Proc. NatI. Acad. Sci. USA Vol. 89, pp. 3835-3839, May 1992 Biochemistry An unusual active site identified in a family of zinc metalloendopeptidases (insulin degrading enzyme/nitochondrial processing protease) ANDREW B. BECKER AND RICHARD A. ROTH Department of Pharmacology, Stanford University School of Medicine, Stanford, CA 94305 Communicated by Brian W. Matthews, January 13, 1992 (received for review December 6, 1991)
ABSTRACT An unusual active site has been identified in dopeptidases, many of which show little overall sequence a family of zinc metalloendopeptidases that includes bacterial identity to thermolysin outside of this domain (4, 5). Re- protease M and the human and Drosophila insidn-degrading cently, through mutagenesis studies, the sequence HEXXH enzymes. All of these enzymes have been characterized as has been confirmed to be at the active site of several ofthese metalloendopeptidases and purified protease HI has been zinc-dependent proteases, including neutral endopeptidase shown to contain stoichiometric levels of zinc. However, all and angiotensin-converting enzyme (6-8). Moreover, this three proteases lack the consensus sequence (EIEXXH) de- consensus sequence has even been used to predict the scribed in the active site of other zinc metalloendopeptidases. peptidase activity ofproteins not previously considered to be Instead, these proteases contain an inversion of this motif, proteases, such as leukotriene A4 hydrolase (9). HXXEH. To determine whether this region could represent the However, some proteases, based on their amino acid active site in these proteins, the two histidines in protease m sequence, do not fit into the currently recognized classifica- were individually mutated to arginine and the glutamate was tion system. One example is the carboxypeptidases, zinc- mutated to glutamine. All three mutants were devoid of dependent metalloexopeptidases, in which the metal binding proteolytic activity toward an exogenous substrate, insulin, as domain consists of the motif HXXE(X)123H (4) as opposed to compared to the wild-type protease. Three lines of evidence the more common active site seen in thermolysin. Another indicate that this loss of activity in the mutants is not due to example is the bacterial enzyme protease III, or pi, from distortion of the three-dimensional structure of the protein: (i) Escherichia coli (10-13). This periplasmic enzyme shows the mutants are secreted into the periplasmic space and chro- 27% overall sequence identity with a human enzyme, called matograph normally; (ia) all three mutants are expressed at insulin-degrading enzyme (14), and 28% identity with a levels nearly identical to wild-type protein and do not appear homologous enzyme from Drosophila melanogaster (15). In to have an increased susceptibility to proteolysis in the bacteria; addition, three regions in these enzymes show >50% se- and (iii) the mutants compete equally with wild-type protein in quence identity (10, 14, 15). Although the physiological roles a radioimmunoassay. The purified wild-type and glutamate of these three enzymes are not known, they all degrade mutants were found to contain stoichiometric amounts of zinc insulin (10-17). All three proteases also have molecular by atomic absorption spectrophotometry, whereas both histi- masses of "100 kDa and, based on their requirement for dine mutants had negligible zinc signals. These findings are divalent cations, have been classified as metalloendopepti- consistent with this region being the active site in this protein, dases. Moreover, protease III has recently been shown to with the contain zinc at stoichiometric levels (18). However, none of histidine residues coordinating the essential zinc atom these enzymes contains the consensus active site sequence and the glutamate involved in catalysis. HEXXH described above for zinc metalloendopeptidases. A careful alignment of the sequences of these three pro- Proteases play a variety of roles in many physiological teases reveals the conservation of 12 histidines and gluta- processes, including fibrinolysis, blood homeostasis, com- mates, which could function in zinc binding (10, 14, 15). In plement activation, digestion, and hormone processing (1). fact, a number of motifs resembling that of the carboxypep- Based on common active sites and mechanisms of action as tidases can be identified in all three enzymes, including determined by primary amino acid sequence, inhibitor and HXXE(X)138H. In addition, all three enzymes