THE JOURNAL OF BIOLOGICAL CHEMISTRY Q 1993 by The American Society for Biochemistry and Molecular Biology, Inc.

Vol. 268, No. 8, Issue of March 15, pp. 5488-5495, 1993 Printed in U. S. A.

The Membrane Domain of a Bacteriophage Assembly Protein TRANSMEMBRANE-DIRECTED PROTEOLYSIS OF A MEMBRANE-SPANNING FUSION PROTEIN*

(Received for publication, September 8, 1992)

Judith K. Guy-Caffey and Robert E. WebsterS From the Department of Biochemistry, Duke University Medical Center, Durham, North Carolina 27710

A tripartite fusion construct encoding the amino- terminal half of EcoRI endonuclease followed by amino acids 217-299 of the filamentous bacteriophage gene I protein (PI) attached to the enzymatic portion of alkaline phosphatase results in the production of two proteins. The larger protein, pIe, is the complete tri- partite fusion protein while the smaller protein, PIt*, results from internal initiation of translation at PI methionine 241. Both pIr and PIt* span the Escherichia coli inner membrane via a 20-amino-acid hydrophobic stretch of PI with their amino termini in the cytoplasm and their carboxyl-terminal alkaline phosphatase do- mains in the periplasm. The alkaline phosphatase moiety of approximately 70% of pIt is released into the periplasm by in vivo proteolysis, but only about 10% of PIP* is cleaved. Neither DegP, OmpT, nor protease I11 are responsible for the cleavage in vivo, and leader peptidase is unable to cleave the fusion protein in vitro. Deletion and substitution analyses demonstrate that the degree of periplasmic cleavage depends on the se- quence of the cytoplasmic domain of the fusion pro- teins. Possible mechanisms for this transmembrane- directed cleavage event are compared to proposed models for signal transduction.

The filamentous bacteriophage fl consists of a circular single-stranded DNA molecule encased in a protein sheath composed of five capsid proteins (Model and Russel, 1988). The phage particle is assembled at a site where the inner and outer membranes of the host bacterium appear to be in close contact (Lopez and Webster, 1985). It is at this site that the membrane-associated capsid proteins assemble around the bacteriophage DNA as it is extruded from the bacteria (Rus- sel, 1991). In addition to the capsid proteins, this assembly process requires the products of the phage gene I and IV proteins as well as the bacterial protein, thioredoxin (Russel and Model, 1986). Gene IV encodes a protein which may be part of the outer membrane portion of the assembly site (Brissette and Russel, 1990) while gene I encodes two proteins, PI’ and PI*, which have been localized to the inner membrane (Guy-Caffey et al., 1992). PI is a 348-amino-acid protein which spans the Escherichia coli inner membrane by a single hydro- phobic stretch with the amino-terminal two-thirds of the

*This work was supported by Public Health Service Grant GM18305 from the National Institute of General Medical Sciences. 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.

$ T o whom correspondence should be addressed. Tel.: 919-684- 3005; Fax: 919-684-8752. ’ The abbreviations used are: PI, bacteriophage gene I protein; kb, kilobase(s).

protein in the cytoplasm and the carboxyl-terminal third of the protein in the periplasm. PI’ is the result of internal translational initiation at methionine 241 of PI. It is identical to the carboxyl third of PI and spans the membrane in the same manner.

The membrane signal sequences of gene I proteins have been shown to reside between amino acids 241 and 295 of the 348-residue PI (Horabin and Webster, 1988). This was deter- mined by analyzing a series of tripartite fusion proteins con- taining different regions of PI flanked by the amino-terminal half of EcoRI endonuclease upstream and by the enzymatic portion of alkaline phosphatase downstream. During the in- vestigations presented here, it was observed that following insertion of the fusion proteins into the membrane, the alka- line phosphatase moiety is cleaved in the periplasm to varying extents depending on the nature of the amino acids in the cytoplasm. Such a transmembrane effect on proteolysis is similar to the signaling responses elicited by a class of proteins represented by certain growth factor receptors (Yardin and Ullrich, 1988) and chemotaxis receptors of Gram-negative bacteria (Dah1 et aL, 1989).

This paper describes experiments investigating the trans- membrane-directed proteolysis of the tripartite fusion pro- teins. We show that plasmids encoding tripartite fusions containing the membrane signal sequence of PI direct the synthesis of two proteins, a full-length fusion protein (pIf) and an internal start fusion protein (PIf*) which lacks the EcoRI endonuclease moiety. Both pIf and PIf* span the mem- brane with their amino termini in the cytoplasm and their carboxyl-terminal alkaline phosphatase portions in the peri- plasm. The alkaline phosphatase moiety of pIf is proteolyzed at the periplasmic face of the membrane to release soluble alkaline phosphatase into the periplasm. However, there is no significant release of soluble alkaline phosphatase from the membrane-bound PI:. Deletion and substitution analyses indicate that the cytoplasmic EcoRI endonuclease portion of PIf is essential for its periplasmic cleavage on the opposite side of the membrane. Possible mechanisms for this cleavage are discussed with regard to providing another model system for investigating transmembrane-signaling processes.

EXPERIMENTAL PROCEDURES

Bacterial Strains, Bacteriophage, and Plasmids-E. coli strain K17 has been described previously (Lyons and Zinder, 1972). K17(DE3) is K17 lysogenized with XDEB carrying the inducible gene for T7 RNA polymerase. E. coli strain IT41 has a temperature-sensitive mutation of leader peptidase (Inada et al., 1989) and was provided by K. Ito (Kyoto University). E. coli strains KS272, KS474, SF100, SF103, and SF120 (Baneyx and Georgiou, 1991) were provided by G. Georgiou (University of Texas). KS474, SF100, and SF103 are deriv- atives of KS272 deficient in E. coli proteases DegP, OmpT, and protease 111, respectively. SF120, also derived from KS272, is deficient in all three proteases. XDEB (Studier and Moffatt, 1986) and plasmids pET3 and pLysS (Rosenberg et al., 1987) were obtained from F. W.

5488

Transmembrane-directed Proteolysis 5489

Studier (Brookhaven National Laboratory). Plasmid C3S::TnphoA no. 6 contains sequences coding for PI amino acids 217-299 flanked by the amino-terminal half of EcoRI endonuclease upstream and the enzymatic portion of alkaline phosphatase downstream (Horabin and Webster, 1988). Plasmid pDNC186 contains rbsB, the structural gene for the ribose-binding protein (Groarke et al., 1983; Lopilato et al., 1984) and was obtained from P. Bassford (University of North Carolina, Chapel Hill). Plasmid pJC1 codes for the EcoRI methylase and was obtained from P. Modrich (Duke University Medical Center).

