The MUC1 SEA module is a self-cleaving domain.

MUC1, a glycoprotein overexpressed by a variety of human adenocarcinomas, is a type I transmembrane protein (MUC1/TM) that soon after its synthesis undergoes proteolytic cleavage in its extracellular domain. This cleavage generates two subunits, alpha and beta, that specifically recognize each other and bind together in a strong noncovalent interaction. Proteolysis occurs within the SEA module, a 120-amino acid domain that is highly conserved in a number of heavily glycosylated mucin-like proteins. Post-translational cleavage of the SEA module occurs at a site similar to that in MUC1 in the glycoproteins IgHepta and MUC3. However, as in the case of other proteins containing the cleaved SEA module, the mechanism of MUC1 proteolysis has not been elucidated. Alternative splicing generates two transmembrane MUC1 isoforms, designated MUC1/Y and MUC1/X. We demonstrated here that MUC1/X, whose extracellular domain is comprised solely of the SEA module in addition to 30 MUC1 N-terminal amino acids, undergoes proteolytic cleavage at the same site as the MUC1/TM protein. In contrast, the MUC1/Y isoform, composed of an N-terminally truncated SEA module, is not cleaved. Cysteine or threonine mutations of the MUC1/X serine residue (Ser-63) immediately C-terminal to the cleavage site generated cleaved proteins, whereas mutation of the Ser-63 residue of MUC1/X to any other of 17 amino acids did not result in cleavage. In vitro incubation of highly purified precursor MUC1/X protein resulted in self-cleavage. Furthermore, addition of hydroxylamine, a strong nucleophile, markedly enhanced cleavage. Both these features are signature characteristics of self-cleaving proteins, and we concluded that MUC1 undergoes autoproteolysis mediated by an N --> O-acyl rearrangement at the cleavage site followed by hydrolytic resolution of the unstable ester and concomitant cleavage. It is likely that all cleaved SEA module-containing proteins follow a similar route.

The MUC1 gene is highly expressed in a number of human epithelial malignancies, including breast, prostate, and colon carcinomas, as well as on the malignant plasma cells of multiple myeloma (1)(2)(3)(4)(5)(6)(7)(8)(9). As a well characterized tumor-associated protein, it has generated considerable interest as a tumor marker for disease prognosis (10 -14) as well as a target for tumor cell killing (15)(16)(17)(18). Although alternative splicing can generate multiple MUC1 protein forms (19 -23), the most intensively studied MUC1 protein is a type I transmembrane protein comprised of a heavily glycosylated extracellular domain containing a tandem-repeat array, a transmem-brane domain, and a cytoplasmic domain (Fig. 1, MUC1/TM) (24 -26). MUC1/TM is proteolytically cleaved soon after its synthesis, generating two subunits, ␣ and ␤, that specifically recognize each other and bind together by a strong noncovalent interaction (27).
Cleavage occurs within the SEA module (28 -30), a highly conserved protein module so-called from its initial identification in a sperm protein, in enterokinase, and in agrin (31), that is found in a number of heavily glycosylated mucin-like proteins that are typically membranetethered. The MUC1 protein is cleaved between glycine and serine residues present in the sequence GSVVV (Figs. 1 and 2) situated within the SEA module (28 -30). We proposed previously that this module likely functions as a site for proteolytic cleavage in all SEA module-containing proteins (30). Subsequently, both MUC3 (32) and IgHepta (33) were found to be cleaved at a GS dipeptide sequence present within their SEA modules, supporting the thesis that the evolutionarily conserved SEA module indeed functions as a proteolytic site. The SEA modules comprise ϳ120 amino acids, and the high degree of conservation in its sequence and structural features suggests that an ordered structure surrounding the cleavage site may be required for proteolysis (34). Furthermore, it appears likely that in all cleaved SEA module-containing proteins the proteolytic mechanism is similar (29,30,32,33).
The 72 amino acids composing the MUC1 cytoplasmic domain can be extensively phosphorylated on its serine and tyrosine residues (35)(36)(37)(38)(39), and with appropriate phosphorylation this domain can interact with second messenger proteins (36, 40 -43) generating a signal transduction cascade that alters gene expression and, thereby, cellular behavior. For MUC1, the incipient event probably involves ligand binding to the MUC1 extracellular domain, leading to allosteric changes on the MUC1 cytoplasmic domain resulting in phosphorylation. The reversible binding of the MUC1 ␣and ␤-subunits may constitute this ligand-receptor interaction (30).
Understanding the proteolytic cleavage mechanism that generates the two interacting MUC1 subunits is an important prerequisite to understanding the role of MUC1 in tumor biology and signal transduction. To learn more about this mechanism, we sought a system that would be amenable to a detailed molecular analysis of the SEA modulemediated cleavage. In the present study we show that the MUC1 alternative splice form, MUC1/X (19), whose extracellular domain comprises essentially only the SEA module, undergoes proteolytic cleavage and that this cleavage is mediated by autoproteolysis.

EXPERIMENTAL PROCEDURES
Materials and Antibodies-Unless otherwise specified, chemicals and reagents were obtained from Sigma. The anti-FLAG antibodies were affinity-purified rabbit polyclonal antibodies. HRP 2 -conjugated * This work was supported in part by the Israel Science Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 To whom correspondence should be addressed: Dept. of Cell Research and Immunology, Tel Aviv University, Ramat Aviv 69978, Israel. Tel.: 972-3-6407425; Fax: 972-3-6422046; E-mail: danielhw@post.tau.ac.il.
goat anti-human Fc (HRP-G␣-hFc) and unlabeled goat anti-human Fc were from Zymed Laboratories Inc. and The Jackson Laboratories, respectively. The anti-MUC1 cytoplasmic domain antibodies were affinity-purified rabbit polyclonal antibodies generated against the human MUC1 cytoplasmic domain. The BOS10B3 antibody is a mouse monoclonal antibody, previously generated by us (44), directed against an epitope found in the sequence STPGGEKETSATQRSS located within the 30 MUC1 N-terminal amino acids following signal peptide cleavage (see Figs. 1 and 2). Cell Culture-Cells were grown at 37°C and 5% CO 2 , in culture media supplemented with 10% heat-inactivated fetal calf serum, 2 mM L-glutamine, 100 IU/ml penicillin, and 25 g/ml streptomycin. DA3 mouse mammary tumor cells and human embryonic kidney cells 293 (HEK) were grown in Dulbecco's modified Eagle's medium. Expression constructs were transfected into the cells using the calcium phosphate procedure.