contain a substrate profiles, and x-ray crystallographic analyses (2), sequence, HXXEH, which could be considered an inversion these proteins are generally classified into one of four fam- ofthe active site consensus sequence seen in thermolysin and ilies. Representative members of each protease class (i.e., other metalloendopeptidases (10, 14, 15). This sequence is cysteine, serine, aspartyl, and metallo-) have been crystal- found in one of the three highly conserved regions present in lized and their active sites elucidated (2). For example, the these enzymes. The surrounding residues of this sequence bacterial enzyme thermolysin is the prototypic zinc metal- also appear to be inverted in comparison to other zinc loendopeptidase. X-ray crystallographic studies have shown metalloendopeptidases (Table 1). The spacing of the two that this enzyme contains two histidines (His-142 and His- histidines and the first histidine to the glutamate in this region 146) and a glutamic acid (Glu-143) at its active site, with the are consistent with the one to three amino acids separating histidines coordinating the binding of a zinc atom and the the zinc binding residues in more than a dozen zinc enzymes glutamate acting as a general base in catalysis (3). Based on whose structures have been analyzed (4) and are the only two studies in thermolysin, it has been proposed that the active histidines and/or glutamates conserved among all three pro- site sequence in metalloendopeptidases is HEXXH (where X teases that exhibit the requisite spacing. is any amino acid, H is histidine, and E is glutamate). This To test whether the residues in the apparently inverted sequence has been identified in well over 15 zinc-dependent sequence HXXEH comprise the active site of protease III, metalloproteases, including aminopeptidases and metalloen- the two histidine residues were individually mutated to arginine and the glutamate was changed to glutamine by The publication costs ofthis article were defrayed in part by page charge site-directed mutagenesis. (The mutants were named by payment. This article must therefore be hereby marked "advertisement" using the single-letter amino acid code and indicating the in accordance with 18 U.S.C. §1734 solely to indicate this fact. number of the residue that was mutated-H88R, histidine-88 3835 Downloaded by guest on September 24, 2021 3836 Biochemistry: Becker and Roth Proc. Natl. Acad. Sci. USA 89 (1992) Table 1. Comparison of the proposed active site residues in Coomassie blue-stained gels, were pooled. The protease peak bacterial protease III, the human (hIDE) and Drosophila was concentrated to 1 ml in an Amicon chamber with a (dIDE) insulin-degrading enzymes, and several other YM-30 membrane (Amicon) and then applied to an Ultrogel zinc metalloendopeptidases AcA34 column (1.5 x 90 cm) equilibrated with a buffer Enzyme Sequence Residues Ref(s). containing 10 mM Hepes (pH 7.6) and 0.02% sodium azide. Again, the peak fractions as determined by Coomassie blue- . * ** Protease III GLA H YL ER MSL 85-95 10 stained gels were pooled and then concentrated in a Centri- hIDE GLS H FC EH MLF 106-116 14 con 30 (Amicon) to 1-1.5 ml. dIDE GLA H FC ER MLF 78-88 15 Insulin-Degradation Assays. The indicated amounts of pro- ** * . tein relative to wild-type protease III (determined by radio- Thermolysin VVA HE LT H AVT 139-149 3 immunoassays and Coomassie blue-stained gels) were incu- Collagenase VAA HE LG H SLG 215-225 19, 20 bated in microtiter wells previously coated with an antibody Serratia protease TFT HE IG H ALG 89-99 21 directed against protease III (18). The wells were washed and Endopeptidase 24.11 VIG HE IT H GFD 580-590 22,23 the immunocaptured enzyme was tested for its ability to Leukotriene A4 degrade 125I-labeled insulin (1251-insulin). Degradation of in- hydrolase VIA HE IS H SWT 292-302 9 sulin was assessed by its loss of ability to bind to the insulin The three active site residues, two histidines and a glutamate, in receptor as described (18). All assays were performed in a number of proteases are marked with an asterisk. Underlined triplicate using 40,000 cpm of 1251-insulin in a final vol of 35 residues indicate residues adjacent to the active site that are inverted p.l for 30 min at 370C. The assays were terminated by addition in the insulin-degrading enzymes and protease III relative to other of bacitracin to a final concentration of 1 mg/ml, and the metalloproteases. The diamond is used to mark residues that are extent of