Media and Chemicals-Bacteria were grown in Luria-Bertani me- dium (LB) supplemented with 0.02% (w/v) calcium chloride and with kanamycin (50 pglml), ampicillin (60 pg/ml), chloramphenicol (15 pglml), and/or tetracycline (20 pg/ml) where appropriate. Isopropyl- /3-D-thiogalactopyranoside and the p-toluidine salt of XP (5-bromo- 4-chloro-3-indolyl phosphate) were obtained from Research Organics. Sigma 104 phosphatase substrate was from Sigma. "'I-protein A (100 pCi/wg) was purchased from Du Pont, New England Nuclear and ICN Biochemicals. Egg white lysozyme, trypsin, NADH, and E. coli alkaline phosphatase were from Sigma. Deoxynucleoside triphos- phates were purchased from P-L Biochemicals. Bal-31 (slow form) was from International Biotechnologies, Inc. Restriction endonucle- ases and other DNA modifying enzymes were purchased from Boeh- ringer Mannheim, New England Biolabs, and Bethesda Research Laboratories. The prokaryotic DNA-directed translation ki t was pur- chased from Amersham Corp. Nitrocellulose paper was from Schleicher and Schuell. Purified leader peptidase was a gift from W. Wickner (UCLA). Purified T7 RNA polymerase was kindly provided by P. Modrich (Duke University Medical Center).

Plasmid Construction-The fusion proteins encoded by each of the plasmids described below are shown in Fig. 1. Standard cloning techniques were used (Maniatis et al., 1982). Plasmid pJKG2 was constructed in the following manner. Site-directed mutagenesis by the two-primer method of Zoller and Smith (1984) was used to create a unique EcoRV site by deleting nucleotides 4062-4064 from the gene I sequences of a pEMBL19 derivative of C3S::TnphoA no. 6, which codes for an EcoRI endonuclease-PI-alkaline phosphatase fusion pro- tein (Horabin and Webster, 1988). From this construct (pJKGl), a 5.9-kb BamHI fragment containing the EcoRI-gene I-alkaline phos- phatase sequences and the Kan' gene from TnphoA (Manoil and Beckwith, 1985) was inserted into the BamHI site of pET3 (Rosen- berg et al., 1987) to create pJKG2, which allows inducible synthesis of the EcoRI-gene I-alkaline phosphatase fusion protein from the bacteriophage T7 promoter.

Plasmid pJKG5 was constructed in the following manner. The XbaI site in the MP19 linker of pJKGl was removed by digestion with XbaI, followed by Bal-31 (slow form) digestion. The resulting linear deletions were blunt-ended with T4 DNA polymerase and religated. Site-directed mutagenesis (Zoller and Smith, 1984) of one of these deleted plasmids, pJKG3, was used to insert a thymine between gene I nucleotides 3892 and 3893 and to convert adenine 3894 to guanine. These two mutations generate a unique XbaI site and alter the reading frame such that translation terminates after PI amino acid 232 at a UAG codon. The 5.9-kb BarnHI fragment from this construct, pJKG4, was inserted into the BarnHI site of pET3 (Rosenberg et al., 1987) to produce pJKG5.

Plasmids pJKG8, pJKG21, and pJKG22 were constructed in the following manner. Site-directed mutagenesis (Zoller and Smith, 1984) of plasmid pJKG4 was used to convert methionine 241 in the gene I sequences to leucine (ATG to CTG), producing pJKG6. The 5.9-kb BarnHI fragment of pJKG6 was inserted into the BarnHI site of pET3 (Rosenberg et al., 1987), giving rise to pJKG7. Plasmid pJKG7 was cut with XbaI, followed by Bal-31 (slow form) digestion. The resulting linear deletions were blunt-ended with T4 DNA polymerase and religated. One of these deletions removed only gene I cytosine 3892, restoring the normal gene I reading frame and creating pJKG8. A second deletion removed sequences coding for EcoRI endonuclease amino acids 135-146 and PI amino acids 217-241, creating plasmid pJKG21. The fusion protein encoded by pJKG21 is designated pIr deletion A 6. A third deletion removed sequences coding for EcoRI endonuclease amino acids 93-146 and PI amino acids 217-237, cre- ating plasmid pJKG22. The fusion protein encoded by pJKG22 is designated pIr deletion A 9.

Plasmid pJKGlO was constructed in the following manner, The 5.2-kb Kan' XbaI-BamHI fragment of pJKG6, which contains se- quences coding for PI amino acids 233-299 fused to alkaline phospha- tase, was ligated to an XbaI-cut polymerase chain reaction fragment encoding TolR amino acids 45-142 with a SacI site at the 5' end and an XbaI site at the 3' end. The resulting TolR-PI-alkaline phospha-

tase fragment, in which the TolR and PI junction is out-of-frame, was cut with SacI and inserted between the SacI and BarnHI sites of the multilinker in plasmid pETBclF, which contains a T7 transla- tional start site (Levengood et al., 1991), creating plasmid pMMM1. The 5.4-kb NdeI-BarnHI fragment of pMMM1, containing the T7 start fused in-frame to the TolR-PI-alkaline phosphatase sequences, was inserted between the NdeI and BamHI sites of pET3cX, a derivative of pET3c (Rosenberg et al., 1987) from which the XbaI site had been removed by XbaI digestion and subsequent filling-in by Klenow. The resulting plasmid, pJKG9, was digested at the XbaI junction of TolR and PI, treated with mung bean nuclease, and religated to bring the TolR and PI-alkaline phosphatase sequences in-frame, creating plasmid pJKG10.

Analysis of Fusion Proteins and Alkaline Phosphatase Actiuity- K17(DE3) or KS272 bacteria containing the plasmid encoding the appropriate fusion protein were grown to 2-3 X 10' cells/ml at 37 "c . Bacteria in a 1-ml sample were collected by centrifugation and resuspended in 0.1 ml of sample buffer (Levengood and Webster, 1989). After boiling for 5 min, samples were subjected to polyacryl- amide gel electrophoresis (Laemmli, 1970) and transferred to nitro- cellulose for Western blot analysis using alkaline phosphatase anti- body and 1251-protein A (Towbin et al., 1979; Yen and Webster, 1981). If the amount of fusion protein/total bacterial protein was to be determined, the bacteria in a 1-ml sample were resuspended in 0.1 ml of 1% (w/v) SDS, and the protein concentration was analyzed using a bicinchoninic acid assay (Pierce Chemical Co.). Twenty-pl samples, adjusted to the same relative protein concentration in sample buffer, were loaded for polyacrylamide gel electrophoresis. Known amounts of E. coli alkaline phosphatase were loaded as standards on the same polyacrylamide gel. After transfer to nitrocellulose and Western blot analysis using alkaline phosphatase antibody and '251-protein A, the nitrocellulose corresponding to the autoradiogram bands for pIf, PI; and alkaline phosphatase was excised and counted in an Intertech- nique SL30 scintillation counter for 10 min. A plot of the known amounts of alkaline phosphatase versus the amount of radioactivity was found to he linear and used to determine the amount of fusion protein in the bacterial samples analyzed on the same Western blot.

