Characterization of a human angiotensinogen cleaved in its reactive center loop by a proteolytic activity from Chinese hamster ovary cells.

Angiotensinogen, the renin (E.C. 3.4.23.15) substrate, belongs to the serpins superfamily and has been classified as a noninhibitory serpin. Using mass spectroscopy, angiotensinogen purified from Chinese hamster ovary cell supernatant shows a broad spectrum. The absence of protease inhibitors throughout the purification leads to an angiotensinogen cleaved within the reactive center loop. This cleavage does not affect the Ang I generation because kinetic parameters are similar to the values of the full-length angiotensinogen. Although cleavage is complete, the cleaved angiotensinogen migrates after deglycosylation on SDS-polyacrylamide gel electrophoresis as a doublet differing by 4 kDa. To test whether the circulating angiotensinogen is cleaved in the reactive center loop, it was purified from a pool of human plasma and was shown to be uncleaved. Its migration was obviously slower than of cleaved angiotensinogen but also consisted of two bands pointing to a so far unexplained residual heterogeneity. We then compared the heat-induced polymerization of full-length- and reactive center loop-cleaved angiotensinogens. Both monomers were able to aggregate, revealing a particular behavior of angiotensinogen distinct from that of reactive center loop-cleaved serpins. Lacking the three-dimensional structure of angiotensinogen, we propose and discuss a structural model of the serpin fold within the renin substrate.

action in human plasma is not a zero-order kinetic reaction (1); the angiotensinogen concentration is therefore rate-limiting in the first step of the enzymatic cascade of the renin-angiotensin system.
The AGT gene has been implicated in human hypertension (2). A common molecular variant in which the methionine residue in position 235 is changed to a threonine residue (M235T) has been found to be associated with a 10 -20% increase in the level of plasma AGT and with an increased prevalence of hypertension.
The present study focuses on the relationship between AGT and serpins. Rat (3) and human (4) AGT genes have been classified in the serpin superfamily (5) because of some limited but significant homologies with sequences from representative members of this superfamily. These include ␣1-antitrypsin (AT) (23%), ovalbumin (21%), and antithrombin III (18%). In addition, rat (6) and human (7) AGT genes show a structural gene organization comprising 4 introns and 5 exons similar to that of the AT gene. The AGT gene could have arisen from the common ancestral serpin gene several million years ago via a duplication followed by a divergent evolution (8).
The members of the serpin superfamily have a common tertiary structure based on a central ␤-sheet (sheet A) surrounded by two other ␤-sheets (sheet B and C) and a mobile reactive center loop (RCL) (9). This RCL acts as a bait for trapping the target protease, and only for inhibitory serpins, the cleavage induces a serpin structural rearrangement. The insertion of the RCL into ␤-sheet A leads to the formation of a covalent complex with the target proteinase, thereby rendering it inactivated (10,11).
Biochemical evidences using pure AGT have shown that renin substrate has kept a serpin-like structure, particularly in having an RCL that has been probed by limited proteolysis. Indeed, two endoproteases, human neutrophil elastase (12) and Staphylococcus aureus endoprotease V8 (13), are able to cleave AGT in the RCL, leading to a proteolysis-resistant shortened protein and to C-terminal peptides, identified by N-terminal sequencing. Cleavage of AGT within the RCL by neutrophil elastase (Thr 410 -Gln 411 ) or V8 protease (Glu 408 -Ser 409 ) does not change its thermal stability contrary to inhibitory serpins (12,13). Thus, AGT has been classified in the non-loop-insertable serpins subfamily (14), characterized by their canonical model, ovalbumin (15). Moreover, the AGT target enzyme is still unknown.
Because AGT purified from human plasma often displays a classical doublet around 60 kDa differing by 4 kDa (16 -18) and has only one N-terminal sequence, it has been proposed that a partial C-terminal cleavage (19) could explain why these two bands persist even after extensive chemical or enzymatic deglycosylation (20). We provide here evidence for an in vitro RCL cleavage of human AGT and analyze its biochemical characteristics. Finally, we propose a three-dimensional model of the serpin fold of the renin substrate.

