Characterization of the Bacterial Metalloendopeptidase Pitrilysin by Use of a Continuous Fluorescence Assay Angela ANASTASI,* C

Home , Astacin, Metalloendopeptidase, Pitrilysin

Biochem. J. (1993) 290, 601-607 (Printed in Great Britain) 601

Characterization of the bacterial metalloendopeptidase pitrilysin by use of a continuous fluorescence assay Angela ANASTASI,* C. Graham KNIGHT and Alan J. BARRETT Department of Biochemistry, Strangeways Research Laboratory, Worts Causeway, Cambridge CB1 4RN, U.K.

Pitrilysin (EC 3.4.99.44) has been purified from an over- with a very low K., and there was no action on the larger expressing strain of Escherichia coli. A 13-residue quenched- proteins tested. Since the activity of pitrilysin is confined to fluorescent-peptide substrate for the enzyme has been synthe- substrates smaller than proteins, it can be described as an sized, and found also to be cleaved by the homologous enzyme, endopeptidase of the 'oligopeptidase' type, and like other such insulinase (EC 3.4.99.45). The action of pitrilysin on peptides enzymes, it did not interact with a2-macroglobulin. The metal- and proteins was studied: insulin B chain was the most rapidly dependence of pitrilysin was confirmed, and it was found to be degraded, small peptides down to 10 residues in length were inhibited by bacitracin, especially in the presence of zinc. cleaved more slowly, intact insulin was cleaved very slowly but

INTRODUCTION Davies et al. (1987). Drosophila insulinase partially purified by a modification of the procedure of Garcia et al. (1988) lacking the Pitrilysin (EC 3.4.99.44) is the product of the ptr gene in hydroxyapatite chromatography and chromatofocusing steps Escherichia coli, and has previously been known as protease III was kindly given by Dr. B. D. Gehm (University of Chicago, IL, protease (Goldberg et al., 1981). (Cheng and Zipser, 1979) and Pi U.S.A.). Pitrilysin was assayed on oligopeptide fragments of,-galactosidase by Cheng and Zipser and on radiolabelled insulin by Goldberg and co-workers. Metalloendopeptidase characteristics Syntheses of the enzyme, including inhibition by chelating agents and N2-Ac-N6-Mca-L-lysine reactivation by metal ions, were described in the early papers. Subsequently, it has been reported that pitrilysin contains zinc Ac-Lys-OMe hydrochloride (48 mg, 0.20 mmol), 7-methoxy- (Ding et al., 1992) and that two histidine residues in the sequence coumarin-4-acetic acid (52 mg, 0.22 mmol) and benzotriazol-l - His-Tyr-Leu-Glu-His are essential for the binding ofzinc and for yloxytris(dimethylamino)phosphonium hexafluorophosphate activity (Becker and Roth, 1992). The amino-acid sequence of (98 mg, 0.22 mmol) were dissolved in dimethylformamide (2 ml) pitrilysin shows it to be homologous with insulin-degrading and di-isopropylethylamine (76 ,ul, 0.44 mmol) was added. The enzyme, insulinase (EC 3.4.99.45), and distantly related to the mixture was stirred for 18 h at 21 'C. The solvents were removed mitochondrial processing peptidases (EC 3.4.99.41) (Rawlings at 40 'C in vacuo and the oil remaining was dissolved in I M and Barrett, 1991). The purpose of the present paper is to NaOH (5 ml) at 0 'C. After 15 min, the solution was acidified describe a new assay for pitrilysin, and to extend the biochemical with acetic acid and applied to a column (15 mm x 400 mm) of characterization of this, first discovered and most readily avail- Vydac 218TPB1520. Elution with a gradient of 5-40% aceto- able, member of the family. nitrile in water containing 0.1 % trifluoroacetic acid yielded the product (61 mg, 75 % yield). N2-Ac-N-Mca-L-Lys (Mr 404.4) had m.p. 159-160 'C (Found: C, 59.0; H, 6.0; N, 6.9; EXPERIMENTAL C20H24N207 requires C, 59.4; H, 6.0; N, 6.9%). Materials N2-Fmoc-N6-Dnp-L-1ysine Chemicals were purchased as follows: glucagon (residues 22-29) from Research Plus Laboratories Inc., Denville, N.J., U.S.A., N6-Dnp-L-lysine (1.0 g, 2.9 mmol) was dissolved in acetonitrile/ bacitracin, Zn-bacitracin, 2,6-pyridinedicarboxylic acid and 4,7- water [1:1 (v/v), 15 ml] with di-isopropylethylamine (1 ml, phenanthroline from Aldrich Chemical Company Ltd., Gilling- 5.8 mmol). The solution was stirred at 0 'C while fluoren-9- ham, Dorset, U.K., and Zincov {[2-(N-hydroxycarboxamido)-4- ylmethylsuccinimidyl carbonate (1.0 g, 3.0 mmol) was added methylpentanoyi-L-alanyl-glycine amide]}, benzotriazol-1-yloxy- over 25 min. After a further 30 min at 0 'C, stirring was continued tris(dimethylamino)phosphonium hexafluorophosphate, benzo- for 16 h at room temperature. The reaction mixture was diluted triazol-l-yloxytripyrrolidinophosphonium hexafluorophosphate with ethyl acetate (100 ml) and the organic phase washed in turn and bromo-tripyrrolidinophosphonium hexafluorophos- with 100% (w/v) KHSO4 (100 ml), water (three times, 100 ml phate from Calbiochem Novabiochem (U.K.) Ltd, Notting- each wash) and saturated NaCl (100 ml). After drying over ham, U.K. Bacitracin was also obtained from Fluka, Glossop, anhydrous MgSO4, the ethyl acetate was removed at 40 'C in Derbyshire, U.K. Other chemicals were from Sigma. vacuo, and the residue was dissolved in chloroform (20 ml) and a2-Macroglobulin was purified essentially as described by applied to a column (2 cm x 7 cm) of silica gel 60 (Merck 9385).

