The Release of Acetylcholine Receptor Inducing Activity (ARIA) from Its Transmembrane Precursor in Transfected Fibroblasts*

Acetylcholine receptor inducing activity (ARIA) is made by motoneurons and is released at the neuromuscular synapse to stimulate the synthesis of acetylcholine receptors by skeletal muscle. ARIA is derived from a transmembrane precursor (pro-ARIA) via proteolytic cleavage of the ectodomain. We studied requirements in the amino acid sequence at the cleavage site with various substitution and deletion mutations. Wild type (WT) and mutant proteins were transiently expressed in COS cells, and release of ARIA into the conditioned medium was measured by tyrosine phosphorylation of its receptor, p185, in L6 cells. Removal of all potential cleavage sites between the extracellular epidermal growth factor domain and the transmembrane domain by substitution and small deletions (<11 amino acid residues out of 21) did not significantly reduce ARIA release, whereas larger deletions abolished it. We propose that cleavage occurs independently of amino acid sequence at a short distance from the epidermal growth factor domain, unless sterically hindered by the nearby secondary structure. A mutant with shorter cytoplasmic domain (“c” isoform) released significantly less ARIA than the WT (“a” isoform), suggesting that the c isoform may be suitable for signaling through direct cell-cell contact. Alternatively, proteolytic conversion of the a isoform to the c isoform may rapidly down-regulate release of ARIA.

via proteolytic cleavage of the ectodomain. Considering that an EGF-like domain, located in the membrane proximal half of the extracellular segment, is crucial for all biological actions of NRGs (reviewed in Ref. 7), proteolytic cleavage of pro-ARIA must occur between the extracellular EGF-like domain and the transmembrane domain in order to release functional ARIA. The C-terminal end of the EGF-like domain can be functionally defined as Met 185 (see Fig. 1a), because a synthetic NRG ending with this residue was fully functional, and this methionine was critical for receptor activation (8).
In transfected CHO cells, significant portion of pro-ARIA is expressed at the cell surface before it is cleaved to release ARIA (9,10). Cleavage of pro-ARIA appears to involve a protein kinase C-dependent step because it can be stimulated by treatment of the cells with phorbol 12-myristate 13-acetate (PMA), a protein kinase C activator (9,10). However, the nature of the processing enzyme remains elusive.
It is clear that NRG plays an important role at the developing and mature nerve-muscle synapses. The number of postsynaptic acetylcholine receptors is reduced in heterozygous NRG knockout (NRGϩ/Ϫ) mice (11). Homozygous knockout mice (NRGϪ/Ϫ) die by embryonic day 10 before neuromuscular junction develops (12,13). It is not clear if this action of NRG depends on an action of cleaved and "soluble" NRG or if the transmembrane form is biologically active. In certain cases, transmembrane form of growth factors are not simply precursors for the soluble counterparts, but they can signal through direct cell-cell contact (juxtacrine signaling) (14 -18). In the case of TNF␣, juxtacrine signaling mediated by pro-TNF␣ may mediate local inflammatory responses, whereas systemic signaling through soluble TNF␣ may be responsible for septic shock or cachexia (14). Therefore, proteolytic cleavage of the transmembrane form of these molecules may be viewed as an important regulatory mechanism in deciding between signaling at a distance (paracrine) versus signaling through contact (juxtacrine) rather than simply regulating the amount of soluble factors.
In this paper, we examine the constraints on the amino acid sequence of pro-ARIA at the cleavage site. We also examine potential differences between different isoforms of pro-ARIA in release of soluble ARIA.

EXPERIMENTAL PROCEDURES
Mutagenesis of Pro-ARIA-Plasmid DNA derived from pcDNAI (Invitrogen, Carlsbad, CA) containing full-length cDNA for pro-ARIA has been described earlier (2). Mutagenesis of the pro-ARIA was performed by Kunkel's method (19) using the Mutagen kit (Bio-Rad). Mutagenic oligonucleotides often contained an additional silent mutation to create a restriction site, which was used for screening of the mutants. All mutations and lack of additional unwanted mutation were confirmed by sequencing the entire coding region of pro-ARIA. Plasmid DNA was purified with a Qiagen kit (Qiagen Inc., Santa Clarita, CA) before sequencing and transfection into COS cells.