EXPERIMENTAL PROCEDURES
Enzymes, Proteins, and Antibodies-Human recombinant renin (21) was provided by Hoffmann-La Roche. Endoglycosidase H, neuraminidase, and O-glycosidase came from Roche Molecular Biochemicals; endoglycosidase F 1 fused to the glutathione S-transferase (22), also provided by Hoffmann-La Roche, were used as a mixture of deglycosidases. Proteins were measured with the Bio-Rad protein assay using bovine serum albumin as standard (23). The Ang I polyclonal antibody (N-1345) used in this study has been previously described (24). One Cterminal antibody (C-1350) was specifically raised against the last six residues of the mature AGT using a synthetic peptide (hAGT Tyr-447-452, NH 2 -Tyr-Asn 447 -Pro-Leu-Ser-Thr-Ala 452 -COOH) in which a tyrosine residue was added for labeling. The peptide was covalently coupled to keyhole limpet hemocyanin as carrier protein and injected into rabbits. The antibody against this peptide were characterized by enzymelinked immunosorbent assay (specific titer: 1/2000) (ANAWA Trading SA, Biomedicals Services and Products-Zurich). The specificity of the reaction against whole AGT was checked by negative controls with preimmune serum from the same rabbit and competition with peptide concentrations in the range 10 nM-1 M.
Large Scale Production of Recombinant AGT-The CHO cell line overexpressing wild type human AGT (Met 235 ) has been previously characterized (25,26). For scale-up cultures, cells were primarily expanded as a layer of adherent cells in medium containing 2% (v/v) fetal calf serum and 500 g/ml Geneticin. Production of recombinant AGT was performed in cell factories (Nunc, Roskilde Denmark) in a serumfree medium called DHI-TIPP, which is a mixture of Dulbecco's modified Eagle's, Ham's F-12, and Iscove's modified Dulbecco's media in respective proportions (1/1/2 (v/v) containing 5 mM L-glutamine and 25 mM glucose. It was supplemented with insulin (5 g/ml), selenite (20 nM), ethanolamine (20 M), human transferrin (6 g/ml), Primatone RL ultrafiltered (2.5 mg/ml), and Pluronic F68 (0.1 mg/ml). On day 0 the supernatant was discarded, cells were washed twice with 1 liter of DHI-TIPP medium containing no fetal calf serum, and AGT production was then performed in this serum-free medium. Supernatants were harvested 3 or 4 days later, and protease inhibitors were immediately added: 1 mM phenylmethylsulfonyl fluoride, 0.5 M aprotinin, and 9.3 mM benzamidine. Supernatants were centrifuged (20 min, 3000 rpm at 4°C), concentrated using a 10-kDa cut-off membrane, and then frozen at Ϫ80°C.
Purification of AGT-All the chromatographic steps were performed at 18°C. Binding of AGT (1-2 ml/min) on to low substitution (20 mol phenyl/ml gel) phenyl-Sepharose (Amersham Pharmacia Biotech) requires the equilibration of the medium with 1.17 M ammonium sulfate in 28.5 mM sodium phosphate buffer (pH 7.4). Elution of bound AGT was performed by decreasing salt concentration using a discontinuous step gradient with MilliQ water (4 ml/min). The pooled fractions were diluted twice with 300 mM sodium acetate buffer (pH 6.0) to equilibrate the sample. Adsorption of AGT on to concanavalin A (Amersham Pharmacia Biotech) was made in the presence of 100 mM NaCl, 1 mM CaCl 2 , and 1 mM MnCl 2 (0.2-0.4 ml/min). Nonbound proteins were washed out by 30 mM sodium acetate (pH 6.0), 1 mM CaCl 2 , 1 mM MnCl 2 buffer. Bound proteins were eluted (0.2 ml/min) by using 200 mM methyl ␣-D-mannopyranoside in the same buffer. Finally, adsorption of AGT onto Mono Q HR5/5 column (Amersham Pharmacia Biotech) (1 ml) was achieved in 100 mM Tris-HCl (pH 8.0). Equilibration of the concanavalin A pool was made by adding 1.5 M Tris-HCl (pH 8.8) (2.5 l/ml). After washing the column (Tris-HCl 100 mM (pH 8.0)), bound proteins were eluted by increasing salt concentration using a linear gradient (60 ml) from 0 to 0.5 M NaCl in Tris-HCl 20 mM (pH 8.0) buffer (1 ml/min). Finally, elution of AGT was achieved in conditions corresponding to 120 -180 mM NaCl. The stock solution was kept at 4°C in the presence of 10% (v/v) glycerol.