Abbreviations used: LeuK-Gly, 5(RS)-amino-7-methyl-4-oxo-octanoic acid; LH-RH, luteinizing hormone-releasing hormone; Mca, (7-methoxycoumarin-4-yl)acetyl; PheK-Gly, 5(RS)-amino-6-phenyl-4-oxohexanoic acid; QF27, Mca-Nle-Ala-Val-Lys-Tyr-Leu-Asn-Ser-Lys(Dnp)-Leu-Asp-D-Lys; VIP, vasoactive intestinal peptide; TBS, Tris-buffered saline; DTT, dithiothreitol; Dnp, 2,4-dinitrophenyl; Fmoc, fluoren-9-ylmethoxycarbonyl. * To whom correspondence should be addressed. 602 A. Anastasi, C. G. Knight and A. J. Barrett

The product was eluted with 40 ml of chloroform/methanol/ until 70 %-90 % cleavage was shown by h.p.l.c. The reaction was acetic acid (9.25:0.5:0.25, by vol.). Removal of the solvents in stopped by the addition of acetic acid to a concentration of 5 % vacuo left an oil that crystallized on trituration with light (v/v). Samples (500 ,ul) were subjected to h.p.l.c. in a gradient of petroleum (1.46 g, 98 %). N2-Fmoc-NM-Dnp-L-lysine (Mr 534.5) 5-60 % (v/v) acetonitrile in 0.1 00 (v/v) trifluoroacetic acid, with had m.p. 81-83 'C. A sample for elemental analysis was recrystal- monitoring at 220 nm. The new peptides formed were collected lized as the dicyclohexylamine salt (Found: C, 65.5; H, 7.0; N, and hydrolysed for amino-acid analysis (Barrett et al., 1989). 9.8; C27 H26N4 08. C12H23N requires C, 65.4; H, 6.9; N, 9.8 %). For the comparison of hydrolysis products of QF27 generated by pitrilysin and Drosophila insulinase, 50,aM substrate was 1 h 0.5 m-unit/ml pitrilysin in Mca-NIe-AIa-Val-Lys-Lys-Tyr-Leu-Asn-Ser-Lys(Dnp)-Leu-Asp-D-Lys (QF27) incubated for at 37 °C with either 50 mM Tris/HCl (pH 7.5) containing 0.05 % Brij-35, or with This was synthesized by the Fmoc-polyamide method (Atherton 0.22 m-unit/ml Drosophila insulinase [activated with 5 mM di- and Sheppard, 1989). Briefly, Fmoc-amino-acid residues thiothreitol (DTT) for 30 min] in 20 mM phosphate buffer, (0.3 mmol) were coupled to Pepsyn KA resin (1.0 g) pH 7.0, containing 0.1 M NaCl. The reaction was stopped with esterified with Fmoc-D-Lys(Boc) (0.1 mmol) by use of benzo- acetic acid and the samples assayed by h.p.l.c. as described triazol-l-yloxytripyrrolidinophosphonium hexafluorophosphate above. (0.3 mmol), 1-hydroxybenzotriazole (0.3 mmol) and di-isopropylethylamine (0.3 mmol). Two couplings were done at each Fluorimetric assays step. The coupling of 7-methoxycoumarin-4-acetic acid remained incomplete after this procedure, and the resin was treated further These were made with the quenched fluorescent substrate QF27 with the acid (0.5 mmol), bromotripyrrolidinophosphonium in a Perkin-Elmer LS-3B spectrofluorimeter controlled by an hexafluorophosphate (0.5 mmol) and di-isopropylethylamine IBM-compatible computer running the FLU-SYS software (1 mmol) in N-methylpyrrolidone. The peptide was released by (Rawlings and Barrett, 1990). The instrument was set to zero treatment with trifluoroacetic acid/water [19:1 (v/v), 40 ml] for against substrate in assay buffer, and calibrated to read 1000 2 h at 21 'C, applied to a column of Vydac 218TPB 1520 and units offluorescence (excitation wavelength 328 nm and emission eluted with a gradient of 5-50 % acetonitrile in 0.1 0% tri- wavelength 393 nm) with N2-Ac-N6-Mca-L-Lys at a concen- fluoroacetic acid. Fractions were analysed by h.p.l.c. and those tration of 100% of that of the substrate. The substrate con- containing homogeneous Mca-Nle-Ala-Val-Lys-Lys-Tyr-Leu- centration was 10 ,uM, except where otherwise stated. Asn-Ser-Lys(Dnp)-Leu-Asp-D-Lys diacetate (Mr 2022.1) were Continuous assays were performed in a total volume of 2.5 ml combined and freeze-dried from 40 % (v/v) acetic acid (35 mg, of assay buffer with about 0.1Im-unit of enzyme, one m-unit of 17 % yield). Solutions were made in dimethylsulphoxide and the activity being defined as that hydrolysing one nmol of substrate/ concentrations determined from the absorbance at 410 nm after min at 37 'C. dilution into water, assuming e 7500 M-1 cm-', the value mea- For the determination of the Km value for QF27, reaction