Transient Transfection of COS Cells and Analysis of Total Cellular Expression-COS cells were maintained in Dulbecco's modified Eagle's medium (DMEM; Life Technologies, Inc.) supplemented with 10% fetal calf serum, L-glutamine, penicillin, and streptomycin (DMEMϩ). Cells were split 1 day before transfection into 24-well plates at a density of 20 -30% confluence. On the day of transfection, 400 ng of plasmid DNA was used to transfect cells in 1 well using LipofectAMINE (Life Technologies, Inc.) according to manufacturer's recommendation. Briefly, the cells were incubated with DNA/LipofectAMINE mixture for 6 h in Opti-MEM (Life Technologies, Inc.) before fetal calf serum was added to 10%. The following day, cells were washed twice with Opti-MEM and incubated for 10 h with Opti-MEM containing 10% fetal calf serum to allow recovery of the cells. To collect conditioned medium for measurement of steady-state release of ARIA, medium was replaced with Opti-MEM containing 2 mM CaCl 2 and further incubated for 48 h. Conditioned medium was filtered through 0.22-m filter to remove floating cells and stored frozen until p185 phosphorylation assay. After collecting conditioned medium, total cell lysates were obtained by adding 100 l of 2ϫ strength SDS-PAGE sample buffer to each well. Total cellular expression of precursor proteins was examined by Western blotting of the cell lysates. Proteins were separated in a 7.5% polyacrylamide gel and transferred to Immobilon-P membrane (Millipore, Bedford, MA). Immunoblotting was performed according to standard procedures. Membrane was blocked for 1 h in 5% nonfat dry milk solution prepared in PBS, incubated with primary antibody for 1 h in blocking solution, washed three times in PBS for 5 min each, incubated for 1 h with horseradish peroxidase-conjugated goat anti-rabbit antibody in blocking solution, and then washed three times in PBS for 5 min each. Proteins were visualized with Renaissance chemiluminescence reagent (NEN Life Science Products) according to the manufacturer's protocol.
p185 Phosphorylation Assay-ARIA was assayed by tyrosine phosphorylation of p185 in L6 muscle cells basically as described (20) with minor modifications. L6 cells were grown in DMEMϩ. For p185 phosphorylation assay, cells were seeded into 48-well plates (Costar, Acton, MA) and grown in DMEMϩ until they became confluent and fused. Cells became confluent in 4 -5 days and began to fuse. In about 10 days after plating, cells fused both laterally and longitudinally forming a large sheet of multinucleated cells, at which time response to ARIA reached maximum sensitivity. Cells maintained maximum responsiveness for about 4 -5 days. All the p185 phosphorylation assays were done during this period. After aspirating the growth medium, assay medium was added in 200 l volume per well and incubated for 45 min at 37°C. At the end of incubation, assay medium was aspirated, and 50 l of preheated 2ϫ strength SDS-PAGE sample buffer was added to each well. Cell lysates were collected into microcentrifuge tubes, heated at 100°C for 5 min, and stored at Ϫ20°C until analysis by Western blotting. Samples were run in a 5% gel until the 130-kDa prestained molecular mass marker (Bio-Rad) reached the bottom of the gel for best separation of the p185 band from non-responding lower molecular weight bands. The rest of the Western blotting was done similarly as described above except that 4G10 antibody (0.3 g/ml final) was used as a primary antibody in 3% nonfat dry milk, and horseradish peroxidaseconjugated goat anti-mouse antibody (Dako, Carpinteria, CA) was used as a secondary antibody. 4G10 antibody was a gift from Dr. T. Roberts (Dana Farber Cancer Institute, Boston).
Quantitative Comparison of Released ARIA and Precursor Expression-To compare amounts of ARIA in the conditioned media from cells transfected with mutant constructs with that from wild type (WT)transfected cells, serial dilutions were made for each conditioned media. p185 phosphorylation assay was performed simultaneously for each of the serially diluted media. L6 cell lysates were run in the same gel for all the samples to be compared. X-ray film containing the Western blotting results was scanned with a flat bed scanner in transmissive mode at 600 dpi, 256 gray scale. The resulting digitized image was imported into ImageQuant software (Molecular Dynamics, Sunnyvale, CA). Band intensity of the p185 from different samples was measured by drawing a rectangular box of identical size around each p185 band and measuring integrated intensity using built-in volume-reporter function. A calibration curve was generated using p185 band intensity by each of the serial dilutions of the WT conditioned medium. The amount of ARIA in undiluted medium was calculated for each mutant using the calibration curve by linear regression. To compare total cellular expression of the precursor protein for each mutant with that of the WT, serial dilutions were made for each cell lysate in SDS-PAGE sample buffer. Western blotting and quantitation was performed in a similar manner as above. Data for qualitative and quantitative comparisons were obtained from separate transfection experiments.
PMA Stimulation of ARIA Release-Transfection of COS cells was performed similarly as described above with minor modifications. Cells were transfected in a 100-mm dish and incubated with DMEMϩ for 24 h after recovery from transfection. On the day of PMA stimulation, cells were washed once and incubated for 1 h with prewarmed DMEM containing 1 mg/ml bovine serum albumin. After collecting the medium for basal release of ARIA, cells were washed once and incubated under the same condition with an additional 100 nM PMA for stimulated release. Collected conditioned media were filtered through a 0.22-m filter to remove floating cells, then concentrated 20-fold using Centricon-10 (Millipore) before serial dilutions were made, and p185 phosphorylation activity was assayed. PMA was initially made in Me 2 SO at 1 mM concentration. Control experiments with purified NRG showed that 100 nM PMA did not interfere with the p185 phosphorylation assay, although a 10-fold higher concentration significantly reduced the phosphorylation response.