The presence of protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 5 mM EDTA, 10 mM benzamidine, 1 M leupeptin, 1 M pepstatin, and 1 M aprotinin) throughout the purification allowed us to obtain the uncleaved AGT. Omitting them gave, reproducibly, one form of AGT cleaved in the RCL.
Purified circulating AGT was made from a pool of human plasma in the same conditions except that blue Sepharose (Amersham Pharmacia Biotech) chromatography was added as a first step using 50 mM Tris-HCl (pH 8.0) buffer, and Superose 12 chromatography was added as the last step using 20 mM Tris-HCl (pH 8.0) containing 150 mM NaCl.
MALDI-TOF Mass Spectroscopy of AGT-Pure AGT (10 -50 g) was submitted to a Superose 12 column (Amersham Pharmacia Biotech) in 100 mM ammonium acetate buffer (pH 6.0). The most concentrated fraction, corresponding to 60 kDa, was lyophilized and resuspended in a 0.1% (v/v) trifluoroacetic acid or 5% (v/v) formic acid just before spotting onto a sinapinic acid matrix (Sigma) for analysis by mass spectroscopy. Spectra were recorded on a Voyager Elite mass spectrometer (Perseptive Biosystems, Framingham, MA) equipped with delayed extraction. Data were acquired at a sampling rate of 2 GHz. Accelerating voltage was 25 kV, grid voltage was 90%, and guide wire was 0.3%. The apparatus was calibrated in linear mode (i) for the determination of the mass of glycosylated AGT using single (m/z ϭ 66431)-and double (m/z ϭ 33216)-charged ions of bovine serum albumin and (ii) for the determination of the mass of AGT C-terminal peptides using Ang I peptide (m/z ϭ 1297.5) and insulin (m/z ϭ 5734.5) as external markers.
Renin Substrate Activity of AGT-Extensive cleavage by recombinant human renin (1.25 nM) was allowed for 1 h at 37°C in a 100 mM sodium citrate buffer (pH 5.7) containing 3 mM EDTA. The velocity parameters were determined at pH 5.7 in the same buffer in the presence of 20 pM recombinant human renin, assumed to be fully active. The reaction was stopped as previously reported (24). The generated Ang I was measured in triplicate in a specific radioimmunoassay (28).
Heat-induced Polymerization of AGT-The thermal stability of the pure AGT was assayed by diluting the stock solution twice in a 100 mM sodium acetate (pH 5.5) buffer or a 100 mM Tris-HCl (pH 8.0) buffer immediately before incubation for 1 h at the indicated temperature. An aliquot was removed to test the ability of AGT to be cleaved by renin (at pH 5.7 for 1 h), and the remaining 2 g were immediately unfolded in denaturing buffer, loaded on to a nonreduced SDS-polyacrylamide electrophoresis gel, and revealed by Coomassie Blue staining.

RESULTS
Cleavage of the Peptide Bond Gln 411 -Gln 412 within the RCL of AGT by a Protease from CHO Cells-Starting from CHO cell supernatants in which AGT represents 20 -40% of the total protein, the renin substrate was purified 2.5-fold with a yield of around 20% using a nondenaturing procedure. MALDI-TOF mass spectroscopy analysis was performed on each preparation. The spectrum of this glycosylated AGT is broad, expanding from m/z ϭ 55000 to more than 60,000 with a mean m/z around 57,000 -58,000 (Fig. 1A). No contaminating proteins were present in the mass range of 20 -80 kDa.
