Proc. Nati. Acad. Sci. USAVol. 83, pp. 6164-6168, August 1986Microbiology

Misread protein creates membrane channels: An essential step inthe bactericidal action of aminoglycosides

(membrane protein/protein export/Escherichia coli/puromycin/streptomycin)

BERNARD D. DAVIS*, LINGLING CHEN*t, AND PHANG C. TAI*t*Bacteral Physiology Unit, Harvard Medical School, Boston, MA 02115; and tBoston Biomedical Research Institute, Boston, MA 02114

Contributed by Bernard D. Davis, April 29, 1986

ABSTRACT Among the pleiotropic effects of aminoglyco-sides, their irreversible uptake and their blockade of initiatingribosomes have appeared to explain their bactericidal action,while the contributions of translational misreading and mem-brane damage and the mechanism of that damage haveremained uncertain. We now present evidence that incorpo-ration of misread proteins into the membrane can account forthe membrane damage. The bactericidal action thus appears toresult from the following sequence, in which each step isessential: slight initial entry of the antibiotic; interaction withchain-elongating ribosomes, resulting in misreading; incorpo-ration of misread protein into the membrane, creating abnor-mal channels; increased (and irreversible) entry through thesechannels, and hence increased misreading and formation ofchannels; and, finally, blockade of initiating ribosomes. Thismechanism can account for several previously unexplainedobservations: that streptomycin uptake requires protein syn-thesis during, but not after, the lag before the membranedamage; that streptomycin-resistant cells, which fail to take upstreptomycin, can do so after treatment by another amino-glycoside; and that puromycin at moderate concentrationsaccelerates streptomycin uptake, while high concentrations(which release shorter chains) prevent it. In addition,puromycin, prematurely releasing polypeptides of normalsequence, also evidently creates channels, since it is reported topromote streptomycin uptake even in streptomycin-resistantcells. These findings imply that normal membrane proteinsmust be selected not only for a hydrophobic anchoring surface,but also for a tight fit in the membrane.

Extensive studies of the mechanism of action of aminogly-cosides, ever since the discovery of streptomycin (Str) in1944, have revealed a remarkably pleiotropic set of effects (1,2). Among these, it has not been clear whether misreading hasa role in the bactericidal action, how the membrane damagearises, and why reversible inhibitors of protein synthesisprevent that damage and the bactericidal action. The dem-onstration that bacteria possess membrane-bound ribosomes(3, 4), involved in protein export, has renewed our interest inthese problems.Anand et al. discovered the membrane damage in 1960 (5).

Addition of [14C]Str to a growing culture of Escherichia coliresulted in immediate adsorption of this highly cationicantibiotic to the cell surface, followed, after a lag of someminutes, by a rapid larger secondary uptake. Since treatmentwith toluene permitted an immediate secondary uptake, itappeared that the cell membrane is relatively impermeable toStr but becomes permeable during exposure to it. Moreover,this effect involves formation of nonspecific channels, per-meable to small molecules regardless of charge: treatmentwith Str permitted entry of citrate (which is normally exclud-

ed by E. colt), and of a S-galactoside in a transport-defectivemutant (6); in the other direction, it impaired the ability toconcentrate ['4C]valine (6) and caused leakage of nucleotides(7) and K+ (8). Inhibition of protein synthesis by chloram-phenicol, and also genetic resistance to Str, prevented themembrane damage (5).

