Purification and characterization of protease Ci, a cytoplasmic metalloendoprotease in Escherichia coli.

Protease Ci, a cytoplasmic metalloprotease in Escherichia coli, has been purified to apparent homogeneity by conventional chromatographic procedures using 125I-labeled oxidized insulin B-chain as a substrate. The purified enzyme behaves as a 54-kDa protein under both denaturing and nondenaturing conditions, suggesting that it consists of a single polypeptide chain. It is inhibited by metal-chelating agents, including o-phenanthroline and NaCN, but not by inhibitors of serine proteases or thiol-blocking agents. Furthermore, protease Ci was found to contain 1.1 mol of zinc per mol of the enzyme upon analysis by HR ICP mass spectroscopy. Thus, protease Ci must be a zinc metalloprotease. Among the polypeptides tested as substrates, oxidized insulin B-chain and glucagon are most rapidly hydrolyzed. Intact insulin is a much poorer substrate than oxidized insulin B-chain, even though the affinity of the enzyme to intact insulin is approximately 100-fold greater than that to the B-chain. Since unlabeled oxidized insulin A-chain is capable of inhibiting the hydrolysis of 125I-labeled insulin B-chain, it also appears to be a substrate. Protease Ci also degrades lysozyme and lactalbumin, although to a much lesser extent than oxidized insulin B-chain. However, it shows little or no activity against proteins larger than 15 kDa (e.g. ovalbumin and denatured bovine serum albumin). Hydrolysis of oxidized insulin B-chain followed by amino acid composition analyses of the cleavage products reveals that as many as 10 of its 29 peptide bonds are hydrolyzed by protease Ci. This ability to hydrolyze relatively small polypeptides suggests that protease Ci may catalyze the later steps in the pathway for intracellular protein breakdown.

Soluble extracts of Escherichia coli contain at least nine distinct endoproteases that appear to be distinct from each other (1,2). Proteases Do (DegP), Re (Tsp), Mi, Fa, So, La (Lon), and Ti (Clp) are serine proteases that hydrolyze relatively large proteins, such as casein and globin. Of these, proteases La and Ti require ATP and Mg 2ϩ for activity and appear to catalyze the rate-limiting steps in the hydrolysis of abnormal proteins and of certain normal proteins (3,4). Two other enzymes, proteases Ci and Pi, are metalloproteases that degrade smaller polypeptides, such as insulin and the N-terminal fragment of ␤-galactosidase with a minimal length of 50 amino acids (called "auto-␣") that were used for complementation assay of ␤-galactosidase upon mixing with appropriate acceptor proteins (2,5). Proteases Mi and Pi are periplasmic enzymes, while all others are localized to the cytoplasm and therefore may play a role in the degradation of intracellular proteins (6).
The periplasmic insulin-degrading protease Pi, also called protease III (5) and pitrilysin (7), has been extensively characterized, although its physiological function is still unknown. Interestingly, protease Pi shares structural and functional homologies with the cytosolic insulin-degrading enzymes present in most mammalian cells (8). Overall amino acid sequence similarity between these enzymes is about 50% when conserved amino acid changes are included as matches. Furthermore, both the proteases contain three highly conserved domains, of which the first domain is characterized by the presence of a metal-binding site (9). Therefore, it has been suggested that the three regions of homology may play an important role in the active site and catalytic mechanism of the proteases (8).
In addition to sharing sequence similarity, both protease Pi and human insulin-degrading enzyme have a molecular mass of about 110 kDa and require a divalent metal ion for activity (5,8,9). Both proteases have an isoelectric point of 5.3 and are maximally active in the pH range of 6.5-8.5. Both are sensitive to inhibition by bacitracin. However, they differ in sensitivity to sulfhydryl blocking agents, such as N-ethylmaleimide and iodoacetamide. While the human insulin-degrading enzyme is highly sensitive to these reagents, protease Pi is not. An insulin-degrading enzyme similar in its properties to the human enzyme and protease Pi has also been isolated from Drosophila melanogaster (10).