In three preparations performed without any protease inhibitor, most of the AGT had a mass of around 52-53 kDa (data not shown), and most importantly, distinct peptides were detected in the range 3500 -6000. Four peptides were detected in a sample originating from a representative purification (Fig. 1B). The observed mass of the largest major peptide (m/z ϭ 4543.1) was fitted with the calculated mass (M r ϭ 4542.2) of the Cterminal sequence beginning at the Gln 412 and ending at Ala 452 , the last residue of mature AGT. The smaller peptide (m/z ϭ 4415.3) could have resulted from the major product through removal of a glutamine residue (Ϫ128.1) Then the sequential removal of the consecutive Leu 413 (Ϫ113.2) and Asp 414 (Ϫ114.1) would explain the presence of the two minor peptides (m/z ϭ 4308.3) and (m/z ϭ 4191.2), respectively. These mass determinations were verified by MALDI-TOF in the reflectron mode and by nanospray mass spectroscopy (data not shown).
These observations were highly suggestive of the presence of an RCL cleavage in the renin substrate. The presence of 4-kDa C-terminal peptides was detected as a faint band after SDS denaturation by using a Tris-Tricine gel, which allows the separation of small peptides in the range 1-20 kDa (27). Reducing and/or heat treatment did not release higher amounts. After transfer onto a polyvinylidenefluoride membrane, the 60-kDa band and the 4-kDa peptide were N terminus-sequenced (Fig. 2B). The 60-kDa band revealed only one sequence corresponding to the mature AGT. For the 4-kDa peptide, three N-terminal sequences were detected; the most abundant began at Gln 412 , the second at Leu 413 , and the third at Asn 414 (Fig.  2B). Thus the initial site of cleavage was between Gln 411 and Gln 412 , in accordance with the previous results obtained by MALDI-TOF mass spectroscopy (Fig. 1B).
Influence of RCL Cleavage on Angiotensin I Generation-Native AGT studied in this experiment has a specific activity of 26 g of Ang I/mg of protein, whereas the RCL-cleaved AGT has a specific activity of 26.7 g of Ang I/mg of protein; both values are close to the theoretical Ang I content of AGT, which is 25 g of Ang I/mg of protein. In addition, initial velocity parameters for renin hydrolysis, determined using the fulllength and the RCL-cleaved AGT, showed no significant difference (Table I). This shows that Ang I generation, located in the N terminus, is not affected by the cleavage in its RCL at the opposite end of the molecule.
Although the RCL Cleavage Was Complete, AGT Migrated as Two Bands-To detect AGT devoid of its 4-kDa C-terminal peptide following cleavage in the RCL, an AGT peptide antibody against the C-terminal end (C-1350) was produced. This antibody together with a N-terminal Ang I antibody (N-1345) (24) allowed the detection of the full-length AGT but not the RCL-cleaved AGT (Fig. 3A). For an unknown reason, the C-terminal antibody (C-1350) did not recognize the isolated 4-kDa peptide generated following cleavage in the RCL of AGT (data not shown). As expected, both native AGT and the RCLcleaved AGT were equally responsive to the Ang I antibody, with a slight difference (around 5 kDa) in their respective apparent molecular masses due to the loss of C-terminal peptides. After extensive enzymatic deglycosylation, which increased the migration of both AGTs (Fig. 3B), well defined   bands were observed. For the full-length AGT, bands at 54 and 50 kDa were both responsive to N-1345 as well as C-1350, and for AGT cleaved within the RCL, bands at 48 kDa and 44 kDa were responsive only to N-1345. Our data reveal that human recombinant AGT migrates on SDS-polyacylamide electrophoresis gels as a doublet, differing by 4 kDa even if it is more than 95% C terminus-cleaved.
