Proprotein Convertase PC7 Enhances the Activation of the EGF Receptor Pathway through Processing of the EGF Precursor*

Although the processing profile of the membrane-bound epidermal growth factor precursor (pro-EGF) is tissue-specific, it has not been investigated at the cellular level nor have the cognate proteinases been defined. Among the proprotein convertases (PCs), only the membrane-bound PC7, the most ancient and conserved basic amino acid-specific PC family member, induces the processing of pro-EGF into an ∼115-kDa transmembrane form (EGF-115) at an unusual VHPR290↓A motif. Because site-directed mutagenesis revealed that Arg290 is not critical, the generation of EGF-115 by PC7 is likely indirect. This was confirmed by testing a wide range of protease inhibitors, which revealed that the production of EGF-115 is most probably achieved via the activation by PC7 of a latent serine and/or cysteine protease(s). EGF-115 is more abundant at the cell surface than pro-EGF and is associated with a stronger EGF receptor (EGFR) activation, as evidenced by higher levels of phosphorylated ERK1/2. This suggests that the generation of EGF-115 represents a regulatory mechanism of juxtacrine EGFR activation. Thus, PC7 is distinct from the other PCs in its ability to enhance the activation of the cell surface EGFR.

Although a wide variety of cancers express EGFR (5), clinical trials of EGFR inhibitors showed moderate beneficial effects (6). A better understanding of EGFR activation may lead to more rational approaches to drug development.
Processing of pro-EGF into the well known 53-amino acid (aa) mature mouse EGF after a basic residue both at the N-(Arg 959 2) and C-terminal (Arg 1012 2) ends is not equally efficient in various tissues. Thus, in the submaxillary gland, pro-EGF is fully and intracellularly processed by kallikreins into mature EGF and then stored in secretory granules (7). In contrast, in kidney, pro-EGF is found at the cell surface mainly as an unprocessed protein (1,8) or as intermediate forms generated by unknown proteases (9 -11). Among the processed forms in kidney, a prominent ϳ115-kDa membrane-bound intermediate (herein called EGF-115) was reported (12), as well as a soluble form in urine generated by an undefined serine protease (10).
Many of the estimated ϳ3500 mammalian secretory proteins, including growth factors and their receptors, are subjected to limited proteolysis by one or more of the nine members of the proprotein convertase (PC) family of serine proteinases of the subtilisin-kexin type as follows: PC1/3, PC2, Furin, PC4, PC5/6, PACE4, PC7, SKI-1/S1P, and PCSK9. Extensive biochemical and in vivo analyses expanded our understanding of some of the physiological functions of these PCs and their roles in embryonic development and in the adult. The first seven members cleave secretory proteins at specific single or paired basic aa within the motif (Arg/Lys)-(Xaa) n -Arg2 (n ϭ 0, 2, or 4 aa) (13)(14)(15)(16). The last two members, SKI-1/S1P and PCSK9, do not cleave at basic aa and are major regulators of cholesterol and lipid metabolism (16,17).
A number of investigations aimed at defining the sequence recognition of PCs revealed some degree of redundancy in their ability to process certain precursor proteins (18). Thus, both Furin and PACE4 redundantly process the TGF␤-like Nodal during early embryonic development (19). However, evidence for PC-specific substrates has also been presented. Thus, although Furin and PC5/6 often cleave the same substrates, the TGF␤-like growth differentiating factor GDF-11 is selectively cleaved by PC5/6 during development (20).
Although the physiological functions of most PCs are now better understood (16,21), the unique functional roles of the seventh member PC7, the most ancestral of the basic aa-specific convertases, are barely explored. Although less efficient than Furin, PC7 specifically cleaves overexpressed substrates at Arg2 residues both in vitro (22)(23)(24)(25)(26)(27)(28)(29) and in cell lines (30 -39). Northern blot analyses revealed a wide expression of PC7 mRNA in all rat tissues and cell lines analyzed (40), suggesting that it may have multiple physiological functions. Quantitative RT-PCR analysis of PC7 expression in adult mouse tissues showed that colon, duodenum, heart, and kidney are the richest sources of PC7 mRNA (41).
We herein characterized the biosynthetic pathway of pro-EGF as well as its enhanced processing by PC7. Our data show that PC7 is the only convertase that enhances the indirect cleavage of pro-EGF into EGF-115, likely via the activation of serine and/or cysteine protease(s). We also show that EGF-115 is more concentrated at the cell surface and more efficiently activates the EGFR than full-length pro-EGF.