Alkaline phosphatase activity was measured essentially as de- scribed by Brickman and Beckwith (1975). Bacteria were harvested by centrifugation, resuspended in the same volume of 1 M Tris-C1 (pH 8.0), and p-nitrophenyl-phosphate (Sigma 104 phosphatase sub- strate) was added to a concentration of 0.04%. The reaction mixture was incubated at 37 "C and 1-ml samples were taken at 5, 10, 15, 30, and 60 min. At each time point, the reaction was terminated by the addition of K2HP04 to 0.1 M, and the absorbance at 420 nm was determined. The amount of alkaline phosphatase fusion protein in the bacteria contained in the 1-ml sample was determined by quan- titative Western blot analysis as described above. The specific activity for each reaction was defined as the change in absorbance at 420 nm/ 10 min/pg of alkaline phosphatase fusion protein.

Cellular and Membrane Fractionation-For cellular fractionation, a 25-ml culture of K17(DE3) containing pLysS and pJKG8 was grown at 37 "C to 2 X 10' cells/ml. After 10 min, the cells were chilled on ice, and a 1-ml whole cell sample was centrifuged. The cell pellet was resuspended in 100 p l of sample buffer (Levengood and Webster, 1989). The remainder of the cells were centrifuged at 3,000 X g for 5 min and the cell pellet was resuspended in 2.5 ml of cold 20% (w/v) sucrose, 50 mM HEPES (pH 7.8), 0.5 mM EDTA. Lysozyme was added to a concentration of 0.04 mg/ml. An equal volume (2.5 ml) of 50 mM HEPES (pH 7.2), 0.5 mM EDTA was added over a period of approximately 3 min. Spheroplast formation was monitored on a Zeiss microscope. When spheroplasting approached 90%, cells were centrifuged at 12,000 X g for 10 min. The periplasmic supernatant was precipitated with trichloroacetic acid and resuspended in 500 pl of sample buffer. The spheroplasts were washed with 2.5 ml of 50 mM Hepes (pH 7.2), 10% (w/v) sucrose, 50 mM EDTA, and the pellet was resuspended in 500 pl of sample buffer. Periplasmic samples were diluted 1:5, while spheroplast samples were diluted 1:2 with sample buffer before loading 20 pl of whole cell, periplasmic, and spheroplast samples for polyacrylamide gel electrophoresis and Western blot analysis with alkaline phosphatase antibody and '251-protein A.

To isolate the inner and outer membrane fractions, 250-ml cultures were grown to 2 X 10' cells/ml at 37 "C. The bacteria were harvested by centrifugation and resuspended in 20 ml of 0.01 M Tris-C1 (pH 7.8), 20% sucrose (w/w). Spheroplast formation and lysis were per- formed as described by Guy-Caffey et al. (1992), and the lysed spheroplasts were layered over a 2-ml cushion of 55% sucrose (w/w) in 0.01 M Tris-C1 (pH 7.8), 0.5 mM EDTA. After centrifugation in an

5490 Transmembrane-directed Proteolysis

SW 41 Beckman rotor a t 35,000 for 3 h, the membrane band at the 55% sucrose interface was collected, and the inner and outer mem- branes were separated on a sucrose gradient as described by Guy- Caffey et al. (1992). The protein concentration and NADH oxidase activity was determined for each fraction. To determine the location of the fusion protein, a portion of each sample was subjected to SDS- polyacrylamide gel electrophoresis and Western blot analysis using alkaline phosphatase antibody and ‘*‘I-protein A as described above.

To analyze cells for regions of membrane-associated PI fusion proteins which were exposed to the periplasm, the cells were shocked with low concentrations of EDTA and treated with trypsin as in Levengood et al. (1991) except that trypsin incubation was for 30 min at 0 “C. The resulting proteins were then analyzed by SDS-polyacryl- amide gel electrophoresis and Western blot analysis using EcoRI endonuclease, TolR, or alkaline phosphatase antibody and 12’I-protein A.

In Vitro Transcription/Translation Reactions-Zn uitro transcrip- tion and translation reactions were done with the Amersham prokar- yotic DNA-directed translation kit according to the manufacturer’s instructions using 2.5 pg of plasmid DNA and 30 pCi of [35S]methio- nine. One pl of purified T7 RNA polymerase (P. Modrich, Duke University) was added to reactions with plasmids containing the T7 promoter. After completion of the transcription and translation re- actions, 2 p1 of purified leader peptidase was added and incubations continued for 1 h at 37 ‘C in the presence of 1% Triton X-100.

RESULTS

Products Encoded by the EcoRI Endonuclease-PI-Alkaline Phosphatase Fusion Construct-Plasmid pJKG2 contains a tripartite fusion construct encoding a protein consisting of the amino-terminal half of EcoRI endonuclease, PI amino acids 217-299, and the enzymatic portion of alkaline phos- phatase (Fig. 1). Previous experiments had shown that such a fusion protein would be inserted into the E. coli inner membrane with its amino terminus in the cytoplasm and its carboxyl terminus in the periplasm (Horabin and Webster, 1988). The region spanning the membrane consists of the hydrophobic stretch of PI amino acids between residues 254- 273. Cells carrying pJKG2 produce three alkaline phospha-

pJKG2

pJKG5

pJKG8

pJKG21

pJKG22

pJKGlO

EcoRl Of

tase-containing proteins (Fig. 2, lane 1 ). The largest is a 73- kDa protein which is the size expected for the full-length EcoRI endonuclease-PI-alkaline phosphatase fusion protein (plf). In addition, two other major proteins of approximately 55 and 47 kDa are produced which react with alkaline phos- phatase antibody. The 55-kDa protein is the appropriate size for a protein resulting from an internal translational initiation event at methionine 241 of pIf. The 47-kDa peptide is the size of mature alkaline phosphatase and is presumably the result of proteolysis of PIf or the 55-kDa protein.

