The Carboxyl-terminal Tail of Kinase Splitting Membranal Proteinase/Meprin β Is Involved in Its Intracellular Trafficking*

The kinase splitting membranal proteinase (KSMP), was recently shown to be identical with the β-subunit of meprin. Meprin is a metalloendoproteinase located in brush border membranes and composed of the two types of subunits, α and β. Despite their high sequence homology and similar domain organization, meprin subunits are differently processed during maturation; meprin α is retained in the endoplasmic reticulum (ER), and undergoes a proteolytic removal of the transmembrane and cytoplasmic domains, prior to its export from this organelle. In contrast, meprin β retains these domains even after reaching its final destination in the plasma membrane. Using truncated mutants of rat meprin β expressed in Cos-7 and human embryonic kidney (HEK) 293 cells, we show here that the cytoplasmic tail is indispensable for its exit from the ER. A meprin β mutant lacking the last 25 amino acids is shown to be transport-incompetent, although it does not contain any of the known ER retention signals. Systematic analysis of the rate of the ER to Golgi transport using a series of mutants with Ala or Pro substitutions in the tail, suggests that while no specific amino acid residue by itself is imperative for normal intracellular trafficking of meprin β, the insertion of a bend at a distinct position in the tail (specifically by a Y685P mutation) suffices to retain this protein in the ER. We propose that the very length of the cytoplasmic tail, as well as its secondary structure are essential for the ER to Golgi transport of meprin β, possibly by allowing an interaction with a cargo receptor.

The presence of a kinase splitting membranal proteinase (KSMP) 1 in brush border membranes was demonstrated as early as 1979 (1,2). The proteinase was shown to clip the catalytic subunit of PKA (C-subunit) with the formation of a distinct cleavage product devoid of kinase activity (2). Recently we demonstrated that the C-subunit degrading activity of KSMP can be quantitatively reproduced by the ␤-subunit of meprin (3) and reported a unique specificity of the enzyme toward substrates containing clusters of acidic (Glu and Asp) amino acid residues, decorated with hydrophobic amino acid residues (4). KSMP/meprin is a Zn 2ϩ /Mg 2ϩ metalloendoproteinase located in brush border membranes of the small intestine (rat, mouse, guinea pig, rabbit, and man) and of the kidney (rat, mouse, guinea pig, and man) (5)(6)(7)(8)(9)(10). Meprins belong to the astacin family of endopeptidases (6,8) and are composed of two types of subunits: ␣ and ␤, that exist as homo-and heterotetramers (7,9). It was previously proposed that meprins might be involved in the degradation of some biologically active peptides such as bradykinin, angiotensin, and substance P (6), emphasizing that meprin subunits differ in their specificity. For example, it was shown that bradykinin is a good substrate for mouse meprin A (which is assumed to be either a heterooligomer of ␣and ␤-subunits or homo-oligomer of ␣-subunits), but not for meprin B (a homo-oligomer of ␤-subunits) (11). Furthermore, we have recently demonstrated that while the rat meprin ␤-subunit cleaves the C-subunit of PKA, the meprin ␣-subunit did not cleave this kinase under the same experimental conditions (4). The physiological targets and the exact role of meprins are still not established. Nevertheless, specificity studies carried out recently in our laboratory showed that the peptide hormone gastrin (which is physiologically accessible to this ecto-proteinase), has a significantly lower K m than any other previously tested substrate of meprin ␤). Since this peptide hormone is inactivated by such cleavage, we proposed that gastrin may well be an in vivo substrate of meprin ␤ (4).
Despite the high sequence homology and similar domain structure, the different meprin subunits undergo different post-translational proteolytic processing. The cDNA structure of both ␣and ␤-subunits of meprin predicts the presence of carboxyl-terminal, cytoplasmic, transmembrane, and epidermal growth factor-like domains, which are all still present in the mature ␤-subunit, but removed from the ␣-subunit in the course of its maturation (10,(12)(13)(14). In addition, the ␣-subunit of meprin also differs from the ␤-subunit by possessing a domain (denoted the I domain) which is both necessary and sufficient for the proteolytic processing of the immature ␣-subunit (15).
It was recently reported that the ␣-subunit of the human analog of meprin is retained in the endoplasmic reticulum until the proteolytic removal of the transmembrane and cytoplasmic domains is completed, and that the cytoplasmic tail of this subunit of human meprin is responsible for its retention in the endoplasmic reticulum (16). Since only the ␤-subunit of meprin possesses the cytoplasmic tail in its mature form, we attempted to establish the functional assignment of this tail, using constructs of meprin ␤ with various mutations in this region (expressed in mammalian cells), and monitoring their properties and their localization. Here we show that the segment Tyr 679 to Asn 700 of the rat meprin ␤ is indispensable for its transport of from the endoplasmic reticulum to the Golgi complex. A meprin ␤ mutant, lacking the last 25 amino acids, is shown to be transport-incompetent, although it does not contain any of the known ER retention signals. It seems, therefore, that not only the two meprin subunits have different mechanisms of maturation and targeting, but that their carboxyl-terminal tail is involved in this process in a different manner. In the ␣-subunit, the tail is involved in its retention in the endoplasmic reticulum (16) to allow it to undergo the proteolytic cleavage(s) necessary for the export of this subunit to the cell exterior. The results presented here show that, in the case of the ␤-subunit, the carboxyl-terminal tail is essential for the exit of the enzyme from the ER and its subsequent transport to the Golgi apparatus.