rates sured with N6-Dnp-lysine hydantoin. were determined in the continuous assay at substrate concentrations of 2-20 ,#M. The rates of hydrolysis of substrate and the corresponding [S] values were fitted to the Michaelis-Menten Protein assay and electrophoresis equation by non-linear-regression analysis (Enzfitter, Elsevier- Protein was assayed with the Bio-Rad assay (Bradford, 1976), Biosoft, Cambridge, U.K.). The values for apparent inhibition with BSA as standard. SDS/PAGE was done as described by constants (K9) were corrected for the effect of substrate, assuming Bury (1981). simple competition, by use of eqn. (2) according to Morrison (1969): (2) H.p.l.c. assays of peptide hydrolysis Ki = Kj'1(I + [S]1K.) The cleavage of peptides by pitrilysin was followed by h.p.l.c. on Strains of E. coli a Varian LC5000 instrument equipped with the Vista 402 The strains used were wild-type E. coli of strain PE004 har- data-processing system and a Techopak IOC18 column bouring the plasmid pPF307 (given by Professor P. T. Emmer- (4.6 mm x 250 mm) (HPLC Technology). A linear gradient of son, Department of Biochemistry, University of Newcastle upon 5-100% acetonitrile containing 0.1 0% trifluoroacetic acid was Tyne, Newcastle NEI 7RU, U.K.) and strain SK7814 containing run over 25 min at a flow rate of 1 ml/min and the eluate was plasmid pCDK35 (given by Dr. S. R. Kushner, Department of monitored at 220 nm. Genetics, Life Sciences Building, University of Georgia, Athens, Rates of hydrolysis of peptides were determined with digestion GA 30602, U.S.A.). mixtures containing 50,M of substrate in 50 mM Tris/HCl, pH 7.5/0.05 % Brij-35 (assay buffer) and 1.5 m-unit ofpitrilysin/ Purffication of pitrilysin ml. Incubation was at 37 °C, and samples were removed period- ically, usually after 0, 15, 30, 45 and 60 min. For each time-point, The two strains of E. coli were grown in Luria-Bertani medium the fraction of substrate unhydrolysed ([S]/[So]) was calculated (Sambrook et al., 1989). For the PE004 strain, the medium was as the area of the peak for unchanged substrate as a fraction of supplemented with 50 ,tg/ml of kanamycin, and the culture was the total area of new peaks plus unchanged substrate. It was initially at 30 °C; at A660 values,of 0.05-4. 1, the temperature was found that plots of ln([S]/[So]) versus time were essentially linear, shifted to 37 °C for 4-5 h, for temperature induction as described and under these circumstances an apparent first-order rate by Yasuda and Takagi (1983). The SK7814 strain was grown in constant was calculated from the slope of the plot according to LB medium supplemented with 170 ,ug/ml of chloramphenicol eqn. (1). for 5-6 h to an A660 value of approx. 2.0. Approx. 12 g (wet weight) of cells was harvested from culture ln([S]/[So]) =-kt (1 ) (3 litre) by centrifugation at 5000 g for 30 min, and washed To identify points of cleavage, peptide (0.1 mM) in assay with 10 mM Tris/HCl, pH 7.5, containing 0.1 M NaCl, followed buffer was treated with 0.25-3 m-unit of pitrilysin/ml at 37 °C by 10 mM Tris/HCl, pH1 7.5. All buffers used subsequently in the Pitrilysin 603 purification of pitrilysin contained 0.05 % Brij-35. For the cordingly, the latter strain was used for further work. Pitrilysin extraction of periplasmic proteins (Ames et al., 1984), the pellet was identifiable in both cytoplasmic and periplasmic fractions of was mixed with 20 ml ofchloroform, allowed to stand for 15 mn the over-expressing cells by assay and by immunoblot. Two- at room temperature, and suspended in 180 ml of 10 mM thirds of the activity was present in the cytoplasmic fraction, but Tris/HCl, pH 7.5. The suspension was centrifuged for 20 min at the specific activity of the periplasmic fraction was 6-fold greater, 6500g. The supernatant was removed, recentrifuged at 8000 g and this fraction was therefore used as the source in the for 20 min, dialysed against 20 mM Tris/HCI, pH 8.0, and run purification, as described in the Experimental section. The results on a column of Q-Sepharose (Sigma) (1.5 cm x 11.5 cm, 20 ml) are summarized in Table 1 and Figure 1. After SDS/PAGE, the equilibrated in the same buffer. The column was eluted with a final preparation of pitrilysin appeared as a single band with gradient to 250 mM NaCl in the Tris buffer at a flow rate of mobility slightly higher than that of,-galactosidase (Mr 116000), 90 ml/h. The fractions (2 ml) were assayed, activity being found consistent with the value of Mr 107000 calculated from the at about 100 mM NaCl. Active fractions were combined and sequence (Figure 1). dialysed against 20 mM potassium phosphate buffer, pH 7.0. The sample was run on a hydroxyapatite column (Bio-Rad) Quenched-fluorescence assay (1 cm x 3 cm, 2.3 ml) and eluted with a gradient (60 ml) of 20-250 mM potassium phosphate buffer, pH 7.0. Active fractions As a result of the finding (see below) of cleavage of vasoactive were combined, dialysed against 25 mM Bistris/HCl, pH 6.3, intestinal peptide (VIP) (16-28) by pitrilysin, a quenched fluores- and run on the Pharmacia f.p.l.c. Mono-P (HR 5/20) column at a flow rate of 0.5 ml/min. The column had been pre-equilibrated with the Bistris buffer, and was developed with Polybuffer 74 Table 1 PurIfication of pitrilysin diluted 10-fold (pH 4.0). Active fractions were combined, dialy- The results are those for a typical preparation from 12 g (wet wt.) of E colicells (strain SK7814) sed into 50 mM Tris/HCl, pH 7.5, glycerol added to give 40 % grown in 3 1 of culture. Activity was determined with QF27 as substrate, and protein (v/v) and stored at -20 'C. concentration was determined by the Bio-Rad assay. In the periplasmic extract, about 17% of the activity was not inhibited by 10 mM EDTA and therefore was attributed to some enzyme Immunological methods other than pitrilysin. An antiserum against pitrilysin was raised in sheep. The antigen Specific was prepared by running the protein obtained from the final step activity of purification on Mono-P, on a 6 % (w/v) polyacrylamide gel. (m-unit/ Puri- The gel was lightly stained with Coomassie Blue and the pitrilysin Protein Activity mg of fication band was excised, rinsed in distilled water and homogenized in Purification step (mg) (m-unit) protein) factor Yield (%) 0.1 0% SDS. Pre-immune serum was collected from sheep to use Periplasmic 174 733 4 (1) (100) as control. For the primary immunization, 100 ,g of the gel extract homogenate was emulsified with an equal volume of Freund's Q-Sepharose 3.4 223 66 15.4 30 complete adjuvant and injected intramuscularly. Two further Hydroxyapatite 0.5 163 326 79 22 injections of 50,ug of antigen were given in the same way at Mono-P 0.05 29 580 152 4 monthly intervals. For immunoblotting, a 10% (w/v) polyacrylamide gel was run and electrophoretically transferred to a nitrocellulose membrane for 2 h with the Bio-Rad Trans Blot System (Bio-Lab lal fbi (c) (d) Laboratories Ltd., Herts., U.K.). The nitrocellulose membrane 10-3 was blocked with 10 % (v/v) donor horse serum in Tris-buffered x Mr saline [TBS; 10 mM Tris/0.9% NaCl, pH 7.4] for 30 min at 37 'C with two changes of buffer. It was rinsed three times in 116 --- TBS containing 0.05% Triton X-100 (washing buffer) before 66 incubating for 16 h at 4 'C with antiserum diluted 1:250 in TBS containing 0.05 % Triton and 100% (v/v) horse serum (diluting buffer). The membrane was again rinsed with several changes of 45 -_ washing buffer, incubated for 1 h at 25 'C with biotinylated 36 -_ antibody [anti-(sheep IgG) antibody] diluted 1:500 in diluting buffer. After extensive washing, the membrane was incubated for 29 -_O 30 min with peroxidase-coupled avidin (Vector Laboratories, Inc., Burlingame, CA, U.S.A.), diluted 1: 300 in diluting buffer, rinsed in TBS and stained with 4-chloro-l-naphthol (Hawkes et 24 --_ al., 1982). The antiserum was run against pure pitrilysin and periplasmic extract in;double immunodiffusion and in immunoblot. Both ...... tests showed a single zone of reaction of the antiserum with pitrilysin, indicating that it was monospecific.