Immunocytochemistry-Two days after transfection, cells were harvested by trypsinization and reseeded into 8-well glass chamber slides (Nunc, Naperville, IL) precoated with laminin. Immunolabeling was performed on the following day. For labeling of cell surface pro-ARIA, cells were washed and prechilled with cold DMEM before they were incubated with 183 antiserum for 1.5 h on ice. Antiserum 183 was raised against the whole pro-ARIA. Cells were washed with cold DMEM twice, fixed in 4% paraformaldehyde for 15 min, and washed thoroughly with PBS. Fixed cells were then incubated with Cy3-conjugated goat anti-rabbit antibody (Molecular Probe, Eugene, OR) for 1 h at room temperature and washed thoroughly with PBS. Antibody 1310, directed against the cytoplasmic domain of pro-ARIA, gave no detectable staining under the same condition. For labeling of total cellular pro-ARIA, cells were first fixed in 4% paraformaldehyde for 15 min and then permeabilized in 0.2% Triton X-100 for 2 min. Primary and secondary antibodies were diluted in PBS containing 10% normal goat serum. Antibody 1310 was used as a primary antibody for labeling of total cellular pro-ARIA. Even when permeabilization with Triton X-100 was omitted, fixation with paraformaldehyde alone apparently allowed access of primary antibody to the intracellular space because clear staining of transfected cells was visible. Slides were mounted with 5% n-propyl gallate in 50% glycerol and then viewed and photographed under rhodamine filter.
Effect of Metal Chelators on ARIA Release-CHO cells stably expressing WT pro-ARIA have been described earlier and were maintained accordingly (10). Cells grown in 12-well plates to near confluence were washed once and incubated for 30 min at 37°C with DMEM containing different treatment reagents. Conditioned medium was centrifuged to remove floating cells, and then treatment reagents were added to adjust for differences. Thus, all conditioned media contained equal concentrations of EDTA, o-phenanthroline (OP), PMA, ZnCl 2 , and CaCl 2 when p185 phosphorylation assay was performed to measure released ARIA. OP was initially made as 5 M concentration in EtOH.

Cleavage of Pro-ARIA Occurs between Extracellular EGF
Domain and the Putative Transmembrane Domain-Cleavage of pro-ARIA must occur between the C-terminal end of the EGF domain (Met 185 ) and the transmembrane domain to release functional ARIA. The transmembrane domain was predicted to begin with Val 207 based on the hydropathy analysis and the presence of adjacent charged residues (see Ref. 2). We call this region including Ala 186 and Arg 206 the "stalk." Deletion of a large portion of the stalk (Ͼ15 amino acid residues out of 21) prevented release of functional ARIA into the medium without affecting expression of the precursor proteins ( Figs. 1 and 2).
WT and mutant proteins were transiently expressed in COS cells, and the ARIA released into conditioned medium was assayed by its ability to phosphorylate tyrosine residues on a p185 in L6 muscle cells (see "Experimental Procedures"). p185 is a broad band containing NRG receptors in extracts of L6 cells that migrates with an apparent molecular mass of 185 kDa. The intensity of p185 band on Western blot was concentrationdependent over a wide range of NRG concentration. We routinely detected NRG at levels as low as 10 pM using this assay, whereas the response did not saturate until 1 nM concentration (see Fig. 7). ARIA in the conditioned medium could not be directly detected by Western blotting method using currently available antibodies, probably due to lower sensitivity or lack of cross-reactivity between species.
Conditioned medium from the WT-transfected cells con-tained high levels of ARIA as measured by p185 phosphorylation activity, whereas those from mock-transfected cells had no activity (Fig. 1b). A doublet of slightly lower molecular mass (about 150 kDa) that did not respond to ARIA served as an internal control for loading of equal amount of proteins. ARIA from the WT conditioned medium was readily detectable even when the conditioned medium was diluted 10-fold (data not shown). On the contrary, conditioned media from cells transfected with large deletion mutants (delA, delB, and delC in Fig.  1a) contained no detectable phosphorylation activity (Fig. 1b), indicating that these mutants were defective in release of functional ARIA. We considered four explanations for lack of functional ARIA from cells transfected with these mutants. First, the mutant proteins might not be expressed in the transfected COS cells. When WT-transfected cell lysates were examined by Western blotting method, two sets of broad bands (each migrating similarly with 83-and 51-kDa molecular mass markers) were observed using an antibody directed against the cytoplasmic domain of pro-ARIA (antibody (Ab) 1310 in Fig. 1c). Only the upper set of bands was recognized by an antibody directed against the extracellular Ig domain (antibody (Ab) HM94 in Fig. 1c). This upper set of bands is most likely the full-length transmembrane precursor (bracket in Fig. 1c) because it contains both extracellular and cytoplasmic epitopes. Heterogeneity in the size of the bands is probably due to variable glycosylation of the protein. The remnant, defined as the transmembrane plus cytoplasmic domain after cleavage of the . Dots indicate identical residues and dashes indicate deletions. The diagram also shows the epitopes for some of the antibodies used in this study. Affinity purified antibody HM94 is directed against a peptide corresponding to a segment within the Ig domain (10). Affinity purified antibody 1310 is directed against a peptide corresponding to the C-terminal end of the c isoform (see Fig. 6) (9). Affinity purified antibody SC348, commercially available from Santa Cruz Biotechnology (Santa Cruz, CA) is directed against a peptide corresponding to the C-terminal end of the WT pro-ARIA (a isoform). Antiserum 183, not shown in this diagram, was raised against the whole pro-ARIA expressed in bacteria (44). b, tyrosine phosphorylation of the p185 from L6 muscle cells by the medium conditioned with the COS cells transiently transfected with WT and different mutant constructs as indicated. Western blotting was done with anti-phosphotyrosine antibody. ectodomain, is expected to run as a single band with a molecular mass of 44 -46 kDa, depending on the exact cleavage site within the stalk. Among the lower set of bands detected by the 1310 antibody but not by the HM94 antibody, a prominent band with an apparent molecular mass of 47 kDa was assigned as the remnant (open arrowhead in Fig. 1c). Other bands may correspond to partial degradation products. A significant level of expression was detected for the precursors of all three mutants (bracket in Fig. 1c, right panel) when expression was measured in the same cells from which conditioned media were collected. One of the mutants (delB) actually showed a higher level of expression than the WT. Therefore, lack of ARIA in the conditioned media from the mutant-transfected cells was not due to lack of expression of the precursor proteins.