Since the leakage of K+ could be detected as early as theinhibition of protein synthesis (9), these findings originallysuggested a direct effect of Str on the integrity of newlyformed membrane. However, Erdos and Ullmann (10)showed that Str decreased protein synthesis in a bacterialextract; and an action on ribosomes, suggested by Spotts andStanier (11) on indirect grounds, was soon demonstrated(12-14). Moreover, ribosomes from resistant mutants did notshow this response (except at excessive Str concentrations).Since it seemed unlikely that a single mutation could alterboth the ribosome and the membrane, the membrane damagewas universally dismissed as secondary to killing. Moreover,at that time there was no basis, conceptual or experimental,for connecting the effects of Str on the ribosome and on themembrane.The action on the ribosome became more complicated

when Gorini and co-workers found that Str can also causemisreading (15, 16). Since misreading and blockade aremutually exclusive, the finding that ribosomes bind only onemolecule of Str (17) created a paradox, which was resolvedby separating the effects of antibiotics on ribosomes indifferent stages of their cycle (18). With ribosomes engagedin chain elongation, Str slows translation and causes mis-reading, while the more flexible free ribosomes bind it in away that allows formation ofunstable initiation complexes (inthe translation of viral RNA) but prevents their transition intochain elongation (19-21). Since this blockade of initiatingribosomes could account for the cessation of protein synthe-sis in killed cells, the misreading effect, although of greattheoretical interest, seemed to be irrelevant to the bacteri-cidal action.However, a possible role of misreading arose again when

Bassford et al. (22) found that certain genetic fusions of/-galactosidase with a periplasmic protein were lethal, ap-parently because the fused protein became stuck in themembrane and blocked the secretion of other proteins. Sinceit seemed possible that misread protein might have a similarfate, we have examined the distribution of the proteinsformed after partial inhibition ofprotein synthesis by Str. Theresults indicate that misread protein is indeed incorporated inthe membrane. However, that incorporation does not appearto be the direct cause of cell death but instead accounts forthe limited membrane damage. This mechanism ofmembranedamage, as an essential part of the bactericidal process,explains several previously puzzling features of aminoglyco-side action, and these will be reviewed in the Discussion.

Abbreviation: Str, streptomycin.

6164

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Proc. Natl. Acad. Sci. USA 83 (1986) 6165

MATERIALS AND METHODS

Strain and Media. E. coli ATCC strain 27257, Met- andconstitutive for alkaline phosphatase, was grown with shak-ing, at 370C, in minimal medium A (23) with 0.2% glucose andsupplemented with 20 gg of methionine per ml.

Fractionation of Radioactively Labeled Protein. Portions (15ml) of a culture in exponential growth, at =4 X 108 cells perml, were incubated with Str as indicated and then werefurther incubated for 5 min with the addition of 100 gCi (1 Ci= 37 GBq) of [35S]methionine. The degree of inhibition ofprotein synthesis by Str was determined by comparison witha similarly incubated culture without Str. The cells wereharvested, washed with 10mM Tris HCl, pH 7.6/50mM KCl,and fractionated as described (24). The periplasmic fractionwas released by EDTA-lysozyme or cold shock, the cellswere lysed by sonication, unbroken cells were removed bycentrifugation at 3000 x g for 5 min, and the cytoplasm wasrecovered as the supernatant after 1 hr of centrifugation at40,000 rpm in a Beckman 50 Ti rotor. Fractions expected toconsist mostly of inner membrane and of outer membranewere obtained by successively extracting the sediment with2% Triton X-100/10 mM Tris HCl, and then with the samecontaining 10 mM EDTA (24), and the residual sediment wassolubilized by hot 1% NaDodSO4/10 mM Tris.HCl/100 mMNaCl. The fractions were made up to equal volumes, andequal portions were measured for their radioactivity and weresubjected to gel electrophoresis and to immunoprecipitation,as described (25).

Reagents. [35S]Methionine was obtained from New En-gland Nuclear, and other chemicals were of reagent grade.The rabbit antiserum to E. coli alkaline phosphatase has beendescribed (3).