Furthermore, we used this C terminus antibody to detect the step at which the RCL cleavage occurred during purification. CHO-produced AGT was intact in the serum-free medium because it was N and C termini, and the C-terminal response of AGT was lost during the third step of our purification (data not shown), suggesting the co-purification of the protease and AGT.
The Circulating Renin Substrate Was Not Cleaved in Its RCL-To test whether a RCL-cleaved AGT may exist in vivo, human plasma AGT was purified and characterized. Plasma AGT was purified 200 times with a yield about 12% (Fig. 4A). It was analyzed using a Tris-Tricine gel (Fig. 4B) and the pair of end-terminal AGT polyclonal antibodies (N1345 (Fig. 4C) and C1350 (Fig. 4D)). In our conditions of purification, no RCL cleavage of circulating AGT was observed because (i) no 4-kDa peptide was detected (Fig. 4B) and (ii) both 60-kDa bands were recognized by the N terminus as well as the C-terminal antibody before (Fig. 4C) and after deglycosylation (Fig. 4D). Although this circulating AGT was not cleaved in the C terminus, a doublet (54 -50 kDa) was still consistently observed, even after N-and O-deglycosylation and after removal of sialic acids residues. This pattern of migration was consistent with that of the recombinant native AGT (Fig. 3).
Formation of SDS-resistant Aggregates on Heating RCLcleaved AGT-Heating native (i.e. full-length) serpins generally leads to the formation of loop-sheet polymers that migrate as a ladder on nondenaturing gels (29). We tested this property by heating the recombinant AGTs at 65°C (Fig. 5). Unexpectedly, the RCL-cleaved AGT had kept its ability to polymerize, suggesting a mechanism of polymerization different from the much-studied loop-sheet interactions for the RCL-cleaved serpins (29). Monomers of AGT disappeared after 15 min of incubation; a longer incubation time had no further effect (data not shown).
The heat-induced polymerization of both AGTs was studied and compared after 1 h of incubation in 2 different conditions: acidic (pH 5.5 in acetate buffer) and basic (pH 8.0 in Tris-HCl buffer). Each was measured at various temperatures ranging from 4°C to 100°C (Fig. 6). The acidic condition corresponded to that previously used to obtain the latent form with partial insertion of the loop of the native plasminogen activator inhibitor-1 (30). The formation of aggregates was followed by their ability to generate Ang I after extensive renin hydrolysis (Fig.   6A) and by Coomassie Blue staining after migration on an SDS-polyacylamide electrophoresis gels (Fig. 6B). When the heat stability of both AGTs was tested in the acidic condition, it was increased by 12°C, as indicated by the shift to the left of the corresponding transition curves. At pH 8.0, although the two curves cannot be completely superimposed, they showed a nonsignificant difference in T m value. In fact, T m was around 56°C when native and RCL-cleaved AGTs were incubated at pH 8.0 and 68°C when both proteins were incubated in acetate buffer. At both pH values, the activity was totally abolished after 1 h at 75°C (Fig. 6A). The dramatic loss of the renin substrate activity was accompanied by the formation of SDSresistant aggregates for the native as well as the RCL-cleaved AGT (Fig. 6B). The formation of aggregates started around 70°C when the pH was acidic and at 60°C with a basic pH. Lowering the pH significantly increased the stability of the monomeric AGT (M). The mechanism of AGT polymerization was characterized by the transient formation of dimers (D) leading to the final formation of large aggregates (Ag) that were so big that they did not enter in the running gel. After 1 h at 100°C, no monomers were stained; thus, all the AGT had been  1, 5 g) was analyzed and compared with the RCL-cleaved recombinant AGT (lane 2, 5 g) for its ability to release the 4-kDa peptide previously characterized (Fig. 2) and indicated by an arrow. C, the circulating AGT was resolved by 9% SDS-polyacylamide gel electrophoresis and analyzed by Western blotting using N-1345 and C-1350 antibodies. D, the circulating AGT was extensively deglycosylated as in the legend of Fig. 3, resolved by 9% SDS-PAGE, and analyzed by Western blotting using N-1345 and C-1350 antibodies. changed into higher molecular weight aggregates. The presence of free cysteines in the AGT sequence could explain this particular behavior of a serpin when heated. In fact, the aggregates were dissociated when a reducing agent was added before loading (data not shown). DISCUSSION We have presented evidence for a new cleavage site in human AGT. Such a cleavage does not affect the generation of Ang I by renin and does not change the tendency of AGT to aggregate.