Second, the mutant proteins might be expressed but not delivered to the cell surface, perhaps due to disruption of a necessary signaling motif. In stably transfected CHO cells, a significant portion of pro-NRG reaches the cell surface before it is cleaved (9,10). We examined the cell surface expression of these proteins following transient transfection into COS cells. Unfixed, transfected COS cells were incubated with antiserum 183, raised against the whole pro-ARIA, without permeabilization to prevent access of the primary antibody to the intracellular space. Incubation was done on ice to prevent internalization of the antibody. All of the mutants (delA, delB, and delC) as well as the WT pro-ARIA were readily detected at the cell surface ( Fig. 2) upon examination of the labeled cells with fluorescence microscopy. The pattern of labeling was consistent with localization at the cell surface and was easily distinguishable from cells that were fixed and permeabilized before incubation with the primary antibody (Fig. 2, top panels). Labeling of permeabilized cells showed staining of a meshwork structure consistent with labeling of the trans-Golgi network (Fig. 2, WT-total). Occasionally, a bright perinuclear staining was visible. On the contrary, cell surface labeling gave numerous small punctate staining patches with the brightest staining at the cell periphery. The pattern of cell surface labeling showed no meshwork but was consistent with the three-dimensional contour of the cell surface. Cotransfection experiments with green fluorescent protein confirmed that those cells that were labeled with pro-ARIA antibody were the cells that received plasmid DNA during transfection (data not shown). There was no noticeable difference in the intensity and frequency of labeled cells among the mutants and the WT. Therefore, lack of ARIA in the conditioned media from the mutant-transfected cells was not due to lack of cell surface expression.
Third, cleavage might have occurred within the EGF domain of the mutants and destroyed the activity of ARIA. Fourth, cleavage of the ectodomain may have been impaired for the mutant proteins either because the cleavage site was eliminated or because of other structural constraints. The results of the Western blotting analysis were more consistent with lack of ectodomain cleavage for these mutants than with destruction of activity by cleavage within the EGF domain. The remnant band, clearly seen from the WT-transfected cell lysate (open arrowhead in Fig. 1c), was almost completely missing in the delC-transfected cell lysate. Other large deletion mutants also showed marked reduction in the intensity of this band.
Small Deletions and Substitutions Did Not Affect Release of ARIA-We attempted to localize the precise cleavage site within the stalk. First, we examined if the pair of basic residues at the juxtamembrane position (Lys 205 -Arg 206 ) was a major cleavage site. Many prohormones and proproteins contain pairs of basic amino acid residues that are cleaved by subtilisin-like proteases (21)(22)(23). To eliminate the potential cleavage site while preserving the putative boundary between the trans-membrane and the extracellular domain, we simultaneously substituted Lys 205 to Glu and Arg 206 to Asp (KR/ED mutant; Fig. 3a). The amount of ARIA released into the medium from KR/ED-transfected cells showed no decrease when compared with WT (Fig. 3b, upper panel). Total cellular expression level of the precursor proteins was similar between WT and the KR/ED mutant (Fig. 3b, lower panel). Therefore, elimination of dibasic residues did not significantly reduce release of ARIA.
Quantitative comparison of the released ARIA and total cellular expression of the precursor between the WT and the KR/ED mutant confirmed this conclusion (Fig. 3d). To quantitate released ARIA, serial dilutions were made for conditioned media, and the p185 phosphorylation activity was assayed for each of them. The amount of ARIA in the conditioned medium for the mutant was quantified using a calibration curve generated by serial dilutions of the WT conditioned medium and, therefore, was expressed as relative to the WT (see "Experi- mental Procedures"). Cellular expression levels of the precursors were also quantified in a similar manner using serial dilutions of the cell lysates. In this transfection experiment, the amount of released ARIA for the KR/ED mutant was reduced to half of WT, but this was explained adequately by the reduced expression of the precursor (Fig. 3d). The ratio of the released ARIA to precursor expression represents efficiency of ARIA release normalized for protein expression. The efficiency of ARIA release thus defined was not decreased for the KR/ED mutant compared with the WT (solid bar in Fig. 3d). We conclude that the juxtamembrane dibasic residues are not a major cleavage site for pro-ARIA.
