Tài liệu Báo cáo khoa học: The role of the SEA (sea urchin sperm protein, enterokinase and agrin) module in cleavage of membrane-tethered mucins pdf

The role of the SEA (sea urchin sperm protein, enterokinase and agrin) module in cleavage of membrane-tethered mucins Timea Palmai-Pallag 1 , Naila Khodabukus 1 , Leo Kinarsky 2 , Shih-Hsing Leir 1 , Simon Sherman 2 , Michael A. Hollingsworth 2 and Ann Harris 1 1 Paediatric Molecular Genetics, Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Oxford, UK 2 Eppley Institute for Research in Cancer and Allied Diseases, University of Nebraska Medical Center, Omaha, NE, USA SEA (sea urchin sperm protein, enterokinase and agrin) modules, minimally comprised of 120 amino acids of which 60% show strong conservation between proteins, are usually found in extracellular domains of dimeric or multimeric membrane-tethered proteins [1,2]. SEA modules contain proteolytic cleavage sites and amino-acid sequences that are important in the noncovalent association of protein subunits [2]. A heavily O-glycosylated domain is located N-terminal to the SEA module in the membrane-tethered mucins MUC1, MUC3, MUC12, MUC13, MUC16 and MUC17 [3]. Membrane-tethered mucins establish selective molecular barriers at epithelial cell surfaces and are implicated in diverse functions, including cell-adhesion, signalling, immune regulation and meta- stasis (reviewed in [3]). MUC1, the best-characterized membrane-tethered mucin, is transcribed as a single precursor protein that undergoes cleavage early in its processing to produce a heterodimer, which is further modified post-translationally and expressed in the cell Keywords MUC3, MUC12, proteolytic cleavage, SEA Correspondence A. Harris, Paediatric Molecular Genetics, Weatherall Institute of Molecular Medicine, John Radcliffe Hospital, Oxford OX3 9DS, UK E-mail: ann.harris@paediatrics.ox.ac.uk (Received 25 January 2005, revised 11 March 2005, accepted 7 April 2005) doi:10.1111/j.1742-4658.2005.04711.x The membrane-tethered mucins are cell surface-associated dimeric or multi- meric molecules with extracellular, transmembrane and cytoplasmic por- tions, that arise from cleavage of the primary polypeptide chain. Following the first cleavage, which may be cotranslational, the subunits remain closely associated through undefined noncovalent interactions. These mucins all share a common structural motif, the SEA module that is found in many other membrane-associated proteins that are released from the cell surface and has been implicated in both the cleavage events and association of the subunits. Here we examine the SEA modules of three membrane-tethered mucins, MUC1, MUC3 and MUC12, which have significant sequence homology within the SEA domain. We previously identified the primary cleavage site within the MUC1 SEA domain as FRPG ⁄ SVVV a sequence that is highly conserved in MUC3 and MUC12. We now show by site- directed mutagenesis that the F, G and S residues are important for the efficiency of the cleavage reaction but not indispensable and that amino acids outside this motif are probably important. These data are consistent with a new model of the MUC1 SEA domain that is based on the solution structure of the MUC16 SEA module, derived by NMR spectroscopy. Fur- ther, we demonstrate that cleavage of human MUC3 and MUC12 occurs within the SEA domain. However, the SEA domains of MUC1, MUC3 and MUC12 are not interchangeable, suggesting that either these modules alone are insufficient to mediate efficient cleavage or that the 3D structure of the hybrid molecules does not adequately recreate an accessible cleavage site. Abbreviations SEA, sea urchin sperm protein, enterokinase and agrin. FEBS Journal 272 (2005) 2901–2911 ª 2005 FEBS 2901 surface membrane [4]. MUC1 subunits remain associ- ated through a poorly understood non–covalent associ- ation, but can be released from the cell surface by additional proteolytic cleavage events [5,6], or other unknown mechanisms. The cleavage and release of membrane-tethered mucins is relevant to mucus accu- mulation in cystic fibrosis, and these molecules are implicated in the pathogenesis of many cancers inclu- ding pancreatic and breast carcinomas, where there is evidence for aberrant expression and processing of a number of mucin molecules [3]. Further, there is accu- mulating evidence that the cytoplasmic tail of MUC1 is involved in signal transduction and control of gene expression (reviewed in [3]), which would be affected by proteolytic cleavage events and the status of the extracellular domains [7]. We previously identified the early cleavage site of MUC1 in the SEA module at the Gly-Ser peptide bond located 58 amino acids upstream of the transmembrane domain (as defined in [8]), by N-terminal sequencing [9]. This cleavage site lies within a FRPG ⁄ SVVV motif that is highly conserved among the SEA modules of MUC1 glycoproteins from different species, and is found in other membrane-tethered mucins including MUC3, MUC12, MUC16 and MUC17 and other proteins with cleaved SEA domains (Table 1). The SEA module of rat MUC3 is cleaved soon after translation in vivo and in vitro at the Gly-Ser peptide bond of the LSKG ⁄ SIVVV motif [10], though additional amino acids C-terminal to this site are also required for efficient cleavage [11]. The solution structure of the SEA domain of the mouse homologue of MUC16 (which in the human encodes the core protein of the ovarian cancer antigen CA125) was recently solved by multidimensional NMR spectroscopy [12]. Similarity in sequence among the SEA modules of MUC1 and MUC16 enabled the gen- eration of a 3D model of MUC1 based on the MUC16 data. This