Multistep Autoactivation of Asparaginyl Endopeptidase in Vitro and in Vivo*

Mammalian asparaginyl endopeptidase (AEP) or legumain is a recently discovered lysosomal cysteine protease that specifically cleaves after asparagine residues. How this unusually specific lysosomal protease is itself activated is not fully understood. Using purified recombinant pro-enzyme, we show that activation is autocatalytic, requires sequential removal of C- and N-terminal pro-peptides at different pH thresholds, and is bimolecular. Removal of the N-terminal propeptide requires cleavage after aspartic acid rather than asparagine. Cellular processing, either of exogenously added AEP precursor or of pulse-labeled endogenous precursor, introduces at least one further cleavage to yield the final mature lysosomal enzyme. We also provide evidence that in living cells, there is clear compartmental heterogeneity in terms of AEP activation status. Moreover, we show that human monocyte-derived dendritic cells harbor inactive proforms of AEP that become activated upon maturation of dendritic cells with lipopolysaccharide.

toxin C fragment, AEP introduces a small number of cleavages that "unlock" the folded protein substrate, making it available for class II MHC binding and, potentially, additional protease action (11,15). 2 In contrast, we have shown that AEP action on myelin basic protein can destroy a T cell epitope, preventing its display on class II MHC molecules and possibly the elimination of autoreactive T cells specific for this epitope (12). AEP can initiate processing of the invariant chain, although this key event in class II MHC maturation can also be performed by other enzymes (16). AEP/legumain has also been implicated in the activation of pro-gelatinase A (17), in ␣-thymosin processing (18), and as an inhibitor of osteoclast formation (19). Very recently, analysis of mice deficient in AEP demonstrated an accumulation of the single chain forms of cathepsins H, B, and L, changes in the activities of several lysosomal exopeptidases, and age-dependent perturbation of kidney proximal tubule lysosome function (20).
Thus AEP is emerging as a lysosomal enzyme whose specificity makes it ideal for the limited cleavages necessary for pro-protein processing and antigen presentation. How AEP is itself activated is of some interest because this may in turn control the activation/processing of downstream substrates including other lysosomal proteases. Like several members of the papain family of lysosomal cysteine proteases, activation of AEP is triggered by acidic pH and appears to be autocatalytic (21)(22)(23). In the case of the Arabidopsis enzyme (␥VPE), Kuroyanagi et al. (21) have shown that sequential removal of C-and N-terminal propeptides by cleavage after Asn occurs, although removal of the N-terminal peptide was not necessary for activity. Other plant VPEs may autoactivate through cleavages after aspartic acid (22). However, somewhat incomplete and conflicting data have appeared regarding the mechanism of autoactivation of mammalian AEP. One report documented autoactivation of an AEP-Ig fusion protein following cleavage after one or more aspartic acid residues (24). Another study, which followed the activation of pro-AEP in cell lysates, reported an obligatory cleavage after asparagine 323 leading to enzyme activation (25).
Here we analyze the autoactivation pathway of pro-AEP in vitro and in vivo. We show that sequential cleavages after both asparagine and aspartic acid are necessary for efficient enzyme activation resulting in the removal of a 110-residue C-terminal and a short 8-residue N-terminal pro-peptide, respectively. The key activating cleavage at Asp 25 only occurs below pH 5.0 and yields active enzyme, which is, however, still significantly larger than the final cellular form of AEP. The fully mature form is generated by living cells either from endogenous pro-AEP or following endocytosis and cellular processing of recombinant pro-AEP. AEP biosynthesis is induced as blood mono-cytes are differentiated into dendritic cells, but only partial activation occurs. Maturation with LPS triggers the appearance of fully mature and active enzyme. This multi-step, pHregulated activation pathway may therefore be used to control processing capacity for antigens and other AEP substrates.