Next, we made a series of small (4 -6 amino acid residues) deletions to cover the rest of the entire stalk (Fig. 3a, delD,  delE, delF, and delG). Unlike the large deletion mutations (delA, delB, and delC), none of these small deletions affected the release of ARIA from the transfected cells (Fig. 3c). Total cellular expression levels of the precursor proteins were similar between the WT and the mutants. Quantitative comparison of the ARIA release and precursor expression between the mutants and the WT (Fig. 3d) confirmed the conclusion that none of these deletions deleteriously affected efficiency of ARIA release. In this experiment, delE mutant showed a 50% increase in the expression of the precursor protein compared with the WT, with a similar increase in ARIA release. Concomitant increase in ARIA release with increased precursor expression also demonstrated that the capacity of the transfected cells to process pro-ARIA was not saturated, validating the method of comparison. Failure to prevent release of ARIA by any of the small deletions and substitutions strongly suggested that cleavage of pro-ARIA is not sequence-specific.
Length of Stalk Affects Efficiency of ARIA Release-Although cleavage of pro-ARIA was sequence-independent, lack of ARIA release by large deletions (delA, delB, and delC mutants) clearly demonstrated the necessity of stalk for such a process. To determine the minimum length of stalk for efficient ARIA release, we made additional deletion mutations within the stalk (Fig. 4a, delH, delI, delJ, and delK).
Deletion of 9 (delJ) or 10 (delH and delI) amino acid residues did not significantly affect ARIA release (Fig. 4b, upper panel). Total cellular expression of the precursor proteins was also comparable to the WT (Fig. 4b, lower panel). When 11 amino acid residues were deleted (delK mutant), release of ARIA was noticeably reduced (Fig. 4b, upper panel) even though total cellular expression of the precursor was comparable to the WT (Fig. 4b, lower panel). Intensity of the p185 band by delK conditioned medium was similar to that by 10-fold diluted WT conditioned medium. Quantitative comparison of the ARIA release and total cellular expression of the precursors confirmed these conclusions (Fig. 4c). Efficiency of ARIA release was decreased moderately (to about 50% of the WT) when 10 amino acid residues were deleted (delH and delI), regardless of the position of the deletion within the stalk. Deletion of 9 amino acid residues (delJ) did not affect the efficiency of ARIA release, whereas deletion of an additional 2 more amino acid residues dramatically reduced it (delK). Considering that the stalk in the WT contains 21 amino acid residues, we conclude that a minimum of 11 amino acid residues is needed in the stalk to support efficient cleavage of the pro-ARIA ectodomain with no or little constraint on the actual amino acid sequence. It is noteworthy that the shortest length of the stalk for naturally occurring isoforms is 13 amino acid residues in the ␤2a isoform, which is known to release its ectodomain efficiently (6). Amino acid sequence of the ␤2a isoform is shown in Fig. 4a as a comparison.
Zinc Chelator Inhibits Release of ARIA from CHO Cells-Because ectodomain cleavage of many cell surface proteins in CHO cells has been shown to be mediated by zinc metalloproteases (24), we tested if similar proteases were responsible for cleavage of pro-ARIA in these cells (Fig. 5). Release of ARIA from stably transfected CHO cells was not affected by EDTA, indicating that pro-ARIA cleaving enzyme does not require Ca 2ϩ . On the other hand, ARIA release was strongly inhibited by 5 mM OP which chelates Zn 2ϩ and other heavy metal ions. The inhibition of ARIA release by OP was reversed by ZnCl 2 but not by the same concentration of CaCl 2 . Thus, the pro-ARIA cleaving enzyme in CHO cells appeared to be a Ca 2ϩindependent zinc metalloprotease. Ectodomain cleavage for many transmembrane proteins is known to be stimulated by protein kinase C activators (24). Similar observations have been made for pro-ARIA cleavage (9,10). These studies measured the change in the amount of immunoreactivity in the conditioned medium upon treatment with PMA, a protein kinase C activator. We confirmed that PMA stimulated ARIA release by measuring p185 phosphorylation activity in the conditioned medium (Fig. 5). PMA-stimulated ARIA release was also inhibited by OP but not significantly by EDTA (Fig. 5), suggesting that PMA-stimulated cleavage of pro-ARIA involved the same class of proteases as those responsible for steady-state cleavage.
Role of Cytoplasmic Domain in the Release of Functional ARIA-Alternative mRNA splicing generates three distinct isoforms of pro-NRG that are different in the length and amino acid sequence of the cytoplasmic domain (a, b, or c isoforms) (6). To examine if isoforms with different cytoplasmic domain have differences in ARIA release, we prepared a deletion mutant (␤1c in Fig. 6a). In this mutant, the C-terminal half of the cytoplasmic domain of pro-ARIA ("a" isoform) was deleted so that the cytoplasmic domain corresponded to that of naturally occurring "c" isoforms. Release of ARIA was significantly decreased from the cells transfected with the ␤1c mutant compared with the WT-transfected cells (Fig. 6b).