Cell Cultures and Transfections-The human cell line HEK 293 and monkey Cos-7 cells were grown on 35 mm plastic tissue culture dishes (Nunc, Napperville, CT) in DMEM with 10% fetal calf serum and standard antibiotics. For transient transfection experiments, subconfluent cells were incubated for 5 h at 37°C with the precipitate prepared from 2 g of the plasmid DNA of interest using the LipofectAMI-NE TM reagent (Life Technologies, Inc.). Cells were analyzed for expression of the recombinant proteins 48 h after adding the precipitate. For the stable expression of the ␤⌬25 mutant, HEK 293 cells were transfected with a ␤⌬25/pcDNA3 construct, and clones were selected as described earlier (3).
Enzyme Activity Assay-The trypsin activation of the expressed meprin ␤ (including mutants), and the C-subunit degradation assay were performed as described earlier (3). Azocaseinolytic activity was measured according to Wolz and Bond (18), using the mock-transfected HEK 293 cells as a control.
Metabolic Labeling and Immunoprecipitation-The HEK 293 cells (transfected as above) were incubated before labeling (5-8 h) in methionine-free DMEM supplemented with 5% fetal calf serum. Cells were then pulse-labeled by incubation in the same medium (0.5 ml/35-mm tissue culture dish) containing 50 Ci/ml [ 35 -S]methionine (Amersham Pharmacia Biotech) for 15 min, and chased by the complete medium for the indicated time. Then the cells were lysed in a radioimmune precipitation buffer (19) containing proteinase inhibitors (Complete TM EDTAfree proteinase inhibitors mixture (Boehringer Mannheim, Germany)) for 30 min at 4°C with constant rocking. Lysates were cleared by centrifugation at 15,000 ϫ g for 15 min at 4°C and incubated with 1 l of the anti-meprin ␤ serum and 10 l of the Affi-Prep TM Protein A Support (Bio-Rad) for 1 h at 4°C. Precipitates were collected by centrifugation and washed three times in the radioimmune precipitation buffer before assay.
Deglycosylation Experiments-Protein samples for the deglycosylation experiments were denatured by boiling for 10 min in a 50 mM sodium acetate buffer, pH 5.5, containing 2% SDS and 10 mM ␤-mercaptoethanol, and divided into equal aliquots. These aliquots were diluted 5-fold with the buffer recommended by the glycosidase manufacturer, containing a proteinase inhibitors mixture (Complete TM EDTA-free mixture, Boehringer Mannheim, Germany). Samples were incubated overnight at 37°C with or without endo-␤-N-acetyglu-cosaminidase H (Endo H, Boehringer Mannheim, Germany), or N-Glycanase® (peptide-N-glycosidase F or PNGase F, Genzyme, Cambridge, MA) and analyzed by SDS-PAGE (20).
Kinetics of the Acquisition of Endo H Resistance-The HEK 293 cells were transiently transfected on 35-mm plastic tissue culture dishes as described above (each series of transfections included the wild type meprin ␤ and five different mutants). Cells then were plated on 24-well plate (Nunc) and metabolically labeled as described above, using a 5-min pulse of [ 35 S]methionine (20 Ci/well), then chased by complete medium for 0 -60 min at 37°C. At different time points, cells were lysed in the radioimmune precipitation buffer, and proteins were immunoprecipitated and treated with Endo H as described above. After resolving of the proteins on SDS-PAGE, the amount of the Endo H-resistant and sensitive forms of the expressed meprin ␤ or its mutants were measured by a PhosphoImager (Fujix Bas 1000, Fuji Foto Film, Ltd., Japan) using MacBas 2.0 software. The ratio between the Endo Hresistant and the sensitive forms for each protein was plotted as a function of time, and a slope was extracted from the linear curve fit. The slope values were designated as Endo H resistance and expressed as percent of that of the wild type meprin ␤.
Immunofluorescence Studies-Cell staining was performed essentially as described in Harlow and Lane (19). Cells growing on glass coverslips were rinsed twice with PBS before fixation in PBS containing 4% paraformaldehyde (fixing solution), and permeabilized (where specified) by treatment for 2 min with the fixing solution containing 0.1% Triton X-100. The cells were subsequently incubated with anti-meprin ␤ immune serum (diluted 1:500 in PBS), and with goat anti-rabbit TRITC conjugated antibodies (Sigma). Immunofluorescence was monitored using a fluorescence microscope (Nikon, Japan). Photographs were taken with a LIS-700 video camera (APPLItec, Israel).