RESULTS AND DISCUSSION Figure 1 Electrophoresis of pitrilysin Purffication of pitrilysin SDS/PAGE with reduction, in 10% (w/v) polyacrylamide is shown, samples being (a) The periplasmic extract of E. coli strain PEOO4 had an initial periplasmic extract, (b) Q-Sepharose fraction, (c) hydroxyapatite fraction and (d) final sample specific activity one-third of that of strain SK7814, and ac- from Mono-P. 604 A. Anastasi, C. G. Knight and A. J. Barrett cent substrate was designed with a related sequence. This was: Table 2 Relative rates of degradation of peptides by pitrilysin Mca-Nle-Ala-Val-Lys-Lys-Tyr-Leu-Asn-Ser-Lys(Dnp)-Leu- The names of the peptides are followed by their amino-acid sequences given in the one-letter Asp-D-Lys, termed QF27. code (except that in the oxidized insulin chain, C = cysteic acid). The values of the apparent When the hydrolysis of QF27 (1 uM) by pitrilysin (0.88,g) first-order rate constants for disappearance of the peptides were determined as described in the was followed to completion over 1 h in the fluorimeter, the Experimental section, and have been normalized by taking the value for insulin B chain as 1. fluorescence increased 20-fold. With nine amino-acid residues Abbreviation used: EGF, epidermal growth factor. between quencher and fluorophore, QF27 is one of the more extended quenched fluorescent substrates so far described. That Relative quenching is still efficient at this separation can be attributed to rate of Bonds the extensive overlap between the Mca emission and Dnp Peptide degradation cleaved absorption spectra (Knight et al., 1992) that allows effective Insulin B-chain (oxidized) energy transfer. (FVNQHLCGSHLVEALYLVCERGFFYTPKA) Assays with QF27 showed accurately measurable rates of Secretin (HSDGTFTSELSRLREGARLQRLLQGLV) 0.3 Multiple hydrolysis, linear with time, for enzyme concentrations in the VIP(1-28) (HSDAVFTDNYTRLRKQMAVKKYLNSILN) 0.23 1 range 0.01-0.1 m-unit/ml (17-170 ng/ml). Previously, pitrilysin VIP(10-28) (YTRLRKQMAVKKYLNSILN) 0.2 4 has been assayed only with radiolabelled insulin (Goldberg et al., QF27 0.2 2 1981; Dykstra and Kushner, 1985) or fragments of f,-galacto- Thyrocalcitonin 0.13 Multiple 1979) as substrates, and the availability (CSNLSTCVLSAYWRNLNNFHRFSGMGFGPETP) sidase (Cheng and Zipser, VIP(16-28) (EMAVKKYLNSVLT) 0.07 3 of a well-characterized substrate permitting continuous assays Substance P (RPKPQQFFGLM) 0.07 2 will facilitate work on the enzyme. Angiotensinogen(1-14) (DRVYIHPFHLLVYS) 0.03 2 The Km of pitrilysin for QF27 was 7.7 + 0.75,M, and k,a, was LH-RH (EHWSYGLRPG) 0.02 2 2.9 s-1, which gave a kcat./Km value of 3.8 x 105 M-1 -s- (taking VIP(1-12) (HSDAVFTDNYTR) 0 Mr = 107000, and assuming that a preparation of 580 m-unit/mg Angiotensin (DRVYIHPFHL) 0 It be noted that two bonds are cleaved Angiotensin II (DRVYIHPF) 0 was fully active). should Dynorphin A(1-13) (YGGFLRRIRPKLK) 0 in QF27, at approx. equal rates (see below), so the rate of Dynorphin A(1-8) (YGGFLRRI) 0 cleavage of each would be about half the figure given. Glucagon(22-29) (FVQWLMNT) 0 QF27 was also found to be a substrate for Drosophila Insulin A chain (GIVEQCCASVCSLYQLENYCN) 0 insulinase. The enzyme was assayed at 37 °C in 20 mM phosphate EGF receptor(1005-1016) (DVVDADEYLIPQ) 0 0.1 mM DTT a-Casein(90-96) (RYLGYLD) 0 buffer containing 0.1 M NaCl, 0.2%o BSA and 0 after activating with 5 mM DTT for 30 min. The Km of Drosophila Bradykinin (RPPGFSPFR) insulinase for QF27 was 18 + 3.3,aM. The products of QF27 hydrolysis by Drosophila insulinase were run on h.p.l.c. and found to be identical to those generated by pitrilysin. Insulin B chain (oxidized) F-V-N-Q-H-L-C-G-S-H-L-V-E-A-L-Y+L-V-C-E-R-G-F-F-Y-T-P-K-A Specificity of pitrilysin VIP (pig) Peptides H-S-D-A-V-F-T-D-N-Y-T-R-L+R-K-Q-M-A-V-K-K-Y-L-N-S-I-L-N The degradation of insulin B chain by pitrilysin was studied by h.p.l.c. Insulin B chain (50,uM) in 50 mM Tris/HCl, pH 7.5, VIP(10-28) (pig) Y-T-R-L-R+K-Q-M-A-V-K-K-Y-L+N+S-I-L-N containing 0.5 % Brij-35 was incubated with 1.5 m-unit of pitrilysin/ml for 15 min at 37 'C. There were two products, VIP(16-28) (chicken) indicating cleavage at a single bond, and 75 % degradation was E-M-A-V-K-K-Y+L+N+S-V-L-T detected after 10 min. When insulin B chain was added to assays of pitrilysin with QF27 as substrate it was found to behave as an Dnp inhibitor with a K1 value of 7.1 ,M. These results suggest that the QF27 insulin B chain is a substrate of pitrilysin, with Km 7.1 ,uM and Mca-Nle-A-V-K-K-Y+L-N+S-K-L-D-k kcat about 4.7 s-'. Figure 2 Points of cleavage of various peptides by pitrilysin Rates of cleavage of a number of smaller peptides were determined as described in the Experimental section, and ex- Points of cleavage were determined as described in the Experimental section. In addition to the pressed relative to that for insulin B chain (Table 2). All of the standard one-letter code for amino acids, C = cysteic acid in the oxidized insulin chain, Nle = norleucine, and k = D-lysine. Peptides isolated and identified by amino-acid analysis rates were lower than that for the B chain (30 residues), despite were, for insulin B chain, 1-16, 17-30; for VIP, 1-13, 14-28; for VIP(10-28), 1-5, 1-14, the fact that most of the peptides were cleaved at two or more 1-15, 6-14, 6-19, 15-19; for VIP(16-28), 1-7,1-8, 1-9, 8-13, 9-13, and for QF27, 1-9, bonds. The smallest of the peptides for which cleavage was 8-14 and 10-14. detectable were substance P (11 residues) and luteinizing hormone-releasing hormone (LH-RH) (10 residues), and higher rates of cleavage were seen with larger peptides. The results indicate that there is a lower limit on the size of a substrate for The points of cleavage of insulin B chain, VIP, VIP(I0-28), pitrilysin close to 10 residues. VIP(16-28) and QF27 were determined as described in the Other potential substrates of less than 10 residues that were Experimental section, with the results shown in Figure 2. With found not to be cleaved by pitrilysin included Abz-Arg-Ala-Leu- each substrate, all susceptible bonds were cleaved at similar rates Tyr-Leu-Val-Lys(Dnp), containing the sequence cleaved by pitri- so that the relative heights of h.p.l.c. peaks for the products did lysin in the insulin B chain, and Mca-Pro-Leu-Gly-Leu-Dpa- not change appreciably with time of incubation. It may well be Ala-Arg-NH2 (known as QF24), a substrate for mammalian that all of the products were of too small a relative molecular matrix metalloendopeptidases (Knight et al., 1992). mass to act as substrates in their own right. Pitrilysin 605