degradation was assessed in a receptor binding invariably hydrophobic in the different metalloendopeptidases. assay (18). Radioimmunoassay of Wild-Type and Mutant Protease III to arginine; H92R, histidine-92 to arginine; E91Q, gluta- Proteins. The indicated amounts of each protein were incu- mate-91 to glutamine.) The mutant proteins were then com- bated with 50,000 cpm of 251I-protease III on microtiter wells pared to wild-type protein in terms of activity and zinc previously coated with an antibody directed against protease content to ascertain whether this region is the active site in III (18). The binding was performed overnight at 40C and the this protein family. decrease in binding of iodinated protein to the antibody relative to a control with no competitor was calculated as the MATERIALS AND METHODS percent inhibition of binding. All assays were performed in triplicate. Materials. The following were purchased: standard SDS Metal Determinations. The amount of zinc in each purified molecular weight markers from Bio-Rad, bacitracin from protein preparation was determined by atomic absorption Sigma, phenyl-Sepharose from Pharmacia, Ultrogel AcA-34 spectrophotometry (Perkin-Elmer, model 2380). All assays from LKB, and zinc standards for atomic absorption spec- were performed in triplicate and a zinc standard curve was trophotometry from Alfa Products, Thiokol-Ventron Divi- established by using a zinc standard solution designed for sion (Danvers, MA). Insulin and purified protease III were atomic absorption spectrophotometry. The amount of zinc in labeled by the Iodo-Gen method (Pierce). each protease preparation was determined after subtracting Site-Directed Mutagenesis of Protease III. The protease III the background readings from an equivalent amount of so- gene (10) was subcloned into the EcoRV and HindIII sites of lution from the parental cell line prepared in the same manner the BlueScript (KS) vector (Stratagene) and then used to as the purified proteins. The amount subtracted represented transform the bacterial strain TG-1. Single-stranded phage- -500 pg of zinc (385 nM). mids of the BlueScript/protease III construct were rescued from the bacteria by using K07 helper phage and then RESULTS AND DISCUSSION purified by a procedure accompanying the T7-Gen in vitro mutagenesis kit (United States Biochemical). Site-directed The three mutant and wild-type protease III proteins were mutagenesis oligonucleotides were synthesized and used in expressed using a bacterial expression plasmid containing a conjunction with the T7-Gen kit to make site-directed mu- Tac promoter, and periplasmic shockates containing these tants of protease III. Confirmation of mutagenesis was ob- proteins were prepared as described (18). Comparable tained by sequencing the mutant clones with Sequenase amounts of wild-type protease III and the three mutant (United States Biochemical). proteins were assayed for proteolytic activity against an Overexpression and Purification of Mutant and Wild-Type exogenous substrate, insulin, after immunoimmobilization Protease m Proteins. The wild-type and mutant protease III (Fig. 1). The amount of each protein in the periplasmic genes were subcloned into the EcoRV and HindIII sites of a shockates was determined relative to the other proteins by a modified Tacterm vector and were used to transform com- radioimmunoassay using a monoclonal antibody directed petent JM101 bacteria as described (18). Expression of the against protease III (18) and these data were subsequently proteins was induced with the addition of 2 mM isopropyl confirmed with Coomassie blue-stained gels (Figs. 2 and 3A). ,B-D-thiogalactopyranoside for 24-48 hr in 4 liters of modified The wild-type protease readily degraded iodinated insulin; M9 medium (18). Bacteria were shifted from hypertonic to however, the three mutants were devoid of activity (Fig. 1). hypotonic media to release periplasmic proteins containing Similar data were obtained when purified proteins in a the wild-type and mutant protease III proteins as described solution-phase assay were used. (18). The loss of enzymatic activity in the mutated protease III The periplasmic extracts were then brought up to 2 M NaCl proteins could have resulted from a general distortion oftheir and applied to a 5- to 10-ml phenyl-Sepharose column pre- three-dimensional structure as a result of mutagenesis. Sev- viously equilibrated with 2 M NaCl in 10 mM Hepes buffer eral lines of evidence argue against this