Cloning of MUC1/X and MUC1/Y in Maltose-binding Protein (MBP) Expression Vector-The MBP-MUC1/X and MBP-MUC1/Y fusion proteins (MBP-Xex-FLAG and MBP-Yex-FLAG) are depicted in Fig. 3, inset F, and described in the legend to the figure. For the construction of the MBP bacterial expression vectors, a cDNA fragment encoding the extracellular domain either of MUC1/X or of MUC1/Y was amplified using pCL-MUC1/X or pCL-MUC1/Y as template DNAs. Reverse and forward DNAs were used that introduced a FLAG epitope at the C terminus of the MBP fusion protein as well as appropriate restrictions sites for ligation to the bacterial MBP expression vector pMAL-c2X (New England Biolabs). DNA sequencing confirmed the identities of all expression constructs. The pMAL-c2X-MUC1 vectors were electroporated into MC1061-competent bacteria.
Purification of Recombinant Bacterial MBP Fusion Proteins-Recombinant bacteria, grown in LB medium and induced with IPTG, were lysed by sonication and the recombinant bacterial MBP-MUC1-FLAG fusion proteins (see Fig. 3, inset F) purified from the supernatant using amylose-agarose beads (Amersham Biosciences) as described by the manufacturer.
Generation of Stable DA3 Mouse Mammary Tumor Cell Transfectants Expressing MUC1/X and MUC1/Y-DA3 cells were cotransfected with the eukaryotic expression plasmid pCL-MUC1/X (or pCL-MUC1/Y) and pSV2neo (coding for neomycin resistance). Stable transfectants were selected with neomycin. Transfectants expressing the MUC1 proteins were identified by immunoblot analysis of cell lysates using affinity-purified polyclonal antibodies directed against the MUC1 cytoplasmic domain.
Generation of MUC1/X and MUC1/Y Eukaryotic Expression Constructs and Fusion Proteins-The MUC1 fusion proteins employed in this study, designated FLAG-Yex-hFc and FLAG-Xex-hFc, are depicted and described in Fig. 1. Standard molecular biology methods were used to generate all constructs. Cloning was conducted with the eukaryotic expression vector pCMV3 (Sigma) via selected restriction sites. This vector codes for the prepro-trypsin signal peptide followed by sequences coding for the FLAG (DYKDDDDK) epitope. DNA coding for the human Fc fragment (hFc) was inserted 3Ј to the DNA coding for the FLAG epitope. This allowed cDNAs coding for MUC1 proteins to be inserted, in-frame, between DNA coding for the FLAG epitope (on the 5Ј side) and DNA coding for hFc (on the 3Ј side) of the pCMV3 vector. Briefly, reverse and forward primer oligonucleotides were used with pCL-MUC1/X or pCL-MUC1/Y as templates to PCR-amplify cDNAs coding for the extracellular domains of MUC1/X or MUC1/Y. This also introduced appropriate restrictions sites for ligation of the MUC1 cDNAs into pCMV3 DNA. cDNA fragments encoding the MUC1/X or MUC1/Y extracellular domains were subcloned in-frame into the pCMV3 (5ЈFLAG-hFc3Ј) vector. To facilitate generation of the mutant FLAG-Xex-hFc proteins containing deletions within the extracellular domain of MUC1/X, an EcoRI restriction site was introduced at nucleotides 210 -216 within the MUC1 sequence (see Fig. 2) by changing the sequence "gaatgc" to the EcoRI site "gaattc." To generate deletion mutants, appropriate forward and reverse primers containing the desired point mutation or deletion were used to PCR-amplify MUC1/X cDNA fragments that were then cloned into the pCMV3 (5ЈFLAG-hFc3Ј) vector. MUC1/X mutants harboring point mutations at the cleavage site were generated by PCR using reverse primers comprising the desired mutation at the cleavage site and extending to the KpnI site at nucleotides 1165-1170 together with an appropriate forward primer comprising the introduced EcoRI site at nucleotides 210 -216. DNA sequencing confirmed the identities of all expression constructs.
Generation of DA3 Mouse Mammary Tumor Cell and HEK293 Transfectants Expressing FLAG-MUC1/Xex-hFc and FLAG-MUC1/ Yex-hFc Fusion Proteins and Purification of hFc-tagged Fusion Proteins-DA3 (mouse mammary tumor) or HEK293 (human kidney) cells were transiently transfected using calcium phosphate with the eukaryotic pCMV3 expression vectors (6 g of DNA/25-cm 2 flask) coding for the FLAG-Xex-hFc, FLAG-Yex-hFc, or mutant MUC1/X proteins. For obtaining stable transfectants, neomycin was added to the culture medium, and neomycin-resistant clones were isolated. Conditioned culture media containing the secreted MUC1 fusion proteins were collected 2-3 days following transfection and spun at 15,000 rpm for 20 min. The supernatant was filtered through a 0.45-m filter and stored at Ϫ75°C. The C-terminally hFc-tagged FLAG-Xex(Yex)-hFc proteins were purified by protein A-Sepharose 4 Fast Flow (Amersham Biosciences) column purification.
SDS-PAGE-SDS-PAGE for protein separation was performed as described previously (20).
Transfer of Proteins to Nitrocellulose Membranes, Immunoblotting, and Other Protein Analyses-Proteins separated on SDS-PAGE were electrotransferred at 0.5 A for 2 h onto nitrocellulose filter paper in transfer buffer. Blots were blocked in 5% skimmed milk followed by incubation with primary antibody (polyclonal rabbit anti-FLAG or BOS10B3). Bound primary antibody was detected with a secondary anti-rabbit or anti-mouse antibody conjugated to horseradish peroxidase followed by enhanced chemiluminescence. The hFc-tagged proteins were directly analyzed with HRP-conjugated goat anti-hFc antibodies followed by enhanced chemiluminescence. For glycosidase treatment, protein A-purified proteins were incubated with 1 unit of PNGase F (New England Biolabs) according to the manufacturer's instructions.
Hydroxylamine Treatment of Mutant FLAG-Xex-hFc Proteins-Protein A-purified mutant FLAG-Xex-hFc proteins were incubated at 37°C in cleavage buffer with hydroxylamine (150 mM), and following various incubation times samples were subjected to SDS-PAGE followed by immunoblot analysis.
ELISA for Determining MUC1 Cleavage (see Fig. 6)-ELISAImmuno-Assay plates (Costar) were coated with polyclonal goat anti-human Fc antibodies (G-␣-hFc, 4 g/ml) followed by washing with PBS/Tween 20 (0.05%) and blocking with PBS/Tween ϩ 5% skimmed milk (Blotto). Spent culture media containing FLAG-Xex(or Yex)-hFc proteins (mutant and wild type, see Figs. 1, 5, 6, 8, and 9 for descriptions of these proteins) was then applied to the wells to allow for binding of the FLAG-Xex(Yex)-hFc proteins. Following incubation, samples were removed, and the wells washed with PBS/Tween. Wells were then incubated either for 30 min with PBS alone or with PBS ϩ 0.1% SDS followed by PBS/Tween washing. As described under "Results," SDS treatment does not interfere with the binding of the well coated G-␣-hFc to the hFc portion of the MUC1 fusion proteins, but it does result in the dissociation of the MUC1 ␣and ␤-subunit (27). Protein remaining bound to the well was then assessed either with HRP-G-␣-hFc or with mouse monoclonal antibody BOS10B3 followed by HRP-conjugated antimouse antibody.