MATERIALS AND METHODS
Cloning and Expression of the AEP Precursor-Total RNA was prepared from the EBV B cell line 4.2 (26) using RNAzol according to the manufacturer's conditions and used as template for reverse transcription-PCR. Specific AEP cDNA was generated using avian myeloblastosis virus reverse transcriptase and the reverse primer 5Ј-CCGCTCG-AGAAGCTTTCAGTAGTGACCAAGGCACAC-3Ј and subsequently amplified using forward primer 5Ј-CCGGAATTCATGGTTTGGAAAG-TAGCTGTATTCCTC-3Ј and reverse primer 5Ј-CCGCTCGAGAAGCT-TTCAGTAGTGACCAAGGCACAC-3Ј under the following PCR conditions: 1 cycle of 95°C for 2 min; 15 cycles of 94°C for 30 s, 65°C for 30 s (decreasing by 1°C/cycle), and 68°C for 3 min; 25 cycles of 94°C for 30 s, 50°C for 30 s, and 68°C for 3 min (ϩ10 s/cycle); and 1 cycle of 68°C for 10 min. AEP cDNA was gel-extracted and ligated using T4 DNA ligase into pCR3 vector (Invitrogen) using the XhoI and EcoR1 restriction sites underlined in the above primers. To generate a suitable vector for methotrexate-driven AEP overexpression the AEP insert was reamplified to include an optimal Kozak sequence at the 5Ј end and a His 6 C-terminal tag using forward primer 1 (5Ј-ACGATATCACCATG-GTTTGGAAAGTAGCTGTATTCC-3Ј) and reverse primer 1 (5Ј-ATCT-AGAGAATTCGCGGCCGCTCAATGGTGATGGTGATGGTGGTAGTG-ACCAAGGCACACGTG-3Ј). The product was cloned into the EcoRV and NotI sites (underlined in primers) of vector pcDNA DHFR (27) to generate the AEP expression vector pDT1. DHFR Ϫ CHO cells (European Collection of Animal Cell Cultures, UK) were transfected using the calcium phosphate method, and colonies were initially selected in Iscove's medium lacking thymidine and hypoxanthine plus dialyzed fetal calf serum. After ϳ2 weeks the surviving colonies were subjected to further selection in 20 nM methotrexate followed by stepwise increases to 100 nM, 500 nM, and finally 1 M. Single colonies were picked at 100 nM. The conditioned medium (5 l) and cell lysates (15 g) were screened by Western blotting using an anti-human AEP antibody raised in sheep (10). More than 20 overproducing cell lines were isolated in this way. For most experiments we used clones 2C3 or 3A5. We estimated up to 70-fold overexpression of active AEP within the cells and up to 10 -20 mg/liter of AEP precursor secreted into the growth medium (see Fig. 1). For production of the AEP precursor, the overproducing cells were adapted to CHO-S-SFMII serum-free medium (Invitrogen) by progressive reduction in the level of fetal calf serum (10%, 5%, 2.5%, 1%, and then 0%) with each step taking about 1 week. To purify the AEP precursor, up to 1 liter of medium, conditioned by cell growth for approximately 1 week, was adjusted to 1 M NaCl and passed over a nickel-agarose column. The column was washed with Tris, pH 7.4, 1 M NaCl, 10 mM imidazole. The AEP precursor was eluted in the same buffer containing 0.5 M imidazole, dialyzed against phosphate-buffered saline and stored at Ϫ20°C in 20% glycerol.
AEP Autoactivation and Assay Conditions-Autoactivation of AEP was performed by diluting aliquots of the 56-kDa AEP precursor into a buffer containing 0.2 M sodium citrate/citric acid plus 1 mM dithiothreitol adjusted to different pH values. After different time periods at 37°C, AEP enzyme activity was assessed by transferring aliquots of the autoactivation reaction to AEP assay buffer (28), which included 10 M Z-Ala-Ala-Asn-NHMec as substrate. Release of NHMec was measured using a Cytofluor plate reader. Note that the pH of the assay buffer (pH 5.8) is optimal for AEP activity measurement but prevents autoactivation, which requires lower pH values. AEP autocatalytic cleavage sites were identified following transfer of 56-, 47-, 46-, and 11-kDa products to polyvinylidene difluoride membrane followed by eight cycles of Edman degradation in an ABI sequencer.
The N-terminal propeptide VPIDDPED and its variant forms (synthesized by Dr. G. Bloomberg, Bristol University) were dissolved in water to a final concentration of 50 mg/ml, and the pH was adjusted by the addition of 2 M NaOH. Different concentrations were tested either on preactivated 46-kDa AEP or for their ability to inhibit the autoactivation of 56-kDa AEP. For the inhibition of autoactivation experiments, the 56-kDa AEP protein was incubated at pH 4.0 at a concentration of 0.5 mg⅐ml Ϫ1 in the presence of the reversible inhibitor Fmoc-AENKamide or a control peptide, Fmoc-AEQK-amide. The autoactivation reaction was stopped by diluting into pH 5.8 buffer and further diluted into assay buffer for measurement of AEP enzyme activity. The final concentration of inhibitor peptide in the assay buffer was 2.3 M, i.e. less than 10% that of the Z-Ala-Ala-NHMec substrate, which was adjusted to 25 M.
Site-directed Mutagenesis-Mutagenesis of Asp 21 and Asp 25 to alanine in pro-AEP was performed by PCR using forward primer 1 (see above) and mutagenic reverse primers that incorporated a unique DraIII site (underlined) located immediately downstream of the Asp 25 codon: Reverse 2, 5Ј-GATCACCACCCAGTGCTTGCCTCCATCTTCAG-GATCAGCTATAGGAAC-3Ј changing Asp 21 to Ala; Reverse 3, 5Ј-GAT-CACCACCCAGTGCTTGCCTCCAGCTTCAGGATCATCTATAGGAA-C-3Ј changing Asp 25 to Ala 25 ; and Reverse 4, 5Ј-GATCACCACCCAGT-GCTTGCCTCCAGCTTCAGGATCAGCTATAGGAAC-3Ј changing both Asp 21 and Asp 25 to Ala. The PCR products were introduced into pDT1 by fragment exchange utilizing EcoRV and DraIII restriction sites. Substitution of Cys 189 and of Asn 323 with Ser and Ala, respectively, was achieved using the QuikChange kit (Stratagene). All of the constructs were checked by DNA sequencing.
Expression of Mutant AEPs-The AEP mutants were expressed in 293T cells by transient transfection using Ca 2ϩ -phosphate. After 17-20 h the medium was replaced with fresh medium (Dulbecco's modified Eagle's medium, 25 mM Hepes, 10% fetal calf serum). Two days later the serum level was reduced to 5%, and medium was collected after 3-5 days. AEP mutants were purified by nickel-agarose chromatography as described above. Autoactivation conditions were as above at pH 4.0 or pH 4.5 for 2-5 h.