possibility. First, the (pH 7.6). The column was washed extensively with 2 M NaCl mutant proteins are secreted into the periplasmic space of E. buffer and then eluted with a decreasing salt gradient (2-0 M coli and chromatograph identically to wild-type protease III NaCl) and an increasing ethylene glycol gradient (0-50%) in on hydrophobic and sizing columns. The mutant proteins are a 10 mM Hepes buffer. Approximately 60 (3 ml) fractions also expressed at levels comparable to the wild-type protein were collected and the peak fractions, as determined by and do not appear to have an increased susceptibility to Downloaded by guest on September 24, 2021 Biochemistry: Becker and Roth Proc. Natl. Acad. Sci. USA 89 (1992) 3837 100 Mutant protease, ,tl 5 10 80 -
*0 60- 0a
40- c ._ 0 U) ._ c * 20- ._-0
0 .q)a0 K) =QF
I I 0 10 20 30 Equivalents of wild-type protease 0 10 20 30 FIG. 1. Degradation of 125I-insulin by wild-type (n) and mutant Wild-type protease, Atl protease III proteins. H88R (n), E91Q (e), and H92R (o). The indicated amounts of protein relative to wild-type protease III FIG. 2. Radioimmunoassay of wild-type (m) and mutant protease (determined by radioimmunoassays and Coomassie blue-stained III proteins. H88R (n), E91Q (o), and H92R (e). The indicated gels) were assayed for the ability to degrade 1251-insulin as described. amounts of each protein were incubated with 50,000 cpm of 1251_ All assays were performed in triplicate. protease III on microtiter wells previously coated with an antibody directed against protease III (18). Upper axis indicates microliters of proteolysis in the bacteria. Finally, the mutant proteins the three mutants and lower axis indicates microliters of wild-type compete equally with wild-type protease in a radioimmuno- protein used in the assay. Results shown are means + SD oftriplicate assay with 125I-protease III and a monoclonal antibody di- determinations. rected against the protease (Fig. 2). The wild-type and mutant proteases were purified to near tained stoichiometric amounts ofzinc, while the two histidine homogeneity by a two-column purification procedure, a mutants had negligible zinc signals. phenyl-Sepharose column followed by an ACA34 sizing These results indicate that the two histidine residues in the column (Fig. 3B). The amount of each purified protein was conserved motif HXXEH are involved in binding the essen- then determined by quantitative amino acid determination of tial zinc molecule in the active site. In contrast, the glutamate 3-7 jig of each preparation. The calculated amount of each did not appear to be required for zinc binding. However, all amino acid as determined through this procedure was nearly three amino acids do appear to play an important function in identical in all preparations and matched closely the amino the catalytic activity of this protein. These findings are acid composition predicted by the DNA sequence for this consistent with the hypothesis that HXXEH represents an protein (10). The amount of zinc in each purified protein was inverted active site sequence in comparison to that present in then determined by atomic absorption spectrophotometry on thermolysin. However, definitive confirmation of these find- two different and independent preparations of each protein ings must await crystallization and elucidation of the struc- (Table 2). The wild-type and glutamate mutant both con- ture of at least one member ofthis unusual class ofproteases. A B a b c d e f a b c d e f
- 200 OF WS 200 ... -116 _m law - 116 PTR __- __ __ lo - 92 PTR ------92
* -66 ' r % -66
_p -45 _b 45
V
FIG. 3. Coomassie blue-stained SDS gels of periplasmic extracts (A) and purified proteins (B). For the extracts, the amount of protease III and the three mutants was determined by radioimmunoassay. Lanes: a, molecular mass markers (kDa); b, wild-type protease III; c, H88R; d, E91Q; e, H92R; f, molecular mass markers. For the purified proteins, the amount of protease III and the three mutant proteins was based on Bradford assays. Lanes: a, molecular mass markers; b, H92R; c, E91Q; d, H88R; e, wild-type protease III; f, molecular mass markers. The location of protease III (PTR) is indicated. Downloaded by guest on September 24, 2021 3838 Biochemistry: Becker and Roth Proc. Natl. Acad. Sci. USA 89 (1992) Table 2. Amount of zinc in purified wild-type protease III and ber of this class of zinc metalloendopeptidases is the mito- three mutants as determined by atomic absorption chondrial-processing protease (26, 