To measure total cellular expression of the precursor proteins, Western blotting of the cell lysates was performed (Fig.  6c). Considering that the ␤1c is shorter than the WT by 216 amino acid residues, it was expected that the precursor for ␤1c would run 24 kDa shorter than the WT precursor. Diffuse set of bands commonly detected by both intracellular and extracellular antibodies (1310 and HM94, respectively) showed expected decrease in molecular weight and, therefore, was assigned as the full-length precursor molecule for ␤1c (precursor-c in Fig.  6c). Interestingly, WT-transfected cell lysate also showed a diffuse set of bands that comigrated with the ␤1c precursor (Fig. 6c), suggesting that parts of the WT (␤1a) precursor proteins are proteolytically converted to the ␤1c forms. 1310 antibody detected additional lower molecular weight bands that were not detected by the HM94 antibody from both WTand ␤1c-transfected cell lysates. Among these, a band with an apparent molecular mass of 25 kDa was in good agreement with the predicted size (20 -22 kDa depending on exact site of cleavage within the stalk) of the remnant for the ␤1c mutant (filled arrowhead in Fig. 6c). Detection of a band from the WT-transfected cell lysate that comigrated with the ␤1c remnant was consistent with the notion that part of the WT precursor proteins (a isoform) were converted to the c isoform via proteolytic processing of the cytoplasmic domain.
Total cellular expression level of the ␤1c precursor molecule was comparable to the WT when measured in the same cells from which the conditioned media were collected (Fig. 6c, right  panel). Therefore, the observed decrease in ARIA release from the ␤1c mutant was not due to a decrease in expression of the precursor. Quantitative comparison of the ARIA release and precursor expression showed that the efficiency of ARIA release from the ␤1c mutant was less than 20% of the WT (Fig.  6d). The intensity of the remnant band relative to that of the  6. Long (␤1a) isoform of pro-ARIA released more ARIA than the short (␤1c) isoform at steady state. a, schematic diagram of the mutants with deletion or substitution in the cytoplasmic domain of pro-ARIA. In the V602R mutant, C-terminal valine (Val 602 ) was substituted to arginine. In the ␤1c mutant, the C-terminal half of the cytoplasmic domain was deleted to convert the cytoplasmic domain of the WT (a isoform) to that of naturally occurring c isoforms. In this mutant, the C-terminal amino acid residue is arginine. b, tyrosine phosphorylation of the p185 from L6 muscle cells by the medium conditioned with the COS cells transiently transfected with WT and different mutant constructs as indicated. precursor was also significantly decreased for the ␤1c mutant compared with the WT (Fig. 6c), consistent with decreased cleavage of the ectodomain. Decreased release of ARIA from the ␤1c mutant did not appear to be due to decreased expression at the cell surface. The ␤1c mutant was also expressed at the cell surface as shown by cell surface immunolabeling method (Fig.  2). The intensity of the cell surface labeling was not decreased from that of WT.
In the case of pro-TGF-␣, efficient ectodomain cleavage requires C-terminal valine or other aliphatic amino acid residues (25). Because the ␤1c mutant contained arginine as a C-terminal residue, we prepared another mutant in which only the C-terminal valine was substituted to arginine (V602R in Fig.  6a) to test if the decreased release of ARIA from the ␤1c mutant was simply due to change of the C-terminal residue. Unlike the ␤1c mutant, however, the V602R mutant did not show any decrease in ARIA release (Fig. 6b), and expression of the precursor protein was not significantly different from the WT (Fig.  6c, right panel). Quantitative comparison of the ARIA release efficiency confirmed this conclusion (Fig. 6d).
Despite the lower steady-state level of ARIA release from the c isoform than from the a isoform, both isoforms were stimulated to release ARIA equally well by PMA treatment (Fig. 7). Comparison of p185 phosphorylation activity between PMAuntreated and -treated conditioned media at each dilution of the media showed that ARIA release from WT-transfected cells was increased by about 5-fold upon treatment with PMA. A similar degree of stimulation was obtained for both V602R and the ␤1c mutant.
We conclude that the identity of the C-terminal residue is not important for PMA stimulation of ARIA release. This is different from PMA-stimulated ectodomain cleavage of pro-TGF-␣, which requires valine at the C-terminal position. We also conclude that the shorter isoform (c isoform) releases less than the corresponding longer isoform (a isoform) and that the membrane distal portion of the cytoplasmic domain plays a role in cleavage of the ectodomain, perhaps by binding to cytoplasmic proteins.