To further investigate the origins of the 55- and 47-kDa proteins, a fusion construct in which PI methionine 241 is converted to leucine was made by site-directed mutagenesis of plasmid pJKG2. Cells carrying this construct (pJKG8) produce both pIf and the 47-kDa alkaline phosphatase moiety (Fig. 2, lane 2), but not the 55-kDa protein. This suggests that the 55-kDa protein made in cells containing pJKG2 is most likely the result of initiation of translation at methionine 241 of pIf and not from degradation of pIf. To confirm this, a plasmid was constructed which contains a terminating codon upstream of the methionine 241 codon (Fig. 1). Bacteria containing this construct (pJKG5) produce an anti-alkaline phosphatase-reactive molecule of 55 kDa (Fig. 2, lune 3 ) , consistent with the hypothesis that the initiating event does occur at methionine 241. The protein produced by this initi- ating event is designated PI;.

The 47-kDa protein is the size of mature alkaline phospha- tase and presumably results from a cleavage event at the periplasmic surface of the inner membrane after insertion of the fusion protein into the membrane. If this is the case, it should be located in the soluble periplasmic material. As shown in Fig. 2B, analysis of the periplasmic material from bacteria containing pJKG8 shows that it contains all of the 47-kDa protein. Therefore, it must result from proteolytic cleavage of the membrane-bound pIf.

BaniHI

- leu

4 5 TO1 R 1 4 2 2 4 1

2 4 1 leu

FIG. 1. Schematic diagram of PI fusion constructs. At the top of the diagram is a schematic drawing of the DNA fragment encoding the PI fusion proteins. The tripartite fusion constructs consist of sequences coding for regions of EcoRI endonuclease (EcoRZ) or TolR (TolR) followed by a portion of the gene I protein (Gene I) fused to the enzymatic portion of alkaline phosphatase (PhoA). Kan’, kanamycin resistance. The amino-terminal portion of the fusion proteins encoded by plasmids pJKG2, pJKG5, pJKG8, pJKG21, pJKG22, and pJKGlO are shown expanded below the construct. The solid black box represents the 20-amino-acid hydrophobic stretch normally located between amino acids 254 and 273 in the 348-amino-acid PI, while the plus symbok represent the five charges in a potential amphiphilic helix located between PI methionine 241 (Met-241) and the hydrophobic stretch (Horabin and Webster, 1988). In plasmid pJKG5, sequences coding for a unique XbaI site were introduced upstream of methionine 241 such that translation terminates after PI amino acid 232 (X). In plasmid pJKG8, sequences coding for methionine 241 were mutated to leucine (Leu-241). For further descriptions, see “Experimental Procedures.”

Transmembrane-directed Proteolysis 5491

Since PIr* contains the same signal sequence as pIf, and therefore is inserted into the inner membrane in the same manner as PIr (Horabin and Webster, 1988; see below), then it should be subject to the same periplasmic cleavage event as PIr. However, synthesis of PIr* alone by pJKG5-containing bacteria results in the production of very little free alkaline phosphatase (Fig. 2, lane 3 ) . Quantitation reveals that while approximately 70% of the alkaline phosphatase moiety of pIr produced in cells carrying pJKG8 is cleaved to a soluble periplasmic form, only about 10% of PI: undergoes this cleavage event in cells carrying pJKG5 (Table I ) .

The Amino-terminal Portion of PI, Is Essential for Its Per- iplasmic Proteolysis-The sequence of PIr in the cleavage region is identical to that of PI:. The only difference between pIr and PIr* is the amino-terminal portion of PIC, which con- sists of EcoRI endonuclease amino acids 1-146 followed by PI amino acids 217-240. This suggests that something in this domain is influencing the periplasmic cleavage of pIr, even though i t is present on the cytoplasmic side of the inner membrane.

To further define the region of the amino-terminal domain of PIr which is required for its cleavage, the proteolysis of fusion proteins produced from three additional constructs was examined. Plasmid pJKG21 encodes a tripartite fusion pro- tein from which EcoRI endonuclease amino acids 135-146 and PI amino acids 217-241 have been removed (Fig. 1). The resulting fusion protein is cleaved with almost the same efficiency as PIr (Fig. 2, lane 4; Table I). Plasmid pJKG22

A. 6. 1 2 3 4 5 6

68 1

43 2

FIG. 2. Analysis of PI fusion proteins with alkaline phos- phatase antibody. A, extracts of cells containing either plasmid pJKG2 (lane I), pJKG8 (lane 2) , pJKG5 (lane 3 ) , pJKG21 (lane 4 ) , pJKG22 (lane 5), or pJKGlO (lane 6 ) were separated by SDS- polyacrylamide gel electrophoresis and subjected to Western blot analysis using alkaline phosphatase antibody and ’“I-protein A. i?, cells containing plasmid pJKG8 were fractionated as described under “Experimental Procedures” and the whole cell (c), spheroplast (s), and periplasmic ( p ) fractions were separated by SDS-polyacrylamide gel electrophoresis and subjected to Western blot analysis using alkaline phosphatase antibody and “‘I-protein A. Molecular weight standards (in thousands) are designated to the left.

encodes a fusion protein from which EcoRI endonuclease amino acids 92-146 and PI amino acids 217-237 have been removed. About 52% of this fusion protein is proteolyzed (Fig. 2, lane 5; Table I). Plasmid pJKGlO encodes a fusion protein in which EcoRI endonuclease amino acids 1-146 and PI amino acids 217-232 have been replaced with amino acids 45-142 of the TolR protein (Fig. l ) , a region which has a net negative charge (Sun and Webster, 1987) compared with the net posi- tive charge found in the amino-terminal portion of EcoRI endonuclease (Newman et al., 1981). Only about 20% of this fusion protein, TolR-PIr, is proteolyzed (Fig. 2, lane 6; Table I). Taken together, these results indicate that the region essential for efficient cleavage of PIr is within EcoRI endo- nuclease amino acids 1-91.

It is possible that the DNA binding activity of the EcoRI endonuclease portion of PIr sequesters PIr in such a way that it becomes available for cleavage. To test this, a plasmid encoding the EcoRI methylase (pJC1) was introduced into pJKG8-containing bacteria so that methylation renders the EcoRI restriction sites resistant to recognition by EcoRI en- donuclease. Fusion protein PIr is cleaved to the same extent in pJCl-containing bacteria (Table I) , arguing that the cleav- age of pIr is not related to the binding of the fusion protein to EcoRI restriction sites in the cellular DNA.

