Proc. Nati. Acad. Sci. USA
`Vol. 86, pp. 343-346, January 1989
`Microbiology
`
`Cloned diphtheria toxin within the periplasm of Escherichija coli
`causes lethal membrane damage at low pH
`
`(membrane insertion)
`
`DONALD 0. O'KEEFE AND R. JOHN COLLIER*
`Department of Microbiology and Molecular Genetics, and The Shipley Institute of Medicine, Harvard Medical School, 200 Longwood Avenue,
`Boston, MA 02115
`
`Communicated by Bernard D. Davis, September 30, 1988
`
`ABSTRACT
`Acidic pH within endosomal vesicles of sen-
`sitive animal cells triggers a conformational change in diph-
`theria toxin (DT) that is believed to cause the B chain to insert
`into the vesicular membrane and the enzymic A chain to be
`released into the cytosol. In artificial lipid bilayers, DT forms
`ion-conductive channels under mildly acidic conditions (pH
`-5). Here we report a related phenomenon in Escherchia coli
`strains that secrete certain cloned DT-related proteins into
`their periplasm: the cells are rapidly killed at pH 5 but remain
`unharmed at pH 7. Expression of full-length DT (an active-site
`mutant, to comply with the National Institutes of Health
`recombinant DNA guidelines) causes acid-sensitivity, whereas
`expression of the A chain alone does not. The killed cells are not
`lysed, but inner-membrane functions are impaired (membrane
`potential, active transport, and ion impermeability). We pro-
`pose that acidification of DT within the periplasm induces its
`insertion into the inner membrane, lethally damaging the
`permeability barrier. This discovery provides a potentially
`important selection procedure for mutations affecting the
`membrane insertion function of DT. Similar approaches may
`be useful in studying other proteins that undergo condition-
`dependent interaction with membranes.
`
`Diphtheria toxin (DT; Mr, 58,342) is representative of a class
`of toxic proteins that are believed to undergo membrane
`insertion and/or penetration under the mildly acidic condi-
`tions prevailing within certain intracellular membrane-bound
`vesicles. This class also includes Pseudomonas aeruginosa
`exotoxin A (1, 2) and modeccin (3, 4), and possibly anthrax
`toxin (5, 6), tetanus toxin (7, 8), and botulinum toxin (8, 9).
`Membrane insertion/penetration steps are not well under-
`stood for any toxin. This paper describes a method for
`studying the problem in bacteria.
`DT is synthesized by Corynebacterium diphtheriae as a
`single-chain protein containing a signal peptide that is re-
`moved during secretion from the bacterium yielding the
`535-residue mature toxin. Before or soon after the toxin
`attaches to sensitive mammalian cells, it is cleaved proteolyt-
`ically within one of its two disulfide loops; this generates an
`amino-terminal A fragment of DT (DTA, 193 residues) and a
`carboxyl-terminal B fragment (DTB, 342 residues), linked by
`a disulfide bond. After binding to its receptor the toxin
`undergoes receptor-mediated endocytosis and is conveyed to
`an endosome (10). Acidification of the endosomal lumen
`triggers a conformational change in the toxin, leading to
`membrane insertion by way of hydrophobic segments within
`DTB. This insertion somehow mediates transfer of DTA to
`the cytoplasmic face of the membrane, where it is believed to
`be released after reduction of the disulfide bond linking it to
`DTB. DTA then catalyzes transfer of the adenosine diphos-
`phate ribose moiety of NAD to elongation factor 2, thereby
`
`inactivating the factor, blocking protein synthesis, and killing
`the cell.
`Various gene fragments encoding nontoxic or hypotoxic
`segments of DT have been cloned and expressed in Esche-
`richia coli (11-13). Fragments containing the signal peptide
`are secreted to the periplasmic space, with the signal peptide
`being removed in the process leaving soluble biologically
`active proteins in this compartment (11, 14). Although the
`intact structural gene for DT has not yet been cloned, owing
`to constraints of the National Institutes of Health Guidelines,
`an intact DT gene containing a 3-base-pair mutation in an
`active-site residue (E148S) has been reconstructed and ex-
`pressed (14). This mutation diminished ADP-ribosyltrans-
`ferase activity and cytotoxicity by several hundred-fold.
`DT-E148S was secreted to the periplasmic space of E. coli,
`where it was processed into a stable and apparently correctly
`folded protein (14).