RESULTS
Proteolytic Cleavage of MUC1 Isoform MUC1/X and Not of MUC1/Y When Expressed as Fusion Proteins in Bacteria-We and others have demonstrated the existence of the alternate MUC1 isoforms, designated MUC1/X and MUC1/Y (19, 22, 23, 44 -46), that are devoid of the tandem repeat array and its flanking sequences (Figs. 1 and 2). Downstream to the splice acceptor sites of MUC1/X and MUC1/Y, both of these proteins retain the same reading frame as the tandem repeat arraycontaining MUC1/TM protein ( Figs. 1 and 2). In addition, the site of proteolytic cleavage between the amino acids glycine and serine within the sequence GSVVV (Figs. 1 and 2) is retained in both MUC1/X and MUC1/Y. The MUC1/Y protein was found not to be cleaved (19,44,45), suggesting that additional upstream sequences, absent in this isoform, may be required for proteolytic cleavage. Downstream to the MUC1/X splice acceptor site, the MUC1/X protein contains the extracellular sequences spanning to the transmembrane domain, which comprise in their entirety the domain described previously, the SEA module ( Figs. 1 and 2). We therefore examined the ability to be cleaved of the MUC1/X protein which, in contrast to MUC1/Y, contains an additional 18 amino acids. A bacterially synthesized protein comprising a maltose-binding protein fused to the MUC1/X extracellular domain at the N terminus and a synthetic FLAG epitope at the C terminus (MBP-Xex-F) migrated markedly faster than the corresponding MBP-Yex-F protein (Fig. 3A, Coomassie stain, compare lanes 2 and 1, and see inset, Although equal amounts of recombinant protein were analyzed, MBP-Yex-F showed a much more intense anti-FLAG signal than the weak signal obtained with MBP-Xex-F (Fig. 3B, lanes 1 and 2).
These results suggested that MBP-Xex-FLAG is proteolytically cleaved, and therefore, a small C-terminal FLAG-tagged cleavage product should be present only in the MBP-Xex-FLAG sample. To obtain better resolution of low molecular mass proteins, the MBP-Xex-F and MBP-Yex-F proteins were analyzed on 12.5% acrylamide gels. Anti-FLAG probing showed a discrete 6 -7-kDa band only in the MBP-Xex-F sample; no signal was seen with MBP-Yex-F (Fig. 3C, lanes 2 and 1,  respectively). This band corresponded to the expected molecular mass of ␤ex-FLAG (Fig. 3, inset F), and tryptic digestion followed by mass spectrometric analysis confirmed its identity (data not shown). These results indicated that in contrast to the noncleavage of MBP-Yex-FLAG, the extracellular domain of MUC1/X protein (Xex) when expressed in bacteria as an MBP-Xex-FLAG fusion protein undergoes proteolytic cleavage.
Additional studies further supported this conclusion. Reprobing the same blot as in Fig. 3B with BOS10B3 antibody, which recognizes an epitope within the N-terminal amino acids of MUC1 (44) (Fig. 3, inset F, FIGURE 1. MUC1 protein isoforms and MUC1 fusion proteins. The tandem repeat array-containing transmembrane MUC1 protein (MUC1/TM) is depicted at left. Cleavage of MUC1/TM generates a large extracellular ␣-subunit that is noncovalently bound to a membrane-tethered ␤-subunit. From the N terminus (N) to the C terminus (C), this protein is comprised of a cleaved-off signal peptide (SP, purple region), followed by N-terminal 30 amino acids (yellow) leading into sequences (green region) that N-terminally flank the tandem repeat array (striped blue region), followed by sequences (green) that C-terminally flank the tandem repeat array leading into regions common to both the MUC1/X and the MUC1/Y proteins (orange and red regions), together representing 62 amino acids. The ␤-subunit extracellular domain is comprised of 58 amino acids (red/white checkered region) immediately N-terminal to transmembrane (blue) and cytoplasmic (pink) domains. The orange, red, and red/white checkered sections together consist of 120 amino acids that form the SEA module. Alternative splicing utilizes a splice donor indicated by S.D.3 and one of two alternative splice acceptors (S.A. Y3 and S.A. X3) that form the MUC1/Y and MUC1/X proteins (2nd and 3rd from left in figure), respectively. The N30 region (yellow region) of these proteins contains the epitope for BOS10B3. As shown experimentally (see text), MUC1/Y does not undergo cleavage, whereas MUC1/X is completely cleaved into X␣-and X␤-subunits. The fusion proteins FLAG-Yex-hFc and FLAG-Xex-hFc (at right of the figure) comprise (from their N termini) the FLAG epitope (dark green, 9 amino acids) followed by N-terminal 30 amino acids (yellow). FLAG-Yex-hFc continues into amino acids 19 -120 (red region), whereas FLAG-Xex-hFc continues into amino acids 1-120 (orange followed by red regions). (For numbering, see italic numbers in Fig. 2 that appear directly below the amino acid sequence; 1 designates the first amino acid following the MUC1/X splice acceptor site). Both proteins are then C-terminally fused to hFc (green vertical line). As shown experimentally (see text), FLAG-Xex-hFc is cleaved into X␣-and X␤ex-subunits, whereas FLAG-Yex-hFc does not cleave. and Fig. 2), showed a strong band derived from MBP-Xex-F. This band was of equal intensity to that of MBP-Yex-F, yet it migrated faster (Fig.  3D, compare lanes 2 and 1), and its molecular mass corresponded with that expected of an MBP-X␣ protein (Fig. 3, inset F). Proteolytic cleavage of MUC1 is known to occur just upstream of serine 63 (29), and its mutation to alanine has been shown to abrogate MUC1 cleavage (28). In accord with this, a point mutation of S63A within MBP-Xex-FLAG to alanine (see Fig Taken together, these results indicate that the bacterially expressed extracellular domain of MUC1/X is proteolytically cleaved, whereas the corresponding MUC1/Y extracellular domain is not cleaved. As the MUC1/X extracellular domain comprises only 30 MUC1 N-terminal amino acids (N30) juxtaposed to the SEA module, it is considerably smaller than the tandem repeat array-containing MUC1/TM protein (Figs. 1 and 2) and affords a much simpler system in which to analyze the proteolytic cleavage mechanism.