Topology of pl Fusion Proteins in the Membrane-One possible explanation for the difference in cleavage of pIr and PIr* is that they are not inserted into the membrane in the same manner. Thus, the topology of PIr, PI:, and TolR-pIf in the membrane was examined by trypsin accessibility experi- ments as described under “Experimental Procedures.” Cells carrying either plasmid pJKG8, pJKG5, or pJKGlO were treated with EDTA to permeabilize the outer membrane. Such treatment causes the release of the soluble periplasmic pro- teins, including any free alkaline phosphatase resulting from cleavage of a fusion protein. The cells were harvested by centrifugation and treated with trypsin such that the portions of membrane-associated proteins exposed to the periplasm would be digested. The resulting cell extracts were electro- phoresed on a SDS-polyacrylamide gel and Western-blotted using EcoRI, TolR, or alkaline phosphatase antibodies (Fig. 3). Probing with the alkaline phosphatase antibody (Fig. 3A), both PIr, PI:, and the majority of TolR-pIr disappear upon trypsin incubation, indicating that their carboxyl termini are digested and thus are exposed to the periplasm. Using the EcoRI antibody (Fig. 3B, left panel), a 24-kDa fragment is detected following trypsin digestion of cells carrying pJKG8, while a 17-kDa fragment is detected with TolR antibodies in cells carrying pJKGlO (Fig. 3R, right panel). Assuming the 20-amino-acid hydrophobic stretch of PI spans the inner

TABLE I Quantitation of in vivo proteolysis and periplasmic alkaline phosphatase (AP) activity of p l fusion proteins

The amount of in vivo proteolysis of the alkaline phosphatase moeity of the pI fusion proteins indicated above was quantitated in uninduced K17 (DM) cells carrying the appropriate plasmids. Uninduced K17 (DE3) cells produce detectable levels of fusion protein, presumably due to the leaky expression of T 7 RNA polymerase from the UV5 promoter of the DE3 lysogen in the chromosome. The percentage of AP cleavage was calculated by comparing the amount of radioactivity in the cleaved alkaline phosphatase band with the combined amount of radioactivity in the intact fusion protein and cleaved alkaline phosphatase bands from a Western blot. The numbers in parentheses refer to the number of experiments included in the calculations. The amount of periplasmic AP activity/total anti-alkaline phosphatase-reactive protein and per cleaved AP was determined as described under “Experimental Procedures.” The numbers given represent the average of two experiments.

Plasrnid(s) Fusion protein % AP cleavaEe AP activity/total AP Units AP/cleaved AP

pdKG8 PIr 69 f 8 (16) pJKG8, PJCl PIr 72 f 7 (5) pJKG5 PIf* 13 * 7 (15) pJKG21 pIf deletion A6 68 * 9 (5) pJKG22 pIr deletion A9 D J K G ~ O

52 * 6 (5) ToIR-DL 21 f 6 (8)

0.87 1.10 1.54 0.85 0.75 0.58

1.20 1.42 7.83 1.69 1.44 1.32

5492 Transmembrane-directed Proteolysis

A. p JKG8 PJKGB pJKGlO 0 - + o - + o - +

97-

68-

43

29.

188

B.

97-

68-

43 -

29-

18-

anti- AP

pJKG8 pJKGlO 0 - + 0 - +

anti-EcoRl anti-TolR FIG. 3. Topology of pIf, PIf*, and TolR-pIr in the membrane.

A, K17(DE3) cells containing plasmid pLysS and either plasmid pJKG8, pJKG5, or pJKGlO were permeabilized with EDTA and treated with trypsin as described under “Experimental Procedures.” After separating the resulting proteins by SDS-polyacrylamide gel electrophoresis, the proteins were subjected to Western blot analysis using alkaline phosphatase antibody and ’*‘I-protein A. Cells were solubilized immediately after EDTA treatment (0) or after a 30 min incubation at 0 “C in the absence (-) or presence (+) of trypsin. Molecular weight standards (in thousands) are designated to the left. The expected positions for pIf, pIf*, and TolR-pIf are shown to the right. R, K17(DE3) cells containing either plasmid pJKG8 ( le f tpanel) or pJKGlO (right panel) were treated with EDTA and trypsin. After separating the resulting proteins by SDS-polyacrylamide gel electro- phoresis, Western blot analysis was performed using ”‘I-protein A and either EcoRI endonuclease antibody (left panel) or TolR antibody (right panel). ( O ) , (-), and (+) are the same as in A.

membrane, these are the sizes of the peptides that would be expected to remain after trypsin proteolysis if the amino- terminal portions of PIr and TolR-pIr are protected in the cytoplasm.

Membrane Localization of pI Fusion Proteins-Another rea- son for the difference in proteolysis of pIr and pIr* might be that they are localized to different regions of the bacterial membrane. The trypsin proteolysis results indicate that both proteins span the inner membrane in the same orientation. It was speculated that perhaps PIr is restricted to membrane regions resembling adhesion zones, where the inner and outer membranes appear to be fused, while pIf* is free to diffuse in the inner membrane. Thus, the inner and outer membranes of cells carrying either plasmid pJKG8 or pJKG5 were sepa- rated as described under “Experimental Procedures.” In ad- dition to the separation of the inner and outer membrane fractions, this procedure results in the isolation of a fraction

of intermediate density which appears to contain native zones of association between the two membranes (Ishidate et al., 1986). The protein profiles of the fractions from the sucrose gradients on which the membranes were resolved are shown in Fig. 4. Also shown is the location of NADH oxidase activity used as an inner membrane marker. Western blot analysis of each fraction showed that both PIr (toppanel) and PI: (bottom panel) fractionate with the inner membrane and no significant amount of either is found in the position expected for the intermediate fraction.

Periplasmic Alkaline Phosphatase Activity in Fusion Pro- tein-containing Bacteria-E. coli alkaline phosphatase is ac- tive only in the periplasm. The soluble enzyme has been shown to require a dimeric structure for activity (McCracken and Meighen, 1980). Since the PI fusion proteins contain a periplasmic alkaline phosphatase domain, it is possible that dimerization of the membrane-bound PIr* is responsible for its relative resistance to the cleavage experienced by pIr. If this is the case, membrane-bound PIr* should exhibit catalytic activity. Periplasmic alkaline phosphatase activity was meas-

P’f 500 1 .o

- 0.9

- 0.8 ”

0.7 t> 0.6 0 0

0.5

0.2

0.1

0.0 0 5 1 0 1 5 2 0 2 5 3 0 3 5

Fraction #

t 400 - - E &300 J v

0

0 0

c 200

.E 100

a 0 2 c aJ

4 I- 0.9

0 5 1 0 1 5 2 0 2 5 3 0 3 5

Fraction #

ing pIf or pIf*. A, K17(DE3) cells containing plasmid pJKG8 and FIG. 4. Inner/outer membrane separation of cells express-

were treated as described under “Experimental Procedures” to sepa- rate their inner and outer membranes. Fractions collected from the final gradient were analyzed for protein concentration and NADH oxidase activity and then subjected to SDS-polyacrylamide gel elec- trophoresis and Western blot analysis with alkaline phosphatase antibody and ”‘I-protein A. The lanes of the blot are aligned with their respective fractions on the graph. B, same as in A, but with K17(DE3) cells containing plasmid pJKG5.