In Situ Hybridization-Kidneys and lacrimal glands of adult wild type mice were frozen and cut into 10-m sections. Tissue slices were fixed in 4% formaldehyde at 4°C for 1 h. In situ hybridization was carried out as reported previously (43) using a complementary mouse pro-EGF or mouse PC7 cRNA probe (552 and 640 nucleotides (nts), respectively) labeled with 35 S-UTP. To generate a mouse pro-EGF cRNA probe, a cDNA segment covering the sequence encoding nts 3051-3602, localized downstream to mature EGF sequence, was PCR-amplified using sense 5Ј-GAAGCATGACATCATGGTGG-3Ј and antisense 5Ј-GACACAGCTTATATATGAATGTCTTG-3Ј oligonucleotides and subcloned into the vector pDrive (Invitrogen). To generate a mouse PC7 cRNA probe, a cDNA segment covering the sequence encoding amino acids 1-639 was amplified using sense 5Ј-TGCTGTTCTGATGCCGAAAGG-3Ј and antisense 5Ј-GGGTCATTAGAGTTGAGGTCATAG-3Ј oligonucleotides and subcloned into pCRII-TA cloning vector (Invitrogen). Hybridization was analyzed on x-ray film (exposure time, 2 days for EGF and 5 days for PC7).
Co-cultures-HEK293 cells expressing either pro-EGF, PC7, or pIRES were co-cultured at an EGF/PC7 cDNA ratio of 1:2. Twenty four hours later, the cells were washed and incubated overnight in serum-free medium before cell lysis. For co-cultures of HEK293 and A431 cells, 24 h post-transfection, HEK293 cells were trypsinized, and 4 ϫ 10 5 cells were mixed in a new P3.5 plate containing 6 ϫ 10 5 A431 cells. Twenty four hours later, the cells were washed and incubated in serum-free medium for 3 h before cell lysis. The incubation of this co-culture with mouse recombinant EGF (20 ng/ml) was used as a positive control for the phosphorylation of ERK1/2.
Enzymatic Digestion of Carbohydrates-Proteins from cell lysates were incubated with endoglycosidase H (endoH) or N-glycosidase F for 1 h at 37°C. Deglycosylated proteins were separated by 8% SDS-PAGE and revealed by immunoblotting.
Cell Surface Biotinylation-For biochemical detection of pro-EGF at the plasma membrane, transfected HEK293 cells were washed with ice-cold phosphate-buffered saline (PBS) adjusted to pH 8 and biotinylated with 0.2 mg of sulfosuccinimidyl-6-(biotin-amido) hexanoate (Pierce) for 30 min at 4°C. Cells were then washed with 100 mM glycine in PBS, pH 8, to quench the reaction. Cell lysates were immunoprecipitated with anti-streptavidin-agarose (Fluka), and immunoprecipitates were then treated as described above.
Biosynthetic Analysis-HEK293 cells were transiently transfected in 60-mm dishes. Biosynthesis was performed 48 h posttransfection, and the cells were washed and pulse-labeled in Cys/Met-free RPMI 1640 medium containing 0.2% BSA for 2 h with 250 Ci/ml of [ 35 S]Cys/Met for all experiments except for sulfation of pro-EGF (2 h with 500 Ci/ml of Na 2 35 SO 4 ). After the pulse, the media were recovered, and the cells were lysed as mentioned previously (45). For pro-EGF pulse-chase experiments, the cells were pulse-labeled with [ 35 S]Met/Cys for 20 min and then chased for various periods. The cell lysates were immunoprecipitated with mAb V5 (1:500, Invitrogen). The immunoprecipitated proteins were resolved by SDS-PAGE on 8% Tricine gels, dried, and autoradiographed, as described previously (45). In some cases, during the 30 min preincubation times for pulse analysis, the cells were pretreated with brefeldin A (2.5 g/ml, Calbiochem), methyl-␤-cyclodextrin (M␤CD, 10 mM, Sigma), or 3-hydroxynaphthalene-2-carboxylic acid (3,4dihydroxybenzylidene)-hydrazide monohydrate (Dynasore; 80 M, Sigma), and this treatment was continued during the 2-h pulse period. Protease inhibitors were added during the 2-h pulse period. The concentrations used, listed in Table 1, were based on previous reports (46 -53). Half of the cellular proteins were precipitated with mAb V5 (1:500, Invitrogen) and the other half with rabbit anti-PC7 (10 l (35)).
Fluorescence-activated Cell Sorting (FACS)-Twenty four hours post-transfection, HEK293 cells were processed for flow cytometry. Cells were washed with buffer A (1ϫ PBS with 7.1 ml of 35% BSA and 2 ml of 250 mg/liter glucose) and then stained with a rat anti-mouse pro-EGF primary antibody (1:250, R&D Systems) for 40 min at room temperature and then with anti-rat IgG conjugated with Alexa 647 (1:750, Molecular Probes) for 20 min at room temperature. FACS analysis was performed using the CyAn ADP cytometer (Beckman Coulter), and the data were analyzed using the Summit software (Beckman Coulter). The mean fluorescence intensity of EGF at the cell surface for a population of 50,000 cells was determined.