Proteolytic Cleavage Occurs on MUC1 Isoform MUC1/X and Not on MUC1/Y When Expressed as Native Proteins in Eukaryotic Cells-To verify that the proteolytic cleavage also pertains to the complete MUC1/X protein and is not an artifact of the recombinant MBP-Xex-FLAG bacterial protein, stable transfectants expressing either MUC1/Y or MUC1/X full-length proteins (see Fig. 1, isoforms shown second and third from left in the figure, respectively) were generated in DA3 mouse mammary tumor cells, and the ability to be cleaved of the two isoforms was assessed. Cell lysates from these stable transfectants as well as from control parental DA3 cells (DA3/PAR) and from neomycin-resistant transfectants not expressing human MUC1 protein (DA3/neo) were subjected to immunoblot analysis with anti-MUC1 cytoplasmic domain antibodies (see "Materials and Antibodies" under "Experimental Procedures"). In cells expressing cleaved MUC1, these antibodies have been shown to detect the cytoplasmic domain cleavage products that migrate as a doublet of approximate molecular mass of 24 -26 kDa (39,47,48). In accord with this, two independent DA3-MUC1/X transfectants displayed characteristic cytoplasmic domain-containing cleavage products (Fig. 4, lanes 3 and 4), indicating that MUC1/X is cleaved in these cells. In contrast, the MUC1/Y-expressing transfectant demonstrated strong diffuse immunoreactivity in the region of 46 -50 kDa (as documented previously) (39,44,45) and lesser intense bands with a molecular mass of about 30 kDa. The 46 -50-kDa MUC1/Y immunoreactive bands are glycosylated, whereas the 30-kDa MUC1/Y bands have been shown to represent incompletely glycosylated versions (19,39,45). Notably, no MUC1/Y immunoreactivity was detected within the 24 -26-kDa region expected for cytoplasmic domain-containing cleavage products (Fig. 4, compare lane 2 with lanes 3 and 4). When expressed in eukaryotic cells as full-length proteins, MUC1/X is thus cleaved, whereas MUC1/Y is not.
Proteolytic Cleavage Occurs on MUC1 Isoform MUC1/X and Not on MUC1/Y When Their Extracellular Domains Are Expressed as Fusion Proteins in Eukaryotic Cells-To establish a convenient system permitting systematic analysis of MUC1 proteolytic cleavage and its mechanism of action, we investigated the ability to be cleaved of secreted fusion proteins comprising the MUC1/Y and MUC1/X extracellular domains. Eukaryotic expression vectors were constructed that encoded a prepro-trypsin signal peptide followed by a FLAG epitope, the MUC1/Y (or MUC1/X) extracellular domains, and C-terminating with the Fc portion of human immunoglobulin (49,50). Transfection of these plasmids into cells led to secretion of fusion proteins comprising MUC1/X and MUC1/Y extracellular domains (Fig. 5, inset), which could then be simply purified from conditioned medium by passage through protein-A-agarose. These proteins were tagged at their N and C termini with epitopes permitting their analysis with anti-FLAG and anti-hFc antibodies, respectively. Probing FLAG-Yex-hFc with anti-hFc (Fig. 5, inset) demonstrated a strong band at about 55 kDa (Fig. 5A, left panel, lane 2); no other bands were detected. Subsequent probing of the same blot (without stripping) with anti-FLAG led to a marked intensification of the 55-kDa band and did not reveal any additional immunoreactive species (Fig. 5A, right panel, lane 2Ј). In contrast, anti-hFc immunoblotting of FLAG-Xex-hFc demonstrated a doublet (indicated by 7 to the left of Fig. 5A) migrating with a mass of about 37 kDa (Fig.  5, A, left panel, lane 1, and B, lane 1); no immunoreactivity was seen in the 55-kDa region (compare Fig. 5A, left panel, lanes 1 and 2). Treatment of FLAG-Xex-hFc with peptide:N-glycosidase F (PNGase F) that digests N-glycosylated proteins between Asn and GlcNAc resulted in disappearance of the 37-kDa doublet and appearance of a discrete band at 34 kDa (Fig. 5B, lane 2), suggesting that the 37-kDa doublet represents proteins that are N-glycosylated to differing extents. The level of inten-sity of the faster migrating component of the 37-kDa doublet was subject to batch variation, and in some cases was not seen at all (data not shown). After removal of N-linked sugars, the mass of the deglycosylated protein (34 kDa) correlated well with the size expected of a nonmodified, cleaved ␤ex-hFc protein. Probing of the blot with antibodies (BOS10D2) recognizing epitopes within ␤ex sequences (44) also revealed the 34-kDa band (Fig. 5C). Finally, N-terminal sequencing of the N-glycosylated 37-kDa band demonstrated that it initiated with the sequence SVVVQLTLAFREGT, matching perfectly with cleavage N-terminal to the serine residue within the sequence GSVVV. Taken together, these results demonstrate that the 37-kDa doublet and deglycosylated 34-kDa proteins are the ␤ex-hFc cleavage products resulting from proteolytic cleavage of FLAG-Xex-hFc.
Reprobing the same blot with anti-FLAG without stripping (Fig. 5A, right panel), or following stripping (Fig. 5B, lane 1Ј), revealed marked differences between the FLAG-Yex-hFc and FLAG-Xex-hFc proteins (Fig. 5A, right panel, lanes 2Ј and 1Ј, respectively). In the FLAG-Xex-hFc sample additional anti-FLAG immunoreactive bands appeared, which migrated with molecular masses of 25 (diffuse band) and 18 kDa (bracketed immunoreactive species indicated by *, Fig. 5, A, right panel, lane 1Ј, and B, lane 1Ј) as well as a very minor band which migrated with a mass of about 20 kDa. These anti-FLAG immunoreactive species were completely absent in FLAG-Yex-hFc (Fig. 5A, compare lanes 1Ј and 2Ј). As these bands were nonreactive with anti-hFc (which recognizes the  Fig. 2 that appear directly below the amino acid sequence; 1 designates the first amino acid following MUC1/X splice acceptor site.) This is followed C-terminally by 9 amino acids comprising the FLAG epitope. Cleavage just upstream to serine 63 renders X␣ and ␤ex. The MUC1/Y fusion protein MBP-Yex-F (abbreviated here to Yex) is identical to MBP-Xex-F, except for deletion of MUC1/X amino acids 1-18 (see Figs. 1 and 2).
C-terminal hFc) and only appeared with anti-FLAG (which recognizes the N-terminal FLAG epitope), we conclude that they represent proteolytically cleaved N-terminal FLAG-X␣ proteins. Their molecular masses were, however, considerably larger than those expected of unmodified FLAG-X␣ protein. N-Glycosidase treatment of FLAG-Xex-hFc protein reduced the molecular mass of the 18-kDa FLAG-X␣ protein to about 11 kDa, which correlates well with that expected of an unmodified FLAG-X␣ protein; the diffuse 25-kDa species converted to diffuse 20 -24-kDa bands (Fig. 5B, lane 2Ј). These results indicate that both the 25-and 18-kDa proteins (Fig. 5, immunoreactive species indicated by *) represent N-glycosylated FLAG-X␣ proteins and that the 25-kDa band, besides N-linked sugars, likely bears additional posttranslational modifications that are the subject of ongoing investigation.