`The cloning and expression of a full-length enzymatically
`inactive form ofDT allows us to investigate structural aspects
`ofthe toxin having participatory roles in receptor binding and
`membrane penetration. In this paper we demonstrate that
`when E. coli containing DT-E148S in the periplasm are
`exposed to acidic pH, the cells die. Several tests for inner-
`membrane function suggest that DT-E148S inserts into the
`membrane and disrupts the permeability barrier. The lethal
`effect creates a positive genetic selection for mutant DT
`molecules incapable of membrane insertion. Such mutant
`toxins may be useful in elucidating the molecular mechanisms
`involved in DT translocation across mammalian cell mem-
`branes.
`
`MATERIALS AND METHODS
`Plasmids. All plasmids are derivatives of pCDptac2 (15)
`and contain coding sequences downstream from the tac
`promoter, a hybrid promoter inducible with isopropyl 18-
`D-thiogalactopyranoside and containing sequences from both
`the trp and lacUVS promoters of E. coli (16). pDO1 contains
`sequences encoding the full-length DT-E148S molecule.
`pDO3 was derived from pDO1 and encodes DTA-E148S plus
`7 amino acids of the amino terminus of DTB, plus 2 amino
`acids added during cloning. ptacF2-E148S encodes F2-
`E148S, a fragment of DT containing all of DTA-E148S and
`55% of DTB. Plasmids used in this study are summarized in
`Table 1. E. coli JM103 was used in all experiments (15).
`Unless otherwise noted, all experiments were performed on
`uninduced cultures.
`Membrane Potential. Cells grown in L broth to an OD595 of
`1.0 were resuspended in 2 mM Tris'HCI/50 mM NaCI, pH
`7.0, to 3 x 109 cells per ml. For each experimental point, the
`cells were diluted 1:10 into 5 mM buffer at various pH values
`
`The publication costs of this article were defrayed in part by page charge
`payment. This article must therefore be hereby marked "advertisement"
`in accordance with 18 U.S.C. §1734 solely to indicate this fact.
`
`Abbreviations: DT, diphtheria toxin; DTA, fragment A of DT; DTB,
`fragment B of DT.
`*To whom reprint requests should be addressed.
`
`343
`
`

`

`344
`
`Microbiology: O'Keefe and Collier
`
`Proc. Natl. Acad. Sci. USA 86 (1989)
`
`markedly inhibited after acid treatment, whereas cells syn-
`thesizing the corresponding mutant DTA (DTA-E148S) or
`containing the vector alone were unaffected (data not
`shown).
`Next we plated E. coli containing plasmids encoding
`various forms of DT-E148S on solid media with various pH
`values and determined cell viability by colony counts. As
`seen in Fig. 1, viability of E. coli synthesizing full-length
`DT-E148S decreased by -50 times at pH 6.0; by 105 at pH
`5.5; and by almost 107 at pH 5.0. Death at acidic pH was
`rapid. A 1-min exposure of cells to pH 5.0, followed by
`plating at pH 8.0, yielded maximal levels of cell killing (data
`not shown). By contrast, cells producing DTA-E148S, or
`cells harboring the plasmid vector lacking DT sequences,
`showed only a modest (-2 times) decrease in viability at pH
`5.0 and were essentially unaffected at higher pH values.
`When cells were induced with 1 mM isopropyl P-D-thioga-
`lactopyranoside at pH 6.0 viability of DT-E148S-producing
`cells decreased by -105 compared to cells induced at pH 7.5,
`whereas cells producing DTA-E148S were unaffected (data
`not shown). This suggests that the lethality observed is a
`function of both the pH and the concentration of DT-E148S
`in the periplasm. Uninduced strains producing DT-E148S and
`DTA-E148S contained stoichiometrically equivalent amounts
`of DT-related protein in the periplasm (-5000 molecules per
`cell).
`A DT fragment (F2-E148S) comprising all of DTA-E148S
`plus 55% of DTB caused a similar, although somewhat lesser,
`decline in viability at pH 5.0, but little loss of viability at
`higher pH values (Fig. 1). A strain expressing wild-type F2
`gave identical results (data not shown). The F2 fragment is 6
`residues longer than CRM45 (23, 24), a chain-termination
`fragment that contains a substantial fraction of the hydro-
`phobic region within DTB (25). CRM45 has been shown to
`form ion-conductive channels under acidic pH conditions
`(26). These results implicate hydrophobic regions of DTB in
`the pH-dependent killing of E. coli. The fact that both the F2
`and F2-E148S fragments were produced in substantially
`lower amounts than DT-E148S, as estimated by ELISA, may
`account for their reduced effect on viability.