27). This protease is com- spectrophotometry posed oftwo separate subunits, the mitochondrial-processing mol of zinc protease (MPP) and the processing-enhancing protein (PEP) Enzyme Zinc, juM Protein, AM per mol of protein (26, 27). Both of these peptides have a significant degree of homology to the first conserved region between insulin- Wild type 0.068 0.55 1.14 ± 0.08 degrading enzyme and protease III (24). However, only PEP Mutant H88R has the proposed active site motifHXXEH (24, 26, 27). A role Preparation 1 <0 1.1 ND for this sequence as the active site of the mitochondrial Preparation 2 <0 1.1 ND enzyme is consistent with the reported requirement of both Mutant E91Q subunits for activity (28, 29) but is inconsistent with some Preparation 1 1.12 1.1 1.02 ± 0.13 claims that a low residual activity is found in preparations of Preparation 2 1.24 1.1 1.09 ± 0.08 purified MPP (30). Additional studies are therefore required Mutant H92R to test the role of these residues in the enzymatic activity of Preparation 1 <0 1.1 ND the mitochondrial enzyme. Preparation 2 <0 1.1 ND Based on biochemical characteristics such as molecular The amount of zinc in each preparation of purified protein was mass and inhibitor profiles, several as yet uncloned metal- determined by atomic absorption spectrophotometry on three inde- loendopeptidases may also be members of the same class of pendent samples. The amount of each protein was determined by proteases as the mammalian and bacterial insulin-degrading amino acid quantitation and a zinc standard curve was established by enzymes. These include using a zinc standard solution designed for atomic absorption spec- enzymes that have been proposed to trophotometry. ND, not determined because the zinc readings were play a role in antigen processing in lymphocytes (31), differ- below the parental cell background. entiation of myoblasts (32), exocytosis in mast cells (33), intracellular membrane fusion in fertilization in the sea urchin While residues other than those studied in this report have (34), degradation of atrial natriuretic factor in kidney (35), been proposed as the metal binding ligands in this class of and, most recently, degradation ofthe antimicrobial magainin proteases (24), it is clear that the two histidines in the peptides found in Xenopus (36). Elucidation ofthe amino acid proposed active site are essential for zinc binding. However, sequences of these enzymes will be necessary to determine if these proteins are analogous to other classes of zinc whether these molecules also belong to the same class of metalloendopeptidases, then one can anticipate that another proteases as bacterial protease III and the human and Dro- glutamate or histidine residue downstream from the active sophila insulin-degrading enzymes and whether they contain site will also contribute to zinc binding (3, 4). There are the proposed active site sequence HXXEH. several glutamate and histidine residues conserved between We thank Dr. Sidney Kushner for the original protease III con- all members of this protein family that are a suitable distance struct, Drs. Paul Galnick and Charles Yanofsky forthe PTAC vector, from the proposed active site to be this third zinc binding and Dr. Edward Solomon for use of the atomic absorption spectro- residue. Which of these residues comprises the third zinc photometer. This work was supported by National Institutes of binding site remains to be determined. Health Grant DK34926. When the residues adjacent to the active site histidines and glutamate in over a dozen zinc-dependent metalloendopep- 1. Neurath, H. (1989) Trends Biochem. Sci. 14, 268-271. tidases were analyzed, it was found that in a region of 10 2. Bond, J. S. & Butler, P. E. (1987) Annu. Rev. Biochem. 56, amino acids, from 3 upstream to 3 downstream of the 333-364. 3. Matthews, B. W. (1988) Acc. Chem. Res. 21, 333-340. conserved active site, there were no charged amino acids 4. Vallee, B. L. & Auld, D. S. (1990) Biochemistry 29,,5647-5659. except for the 3 active site residues (5). However, 2 recently 5. Jongeneel, C. V., Bouvier, J. & Bairoch, A. (1989) FEBS Lett. cloned metalloendopeptidases, meprin A and N-benzoylL- 242, 211-214. tyrosylp-aminobenzoic acid hydrolase, are exceptions to this 6. Devault, A., Sales, V., Nault, C., Beaumont, A., Roques, B., rule; each contains a charged glutamate residue adjacent to Crine, P. & Boileau, G. (1988) FEBS Lett. 231, 54-58. the first histidine of the active site (25). In addition, further 7. Devault, A., Nault, C., Zollinger, M., Fournie-Zaluski, M.