DISCUSSION
Many precursor proteins are cleaved at sites of dibasic or tetrabasic residues by eukaryotic subtilisin-like proteases, or furins, as they mature into fully active molecules (reviewed in Ref. 23). Pro-ARIA and related NRGs contain dibasic residues at the extracellular juxtamembrane position, which has been proposed to be a cleavage site (2,4). However, our results show that this is not a major cleavage site at least when the precursor is expressed in fibroblasts. We attempted to remove a potential cleavage site by making a series of small (4 -5 amino acid residues) deletions (delD, delE, delF, or delG) to cover the rest of the entire stalk (between extracellular EGF domain and the transmembrane domain). None of these deletions prevented ARIA release, indicating that cleavage was not sequence-specific. Alternatively, cleavage may occur at more than one specific site. Once a primary cleavage site is removed by deletion, a secondary cleavage site may be utilized as has been shown to occur with pro-TGF-␣ (16) and pro-TNF␣ (26,27). Quantitative analysis of the released ARIA and expression of the precursor proteins showed no decrease in efficiency of ARIA release for any of these mutants, indicating that efficiency of cleavage is not different between the putative primary and the secondary cleavage sites. Cleavage at multiple sites with equal efficiency would be considered as a special case of cleavage with no sequence specificity. Therefore, we conclude that cleavage of pro-ARIA is not sequence-specific.
Although cleavage of pro-ARIA may not require a specific amino acid sequence at the cleavage site, release of ARIA does not appear to be a result of random degradation of the pro-ARIA. Marked reduction in ARIA release by deletion of 11 amino acid residues within the stalk (delK mutant) indicates that cleavage of pro-ARIA is a specific process subject to structural constraints. C-terminal sequencing of the purified NRGs also showed predominance of a specific amino acid residue at the C terminus (28), indicating cleavage at a specific site. Two different mechanisms have been proposed for specific cleavage of transmembrane precursor proteins in a sequence-independent manner. Cleavage of amyloid precursor protein by ␣-secretase, which we term as "molecular ruler mechanism," has been shown to occur at a certain distance from the membrane regardless of the exact amino acid sequence (29). Proteolytic cleavage of ectodomain of the transmembrane form of angiotensin-converting enzyme was also shown to be independent of the amino acid sequence at the cleavage site (30,31). In this case, which we term as "modified molecular ruler mechanism," cleavage occurred at a certain distance from the membraneproximal extracellular folded domain rather than from the membrane. Lu et al. (28) have expressed various pro-NRG isoforms in CHO cells and purified the NRGs from the conditioned media. The C-terminal amino acid residue of all the purified NRGs was close to the C-terminal end of the EGF domain but not close to the N terminus of the transmembrane domain despite wide differences in the amino acid sequence and length of the stalk. From our results and those of Lu et al. (28), we propose that cleavage of pro-NRG occurs via the "modified ruler mechanism." In this model, cleavage of pro-ARIA occurs independently of the amino acid sequence within a short distance (3-4 amino acid residues) from the extracellular EGF domain but requires a specific length of the stalk for efficient cleavage.
The substrate for those "ruler" proteases would be a stretch of amino acid residues not sterically hindered by the local secondary structure or by the nearby bulky structures. The minimum length of such an unhindered stretch of amino acid residues appears to be 11 amino acid residues for pro-ARIA considering a sharp decrease in the efficiency of ARIA release when the length of the stalk was decreased from 11 to 10 amino acid residues. Ehlers et al. (30,31) have also proposed the same length of stalk for optimal ectodomain cleavage of membraneanchored angiotensin-converting enzyme. It is interesting to note that despite wide variation in the length and sequence of the stalk among different pro-NRG isoforms, all isoforms have longer than 11 amino acid residues in their stalk, and they all efficiently release the ectodomain (6). The shortest naturally occurring stalk is that for the the ␤2a isoform with 13 amino acid residues.
Diverse transmembrane proteins undergo ectodomain cleavage (reviewed in Refs. 32 and 33). Although amino acid sequences around the cleavage site do not show a high degree of homology, a common mechanism appears to operate. A group of metalloprotease inhibitors, OP and TAPI-2 but not EDTA, inhibited ectodomain cleavage of diverse transmembrane proteins from CHO cells suggesting a Ca 2ϩ -independent metalloprotease as a cleaving enzyme for diverse membrane proteins (24). Considering that ARIA release from transfected CHO cells was also inhibited by OP but not by EDTA, it appears likely that cleavage of pro-ARIA is also mediated by the similar protease in these cells.
A candidate has been identified. Cloning of cDNA for a protease responsible for cleavage of pro-TNF␣ (TACE) revealed that it belonged to an ADAM family of transmembrane zinc metalloprotease (34,35). Cells derived from mice lacking both copies of this gene were defective in ectodomain cleavage of not only pro-TNF␣ but also many transmembrane proteins including pro-TGF-␣, L-selectin, and TNF receptor (36). It is not known how TACE recognizes these diverse proteins. TACE may be a ruler protease that recognizes an unhindered stretch of amino acid residues near the membrane rather than a specific amino acid sequence of each of these proteins. Being anchored to the membrane, TACE would be in a good position to measure distance from the membrane or from a membraneproximal folded domain. Such a possibility has not been tested yet.