model predicts that the FRPG ⁄ SVVV motif is contained in a surface-exposed loop. To further investigate the role of the SEA domain in cleavage of membrane-tethered mucins we carried out site-directed mutagenesis of the FRPG ⁄ SVVV site and evaluated hybrid mucins in which the SEA module of MUC1 was replaced by that from MUC3 and MUC12. Our aim was to determine whether cotranslational cleavage of membrane-tethered mucins was solely dependent on the FRPG ⁄ SVVV motif or whether additional sequen- ces, within or outside the SEA module, were required. We show that mutagenesis of the FRPG ⁄ SVVV motif does not completely prevent cleavage of MUC1. Though mutation of the Phe, Gly and Ser residues has a significant effect on the efficiency of the cleavage reaction, it is not completely abolished, suggesting that the precise sequence of this motif is not critical for clea- vage. In addition we determined that the SEA domains of human MUC3 and MUC12 are not interchangeable with that of MUC1, as substitution of these yielded cleaved products inefficiently. This suggests that motifs outside the SEA module may contribute to the cotrans- lational cleavage event in vivo or that the structures of the hybrid molecules do not enable efficient cleavage. Results Mutation of the FRPG ⁄ SVVV motif does not completely inhibit cleavage of MUC1FDTR MUC1FDTR undergoes early proteolytic cleavage at the FRPG ⁄ SVVV site (Fig. 1), which generates two Table 1. Conservation of the G ⁄ S cleavage site sequence in MUC1 from several species, other membrane-tethered mucins and SEA domain containing proteins. Protein Amino acid sequence Accession number MUC1 (human) GLSNIKFRPGS VVVQLTLAFREGTIN J05582 Muc1 (mouse) GISSIKFRSGS VVVESTVVFREGTFS U16175 Muc1 (cow) GLSEIKFRPGS VVVELTLAFREGTTA L41543 Muc1 (hamster) GISTIEFRSGS VVVDSTVIFREGAVN U36918 Muc1 (guinea pig) GLLNIKFRPGS VAVESTVIFRKNAVN L41546 MUC3 (human) GVEILSLRNGS IVVDYLVLLEMPFSP AF143371 Muc3 (rat) GVIIKNLSKGS IVVDYDVILKAQYTP U76551 MUC12 (human) GVNIRRLLNGS IVVKNDVILEADYTL AF147790 MUC16 (human) DCQVSTFRSVPNRHHTGVDSLCNFS PL AF361486 Muc16 (mouse) DCQVLAFRSVSNNNNHTGVDSLCNFS-PL [12] MUC17 (human) GVNITKLRLGS VVVEHDVLLRTKYTP AF430017 Ig-Hepta SVTVTQFTKGS VVVDYIVEVASAPLP AB019120 SPACR QLEILNFRNGS VIVNSKMKFAKSVPY AF017776 SPACRCAN QNLEILFRNGS¼¼¼IVVNSRMKFANSVPP AF157624 SEA modules and mucin cleavage T. Palmai-Pallag et al. 2902 FEBS Journal 272 (2005) 2901–2911 ª 2005 FEBS subunits: an extracellular component and a cell surface associated subunit containing the transmembrane region and the cytoplasmic tail. These two parts of the protein remain associated by noncovalent interac- tion, probably involving sequences within the SEA domain. The contribution of five different amino acids at the proteolytic cleavage site (shown in bold) FRPG ⁄ SVVV to the efficiency of cleavage of MUC1FDTR was investigated by mutating these resi- dues individually to alanine (A). The glycine (G) and serine (S) residues at the proteolytic cleavage site (resi- due numbers 185 and 186 in MUC1FDTR) were mutated first as these residues are conserved in MUC1 from several species, and in MUC3, MUC12, MUC17 and IgHepta (Table 1). Three additional residues close to the cleavage site, phenylalanine (F) 182, arginine (R) 183 and valine (V) 189, that are highly conserved amongst different mucins, were also mutated and evaluated. Caco2 colon adenocarcinoma cells were stably trans- fected with the wild-type (WT) and mutant constructs (G ⁄ A, S ⁄ A, F ⁄ A, R ⁄ A and V ⁄ A) and clones expres- sing high levels of the individual mutant MUC1FDTR constructs were identified. In Fig. 2 each lane contains mucin immunoprecipitated with an excess of M2 or CT-2 from 300 lg of total cell lysate. When inter- preting these data comparison should be made of the ratio of cleaved to noncleaved protein within each clone as the total amount of M2-reactive material var- ies between clones due to different expression levels. Figure 2A shows a western blot of MUC1FDTR pro- tein immunoprecipitated with M2 (anti-FLAG) from each clone, separated by 6% SDS ⁄ PAGE, and probed with M2. The major M2- reactive species in each clone migrates at around 75 kDa though resolution of this gel does not enable discrimination between cleaved and noncleaved protein. (Additional M2-reactive spe- cies, such as those at about 50 kDa are probably incompletely processed or differentially glycosylated forms of the epitope-tagged mucin as described previ- ously [9,13]). However, in Fig. 2B a western blot of a 10% SDS ⁄ PAGE gel of the same material probed with the CT-2 anti-cytoplasmic tail revealed both cleaved and noncleaved mutant MUC1FDTR. Wild-type MUC1FDTR generates CT-2 reactive species migrating between 15 and 30 kDa corresponding to the cytoplas- mic tail following cleavage at the G ⁄ S site. Multiple cytoplasmic tail species are routinely observed with the CT-2 antibody [13], which may be due to additional cleavage events and ⁄ or other modifications of the pro- tein that affect migration in the SDS ⁄ PAGE gel. CT-2 reactive species between 15 and 30 kDa are seen for all the mutants hence they must all undergo at least par- tial cleavage at the G ⁄ S motif. However, the G ⁄ A, MUC1 MUC1F∆ ∆ TR MUC1F∆TR /MUC3-CL MUC1F∆TR /MUC3-GA MUC1F∆TR /MUC12-CL MUC1F∆TR /MUC12SEA MUC1F∆TR /MUC3SEA MUC1F∆TR /MUC12FREG TR 926 aa TM CTSEAS SSEATMCT FRPGSVVV ** ** * STMCT LRNGSIVV STMCT LRNGSIVV * TM CTS LLNGSIVV STMCT LLNGSIVV CTTMS LRNGSIVV TM CTS LLNGSIVV FREG Fig. 1. Schematic representation of the dif- ferent MUC1FDTR constructs.Full length MUC1 is at the top of the figure and MUC1FDTR, derived from it but lacking the tandem repeat (TR) domain and containing a FLAG epitope (black box with flag) is shown below. All constructs are based on MUC1FDTR and contain the MUC1 signal sequence (S), transmembrane domain (TM) and cytoplasmic tail (CT) with the C-terminal black box representing the epitope for the CT2 antibody. The SEA domain (SEA) in each construct is as follows: MUC1, blank box, MUC3 diagonal stripe box, MUC12 diamond hatched box. The closed arrowheads denote the FRPG ⁄ SVVV cleavage site in MUC1 and predicted cleavage sites in MUC3 (LRNG ⁄ SIVV) and MUC12 (LLNG ⁄ SIVV); stars denote the sites of mutations. The open arrowhead denotes the LEAD to FREG change in MUC12. T. Palmai-Pallag et al. SEA modules and mucin cleavage FEBS Journal 272 (2005) 2901–2911 ª 2005 FEBS 2903 S ⁄ A and F ⁄ A mutants also generate a CT-2 reactive noncleaved species that migrates close to 75 kDa (Fig. 2B). Thus mutating G185, S186 or F182 to alan- ine partially inhibits the cleavage at the FRPG ⁄ SVVV motif whereas mutating R183 or V189 to alanine does not impair cleavage. Comparison of the relative inten- sity of the noncleaved and cleaved forms of the S ⁄ A, G ⁄ A and F ⁄ A mutants suggests that the S ⁄ A mutation has the most dramatic effect on the cleavage event. This observation was confirmed when mutants were transfected transiently into COS-7 cells as the results were similar to those observed in Caco2 cells with the exception of the S ⁄ A mutant, which showed almost complete inhibition of cleavage (data not shown). We have previously shown that low levels of MUC1F are released from Caco2 cells into the culture supernatant [9,13]. To evaluate the effect of the cleavage site muta- tions on release of MUC1FDTR, media conditioned by these clones was immunoprecipitated with M2 (with M2-beads in excess) (Fig. 2C,D). These data are not quantitative as the individual clones express different levels of the MUC1FDTR glycoprotein and some material enters the media via routes other than by membrane-tethered release (as demonstrated by CT-2 binding to cleaved MUC1FDTR cytoplasmic tail in the media, Fig. 2D). However, it is clear that despite the inefficient cleavage of the F ⁄ A, G ⁄ A and S⁄ A mutants, MUC1FDTR carrying these mutations is released from the cell (Fig. 2C). The cleavage site of MUC1 is predicted to lie in a surface loop on the SEA domain The recent publication of the solution structure of the SEA domain from murine MUC16, derived by NMR spectroscopy [12] enabled computer modelling of the MUC1 SEA domain (Fig. 3). The model has the same a ⁄ b sandwich fold as the template SEA domain of mu- rine MUC16 [12] and consists of three a-helices and six b-strands. This model predicts that the FRPG ⁄ SVVV cleavage site of MUC1 is located in a surface loop (residues marked) between b2 and b3 strands of the SEA module. Interestingly, the glycine (G185) and serine (S186) residues, which when mutated to alanine have the greatest impact on MUC1 cleavage, lie in the most exposed part of this loop. The phenylalanine (F182), which also affects MUC1 clea- vage when mutated to alanine, is located at the base of the loop and may interact with other residues in the adjacent b2 and b3 strands, affecting conformation and tightness of the loop. Next to this phenylalanine, the arginine residue (R183), which is located on the WT F/A R/A U G/A V/A S/A 50 75 105 160 250 U WT F/A R/A G/A S/A V/A 250 50 75 105 35 30 25 WT F/A R/A G/A S/A V/A WT F/A R/A G/A S/A V/A 75 105 30 25 A B C D Fig. 2. Expression and release of wild-type and cleavage site mutants of MUC1FDTR in Caco2 cells. Western blots of M2-immunopurified material from Caco2 clone cell lysates (A, B) or cell culture supernatant (C, D) from clones expressing wild- type (WT) or mutant (R ⁄ A, F ⁄ A, G ⁄ A, S ⁄ A, V ⁄ A) MUC1FDTR as shown. Samples were run on a 6% (A) or 10% (B, C, D) SDS ⁄ PAGE gel. The western blots was probed with M2 (A, C) or CT-2 (B, D). U, untransfected Caco2 cells. Mucins immunopre- cipitated with an excess of M2 beads or CT-2 from 300 lgof total cell lysate were loaded in each lane (A, B). Volumes of cell culture supernatant-derived M2 immunoprecipitated material loa- ded on gels in panels C and D were based on the intensity of the M2-reactive species in the corresponding whole cell lysate (not shown). SEA modules and mucin cleavage T. Palmai-Pallag et al. 2904 FEBS Journal 272 (2005) 2901–2911 ª 2005 FEBS N-terminal side of the loop, faces solvent and its muta- tion to alanine has little if any effect on the cleavage event. On the other side of the loop, the valine residue (V189) that interacts only with the C-terminal residues of the b1 strand, also has little impact on a loop con- formation and can be mutated to alanine without sig- nificant effect on the cleavage. It should be noted that relatively low overall sequence homology ( 25%) between the SEA modules of MUC1 and MUC16 and the shortening of the predicted loop in MUC1 (FRPGSV_ _ _VV in MUC1 compared to FRS- VSNNNNHT in MUC16) makes detailed structural analysis of specific interactions within this loop some- what speculative. Substitution of the MUC1 FRPG ⁄ SVVV motif with LRNG ⁄ SIVV from MUC3 or LLNG ⁄ SIVVV from MUC12 largely abolishes cleavage The conservation of the known proteolytic cleavage sites of MUC1 (FRPG ⁄ SVVV) and human MUC3 (LRNG ⁄ SIVV) and the predicted cleavage site of MUC12 (LLNG ⁄ SIVV) raises the possibility that a common protease and cleavage mechanism might be involved in the processing of these three membrane- associated mucins (Tables 1,2). To evaluate this hypo- thesis, 33 amino acids encompassing the FRPG ⁄ SVVV site (12 amino acids N-terminal and 13 amino acids C-terminal) of