Transmembrane-directed Proteolysis 5493

ured in intact bacteria containing the various fusion proteins under conditions in which cellular alkaline phosphatase is repressed as described under “Experimental Procedures.” The amount of anti-alkaline phosphatase reactive material present as intact fusion protein or as the cleaved soluble form was determined by quantitative Western blot analysis. This al- lowed the calculation of the specific alkaline phosphatase activity per microgram of all alkaline phosphatase-containing protein species present in the bacteria (Table I, column 4). All strains tested showed similar specific activities, despite different degrees of cleavage of the various fusion proteins. This suggests that membrane-bound PI;“ must be active, presumably because it can form a dimer or other active oligomeric structure. This conclusion is substantiated by the calculation of specific alkaline phosphatase activity per mi- crogram of cleaved soluble alkaline phosphatase/cell (Table I, column 5). If only the soluble periplasmic alkaline phos- phatase were active, then the specific activity should be the same for all the bacteria tested. The significantly larger activ- ity found for bacteria containing PIf* is consistent with the conclusion that membrane-bound PIf* must be active.

Leader Peptidase, DegP, OmpT, and Protease I I I Do Not Cleave PI,-Because the cleavage of pIr occurs at the peri- plasmic face of the inner membrane, the protease responsible could either be localized to the periplasm or to the inner or outer membrane with its active site facing the periplasm. Of the known E. coli proteases located in one or the other of these compartments, mutations abolishing activity have been isolated in leader peptidase (Inada et al., 1989), DegP (Strauch et al., 1989), OmpT (Baneyx and Georgiou, 1990), and pro- tease I11 (Baneyx and Georgiou, 1991). Each of these proteases was tested for their ability to cleave pIf, pIf*, and TolR-pIf.

Leader peptidase, encoded by the lep gene, is an inner membrane protein with its active site a t the periplasmic face of the membrane (Zimmermann et al., 1982). We tried to investigate the possible degradation of PIr by leader peptidase using a temperature-sensitive lep mutant, IT41 (Inada et al., 1989). However, the lep mutant cells which were induced to express PIf a t the non-permissive temperature failed to incor- porate label, presumably due to the effect of the lep mutation a t 42 “C combined with the expression of PIr, which inhibits cell growth. Thus, an alternate approach was employed. Plas- mids pJKG8, pJKG5, and pJKGlO were used to direct i n vitro transcription and translation reactions to produce pIf, pIf*, and TolR-PIf, respectively. Plasmid pDNC186, which encodes the known leader peptidase substrate ribose-binding protein, was used as a control. Purified leader peptidase was added exogenously to determine whether it could cleave any of the fusion proteins in vitro (Fig. 5). Quantitation shows that 90- 100% of intact PIf, PI:, and TolR-pIf produced i n vitro remain after incubation with leader peptidase, while only about 20% of the ribose-binding protein remains after incubation with leader peptidase under identical conditions. Thus, leader pep- tidase does not appear to be able to cleave pIf, PI:, or TolR- PIr in vitro, although this does not rule out the possibility that it does so in vivo.

DegP, OmpT, and protease I11 have all been shown to be involved in the degradation of fusion proteins secreted into the periplasmic space (Strauch and Beckwith, 1988; Baneyx and Georgiou, 1990, 1991). Thus, strains deficient in each of these proteases, as well as a triple mutant, were tested for their ability to cleave pIf, PIf*, and TolR-PIf. The amount of proteolysis of the three fusion proteins seen in the mutant strains was not significantly different than that seen with the parent strain (Table 11). Thus, it appears that neither DegP, OmpT, nor protease I11 is involved in the cleavage of PIf.

1 2 3 4 5 6 7 8 97*

68 - -a-

43*

29,

18*

dase in vitro. Plasmids pJKG8, pJKG5, pJKG10, and pDNC186 FIG. 5. Incubation of PI fusion proteins with leader pepti-

were used to direct in vitro transcription and translation reactions to produce [R5SS]methionine-labeled proteins as described under “Exper- imental Procedures.” The reaction mixtures were divided in half and incubated with or without purified leader peptidase ( L P ) in the presence of 1% Triton X-100. After 1 h of incubat.ion, an equal volume of 2 X sample buffer (Levengood and Webster, 1989) was added, and the samples were subjected to SDS-polyacrylamide elec- trophoresis. The fixed and dried gel was exposed to film. Lanes I and 2, plasmid pJKG8-directed reactions incubated without and with LP, respectively; lanes 3 and 4, plasmid pJKG5-directed reactions incu- bated without and with LP, respectively; lanes 5 and 6, plasmid pJKG10-directed reactions incubated without and with LP, respec- tively; lanes 7 and 8, plasmid pDNC186-directed reactions incubated without and with LP, respectively. The lower case letters designate bands which are the appropriate size for the following proteins: (a ) pIf, ( b ) PI;, (c ) TolR-pIf, ( d ) pre-ribose-binding protein, ( e ) ribose- binding protein. Molecular weight standards (in thousands) are des- ignated to the left.

DISCUSSION

A plasmid was constructed that encodes a tripartite fusion protein consisting of the amino-terminal half of EcoRI endo- nuclease followed by amino acids 217-299 of the PI attached to the enzymatic portion of alkaline phosphatase. This fusion construct results in the production of two proteins, the com- plete tripartite protein PIr and a smaller protein PI:, which only contains sequences from PI and alkaline phosphatase. The smaller protein is the result of internal initiation at the codon for PI methionine 241. Both PIr and PIr* are inserted into the bacterial inner membrane with their carboxyl-ter- minal alkaline phosphatase portion in the periplasm. This is consistent with the hydrophobic stretch of PI amino acids 254-273 spanning the membrane.