Antibodies-Antisera production was performed under contract with Diagnostics Scotland (Carluke). Sheep and mice were immunized with a mixture of 56-kDa AEP precursor and fully autoactivated 46-kDa AEP. Rabbits were immunized with the N-terminal peptide VPID-DPEDGGKHWV coupled through a C-terminal cysteine residue to maleimide-activated keyhole limpet hemocyanin. After two successive boosts, both the rabbit and sheep antibodies were affinity purified on the same peptide coupled to Sulfolink gel (Pierce). The sheep antibodies were subjected to repeated rounds (four in all) of purification until the reactivity to activated forms of AEP was removed. This antibody is named ␣AEP 18 -31 . Sheep antibodies were also affinity purified on immobilized 56-kDa AEP. Murine monoclonal antibodies were obtained by standard methods, screened by enzyme-linked immunosorbent assay, Western blotting, and immunofluorescence microscopy and were recloned twice.
Radiolabeling of 56-kDa AEP and [ 35 S]Met/Cys Pulse-Chase Labeling-Purified 56-kDa AEP precursor (6 -10 g) was radioiodinated in a final volume of 20 l of 0.25 M sodium phosphate, pH 7.2 containing 0.5 mCi of 125 I. After 15 min the reaction was terminated by the addition of 5 l of 20 mM tyramine and 5 l of 1 M KI. The labeled protein was isolated by chromatography on a 10-ml PD-10 column (Amersham Biosciences). To detect uptake of 125 I-AEP by wild type CHO cells, 10 6 cells were seeded in 6-well culture plates, and 0.2 g/ml 125 I-AEP was added in a final volume of 2 ml complete Dulbecco's modified Eagle's medium. Where indicated, inhibitors were added 1 h before the addition of 125 I-AEP. After different times the cells were washed with ice-cold phosphate-buffered saline, detached with phosphate-buffered saline with 10 mM EDTA, and collected by centrifugation. After two more washes, the cells were lysed in AEP assay buffer (28) but containing 1% CHAPS, 0.5% Triton instead of 0.1% CHAPS. The nuclei were removed by centrifugation, and the supernatant was analyzed by SDS-PAGE and autoradiography. In some experiments, labeled AEP was immunoprecipitated from postnuclear supernatants with affinity-purified antisera raised either in sheep or in rabbits. For [ 35 S]Met/Cys labeling, the 3A5 overproducing cell line was preincubated in RPMI 1640 lacking methionine and cysteine plus 10% dialyzed fetal calf serum for 1 h. 5 ϫ 10 6 cells were labeled with 0.25 mCi of [ 35 S]Met/Cys (Amersham Biosciences) for 5 min. The medium was removed, and after washing the cells were chased in complete Dulbecco's modified Eagle's medium including 5 mM methionine. After different times the cells were washed and lysed at 0°C in 20 mM Tris, pH 7.0, 0.15 M NaCl, 1% Triton X-100, 5 mM MgCl 2 plus protease inhibitors (Complete mixture; Roche Applied Science). After centrifugation to remove nuclei, labeled AEP was immunoprecipitated as above and analyzed on 12% Bis-Tris gels (Nu-PAGE; Invitrogen).
Dendritic Cell (DC) Culture-Blood from healthy donors was used to prepare peripheral blood monocytic cells. CD14 ϩ monocytes were isolated with magnetic beads (Miltenyi) according to the manufacturer's protocol and cultured at 4 ϫ 10 5 /ml with 800 units/ml granulocyte macrophage colony-stimulating factor (Leucomax, Sandoz) plus 1000 units/ml interleukin-4 (BD Biosciences) essentially according to the protocol of Sallusto and Lanzavecchia (29). DCs were CD1a-positive, and expression of CD86 and class II MHC increased following LPSdriven maturation (1 g/ml; Sigma).

RESULTS
Production and Purification of Homogenous AEP Proenzyme-To resolve some of the issues concerning the autoactivation of asparaginyl endopeptidase, we developed an expression system that permitted the purification of large amounts of the proenzyme. We took advantage of the potential that gene amplification offers in CHO cells for protein overproduction (30). The AEP cDNA was cloned into the vector pcDNA DHFR (27), DHFR Ϫ CHO cells were transfected, and initial selection was performed as described under "Materials and Methods." Selected colonies were then subjected to stepwise selection with increasing concentrations of the DHFR inhibitor methotrexate until the desired level of AEP production had been achieved. Amplification of the DHFR gene was confirmed by fluorescence-activated cell sorter analysis of cells labeled with fluorescinated methotrexate (Ref. 31 and data not shown), and parallel AEP overproduction was demonstrated by Western blotting of both the cells and growth medium. As shown in Fig.  1 AEP was readily detected in 5 l of conditioned growth medium and in cell lysates from various selected clones. Parallel electrophoresis of purified pig kidney AEP indicated that overproduction in the cells was considerable (Fig. 1a). The AEP protein found in the growth medium was the proenzyme (56 kDa). We found no evidence of secretion of the mature enzyme (36 kDa), which was confined to cell lysates and migrated somewhat more slowly than purified mature pig kidney enzyme run in parallel (Fig. 1). To estimate the overproduction of AEP more accurately, we measured the hydrolysis of the AEP substrate Z-Ala-Ala-Asn-NHMec by CHO cell lysates. As shown in Fig. 1b, the initial rates of hydrolysis of this substrate indicated that at least 40-fold overproduction of active enzyme had been achieved in the cell lysates alone. We included a C-terminal histidine tag to aid purification of the proenzyme from CHO cell supernatants.