Immunofluorescence-Cells were plated on collagen-coated culture dishes (MatTek) and then transfected the following day. Twenty four hours post-transfection, the cells were fixed for 15 min with 4% paraformaldehyde ϩ 1% methanol for PC7 and pro-EGF cell surface labeling to slightly permeabilized cell to have access to the V5 tag or 5 min at Ϫ20°C with 100% methanol for intracellular staining. Cell fixed with paraformaldehyde were incubated 5 min in 150 mM glycine to quench aldehyde groups. Cells were then incubated overnight with primary antibodies. PC7 was detected with rabbit anti-PC7 (1:1,000 (35)) and pro-EGF with mAb V5 (1:200, Invitrogen). Antigen-antibody complexes were revealed by 1 h of incubation with the corresponding species-specific Alexa Fluor (488, 555, or 647)tagged antibodies (Molecular Probes), and the nuclei were stained by Hoechst 33258 for 1 min (100 g/ml, Sigma). Cells were covered with 1,4-diazabicyclo[2.2.2]octane (Sigma) in PBS, 90% glycerol, and immunofluorescence analyses were performed with a confocal microscope (Zeiss LSM-710).
In Vitro Activity Assay-Enzymatic in vitro assays were performed in 100 l of buffer (1 mM CaCl 2 , 25 mM Tris-HCl, pH 6.7) at 37°C for 1 h in the presence of 100 M of the substrate pyroglutamic acid-RTKR-7-amido-4-methylcoumarin (23) and 5 l of concentrated media from HEK293 cells expressing either sPC7 or pIRES. The media were first preincubated for 20 min with the protease inhibitors, and then the fluorogenic substrate was added for a further incubation of 1 h. The release of free 7-amino-4-methylcoumarin was detected with a spectra MAX GEMINI EM microplate spectrofluorimeter (Molecular Devices, excitation, 360 nm; emission, 460 nm). The mean of the relative fluorescence unit of three independent experiments is reported.
Statistical Analysis-All experiments were performed at least in triplicate. Quantifications are defined as mean Ϯ S.E. In some cases, results were expressed as percentage of control, which was considered to be 100% or normalized to EGF ϩ pIRES, which was considered as 1. The statistical significance of differences between various groups was evaluated by one-way analysis of variance test followed by the Dunnet test. A difference between experimental groups was considered statistically significant whenever the p value was Ͻ0.05.

PC7 Processes Pro-EGF into an ϳ115-kDa Membrane-bound
Form-We first tested the possible implication of PCs in the processing of the cell surface pro-EGF, which may mimic the undefined kidney protease reported to be implicated in pro-EGF maturation (12). Thus, we co-expressed pro-EGF in HEK293 or Neuro2A cells with PC1/3, Furin, PC5/6A, PACE4, or PC7, the only basic aa-specific PCs that can function in the trans-Golgi network and/or cell surface (16). The data demonstrated that under our experimental conditions, in the two cell lines, only PC7 significantly processed (ϳ45%) the membranebound pro-EGF, resulting into an ϳ115-kDa nonsecreted V5-tagged product (EGF-115; Fig. 1A). In addition, no smaller cellular V5-tagged fragment(s) was detected in cells (data not shown), suggesting that pro-EGF is not shed into the HEK293 media. To check and compare the protein levels of the overexpressed PCs, V5-tagged PCs, except for PC1 that was detected with a C-terminal PC1-specific antibody (44), were co-expressed in HEK293 cells. The data show that all PCs are similarly expressed (supplemental Fig. S1). This conclusion was further supported by the analysis of the EGF immunoreactivity in the media, which showed that no PC can generate the 53-aa mature EGF or any shed form of pro-EGF (supplemental Fig.  S2). Intriguingly, we also noted that in COS-1 cells PC7 was much less efficient in generating EGF-115 because the cleavage of pro-EGF represented only ϳ25% compared with ϳ45% in Neuro2A and HEK293 cells (Fig. 1A). We confirmed that the PC7 activity is necessary for the observed pro-EGF processing by treating cells with the general PC inhibitor decanoyl-RVKRcmk ( Fig. 1B) (42), which resulted in much reduced levels of EGF-115 similar to basal levels (ϳ14%). It is thus possible that the EGF-115-like band seen in the absence of overexpressed PC7 may not be exclusively PC7-generated. This is supported by the treatment of Neuro2A cells with an siRNA against PC7, which completely abolished endogenous PC7 expression but the background EGF-115-like band still remained (supplemental Fig. S3).
In situ hybridization of pro-EGF and PC7 mRNA showed that whereas lacrimal glands express high levels of pro-EGF, they only exhibit low levels of PC7 transcripts (Fig. 2). In contrast, both are highly expressed and co-localized in the medulla and cortex of the kidney (Fig. 2). Because kidney pro-EGF is cleaved into an EGF-115-like product (12), the above data suggest that PC7 may be involved in pro-EGF processing in the kidney. Furthermore, in HEK293 cells, pro-EGF and PC7 co-localize at the cell surface under slightly permeabilizing conditions ( Fig. 3A) and in subcellular compartments just under the cell surface under permeabilizing conditions (Fig. 3B). Altogether, these data support the hypothesis that PC7 may process pro-EGF ex vivo and likely in vivo into EGF-115.