Immunoelectron Microscopy-HEK 293/␤ (3) and 293/␤⌬25 cells were grown to confluence on 15-cm tissue culture dishes in DMEM, supplemented with 10% fetal calf serum, a standard mixture of antibiotics, and 0.2 mg/ml G418. Cells were fixed on the plates with a PBS buffer (containing 1% glutharaldehyde and 3% paraformaldehyde) for 1 h at room temperature, and the ultrathin frozen sections were prepared as described by Himmelhoch (21). The anti-meprin ␤ immune serum or normal rabbit serum (diluted 1:25), and goat anti-rabbit IgG (10-nm gold-conjugated, Zymed Laboratories Inc. Laboratories, San-Francisco, CA), were used for the detection of the immunocomplexes. Electron micrographs were taken at magnifications of 15,000 -34,000 with a Philips 410 electron microscope operated at 80 kV.
SDS-PAGE and Western Blotting-Protein samples were analyzed by 10% SDS-PAGE under reducing conditions, as described in the literature (20). The proteins were either stained by Coomassie Blue R-250 (Serva, Germany), or subjected to Western blotting according to Harlow and Lane (19). Typically, a dilution of 1:15,000 of anti-meprin ␤ immune serum was used. For visualization of the cross-reactive material, horseradish peroxidase-conjugated goat anti-rabbit antibodies (Transduction Laboratories, Lexington, KY), and an ECL detection system (Amersham, UK) were used.

Construction and Expression of the Meprin ␤ Mutants Truncated in the Cytoplasmic Domain-
In an attempt to assess the possible functional assignment of the cytoplasmic tail of KSMP/ meprin ␤ we prepared meprin ␤ mutants systematically truncated in their carboxyl-terminal part (see Scheme 1). A comparison of the carboxyl-terminal sequences of the human (PPH␤) and of rat meprin ␤ (9, 22) revealed three distinct segments within the tail having a different level of sequence identity, a highly conserved cluster of basic amino acids following the transmembrane domain, another conserved region in the last six-amino acid fragment of the tail, and an in-between segment with no homology. On this basis, stop codons were introduced by PCR into positions between these regions, and the mutant enzyme molecules were then subcloned into a pcDNA3 vector for expression in mammalian cells.
To test the general integrity of the cloned proteins, we translated them in vitro using the rabbit reticulocyte lysate system (Fig. 1A). By supplementing the reaction mixture with the purified canine microsomal membranes (CMM), and subsequent purification of the membranal fraction by centrifugation, we monitored the co-translational translocation of the proteins into the membrane. As shown in Fig. 1A, all the proteins exhibited an ability to translocate to the microsomal membrane. The increase in apparent molecular weight of the proteins translated in the presence of CMM, reflects posttranslational glycosylation by the microsomal membrane preparation, since this difference is eliminated upon treatment with PN-Gase F (the broad specificity glycosidase). We therefore con-cluded that the carboxyl-terminal truncations did not impede the initial maturation of the meprin ␤ mutants.
The constructs thus obtained were utilized to transfect HEK 293 fibroblasts, monitoring the expression of the cloned proteins with polyclonal antibodies raised against purified meprin ␤ (4). In an attempt to assess the effect of carboxyl-terminal truncations on the enzyme activity, we tested the proteolytic activities of nonactivated and of trypsin-activated wild type meprin ␤ and of its mutants on two previously characterized KSMP/meprin ␤ substrates, the catalytic subunit of PKA (3) and azocasein (18). The level of expression of each construct was determined by enzyme-linked immunosorbent assay with polyclonal antibodies raised against meprin ␤, using as a reference the meprin ␤ purified from the stably transfected cell line (described previously in Chestukhin et al. (4)). As shown in Table I, all the mutants were found to have a specific activity toward azocasein and toward the C-subunit of PKA, which was similar to that of the wild type meprin ␤, clearly showing that the carboxyl-terminal truncations caused no major misfolding in the catalytic domain of the enzyme.