Stepwise Autoactivation of AEP at Acidic pH-We next attempted to autoactivate the purified proenzyme. As shown in the Coomassie Blue-stained gels in Fig. 2a, exposure to pH values in the range 6.0 -5.5 resulted in the partial disappear-ance of the 56-kDa precursor and the appearance of 47-and 11-kDa products. Interestingly, little enzyme activity appeared, even at pH 5.0 where a significant amount of the 56-kDa precursor became converted. Similarly, below pH 3.5 we observed conversion of the 56-kDa precursor, but again enzyme activity failed to appear. As shown in Fig. 2b, active enzyme only appeared in a pH "window" between 4.5-3.5 and paralleled a small but clear increase in mobility of the major product of autocatalysis from 47 to 46 kDa (Fig. 2a). We sequenced the N termini of all the products generated at pH 5.8 and 4.5. As shown in Fig. 2a, whereas the inactive 56-and 47-kDa forms of AEP had the N-terminal sequence VPIDDP, the active 46-kDa form of AEP began with the sequence GGKHWV. Thus the inactive 47-kDa form of AEP was extended by 8 residues at the N terminus compared with the active 46-kDa form (Fig. 2c). We also sequenced the smaller 11-kDa processing product generated over a wide range of pH values (Fig. 2a). This product began with the sequence DLEESR demonstrating autocatalytic cleavage after Asn 323 as previously reported by Chen and colleagues (25). However, in our hands it appeared to be the subsequent N-terminal cleavage after Asp 25 that led to enzyme activation ( Fig. 2 and see below).
The Final Step in AEP Maturation Is Not Autocatalytic but Occurs Following Normal Cellular Processing-AEP found in living cells has a mass of ϳ36 kDa, considerably smaller than the autocatalytically activated 46-kDa material. No processing beyond the 46-kDa product occurred in vitro even upon prolonged incubation. To test whether the purified recombinant 56-kDa proenzyme was capable of being fully processed by living cells, we labeled it with 125 I and offered this as a substrate for endocytic capture and processing by living CHO cells. We anticipated that the secreted 56-kDa precursor would still carry the necessary targeting signals, most likely mannose-6phosphate, and that they might be recognized by receptors cycling between the cell surface and endosomes (32). As shown in Fig. 3a  added exogenously to living cells. We conclude that final maturation of AEP is most likely not autocatalytic but mediated by other lysosomal proteases. However, the identity of these enzymes remains uncertain. Neither leupeptin, E64 (Fig. 3a), or pepstatin (not shown) had much impact on the maturation of 56-kDa AEP, although in some experiments evidence of accumulation of a form slightly larger than 36 kDa was obtained with high concentrations of leupeptin (Fig. 3a). These inhibitors were active because the 11-kDa C-terminal propeptide accumulated to a greater extent in leupeptin and E64-treated cells (Fig. 3a). In addition, there was a noticeable increase in the amount of 56-kDa precursor associated with the cells. We suggest that a small proportion of the iodinated 56-kDa precursor fails to autoactivate and that this material is normally destroyed by leupeptin-and E64-sensitive lysosomal proteases that also degrade the 11-kDa C-terminal propeptide.
We were surprised to find no evidence for intermediate 46/ 47-kDa forms of AEP in the experiments described above. Most likely in the minimum time needed (1 h) to see accumulation of endocytosed AEP, conversion has already occurred. To establish that the 47-and 46-kDa forms are bona fide intermediates in AEP autoactivation in living cells, we also examined the maturation of biosynthetically labeled AEP. To distinguish between inactive (56 and 47 kDa) and active (46 and 36 kDa) forms of AEP, we raised an antiserum in rabbits against a synthetic peptide corresponding to the N terminus of the 56/ 47-kDa forms and a sheep antiserum that recognizes all forms of AEP. As expected, the rabbit serum did not recognize the active 46-kDa (or 36-kDa) species, whereas the sheep serum did (Fig. 3b). The AEP-overexpressing CHO cell line 3A5 was pulsed with [ 35 S]Met/Cys for 5 min and chased for various times, and the status of AEP was examined by sequential immunoprecipitation with the rabbit followed by the sheep antiserum. After the short pulse, most of the AEP migrated at the 56-kDa precursor size and over a 4-h period chased to the fully mature 36-kDa form, which was precipitated by the sheep but not the rabbit serum (Fig. 3c). However, the rabbit serum, raised against the N-terminal peptide, was able to precipitate a processed form of AEP that appeared transiently after 15 min of chase and that migrated around 47 kDa (Fig. 3c). Thus the C-terminally processed but N-terminally intact 47-kDa intermediate seen in vitro is also seen in living cells. Intermediates between the 46/47-kDa forms and the 36-kDa form were seen in this and other experiments, suggesting that the 36-kDa form may arise by processive C-terminal trimming of the 46-kDa form (Fig. 3c). As noted above, however, the enzyme or enzymes involved remain to be identified. In addition, this additional processing seems not to alter enzyme activity because the specific activity of 46-kDa human AEP activated in vitro was very similar to that of the fully mature 36-kDa enzyme purified from pig kidney (1 and data not shown).