`
`0
`
`100
`
`1
`
`1
`
`0
`
`10-2 -
`
`10-4
`10-5
`
`o
`
`0
`
`0N
`
`0
`
`7.0
`
`5.0
`
`6.0
`pH
`Viability of cells harboring plasmids encoding all or part
`FIG. 1.
`of DT-E148S when plated on media of various pH values. E. coli
`JM103, containing plasmids, were grown to midlogarithmic phase
`and plated on L agar containing either 100 mM Pipes (pH 7.5-6.5) or
`50 mM sodium citrate/50 mM sodium succinate (pH 6.0-5.0). One
`day later viability was assessed and scored relative to cells plated at
`pH 7.5. Identical results were seen at pH 5.5 when 100 mM sodium
`acetate was used instead of sodium citrate/sodium succinate. Plas-
`mids are as follows: o, pDO1 (DT-E148S); A, pDO3 (DTA-E148S);
`o, ptacF2-E148S (F2-E148S).
`
`Plasmids used in this study
`Table 1.
`Encoded protein
`Plasmid
`pDO1
`DT-E148S, a full-length DT mutated at the active
`site
`DTA-E148S, a DTA mutated at the active site
`plus 7 amino acids of the amino terminus of
`DTB plus the dipeptide Lys-Ser
`F2, an enzymically active protein containing
`wild-type DTA and 55% of the amino terminus
`of DTB
`F2-E148S, an F2 mutated at the active site
`
`pDO3
`
`ptacF2
`
`ptacF2-E148S
`
`containing 50 mM NaCl. From this point, the cells were kept
`at 37°C in a thermoregulated magnetically stirred cuvette.
`The following buffers were used: Pipes (pH 7.0 and 6.5), Mes
`(pH 6.0), and sodium citrate/sodium succinate (pH 5.5 and
`5.0). The cells were incubated for 1 min and then brought to
`100 mM Tris-HCl (pH 7.5). One minute later 3,3'-dipropyl-
`thiadicarbocyanine iodide (Molecular Probes, Eugene, OR)
`was added to a final concentration of 1 ,ug/ml. The fluores-
`cence, measured when the signal stabilized, was divided by
`the fluorescence at pH 7.0 to obtain relative fluorescence.
`Fluorescence was measured in a SLM AMINCO (Urbana,
`IL) SPF-500C spectrofluorometer with excitation at 645 nm
`and emission at 668 nm. Both slits were set at 5 nm.
`Proline Transport. Cells grown overnight in potassium-free
`M9 medium (sodium phosphate was substituted for potas-
`sium phosphate) were diluted 1:100 in the same medium.
`After growing for 5 hr, the cells were resuspended in growth
`medium to an OD595 of 1.0. Cells (0.1 ml) were then added to
`1.9 ml of 20 mM buffer at various pH values containing 4.8
`,uCi (1 Ci = 37 GBq) of L-[2,3,4,5-3H]proline. After 10 min at
`37°C, the cells were filtered through Millipore HA filters and
`washed with 5 ml of potassium-free M9 salts. The filters were
`dried and dissolved in OCS (Amersham), and the radioac-
`tivity was measured in a scintillation counter. Buffers were
`identical to those used in the membrane potential experi-
`ments.
`86Rb EMux. Cells were grown as for the transport assays.
`After resuspending to an OD595 of 1.0, the cells were
`incubated in the presence of 86RbCl (20 ,Ci/ml) for 1 hr at
`37°C. Cells (0.1 ml) were then added to 1.9 ml of 20 mM buffer
`at various pH values and incubated for 10 min at 37°C. The
`cells were then filtered through Millipore HA filters pre-
`soaked in wash buffer. The filters were washed with 5 ml of
`M9 salts supplemented with 10 mM RbCl and dried, and
`radioactivity was determined in a y counter. Buffers were
`identical to those used in the membrane potential experi-
`ments.
`
`RESULTS
`When exposed to acidic pH, DT and certain truncated forms
`of DT undergo a conformational change that exposes latent
`hydrophobic regions within the B chain (17-19). If the pH is
`reduced in the presence of artificial planar lipid bilayers, the
`toxin undergoes membrane insertion and forms ion-con-
`ductive channels (20). Similarly, if toxin that is bound to the
`mammalian cell surface is exposed to acidic conditions, it is
`able to translocate the DTA moiety across the membrane into
`the cytosol (21, 22).