Once pIf is inserted into the membrane, a cleavage event occurs resulting in the release of approximately 70% of the alkaline phosphatase domain as a soluble periplasmic form. Presumably, this cleavage event occurs a t the periplasmic face of the inner membrane. The protease involved in this cleavage event is unknown. The most likely candidate is E. coli leader peptidase, an inner membrane protein whose active site faces the periplasm (Zimmerman et al., 1982). However, purified leader peptidase was shown not to cleave PIr synthesized by i n vitro transcription and translation reactions under condi- tions in which a known substrate, the ribose-binding protein precursor, was proteolyzed. This does not rule out the possi- bility that leader peptidase cleaves PIr in vivo when it is located in the bacterial inner membrane. E. coli proteases DegP, OmpT, and protease I11 have all been shown to be involved in the degradation of fusion proteins secreted into the peri- plasmic space (Strauch and Beckwith, 1988; Baneyx and

5494 Transmembrane-directed Proteolysis

TABLE I1 Quantitation of in vivo proteolysis of pl fusion proteins in protease-mutant strains

The amount of in vivo proteolysis of the PI fusion proteins indicated above was quantitated in the protease-mutant strains listed above as well as the wild-type parent strain KS272. Bacteria containing pJKG8, pJKG5, and pJKGlO were used to measure cleavage of pIr PIf* and TolR-pIf, respectively. These cells produce detectable levels of fusion protein in the absence of a source of T7 RNA polymerase, presumably due to the ability of E. coli RNA polymerase to recognize the T7 promoter present in the plasmids. The percentage of proteolysis was calculated by comparing the amount of radioactivity in the alkaline phosphatase band with the combined amount of radioactivity in the intact fusion protein and alkaline phosphatase bands from a Western blot. The numbers in parentheses refer to the number of experiments included in the calculations.

% AP cleavage Strain Genotype

PIr PIf* TdR-pIr KS272 F- lacX74 galE galK thi rpsL(strA) PhoA(PvuII) 70 k 2 (2) 9 19 KS474 KS272 degP41 (PstI-Kan') 70 f 2 (2) 4 20 SF100 KS272 ompT 72 f 2 (2) 7 17 SF103 KS272 ptr-32::Cm' 71 k 2 (2) 5 15 SF120 KS272 ptr-32::Cm' degP41 (PstI-Kan') ompT 70 f 2 (2) 5 8

Georgiou, 1990, 1991). However, strains mutant in these pro- teases were still able to process PIf to the same extent as the wild-type parent strain. Thus, it appears that none of these proteases are involved in processing PIf. The cleavage of pIr may prove useful in identifying E. coli mutants which are defective in proteolysis by screening for mutations which stabilize PIf. This general approach was used successfully in the isolation of degP (Strauch and Beckwith, 1988).

The cytoplasmic EcoRI endonuclease portion of pIf appears to govern the ability of the protease to cleave and release the alkaline phosphatase moiety into the periplasm efficiently. While PI: only differs from PIf in that it lacks EcoRI endo- nuclease amino acids 1-146 and PI residues 217-240, less than 10% of the alkaline phosphatase is released from PIf* as compared to 70% cleavage for PIf. Deletion experiments dem- onstrate that only EcoRI endonuclease amino acids 1-92 need be present in the cytoplasm for efficient cleavage to occur in the periplasm.

If the EcoRI endonuclease sequences of PIf are replaced by a 98-amino acid portion of another protein, TolR, the alkaline phosphatase portion of PIf is cleaved to the soluble periplasmic form less than 20% of the time. This portion of TolR has a net negative charge (Sun and Webster, 1987) as compared to the net positive charge found in the amino-terminal portion of EcoRI endonuclease (Newman et al., 1981). These results suggest that it is not only the presence of the cytoplasmic portion of the protein that governs the cleavage event, but also the structure that it forms and/or interactions it may have with other bacterial proteins. Whatever these cyto- plasmic interactions may be, they must influence the structure of the periplasmic domain to affect the exposure or conceal- ment of the protease cleavage site(s).

The ability of the cytoplasmic portion of the fusion protein to affect the conformation of the periplasmic domain is anal- ogous to examples of signal transduction where a stimulus on one side of a membrane is able to produce a response on the other side of the membrane. A number of transmembrane proteins have been identified which are able to perform this signaling function. One class of these signaling molecules consists of an extracellular and a cytoplasmic domain con- nected by one or two transmembrane hydrophobic sequences. Examples of this class of proteins are certain growth factor receptors (Yardin and Ullrich, 1988) and chemotaxis receptors of Gram-negative bacteria (Dah1 et al., 1989). Recent studies have elucidated much information about the structure of these receptors, especially in the case of the chemotaxis receptor for aspartate (Milburn et al., 1991; Pakula and Simon, 1992a). However, the nature of any conformational change in these molecules which facilitates the transmission of a signal across

the membrane is still unclear (Pakula and Simon, 1992b). Based on the structure and function of this class of receptors, two general types of transduction mechanisms have been proposed. One mechanism involves the stimulus-induced oli- gomerization of two or more receptor proteins as i s proposed to occur with the growth factor receptors (Schlessinger, 1988). The other mechanism invokes the propagation of a stimulus- induced signal through the transmembrane domain of an individual receptor molecule in such a way as to influence the conformation of the portion of the signaling molecule on the opposite side of the membrane. This latter type of mechanism has been suggested for the bacterial chemotaxis receptors (Milligan and Koshland, 1988, 1991).

A similar mechanism could be involved in the transmem- brane-directed proteolysis observed for the PI fusion proteins. The most reasonable hypothesis for how the cytoplasmic domain of pIf affects its periplasmic cleavage is that this domain regulates the dimerization of the alkaline phosphatase moeity in the periplasm. Since soluble alkaline phosphatase is active as a dimer (McCracken and Meighen, 1980), it is expected that dimerization of membrane-bound alkaline phosphatase fusion proteins should be able to occur under the appropriate circumstances. This is most likely what occurs with pIf* since it exhibits significant alkaline phosphatase activity in its membrane-bound form. The paucity of cleavage of pIf* suggests that the sites of recognition for the protease are not accessible when the alkaline phosphatase moiety is in the dimeric state. Presumably, molecules containing the EcoRI endonuclease moiety in the cytoplasm, such as pIf, have more difficulty forming dimers, thus allowing the protease time to recognize the cleavage site. The difference in the overall rates of dimerization uersus cleavage of the fusion proteins would then determine how much of the alkaline phosphatase would be solubilized. In the case of PIf, this would be approximately 70%. Changing the cytoplasmic domain from EcoRI endonuclease to TolR sequences would alter the kinetics of dimerization in such a way as to give the observed 20% cleavage. This type of dimerization mechanism would be similar to that proposed for signal transduction by certain growth factor receptors (Schlessinger, 1988).