Cleavage at Asp 25 Correlates with Appearance of AEP Enzyme Activity-We investigated further the relationship be-tween removal of the N-and C-terminal propeptides and the appearance of enzyme activity in vitro. We generated various mutant forms of AEP, expressed them in 293T cells, and purified them from the culture supernatant. Following a shift to pH 4.0, we assessed autocatalytic processing and the appearance of enzyme activity. As previously reported (25) mutation of the active site cysteine residue Cys 189 or of Asn 323 , the target for removal of the C-terminal propeptide, blocked autocatalytic processing and the appearance of enzyme activity (Fig. 4, a and  b). However, in the case of the Asn 323 mutant, we consistently observed the appearance of enzyme activity following long incubations at pH 4.0 (Fig. 4b, filled circles). N-terminal sequencing of this mutant indicated that the appearance of enzyme activity correlated with cleavage at Asp 25 . For example, after a 16-h incubation, 34% of the protein had the mature N terminus (GGKHW), and 36% of wild type enzyme activity had appeared. This suggests that cleavage at Asn 323 , although profoundly rate-limiting, is not essential for enzyme activity. To test directly the need for cleavage at Asp 25 , we performed a further experiment with the N323A mutant where the reversible inhibitor of AEP, Fmoc-AENK-amide, was included during the slow time course of activation and then diluted out, and enzyme activity was assessed at pH 5.8. As shown in Fig. 4c, the appearance of enzyme activity was abolished under these conditions (Fig. 4c), providing further evidence that cleavage at Asp 25 is required. As expected, the inhibitor also profoundly slowed the activation of the wild type enzyme as well.
We next tested the effect of mutating Asp 25 . Because after shorter times of autoactivation we also observed a small mole fraction of the N-terminal sequence DPEDG indicating cleavage after Asp 21 (Fig. 2c), we also mutated this residue. In fact this mutation made little difference either to autocatalytic processing (Fig. 4a) or to the appearance of enzyme activity (Fig. 4b). In contrast mutation of Asp 25 (with or without simultaneous mutation of Asp 21 ) led to the appearance of a slower migrating 47-kDa species upon shift to pH 4.0, confirming that Asp 25 is recognized during autocatalytic cleavage of the Nterminal propeptide in vitro (Fig. 4a). When we examined the appearance of enzyme activity, the rate of autoactivation of the D25A and D21A,D25A mutants was significantly slower (Fig.  4b). Nonetheless, given the data already described correlating cleavage at Asp 25 with appearance of enzyme activity, we were surprised that the Asp 25 mutants activated as well as they did. One possibility is that the mutation not only prevents cleavage but also disrupts an important interaction of the N-terminal propeptide with the enzyme. To try to address this we tested the effect of adding back to activated enzyme the wild type 8-mer peptide or variants thereof. As shown in Fig. 4d, the wild type sequence VPIDDPED was able to inhibit AEP hydrolysis of the Z-AAN-NHMec substrate and indeed the pH-triggered autocatalytic cleavage of wild type AEP (Fig. 4e). In contrast, a randomized version of the same sequence was completely inactive. Moreover, a peptide that corresponded to the D21A,D25A mutant (VPIADPEA) also failed to inhibit AEP activity as did the Asp to Glu variant VPIEDPEE. These results show that the N-terminal propeptide can act as a specific inhibitor of the active enzyme, although fairly high concentrations are required. In addition, this experiment is consistent with the hypothesis that mutagenesis of Asp 25 may not only block autocatalytic cleavage but also the ability of this sequence to interact effectively with the enzyme. This may explain why enzyme activity appeared to the extent it did with the Asp 25 mutant.
Visualization of Active and Inactive AEP in Living Cells-We asked whether inactive and active forms of AEP might be visualized in living cells. Because the levels of AEP in most FIG. 3. AEP maturation in living cells. a, 125 I-labeled 56kDa AEP precursor (0.2 g/ml) was added to living CHO cells for the times indicated with or without the cysteine protease inhibitors E64 (0.1 mM) or leupeptin (1 mM). Capture and maturation of AEP was assessed by SDS-PAGE and autoradiography. The ability of the radioiodinated 56-kDa precursor to autoactivate in vitro is shown in the leftmost two tracks. b, to verify the specificity of the two affinity-purified antisera, 56-and 46-kDa AEP were transferred to nitrocellulose, probed with the rabbit (N terminus-specific) antiserum, and revealed using a peroxidase-conjugated secondary. The blot was stripped and reprobed with the sheep antiserum. The corresponding Coomassie-stained gel is shown on the left. c, pulse (5 min) and chase (times shown) analysis of AEP maturation in CHO-3A5 cells. AEP products were immunoprecipitated (IP) as shown and analyzed by SDS-PAGE. cells are too low to permit visualization of transient maturation intermediates, we used a previously described melanoma cell line SD1 (12) that overexpresses AEP, although not to the extent seen in the CHO cell lines. Because of background problems with the N-terminus-specific rabbit serum, we purified a subfraction of the polyclonal sheep serum using the  ‚). The autoactivation reaction was stopped by diluting into pH 5.8 buffer and further diluted into assay buffer for measurement of AEP enzyme activity. The final concentration of inhibitor peptide in the assay buffer was 2.3 M, i.e. less than 10% that of the Z-Ala-Ala-NHMec substrate, which was adjusted to 25 M. As control for this residual inhibitor, the control peptide-treated samples were also adjusted to 2.3 M AENK (q and ⅜). d and e, the N-terminal propeptide VPIDDPED and variant forms VPIEDPEE, VPIADPEA, and DDVPIEDP were included at the concentrations shown in a standard AEP assay at pH 5.8 (d) and added to an AEP autoactivation experiment (e). The extent of autoactivation was assessed on Coomassie Blue-stained gels.