Little has been known about the specificity of pitrilysin for oxyhaemoglobin (Mr 16000), casein (M, 27000), BSA (Mr 66000), cleavage of peptide bonds. Cheng and Zipser (1979) showed phosphorylase A (Mr 97000), and gelatin (Mr 110000). Each cleavage ofinsulin B chain at the bond Tyr-16-Leu-17 (which we protein (100 ,ug) was incubated with 0.15 m-unit of pitrilysin in confirm) and more slowly at Phe-24-Tyr-25. Our results give an 0.1 ml of assay buffer for 2 h at 37 'C. The reaction was stopped indication of a preference for a bulky hydrophobic residue in the by boiling, after which the samples were run on SDS/PAGE. P1 position, but otherwise there was no readily definable specifi- None ofthe proteins was seen to be cleaved. Previously, Goldberg city for amino-acid residues around a scissile bond in the small et al. (1981) were unable to detect cleavage of casein or globin, set of cleavage sites we identified. We saw no cleavage of any but Baneyx and Georgiou (1991) suggested that pitrilysin hydro- peptide bond less than four residues from an N- or C-terminus. lysed a fusion protein in living E. coli. To test whether proteins larger than Mr 7000 might behave like insulin, for which cleavage was also not detectable at high Insulin concentrations of substrate, cytochrome c (Mr 12700) (up to The degradation ofinsulin (5700Mr) by pitrilysin was looked for 5 ,uM), lysozyme (Mr 14400) (up to 20 #uM) and BSA (Mr 66000) by h.p.l.c. as had been done for insulin B chain. At a concentration (up to 100 ,uM) were added to the assay with QF27. No significant of 50 ,uM insulin no cleavage of the protein was detected after inhibition was detected. 2 h. When the insulin concentration was decreased to 1 ,uM, however (which necessitated loading of a larger volume of Lack of inhibition of pitrilysin by ;2-macroglobulln sample on to the h.p.l.c. system), 71 % cleavage of the protein Pitrilysin (10 ,ug) was incubated with a2-macroglobulin (350 ,ug) was detected in 1 h. Insulin was also added to assays of pitrilysin for 2 h at 30 °C. The mixture was run on an Ultrogel AcA 34 with QF27 as substrate, and found to behave as an inhibitor with column, and fractions (1 ml) were collected and assayed against a K1 of 1.1 tM. These results suggest that insulin is a substrate of QF27. Pitrilysin and a2-macroglobulin were treated and chroma- pitrilysin, with a Km of 1.1 ,uM and kcat of 0.03 s-5. From these tographed separately under identical conditions as controls. values, it can be calculated that only 4 % of the substrate would It was found that the activity of pitrilysin that had been pre- have been hydrolysed in the 50 #M experiment, which was at the incubated with a2-macroglobulin eluted in the same position as limit of detection. However, it must be recognized that the effect that of the untreated pitrilysin, showing lack of binding of of the concentration of insulin on the rate of its hydrolysis by pitrilysin to the inhibitor. In a separate experiment, pitrilysin pitrilysin may be somewhat dependent on the oligomerization of activity was assayed with QF27 in the presence of excess a2- insulin at higher concentrations (Pocker and Biswas, 1981). macroglobulin, and no inhibition was detected. The low apparent values ofboth kcat and Km for pitrilysin with We conclude from these results that pitrilysin does not interact insulin may account for the fact that it is only at very low with a2-macroglobulin, which is consistent with the fact that the concentrations of substrate that a substantial fraction of the enzyme does not act on large proteins since the interaction substrate is hydrolysed under typical experimental conditions. requires the trapping reaction that is initiated by cleavage of the Much the same has been found for insulinase, for which Km macroglobulin in the bait region (Starkey and Barrett, 1977). values in the order of 0.1 ,uM have been reported (Duckworth, 1988). For both enzymes, radiolabelled insulin at very low concentrations can be used as a test substrate (Goldberg et al., Metal-dependence of pitrilysin 1981; Duckworth, 1988). The characteristics of low kcat and low Effect of chelating agents on pitrilysin Km are not unlike those of small protein inhibitors of proteinases (Laskowski and Kato, 1980), and also explain why it is easy to Solutions of pitrilysin in 50 mM Tris/HCl buffer, pH 7.5, con- form covalently cross-linked enzyme-substrate complexes with taining 0.5 % Brij-35 were made 10 mM EDTA, 10 mM tetra- these enzymes (Ding et al., 1992). ethylenepentamine or 1 mM 2,6-pyridinedicarboxylic acid. At Like insulin, glucagon (29 residues) at a concentration of 60 min, samples were removed for assay in the standard system 50 ,tM was not seen, by h.p.l.c., to be cleaved by pitrilysin after with 4,uM QF27, 0.02 m-unit of enzyme/ml and the same 2 h. On reducing the glucagon concentration to 5 ,M, 590% concentration of chelating agent as was used in the pre- cleavage was detected in 30 min. When added to QF27 assays of incubation. Complete inactivation of pitrilysin was observed in pitrilysin, glucagon, like insulin, behaved as an inhibitor, with a all cases. 1,10-Phenanthroline (1 mM), added directly to the K, of 12.7 ,uM. These results suggest that glucagon is a substrate assay, caused complete and instantaneous inactivation. In con- of pitrilysin with a Km value of 12.7 ,uM and a kcat of 0.5 s-'. trast, the non-chelating analogue, 4,7-phenanthroline showed From these values, it is calculated that 78 % of the substrate only a slight (13 %) and instantaneous inactivation at I mM. should have been hydrolysed in the h.p.l.c. experiment. The fact that no hydrolysis was detected may be attributable to the Reactivation with metals tendency of glucagon to aggregate at high concentrations, The reactivation of pitrilysin by metal ions was followed by the limiting susceptibility to proteolysis (Rose et al., 1988). method described by St6cker et al. (1988) for astacin. The Dynorphin A(I-13) proved to be an inhibitor of pitrilysin in enzyme was inactivated with 0.3 mM 2,6-pyridinedicarboxylic the QF27 assay, with a K, of 13.3 ,tM. No hydrolysis was acid, for 10 min before metal ions were introduced (Table 3). observed on h.p.l.c. after incubating 50 ,uM dynorphin A(1-13) Generally, all metal ions were used at a concentration of0.5 mM, with pitrilysin at 37 °C for 2 h, but we do not exclude the which was found to give maximal activation. For Zn2+, the possibility that this may also be a substrate of low Km and low activity was 200% lower at 0.5 mM than at 0.1 mM. At higher keat.' concentrations none of the metal ions showed increased activation except Ca2 , which gave 100 % reactivation at 1 mM. Zn2+ was the most effective ion in reactivating the enzyme Proteins [consistent with the recent finding that pitrilysin contains zinc To determine whether larger proteins are cleaved by pitrilysin, (Ding et al., 1992)]. Less effective metals were Co2+, Ca2` and tests were made with several polypeptide chains of varying Mr: Mn2+. Activation appeared instantaneous with Zn2+, Co2' and 606 A. Anastasi, C. G. Knight and A. J. Barrett