Removal of 25, 15, or 10 Carboxyl-terminal Amino Acids Results in an Altered Intracellular Transport of Meprin ␤-To find out whether the carboxyl-terminal tail of KSMP/meprin ␤ is involved in the maturation of the enzyme, we compared the intracellular transport of meprin ␤ and mutants. Transiently expressed in 293 cells proteins were subjected to the [ 35 S]methionine pulse-chase labeling, followed by immunoprecipitation with anti-meprin ␤ antibodies (for details, see "Materials and Methods"). The precipitated proteins were then treated by Endo H and PNGase F in order to analyze their processing. Since Endo H is able to remove only the high mannose oligosaccharide side chains added to the proteins upon maturation in the endoplasmic reticulum, the appearance of the Endo H-resistant forms indicates the processing of the oligosaccharides by Golgi-associated enzymes. As shown in Fig. 1B, whereas most of the amount of the full-length meprin ␤ and ␤⌬5 mutant protein exhibited an Endo H-resistant glycosylation (this is evident from the appearance of slow migrating forms of the protein sensitive to the treatment with PNGase F), essentially no Endo H-resistant form of the ␤⌬25, ␤⌬15, and ␤⌬10 mutant proteins was accumulated under the same experimental conditions. The lack of complex glycosylation of the ␤⌬25, ␤⌬15, and ␤⌬10 proteins suggests that they are transport-incompetent, i.e. that they retained in the endoplasmic reticulum.
This result was further confirmed by immunocytochemical visualization of the distribution of the truncated and the nontruncated meprin ␤ within the cell. For indirect immunofluorescence studies we used transiently transfected Cos-7 cells (see "Materials and Methods"), since these cells were found to be more suitable than the HEK 293 fibroblasts for microscopy studies. The full-length meprin ␤ is detected on the surface of the nonpermeabilized cells (Fig. 2a), and the pattern of staining was essentially not changed upon permeabilization of the cell membrane before incubation with antibodies (Fig. 2b). The distribution of the ␤⌬5 mutant was similar to that of the wild type, but with the highlighted Golgi-like compact structure. In agreement with the Endo H assay results, we observed that the ␤⌬25, ␤⌬15, and ␤⌬10 mutant proteins were accumulated in the perinuclear zone of the cell exhibiting a characteristic localization in the endoplasmic reticulum (Fig. 2). Thus, the carboxyl-terminal truncation of more than 10 amino acids resulted in accumulation of meprin ␤ in the ER.
Stable Expression of the Meprin ␤ Mutant Lacking the 25 Carboxyl-terminal Amino Acids-To gain an insight into the possible reason for the retention of the meprin ␤⌬25 mutant in SCHEME 1. Structures of meprin ␤ (human and rat) and of the mutants of rat meprin ␤ truncated in the cytoplasmic tail which were prepared for the work described here. The domain structure of meprins is designated as follows: S, signal peptide; P, prosequence segment, catalytic domain; TM, transmembrane domain, CYT, cytoplasmic tail. The alignment of the carboxyl-terminal segments (the "tails") of the human and the rat meprin ␤ was performed using the Bestfit program (GCG, Madison, WI). Double dots indicate identical, and single dots, homolgous amino acid residues. The structure of the deletion (truncated mutants ␤⌬5, ␤⌬10, ␤⌬15, and ␤⌬25 is given.

FIG. 1. Glycosylation of meprin ␤ and its truncated mutants in vitro and in vivo.
A, cell-free expression of the full-length meprin ␤ and its truncated mutants. The proteins were translated in vitro in the presence or absence of CMM, using the rabbit reticulocyte lysate system described under "Materials and Methods." The membranal fraction (where present) was separated from the reaction mixture by centrifugation for 30 min at 22,000 ϫ g. The samples prepared from translation mixture without CMM, and the isolated CMM (intact or treated with PNGase F) were analyzed by SDS-PAGE (10% gel) (see "Materials and Methods" for details). Note the deglycosylation-sensitive increase in the molecular mass of the translation product when CMM is included in the reaction mixture. B, glycosylation of the truncated meprin ␤ mutants expressed in mammalian cells. Transiently transfected HEK 293 fibroblasts were labeled with [ 35 -S]methionine for 10 min, and after 1 h of chase in the complete medium the cells were lysed in the immunoprecipitation buffer containing 0.1% SDS. After immunoprecipitation by anti-meprin ␤ antibodies, the samples were denatured and incubated without (lanes "Ϫ") or with Endo H (lanes H), or PNGase F (lanes F) before analysis by SDS-PAGE followed by autoradiography of the dried gel.
the ER and to minimize the possible effect of overexpression, we prepared the HEK 293 fibroblasts cell line stably expressing this mutant (293/␤⌬25 cells). The meprin-␤⌬25 protein was expressed at approximately the same level as the wild type meprin ␤ in 293/␤ cells (Fig. 3A) and was found to be associated with the total membranal fraction of these cells. It was not present in the culture medium or the cytoplasm, thus exhibiting essentially the same distribution as the wild type enzyme (data not shown). The apparent molecular mass of the ␤⌬25 mutant expressed in 293 cells was calculated to be 78 kDa by SDS-PAGE (Fig. 3A). Such a 37-kDa decrease in molecular mass (from 115 kDa for the full-length meprin-␤) cannot be attributed merely to the removal of 25 amino acids. The susceptibility of the ␤⌬25 mutant to the Endo H treatment proved the absence of the complex glycosylation (Fig. 3A), which is in agreement with the results on the transiently expressed ␤⌬25 described above. The possibility that the molecular weight of the protein is reduced due to an additional deletion from the amino terminus was ruled out by NH 2 -terminal amino acid sequence of the meprin ␤ and of the ␤⌬25 mutant purified from the cell lines described above. In both cases, the analysis revealed an identical amino acid sequence which started from Leu 21 , the amino acid following the predicted signal peptide cleavage site (23).
