linked dimers. The different protein bands observed with pMCcH are the result of dimer formation among the two monomeric species observed under reducing conditions. These results also indicate that the half-cystine residues involved in forming interchain disulfide bonds in disulfide-rich domain dimers (15) are among the 11 half-cystines contained in the last 90 amino acids from the carboxyl terminus of the domain. This conclusion is consistent with the observation that transfection of COS-7 cells with 2 other plasmids, pMCab and pMCa, which lacked DNA encoding the last 90 and 150 carboxyl-terminal residues, respectively, in the disulfide-rich domain, did not secrete any proteins into the medium as judged by SDS-gel electrophoresis of immunoprecipitates (Fig. 1C).

Expression and Secretion of Truncated Forms of the Carboxyl-terminal Disulfide-rich Domain of Mucin-The expression vector pMC
Secretion and Dimer Formation of Half-cystine to Serine Mutants of the Disulfide-rich Domain-By site-specific mutagenesis of pMC, each of the codons for the 11 half-cystine residues contained in the last 90 residues of the disulfide-rich domain was changed one at a time to the codon for serine. The proteins expressed by each of the mutant plasmids were then determined after transfection of COS-7 cells with the plasmids followed by metabolic labeling of the cells with Tran 35 S-label, immunoprecipitation of the proteins from the medium, and autoradiography of the proteins after gel electrophoresis. As shown in Fig. 2A, all mutants expressed and secreted into the medium five proteins species ranging in size from M r ϭ 33,000 to 47,000 on gels run in the presence of reducing agent. The five proteins differ in the extent of N-glycosylation as shown earlier (15). The amounts of protein expressed by each mutant appears to be about the same as that for pMC (lane 2) except C13223S (lane 11), which secreted lesser amounts of the domain. C13279S (lane 4) showed more radioactivity at the interface between the stacking and running gels than the amounts seen normally with unmutated proteins (lane 2) and with preimmune serum (lane 1). This suggests that aggregates of the disulfide-rich domain are expressed by these plasmids, especially in the case of C13200S (lane 9), C13214S (lane 10), and C13247S (lane 7). The nature of these aggregates is unknown, but they likely result from incorrect formation of interchain and intrachain disulfide bonds that contain many covalently linked monomeric polypeptides. Fig. 2C shows the autoradiographs of the proteins on nonreducing gels that were expressed by COS-7 cells transfected with two mutant plasmids not shown in Fig. 2B. The majority of the protein expressed by these mutants, C13244S (lane 4) and C13246S (lane 3), were of about the size (M r ϭ 73,000) as that of disulfide-bonded dimers expressed by pMC (lane 2), the unmutated plasmid. Moreover, there was no more radioactivity between the running and stacking gels than seen with the pMC (lane 2) and with preimmune serum (lane 1), indicating that they did not express aggregates. However, each mutant plasmid expressed five protein bands, when analyzed on nonreducing gels, that are characteristically seen on reducing gels ( Fig.  2A). These results suggest that a small but significant amount of non-disulfide-bonded monomer is secreted into the medium of the cells transfected with these two plasmids. The unreduced monomeric species expressed by C13244S migrated slightly faster than those from C13246S.  2 and 4) and was present in reducing gels as five species of protein (lane 2) resulting from differences in the extent of N-glycosylation. When transfected cells were radiolabeled for 4 or more hours, immunoprecipitates of the cell lysates from transfected COS-7 cells also showed a faint band with M r ϭ 80,000 (Fig. 3B, lane  1). This protein was neither absorbed by the pre-immune serum (lane 3) nor observed in immunoprecipitates from cells expressing the unmutated domain (lane 2). Moreover, it is not found in the medium (Fig. 3A, lane 1). The nature of this protein is not known, but it did not immunoprecipitate with antiserum to BiP (GRP78) or to calnexin, chaperones (21) that could be involved in folding of the domain. Pulse-chase studies with the mutant plasmid as shown in Fig. 3C indicated that the expressed protein was still present in transfected COS-7 cells 3 h after synthesis of the protein commences, the time when they were endoglycosidase H-sensitive. This result suggests that intracellular precursors of C13223S are retained in the endoplasmic reticulum during the chase period. Proteins from unmutated plasmids that express the disulfide-rich domain were found earlier (15) to be completely secreted after 2 h. Thus, it is very likely that the mutant expressing C13223S impairs dimerization and secretion of the disulfide-rich domain of mucin. Figs. 2 and 3 indicated that three of the half-cystine to serine mutant plasmids, C13223S, C13244S and C13246S, expressed proteins from transfected cells that were different in some respect from that expressed by the other mutant and unmutated plasmids. To determine whether an amino acid other than serine produced the same kinds of effects on the disulfiderich domains, these three half-cystines were changed to alanine by site-directed mutagenesis, and the proteins produced were analyzed as described above. Fig. 4A shows that the proteins expressed by C13223A, C13244A, and C13246A behave exactly like the corresponding proteins that contained serine instead of alanine. The secreted forms of C13244A (lane 9) and C13246A (lane 8) showed substantial amounts of dimers on nonreducing gels but also had small but significant amounts of the five monomeric species that differ in the extent of N-glycosylation and are characteristic of the monomeric species observed on reducing gels (lanes 3 and 4). Moreover, the protein expressed by the mutant C13223A is poorly secreted from the cells (lanes 5 and 10) and is recovered from cell lysates as disulfide-bonded aggregates (lane 6), with no indication of dimeric species. Fig. 4B shows autoradiograms of the proteins expressed by NIH-3T3 cells that were stably transfected with the mutant plasmids that have the C13244A (lane 4) and the C13246A (lane 3) mutation. The behavior of the mutant disulfide-rich domain is essentially identical to that obtained from transiently transfected COS-7 cells. Thus, the mammalian expression system appears to have no effect on the nature of the disulfide-rich domains expressed.  