Establishing an ELISA Monitoring Proteolytic Cleavage of MUC1 Proteins-We established a simple assay for monitoring MUC1 cleavage. This comprised capture of C-terminally tagged MUC1-hFc proteins (secreted into the culture medium of cells transfected with plasmids encoding MUC1-hFc-tagged proteins) in ELISA plate wells previously coated with polyclonal anti-hFc antibodies (Fig. 6). Binding of MUC1 (C-terminally tagged)-hFc proteins could then be assessed by probing with either HRP-conjugated polyclonal anti-hFc antibodies or with mouse monoclonal antibody BOS10B3, which recognizes a MUC1 epitope present in the N-terminal 30 amino acids of the MUC1 protein (44) (see Figs. 1 and 3F for the site of BOS10B3 immunoreactivity and Fig. 6 for description of the ELISA).
It has been demonstrated previously that the complex composed of MUC1 ␣-subunit bound to the ␤-subunit can be dissociated by 0.1% SDS (27). To see whether we could take advantage of this finding, wells containing FLAG-Xex-hFc (or FLAG-Yex-hFc) bound to anti-hFc antibodies were washed with PBS containing varying concentrations of SDS. The remaining protein that contained C-terminal hFc was then assessed with HRP-conjugated polyclonal anti-hFc antibodies. Washing with SDS concentrations up to 0.1% gave an identical signal to that seen in its absence (Fig. 7, dotted line joined by black squares), demonstrating that the interaction between polyclonal anti-hFc and the C-terminal hFc portion of MUC1 fusion proteins is completely resistant to 0.1% SDS. In contrast, the amount of FLAG-Xex-hFc protein remaining bound after SDS washing as assessed with antibody BOS10B3 was drastically decreased at 0.04% SDS with almost no signal seen at 0.1% SDS (Fig. 7, black line joined by black diamonds). Thus 0.1% SDS completely dissociates the N-terminal FLAG-X␣ portion from the well bound C-terminal ␤ex-hFc part (see Fig. 7, graph and histogram inset). In accord with this, when noncleaved FLAG-Yex-hFc protein was assayed, washing with 0.1% SDS had no effect on BOS10B3 signal (Fig. 7, histogram inset). The ELISA therefore represents a convenient means of monitoring cleavage of hFc-tagged MUC1 proteins (schematically presented in Fig. 6).
SEA Module Ser-63 Mutations to 17 Amino Acids Abrogates MUC1 Cleavage, whereas Mutation to Threonine or Cysteine Retains Cleavage-As described above, the MUC1/X extracellular domain undergoes proteolytic cleavage in both bacteria and eukaryotic cells, and cleavage occurs at the same site as in the tandem repeat array-containing MUC1 protein. These results indicate that the minimum sequences required for proteolytic cleavage are contained within the 120 amino acids of the MUC1/X extracellular domain that form the SEA module and that the cleavage mechanism is likely identical for both the tandem repeat array-containing MUC1 and MUC1/X proteins. Furthermore, the MUC1 protein is also cleaved when expressed in reticulocyte lysates (27). If proteolytic cleavage is mediated by a separate protease entity acting on the MUC1 protein in trans, rather than an intramolecular cis-mediated process, one would have to invoke a protease of exquisite and identical specificity (G/ 63 SVVV) in bacteria, in reticulocyte lysates, and in eukaryotic cells.
To learn more about the MUC1 proteolytic machinery and molecular identity of the protease, we proceeded to investigate cleavage site specificity. A panel of FLAG-MUC1/Xex-hFc fusion proteins was generated where the Ser-63 cleavage site (proteolytic site ϩ 1, P ϩ 1) was individually mutated to all other 19 amino acids. Significantly, Ser-63 mutation to 17 amino acids S63(A/D/E/F/G/H/I/K/L/M/N/P/Q/R/V/W/Y, Fig.  8A, left panel) totally abrogated cleavage. As in the wild type protein, MUC1/X proteins harboring S63C or S63T mutations were entirely cleaved (Fig. 8A).
The cleavage characteristics of Ser-63 point mutants were highly informative. The common denominator of the three cleavage-supporting amino acids is that they all contain a terminal hydroxyl (serine and threonine) or thiol (cysteine) residue. Searching for proteases harboring specificity for these three amino acids revealed that hedgehog proteins (51), inteins (52,53), and other Ntn (N-terminal nucleophile) hydrolases (54 -65) are all cleaved at an internal peptide bond immediately preceding a cysteine, serine, or threonine residue. Furthermore, these proteins undergo an autoproteolytic reaction in which the hydroxyl or thiol groups of serine, threonine, or cysteine act as nucleophiles generating an ester bond at the point of scission via an N 3 O-acyl shift that can then be resolved by hydrolysis (66,67).
Because of the difficulty of invoking a separate protease acting in trans on the MUC1 protein (see above) and because of the absolute requirement of serine, threonine, or cysteine at P ϩ 1, we entertained the possibility of MUC1 autoproteolysis.  3 and 4). In contrast, specific immunoreactive bands appeared in DA3-MUC1/X as a doublet migrating at about 25 kDa, as expected of cytoplasmic domain-containing MUC1 cleavage products. These were not seen in MUC1/Y transfectants.
To see whether MUC1/X cleaves via an autoproteolytic mechanism, we sought to identify a partially cleaving mutant which would allow purification to homogeneity of uncleaved precursor. Generation of proteolytically cleaved products following in vitro incubation of pure uncleaved precursor without any extraneous additions would constitute strong evidence for a self-cleaving mechanism. Enhancing the rate of processing of a mutant protein by a robust nucleophile such as hydroxylamine would provide additional proof for an autoproteolytic mechanism (67).
The search for a partially cleaving mutant was initiated by assessing the effect of N-terminal deletions within the SEA module on proteolytic cleavage. Results indicated that although a mutant with deletions of amino acids 1-7 retained full cleavage activity (Fig. 8B, ⌬1-7, right panel), the mutant with deletion of amino acids 1-11 was completely uncleaved. We therefore used the ⌬1-7 construct as a starting point for generating a partially cleaving mutant. By comparing interspecies MUC1 SEA module sequences, we observed that serine and threonine residues located within the first 30 amino acids of these domains are highly conserved. To obtain a partially cleaving mutant we initiated systematic mutations of these residues. We found that mutant ⌬1-7, S11A/S15A (FLAG-Xex-hFc harboring this deletion and mutations) was cleaved to the same extent as wild type protein (data not shown). In contrast, mutant protein Fi17 (⌬1-7, S11A/S15A/S21A/S22A) was 70% cleaved, and Fi18 (⌬1-7, S11A/S15A,S27A, and T28A) was only 30% cleaved (Fig. 8B). The partial cleavage of these mutants was confirmed by anti-hFc immunoblotting, which showed good concordance with the ELISA analysis (Fig. 9, A and B).