immobilized N-terminal peptide as adsorbent (see "Materials and Methods"). We were able to isolate antibodies that only reacted with the 56-and 47-kDa forms of AEP and that labeled a distinct set of compartments in living SD1 cells. All of these compartments were also labeled by a mixture of two monoclonal mouse anti-AEP antibodies, 11B7 and 6E3, reactive with all forms of AEP ( Fig. 5 and data not shown). However, the latter antibodies also identified a second set of compartments that were not labeled by the N-terminal pro-peptide-specific antibodies. These vesicles occupied a more perinuclear position and, unlike most of the compartments labeled by the propeptide specific antibodies, were LAMP-1-positive and therefore presumably lysosomes (Fig. 5). Thus there is heterogeneity among the endosomal/lysosomal compartments harboring the AEP protease; some clearly contain significant levels of the precursor forms, whereas others, which are mostly LAMP-1 positive, appear to have fully converted the enzyme to the mature form by removal of the N-terminal propeptide.
Evidence That AEP Autoactivation Can Proceed in a Bimolecular Fashion-In the case of the papain-like C1 family of lysosomal cysteine proteases, the precise mechanism of autoactivation has been difficult to pin down with both unimolecular and bimolecular mechanisms being proposed (33)(34)(35)(36). In a bimolecular reaction, the rate of appearance of enzyme activity should accelerate over time and be sensitive to the concentration of proenzyme. 56-kDa AEP, at various concentrations, was shifted to pH 4.5, and autoactivation was allowed to proceed for different times, after which the enzyme was diluted in pH 5.8 assay buffer, preventing further autoactivation. As shown in Fig. 6a, the rate of appearance of enzyme activity accelerated with time, consistent with a bimolecular mechanism. In addition, the rate of appearance of enzyme activity was markedly sensitive to pro-enzyme concentration. At 40 g/ml almost 10% of the enzyme had become activated in 10 min, whereas at 1 g/ml only about 0.5% had been activated (Fig. 6b).
A bimolecular mechanism should allow catalytically inactive enzyme to be processed in the presence of active enzyme. We tested this in living CHO cells by feeding exogenous 125 I-labeled wild type AEP or mutant C189S AEP. As shown in Fig.  6c, both forms of the enzyme were taken up, were processed to the mature 36-kDa size, and were precipitable with sheep serum but not with the rabbit serum raised against the Nterminal pro-peptide. Presumably, the C189S mutant is processed by endogenous active AEP in a bimolecular fashion. We performed the same experiment in the monocytic cell line THP-1, which has low levels of endogenous AEP. 3 These cells took up the AEP precursor very inefficiently but nonetheless revealed that although there was some processing of the mutant enzyme, a partially processed form precipitable with the rabbit antiserum also accumulated, i.e. retaining the N-terminal propeptide (Fig. 6c). Thus the rate of AEP processing in vivo is also dependent upon AEP activity. In the case of C189S, which has no intrinsic enzyme activity, there is sufficient endogenous AEP to catalyze efficient bimolecular processing in CHO cells but not in THP-1. Conversely, the wild type enzyme is equally efficiently processed in either cell type.
Taken together the data are consistent with a bimolecular mechanism of AEP activation both in vitro and in vivo. This does not, however, exclude the possibility that a unimolecular process can proceed at a slower rate and may precede a bimolecular phase. Activation of AEP Is Regulated during Dendritic Cell Maturation-Cells may need to regulate the proteolytic capacity of their endosome/lysosome system under different conditions. This could be controlled at the level of protease expression by expression of endogenous inhibitors such as cystatins (37) or by regulation of zymogen activation, which, as shown above, can progress in a stepwise fashion with different potential regulatory points. There is evidence that DC may use cystatins to modulate their cysteine protease activity and that this may contribute to regulation of class II MHC expression following LPS-driven maturation of immature DC (38). Using serial analysis of gene expression Hashimoto et al. (39) have observed that AEP/legumain expression is induced as monocytes differentiate into macrophages. We therefore asked whether AEP maturation was regulated during human DC development and maturation.