Ectodomain cleavage of many transmembrane proteins can be up-regulated by treatment with PMA (24,37,38). In the case of pro-TGF-␣, the C-terminal residue in the cytoplasmic domain is crucial for such up-regulation. Substitution of the Cterminal valine with polar amino acid residues significantly impaired stimulation of ectodomain cleavage in response to PMA (25). Although many other transmembrane precursor proteins including pro-ARIA, CSF-1, and the c-kit ligand also contain valine at the C terminus (2,39,40), the requirement of the C-terminal valine for PMA-stimulated ectodomain cleavage does not appear to be universal. Burgess et al. (9) have shown that an ␣2c isoform of pro-NRG with arginine as a C-terminal residue undergoes PMA-stimulated ectodomain cleavage equally well or even faster than a ␤4a isoform of pro-NRG with valine as a C-terminal residue. It remained possible, however, that the two different isoforms are under different regulatory mechanisms considering the many differences in the sequence and length of both the extracellular and intracellular domains. To eliminate such possibilities, we changed the cytoplasmic sequence of pro-ARIA without affecting the rest of the molecule. The C-terminal valine of pro-ARIA could be substituted to arginine without affecting PMA stimulation of ARIA release. Deletion of the C-terminal half of the cytoplasmic domain to convert the pro-ARIA (a isoform) to a c isoform (␤1c mutant) also did not affect PMA stimulation of ARIA release despite the large difference in the size of the cytoplasmic domain. From these comparisons, we conclude that a membrane-distal portion of the cytoplasmic domain, including the C-terminal valine, is not required for PMA stimulation of ARIA release.
At steady state, the c isoform (␤1c mutant) released much less ARIA than the corresponding a isoform (WT), suggesting that the C-terminal half of the cytoplasmic domain contributes to efficiency of pro-ARIA cleavage. Consistent with our findings, Wen et al. (6) have also reported that, despite higher expression of the ␣2c isoform than the ␣2a isoform in the transfected COS cells, the amount of released NRG was comparable between the two. The difference in the efficiency of ARIA release between the two isoforms raises the possibility that the a isoforms of pro-NRGs may serve primarily as a precursor to release soluble NRGs (paracrine signaling), whereas the c isoforms may stay as a membrane-bound form to signal through direct cell-cell contact (juxtacrine signaling). Alternatively, the c isoforms may be degraded faster and release less soluble NRG than the a isoforms. In either case, proteolytic cleavage of the cytoplasmic domain may instantaneously convert a isoforms of pro-NRGs to c isoforms without involving mRNA splicing. Such proteolytic processing of pro-NRG may provide a mechanism for rapid conversion of the signaling mode from paracrine to juxtacrine or for rapid downregulation of NRG release. Observation of bands from the WT (a isoform)-transfected cell lysates that comigrated with those from the ␤1c-transfected cell lysate supports the notion that the a isoform can be proteolytically converted to the c isoform. Wang et al. (41) have reported that the cytoplasmic domain of pro-NRG interacts with LIM kinase 1. It will be interesting to find out if the LIM kinase is involved in regulation of NRG release or signaling by the pro-NRG.
In contrast to our findings that the membrane-distal portion of the cytoplasmic domain is important for ARIA release, Liu et al. (42) have recently proposed that the membrane-proximal half of the cytoplasmic domain was critical for proteolytic cleavage of the ectodomain. They reported that a mutant lacking the whole cytoplasmic domain did not release detectable amounts of NRG, whereas the ␣2a and ␣2c isoforms released similar amounts of NRG. This observation was inconsistent with our findings that the c isoform (␤1c mutant) released significantly less amount of ARIA than the corresponding a isoform (WT). This discrepancy may not be attributed to differences between the ␣ and the ␤ isoforms because Wen et al. (6) have also reported that the release of soluble NRG from the ␣2c isoform was less than that from the ␣2a isoform. Burgess et al. (9) also showed that the amount of radioactive NRG in the medium 4 h after pulse labeling with 35 S was much higher for the ␤4a isoform than the ␣2c isoform. Although the half-life of the precursor was shorter for the ␣2c isoform than the ␤4a isoform, quicker disappearance of the ␣2c may be due to degradation of the precursor without release of functional NRG. Therefore, data from our experiments and others (6,9) all indicate that the c isoform releases less NRG than the a isoform. We conclude that the membrane-distal portion of the cytoplasmic domain (unique to the a isoform) plays an important role in the steady-state level of ARIA release. The membrane-proximal portion of the cytoplasmic domain may make an additional contribution to proteolytic cleavage of the ectodomain (43), although we have not tested this possibility.
Why is ARIA initially made as a transmembrane precursor? Aside from the possibility that pro-ARIA may itself serve as a signaling molecule through cell-cell contact, there is a unique advantage to being made as a transmembrane precursor form. Release of many proteins and peptides is under tight regulation. Insulin, for example, is released only when blood glucose level rises after a meal. These proteins are made by cells that have secretory vesicles designed for regulated release of proteins. ARIA, on the other hand, is made not only in neurons but also in many other cell types including cells of mesenchymal origins. Most cell types do have constitutive secretory pathways, but not all of them have a regulated secretion mechanism. Delivering pro-ARIA to the cell surface via a constitutive pathway and then regulating cleavage of the ectodomain would be an economical solution to achieve the regulated release of ARIA in these cells without implementing a regulated secretory pathway.