`The behavior of DT under acidic conditions suggested that
`membrane-insertion-competent forms of the toxin residing in
`the periplasm of E. coli might insert into the plasma mem-
`brane when the bacteria were exposed to acid pH. Initially
`we monitored cell growth spectrophotometrically. Cultures
`were adjusted to pH 5.0, incubated for 10 min, and neutral-
`ized; and 2 hr later the OD595 was measured. The results
`indicated that growth of cells synthesizing DT-E148S was
`
`

`

`Microbiology: O'Keefe and Collier
`
`Proc. Natl. Acad. Sci. USA 86 (1989)
`
`345
`
`140
`
`120
`
`100
`
`80
`
`60
`
`40
`
`20
`
`0
`
`7.0
`
`6.0
`pH
`
`5.0
`
`0c0 nC
`
`6
`
`0 0a-
`
`ID
`0
`
`° 120
`
`100
`
`80
`
`C0o
`
`0*
`
`a 60
`a,
`-Y
`
`.C
`zL
`
`I
`
`40
`
`20
`6
`
`0
`
`U)
`U)
`
`0rC)
`
`7.0
`
`6.0
`pH
`
`5.0
`
`Proline transport (A) and 86Rb efflux (B) in E. coli JM103
`FIG. 3.
`containing various plasmids after acid pH treatment. Cell-associated
`radioactivity is expressed relative to samples incubated at pH 7.0. o,
`pDO1 (DT-E148S); A, pDO3 (DTA-E148S); and a, pCDptac2
`(vector).
`toxin receptor found on sensitive mammalian cells is not
`surprising, since the toxin is known to insert into protein-free
`artificial lipid bilayers under acidic conditions. Loss of
`inner-membrane function, as demonstrated, could account
`for the lethal effects of acidic pH on toxin-containing cells
`and seems likely to be the primary effect of the toxin. We
`have not yet studied the possibility that outer-membrane
`integrity or function is altered. It may be that channels similar
`to those observed in artificial bilayer systems are formed in
`the inner membrane (diameter estimates are in the range of 18
`A) (26), but we lack specific information about this. Prelim-
`inary results indicate that ultraviolet-absorbing material of
`low molecular mass is released when DT-E148S-producing
`cells are treated at low pH.
`Although acidification of the extracellular milieu of E. coli
`should cause an approximately parallel decrease in pH of the
`periplasm, it is unlikely that the periplasmic pH is precisely
`the same as that ofthe external medium. Stock et al. (31) have
`predicted that the periplasmic pH should be significantly
`below that in the medium, due to the Donnan potential across
`the outer membrane and the buffering capacity of ionic
`species within the periplasm. Regardless ofthe precise values
`of periplasmic pH, one would only expect an approximate
`correlation between pH profiles of cell killing with those of
`membrane insertion activity (20) or membrane translocation
`(21) in other systems. Membrane insertion in some systems
`represents initial rate measurements whereas in others it does
`not, and all the activities depend on variables besides pH
`(e.g., ionic strength) that can not be precisely controlled in
`the bacterial periplasm.
`Although our results do not entirely exclude the possibility
`that the DTA moiety of DT-E148S and F2-E148S might play
`some role in generating the acid-sensitivity phenotype, ADP-
`ribosyltransferase activity is almost certainly irrelevant. No
`macromolecular substrates for DTA have been found in E.
`coli, and DTA expression in the cytoplasm, as well as its
`secretion to the periplasm, is without apparent detrimental
`effect (32). More importantly, we have found that the
`pH-sensitivity profiles of strains synthesizing wild-type
`(E148) and mutant (E148S) forms of the F2 fragment are
`identical. Finally, it is noteworthy that peptides contained
`within DTB are sufficient for channel formation in vitro (33).
`An important implication of these studies is that they
`define an approach to isolating additional classes of mutants.
`For studies of DT structure and activity the system can be
`used as a positive genetic selection of toxin mutations
`affecting membrane insertion and/or formation of ion-
`conductive channels. Directed and/or random mutagenesis
`
`7.0
`
`6.0
`pH
`
`5.0
`
`2.6
`
`2.4
`
`2.2
`
`2.0
`
`1.8
`
`1.6
`
`1.4
`
`1.2
`
`1.0
`
`ci)
`)
`Ql)
`
`Uc
`
`0 c
`
`i)
`
`FIG. 2. Membrane potential of E. coli JM103 containing various
`plasmids after exposure to acid pH. o, pDO1 (DT-E148S); A, pDO3
`(DTA-E148S); a, pCDptac2 (vector).