MUC1FDTR were replaced with equiv- alent sequences from MUC3 and MUC12 (Table 2). These hybrid constructs (MUC1FDTR ⁄ MUC3-CL and MUC1FDTR ⁄ MUC12-CL) were stably transfected into Caco2 colon adenocarcinoma cells and clones expressing high levels of the hybrid glycoproteins identified. Figure 4A and B show western blots of M2-immunoprecipitated MUC1FDTR ⁄ MUC3-CL and MUC1FDTR ⁄ MUC12-CL glycoproteins separated on 10% SDS ⁄ PAGE and probed with M2 antibody (Fig. 4A) or CT-2 (Fig. 4B). As shown in Fig. 2, the MUC1FDTR glycoprotein migrates at about 75 kDa (weakly evident in Fig. 4A) and a CT-2-reactive 15–30 kDa cleavage product identifies the cytoplas- mic tail after cleavage (Fig. 4B). Three forms of MUC1FDTR ⁄ MUC3-CL glycoprotein are seen bet- ween about 30 and 50 kDa (apparent MW) (Fig. 4A, MUC3-CL) in addition to a diffuse form at around 75 kDa that is seen more clearly in Fig. 4B. The fas- ter migrating glycoforms may include incompletely processed and differentially glycosylated forms. Similar migration profiles are seen for the MUC1FDTR ⁄ MUC3GA glycoprotein (Fig. 4A, MUC3-GA), which contains a Gly ⁄ Ala mutation in the cleavage site motif, Fig. 3. Model of SEA domain of MUC1. Ribbon representation of the modelled SEA domain of human MUC1. The model consists of three a-helices and six b-strands forming an a ⁄ b sandwich fold. The FRPG ⁄ SVVV residues of the MUC1 cleavage site, which is located in a loop between b2andb3 strands, are illustrated. Table 2. The SEA domains of MUC1, MUC3 and MUC12. The predicted cleavage site of each SEA domain is bold underlined. The FREG sequence that is conserved in MUC1 SEA domains from many different species and corresponding sequences from MUC3 and MUC12 are underlined. The numbered amino acids denote the limits of the SEA domain substitutions in MUC1 (J05582) by MUC3 (AF143371) and MUC12 (AF147790) sequences. Protein Sequence MUC12SEA E 273 KLN ATLGMTVKVTYRNFTEKMNDASSQEYQNFSTLFKNRMDVVL MUC3 SEA D 61 VVE TEVGMEVSVD.QQFSPDLNDNTSQAYRDFNKTFWNQMQKIF MUC1SEA PQLS TG V 128 SFFFLSFHISNLQFNSSLEDPSTDYYQELQRDISEMFLQIY MUC12SEA KGDNLPQYRGVNIRR LLNGSIVVKNDVILEADYT LEYEELFENLAEIVKA MUC3 SEA ADMQGFTFKGVEILS LRNGSIVVDYLVLLEMPFS PQLESEYEQVKTTLKE MUC1SEA KQG GFLGLSNIK FRPGSVVVQLTLAFREGTINVHDVETQFNQYKTEAAS MUC12SEA KIMNETRTTLLDPDSCR.KAILCY S 391 MUC3 SEA GLQNAS QDVNSCQDSQTLCF KPD S 178 MUC1SEA RYNLTIS DVSVSDVPFPFSA 238 Q T. Palmai-Pallag et al. SEA modules and mucin cleavage FEBS Journal 272 (2005) 2901–2911 ª 2005 FEBS 2905 and for the MUC1FDTR ⁄ MUC12-CL glycoprotein (Fig. 4A, MUC12-CL). When the same M2 immuno- precipitated proteins were western blotted and probed with CT-2 the major species detected for each of the hybrid proteins was noncleaved protein migrating close to 75 kDa (Fig. 4B, MUC1FDTR ⁄ MUC3-CL [MUC3-CL], MUC1FD TR ⁄ MUC12-CL [MUC12-CL], MUC1FDTR ⁄ MUC3GA [MUC3-GA]). However, a minor component migrated close to the cleaved cytoplasmic tail of MUC1FDTR (15–30 kDa) for MUC1FDTR ⁄ MUC3GA (Fig. 4B, MUC3-GA) and MUC1FDTR ⁄ MUC12-CL (Fig. 4B, MUC12-CL) gly- coproteins (15–30 kDa, marked by a bracket) but no cytoplasmic tail peptides were detected in MUC1FDTR ⁄ MUC3-CL (Fig. 4B, MUC3-CL). These data suggest that the replacement of the MUC1FDTR cleavage site by that from MUC12 enables an ineffi- cient cleavage event to occur but the MUC3 replace- ment largely abolishes cleavage. Next, the CT-2 antibody was used to immunopurify the glycopro- teins prior to western blotting (Fig. 4C,D). For the MUC1FDTR ⁄ MUC3-CL, MUC1FDTR ⁄ MUC12-CL and MUC1FDTR ⁄ MUC3GA constructs both M2 and CT-2 antibodies reacted with the 75 kDa species, sug- gesting that the majority of the glycoprotein from each construct is not cleaved (Fig. 4C). However, for all constructs, minor species were detected between 15 and 30 kDa (marked by the bracket) that reacted with the CT-2 antibody (Fig. 4D, MUC3-CL, MUC12-CL and MUC3-GA), but were not reactive with M2 (data not shown). More than one cleaved form of the cytoplasmic tail peptide is seen for both MUC3 and MUC12-containing hybrid molecules. These data show that the replacement of 33 amino acids flanking the MUC1FDTR cleavage site by equivalent sequences from MUC3 and MUC12 enables a cleavage event to occur within these regions but does not provide an efficient substrate for prote- ase-mediated cleavage at the G ⁄ SIVV motif. It may be relevant that the G ⁄ A mutant of the MUC3 clea- vage site is apparently more efficiently cleaved than the wild-type sequence in these hybrid molecules implicating structural constraints on the cleavage motif. U MUC1F∆TR MUC3-GA MUC3-CL MUC12-CL MUC3SEA MUC12SEA MUC12FREG 105 kDa 75 kDa 50 kDa 35 kDa 30 kDa 25 kDa 15 kDa A U MUC1F∆TR MUC3-GA MUC3-CL MUC12-CL MUC3SEA MUC12SEA MUC12FREG 105 kDa 75 kDa 50 kDa 35 kDa 30 kDa 25 kDa 15 kDa B 105 kDa 75 kDa 50 kDa 35 kDa 30 kDa 25 kDa 15 kDa U MUC1F∆TR MUC3-GA MUC3-CL MUC12-CL MUC3SEA MUC12SEA MUC12FREG C U MUC1F∆TR MUC3-GA MUC3-CL MUC12-CL MUC3SEA MUC12SEA MUC12FREG 105 kDa 75 kDa 50 kDa 35 kDa 30 kDa 25 kDa 15 kDa D } } Fig. 4. Expression and cleavage of the MUC1FDTR hybrid SEA domain mucins in Caco2 cells. Western blots of hybrid mucin, immunopuri- fied from stably transfected Caco2 clones with M2 anti-FLAG antibody (A, B) or with CT-2 anti-cytoplasmic tail antibody (C, D). Mucin was separated on 10% SDS ⁄ PAGE gels and probed with M2 (A, C) or with CT-2 (B, D). The brackets in B and D denote C-terminal cleavage products of the hybrid mucins. Amounts of mucin per lane are as described in the legend for Fig. 2. U denotes material immunopurified with the relevant antibody from untransfected Caco2 cells. The clones carry the constructs described