Cleavage of recombinant AGT was observed between Gln 411 and Gln 412 during a nondenaturing purification, generating a major 4-kDa peptide that was N terminus-degraded, presumably by an aminopeptidase. The residue 415, a proline, stops its action. No other AGT peptides were detected from the purified recombinant AGT by mass spectroscopy or in Tris-Tricine gel, showing that the full-length protein can be specifically proteolysed in its RCL. In addition, we have shown that the generated C-terminal 4-kDa peptides remain associated with the rest of the protein unless AGT has been unfolded by SDS.
This site of proteolysis, between two consecutive glutamine residues, is very unusual. Our data allow the assignment of a new P1-P1Ј cleavage site in the RCL of human AGT and prompt a reexamination of the comparison of the reactive center sequences of others serpins (Fig. 7A). According to this alignment, the doublet of glutamine residues Gln 411 -Gln 412 is conserved in marmoset, rat, and mouse but not in sheep. Thus, it may constitute a physiological P1-P1Ј site, the target of an unknown protease. For inhibitory serpins, the P1 residue plays an important role in determining the target specificity of the protease involved (31). In this regard, it should be noted that the P1 residue in human AGT is a glutamine, although it is very unlikely that it can form an acyl-enzyme with the target protease. Nevertheless, our alignment also showed the conservation of a proline residue in P6 position, except in sheep, and the strict conservation of residues present at position P3, P5Ј, and P6Ј. They may have a role in the presentation of the P1-P1Ј peptide bond for cleavage. Most importantly, human AGT has a large charged residue, a glutamate, in P14 in accordance with its inability to insert the reactive center loop (12,13). Indeed, the presence of an arginine charged residue in P14 in ovalbumin has been proposed to explain the absence of loop insertion (32).
A number of other alignments of this C terminus part of human AGT have been previously described (14,33,34). In addition, Patston and Gettins (14) have recently proposed a classification of the serpins based on the prediction of the secondary structure outside the RCL. A requirement for an alanine residue between P9 and P12 rather than a structural requirement for an ␣-helix was proposed for loop insertion. However, AGT did not match with these predictions, especially in the ␤-sheet, which immediately precedes their putative RCL cleavage site, Pro 416 -Glu 417 .
AGT is heavily glycosylated, leading to a heterogeneous protein, which migrates as a fuzzy band around 60 kDa in SDSpolyacrylamide electrophoresis gels (19). This heterogeneity of the circulating renin substrate has been well documented (16 -19); a partial C-terminal cleavage has been invoked to explain the presence of two bands, differing by 4 kDa, which remain after deglycosylation. In our experiments, human plasma AGT FIG. 6. Heat-induced polymerization of the native and the RCL-cleaved AGTs followed at pH 5.5 and pH 8.0; evidence for SDS-resistant aggregates. A, the ability of native and RCL-cleaved AGT to generate Ang I after extensive cleavage by renin (1 h at pH 5.7) was measured after a 1-h incubation of the renin substrate at the indicated temperature in two pH conditions. The value obtained at 50°C was taken as 100% because no significant differences were observed at 4°C, 37°C, and 50°C at either pH. B, Coomassie Blue-stained gels of AGTs (2 g/lane) incubated for 1 h at the indicated temperatures at pH 5.5 (two left panels) and pH 8.0 (two right panels) just before denaturation and loading on SDS-polyacylamide electrophoresis gels. Arrows indicate monomeric form (M), dimers (D), and aggregates (Ag), which remain on the top of the running gel. purified by a new nondenaturing method was not cleaved in its C terminus. Nevertheless, after removal of N-and O-glycans and sialic acid residues, it migrates as two equal intensity bands (54 -50 kDa), both stained by Coomassie Blue. This result is in accordance with our previous data showing a similar behavior for deglycosylated wild type recombinant AGT (52-50 kDa) detected by antibody in serum-free medium (26) but seems to contradict the production by CHO cells of a modified AGT, devoid of N-glycosylation sites, which migrates as a unique species at 48 kDa (35). Although we cannot completely exclude an incomplete deglycosylation and/or a specific migration on SDS-polyacylamide electrophoresis gels of the AGT, taken together our results point to a residual heterogeneity of the renin substrate. We hypothesize that the wild type AGT produced from CHO cells and the circulating renin substrate may carry some as yet unidentified compounds added after translation. The complete cleavage of the AGT in its C terminus does not lead to the disappearance of the doublet, suggesting that this compound may be linked to the rest of the protein and not to the C-terminal peptides.