E. coli in its cytoplasm contains an additional enzyme, protease Ci, that can degrade insulin (1,2). Aside from its subcellular location, protease Ci shows remarkable similarities to protease Pi. Like Pi, it appears to be 110 -130 kDa, to have a pH optimum at 7.5, and to be sensitive to inhibition by metalchelating agents and by the antibiotic bacitracin. However, all these properties were determined only on the partially purified protease Ci, whose physiological role is totally unknown. In the present studies, therefore, we purified protease Ci completely to carry out a detailed characterization of its biochemical and physicochemical properties, for determination of its cleavage specificity, and ultimately for elucidation of its physiological function.

EXPERIMENTAL PROCEDURES
Materials-E. coli 3302 strain (HfrH ptr-3 lacZ727) (11) was grown in Luria broth, harvested at late log phase, and kept at Ϫ70°C until use. Protein substrates including insulin and oxidized insulin B-chain were radioiodinated using Iodo-Beads as described (12). Denatured * This work was supported by grants from The Korea Science and Engineering Foundation through Research Center for Cell Differentiation and The Ministry of Education. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
bovine serum albumin (BSA) 1 was prepared by reduction of disulfide bonds in the presence of 6 M guanidine HCl and alkylation with iodoacetamide (13). Iodo-Beads were obtained from Pierce; HiLoad Q, heparin-Sepharose, and phenyl-Sepharose were from Pharmacia Biotech Inc.; butyl-Toyopearl and TSK gel ODS-80T M column were from TOSOH Corp. (Japan). All others were purchased from Sigma.
Assays-Proteolysis was assayed as described (14). The reaction mixtures (final volume, 0.1 ml) contained 50 mM Tris-HCl (pH 8), 5 mM MgCl 2 and proper amounts of 125 I-labeled protein substrates and the protease preparations. Incubations were performed at 37°C for 30 -60 min. The amount of radioactive materials soluble in 10% (w/v) trichloroacetic acid was then determined using a ␥ counter. Proteins were assayed as described by Bradford (15) using BSA as a standard. Isoelectric focusing of proteins on 5% (w/v) polyacrylamide gels was carried out as described (16).
Preparation of Crude Extract-The frozen E. coli cells (140 g) were thawed and suspended in 50 ml of buffer A (20 mM Tris-HCl (pH 7.8) and 5 mM MgCl 2 ) containing 100 mM NaCl. The cells were disrupted with a French press at 14,000 psi and centrifuged at 100,000 ϫ g for 3 h. The supernatant was dialyzed against the same buffer and referred to as crude extract.
Preparation of Antibody-The purified protease Ci preparation obtained from the final purification step (see below) was subjected to electrophoresis on 10% polyacrylamide slab gels containing SDS and 2-mercaptoethanol (17). After briefly staining the gels with Coomassie Blue R-250, the protein band of the 54-kDa protease Ci was cut off and crushed. The sample was then subcutaneously injected into albino rabbits for the production of anti-protease Ci antiserum. IgGs were isolated from the antiserum by sodium sulfate fractionation (18). Immunoblot analysis was performed using the antiserum or the anti-IgG as described (19).
Determination of Cleavage Specificity-Oxidized insulin B-chain was digested with the purified protease Ci for 0.5-15 h at 37°C. The cleavage products were subjected to reversed-phase HPLC using a C 18 column (TSK gel ODS-80T M ; 4.6 ϫ 150 mm). The peptides bound to the column were eluted with a linear gradient of 1-35% (v/v) acetonitrile containing 0.1% (v/v) trifluoroacetic acid. Elution of the peptides was monitored by their absorbance at 206 nm. Each of the peptide peaks was pooled, concentrated by evaporation, and again subjected to the reversed-phase HPLC. Homogeneous peptides were isolated, dried, and hydrolyzed in 6 N HCl and 0.1% (v/v) phenol at 140°C for 4 h under vacuum. The amino acid compositions of the peptides were then determined using an HPLC system (TOSOH Corp., Japan) equipped with a fluorescence detector (model FS8010) and a system controller (data processor, model SC8020).