Generation of EGF-115 Requires Endocytosis and Cholesterol-rich
Microdomains-Because pro-EGF and PC7 are highly expressed in the kidney and co-localize in HEK293 cells, all subsequent experiments were performed in the latter cells. To define the processing kinetics and the compartment(s) where such cleavage occurs, cells expressing PC7 and pro-EGF were pulsed with [ 35 S]Met/Cys for 20 min and then chased for 1-5 h, followed by SDS-PAGE analysis of the cell lysates immunoprecipitated with the mAb V5 (Fig. 4A). Although EGF-115 was not detected at 30 min (data not shown), it appeared by 1 h and reached maximum levels by 2 h of chase, followed by a slow decline, suggesting that processing occurs after pro-EGF exits from the endoplasmic reticulum (ER). Indeed, pro-EGF is partially sensitive to endoH digestion (Fig. 4B), and only ϳ50% of it, compared with [ 35 S]Met/Cys-labeling, is sulfated (Fig. 4C), suggesting that only the latter fraction exited the ER. In contrast, EGF-115 is both endoH-insensitive (Fig. 4B) and is fully sulfated (Fig. 4C), supporting the notion that it is produced in a post-ER compartment.
We next tested whether retention of pro-EGF and/or PC7 in the ER would permit processing. Thus, HEK293 cells expressing pro-EGF and PC7 were incubated with brefeldin A, which inhibits ER to Golgi transport of proteins (54). This treatment prevented pro-EGF processing because it reduced the level of EGF-115 to the basal one (ϳ14%). (Fig. 4D). Similarly, a PC7-KDEL construct, which retains the enzyme in the ER because of its C-terminal KDEL sequence (55), cannot process pro-EGF (Fig. 4E). We conclude that processing of pro-EGF does not occur in the ER but rather in a post-ER compartment.
To further identify such a compartment, we tested the implication of cholesterol-rich microdomains or endosomes as follows. 1) Accordingly, we first evaluated whether depletion of the cholesterol cell content would affect the ability of PC7 to generate EGF-115 from pro-EGF. The extraction of cholesterol was achieved with M␤CD that has been shown to selectively eliminate cholesterol from the plasma membrane, in preference to other lipids (56). Thus, cells expressing pro-EGF and either pIRES or PC7 were preincubated with 10 mM M␤CD for    4F). These data suggest that processing of pro-EGF requires the presence of cholesterol-rich microdomains. 2) Incubation of the same cells with 80 M dynasore, a cell-permeable inhibitor of dynamin GTPase activity that blocks clathrin-dependent endocytosis (57), also reduced to basal levels (ϳ14%) the generation of EGF-115 (Fig. 4F). 3) Finally, the requirement of endocytosis for pro-EGF processing is supported by the lack of generation of EGF-115 in confluent co-cultures of pro-EGFexpressing cells with those expressing either PC7 or its soluble secretable form sPC7 (supplemental Fig. S4). These data suggest that pro-EGF processing by full-length membrane-bound PC7 does not occur at the cell surface, but rather it requires endocytosis of pro-EGF into clathrin-coated pits and likely localization in cholesterol-rich microdomains.
Transmembrane Full-length PC7 Can Process Membranebound and Soluble Pro-EGF-Interestingly, the transmembrane (TM)-cytoplasmic (CT) domains of pro-EGF are dispensable, because PC7 can process both wild type (WT) and soluble pro-EGF (Fig. 5A). However, soluble sPC7 cannot process WT pro-EGF (Fig. 5B), although sPC7 is active in in vitro assays (23). These data suggested that the TMCT domain of PC7 is required for pro-EGF processing.
PC7 is the only convertase that contains Cys-palmitoylation sites within its cytosolic domain (Fig. 6A) (41,58). Thus, we tested whether the CT-Cys-palmitoylation regulates pro-EGF processing. We note that both human and rat CTs contain five and eight Cys residues, respectively. Earlier work indicated that in h-PC7, Cys 699 and Cys 704 are palmitoylated (58), and our [ 3 H]palmitate incorporation studies with the double PC7 mutant C699A,C704A confirmed that these are the only palmitoylation sites in h-PC7 when expressed in HEK293 cells (41). Both WT and PC7-(C699A,C704A) double mutant process pro-EGF equally well (ϳ47%, Fig. 6B). In addition, 2-bromopalmitate, a palmitoylation inhibitor, does not prevent pro-EGF processing by PC7 (Fig. 6C). Altogether, these data demonstrate that Cys-palmitoylation is not required for PC7 activity on pro-EGF.
To determine whether specific sequences within the TM or TMCT domains of PC7 are critical for its activity on pro-EGF,  we swapped these domains in h-PC7 with the corresponding ones of h-Furin (PC7 CT-Furin ; PC7 TMCT-Furin ) and vice versa (Furin CT-PC7 and Furin TMCT-PC7 ) (Fig. 7A) (41). None of these swaps affected the extent of PC7-processed pro-EGF, which is maintained at ϳ47%, or allowed cleavage of pro-EGF by Furin (Fig. 7B). Thus, although pro-EGF processing requires a membrane-bound form of PC7, the TMCT of the latter can be replaced by that of Furin.