An unexpected result shows that the RCL-cleaved protein does not differ from the intact AGT in its ability to polymerize on heating. We used native and also denaturing gel to characterize the polymerization of AGT. Neither experiment revealed any significant difference between the native and the RCL-cleaved AGT. In our conditions, both AGTs were able to polymerize on SDS-polyacylamide electrophoresis gels, leading to a dramatic loss of their ability to serve as a substrate for human renin. Moreover, both proteins showed the same thermal stability when they were incubated at pH 8.0 and pH 5.5. In summary, we have demonstrated that the heat-induced inactivation of human AGT is concomitant with the formation of SDS-resistant polymer, independently of its RCL cleavage. As distinct from the vast majority of native serpins that are able to form noncovalent aggregates, we describe AGT as able to form SDS-resistant aggregates. We suspect a predominant role for free cysteines (Cys 232 and/or Cys 308 ) (24) in this mechanism of polymerization.
Finally, our results lead us to propose a structural model for human AGT. A search in a data base for proteins with known three-dimensional structures revealed highest similarity between human AGT and AT (36). We have constructed a threedimensional model of the renin substrate using AT as a template structure (Fig. 7B). The functional N terminus (1-69) had to be excluded from the model because no template structure with this N-terminal extension exists. Important structural features of angiotensinogen are the central ␤-sheet A consisting of 4 antiparallel ␤-strands (Ala 168 -Pro 182 ; Asp 240 -Lys 256 ; Glu 350 -Gln 360 ; Gly 389 -Ala 401 ) and the C-terminal peptide (Gln 412 -Ala 452 ), which FIG. 7. Human AGT; comparison of the RCL sequences and modeling its serpin structure. A, human AGT sequence in the RCL region is reported and compared with the four other cloned mammalian AGTs and with sequences from a number of serpins. The nomenclature P14-P10Ј is from Schechter and Berger (43). AGT sequences from human (hAGT) (4), rat (rAGT) (3), mouse (mAGT) (44), and sheep (sAGT) (45) were previously reported. The marmoset AGT sequence (maAGT) is from Olivier Valdenair (personal communication). The alignment between the five mammalian AGTs was made using the TREEALIGN software. Amino acid sequences of the reported serpins and their classification were taken from Patston and Gettins (14). ACT, ␣-antichymotrypsin; AP, ␣2 antiplasmin; ATIII, antithrombin III; PEDF, pigment epithelium-derived factor; TBG, thyroxin binding globulin; CBG, corticosteroid binding globulin; PAI-1, plasminogen activator inhibitor-1; PAI-2, plasminogen activator inhibitor-2; HC-II, heparin cofactor II. B, modeling the serpin structure of human AGT in its native conformation: a ball and stick representation under two orientations. The structure of uncleaved AT (PDB:1PSI (36)) served as a template to build a C-␣ model of AGT. The model was built using the Roche in-house modeling package Moloc (46). Because of the rather low sequence identity between template and target sequence, we refrained from introducing any side chains. The final model comprises C-␣ positions of amino acid 70 -452 of human AGT. Some key residues are indicated as bigger balls: Ala 70 , the first residue, and Ala 452 , the last residue of the mature protein; three out of the four AGT cysteines; and the Met 235 residue. Gln 411 and Gln 412 in the middle of the RCL and the first residue of each of the four strands of the central large ␤-sheet A are indicated. The C-terminal peptide (Gln 412 -Ala 452 ) is represented as a bolder line. The figure was generated using RasMol 2.6. is inserted into the core of the protein. The two consecutive glutamines (Gln 411 -Gln 412 ), although not localized at the very top of the RCL, are in quite a good position for cleavage