Determination of Metal Ion-The purified protease Ci was extensively dialyzed against Milli-Q water with resistance of 18 megaohm. The Milli-Q water from the last dialysis was used as the reference. All the glasswares used in this study were boiled in 5% (v/v) HCl for 3 h to remove adventitious metal ions. Metal contents of the enzyme were then determined using an HR ICP mass spectrometer (model VG Plasma Trace).

RESULTS
Purification-The crude extract (17.7 g of protein) was obtained from 140 g of E. coli 3302 strain that lacks protease Pi, degrading insulin and existing in the periplasmic space (2,5). The extract was loaded onto a DEAE-Sepharose column (5 ϫ 20 cm) equilibrated with buffer A containing 100 mM NaCl. After washing the column, proteins were eluted with a linear gradient of 100 -250 mM NaCl and assayed for proteolysis. As shown in Fig. 1, a single peak of proteolytic activity appeared at about 170 mM NaCl, whether 125 I-labeled insulin or oxidized insulin B-chain was used as the substrate. However, oxidized insulin B-chain was hydrolyzed at a rate at least 50-fold faster than intact insulin (also see below). Therefore, we used oxidized insulin B-chain as the substrate for further purification and characterization of protease Ci, although this enzyme had initially been identified using intact insulin as its substrate (2,20).
The fractions containing high activity were pooled and applied to a HiLoad Q column (2.6 ϫ 10 cm) equilibrated with buffer A containing 150 mM NaCl. Proteins bound to the column were eluted with a linear gradient of 150 -300 mM NaCl. The active fractions were pooled, dialyzed against buffer A, and loaded onto a heparin-Sepharose column (1.5 ϫ 8 cm) equilibrated with the same buffer. Proteins that did not bind to the column were collected, and a saturated (NH 4 ) 2 SO 4 solution (pH 7.8) was added to a final concentration of 1 M. The sample was then loaded onto a phenyl-Sepharose column (1 ϫ 6 cm) equilibrated with buffer A containing 1 M (NH 4 ) 2 SO 4 . Proteins bound to the column were eluted by decreasing linearly the salt concentration to 0.01 M. The peak of oxidized insulin B-chaindegrading activity was eluted at about 0.17 M (NH 4 ) 2 SO 4 ( Fig.  2A). The fractions containing high activity were pooled, adjusted to 1.2 M (NH 4 ) 2 SO 4 , and loaded onto a butyl-Toyopearl column (1 ϫ 4 cm) equilibrated with buffer A containing 1.2 M (NH 4 ) 2 SO 4 . Proteins were eluted by decreasing the salt concentration to 0.1 M. Active fractions were pooled, adjusted to 1.5 M (NH 4 ) 2 SO 4 , and again loaded onto the same butyl-Toyopearl column but equilibrated with 1.5 M (NH 4 ) 2 SO 4 . Proteins were eluted as above but by decreasing the salt concentration to 0.9 M. The fractions containing high activity (Fig. 2B) were pooled, added with glycerol to a final concentration of 20% (v/v), and kept at Ϫ70°C for further use. Summary of the purification of protease Ci is shown in Table I.
Physicochemical Properties-The subunit size of protease Ci was estimated to be 54 kDa upon analysis by polyacrylamide gel electrophoresis in the presence of SDS and 2-mercaptoethanol (Fig. 3A, lane b). Only a single band could be seen in the

. Elution profile of insulin and oxidized insulin B-chain-degrading activities from a DEAE-Sepharose column.
The crude extract was chromatographed on a DEAE-Sepharose column as described in the text. Fractions of 20 ml were collected at a flow rate of 250 ml/h. Proteolysis was assayed at 37°C using 125 I-labeled insulin and oxidized insulin B-chain as the substrates as described under "Experimental Procedures." The dotted and slashed lines indicate protein profile and NaCl gradient, respectively. gel, indicating that protease Ci was purified to apparent homogeneity. To determine the size of the protease under nondenaturing conditions, the purified enzyme was subjected to gel filtration on a Superose-12 column (1 ϫ 30 cm). Fig. 4 shows that the peak activity of purified protease Ci (closed circles) is eluted in the fractions corresponding to about 54 kDa. Thus, the enzyme consists of a single polypeptide chain. The isoelectric point of protease Ci was estimated to be about 5.2 upon analysis by isoelectric focusing on a 5% polyacrylamide gel.