in Fig. 1. SEA modules and mucin cleavage T. Palmai-Pallag et al. 2906 FEBS Journal 272 (2005) 2901–2911 ª 2005 FEBS In a further attempt to improve the efficiency of cleavage of MUC1FDTR ⁄ MUC12-CL, we next intro- duced the highly conserved MUC1 FREG sequence into the partial MUC12 SEA module of the hybrid protein to replace the native LEAD sequence (Table 2). Evaluation of the hybrid mucin immunopre- cipitated with M2 or CT-2 from stably transfected Caco2 cells showed that substitution of the FREG sequence did not markedly improve the efficiency of cleavage of this glycoprotein (Fig. 4A–D, MUC12- FREG). Substitution of the SEA domain of MUC1 with those of MUC3 or MUC12 provide substrates for mucin cleavage Rodent MUC3 is apparently cleaved twice during its post-translational processing, first at the G ⁄ SIVV resi- due within the SEA domain [11] and subsequently at a second site that is currently undefined, but is located at least 10 amino acids C-terminal to the G ⁄ S site [10]. Based on these data and the current observations that replacement of the small fragment of the SEA domain (33 amino acids flanking the G ⁄ SVVV motif) resulted in substantial inhibition of the cleavage, the entire SEA domain (boundaries established by [1], Table 2) of MUC1FDTR was replaced with the SEA modules from human MUC3 and MUC12. The MUC1FDTR ⁄ MUC3SEA and MUC1FDTR ⁄ MUC12SEA constructs were stably expressed in Caco2 clones. Western blots of M2-purified MUC1FDTR ⁄ MUC3- SEA and MUC1FDTR ⁄ MUC12SEA glycoproteins showed a major M2-reactive and CT-2 reactive non- cleaved glycoform of about 105–120 kDa (Fig. 4A,C, MUC3SEA and MUC12SEA) and a minor population of a CT-2 reactive, M2 nonreactive cleavage product, at about 25 kDa for MUC3 and 30 kDa for MUC12 (Fig. 4B,D, MUC3SEA and MUC12SEA). There appear to be 2 forms of the CT-2 reactive peptide in the MUC3-containing construct, consistent with previ- ous observations [10]. A slight difference in the mobi- lity of the major glycoform of MUC1FDTR and MUC1FDTR ⁄ MUC3SEA or MUC1FDTR ⁄ MUC12- SEA may be accounted for by the introduction of an additional N-glycosylation site. These data suggest that the SEA domains are not completely interchangeable between different mucins; however, they form a mod- ule that can be cleaved even when substituted into another mucin. The cleavage is apparently more effi- cient in the hybrid mucins carrying the whole SEA domains rather than the 33 amino acid fragments encompassing the G ⁄ SIVV motif alone as the ratio of cleaved to uncleaved protein is greater (Fig. 4B,D), however, it is still highly inefficient in comparison to MUC1FDTR (Fig. 4B,D). Discussion The function of the SEA domain in the cleavage and sub–unit association of membrane-tethered glycopro- teins is well documented but poorly understood. Our aim was to evaluate the role of this domain in the processing of membrane-tethered mucins. The mech- anism by which the heavily glycosylated portion of these molecules is released at epithelial surfaces within the airway, the intestine and pancreatic ducts is of considerable significance in the pathology of a number of diseases including cystic fibrosis and cancer. Equally important is the fate of the cytoplasmic tail of the mucins once the extracellular domain is lost. This peptide has been shown to associate with b-cate- nin [7,14], Grb2 [15] and erbB family members [16] and to be involved in signal transduction (reviewed in [3]). We previously determined that there is an important early cleavage of the MUC1 mucin at an FRPG ⁄ SVVV motif within the SEA module of the protein [9]. Herein we show that mutating G185, S186 or F182 to alanine inhibits but does not completely block cleavage at the FRPG ⁄ SVVV motif, whereas mutating R183 or V189 to alanine does not impair cleavage. Compari- sons of the relative intensities of the noncleaved and cleaved forms of the S ⁄ A, G ⁄ A and F ⁄ A mutants sug- gests that the S ⁄ A mutation has the most dramatic effect on the cleavage event. Interestingly, the ineffi- cient cleavage reduces, but does not abolish, the sur- face membrane targeting of these mucin glycoproteins (data not shown). The FRPG ⁄ SVVV site and homo- logous sites in other membrane-tethered mucins, have recently been studied by several other groups. Muta- genesis studies by Lillehoj et al. [17] on the MUC1 cleavage site FRPG ⁄ SVVV yielded data that are partly but not completely consistent with our results. Lillehoj et al. [17] observed that the serine to alanine mutation at the cleavage site completely abolished cleavage in COS7 cells and in several human airway and breast cancer cell lines that express endogenous MUC1. We observed the same results in COS7 cells; however, in contrast to the published findings with human airway and breast cancer cell lines, when the Ser to Ala mutant was transfected into the Caco2 colon carci- noma cell lines that expresses very low levels of endo- genous MUC1, we observed partial cleavage of the protein. It is currently not clear whether cleavage at the G ⁄ SVVV site is protease-mediated or autocatalytic. If the former is true, then the difference between our T. Palmai-Pallag et al. SEA modules and mucin cleavage FEBS Journal 272 (2005) 2901–2911 ª 2005 FEBS 2907 data and those of Lillehoj et al. [17] may reflect the different spectrum of proteases in different cell types or species. Alternatively, the failure of Lillehoj et al. [17] to see cleavage of the Ser to Ala mutant in the breast and airway lines may be explained by the possi- bility that the antibodies employed in that study detect endogenous MUC1 in addition to the transfected mutant construct. In our