Other types of interactions within the cytoplasm could also lead to the cleavage of the alkaline phosphatase moiety in the periplasm. The interaction of the EcoRI endonuclease domain of PIf with cellular DNA does not seem likely since methyla- tion of the EcoRI restriction sites does not affect the cleavage of pIf. However, there could be a specific interaction between the EcoRI endonuclease domain of PIf and some other protein, perhaps the protease itself. Since leader peptidase has a cytoplasmic domain important for its periplasmic enzymatic

Transmembrane-directed Proteolysis 5495

activity (Ohno-Iwashita and Wickner, 1983), it is possible than an interaction between the EcoRI endonuclease portion of pIr and the cytoplasmic domain of leader peptidase might result in the cleavage of PIC by leader peptidase at the peri- plasmic face of the inner membrane. Such an interaction might not be sterically possible in the in vitro transcription/ translation reaction mixture, explaining why no cleavage of pIf was detected by leader peptidase under these conditions.

It also is possible that the cytoplasmic sequence of amino acids may directly transmit via the transmembrane portion a subtle conformational change to the periplasmic domain which is recognized by a protease. Further study of this system will be aided by knowledge of the protease(s) involved in the cleavage event. These studies could potentially give a clearer picture of the sites recognized by the protease as well as possible add to our knowledge of how a signal might be transmitted across a membrane via conformational changes in a membrane protein.

Acknowledgments-We thank Mary Jo Outlaw for help in process- ing the manuscript and Gerda Vergara for assistance with the figures.

REFERENCES

Baneyx, F., and Georgiou, G. (1991) J. Bacterid 1 7 3 , 2696-2703 Baneyx, F., and Georgiou, G. (1990) J. Bacteriol. 172 , 491-494

Brissette, J. L., and Russel, M. (1990) J. Mol. Biol. 2 1 1 , 565-580 Brickman, E., and Beckwith, J. (1975) J. Mol. Biol. 96,307-316

Dahl, M. K., Boos, W., and Manson, M. D. (1989) J. Bacteriol. 171,2361-2371 Groarke, J. M., Mahoney, W. C., Hope, J. N., Furlong, C. E., Robb F. T

Zalkin, H., and Hermodson, M. A. (1983) J. Biol. Chem. 2 5 8 , 12952112956" Guy-Caffey, J. K., Rapoza, M. P., Jolley, K. A,, and Webster, R. E. (1992) J.

Bacteriol. 174 , 2460-2465

Inada, T., Court, D. L., Ito, K., and Nakamura, Y. (1989) J. Bacteriol. 171, Horabin, J. I., and Webster, R. E. (1988) J. Biol. Chem. 2 6 3 , 11575-11583

585-587

Ishidate, K., Creeger, E. S., Zrike, J., Deb, S., Glauner, B., MacAlister, T. J.,

Laemmli, U. K. (1970) Nature 227,680-685 Levengood, S. K., and Webster, R. E. (1989) J. Bacteriol. 171,6600-6609 Levengood, S. K., Beyer, W. F., Jr., and Webster, R. E. (1991) Proc. Natl. Acad.

Lopez, J., and Webster, R. E. (1985) J. Bacteriol. 163 , 1270-1274 Lopilato, J. E., Garwin, J. L., Emr, S. D., Silhavy, T. J., and Beckwith. J. R.

Lyons, L. B., and Zinder, N. D. (1972) Virology 49,45-60 Maniatis, T., Fritsch, E. F., and Sambrook, J. (1982) Molecular Cloning: a

Laboratory Manual, Cold Spring Harbor Laboratory, Cold Spring Harbor,

Manoil, C., and Beckwith, J. (1985) Proc. Natl. Acad. Sci U. S . A. 8 2 , 8129- NY

McCracken, S., and Meighen, E. (1980) J. Biol. Chem. 265 , 2396-2404 8133

Milburn, M. V., Prive, G. G., Millian, D. L., Scott, W. G., Yeh, J., Jancarik,

Milligan, D. L., and Koshland, D. E., Jr. (1988) J. Biol. Chem. 263,6268-6275

Model, P., and Russel, M. (1988) The Bacteriophages (Calander, R., ed) pp. Milligan, D. L., and Koshland, D. E., Jr. (1991) Science 254,1651-1654

Newman, A. K., Rubin, R. A,, Kim, S-H., and Modrich, P. (1981) J Biol. Chem.

Ohno-Iwashita, Y., and Wickner, W. (1983) J. Bioi. Chem. 258,1895-1900 Pakula, A. A., and Simon, M. I. (1992a) Proc. Natl. Acad. Sei U. S. A. 89,4144-

and Rothfield, L. I. (1986) J. Biol. Chem. 2 6 1 , 428-443

Sei. U. S. A. 88,5939-5943

(1984) J. Bacteriol. 158 , 665-673

J., Koshland, D. E., Jr., and Kim, S.-H. (1991) Science 254,1342-1347

375-346, Plenum Publishing Corp., New York

256,2131-2139

A 1 A5 Pait;:, A. A., and Simon, M. I. (1992b) Nature 355,496-497 Rosenberg, A. H., Lade, B. N., Chui, D.-S., Lin, S.-W., Dunn, J. J., and Studier,

F. W. (1987) Gene (Amst.) 56,125-135 Russel, M. (1991) Mol. Microbioi. 5,1607-1613 Russel, M., and Model, P. (1986) J. Biol. Chem. 261,14997-15001 Schlessinger, J. (1988) Trends. Biochem. Sci. 13 , 443-447 Strauch, K. L., and Beckwith, J. (1988) Proc. Natl. Acad. Sci. U. S. A. 8 5 ,

Strauch, K. L., Johnson, K., and Beckwith, J. (1989) J. Bacteriol. 171 , 2689-

Studier F. W., and Moffatt B. A. (1986) J. Mol. Biol. 189,113-130 Sun, T-'P., and Webster, R.'E. (1987) J. Bacteriol. 169,2667-2674 Towbin, H., Staehelin, T., and Gordon, J. (1979) Proc. Natl. Acad. Sci. U. S. A.

1576-1580

2696

76,4350-4354 Yardin, Y., and Ullrich, A. (1988) Annu. Reu. Biochem. 5 7 , 443-478

Zimmerrnann, R., Watts, C., and Wickner, W. (1982) J . Biol. Chem. 257,6529- Yen, T. S. B., and Webster, R. E. (1981) J. Bid. Chem. 256 , 11259-11265

6536 Zoller, M. J., and Smith, M. (1984) DNA 3,479-488