However, the native size of protease Ci had been reported to be in the range of 110 -130 kDa, although the prior studies used crude preparations of the enzyme for the size estimation (2). To resolve this discrepancy, the enzyme preparations obtained from the DEAE-Sepharose column (see Fig. 1) were also subjected to gel filtration using the Superose-12 column. As was reported earlier, the activity of protease Ci was eluted in the fractions corresponding to about 110 kDa (Fig. 4, open  circles). On the other hand, the enzyme preparation obtained after the HiLoad Q column, which removed more than 98% of total proteins, consistently ran as a 54-kDa protein, just as the purified protease did (data not shown). However, when the gel filtration chromatography was performed in the presence of 8 M urea and the resulting column fractions were subjected to immunoblot analysis using the antiserum raised against the purified protease Ci, the peaks of both the DEAE and purified enzyme preparations were found to elute in the fractions corresponding to about 54 kDa (data not shown). These results suggest that the size change may be due to dissociation of the subunits or removal of certain protein(s) that might have been nonspecifically bound to protease Ci.
We also tested whether the change in the size of protease Ci is due to limited proteolysis of the enzyme during purification. The E. coli cells grown to mid-log phase were harvested, immediately boiled, electrophoresed in the presence of 2-mercaptoethanol, and subjected to immunoblot analysis as above. This antiserum mainly interacted with the 54-kDa protein and not with any protein of 110 kDa (Fig. 3B, lane a). The minor bands seen in the gel could also be detected with the preimmune serum (data not shown). The identical data were obtained when the same experiments were performed under nonreducing conditions. Thus, the change in the size of protease Ci is not due to limited cleavage during purification of the enzyme either by itself or by other contaminating E. coli proteases. These results indicate that the 54-kDa protein responsible for the insulin degradation is protease Ci.
Effect of Site-specific Reagents and Protease Inhibitors-The effects of various protease inhibitors and site-specific reagents were tested by incubating the purified protease Ci for 20 min at 37°C prior to the addition of 125 I-labeled insulin B-chain. As shown in Table II, o-phenanthroline at 1 mM inhibited the enzyme activity by about 50%. NaCN, which also is known to inhibit metalloproteases (21), reduced the activity by about 80% at 1 mM. In addition, the antibiotic bacitracin at 0.7 mM inhibited the enzyme by about 90%. However, protease Ci was not sensitive to inhibitors of serine proteases, such as diisopropyl fluorophosphate and phenylmethylsulfonyl fluoride, or of sulfhydryl proteases, including N-ethylmaleimide and iodoacetamide. It is also not sensitive to various antibiotic peptide aldehyde inhibitors tested, such as leupeptin, antipain, chymostatin, and bestatin, nor to the epoxide, E-64. Thus, protease Ci appears to be a metalloprotease. Nearly identical results were obtained when intact insulin was used as the substrate (data not shown).
To clarify further the requirement for metal ions, protease Ci was preincubated with 0.2 mM NaCN for 20 min, which blocked the initial enzyme activity by about 65%. Various metal ions were then added to determine whether any of them can reactivate the enzyme. As shown in Table III, protease Ci could be reactivated by Co 2ϩ , Mn 2ϩ , and Cu 2ϩ in the order of their effectiveness but not by Ca 2ϩ , Mg 2ϩ , and Zn 2ϩ at the concentrations tested. Without the preincubation, however, Zn 2ϩ , Co 2ϩ , and Cu 2ϩ strongly inhibited the activity of protease Ci while Mg 2ϩ , Ca 2ϩ , and Mn 2ϩ were stimulatory (data not shown). Thus, it appeared as if protease Ci uses Co 2ϩ as its metal cofactor.