PC7 Indirectly Enhances the Processing of Pro-EGF at VHPR 290 -Because Leu is an abundant aa within the pro-EGF sequence, we tried to microsequence the [ 3 H]Leu-labeled EGF-115. However, our multiple attempts failed to give a clear protein sequence (data not shown). This may be due to the absence of Leu within the first 20 residues of the EGF-115 or to the presence of a Gln early in the sequence that would cyclize and block sequencing.
To identify the pro-EGF cleavage site, we generated various secretory pro-EGF deletants by fusing the signal peptide to either aa 160, 202, 310, or 341 (Fig. 8A) and estimated their molecular mass compared with the EGF-115 (Fig. 8B). The data showed that cleavage seems to occur close to the LKQR 310 site (Fig. 8B). However, systematic mutations of Arg around this site in mouse pro-EGF, including the mutants R294A, R310A, R312A, R315A, R321A, did not affect processing (data not shown). Because microsequencing did not yield informative results, we suspected that cyclization of a Gln may have hindered sequencing by Edman degradation. Accordingly, among other mutants, we concentrated on the double mutant Q292M,K300M because Lys 300 is not conserved, and the specific activity of [ 35 S]Met is much higher than that of [ 3 H]Leu. Microsequencing of [ 35 S]Met-labeled EGF-115 resulting from the pro-EGF-(Q292M,K300M) double mutant gave a clear sequence with a Met at positions 2 and 10. This unique sequence unambiguously identified Arg 290 as the cleavage site, pointing to an unusual processing motif VHPR 290 2A (Fig. 8C), which is not a normal cleavage site for any PC, as it does not display a second basic residue preceding Arg 290 (13). It was thus not surprising that the R290A mutation did not affect processing of pro-EGF, which is maintained at ϳ50% (Fig. 8D). This suggests that PC7 triggers the cleavage activation of an unknown latent protease, which then cleaves pro-EGF at Arg 290 . Furthermore, the PC7 dose-dependent processing of pro-EGF was not sensitive to PC7 levels between 3 and 300 ng of cDNA transfected, for a constant level of pro-EGF cDNA of 300 ng. This represents a similar efficacy of processing for PC7/EGF cDNA ratios of 1:1 up to 1:100 (Fig. 8E), suggesting that the level of the latent protease is limiting, and a small amount of PC7 activity, inhibitable by dec-RVKR-cmk (Fig. 1B), is enough to activate it. We also presume that the low efficacy of pro-EGF processing in COS-1 cells (Fig. 1A) may be due to an even lower level of the cognate latent enzyme in the latter cells versus HEK293 or Neuro2A cells.
To define the nature of the intermediate protease, we first tested the effect of a wide range of protease inhibitors on pro-EGF processing. HEK293 cells co-expressing pro-EGF and either pIRES or PC7, pulse-labeled with [ 35 S]Met/Cys, were incubated with different protease inhibitors (Table 1). Our data showed that the membrane-permeable cysteine protease inhibitor (2S,3S)-trans-epoxysuccinyl-L-leucylamido-3-methylbutane ethyl ester (E64d, 25 M) and the serine protease inhibitors 4-(2-aminoethyl) benzenesulfonyl fluoride hydrochloride (AEBSF, 300 M) and ␣ 1 -antitrypsin (0.5 mg/ml) significantly decreased (60 -70%) the processing of pro-EGF into EGF-115 (Fig. 9A). These inhibitors did not inactivate PC7 in vitro even at high concentrations (Fig. 9B). Altogether these data suggest that the intermediate enzyme is either a cysteine and/or serine protease or that PC7 concomitantly activates cysteine and/or serine proteases. All other protease inhibitors had no significant effect on pro-EGF processing (Fig. 9, A and C). The presence of a proline in P2 is unusual and raised the possibility that thrombin, granzyme M, or neurolysin (59, 60) may be the intermediate enzyme. We thus tested the importance of this proline by mutagenesis (P289A). The cleavage of pro-EGF into EGF-115 represented ϳ45% even with pro-EGF mutant P289A, demonstrating that the proline in P2 is not critical for the intermediate enzyme (Fig. 9D). Furthermore, treatment of HEK293 cells co-expressing pro-EGF and PC7 with specific thrombin (antithrombin III) and neurolysin (Pro-Leu) inhibitors had no effect on pro-EGF processing (Fig. 9A).

EGF-115 Exhibits Higher Cell Surface
Localization and Activity on EGFR than Pro-EGF-To increase the level of the EGF-115, we generated an RRRR 290 EL mutant at the PC7-enhanced cleavage site (Fig. 10A) exhibiting an optimal Furin-like motif, which would be expected to be cleaved efficiently by endogenous PCs in HEK293 cells (61). Indeed, although EGF-115 represents ϳ40% of total wild type pro-EGF indirectly produced by PC7, its level rises to ϳ55% in the RRRR 290 EL mutant (Fig.  10A). Accordingly, we compared the cell surface expression of pro-EGF and pro-EGF co-expressed with PC7 with that of the RRRR 290 EL mutant by FACS (Fig. 10B), cell surface biotinyla-tion (Fig. 10C), and immunocytochemistry (Fig. 10D) analyses. The FACS experiment showed that although the number of cells expressing cell surface pro-EGF was similar in all transfections (data not shown), the mean cell surface fluorescence intensity of pro-EGF significantly increased ϳ1.7-fold in the RRRR 290 EL mutant (Fig. 10B). Interestingly, immunohistochemistry shows a clear enrichment of pro-EGF at the cell surface in pro-EGF-(RRRR 290 EL)-overexpressing cells (Fig. 10D). Finally, in the presence of PC7, cell surface biotinylation data show that EGF-115 is the major cell surface protein (ϳ66% of total cell surface pro-EGF) as compared with pro-EGF (Fig.  10C).