Table 3 ReactIvation of pitrilysin by metal ions in the presence of 0.3 mM 100' 2,6-pyridinedicarboxylic acid Activity is expressed relative to that of the untreated enzyme. 80 Concentration Activity Metal ion (mM) (% of control value) 0-0 60

None 6 ._1 Zn2+ 0.1 96 40 co2+ 0.5 75 Ca2+ 0.5 70 1 100 20 Mn2+ 0.5 52 cu2+ 0.5 6 Cd2+ 0.5 6 0 200 400 600 800 1000 Mg2+ 0.5 6 Metal ions (uM) Figure 3 InhIbition of pitrilysin by metal ions

Pitrilysin activity was assayed with QF27 as described in the Experimental section. Various concentrations of Zn2+(@), Co2+ (A), Mn2+ (U) and Ca2+ (0) were added to the assay.

Mn2+, but not with Ca2+, for which the activity increased pro- gressively over 2-3 min. Further activation by Zn2+ was observed Table 4 The effect of zinc on the Inhibition of pitrilysin by bacitracin following Mn2+ and Ca2 , giving 96 and 90 % activity respectively. Enzyme activity was determined with QF27 as substrate. Zinc (7%, w/w) was already present Although Cu2+, Cd2+ and Mg2+ did not reactivate the enzyme, in the bacitracin sample (Aldrich no. 85187-6) as supplied. The other samples, Aldrich no. they did not prevent subsequent activation by 0.1 mM Zn2+, 85186-8 and Fluka no. 11702, did not contain zinc, and ZnCI2 was added as indicated before which resulted in 88, 49 and 105 % activity respectively. the bacitracin was introduced into the assay. The two values marked (*) should be regarded The effects of metal ions on pitrilysin may be compared with as IC50 values, since the results did not fit the Morrison (1969) equation satisfactorily. those for other metallopeptidases. Neprilysin (EC 3.4.24.11) is reactivated by Zn2+, Mn2+, Co2+ and Ca2+ (in order of decreasing Supplier Zinc (atom/mole) K (guM) effectiveness) after treatment with EDTA (Kerr and Kenny, 1974). For thimet oligopeptidase (EC 3.4.24.15), the corre- Fluka (11702) 149* sponding order was Zn2+, Mn2 , Ca2 , Co2+, Cd2+ (Barrett and 1.6 (added) 32 Brown, 1990). Astacin (EC 3.4.24.21) was reactivated by Zn2+, Aldrich (85186)8) 168* Co2+ or Cu2+, but not by Mn2+ or Ca2+ (Stocker et al., 1988). 0.8 (added) 41 1.6 (added) 23 Aldrich (85187-6) 1.6 (present) 12.9 Inactivation of pitrilysin by metal ions

At 1 mM, several metal ions showed inhibition ofpitrilysin. Thus Zn2+, Co2+, Mn2+ and Ca2+ inhibited by 90, 32, 15 and 90% respectively (Figure 3). Inhibition by excess zinc and other metal is seen with thermolysin (K1 0.48 ,uM) (Nishino and Powers, ions has been reported for many other metallopeptidases, and a 1979). mechanism for the effect of zinc on carboxypeptidase A has been proposed (Larsen and Auld, 1989). Bacitracin The antibiotic bacitracin was found to inhibit pitrilysin (Table 4). The preparation supplied containing zinc (Aldrich 85187-6) Inhibitors of pitrilysin was much more inhibitory than the others. Addition ofzinc to the DTT preparations not originally containing the metal ion increased their inhibitory activity greatly, whereas the same concentrations DTT (5 mM) caused slowly progressive inhibition of pitrilysin, of Zn2+ added in the absence of bacitracin did not inhibit. An reaching 93 % at 30 min. It was previously reported that 5 mM interpretation of these results would be that the conformation of DTT inhibited the enzyme by 60 % (Goldberg et al., 1981). Since the zinc-bacitracin complex is such that it is a more potent pitrilysin contains only a single residue of cysteine (Finch et al., inhibitor than free bacitracin. 1986), it can contain no disulphide bond, so the inhibition by Inhibition of insulinase has been reported with 0.5-0.7 mM DTT is presumably due to binding to the zinc atom. bacitracin, suggesting that inhibition is of the same order as for pitrilysin (Shii et al., 1986; Garcia et al., 1988; Ding et al., 1992).

Zincov Antiserum to pitrilysin Pitrilysin was weakly inhibited by the dipeptide hydroxamic acid The antiserum to pitrilysin was found to inhibit pitrilysin activity Zincov. At 0.5, 1.0 and 2.0 mM, 48, 65 and 750% inhibition when assayed with QF27. The enzyme (0.14 m-unit in 0.01 ml respectively, were obtained. A metalloendopeptidase from venom assay buffer) was pre-incubated for I h at 25 °C with an equal of the Southern Copperhead snake also was weakly inhibited volume of antiserum. Inhibition of activity (69 %) was detected, (Ki 2.9 mM) (Guan et al., 1991), but much more potent inhibition whereas no inhibition was produced by pre-immune serum. Pitrilysin 607