2 and 4), obtained by lysis in buffered Triton X-100, and medium (lanes 1 and 3) were purified as in Fig. 1A and analyzed by SDS-gel electrophoresis in the presence (lanes 1 and 2) or absence (lanes 3 and 4)  Site-specific mutagenesis was also used to prepare a plasmid that encoded a disulfide-rich domain with both half-cystine 13244 and 13246 changed to alanine. Fig. 4C shows SDS gel electrophoresis of the proteins secreted into the medium by COS-7 cells transfected with this plasmid and then metabolically labeled with Tran 35 S-label. Except for the presence of aggregates, this plasmid produced the same pattern of proteins on unreduced gels as those found when cells were transfected with plasmids with only one of either of the two codons for half-cystine changed to alanine. Dimers were the predominant species secreted, but significant amounts of the five monomeric species and some aggregates were observed (lane 2). This suggests that half-cystines other than C13244 and C13246 may be involved in formation of disulfide-bonded dimers.

Secretion and Dimer Formation of Half-cystine to Alanine Mutants of the Disulfide-rich Domain-The studies shown in
Two mutants of pMCcH, which encodes the carboxyl-terminal 90 residues of the disulfide-rich domain, were also subjected to site-specific mutagenesis to attempt to gain further insight into the possible role of half-cystine 13244 and 13246 in dimer formation. Fig. 5 shows the autoradiographs of the proteins secreted into the medium by COS-7 cells transfected with mutant plasmids of pMCcH that have the C13244A or the C13246A mutation. The amounts of protein expressed by the cells transfected with the mutant plasmids were lower than those from cells transfected with normal plasmid, but the species secreted into the medium by the cells were the same, except that monomers were not found in the media of cells transfected with normal plasmid. Cells transfected with the mutant plasmids secreted dimers and monomers of the disul-fide-rich 90-residue domain as observed on nonreducing gels (Fig. 5). The mutant protein produced by C13244A (lane 8) had three monomeric species on nonreducing gels. The fastest moving species may have a different pattern of disulfide bonds than the other two. A third mutant vector encoding the carboxylterminal 90 residues of mucin with C13244 and C13246 replaced by alanine expressed and secreted the same species as the single alanine mutants for each half-cystine (data not shown). This is further evidence that the carboxyl-terminal 90 residues contain the information necessary for dimer formation.
Determination of the Free Thiol Content of the Disulfide-rich Domain-To understand more exactly the nature of the halfcystine residues in the disulfide-rich domain, it was important to know whether the domain contained any free thiol groups. Thus, cells were transiently transfected with pMCH (15), which encoded the same domain as pMC plus six histidine residues at its carboxyl terminus. The cells were metabolically labeled, and the proteins secreted into the medium isolated by absorption onto and elution from TALON-IMAC adsorbant. After denaturation in 6 M urea, the proteins were applied to a column of thiolpropyl-Sepharose. Fig. 6 shows the elution profile of the column. It can be seen that all of the radioactivity emerged unretarded from the column. Autoradiograms are also shown in Fig. 6B of the different fractions emerging from the column. Only monomeric species that had been N-glycosylated to different extents were observed. Fig. 6A also shows that the domain is completely absorbed to the thiolpropyl-Sepharose column if first reduced with 2-mercaptoethanol to break the disulfide bonds and can be eluted from the column with dithiothreitol. These results indicate that the 30 half-cystines in the disulfide-rich domain are in disulfide bonds, and the domain is devoid of free thiol groups. DISCUSSION The present studies suggest that formation of disulfidebonded dimers of the carboxyl-terminal disulfide-rich domain of porcine submaxillary mucin is dependent only on structures in the 90 carboxyl-terminal residues of mucin, including 11 half-cystines (Fig. 1). By site-specific mutagenesis, each of the codons for the 11 half-cystines was changed one at a time to serine, and the ability of the resulting mutant proteins to form disulfide-bonded dimers examined (Fig. 2). Two of the mutants, C13227S and C13283S, were identical to the normal domain. Six others, C13200S, C13214S, C13247S, C13261S, C13277S, and C13279S, formed dimers indistinguishable from normal protein although they also formed aggregates of disulfidebonded mutant protein. The aggregates are thought to be improperly folded molecules with scrambled disulfide bonds because they give monomers on reduction. These results suggest that these eight half-cystines are not involved in forming the interchain disulfide bonds of dimers. Moreover, these eight half-cystines likely form intrachain disulfide bonds because the disulfide-rich domain is devoid of free thiol groups (Fig. 6). Two other mutant proteins, C13244S and C13246S, form substantial amounts of monomers and dimers under nonreducing conditions, suggesting that interchain disulfide bond formation is impaired in these mutants. They did not, however, form aggregates. The final mutant, C13223S, does not form dimers but forms disulfide-bonded aggregates that are very poorly secreted from the cells (Fig. 3). Thus, the three half-cystines at residues 13223, 13244, and 13246 would seem to be likely candidates for forming interchain disulfide bonds.