To determine whether protease Ci indeed contains Co 2ϩ , HR-ICP mass spectral analyses of two different preparations of the purified enzyme were performed. The enzyme contained 1.51 and 1.26 g of Zn 2ϩ per mg of protein. Assuming a size of 54 kDa, this is equivalent to a mean of 1.1 mol of Zn 2ϩ per mol of protease Ci. Although the NaCN-treated enzyme could be reactivated most effectively by Co 2ϩ and not by Zn 2ϩ , it is a well known fact that metalloproteases that use Zn 2ϩ as the natural metal cofactor can be inhibited when incubated with excess Zn 2ϩ . In addition, in some cases (e.g. metallocarboxypeptidases), Co 2ϩ activates the enzyme severalfold over the activity seen with the native Zn 2ϩ -containing enzyme (22)(23)(24). Thus, protease Ci is likely to be a Zn 2ϩ metalloprotease.
Hydrolysis of Protein Substrates-To determine the time course for hydrolysis of 125 I-labeled insulin and oxidized insulin B-chain, the purified protease Ci was incubated with these substrates for varying periods. As shown in Fig. 5, hydrolysis of both polypeptides increased linearly for the duration of incubation. In addition, this result again shows that oxidized insu-

FIG. 2. Elution profiles of the protease Ci activity from phenyl-Sepharose (A) and the second butyl-Toyopearl columns (B).
Proteins obtained from heparin-Sepharose column (17.2 mg) were chromatographed on a phenyl-Sepharose column. Fractions of 1 ml were collected at a flow rate of 10 ml/h. Proteins obtained from the first butyl-Toyopearl column (0.66 mg) were again chromatographed on the same column after equilibration with buffer A containing 1.5 M (NH 4 ) 2 SO 4 . Fractions were collected as above, and their ability to hydrolyze 125 I-labeled insulin B-chain (q) was then assayed by incubating 5-l aliquots of them for 30 min at 37°C. The dotted and slashed lines indicate protein profile and (NH 4 ) 2 SO 4 gradient, respectively. lin B-chain is much more rapidly degraded than intact insulin. From these data, the specific activity of protease Ci against intact insulin was estimated to be 2.5 nmol/min/mg, which is comparable to that of Drosophila insulin-degrading enzyme (3.3 nmol/min/mg) (10).
We then examined whether protease Ci can also hydrolyze oxidized insulin A-chain. The activity of the purified enzyme against 125 I-labeled B-chain was measured in the presence of increasing amounts of the unlabeled A-chain. As shown in Fig.  6, the insulin A-chain (like unlabeled intact insulin and oxidized B-chain) could competitively reduce breakdown of the radiolabeled B-chain. These results suggest that protease Ci is also capable of hydrolyzing oxidized insulin A-chain. In addition, unlabeled intact insulin molecules were able to inhibit the hydrolysis of 125 I-labeled B-chain, even more effectively than any of A-or B-chain. From these data, the K i values of the competing, unlabeled intact insulin and oxidized insulin Bchain molecules were estimated to be 0.13 and 12.4 M, respectively. These results suggest that the affinity of protease Ci for intact insulin is about 100-fold higher than that of oxidized insulin B-chain, despite the fact that the latter polypeptide is much more rapidly hydrolyzed by the enzyme than intact insulin.