by a proteinase. In addition, our model predicts that the Cys 232 -Met 235 region is solvent-accessible. Our previous work has shown that this region bearing the M235T polymorphism is localized at the outside of the protein (37) and thus able to interact with the proform of eosinophil granule major basic protein (24). For other natural mutations of the AGT gene (2), only the T174M and the L359M changes could affect the stability of the predicted ␤-sheet A, which seems to be essential for the serpin fold of hAGT. However, it is very unlikely that these two amino acid substitutions would change the noninhibitory into an inhibitory angiotensinogen. More generally, our model predicts that none of the nine described missense mutations would drastically modify the serpin characteristics of hAGT described in this paper because they are located far away of the RCL region. Particularly the P14 position, crucial for loop insertion (32), and the cleavage site, important for the formation of an acyl-enzyme, are not the site of any substitution in hAGT .
The part (Asp 1 -Arg 69 ) that is excluded from our model, and that contains the Cys 18 partner of the Cys 138 (24) may play a crucial role in linking the Ang I N-terminal part to the des(Ang I)AGT. Des(Ang I)AGT is involved in renin cleavage because it cannot be suppressed (38) or substituted by the AT structure (39) for an efficient reaction. In fact, the Asp 1 -Asn 14 tetradecapeptide and also a chimeric protein AGT-AT are not good substrate for human renin. In addition, we describe how renin hydrolysis does not require the integrity of the serpin RCL, suggesting that the C-terminal part of angiotensinogen is not contacted by this aspartyl proteinase. The Ang I moiety is most likely to be located at the other end of the molecule where, in our model, Cys 138 is also found (the base of the protein in our representation).
A protein with a serpin fold bearing its physiological function at the N-terminal end is pigment epithelium-derived factor, a 375-residue protein. In this case, deletion of the serpin structure from the C-terminal end led to the production of truncated proteins still bearing the neurite-promoting activity (40). Contrary to that, we tried to express a truncated AGT (Asp 1 -Glu 403 ) in which the serpin RCL and the C-terminal peptide were lacking and failed to produce any expressed AGT in COS cells (data not shown), suggesting a predominant role of the serpin part in the overall folding of the renin substrate.
In conclusion, relationships between AGT and the serpin family can be summarized as a combination of Ang I generation function located at the N-terminal end and the presence of a serpin structure with no apparent function at the opposite end. The identification of a new site of proteolysis in its RCL suggests that AGT may be able to act as the substrate of a new protease. Interaction and cleavage within the RCL in serpins is not sufficient to obtain an inhibition of the target protease. In this respect it should be added that our previous work 2 has failed to detect any inhibitory activity of human AGT toward some classical serine proteases. Nevertheless, we cannot completely exclude the possibility that AGT can inhibit one nonserine protease. Indeed, new serpins that were recently characterized are able to inhibit pepsin, an aspartyl protease (41), and interleukin-converting enzyme, a cysteyl protease (42). The protease from CHO cells able to interact and cleave human AGT at this particular P1-P1Ј site has not yet been identified and might also be able to cleave other mammalian AGTs because the two consecutive glutamine residues are largely conserved. Although the circulating renin substrate that we have purified has an intact RCL, we hypothesize that a cleavage in the RCL of AGT may occur in vivo only in some AGT-rich tissues or in some physiological fluids.