The mutant proteins, C13244A and C13246A, were indistinguishable from their serine counterparts and showed both monomers and dimers and no aggregates under nonreducing conditions (Fig. 4). Surprisingly, a double mutant construct, C13244A/C13246A, expressed and secreted monomeric and dimeric species in addition to aggregates. These results suggest that one other half-cystine is involved in dimer formation and that at least two interchain disulfide bonds may exist in dimers. It is possible that half-cystine 13223 is also involved, but its role could not be tested by site-specific mutagenesis because proteins with this half-cystine mutated are poorly secreted from the cells. Fig. 7 compares the amino acid sequences of the disulfiderich carboxyl-terminal domain of porcine submaxillary mucin 1 with the corresponding domains in human mucins MUC2, 3 MUC5AC, 4 MUC5B, 5 and MUC6 6 and in human prepro-von Willebrand factor 2 and human norrin. 7 Noteworthy is the almost identical position of the 11 half-cystine residues in each sequence. Experimental proof for dimer formation is lacking for the human mucins, but both von Willebrand factor (14) and porcine submaxillary mucin (15) form dimers, and norrin forms disulfide-linked oligomers (16). Half-cystine 13246 in submaxillary mucin is in an identical position to half-cystine 2010 in mature von Willebrand factor and to half-cystine 95 in norrin. Moreover, a mutant form of von Willebrand factor containing an arginine instead of half-cystine 2010 fails to form dimers (22). In addition, a mutant form of norrin containing alanine instead of half-cystine 95 has impaired oligomer formation (16). Taken together, these observations argue strongly for the involvement of half-cystine 13246 of porcine submaxillary mucin in formation of disulfide-bonded dimers. Similar support for the involvement of porcine mucin half-cystine 13244 in interchain disulfide bonds of dimers is not available, but noteworthy is the conservation of the sequence CXCC in the seven proteins shown in Fig. 7, with the three half-cystines corresponding to half-cystines 13244, 13246, and 13247, respectively, in porcine mucin. It is possible that substitutions of half-cystine 13244 result in changes in conformation that affect the interchain disulfide bonds formed by half-cystine 13246. This view is consistent with the observation that mutant proteins C13244S and C13244A migrate slightly faster on SDS-gel electrophore- sis under nonreducing conditions than mutant proteins C13246S and C13246A (Figs. 2C, and 4, A and B).
Support for the involvement of half-cystine 13223 in formation of disulfide-bonded dimers in mucin also comes from comparison of the sequences shown in Fig. 7. Half-cystine 13223 is the first half-cystine in the sequence VEMARCVGECKK, which is one of the most highly conserved sequences in these proteins. Substitution of the half-cystine in norrin that corresponds to half-cystine 13223 by either tryptophan (23) or tyrosine (24) produces Norrie disease in humans. Moreover, when valine 60 of norrin, which corresponds to valine 13218 in submaxillary mucin, is substituted by glutamic acid, the secretion of the mutant norrin is impaired (16), and humans with this mutation have a severe form of Norrie disease (13). Valine 13218, five residues removed from half-cystine 13223, may well be in a sequence that is directly involved in dimer formation through disulfide bonds.
The recent report (16) describing the oligomerization of human norrin, in conjunction with the studies reported here, strongly suggests that this type of domain with 11 half-cystines functions to form interchain disulfide bonds. This view is supported by the recent report (10) that human MUC6 contains a carboxyl-terminal domain with only 91 residues, including 11 half-cystines, that has significant sequence identity with the carboxyl-terminal, disulfide-rich domain of porcine submaxillary mucin (Fig. 7).