Cycloheximide treatment of HEK293 cells overexpressing pro-EGF and PC7 showed that the levels of pro-EGF and its EGF-115 derivative were stable during the first 4 h of treatment. Following this initial period, they were both degraded with similar kinetics of ϳ7% decrease/hour with estimated half-lives of ϳ8 -9 h (Fig. 10E). We conclude that the increased presence of the EGF-115 at the cell surface is not due to a longer half-life versus pro-EGF but rather to a preferred localization of this product at the cell surface, which may thus enhance activation of the EGFR.
To verify this hypothesis, we analyzed EGFR activation via the downstream phosphorylation of the MAPK ERK1 and ERK2 (62). Transfected HEK293 cells were mixed with A431 cells, known to express high levels of EGFR (63), and hence represent better responders to EGF. The amount of phosphorylated and nonphosphorylated ERK1/2 was estimated by Western blotting (Fig. 10F). We demonstrate that cells co-expressing pro-EGF and an empty vector (control) or sPC7 induce a low level of ERK1/2 phosphorylation (Fig. 10F). In contrast, co-expression of pro-EGF with PC7 or expression of the pro-EGF-RRRR 290 EL mutant induce a higher level of ERK1 phosphorylation (ϳ2-and ϳ5-fold increase, respectively) than cells The deduced sequence positions of the 35 S-radiolabeled residues are shown. D, processing of pro-EGF or its R290A mutant into EGF-115 in the absence or presence of rat PC7 was tested as described above. E, Western blot of lysates obtained from cells expressing pro-EGF (300 ng) and three different doses of PC7 (300, 30, and 3 ng) shows that pro-EGF processing requires less than 100-fold lower cDNA input of PC7 compared with pro-EGF. The percent of pro-EGF cleavage into EGF-115 is indicated below each lane. These data are representative of three independent experiments. Antithrombin III, human plasma co-expressing pro-EGF and pIRES (control). ERK2 phosphorylation was also increased but was detected at very low levels.
Collectively, these data demonstrate that PC7 processing favors the cell surface localization of EGF-115, and consequently it enhances EGFR activation.

DISCUSSION
Although EGF has strong biomedical significance, little is known about its biosynthesis and processing, especially the enzymes responsible for the soluble and transmembrane intermediate EGF forms (9,10). Here, we examined the possible implication of the proprotein convertases in pro-EGF maturation. The data showed that none of the basic aa-specific PCs were able to generate soluble EGF forms in media, including the 53-aa EGF hormone (supplemental Fig. S2), suggesting that the serine protease involved in release of soluble EGF (10) is not a member of PC family. However, we did find that only fulllength membrane-bound PC7 can enhance the processing of pro-EGF into an intermediate ϳ115-kDa transmembrane form (EGF-115; Fig. 1), which had already been observed in kidney, but not in lacrimal gland, extracts (12). Interestingly, we demonstrate that in kidney both PC7 and EGF mRNAs are abundant, but PC7 expression is very low in lacrimal glands (Fig. 2), in which EGF is only stored as its full-length membrane precursor (12). Thus, the low level of PC7 in lacrimal glands may explain the absence of EGF-115 in this tissue. We also showed that PC7 co-localizes with pro-EGF in HEK293 cells (Fig. 3). These data support the notion that PC7 induces pro-EGF processing ex vivo and may play the same function in vivo.
The proprotein convertase PC7, the most ancient and conserved PC-like enzyme, is the least studied member of the mammalian basic aa-specific PCs. When it was first discovered in 1996 (40,64), it was thought to have similar properties to the other members of the family in terms of its zymogen activation, cellular trafficking, and cleavage specificity (23,32). Although Furin knock-out mice exhibited early embryonic death during development, with multiple endothelial and heart defects (65), PC7 knock-out mice were viable and did not exhibit visible anatomical differences (66). Furthermore, other differences between PC7 and Furin began to surface upon analysis of their subcellular localization, which suggested that PC7 concentrates in less dense compartments than Furin (67). Therefore, it was of interest to identify specific PC7 functions, which would differentiate this enzyme from the other PCs and define its nonredundant physiological roles. Pro-EGF processing by PC7 was the first indication of a possible unique processing pathway of PC7. Whether this singular mode of action of PC7 is representative of its general physiological function(s) has yet to be explored in animal models and in human. In this context, we attempted to detect the EGF-115 form in the kidney of PC7 knock-out mice (data not shown). Unfortunately, the available commercial antibodies generated against mature 53-aa EGF (supplemental Fig. S2) do not recognize full-length mouse pro-EGF, and the only commercial mouse pro-EGF antibody (R&D Systems) does not work on Western blot (data not shown). Therefore, a CT-specific mouse pro-EGF antibody would be needed to show the implication of PC7 in pro-EGF processing in vivo. In addition, the lack of endogenous PC7-generated EGF-115 production in various cell lines (Neuro2A, HEK293, and ␤-TC3) that we tested did not allow the use of an effective siRNA approach (supplemental Fig. S3).