`
`The acid-induced death seen in cells synthesizing DT-
`E148S or F2-E148S might result from the formation of
`ion-conducting channels in the inner membrane, similar to
`those formed when cells are treated with certain colicins (27).
`In vitro, colicin El forms channels in artificial lipid mem-
`branes at low pH (28). In vivo, it depolarizes the inner
`membrane of E. coli, impairing active transport, and there is
`an accompanying loss of intracellular ions (27).
`To determine if there were effects of acidic pH on inner
`membrane function in cells expressing DT-E148S, we mea-
`sured membrane potential, active transport, and Rb efflux.
`To measure membrane potential we used the fluorescent dye
`3,3'-dipropylthiadicarbocyanine iodide, which undergoes
`membrane potential-sensitive concentration and quenching
`within cells (29, 30). Fig. 2 shows that at pH 6.0 or lower the
`potential in cells producing DT-E148S was dissipated. A
`minimal effect was observed in cells containing either the
`DTA-E148S plasmid or the vector alone. When cells produc-
`ing DTA-E148S were treated at pH 5.0 in the presence of
`externally added DT at 0.1 mg/ml, no effect was observed
`(data not shown). This demonstrates that dissipation of the
`membrane potential is due to cell-associated toxin and not to
`toxin released from spontaneously lysing cells.
`Fig. 3A shows that the acid-induced loss of membrane
`potential in cells producing DT-E148S is accompanied by a
`loss in active transport as measured by proline uptake, and
`Fig. 3B shows that the toxin-producing cells are unable to
`retain 86Rb at pH 6.0 or below. For each of these assays little
`or no effect of acid pH was found on cells containing the
`plasmid vector alone or those synthesizing only DTA-E148S.
`Microscopic examination of cells containing the whole
`toxin treated at pH 5.0 showed that cellular integrity was
`maintained. Furthermore, we found that glucose-6-phos-
`phate dehydrogenase, a cytoplasmic enzyme, remained cell-
`associated after treatment of DT-E148S-producing cells at
`low pH.
`
`DISCUSSION
`The results presented are consistent with a model whereby an
`acidic environment, inducing a conformational change in the
`B moiety of periplasmic DT-E148S, enables the protein to
`insert into the inner membrane and disrupt membrane po-
`tential. Insertion of the toxin into a membrane devoid of the
`
`

`

`346
`
`Microbiology: O'Keefe and Collier
`
`Proc. Natl. Acad. Sci. USA 86 (1989)
`
`of the cloned toxin gene, followed by transformation into E.
`coli and analysis of survivors plated onto acidic plates, may
`yield various classes of mutants useful in analyzing the poorly
`understood processes of membrane insertion and transloca-
`tion. Membrane perturbations caused by DT may not be
`restricted to acid induction; other variables (e.g., ionic
`strength, specific ions, or osmolarity) might be used to induce
`lethality, and hence be used for selection. Furthermore, ifcell
`death is due to periplasmic DT-E148S and not to a minor
`population of toxin molecules left in the cytoplasm then this
`system might also be used to isolate conditional secretory
`mutants, either of the toxin or of bacterial constituents.
`The general approach described might also be adapted to
`the study of other proteins that normally interact with
`biological membranes at low pH or under other conditions.
`Besides the other toxic proteins that undergo pH-dependent
`membrane interactions, there are proteins present on certain
`enveloped viruses (e.g., the hemagglutinin of influenza) that
`mediate fusion between the viral envelope and target mem-
`branes when exposed to a low pH (34) and a number of other
`proteins that induce fusion ofphospholipid vesicles at low pH
`(35).
`
`This work was supported by National Institutes of Health grants
`(AI-22021 and AI-22848). Partial support was also received from the
`Shipley Institute of Medicine. D.O.O. is the recipient of a National
`Research Service Award from the National Institute of Allergy and
`Infectious Diseases.
`
`1.
`
`2.
`
`3.
`
`4.
`5.
`6.
`
`7.
`8.
`9.
`
`10.
`
`Morris, R. E. & Saelinger, C. B. (1986) Infect. Immun. 52,445-
`453.
`Farahbakhsh, Z. T., Baldwin, R. L. & Wisnieski, B. J. (1986)
`J. Biol. Chem. 261, 11404-11408.
`Draper, R. K., O'Keefe, D. O., Stookey, M. & Graves, J.
`(1984) J. Biol. Chem. 259, 4083-4088.
`Sandvig, K. & Olsnes, S. (1982) J. Biol. Chem. 257, 7504-7513.