Deletion of the Cytoplasmic Tail Does Not Affect the Life Span and Folding of Meprin ␤-It is now known that the ER can be regarded as a "quality control station" for newly synthesized proteins, rapidly disposing of those with incorrect folding (24). a Activity was measured in total 1% octyl-␤-D-glucopyranoside extracts of transfected HEK 293 cells according to Wolz and Bond (18). Proteolytic activity of the extract of mock-transfected cells was used as a background. Specific activity was calculated per mg of expressed meprin ␤ protein (quantitated by enzyme-linked immunosorbent assay using meprin ␤ purified from stably transfected HEK 293 cells as a reference). Values are mean of the two independent measurements. b C-degrading activity was measured as described previously (2,3). Briefly, 0.5 g of purified bovine catalytic subunit of PKA was incubated with 0.2 g of the total 1% octyl-␤-D-glucopyranoside extracts of transfected HEK 293 cells for different times. The samples then were subjected to SDS-PAGE and visualized with Coomassie Blue staining. Activity was determined by densitometric quantitation of the CЈ to C (clipped form) conversion according to equation: activity (%) ϭ CЈ C ϩ CЈ . One unit of the C-degrading activity was defined as amount of the enzyme degrading 1 g of the catalytic subunit of PKA/min at 22°C. C-degrading activity of the extracts without trypsin activation was below the level of detection. c ND, not detectable.

FIG. 2. Intracellular localization of meprin ␤ and its truncated mutants by immunofluorescent microscopy.
Transiently transfected Cos-7 cells were fixed (panel a), or fixed and permeabilized (panels b-f), before incubation with an anti-meprin ␤ serum. Cross-reactive material was visualized by staining with goat anti-rabbit TRITC-conjugated antibodies and photographed (for experimental details see "Materials and Methods"). Note the presence of cells with different levels of expression of the transfected construct in panels a and d, and the apparently nontransfected cells (dark shadows) in panels b and c.
Considering the possibility that the ␤⌬25 mutant may be improperly folded and therefore destined for degradation, we performed a long [ 35 S]methionine pulse-chase labeling experiment of the HEK 293/␤ and 293/␤⌬25 cells and monitored the content of both the full-length meprin ␤ and of the ␤⌬25 mutant by immunoprecipitation with anti-meprin ␤ antibodies. Since we did not observe any significant decrease in the amount of both proteins in the course of a 12-h chase (Fig. 3B), we concluded that the mutant protein does not have a significantly shorter life span. It should also be noted that, even after the 12-h chase, the molecular mass of meprin ␤⌬25 did not increase, whereas 1 h of chase was sufficient to achieve a complete glycosylation of meprin ␤, as evident from the increase of the apparent molecular mass from 97 to 115 kDa.
Since it is known that for oligomeric proteins the process of subunit assembly is tightly associated with the folding of its subunits, and their interaction with the chaperones, we attempted to compare the kinetics of dimerization of the fulllength meprin ␤ and the ␤⌬25 mutant. We monitored the formation of the dimer after the pulse-labeling of the 293/␤ and 293/␤⌬25 cells with [ 35 S]methionine followed by the chase for different times and immunoprecipitation by anti-meprin ␤ an-tibodies (for details, see "Materials and Methods"). The immunoprecipitated proteins were analyzed by SDS-PAGE (under nonreducing conditions) followed by autoradiography. As seen in Fig. 3C, no significant difference in the kinetics of accumulation of the dimer form of the protein was observed, suggesting that the truncation of the tail did not affect the overall folding of the mutant protein. Interestingly, the dimer form of both wild type and truncated meprin ␤ migrates on the gel as two bands. A possible explanation for this observation might be that meprin undergoes core glycosylation at the same time as it forms the disulfide-linked dimers (as previously shown for human meprin by Sterchi et al. (25)), and that the two bands seen in Fig. 3C may represent differently glycosylated forms of the meprins.