As a result of these findings, Fi18 was used as a source of partially cleaved precursor. When incubated in vitro, highly purified Fi18 protein demonstrated a time-dependent decrease in the amount of 55-kDa precursor protein and corresponding increase in ␤ex-hFc levels (Fig. 9C,  lanes 1 and 2) indicating that the purified Fi18 precursor protein undergoes spontaneous cleavage, albeit at a slow rate. A kinetics analysis (Fig.  9D, lanes 2-4) demonstrated spontaneous cleavage with a half-activation time () of slightly less than 24 h. In contrast, purified FLAG-Yex-hFc incubated for 24 h (Fig. 9D, lane 1) did not produce any proteolytic cleavage products, suggesting that in vitro proteolysis of the partially cleaved Fi18 mutant is an inherent property of the MUC1/X protein and is not because of any contaminating proteases acting in trans.
Hydroxylamine (NH 2 OH) is a strong nucleophile that is highly reactive against (thio)esters and can significantly enhance the rate of autoproteolysis by facilitating hydrolysis of the (thio)ester intermediate, the  1 and 1Ј) or MUC/Y (lanes 2 and 2Ј) that were tagged at their N termini with a FLAG epitope and fused at their C termini to the Fc portion of human IgG (constructs of FLAG-Xex-hFc and FLAG-Yex-hFc fusion proteins are illustrated in the inset below the blots). The MUC1 fusion proteins were purified from conditioned media (culture medium) using protein A-agarose beads. The blots were probed first with HRP-conjugated anti-human Fc. Following ECL development (A, left panel), the blot (without stripping) was sequentially probed with anti-FLAG antibodies (A, right panel). Anti-hFc as well as anti-FLAG probings (lanes 2 and 2Ј) detect FLAG-Yex-hFc protein at identical locations (about 50 -55 kDa, indicated by filled circle-ended arrow). In contrast, anti-hFc probing of FLAG-Xex-hFc shows a doublet (bracketed, indicated by 7) migrating with a relative mass of about 37 kDa (lanes 1) representing the ␤ex-hFc cleaved proteins. Following anti-FLAG probing (lanes 1Ј), a diffuse 25-kDa region and major 18-kDa band are seen (bracketed, indicated by *), representing FLAG-X␣ cleavage products. Molecular masses (kDa) are indicated to the right of each figure. B, Western blot analysis of FLAG-Xex-hFc fusion proteins treated with PNGase F. A similar experiment was performed as described for A, except that after probing with HRP-conjugated anti-human Fc (lanes 1 and 2), the blot was stripped and reprobed with anti-FLAG antibodies (lanes 1Ј and 2Ј). FLAG-Xex-hFc proteins treated with PNGase F to remove N-linked sugars are shown in lanes 2 and 2Ј. C, Western blot analysis of FLAG-Xex-hFc fusion protein not treated (Ϫ) or treated (ϩ) with PNGase F. Blots were probed with HRP-conjugated anti-human Fc (left panel) or with BOS10D2 mouse monoclonal antibodies that recognize an epitope contained within the sequence SDVSVSDVPFPFSAQ (see Fig. 2, amino acids 397-411; amino acid numbering is indicated at the right of the figure (44)). IB, immunoblot.
rate-limiting step in autoproteolysis (67). Hydroxylamine markedly enhanced proteolytic cleavage of the precursor MUC1/X protein (Fig.  9D, lanes 5 and 6), and densitometric analysis showed that after a 6-h incubation, 80% of uncleaved precursor had been converted into the proteolytic cleavage product (Fig. 9D, lane 5 and bar histogram). This hydroxylamine-enhanced cleavage resulted in a of less than 6 h, which was in marked contrast to levels of proteolysis seen in its absence (Fig.  9D, compare lanes 2 and 3 with lanes 4 and 5). By 24 h, hydroxylamineenhanced cleavage resulted in almost complete cleavage of precursor MUC1/X (Fig. 9D, lane 6 and bar graph).

DISCUSSION
Restricted proteolysis is mandatory for a diverse array of biological processes. These include apoptosis (68), tissue remodeling (69), coagulation cascades (70), complement activation (71), release of cell-and matrix-associated growth factors (72), cell-fate determination (73), and ligand-induced receptor activation (74,75). The proteases involved are usually site-specific, and often their own activities are, in turn, regulated by limited proteolysis.
Self-cleaving proteases are gaining increasing recognition as components of key regulators of cellular processes. For example, hedgehog proteins (51,76,77) are signaling molecules essential for embryonic pattern formation that are first synthesized as inactive precursors. These undergo self-cleavage to produce an N-terminal signaling domain. Autoproteolysis also participates in the formation of a group of enzymes collectively referred to as N-terminal nucleophilic (Ntn) hydrolases (62,78,79). The Ntn hydrolases are formed by self-cleavage immediately upstream to a nucleophilic residue, which then serves as a single enzymatic N-terminal nucleophile active site. In prokaryotes, intein-containing proteins are modified by self-catalyzed cleavage and protein-splicing reactions (53,80).
Two outstanding features are characteristic of these self-cleaving enzymes. First, no extraneous proteins are required for cleavage; second, cleavage invariably takes place just upstream to either a serine, threonine, or cysteine residue. The self-cleavage mechanism comprises an N 3 O-or N 3 S-acyl rearrangement that moves the carbonyl carbon of the N-terminal sequences to the oxygen (or sulfur) atom of the hydroxyl (or thiol) side chain of the downstream Ser/Thr/Cys generating a linear ester (or thioester) intermediate (81). Hydrolysis of the ester (or thioester) bond generates cleaved N-terminal and C-terminal subunits. In some cases such as for the self-cleaving enzyme glycosyl asparaginase, the two subunits generated by autoproteolysis recognize To assess the presence of fusion protein having hFc at its C terminus, HRP-conjugated polyclonal goat anti-human Fc antibodies were added to the wells followed by substrate (see lefthand corner of each panel). To assess the presence of fusion protein extending from the C-terminal hFc to the MUC1 N terminus, BOS10B3 mouse monoclonal antibodies (which recognize an epitope located in the MUC1 N-terminal 30 amino acids, see Fig. 1) were added to wells followed by HRP-conjugated anti-mouse antibodies and substrate. As schematically illustrated here and experimentally shown in Fig. 7, SDS does not disrupt binding of the hFc portion to well coated polyclonal goat anti-human Fc antibodies (AЉ and BЉ). In contrast, interaction of MUC1/X ␣ and ␤-subunits is disrupted by 0.1% SDS (schematically depicted here in BЉ; see Fig. 7). and bind to each other (54). A similar mechanism of self-cleavage has been reported recently for proteins containing a site designated GPS (G-protein-coupled receptor proteolytic site) that is located in a G-protein-coupled receptor subfamily comprising a long N-terminal extracellular domain that contains a serine/threonine-rich mucin-like domain (82). The GPS consensus site is not only found in 7-transmembrane G-protein-coupled receptors but also in type I membrane proteins such as polycystin (83), which transverse the cell membrane only once. As with cleavage of MUC1, autoproteolysis of EMR2 at the GPS site likewise generates extracellular ␣and ␤-subunits that interact with each other.