A variety of 125 I-labeled proteins of different sizes was also tested for susceptibility to the purified protease Ci. Glucagon was also rapidly degraded by the enzyme but approximately 20% as fast as oxidized insulin B-chain (Table IV). Protease Ci was also capable of degrading polypeptides with sizes smaller

TABLE II Effects of various site-specific agents and protease inhibitors
on the activity of protease Ci Protease Ci (0.1 g) was incubated for 20 min at 37°C with the inhibitors prior to the addition of 5 g of 125 I-labeled insulin B-chain. Proteolysis was then assayed by further incubation of the mixtures for 30 min as described under "Experimental Procedures." The enzyme without added inhibitor converted 1.9 g of the B-chain into trichloroacetic acid-soluble products. This activity was expressed as 100%, and the others were as their relative values. than 15 kDa, such as lactalbumin and lysozyme, although to a much lesser extent than oxidized insulin B-chain. However, it showed little or no activity against any larger proteins tested, such as BSA, denatured BSA, ovalbumin, carbonic anhydrase, and globin. In addition, protease Ci cleaved angiotensin I and the synthetic peptide that is used as a substrate of a tyrosineprotein kinase (RRLIEDAEYAARG) (25) upon analysis of their incubation mixture on HPLC. However, it failed to cleave Leuenkephalin and bradykinin. Therefore, it is tempting to speculate that protease Ci may preferentially hydrolyze the peptides with amino acids ranging from 10 to 150. Peptide Bond Specificity-To determine the peptide bond specificity of protease Ci, oxidized insulin B-chain was incubated with the enzyme for various periods, and the reaction products were separated by reversed-phase HPLC using a C 18 column. Whether the incubations were performed for 2 or 15 h, the same 14 peptide peaks were detected by absorbance at 206 nm (Fig. 7). Upon prolonged incubation, however, the areas under the peaks 5, 8, 9, 10, 12, and 14 markedly decreased, while those under peaks 1, 2, 3, 4, 5, 11, and 13 increased significantly. Thus, it appears likely that the peptides in the former peaks behave as intermediates in degradation by protease Ci. When the individual peaks were rechromatographed on the same column but using a shallow gradient of acetonitrile, each of the peaks 4, 8, and 12 was further separated into two distinct peaks (data not shown).
The cleavage sites on oxidized insulin B-chain were then determined by analysis of the amino acid composition of each of the purified peptides. Fig. 8 summarizes the amino acid sequences of the peptides deduced from their amino acid compositions. In accord with the results from the reversed phase chromatography, a number of peptides appeared to behave as intermediates in degradation, such as peptide 8b for peptides 1 and 4a, 12b for 2 and 11, and 14 for 2 and 13. Peptide 5 may also be an intermediate in the generation of peptide 1 and LCGSH. However, the latter pentapeptide was not found among the 14 peaks and may correspond to any of the minor peaks whose amino acid composition was not determined. These results strongly suggest that protease Ci preferentially cleaves on the amino side of hydrophobic amino acids (i.e. the peptide bonds of His 5 -Leu 6 , His 10 -Leu 11 , Leu 11 -Val 12 , Glu 13 -  1 g) was incubated for 20 min with 0.2 mM NaCN and then with the indicated metal ions for the next 10 min at 37°C. After the incubation, proteolysis was assayed by adding 125 I-labeled insulin B-chain. FIG. 5. Time-dependent hydrolysis of insulin and oxidized insulin B-chain by the purified protease Ci. Assays were performed by incubating 1 g of 125 I-labeled insulin (E) and 0.25 g of the purified protease Ci at 37°C for various periods. Incubations were also carried out with 5 g of oxidized insulin B-chain (q) and 0.05 g of the enzyme.  To examine whether peptide 4b is produced by trace contam-inating proteases under the incubation conditions, the same experiment was performed using the purified protease Ci that had been treated with a mixture of protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 1 mM N-ethylmaleimide, 5 g/ml leupeptin, and 5 g/ml E-64), which do not inhibit protease Ci (see Table II), followed by dialysis against buffer A to remove the inhibitors. However, peptide 4b was consistently generated with the others. On the other hand, when the cleavage of insulin B-chain was carried out with the supernatant fraction obtained after immunoprecipitation of the purified enzyme using anti-protease Ci IgG, no peptide was generated. Therefore, it remains uncertain whether peptide 4b is one of the authentic cleavage products or generated by trace contaminating proteases in the purified protease Ci preparation.

DISCUSSION
In the present studies, a cytoplasmic metalloprotease in E. coli, protease Ci, was purified to apparent homogeneity using 125 I-labeled oxidized insulin B-chain as a substrate. Since E. coli contains an additional insulin B-chain-degrading enzyme, protease Pi, in the periplasm (2, 6), we used E. coli 3302 strain, which lacks the periplasmic enzyme. From 140 g of the frozen cells, we obtained 90 g of purified protease Ci with a final yield of about 1.5%. Thus, protease Ci accounts for at least 0.03% of the soluble protein of E. coli.