CD14 ϩ monocytes from human peripheral blood were cultured in granulocyte macrophage colony-stimulating factor plus interleukin-4. At intervals, the cells were removed for analysis of AEP induction and maturation state by Western blotting. After 6 days, half of the cells were exposed to LPS to drive maturation. For reasons that are not clear, the results of this analysis were somewhat variable from donor to donor in terms of levels of AEP induction. Nonetheless, as shown in Fig.  7 (left panel), in DC from some donors we could observe a clear induction of AEP expression during the culture period consistent with the serial analysis of gene expression of Ref. 39. Interestingly, following LPS challenge, there was a clear disappearance of the 56-kDa inactive form and a corresponding increase in the fully active 36-kDa form. This general pattern was replicated in a second experiment (Fig. 7, right panel) in which the effect of LPS was examined after 16 and 40 h. Once again, there was a clear conversion of inactive AEP to active enzyme. Thus in primary dendritic cell cultures the induction and activation of AEP can be regulated by appropriate physiological stimuli.

DISCUSSION
Mammalian AEP, also known as mammalian legumain, is a lysosomal cysteine protease unrelated to the papain-like lysosomal cathepsins. It has unusually precise specificity for cleavage after asparagine and selectivity even among asparagine containing sequences (1,10,41). It is involved in both immune and autoimmune antigen processing reactions (10 -14) and also can perform activating cleavages in other substrates such as progelatinase A and prothymosin ␣ (17,18). All of the evidence to date indicates that AEP has evolved to introduce limited cleavages into folded protein substrates. The plant form of AEP, legumain, has also been shown to be responsible for pro-protein activation (42). How this enzyme is itself activated is therefore of some interest. In agreement with previous work, we find that activation is autocatalytic and triggered by acidic pH (21)(22)(23)(24)(25). Using highly purified human proenzyme, we show in addition that activation is bimolecular and requires the sequential removal of C-and N-terminal propeptides at different pH thresholds by cleavage after asparagine 323 and aspartic acid 25, respectively. Several considerations support the conclusion that the latter is the key activating cleavage. First, the inactive 47-kDa form was extended by 8 N-terminal residues relative to the active 46-kDa form. Second, incubation of the N323A AEP mutant at activating pH resulted in very slow appearance of enzyme activity that nonetheless correlated with cleavage at Asp 25 . The reversible inhibitor AENK blocked both this cleavage and the appearance of enzyme activity. Third, the addition of the 8-mer N-terminal propeptide, VPIDDPED, to active enzyme inhibited the enzyme in a dose-dependent fashion. Randomization of this peptide or alteration of the key terminal Asp residue prevented inhibition. Finally, the N-terminal sequence we observe for the active 46-kDa form fits precisely with the known N terminus of the enzyme isolated from mammalian cells (1).
We believe that our results can reconcile the somewhat dis-FIG. 6. AEP autoactivation is a bimolecular reaction. a, the AEP precursor (1 g/ml) was shifted to pH 4.5 to trigger autoactivation. At each time point identical amounts of enzyme (5 ng) were transferred to pH 5.8 assay buffer, and an AEP activity assay was performed for 20 min. b, AEP precursor was shifted to pH 4.5 at the different concentrations shown, and autoactivation was allowed to proceed for 10 min. Equal amounts of enzyme were transferred to an AEP assay at pH 5.8 as in a. To compare the extent of activation at different concentrations of 56-kDa precursor, the activity obtained in 2 h at a concentration of 40 g/ml was set to 100% because under these conditions AEP was fully activated. c, CHO cells and THP-1 cells were incubated for 6 h in the presence of 0.2 g/ml 125 I-labeled 56-kDa wild type (WT) or C189S AEP. Captured AEP was sequentially immunoprecipitated by first rabbit (R) and then sheep (S) antisera and visualized by SDS-PAGE and autoradiography. crepant earlier reports on activation of mammalian AEP/legumain. These studies reported autocatalytic cleavages after either aspartic acid or asparagine (24,25). We find that both are required. Halfon et al. (24) observed several autocatalytic cleavages at aspartic residues in an AEP-Ig fusion protein, but it was not clear which were activating. In contrast Chen et al. (25) demonstrated the same autocatalytic cleavage at Asn 323 that we report here, and they also detected an additional increase in mobility below pH 5.0 concomitant with enzyme activation. However, they attributed this to further processing at the C terminus and concluded that removal of residues at the N terminus was not necessary for activity. However, a second C-terminal processing event to account for enzyme activation seems more difficult to envisage because there is no Asn (or Asp) residue upstream of Asn 323 that could be autocatalytically cleaved to account for the size reduction observed from 47 to 46 kDa. The present data bear comparison with those of Kuroyanagi et al. (21), who reported a similar stepwise removal of C-terminal followed by N-terminal propeptides during autoactivation of the Arabidopsis AEP, ␥VPE. Interestingly, however, this plant enzyme differs from mammalian AEP in that both cleavages are made after Asn, and the acquisition of enzyme activity correlates solely with cleavage at the C terminus.