`Friedlander, A. M. (1986) J. Biol. Chem. 261, 7123-7126.
`Gordon, V. M., Leppla, S. H. & Hewlett, E. L. (1988) Infect.
`Immun. 56, 1066-1069.
`Roa, M. & Boquet, P. (1985) J. Biol. Chem. 260, 6827-6835.
`Simpson, L. L. (1983) J. Pharmacol. Exp. Ther. 225, 546-552.
`Donovan, J. J. & Middlebrook, J. L. (1986) Biochemistry 25,
`2872-2876.
`Marnell, M. H., Shia, S.-P., Stookey, M. & Draper, R. K.
`(1984) Infect. Immun. 44, 145-150.
`
`11.
`
`12.
`
`13.
`
`14.
`
`15.
`
`16.
`17.
`18.
`
`19.
`
`20.
`
`21.
`22.
`23.
`
`24.
`
`25.
`
`26.
`
`27.
`28.
`
`29.
`30.
`
`31.
`
`32.
`
`33.
`
`34.
`
`35.
`
`Tweten, R. K. & Collier, R. J. (1983) J. Bacteriol. 156, 680-
`685.
`Leong, D., Coleman, K. D. & Murphy, J. R. (1983) Science
`220, 515-517.
`Kaczorek, M., Delpeyroux, F., Chenciner, N., Streeck, R. E.,
`Murphy, J. R., Boquet, P. & Tiollais, P. (1983) Science 221,
`855-858.
`Barbieri, J. T. & Collier, R. J. (1987) Infect. Immun. 55, 1647-
`1651.
`Douglas, C. M., Guidi-Rontani, C. & Collier, R. J. (1987) J.
`Bacteriol. 169, 4962-4966.
`Russell, D. R. & Bennett, G. N. (1982) Gene 20, 231-243.
`Sandvig, K. & Olsnes, S. (1981) J. Biol. Chem. 256, 9068-9076.
`Collins, C. M. & Collier, R. J. (1987) in Membrane Mediated
`Cytotoxicity, eds. Bonavida, B. & Collier, R. J. (Liss, New
`York), pp. 41-52.
`Blewitt, M. G., Chung, L. A. & London, E. (1985) Biochem-
`istry 24, 5458-5464.
`Donovan, J. J., Simon, M. I., Draper, R. K. & Montal, M.
`(1981) Proc. Natl. Acad. Sci. USA 78, 172-176.
`Draper, R. K. & Simon, M. I. (1980) J. Cell Biol. 87, 849-854.
`Sandvig, K. & Olsnes, S. (1980) J. Cell Biol. 87, 828-832.
`Greenfield, L., Bjorn, M. J., Horn, G., Fong, D., Buck, G. A.,
`Collier, R. J. & Kaplan, D. A. (1983) Proc. Natl. Acad. Sci.
`USA 80, 6853-6857.
`Uchida, T., Pappenheimer, A. M., Jr., & Greany, R. (1973) J.
`Biol. Chem. 248, 3838-3844.
`Boquet, P., Silverman, M. S., Pappenheimer, A. M., Jr., &
`Vernon, W. B. (1976) Proc. Natl. Acad. Sci. USA 73, 4449-
`4453.
`Kagan, B. L., Finkelstein, A. & Colombini, M. (1981) Proc.
`Natl. Acad. Sci. USA 78, 4950-4954.
`Konisky, J. (1982) Annu. Rev. Microbiol. 36, 125-144.
`Davidson, V. L., Cramer, W. A., Bishop, L. J. & Brunden,
`K. R. (1984) J. Biol. Chem. 259, 594-600.
`Waggoner, A. S. (1979) Annu. Rev. Biophys. Bioeng. 8, 47-68.
`Laszlo, D. J. & Taylor, B. L. (1981) J. Bacteriol. 145, 990-
`1001.
`Stock, J. B., Rauch, B. & Roseman, S. (1977) J. Biol. Chem.
`252, 7850-7861.
`Greenfield, L., Dovey, H. F., Lawyer, F. C. & Gelfand, D. H.
`(1986) BiolTechnology 4, 1006-1011.
`Deleers, M., Beugnier, N., Falmagne, P., Cabiaux, V. &
`Ruysschaert, J.-M. (1983) FEBS Lett. 160, 82-86.
`White, J., Kielian, M. & Helenius, A. (1983) Q. Rev. Biophys.
`16, 151-195.
`Kim, J. & Kim, H. (1986) Biochemistry 25, 7867-7874.
`
`