experiments, the use of anti- bodies against a FLAG epitope tag in MUC1FDTR enabled discrimination between endogenous MUC1 and MUC1FDTR. The rat MUC3 mucin is cleaved at a site homolog- ous to the MUC1 site, LSKG ⁄ SIVV and mutagenesis of the glycine residue to alanine in this sequence reduced the efficiency of cleavage [10], whereas a serine to alanine mutation abolished cleavage in COS 7 cells. In addition to the LSKG ⁄ SIVV mediated cleavage, a second proteolytic event has been reported in rat MUC3 [11] and though the precise cleavage site has not been mapped it is at least 10 amino acids C-ter- minal to the first site. This is of particular interest as we, and others, have detected multiple forms of the cytoplasmic tail of MUC1, MUC1FDTR and its deriv- atives [9,13]. Further, the migration of the CT2-react- ive forms of the hybrid mucins shown in the current work predicts more than one species derived from the MUC3- and possibly the MUC12-SEA domain. This raises the possibility of additional cleavage events within the SEA domain that are probably subsequent to that at the G ⁄ SVVV. Inspection of the predicted 3D structure of the SEA domain of MUC1 suggests that more than one cleavage event is required to release the extracellular domain of the protein from its cytoplas- mic tail. Recent data on additional cleavage events for the MUC1SEA domain in uterine epithelial cells, mediated by defined proteases including TACE and MT1-MMP, which are not active at the G ⁄ SVVV site, provide evidence for this [5,6]. However, it is also poss- ible that some of the multiple forms of CT2-reactive peptides seen on western blots are modified by other post-translational modifications, for example by phos- phorylation. When we substituted 33 amino acids flanking the cleavage site of MUC1FDTR or the whole SEA mod- ule with equivalent sequences from the SEA module of MUC3 or MUC12, the efficiency of the cleavage event was greatly reduced in comparison to WT MUC1FDTR. Nonetheless, small amounts of cleaved cytoplasmic tail were detected on western blots for all hybrid molecules. These observations support the data from Wang [10] and Khatri [11] that clearly demon- strate the cleavage of rat MUC3 at the LSKG ⁄ SIVV site and provide in vivo confirmation for cleavage of human MUC3 in the SEA domain. Further, our data provide the first evidence for the cleavage of the MUC12 mucin SEA domain that would be predicted to occur at the LLNG ⁄ SIVV site. The greatly reduced efficiency of the cleavage in the hybrid mucins in com- parison to that observed for MUC1FDTR might be accounted for by the 3D structure of the hybrid mole- cules, which may not allow proper folding and accessi- bility of the cleavage site. This hypothesis is consistent with our observations that substitution of the whole SEA domain is marginally more effective at generating a substrate for cleavage than are the 33 amino acid substitutions. Another feature of the hybrid mucins is that that they lack motifs of the native MUC3 and MUC12 that may contribute to the generation of the cleavage substrate, for example the flanking EGF domains. Alternatively, the SEA modules of the differ- ent membrane-tethered mucins may not be sufficiently homologous to allow substitution of one for another. It is relevant that, due to poor sequence homology, we were not able to generate a reliable model of the SEA domain of MUC3 or MUC12 based on the solution structure of the MUC16 SEA domain derived by NMR spectroscopy. In summary, the results presented here demonstrate that cleavage of membrane associated mucins in the SEA domain is not entirely sequence dependent, but is instead related to general structural features of the SEA domain. Molecular modelling of the MUC1 SEA domain supports the hypothesis that the cleavage site exists as an exposed loop that protrudes outward from an otherwise folded compact structure, which we pre- dict provides molecular topological features enabling its recognition and cleavage by cellular proteases. Experimental procedures Site directed mutagenesis of the G ⁄ SVVV site of MUC1 and the LRNG ⁄ SIVV site of human MUC3 Site-specific mutations were carried out using the Quik- Change site-directed mutagenesis kit (Stratagene, Cedar Creek, TX, USA). For all constructs a FLAG-epitope-tagged MUC1 cDNA that lacked native tandem repeat sequences (MUC1FDTR [18], [13], Fig. 1) was used. The F182, R183, G185, S186 and V189 of the FRPG ⁄ SVVV site in MUC1FDTR were mutated individually to alanine using the following prime r pairs: G ⁄ A substitution: 5¢-GTTCAGGC CAGCATCTGTGGTGGTACAATTG-3¢ (sense), 5¢ -CAAT TGTACCACCACAGATGCTGGCCTGAAC-3¢ (antisense), S ⁄ A substitution: 5¢-GTTCAGGCCAGGAGCTGTGGTG GTACAATTG-3¢ (sense), 5¢-CAATTGTACCACCACAG CTCCTGGCCTGAAC-3¢ (antisense), R ⁄ A substitution: SEA modules and mucin cleavage T. Palmai-Pallag et al. 2908 FEBS Journal 272 (2005) 2901–2911 ª 2005 FEBS 5¢-CTCCAATATTAAGTTCGCGCCAGCATCTGTGGT GG-3¢ (sense), 5¢-CCACCACAGATCCTGGCGCGAACTT AATATTGGAG-3¢ (antisense), V ⁄ A substitution: 5¢-CCAGG ATCTGTGGTGGCACAATT GACT CTG-3 ¢ (sense), 5¢-CAG AGTCAATTGT GCCACCACAGATCCTGG-3¢ (antisense), F ⁄ A substitution 5¢-TCTCCAATATTAAGGCCAGGCCA GGATCTGTGGTG-3¢ (sense), 5¢-CACCACAGATCCTG GCCTGGCCTTAATATTGGAGA-3¢ (antisense). In the MUC1FDTR ⁄ MUC3 chimera the glycine in the LRNG ⁄ SIVV motif was mutated to alanine using the following pri- mer pairs 5¢-CCTGTCCCTGAGGAATGCCAGCATCGT GGTGGAC-3¢ (sense) and 5¢-GTCCACCACGATGCTGG CATTCCTCAGGGACAGG-3¢ (antisense). Nucleotides in bold and underlined represent the base-pair changes required to generate the appropriate amino acid substitu- tion. The mutations were confirmed by DNA