Our data support a scheme for AEP autoactivation summarized in Fig. 8. The key autoactivating cleavage occurs not after asparagine but after aspartic acid. Moreover, this cleavage only occurs at lower pH values. One possibility is that lower pH is required to protonate the aspartate side chain, thus making it resemble an asparagine side chain. Indeed, a synthetic peptide corresponding to residues 18 -31 of pro-AEP (VPIDDPEDG-GKHWV) was hydrolyzed more efficiently by activated AEP at pH 4.0 than at pH 5.8 (data not shown). Other residues in this very acidic sequence may also need to be protonated. Both the major Asp 25 and the minor Asp 21 processing sites we observed are preceded in the P3 position by proline. Interestingly, other studies have noted some preference for proline in the P3 posi-FIG. 8. Suggested scheme for cellular processing of AEP. The inactive 56-kDa proenzyme, resident in prelysosomal compartments, is progressively converted to active 46-kDa AEP by sequential autocatalytic cleavage at Asn 323 followed by Asp 25 . Subsequent C-terminal trimming by other lysosomal proteases yields the final mature 36-kDa form but does not result in any further increase in enzyme activity. Autoactivation is highly pH-dependent and may in part be controlled by endosomal acidification or progress through the endosome/lysosome system, because the 56/47-kDa species are not present in LAMP-1 ϩ vesicles. This represents a possible mechanism for the control of AEP activation during maturation of monocyte-derived dendritic cells. tion before canonical Asn cleavage sites (41,43). Detection of cleavage at Asp 21 suggests that two sequential cleavages at Pro-Xaa-Asp motifs can occur, each removing 4 amino acids. Because the 8-mer peptide is inhibitory, cleavage in two steps might serve to inactivate it. However, mutagenesis of Asp 21 did not affect the appearance of enzyme activity, so cleavage at this site is not obligatory. Below pH 3.0, the activating cleavage at Asp 25 was no longer observed. Possibly the titration of other residues around the active site now precludes autocatalysis at Asp 25 . We have also noted that AEP can cleave some protein substrates at Asp residues. 2 Lysosomal cysteine proteases of the papain family are generally autoactivated following removal of a single long (50 -100 residues) N-terminal propeptide (44 -47). In this context AEP is unusual in requiring the removal of two propeptides, one C-terminal and one N-terminal. Nor are there close similarities with the activation pathways of the other cysteine proteases in clan CD. These include the caspases, separase, and gingipain. Caspases differ in their mode of activation with executioner caspases (e.g. caspases 3 and 7) being activated by single proteolytic cleavages performed by upstream "apical" caspases (e.g. caspases 8 and 9) (48). Recent data suggest that the latter are not primarily activated by cleavage but rather by dimerization, with cleavage conferring other properties such as stability and sensitivity to inhibitors (49). In the case of separase, cleavages occur but may not be essential for activity (50). The activation of Arg-gingipain has also recently been analyzed in vitro. Like AEP, activation requires autocatalytic removal of both N-and C-terminal propeptides (51). However, the order of removal is reversed, with removal of a 229-residue N-terminal peptide preceding removal of 72 residues at the C terminus. Unlike AEP, the zymogen form of Arg-gingipain showed significant catalytic activity (51).
Similar to Arg-gingipain, the rate of AEP autoactivation was sensitive to the concentration of the precursor, suggesting that activation proceeds most efficiently as a bimolecular (i.e. intermolecular rather than intramolecular) process. In the case of the lysosomal cathepsins whose autoactivation has been studied, both unimolecular and bimolecular mechanisms have been proposed (33)(34)(35)(36). However, for those procathepsins whose structures are known (cathepsins B, L, and K) (44 -47), a bimolecular mechanism is favored because the active site, which is fully formed in the proenzyme, is distant from the pro/mature junction site(s) that are cleaved during autoactivation. In addition the propeptide passes through the active site cleft in the opposite orientation to a normal substrate peptide. Nonetheless, evidence has been recently presented for an activated intermediate for cathepsin B and S where the propeptide was cleaved not at the pro/mature junction but rather upstream at sites that are known (cathepsin B) or predicted (cathepsin S) to be in close proximity to the catalytic center (36). This cleavage, despite the reversed orientation, was postulated to be a unimolecular event. Thus the relative contributions of unimolecular and bimolecular events in cathepsin activation are not fully resolved. In the case of AEP, evidence for a bimolecular process was found both in vitro and in vivo. It remains possible, however, that there is also a unimolecular pathway. Structural analysis of the different intermediate forms should help resolve some of these outstanding questions.
It is intriguing to speculate that cells might exploit the pH-staged activation of enzymes like AEP to regulate proteolysis. For example, it is known that in immature DC antigen can be endocytosed and stored for several days in an unprocessed form (52,53). Upon maturation of the cells, for example with LPS, the antigen (hen egg lysozyme in the best studied case) is processed and the peptides are loaded onto class II MHC. In addition to differential expression of endogenous protease inhibitors such as cystatin C (38) and variations in class II MHC trafficking patterns (54,55), regulation of protease activation might also contribute to altered antigen processing and presentation as dendritic cells mature. We show here that AEP biosynthesis is induced as monocytes differentiate into dendritic cells, confirming earlier results from serial analysis of gene expression (39). However, we found that immature DC can harbor significant amounts of inactive AEP. Upon LPSdriven maturation, AEP activation is completed. Trombetta et al. (56) have recently reported a similar phenomenon in murine DC and have also shown that such enhanced proenzyme processing and enhanced antigen processing capacity is due to enhanced acidification of the endosome/lysosome system. Thus by setting pH thresholds for activation of protease zymogens, cells may fine tune the proteolytic capacity of the endocytic pathway to meet particular biological requirements. The pathway of AEP activation described here offers a clear example of this.