What are the domains of PC7 that confer to this enzyme its specific function on pro-EGF? We herein demonstrate that whereas the transmembrane domain of pro-EGF is dispensable for its processing by PC7 (Fig. 5A), anchorage of PC7 to the membrane is indispensable for its activity. To better understand the specificity of PC7 on pro-EGF processing, we examined the effect of post-translational modification of its cytosolic tail. Previously, Cys-palmitoylation of PC7 was reported to stabilize the enzyme without affecting its subcellular localization (58). Our data further show that the nonpalmitoylated PC7-(C699A,C704A) mutant (41) is as active on pro-EGF processing as the WT enzyme, and 2-bromopalmitate does not prevent pro-EGF processing (Fig. 6). These data demonstrate that Cyspalmitoylation does not regulate PC7 activity. Although the TMCT of PC7 is not dispensable, it can be replaced by that of Furin, at least for pro-EGF processing (Figs. 5B and 7B). In contrast, WT Furin or its PC7 chimeras cannot process pro-EGF (Fig. 7B). We conclude that because PC7 does not enhance the cleavage of pro-EGF (Fig. 5B), but PC7 activates the cleavage of both pro-EGF and its soluble pro-EGF forms (Fig. 5A), the production of EGF-115 requires the TMCT domain of PC7 but not that of pro-EGF. However, the fact that the TMCT of PC7 can be replaced by that of Furin, which is well known to be required for endocytosis of its cognate enzyme (14), also suggests that not just endocytosis is critical but that PC7 possesses specific catalytic properties to generate EGF-115 via its unique ability to activate a presumably latent proteinase.
Attempts to identify the PC7-enhanced cleavage site in pro-EGF were fraught with difficulties in view of the sensitivity of the folding of this protein to domain deletions and the presence of Gln within the N-terminal sequence of the EGF-115 product that blocked its Edman degradation. Nevertheless, combining two mutations, Q292M and K300M, allowed the identification of the cleavage site as Arg 290 2 within the sequence MVVHPR 290 2AQ (Fig. 8C). We suspected that this processing may not be performed by PC7 but rather by a PC7-activated latent intracellular protease because this motif is not a typical PC-cleavage motif (13). Indeed, the basic aa-specific PC7 still enhanced the processing of the pro-EGF mutant R290A (Fig.  8D), and pure soluble PC7 was not active in vitro on immunopurified pro-EGF (supplemental Fig. S5A) or when incubated with cells expressing pro-EGF at the cell surface (supplemental Fig. S5B). We conclude that the production of EGF-115 is performed by a PC7-activated proteinase, expressed in both HEK293 and Neuro2A cells, but much less so in COS-1 cells (Fig. 1A). The hypothesis of an indirect cleavage was supported from inhibitory studies whereby the blockade of EGF-115 production was achieved by a cysteine (E64d) and two serine protease (AEBSF and ␣ 1 -antitrypsin) inhibitors (Fig. 9A), which do not inactivate PC7 in vitro (Fig. 9B). The protein ␣ 1 -antitrypsin is the major inhibitor of neutrophil elastase and proteinase 3 (68,69), and it has been shown to also inhibit the type II transmembrane serine protease matriptase (70). Because matriptase-2 is expressed in kidney (71) and is processed at multiple Arg residues (72), this raises the possibility that latent matriptase-2 (also known as TMPRSS6), or a functional homologue, may be a potential candidate for activation by PC7, resulting in pro-EGF processing. Finally, we do not believe that the cognate enzyme activity requires a glycosylphosphatidylinositol-anchored cell-surface enzyme, like kidney prostasin (73), because production of EGF-115 was not affected by treatment of cells with phospholipase C (Fig. 9C).
We next demonstrated that processing occurs in a lipid-rich post-ER compartment and requires endocytosis from clathrincoated-pits of either pro-EGF, PC7, and/or the latent enzyme(s) to generate EGF-115 (Fig. 4F). However cleavage of pro-EGF does not require an acidic compartment, because incubation of cells with 5 mM NH 4 Cl had no effect on its processing (data not shown). Therefore, although the cleavage site of pro-EGF and the class of the intermediate enzyme(s) have been identified, the specific enzyme(s) and subcellular localization of the PC7activated processing reaction remain elusive.