Immunoelectron Microscopy of the Meprin ␤and ␤⌬25-expressing Cells-Using the HEK fibroblast cell lines expressing the intact (293/␤) and the truncated (293/␤⌬25) we could establish and compare the intracellular localization of these proteins by immunoelectron microscopy. Immunogold-conjugated goat anti-rabbit antibodies were used for visualization of the proteins cross-reacting with anti-meprin ␤ polyclonal antibodies. The 293/␤ cells exhibited an intense immunostaining of the plasma membrane (Fig. 4, panel 1A), while in the cells expressing the mutant protein, only occasional staining of the outer plasma membrane structures was observed (not shown). As seen in Fig. 4, panel 2A, there is a significant accumulation of the ␤⌬25 mutant protein in the ER tubules and in the outer nuclear membrane. It also seems to us that there is an indication for differential staining of the endoplasmic reticulum subcompartments in the ␤⌬25 expressing cells: immunoreactive material was preferentially found in the structures which appear to represent the rough ER (tubules on Fig. 4, panel 2A), whereas smooth ER (vesicular structures in the same photograph) were stained much less efficiently. This result might indicate that the mutant protein is recognized as a transportincompetent protein already in this early stage of maturation. Interestingly, besides the expected localization on the plasma membrane, meprin ␤ immunoreactivity was also associated with the large (0.2-0.5 m) vacuolar structures (Fig. 4, panel  1A). Considering the size of those vacuoles, we believe that they may represent either a kind of endocytotic vesicle or structures formed as a result of the fusion of a number of the transport vesicles derived from the Golgi apparatus. Such structures were not observed in the ultrathin sections (prepared following the same procedure) of 293/␤⌬25 cells, or of the HEK 293 fibroblasts transfected with pcDNA3 vector and selected by G418 resistance (not shown). Therefore we believe that their appearance is associated with the expression of meprin-␤ in this cells. The good agreement between the results obtained with Cos-7 and HEK 293 cells indicates that the difference in cellular distribution of the full-length and mutant meprin ␤ is independent of the cell line and the method of transfection used.
Evidence Suggesting That No Specific Amino Acid Residue in the Tail of Meprin ␤ Is Imperative for the Successful ER to Golgi Transport-From a comparison of the maturation of the meprin ␤⌬5 and ␤⌬10 mutants, it appeared that the segment 695 AMNLE 699 in the cytoplasmic tail, which is chopped off in ␤⌬10 but not in ␤⌬5 (see Scheme 1 and Fig. 5A) may accommodate at least an important constituent of the signal sequence necessary for the exit of this protein from the ER. In an attempt to identify the specific amino acids which might constitute such a sorting signal, we introduced a series of mutations in this region by replacing the wild type amino acid residues by ala-nine (Fig. 5B), and then tested the kinetics of maturation of the mutant proteins transiently expressed in HEK 293 cells. In this set of experiments, maturation was monitored by the ratio between Endo H-resistant and -sensitive forms of the mutants (taking the wild type meprin ␤ as a reference in each series of experiments) with time (0 -60 min). The bars in Fig. 5 represent the slopes calculated from a linear curve fitting of the results and expressed as percent of the value obtained in the same experiment with the wild type meprin ␤. As evident from Fig. 5B, alanine substitutions in the 695 AMNLE 699 region caused no dramatic effect on the ability of the expressed mutant to translocate from the ER to the Golgi (see mutants M696A/N697A and L698A/E699A), although the mutant M696A/N697A exhibited a slightly decreased rate of acquisition of an Endo H resistance.
In view of this finding we considered the possibility that a putative sorting signal may be located upstream of Ala 694 , and that the ␤⌬10 truncation may render meprin ␤ transport-incompetent by placing this signal too close to the COOH terminus. Therefore we also tested other mutants in the segment between the Tyr 679 and Ala 694 . As seen in Fig. 5B we found a ϳ25% decrease in the translocation rate in the R686A/K687A/ K688A mutant. This effect was even more pronounced when all the amino acid residues creating the cluster of basics in the tail (as in the RRK,RKK to Ala mutant) were substituted with alanine. Replacement by alanine of the Tyr 685 residue in the middle of the cluster of basics (a cluster conserved also in the human meprin ␤; Scheme 1), also resulted in no impedance the ER to Golgi transport. It should be noted that in the case of the R682A/R683A/K684A substitution, the mutant protein was actually able to undergo glycosylation even faster than the wild type meprin ␤ as judged by the Endo H resistance. The reason for this apparent acceleration is not clear to us at this stage, but from the mechanistic point of view it is quite intriguing. It should be noted, however, that most of the mutants in which charged amino acids were replaced by the nonpolar alanine residue (K692A/T693A, L698A/E699A, and R682A/R683A/ K684A) exhibited a somewhat increased rate in the ER to Golgi translocation. In conclusion it can be said that this series of experiments indicate that the substitution of the any of the tested amino acid residues in the cytosolic tail of meprin ␤ does not dominantly affect the transport of this protein from the ER to the Golgi apparatus.