Protease Ci was found to have much higher affinity for intact insulin than for oxidized insulin B-chain, despite the fact that the latter polypeptide is hydrolyzed approximately 50-fold more rapidly than intact insulin. Similar results have been reported for protease Pi (7). Therefore, it appears possible that dissociation of intact insulin and/or its cleavage products from the enzyme may be a rate-limiting step for degradation of additional insulin molecules. In this regard, the characteristics of low k cat and low K m are not unlike those of small protein inhibitors of proteases (26), as was noted by Barrett and coworkers (7) for protease Pi.
Protease Ci, like protease Pi (5,9), is not inhibited by sulfhydryl blocking agents, such as N-ethylmaleimide and iodoacetamide. This insensitivity of the bacterial enzymes distinguishes them from the human and Drosophila insulindegrading enzymes that are highly susceptible to inactivation by such reagents (9,10). The predicted amino acid sequence of protease Pi contains a single Cys residue (27), whereas the human insulin-degrading enzyme has 12 Cys residues (8). One of the Cys residues in the human enzyme is present in the metal-binding site (HXCXH) in the first highly conserved domains of insulin-degrading enzymes (28). Protease Pi also contains the metal-binding motif (HXXXH) but lacks the Cys residue (27). Protease Ci may contain a similar metal-binding site to that of protease Pi, which might explain its lack of sensitivity to sulfhydryl-modifying agents. Isolation and sequence determination of the DNA clone for protease Ci should reveal whether this is true.
Protease Ci is capable of hydrolyzing as many as 10 sites among the total 29 peptide bonds in oxidized insulin B-chain. In contrast, protease Pi cleaves only the Tyr 16 -Leu 17 bond (7), which is also hydrolyzed by protease Ci. Like protease Ci, the insulin-degrading enzymes from Drosophila and rat also hydrolyze multiple peptide bonds in the insulin B-chain (29), although protease Ci is able to cleave many more sites. In addition, all the peptide bonds hydrolyzed by the various eukaryotic insulin-degrading enzymes are also sensitive to degradation by protease Ci, except for the Ser 9 -His 10 bond and the Phe 24 -Phe 25 bond, which are cleaved only by the rat enzyme (29).
Protease Ci appears to preferentially degrade relatively small-sized polypeptides, in contrast with other soluble pro- FIG. 7. HPLC separation of the peptide fragments of oxidized insulin B-chain. The purified protease Ci (1.5 g) was incubated with 100 g of oxidized insulin B-chain at 37°C for 0, 2, and 15 h. The cleavage products were then separated by reversed-phase HPLC on a C 18 column as described under "Experimental Procedures." The peaks of peptide fragments were numbered in the order of their elution. The capital letter B indicates oxidized insulin B-chain.  Fig. 7 were rechromatographed on the same C 18 column but using a shallow gradient of acetonitrile, each of the peaks 4, 8, and 12 was further separated into two peaks indicated as a and b. Therefore, total 17-peptide fragments were subjected to analysis for their amino acid composition as described under "Experimental Procedures." The amino acid sequences of the peptides (bars) and then the cleavage sites on oxidized insulin B-chain (arrows) were deduced from the amino acid composition data. teases in the cytoplasm of E. coli, proteases Do (DegP), Re (Tsp), Mi, Fa, So, La (Lon), and Ti (Clp) (1,2). In addition to oxidized insulin B-chain, protease Ci rapidly hydrolyzed glucagon and degraded lactalbumin and lysozyme, although at much slower rates. However, little or no trichloroacetic acid-soluble products are generated by the purified protease from any protein tested that is larger than 15 kDa. Interestingly, when proteins larger than 15 kDa, such as casein, were incubated with the ATP-dependent protease La or Ti in the presence of protease Ci, the production of acid-soluble products was over 2-fold faster than in its absence. 2 The ATP-dependent proteases appear to catalyze the initial rate-limiting steps in the degradation of highly abnormal polypeptides and of certain normal proteins (3,4). Therefore, protease Ci may function subsequently in the pathway for intracellular protein breakdown and digest the oligopeptides generated by the ATP-dependent endoproteases.