The tandem repeat array-containing MUC1 protein is proteolytically cleaved within its SEA module at the proteolytic cleavage site G2SVVV (29). Here we show that the MUC1 isoform MUC1/X, which lacks the tandem repeat array, is proteolytically cleaved at a site identical to that occurring on the tandem repeat array-containing MUC1/TM protein. This is significant because in its extracellular domain the MUC1/X protein comprises only the 120-amino acid SEA module fused to the MUC1 30 N-terminal amino acids. Another MUC1 isoform designated MUC1/Y, which is identical to MUC1/X except for an 18-amino acid deletion at the SEA module N terminus, does not undergo cleavage, suggesting that an intact SEA module is the minimal requirement for proteolytic cleavage. Cleavage of MUC1/X and absence of cleavage for MUC1/Y was demonstrated as follows: (a) for the complete proteins, (b) for their extracellular domains expressed as fusion proteins in bacteria, and (c) for their extracellular domains expressed as fusion proteins in eukaryotic cells. These results demonstrate that the presence of the restricted cleavage site sequence itself, as in MUC1/Y, is not sufficient to maintain proteolysis. On the contrary, sequences more than 50 amino acids upstream to the G2SVVV MUC1 cleavage site, as in MUC1/X  Plasmid constructs were generated that encode mutant FLAG-Xex-hFc proteins containing deletions from MUC1/X amino acids 1-7 or 1-11 (⌬1-7 and ⌬1-11, respectively; see Fig. 2 for numbering of MUC1/X extracellular domain amino acids) or encode mutant ⌬1-7 that additionally contained serine to alanine mutations as indicated (Fi17 with S11,15,21,223 A, and Fi18 with S11,15,27,T283 A). ⌬1-7 was cleaved to the same extent as nonmutated FLAG-Xex-hFc protein, and no detectable cleavage was seen with ⌬1-11. Mutants Fi17 and Fi18 demonstrated ϳ70 and 30% cleavage, respectively. and in the ⌬1-7 MUC1/X mutant, are indispensable for proteolysis. Further N-terminal deletions as in the ⌬1-11 MUC1/X mutant abrogate cleavage. This suggests that in order to sustain proteolysis, an intact SEA module having a specific protein structure is required and that particular N-terminal amino acids are of critical importance for the correct protein folding of a cleavage-competent SEA domain structure.
The SEA domain was initially identified in a sperm protein (31), in enterokinase, and in agrin and subsequently found in a number of heavily O-glycosylated mucin-like proteins that in all cases are extracellular proteins and, in many cases, membrane-bound (31). From the documented cleavage of MUC1, MUC3 (32), and IgHepta (33), it appears that proteins comprising SEA modules, similar to those found in the cleaved proteins, are also proteolytically cleaved. However, the SEA domains cluster into several subfamilies, each with common features, and it may be that some members of the SEA module-containing proteins cannot be cleaved. The structure of a SEA module derived from the ovarian cancer antigen MUC16 has been recently elucidated using multidimensional NMR spectroscopy and consists of a distinctive ␣/␤sandwich fold of two ␣-helices and four anti-parallel ␤-strands (34). A characteristic turn for all SEA modules, designated the TY-turn, appears between ␣-helix 1 and 2. The TY-turn is thus designated because it consists of highly conserved threonine (or serine) and tyrosine (or phenylalanine) residues, suggesting that this region may be important for cleavage of the SEA module (34).
Although the functional significance of MUC1 cleavage within its extracellular domain is uncertain, it is becoming increasingly evident that the transmembrane MUC1 protein has signaling properties because its cytoplasmic domain undergoes tyrosine and serine phos-  2 and 3, respectively). The extent of cleavage within each lane can be appreciated by comparing the relative intensities of the intact protein to its cleaved product. C, immunoblot of highly purified Fi18 and FLAG-Xex-hFc proteins that had not been incubated (lanes 1 and 3, respectively) or that were incubated for 48 h at 37°C (lanes 2 and 4, respectively). D, FLAG-Yex-hFc (lane 1) or Fi18 proteins were incubated at 37°C for the times indicated in the absence (Ϫ) or presence (ϩ) of 200 mM hydroxylamine (HA), and analyzed by immunoblotting. No cleavage was noted for FLAG-Yex-hFc protein (lane 1); partial cleavage for Fi18 protein was incubated in the absence of hydroxylamine, which increased with time (lanes 2-4, ␤ex-hFc cleavage products indicated by 7), and markedly enhanced cleavage when Fi18 protein was incubated in the presence of hydroxylamine (lanes 5 and 6). The percentage cleavage of each sample was derived by quantitative densitometry scanning of each lane and calculated by the formula {(C:(I ϩ C)) ϫ 100}, where C and I are densitometry values for cleaved (C) and intact (I) protein in each lane. These results are presented in the bar graph below the immunoblot. phorylation and interacts with cytoplasmic second messenger signaling proteins (35)(36)(37)(38)(39)(40)(41)43). However MUC1, because of its cleavage, does not represent a classical type I transmembrane protein wherein a ligand binds in trans to a contiguous receptor moiety; MUC1 is not a contiguous molecule. As the cleaved extracellular ␣-subunit is not covalently bound to the ␤-subunit and is free to dissociate and rebind, we speculated previously that this ␣/␤-subunit interaction might be a special case of a ligand-receptor interaction where the ␣and ␤-subunits represent the ligand and receptor molecules, respectively (30). This model merits further investigation as it could provide the basis for modulating cell phenotype by selectively intervening in the ␣/␤-subunit interaction.
As for most Ntn hydrolases, cleavage of the MUC1/X extracellular domain was reproduced in the heterologous Escherichia coli bacterial system. Similar to the bacterial wild type protein and the corresponding MUC1/X mutants expressed in eukaryotic cells, the bacterial S63T and S63C mutant MUC1/X proteins also demonstrated cleavage, albeit with reduced efficiency. This indicates that the bacterial system is likely not the most favorable one for supporting SEA module proteolytic cleavage, and it may be that post-translational modifications occurring in MUC1 in eukaryotic cells are required for optimal cleavage. In a similar way, the self-cleaving EMR2 protein undergoes autoproteolysis only when expressed in eukaryotic cells, whereas bacterially produced EMR2 is cleavage-incompetent.