sequencing. Constructs were subcloned into the mammalian expression vector pHb-APr1-neo [19] at the BamH1 site. Generation of MUC1FDTR constructs containing MUC3- and MUC12- cleavage sites and the MUC3 and MUC12 SEA domains All constructs were based on the MUC1FDTR backbone [18,13], and are illustrated in Fig. 1. Using unique PsiI (J05582: 3351) and Dra III (J05582: 3421) restriction sites, 33 amino acids (from K170 to V202 inclusive, with amino acids numbered according to [13]) including the FRPG ⁄ SVVV proteolytic cleavage site of MUC1FDTR were replaced with PCR amplified cDNA sequences from human MUC3 (AF143371) and human MUC12 (AF147790). The primer pair used to produce the MUC3 cDNA insert was (sense) 5¢-ATTTATAAGGGCTTCACC TTCAAG-3¢ (AF143371 324–339) and (antisense) 5¢-CTCC ACGTCGTGCTGGGGGCTGAAGGG-3¢ (AF143371 406– 420), the MUC12 cDNA insert was (sense) 5¢-ATTTAT AATCTTCCTCAGTATAGAGGG-3¢ (AF147790 965–983), and (antisense) 5¢-CTCCACGTCGTGCTCTAAAGTGTA GTC-3¢ (AF147790 1047–1061). Mucin-specific sequences (bold) were extended by eight nucleotides carrying a PsiI recognition site in the sense primers, or 12 nucleotides with a DraIII site in the antisense primers. PCR parame- ters were 30 cycles of 94 °C for 30 s, 60 °C for 30 s, 72 °C for 1 min with a final extension at 72 °C for 5 min. The PCR generated fragments were digested with PsiI and DraIII and cloned into the respective sites of the plasmid pUC18 ⁄ MUC1FDTR. The mucin constructs were then transferred as BamHI fragments into the BamHI site of the mammalian expression vector pHb-APr1-neo [19]. MUC1FDTR ⁄ MUC12FREG was identical to the MUC1FDTR ⁄ MUC12-CL construct but with the LEAD sequence 11 amino acids C-terminal to the G ⁄ S cleavage site replaced by the FREG sequence that is conserved in MUC1 proteins from many different species. The following oligonucleotides were synthesized: M12FREG-F (sense) 5¢-TAATCTTCCTCAGTATAGAGGGGTGAACATTCG GAGATTGCTCAACGGTAGCAT CGTGGTCAA GAAC GATGTCATCTTCCGAGA AGGTTACA CTT TAGAGCA CGAC-3¢ and M12FREG-R (antisense) 5¢-GTGCTCTAA AGTGTAACCTTCTCGGAAGAT GACATCGTT CTTGA CCACGATGCTACCGT TGAGCAATCTCCGAATGTTC ACCCCTCTATACTGAGGAAGATTA)3¢ (AF147790 965– 1061) where the PsiI and DraIII sites are in italics and the FREG motif in bold. These oligonucleotides were annealed and cloned into the PsiI ⁄ DraIII sites of MUC1FDTR. To replace the SEA domain of MUC1 with those of MUC3 or MUC12, XcmI sites were used to remove 111 amino acids from V128 to A238 of MUC1FDTR. These were replaced by 118 amino acids from D61 to S178 of MUC3 (AF143371) or by 119 amino acids from E273 to S391 of MUC12 (AF147790). Oligonucleotides were syn- thesized to encode an XcmI-SpeI-EcoRI-XhoI-XcmI poly- linker. SEALINK1 (sense) 5 ¢-CTACTGGACTAGTGAA TTCCTCGAGCCAGTCTG-3¢ and SEALINK2 (antisense) 5¢- AGACTGGCT CGAGGA ATTCAC TAGTCCA GTAGA-3 ¢ were annealed and directly cloned into the XcmI site of MUC1FDTR. The MUC3 and MUC12 SEA domains were amplified by PCR using gene-specific primers 3-SEA-F (sense) 5¢-GGACTAGTAGATGTAGTGGAGACCGAG-3¢ (AF143371 180–198), 3-SEA-R (antisense) 5¢-CCGCTC GAGTCAGGCTTAAAACACAGG-3¢ (AF143371 513–530), 12-SEA-F 5¢-GGACTAGTGGAAAAACTCAAGGCCACT TTAGG-3¢ (AF147790 818–841) and 12-SEA-R 5¢-CCGCT CGAGTAGCACAGTATGGCCTTTC-3¢ (AF147790 1153– 1171). SpeI and XhoI sites were incorporated at the 5¢ end of the forward and reverse primers, respectively, and MUC3 and MUC12-specific sequences are shown in bold. PCR conditions were as described previously. The frag- ments were then cloned into the SpeI ⁄ XhoI sites of the polylinker in MUC1FDTR. The remaining portions of the polylinker introduced leucine and valine residues N-ter- minal to the MUC3 and MUC12 specific sequences and a serine residue C-terminal to them. Transient expression of epitope-tagged mucin proteins in Cos-7 cells and generation of epitope- tagged mucin-expressing stable Caco2 cell clones Cos7 and Caco2 cells were cultured in DMEM containing 10% (v ⁄ v) FBS, penicillin (100 UÆmL )1 ), streptomycin (100 lgÆmL )1 ). For transient expression, Cos7 cells were grown to 50% confluence in 90 cm dishes. For each dish, 20 lg of plasmid DNA was transformed by standard meth- ods using Lipofectin (Invitrogen, Carlsbad, CA, USA) or calcium phosphate. Cells were lysed 48 h post-transfection and proteins analysed by western blot. For stable clones Caco2 cells were transfected with Lipofectin (Invitrogen) by standard methods and clones selected in G418 (Invitrogen) at 600 lgÆmL )1 . T. Palmai-Pallag et al. SEA modules and mucin cleavage FEBS Journal 272 (2005) 2901–2911 ª 2005 FEBS 2909 Preparation of cell lysates, immunopurification of epitope-tagged hybrid-mucins and Western blot analysis Cells were lysed in NET buffer (50 mm Tris HCl pH 7.5, 5mm EDTA, 150 mm NaCl) with complete protease inhi- bitor cocktail (Sigma, St. Louis, MO, USA) at 2–3 day post- confluence. Epitope tagged mucins were immunopurified with M2 anti-FLAG conjugated agarose beads (Sigma) as described previously [13] or with CT2 antibody (directed against the last 17 amino acids [SSLAYTNPAVAATSANL] of the cytoplasmic tail of MUC1, kindly donated by S. Gendler, Mayo Clinic, Scottsdale, AZ, USA [16]) in the presence of Protein G ⁄ Sepharose (Sigma) at 4 °C. Mucins were eluted from the beads ⁄ Sepharose with specific peptides. Immunoprecipitates were subjected to SDS ⁄ PAGE under reducing conditions, and transferred to Hybond ECL mem- branes. Mucins immunoprecipitated with an excess of M2 beads or CT2 from 300 lg of total cell lysate were loaded in each lane. Rainbow marker (Amersham, Chalfont St. Giles, Bucks, UK) was used as a molecular mass standard. Mem- branes were incubated with M2 or CT2 and