A very intriguing observation we made in the course of these experiments was with the ⌬Ala 689 -Met 696 mutant (Fig. 5A). As a result of this mutation, the rate of the ER to Golgi translocation of the protein was decreased significantly (40% of the wild type), although the deleted amino acid residues by themselves were found to be nonessential for transport by alanine scanning. Therefore, we considered the possibility that the retention of the ␤⌬15 and ␤⌬10 mutants might be due to a recognition of the 687 KK 688 segment as an ER-retrieval signal as reported for several proteins that are known to reside in the endoplasmic reticulum (26,27). To test this possibility, we constructed two mutants in which the segment accommodating the last 10 amino acid residues (which is truncated in ␤⌬10), or the segment Ala 689 -Met 696 , were deleted alongside with the substitution of the 686 RKK 688 by alanine residues. By analysis of the kinetics of Endo H resistance, we found that, indeed, in those mutants, the same deletion and truncation that caused an impaired transport, resulted now in no ER retention (Fig. 6). In both cases, the ability to undergo complex glycosylation was reconstituted, showing that, although the dilysine motif is localized in an unfavorable position in relation to the carboxylterminal tail (KK(X) 6 and KK(X) 8 in meprin ␤⌬10 and ⌬Ala 689 -Met 696 mutants, respectively), it is able to cause retrieval of these mutants from the ER, therefore preventing them from further transport to the plasma membrane.
Possible Contribution of the Secondary Structure of the Tail and Its Flexibility to the Successful ER to Golgi Transport-Since the alanine substitutions of the amino acids in the tail which were considered likely to participate in protein-protein interaction (charged, polar or hydrophobic residues), did not influence significantly the ER to Golgi rate of translocation, we attempted to find out whether a mutation which is known to affect the local secondary structure would affect this translocation. Indeed, as seen in Fig. 5C a specific single amino acid mutation (Y685P, but not Y685A) in the middle of the cluster of basic amino acids resulted in a dramatic decrease in the rate of exit from the ER, reducing the rate of the translocation to essentially that of the ␤⌬25 mutant. The structural importance of this exact position in the sequence was indicated (i) by the finding that mutation to proline of its two adjacent amino acids (K684P and R686P) also caused a significant delay (about 50% of the wild type) in the ER to Golgi transfer, but did not essentially abolish the transport completely as found in case of the Y685P mutant, and (ii) by the fact that the introduction of this mutation at the position of two other alanines in the tail (A691P and A695P) did not result in a significant decrease in the rate of the translocation. It should be noted that the clearcut effect of the Y685P mutation on the transfer could not be attributed to an interference with the dimerization of meprin ␤ since the rate of dimerization of this mutant protein was found to be essentially identical to that of the wild type meprin ␤ (data not shown).
On the basis of the results presented above we conclude that the introduction of the kink forming proline residue at position 685 or at its immediate vicinity alters the secondary structure and maybe also the flexibility of the tail, and that such alterations interfere with the ER-to-Golgi transfer. This conclusion is in line with the analysis of the alanine and proline substitutions in this region, which show that while alanine substitution of the one (Y685A), three (R682A/R683A/K684A and R686A/K687A/K688A) and even six (RRK,RKK to Ala) amino acids can be tolerated, single-site mutations such as K684P, R686P, and especially Y685P affect the transport dramatically. In line with this proposal we would like to point out that the prediction of the secondary structure of the cytoplasmic tail by the method of Chou and Fasman (28) (Pepplot software, GCG, Madison, WI) suggests a high probability for the existence of an ␣-helix in the tail of meprin ␤ that is reduced significantly upon introduction of a helix-breaking proline residue in the sequence of the tail. DISCUSSION Until recently it was commonly accepted that the transport of proteins to the plasma membrane through the ER to Golgi vesicular pathway occurs "by default," or by "bulk flow." However, it now seems that there are diverse routes of transport, and that proteins may be specifically sorted before they are targeted to the appropriate membranal compartment or organelle (for a recent review see Kuehn and Schekman (29)). In addition, it was recently shown that membranal, as well as secreted, proteins are first concentrated in the ER before being transferred to the Golgi apparatus (30,31). This finding raises the possibility that there may be specific receptors for cargo proteins that selectively anchor to these molecules upon interaction with appropriate transport signals. For example, the Emp24 protein, a known component of the COPII-coated vesicles, was shown to function as such a receptor in yeast cells (32). A family of structurally related proteins (the p24 family), has already been identified, indicating that the occurrence of such specific receptors may well be more widespread than previously thought (33). Many of the transport studies were focused on the protein traffic through the early secretory pathway, namely from the ER to Golgi apparatus and inside the Golgi apparatus. It is now believed that this transport is mediated by the coatomer complexes COPI (participating both in the anterograde and retrograde transport of vesicles between the ER and Golgi apparatus) (34), and COPII (primarily involved into the anterograde vesicular flow) (35). Several anter- FIG. 6. Analysis of the role of the RKK motif in the ER retention of the meprin ␤ "short-tail" mutants. A, the kinetics of the Endo H resistance was monitored in the same experiment for meprin ␤, ␤⌬10, and ⌬Ala 689 -Met 696 mutants, and two mutants with the same truncation and deletion in the cytoplasmic tail but also carrying the R686A/K687A/K688A substitution. Meprin ␤ and mutants were transiently expressed in HEK 293 cells, metabolically labeled with [ 35 S]methionine for 5 min, and chased in a complete culture medium for the indicated time before immunoprecipitation and the Endo H resistance assay. The formation of the Endo H-resistant (black triangle) and sensitive (empty triangle) forms of meprin ␤ and its mutants was monitored after SDS-PAGE separation of the immunoprecipitated proteins and autoradiography of the dried gel. The structure of the cytoplasmic tail of the mutants is shown in B.