Although mature EGF can effectively bind and activate the EGFR, the membrane-bound and shed EGF forms are also FIGURE 10. EGF-115 is highly expressed at the cell surface and enhances EGFR activation. A, expression of the pro-EGF-(RRRR 290 EL) mutant, which is cleaved efficiently by endogenous PCs in HEK293 cells, increases the production of EGF-115, as compared with either pro-EGF ϩ pIRES or co-expressed with PC7. B, cell surface EGF immunoreactivity (pro-EGF antibody) obtained by FACS analysis of HEK293 cells expressing PC7, pro-EGF, pro-EGF ϩ PC7, or pro-EGF-(RRRR 290 EL) shows a clear enrichment of immunoreactivity of the pro-EGF-(RRRR 290 EL) construct, likely because of the enhanced generation of EGF-115 at the cell surface. The mean intensity of fluorescence (MIF) values are plotted as fractions relative to the control sample (EGF ϩ pIRES) for each condition. Error bars represent the S.E. for triplicate measurements, and oneway analysis of variance followed by the Dunnett test was used (*, p Ͻ 0.05). C, cell-surface biotinylation of pro-EGF in HEK293 cells co-transfected with pro-EGF and either pIRES or PC7. Proteins were immunoprecipitated (IP) with streptavidin-agarose (SA) and analyzed by Western blotting (WB) using mAb V5-HRP. D, cell surface immunofluorescence of pro-EGF in cells expressing pro-EGF, pro-EGF ϩ PC7 or pro-EGF-(RRRR 290 EL) (red labeling) under slightly permeabilized cells, showing an increased staining at the cell surface in cells expressing pro-EGF-(RRRR 290 EL). Nuclei of transfected cells were marked by Hoechst 33258 staining (blue labeling). Bar, 10 m. E, half-life of pro-EGF and EGF-115 measured by Western blotting using mAb V5 after incubation of HEK293 cells expressing pro-EGF ϩ PC7 with 10 M cycloheximide for 2, 4, 8, and 16 h. The signal intensities are plotted as percentage of control samples (t ϭ 0). Error bars represent the S.E. for triplicate measurements. F, co-cultures of A431 cells, which express high levels of EGFR, with HEK293 cells in the presence of 20 ng of mouse recombinant 53-aa EGF (Rec EGF) or HEK293 cells co-expressing pIRES ϩ PC7, or pro-EGF and either pIRES, PC7, or sPC7, pIRES ϩ pro-EGF-(RRRR 290 EL). Western blot analysis of pro-EGF with mAb V5, compared with that of phosphorylated (P-) and nonphosphorylated ERK1/2 revealed that the generation of EGF-115 correlates with an induction of phosphorylation of ERK1/2 as quantitated at the bottom of the figure. The ratio of P-ERK/ERK was estimated by Scion image analysis and normalized to that obtained using pIRES ϩ pro-EGF (value ϭ 1). This is a representative example of an experiment repeated at least three times with consistently similar result. Stars point out on the increase of ERK1/2 phosphorylation. The percent of pro-EGF cleavage into EGF-115 is indicated below each lane. potent activators of this pathway (11, 74 -76). The pro-EGF mutant RRRR 290 EL, which is very efficiently processed by endogenous PCs in HEK293 cells into the EGF-115 (Fig. 10A), is more active than wild type EGF in stimulating the EGFR pathway, as evidenced by the increased levels of phospho-ERK1/2 (Fig. 10F). Our data suggest that activation of the EGFR is due to the presence of a high level of EGF-115 at the cell surface (Fig.  10C) rather than its higher ability to activate its receptor. We obtained very similar results with a RRRR 315 EL mutant of pro-EGF (data not shown), suggesting that cleavage of pro-EGF between aa 290 or 315 increases the activity of the remaining product through its enhanced cell surface localization.
Intriguingly, the biosynthetic pathway of pro-EGF maturation into different intermediate soluble and transmembrane forms has not been studied in detail (77)(78)(79). The cytoplasmic domain of pro-EGF has been described to have its intrinsic biological activity. For example, it negatively modulates the regulation of motility and elastinolytic invasion in human thyroid cancer cells (80) and induces a reduction in body and organ weights in transgenic mice, including smaller kidney and brain (81). In contrast, the role of the extracellular region of mouse pro-EGF, which contains nine EGF-like domains and eight LDL receptor class B repeats (82), has not been defined. Furthermore, no previous work addressed the functional role of aa 26 -290 of pro-EGF, a domain removed by the PC7-activated protease in our study. Because cleavage of this domain enhanced the cell surface localization of EGF-115, and its activity on EGFR, we presume that the presence of this N-terminal sequence is a negative modulator of pro-EGF cell surface localization, either by a conformational effect or through binding to an intracellular partner. Conceptually, because pro-EGF processing requires endocytosis into clathrin-coated vesicles (Fig.  4F), the N-terminal domain of pro-EGF may prevent its recycling to the cell surface. Hence, processing would release the N-terminal domain and result in EGF-115 as the major pro-EGF form at the cell surface (Fig. 10C).
In conclusion, our present data show the uniqueness of the ability of PC7 to enhance pro-EGF processing into EGF-115, which is found in the kidney but not in lacrimal glands, upon prior activation of a latent serine and/or cysteine protease(s). The data also suggest that the PC7-induced generation of EGF-115 may represent a regulatory mechanism of juxtacrine EGFR activation, at least in kidney.