reactive protein species were detected with the Hybond ECL western Blot Detection Kit (Amersham). Mucins released from the cells were collected from cell culture medium conditioned for 4 days postconfluence. The media was cleared by centrifugation at 300 g for 10 min and then immunoprecipitated with 100 lL M2 agarose beads for 30 mL of culture medium. Immunoprecipitates were subjected to SDS ⁄ PAGE as described above and vol- umes were loaded according to the intensity of M2-reactive species in the corresponding whole cell lysate. Molecular modeling of SEA domain of MUC1 The composer and biopolymer modules of the sybyl 6.91 software package (Tripos Inc., St Louis, MO, USA) were used for comparative molecular modelling of the human MUC1 SEA domain based on the NMR-derived structure of the murine MUC16 SEA domain (PDB code:1IVZ). Sec- ondary structure-based multiple sequence alignment [12] was used to build a model. The modelled structure of the MUC1 SEA domain was refined by energy minimizations using the amber (Kollman all-atom) force field implemen- ted in sybyl. The quality of the structural model was tested with the protable module of sybyl. Acknowledgements We thank Dr S. Williams for the MUC3 and MUC12 partial cDNAs and Dr S. Gendler for the CT2 anti- body. This work was supported by the Cystic Fibrosis Trust, UK, a Wellcome Trust Biomedical Research collaboration grant and NIH Grants 1R01 CA84106 to S.S and CA57326 to M.A.H. References 1 Bork P & Patthy L (1995) The SEA module: a new extracellular domain associated with O-glycosylation. Protein Sci 4, 1421–1425. 2 Wreschner DH, McGuckin MA, Williams SJ, Baruch A, Yoeli M, Ziv R, Okun L, Zaretsky J, Smorodinsky N, Keydar I et al. (2002) Generation of ligand-receptor alliances by ‘SEA’ module-mediated cleavage of mem- brane-associated mucin proteins. Protein Sci 11, 698– 706. 3 Hollingsworth MA & Swanson B (2004) Mucins in Can- cer: protection and control of the cell surface. Nat Rev Cancer 4, 45–60. 4 Ligtenberg MJ, Kruijshaar L, Buijs FM, van Meijer M, Litvinov SV & Hilkens J (1992) Cell-associated episialin is a complex containing two proteins derived from a common precursor. J Biol Chem 267, 6171–6177. 5 Thathiah A, Blobel CP & Carson DD (2003) Tumor necrosis factor-alpha converting enzyme ⁄ ADAM 17 mediates MUC1 shedding. J Biol Chem 278, 3386– 3394. 6 Thathiah A & Carson DD (2004) MT1-MMP mediates MUC1 shedding independent of TACE ⁄ ADAM17. Biochem J 382, 363–373. 7 Wen Y, Caffrey TC, Wheelock MJ, Johnson KR & Hollingsworth MA (2003) Nuclear association of the cytoplasmic tail of MUC1 and beta-catenin. J Biol Chem 278, 38029–38039. 8 Lan MS, Batra SK, Qi WN, Metzgar RS & Hollings- worth MA (1990) Cloning and sequencing of a human pancreatic tumor mucin cDNA. J Biol Chem 265, 15294–15299. 9 Parry S, Silverman HS, McDermott K, Willis A, Hollingsworth MA & Harris A (2001) Identification of MUC1 proteolytic cleavage sites in vivo. Biochem Bio- phys Res Commun 283, 715–720. 10 Wang R, Khatri IA & Forstner JF (2002) C-terminal domain of rodent intestinal mucin Muc3 is proteolyti- cally cleaved in the endoplasmic reticulum to generate extracellular and membrane components. Biochem J 366, 623–631. 11 Khatri IA, Wang R & Forstner JF (2003) SEA (sea- urchin sperm protein, enterokinase and agrin) –module cleavage, association of fragments and membrane tar- geting of rat intestinal mucin Muc3. Biochem J 372, 263–270. 12 Maeda T, Inoue M, Koshiba S, Yabuki T, Aoki M, Nunokawa E, Seki E, Matsuda T, Motoda Y, Kobaya- shi A et al. (2004) Solution structure of the SEA domain from the murine homologue of ovarian cancer antigen CA125 (MUC16). J Biol Chem 279, 13174– 13182. 13 Silverman HS, Parry S, Sutton-Smith M, Burdick MD, McDermott K, Reid CJ, Batra SK, Morris HR, SEA modules and mucin cleavage T. Palmai-Pallag et al. 2910 FEBS Journal 272 (2005) 2901–2911 ª 2005 FEBS [...]... Hollingsworth MA, Dell A & Harris A (2001) In vivo glycosylation of mucin tandem repeats Glycobiology 11, 459–471 14 Yamamoto M, Bharti A, Li Y & Kufe D (1997) Interaction of the DF3 ⁄ MUC1 breast carcinoma-associated antigen and beta-catenin in cell adhesion J Biol Chem 272, 12492–12494 15 Pandey P, Kharbanda S & Kufe D (1995) Association of the DF3 ⁄ MUC1 breast cancer antigen with Grb2 and the Sos... with Grb2 and the Sos ⁄ Ras exchange protein Cancer Res 55, 4000–4003 16 Schroeder JA, Thompson MC, Gardner MM & Gendler SJ (2001) Transgenic MUC1 interacts with epidermal growth factor receptor and correlates with mitogen-acti- FEBS Journal 272 (2005) 2901–2911 ª 2005 FEBS SEA modules and mucin cleavage vated protein kinase activation in the mouse mammary gland J Biol Chem 276, 13057–13064 17 Lillehoj... (2003) Mutagenesis of a Gly-Ser cleavage site in MUC1 inhibits ectodomain shedding Biochem Biophys Res Commun 307, 743–749 18 Burdick MD, Harris A, Reid CJ, Iwamura T & Hollingsworth MA (1997) Oligosaccharides expressed on MUC1 produced by pancreatic and colon tumor cell lines J Biol Chem 272, 24198–24202 19 Gunning P, Leavitt J, Muscat G, Ng SY & Kedes L (1987) A human beta-actin expression vector... pancreatic and colon tumor cell lines J Biol Chem 272, 24198–24202 19 Gunning P, Leavitt J, Muscat G, Ng SY & Kedes L (1987) A human beta-actin expression vector system directs high-level accumulation of antisense transcripts Proc Natl Acad Sci USA 84, 4831–4835 2911 . The role of the SEA (sea urchin sperm protein, enterokinase and agrin) module in cleavage of membrane-tethered mucins Timea Palmai-Pallag 1 ,. (2003) SEA (sea- urchin sperm protein, enterokinase and agrin) module cleavage, association of fragments and membrane tar- geting of rat intestinal mucin