ograde (ER to Golgi) transport signals have been identified in the cytoplasmic domains of membranal proteins, for example the conserved diphenylalanine motif in members of the p24 family of putative cargo receptors (36), and the diacidic signal (DXE, X represents any amino acid) localized at carboxyl-terminal side of a YXX⌽ motif found in several membranal proteins (X is any amino acid, ⌽ represents a bulky hydrophobic residue) (37). Despite these extensive studies, it seems that the nature of the mechanisms involved in intracellular traffic are more complicated than it was initially thought and that there is a need to expand such studies on a variety of proteins and of cells before we can establish the traffic laws and their general validity.
Here we report an analysis of the maturation and the intracellular traffic of rat meprin ␤ expressed in HEK 293 fibroblasts and Cos-7 cells. One of the major messages of this report is that the cytoplasmic tail of the ␤-subunit of meprin (which does not contain any of the known anterograde transport signals mentioned above) is indispensable for transferring it from the ER to the Golgi apparatus. This is based on the following observations: (i) that deletion of this carboxyl-terminal tail (mutation ␤⌬25) traps the mutant protein in the ER; (ii) that this truncation does not lead to the exposure of any of the known ER retention signals (KKXX, KXKXX or K(or H)DEL (X is any amino acid) (26) in the newly formed carboxyl terminus; and (iii) that the impeded exit of the truncated mutant from the ER is probably not due to its being a misfolded protein, since the truncated mutant is shown to have essentially the same stability, dimerization rate, and catalytic activity as the wild type protein.
Marchand et al. (15) proposed that the retention of the ␣-subunit of meprin in the ER occurs by an interaction between a motif of repeating glycine residues in the transmembrane domain of meprin ␣ with a similar motif found in some ER resident protein(s). Since such a motif of repeating glycine residues is not found in meprin ␤, this mechanism cannot account for the retention in the ER of the meprin ␤⌬25 mutant. Moreover, the ␣/␤ tail-switch mutant (without the I domain) which was described in the work of Marchand et al. (15), was not retained in ER and was found to undergo a complex glycosylation. This finding indicates that there is no retention signal in the ␤-subunit region accommodating epidermal growth factor-like, transmembranal, and cytoplasmic domains.
The effects of the systematic deletions and substitutions in the cytoplasmic tail on the ER to Golgi translocation of meprin ␤ might be summarized as follows: (i) the truncation of the 25 carboxyl-terminal amino acid residues, and the single site mutation Y685P (but not Y685A) result in a complete blocking of this process; (ii) substitution of the basic amino acids stretch ( 682 RRKYRKK 688 ) in the vicinity of the transmembranal domain to alanines results in a decreased rate of ER to Golgi translocation; and (iii) the amino acid residues in the segment 689 ASAKTAMNLENQHAF 704 do not contribute significantly to the successful ER to Golgi transport. It appears, therefore, that the tail length and perhaps its secondary structure in the region proximal to the transmembrane domain are more important for the successful transport than a specific amino acid sequence. The fact that a single-site mutation (Y685P) blocks the export of this mutant molecule from the ER is very impressive in that respect, since the tail length is not changed in this case, but the secondary structure and flexibility of the cytoplasmic tail might be altered. The possibility to entrap a protein by the introduction of a conformational alteration such as a kinkforming proline at a given position in the sequence may shed light from a new angle on the physiological traffic of proteins within the cell and possibly on aberrated transport pathways leading to pathological traffic jams. It may indicate, for example, that the interaction of a putative sorting or transport protein with the tail of meprin ␤ is particularly sensitive to the conformation of the juxtamembranal region accommodating the cluster of basic amino acids. Another possibility that we cannot neglect at this stage is that the putative sorting signal may be composed of a unique combination of distant amino acids in the sequence. An extensive and complex, pairwise scanning of mutations in the tail may be required to unveil this signal.
The results reported here indicate that in the case of meprin ␤, the carboxyl-terminal tail plays a role which is actually opposite to the role of the corresponding region in meprin ␣, specifically we propose that in meprin ␤ the tail is necessary for the export of the newly synthesized enzyme molecule from the ER to Golgi apparatus, possibly through an interaction with one or several cargo receptor proteins. The question of why these two structurally similar subunits have such different mechanisms for their intracellular transport and maturation is not clear at this stage, but it seems to us that our findings shed light on the trafficking of meprin ␣ and meprin ␤ in a way that may have a general mechanistic significance.