In addition to serine 63, threonine or cysteine residues, and only these residues, at the P ϩ 1 site supported proteolytic cleavage. Ser-63 mutations to all other 17 amino acids, excluding threonine and cysteine, completely abrogated cleavage. Coupled with the finding of MUC1 cleavage in reticulocyte lysates (27), our results of MUC1/X cleavage both in bacterial and eukaryotic cells lead to the rather unexpected conclusion that an exquisitely specific protease must be present in all three systems that recognizes, in each case, the G2SVVV MUC1 cleavage site (82). Together with the difficulty of invoking a protease acting in trans, the finding of an unconditional MUC1 requirement for Ser, Thr, or Cys at the P ϩ 1 site suggested that, just like the self-cleaving hedgehog proteins, Ntn hydrolases, GPS-containing proteins, and the inteins, which all require an Ser, Thr, or Cys residue at P ϩ 1, MUC1 also undergoes autoproteolysis.
In addition to the requirement for a P ϩ 1 amino acid residue comprising an hydroxyl-or thiol-containing side chain, autoproteolysis stipulates two additional unique characteristics for the MUC1 protein, namely self-cleavage in the absence of any additional proteins, and nucleophile-enhanced cleavage of a slow-cleaving mutant MUC1 protein. Demonstration of these features would constitute strong evidence for MUC1 autoproteolysis. In fact, cleavage of in vitro incubated purified intein-containing proteins established that autoproteolysis mediates their cleavage (52). Intein-containing proteins isolated from thermophilic bacteria were used to ensure that protein synthesized and purified at reduced temperatures would result in uncleaved precursor protein, which could then be incubated in vitro to demonstrate autoproteolysis (52). Alternatively, selective mutation of amino acids required for the autoproteolytic reaction can result in slow cleaving mutant proteins, which upon purification and in vitro incubation can be used to demonstrate self-cleavage. This was the route taken to show cleavage both of glycosyl asparaginase (84) and the GPS-containing EMR2 protein (82).
We identified a partially cleaving mutant MUC1 protein by systematically mutating interspecies conserved serine and threonine residues located at the SEA module N terminus. Most interestingly, the mutations that produced the partially cleaving mutant Fi18 MUC1/X protein (30% cleavage) were located within the TY-turn described previously (34), which has been shown to be highly conserved in all SEA modules, suggesting a functional role for this region in the cleavage reaction.
Incubation of highly purified partially cleaving MUC1/X mutants in the absence of supplementary protein demonstrated self-cleavage of the MUC1 protein. This cleavage was remarkably enhanced by addition of the strong nucleophile hydroxylamine. These findings are compatible with features required of a self-cleaving protein, are in agreement with the autoproteolytic mechanism documented for Ntn hydrolases (85) and glycosylasparaginase, and provide firm evidence for autoproteolysis of the MUC1 protein. As documented for other self-cleaving proteins, the role of the nucleophile is to hydrolyze the ester bond formed by the N 3 O-acyl rearrangement, thereby leading to cleavage of the protein (67). The presence of the nucleophile with the slow-cleaving mutant facilitates hydrolysis of the intermediate and thereby shifts the reaction toward the direction of hydrolysis. It is very likely that the initial N 3 O-acyl rearrangement occurring at the cleavage site between glycine and serine residues of the GSVVV sequence does not occur in isolation (as depicted, for simplicity, in the model in Fig. 10) and that other constraints imposed by the spatial configuration of the SEA module as a complete unit are most likely required to lead to this initial event. Step 1, upon completion of MUC1/X (and MUC1/TM) synthesis, the protein assumes a cleavage-competent conformation, immediately followed by an N 3 O-acyl rearrangement mediated by attack on the glycine carbonyl carbon by the hydroxyl residue present on the P ϩ 1 serine side chain. The complete SEA module structure likely introduces constraints on peptide bonds within the vicinity of the cleavage site rendering them conducive to the N 3 O rearrangement.
Step 2, subsequent ester hydrolysis and cleavage generates the newly formed N-terminal hydroxyl group on Ser-63 and a C-terminal carboxyl group on Gly-62. The possibility that additional nucleophile(s) besides a water molecule attack the ester bond has not been ruled out, and this is indicated by "other nucleophiles" in Step 2.
We do not know the nature of the natural nucleophile within the cell that resolves the ester bond between glycine and serine residues leading to cleavage. The simplest scenario would invoke the nucleophilic oxygen atom of a water molecule that hydrolyzes the ester bond. However, there are other possibilities. For example, in the case of hedgehog proteins, the hydroxyl group of a cholesterol molecule attacks the ester bond formed by the N 3 O-acyl shift, leading to protein bisection and covalent addition of the cholesterol moiety to the final residue of the N-terminal fragment (86). In the case of intein excision, the ester bond formed by an N 3 O-acyl rearrangement at the N-terminal exteinintein junction is attacked by the hydroxyl group of a downstream serine residue in a transesterification reaction, and a peptide bond is then reformed from the new ester bond by the reverse O 3 N reaction (53).
Because of the similarity between the formation of the cleaved MUC1 proteins and the Ntn hydrolases, we speculate that the MUC1 ␤-subunit may also possess enzymatic activity. Ntn enzymes are amidases that autoproteolyze to generate a newly exposed N-terminal threonine, serine, or cysteine residue, which act as nucleophiles in hydrolytic amidase reactions. Members of the Ntn family include glucosamine-6-phosphate synthase (58) and asparagine synthase (59) (which use cysteine as the N-terminal nucleophile), penicillin acylase (60) (which uses serine), taspase1 (65), ␥-glutamyltranspeptidase (64), glycosylasparaginase (56,62), and the proteosome ␤-subunit (79) (which all use threonine). We are presently investigating whether, in a similar fashion to the Ntn clan, the MUC1 ␤-subunit also harbors Ntn-like hydrolytic activity.
In light of our findings we propose the following sequence of events (Fig. 10). Initially, the MUC1/TM and MUC1/X proteins are synthesized as a continuous polypeptide from their N terminus to the C terminus. Soon after synthesis, and at any rate still within the cytoplasm, the newly synthesized MUC1 proteins fold into a "cleavage-competent" conformation, wherein the hydroxyl group of the P ϩ 1 serine residue initiates an in cis nucleophilic attack on the carbonyl carbon of the preceding glycine residue to form an ester intermediate via an N 3 O shift. Hydrolysis of the ester bond (Fig. 10) generates ␣and ␤-subunits that specifically recognize and bind tightly to each other, thereby concluding the self-cleavage process. The MUC1 protein complex then progresses to its final location at the cell membrane as an ␣/␤-heterodimer (that is two asymmetric parts derived from the same parent MUC1 protein), tethered by the ␤-subunit transmembrane domain.
In summary, we have demonstrated that cleavage of the MUC1 proteins is an autoproteolytic reaction similar to the cis autoproteolysis seen in the self-cleaving hedgehog proteins, Ntn hydrolases, GPS-containing proteins, and the inteins. Exactly as in those self-cleaving proteins, MUC1 requires a Ser, Thr, or Cys residue at the P ϩ 1 site. Furthermore, an intact SEA module is required for MUC1 cleavage. In light of these findings, we suggest that all cleaved SEA module-containing proteins undergo a similar autoproteolytic reaction to that observed here for MUC1.