Potential inhibitors tested Drosophila insulinase. Mrs. M. A. Brown and Mrs. L. Handford provided excellent technical assistance. Other inhibitors of metallopeptidases that were tested but found not to be inhibitory at the concentrations used were: captopril (100 ,uM), thiorphan (50 ,uM) bestatin (100 ,aM), amastatin REFERENCES (100 ,uM), (2S,3R)-3-amino-2-hydroxy-4-(4-nitrophenyl)butan- Ames, G. F.-L., Prody, C. and Kustu, S. (1984) J. Bacteriol. 160, 1181-1183 oyl-L-leucine (100,M), Bz-PheK-Gly-Pro-Ala (100,uM), cinna- Atherton, E. and Sheppard, R. C. (1989) Solid-Phase Peptide Synthesis. A Practical moyl-LeuK-Gly-Pro-Ala (100 jcM), cinnamoyl-LeuK-Gly-Pro- Approach, IRL Press, Oxford Leu (100 ,uM), cinnamoyl-Leu'-Gly-Pro-Leu-OMe (100 ,uM), Baneyx, F. and Georgiou, G. (1991) J. Bacteriol. 173, 2696-2703 cinnamoyl-Leu-Gly-Pro-Arg (100 ,uM), Barrett, A. J. and Brown, M. A. (1990) Biochem. J. 271, 701-706 Z-Pro-Phe'-Gly-Pro- Barrett, A. J. and Rawlings, N. D. (1992) Biol. Chem. Hoppe-Seyler 373, 353-360 Ala (100 ,uM), cinnamoyl-Leu'-Gly-Pro-Arg-OMe (100 ,uM), N- Barrett, A. J., Knight, C. G., Brown, M. A. and Tisljar, U. (1989) Biochem. J. 260, 259-263 [I (RS)-carboxy-3-phenylpropyl]-Ala-Ala-Phe-p-aminobenzoate Becker, A. B. and Roth, R. A. (1992) Proc. Natl. Acad. Sci. U.S.A. 89, 3835-3839 (50,tM), N-[ 1 (RS)-carboxy-2-pheniylethyl]-Ala-Ala-Phe-p- Bradford, M. M. (1976) Anal. Biochem. 72, 248-254 aminobenzoate (50 ,tM), N-[l(RS)-carboxy-3-phenylpropyl]- Bury, A. F. (1981) J. Chromatogr. 213, 491-500 Ala-Ala-Tyr-p-aminobenzoate (50 ,uM), N-[1(RS)-carboxybu- Cheng, Y.-S. E. and Zipser, D. (1979) J. Biol. Chem. 254, 4698-4706 tyl]-Ala-Ala-Phe-p-aminobenzoate (50 ,M), and tissue inhibitor Davies, M. E., Coughlan, R. and Barrett, A. J. (1987) Arthritis Rheum. 30, 872-877 Ding, L., Becker, A. B., Suzuki, A. and Roth, R. A. (1992) J. Biol. Chem. 267, 2414-2420 of metalloproteinases (0.3 ,uM). Duckworth, W. C. (1988) Endocr. Rev. 9, 319-345 Dykstra, C. C. and Kushner, S. R. (1985) J. Bacteriol. 163, 1055-1060 Finch, P. W., Wilson, R. E., Brown, K., Hickson, I. D. and Emmerson, P. T. (1986) Nucleic Conclusions Acids Res. 14, 7695-7703 Pitrilysin is an oligopeptidase, in that it acts on oligopeptides and Garcia, J. V., Fenton, B. W. and Rosner, M. R. (1988) Biochemistry 27, 4237-4244 Goldberg, A. L., Swamy, K. H. S., Chung, C. H. and Larimore, F. S. (1981) Methods polypeptides, but not on proteins. It thus a group of such joins Enzymol. 80, 680-702 endopeptidases restricted in substrate size (Barrett and Rawlings, Guan, A. L., Retzios, A. D., Henderson, G. N. and Markland, F. S., Jr. (1991) Arch. 1992). The lack of inhibition by a2-macroglobulin is a general Biochem. Biophys. 289, 197-207 property of oligopeptidases that distinguishes them from the Hawkes, R., Niday, E. and Gordon, J. (1982) Anal. Biochem. 119, 142-147 great majority of endopeptidases. Kerr, M. A. and Kenny, A. J. (1974) Biochem. J. 137, 489-495 Because of the lack of the HEXXH consensus sequences in Knight, C. G., Willenbrock, F. and Murphy, G. (1992) FEBS Lett. 296, 263-266 Larsen, K. S. and Auld, D. S. (1989) Biochemistry 28, 9620-9625 pitrilysin, we re-examined the metal dependence of the enzyme, Laskowski, M., Jr. and Kato, I. (1980) Annu. Rev. Biochem. 49, 593-626 but our findings are consistent with the general view that this is Morrison, J. F. (1969) Biochim. Biophys. Acta 185, 269-286 a zinc metalloenzyme. However, many compounds that are Nishino, N. and Powers, J. C. (1979) Biochemistry 18, 4340-4347 potent inhibitors of other metalloendopeptidases had only weak Pocker, Y. and Biswas, S. B. (1981) Biochemistry 20, 4354-4360 effects on pitrilysin. Rawlings, N. D. and Barrett, A. J. (1990) Comput. Appl. Biosci. 6, 118-119 The synthetic substrate, QF27, greatly facilitates work with Rawlings, N. D. and Barrett, A. J. (1991) Biochem. J. 275, 389-391 pitrilysin, and is likely to be valuable for mammalian insulinase Rose, K., Savoy, L.-A., Muir, A. V., Davies, J. G., Offord, R. E. and Turcatti, G. (1988) Biochem. J. 256, 847-851 as well. Sambrook, J., Fritsch, E. F. and Maniatis, T. (1989) Molecular Cloning. A Laboratory Manual, Cold Spring Harbor Laboratory Press, New York We thank Professor P. T. Emmerson (University of Newcastle upon Tyne, Shii, K., Yokono, K., Baba, S. and Roth, R. A. (1986) Diabetes 35, 675-683 Newcastle, U.K.) and Dr. S. R. Kushner (University of Georgia, Athens, GA, U.S.A.) Starkey, P. M. and Barrett, A. J. (1977) in Proteinases in Mammalian Cells and Tissues for providing strains of E colias described in the text. Drs. R. E. Galardi (University (Barrett, A. J., ed.), pp. 663-696, North-Holland Publishing Company, Amsterdam of Kentucky, KY, U.S.A.), M. Orlowski (Mount Sinai Medical School, New York, N.Y., Stocker, W., Wolz, R. L., Zwilling, R., Strydom, D. J. and Auld, D. S. (1988) Biochemistry U.S.A.) and G. Murphy (this laboratory) kindly gave potential inhibitors. Dr. B. D. 27, 5025-5032 Gehm (University of Chicago, IL, U.S.A.) generously donated the sample of Yasuda, S. and Takagi, T. (1983) J. Bacteriol. 154, 1153-1161

Received 13 July 1992/27 August 1992; accepted 22 September 1992