-C., analysis ofthe different indicates that Roques, B. P., Crine, P. & Boileau, G. (1988) J. Biol. Chem. metalloendopeptidases 263, 4033-4040. there is invariably a hydrophobic residue 2 amino acids 8. Wei, L., Alhenc-Gelas, F., Corvol, P. & Clauser, E. (1991) J. downstream from the second active site histidine (5). This Biol. Chem. 266, 9002-9008. hydrophobic residue is conserved in meprin A and in paba- 9. Medina, J. F., Wetterholm, A., Radmark, O., Shapiro, R., peptide hydrolase (25). These findings suggest that the sig- Haeggstrom, J. Z., Vallee, B. L. & Samuelsson, B. (1991) nature sequence at the active site of most zinc-dependent Proc. Natl. Acad. Sci. USA 88, 7620-7624. metalloendopeptidases is UUZHEUUHUBU (where U rep- 10. Finch, P. W., Wilson, R. E., Brown, K., Hickson, I. D. & resents uncharged residues, B signifies hydrophobic resi- Emmerson, P. T. (1986) Nucleic Acids Res. 14, 7695-7703. 11. Dykstra, C. C. & Kushner, S. R. (1985) J. Bacteriol. 163, dues, and Z represents a residue that can be either charged 1055-1059. or uncharged). In the proposed active site ofprotease III and 12. Cheng, Y.-S. E. & Zipser, D. (1979) J. Biol. Chem. 254, the insulin-degrading enzymes, these 10 amino acids are all 4698-4703. uncharged except for the 3 active site residues. However, the 13. Swamy, K. H. S. & Goldberg, A. L. (1982) J. Bacteriol. 149, position of the adjacent invariant hydrophobic residue is also 1027-1033. inverted in the protease III active site relative to the active 14. Affholter, J. A., Fried, V. A. & Roth, R. A. (1988) Science 242, site seen in other metalloendopeptidases, with the hydropho- 1415-1418. bic residue being 2 amino acids upstream from the first 15. Kuo, W.-L., Gehm, B. D. & Rosner, M. R. (1990) Mol. En- histidine (Fig. 1). These findings lead to a completely inverted docrinol. 4, 1580-1591. in III and the related insulin- 16. Duckworth, W. C. (1990) in Handbook ofExperimental Phar- signature sequence protease macology: Insulin, eds. Cuatrecasas, P. & Jacobs, S. (Springer, degrading enzymes relative to other known zinc metalloen- Berlin), pp. 143-165. dopeptidases. 17. Duckworth, W. C. (1988) Endocr. Rev. 9, 319-345. The proposed active site, HXXEH, is conserved in a 18. Ding, L., Becker, A. B., Suzuki, A. & Roth, R. A. (1992) J. bacterial enzyme (protease III) and in the human and Dro- Biol. Chem. 267, 2414-2420. sophila insulin-degrading enzymes. Another potential mem- 19. McKerrow, J. H. (1987) J. Biol. Chem. 262, 5943. Downloaded by guest on September 24, 2021 Biochemistry: Becker and Roth Proc. Nadl. Acad. Sci. USA 89 (1992) 3839
20. Goldberg, G. I., Wilhelm, S. M., Kronberger, A., Bauer, E. A., 27. Hartd, F.-U. & Neupert, W. (1990) Science 247, 930-938. Grant, G. A. & Eisen, A. Z. (1986) J. Biol. Chem. 261, 6600-665. 28. Geli, V., Yang, M., Suda, K., Lustig, A. & Schatz, G. (1990) 21. Nakahama, K., Yoshimura, K., Marumoto, R., Kikuchi, M., J. Biol. Chem. 265, 19216-19222. Lee, I. S., Hase, T. & Matsubara, H. (1986) Nucleic Acids Res. 29. Yang, M., Geli, V., Oppliger, W., Suda, K., James, P. & 14, 5843-5855. Schatz, G. (1991) J. Biol. Chem. 266, 6416-6423. 22. Malfroy, B., Schofield, P. R., Kuang, W.-J., Seeburg, P. H., 30. Schneider, H., Arretz, M., Wachter, E. & Neupert, W. (1990) Mason, A. J. & Henzel, W. J. (1987) Biochem. Biophys. Res. J. Biol. Chem. 265, 9881-9887. Commun. 144, 59-66. 31. Semple, J. W., Ellis, J. & Delovitch, T. L. (1989) J. Immunol. 23. Devault, A., Lazure, C., Nault, C., Le Moual, H., Seidah, 142, 4184-4193. N. G., Chretien, M., Kahn, P., Powell, J., Mallet, J., Beau- 32. Kayalar, C. & Wong, W. T. (1989) J. Biol. Chem. 264, 8928- mont, A., Roques, B. P., Crine, P. & Boileau, G. (1987) EMBO 8934. J. 6, 1317-1322. 33. Mundy, D. I. & Strittmatter, W. J. (1985) Cell 40, 645-656. 24. -Rawlings, N. D. & Barrett, A. J. (1991)Biochem. J. 275,389-391. 34. Farach, H. A., Jr., Mundy, D. I., Strittmatter, W. J. & Len- 25. Dumermuth, E., Sterchi, E. E., Jiang, W., Wolz, R. L., Bond, narz, W. J. (1987) J. Biol. Chem. 262, 5483-5487. J. S., Flannery, A. V. & Beynon, R. J. (1991) J. Biol. Chem. 35. Toll, L., Brandt, S. R., Olsen, C. M., Judd, A. K. & Almquist, 266, 21381-21385. R. G. (1991) Biochem. Biophys. Res. Commun. 175, 886-893. 26. Witte, C., Jensen, R. E., Yaffe, M. P. & Schatz, G. (1988) 36. Resnick, N. M., Maloy, W. L., Guy, H. R. & Zasloff, M. EMBO J. 7, 1439-1447. (1991) Cell 66, 541-554. Downloaded by guest on September 24, 2021