RESULTS

Altered Distribution of Misread Proteins. To test the pos-sibility that the misreading caused by Str might impair thesecretion of proteins across the plasma membrane, wetreated a growing culture of E. coli, constitutive for alkalinephosphatase, with Str, and after it reached either 50% or 95%inhibition of protein synthesis we briefly incubated portions(along with an untreated sample) with [35S]methionine. Thecells were fractionated into periplasm, cytoplasm, the bulk ofthe inner and the outer membrane, and a residue insoluble innonionic detergent.As Table 1 shows, treatment with Str did alter the distri-

bution of the subsequently formed protein. The proportionsecreted into the periplasm was decreased 40-50%o, indicat-ing that misreading prevents some of the secretory proteinfrom completing the transfer across the membrane. The

proportion in the cytoplasm was decreased slightly, whilethat in the poorly soluble residue was increased 3-fold. Thisincrease evidently represents misread proteins (partly secre-tory, but probably also from other classes) whose abnormalfolding decreases their solubility in Triton.

Table 1 also shows that the degree of interference withprotein secretion did not increase significantly when theinhibition of synthesis by Str was increased from 50% to 95%:the distribution of the newly formed protein was essentiallythe same. The implications for the mechanism of the lethalaction of the antibiotic are discussed below.Accumulation of Alkaline Phosphatase in the Membrane. To

verify the accumulation of a secretory protein in the mem-brane, we used gel electrophoresis and immunoprecipitationto observe the distribution of alkaline phosphatase (which isnormally periplasmic) in the fractions from the cells of Table1A. Gel electrophoresis (Fig. 1) showed that in theperiplasmic fraction the band of newly formed radioactivealkaline phosphatase monomers (47 kDa) was decreased inthe cells treated with Str for 20 min (25% inhibition of totalprotein synthesis), and it was virtually absent from the 30-minsample (50% inhibition). [This band is seen to lie immediatelyabove a heavy band of elongation factor EFTu, as observedin periplasmic fractions prepared by cold shock (26); we didnot observe the EFTu band in an EDTA-lysozyme prepara-tion (not shown).] Conversely, in the residual fractions Fig.1 shows that a relatively weak band appeared at that positionin the 20-min sample, and more in the 30-min sample. Thetreatment with Str affected only the proteins formed in itspresence, since there was no change in the distribution of thetotal proteins (mostly formed before treatment), detected byCoomassie blue staining (not shown).

Immunoprecipitation of these fractions with antiserum toalkaline phosphatase, followed by solubilization and gelelectrophoresis, confirmed the shift in the distribution of thisprotein. As Fig. 2 shows, in the immunoprecipitates theuntreated cells showed a heavy band in the periplasmicfraction at the position of alkaline phosphatase and none inthe residual fraction, while in the cells treated with Str theperiplasmic band was reduced by 80-90%, and the residualfraction yielded a weak band at the same position.

It is not surprising that the residual fraction of treated cellsyielded a much weaker band than the periplasm of untreatedcells. The generalized misreading caused by Str shouldchange both the size and the sequence of polypeptides, andamong the alkaline phosphatase molecules altered in this wayone should detect in the membrane only those that combinethree features: they would be sufficiently altered to beretained in the membrane, would not be altered in size, andwould retain enough of the native immunological determi-nants to be able to react with polyclonal antibody. While

Table 1. Distribution of newly synthesized proteins in cells treated with Str

% radioactivity in

Exp. Treatment Periplasm Cytoplasm IM OM ResidualA - Str 13.4 53.2 21.0 5.8 6.6

+ Str (20 pg/ml)(50% inhibition) 7.9 44.0 21.3 6.0 20.9

B - Str 20.8 40.0 22.0 6.1 11.1+ Str (200 jig/ml)(95% inhibition) 11.6 34.1 14.3 7.9 32.1

Portions of a growing culture of E. coli, as described in Materials and Methods, were incubated withor without Str (20 ,ug/ml) for 30 min (Exp. A), or with or without Str (200 ug/ml) for 8 min (Exp. B),and then were further incubated with [35S]methionine for S min. The cells were harvested andfractionated as noted in the text, equivalent volumes of each fraction were precipitated with hot 5%trichloracetic acid, and the radioactivity of each precipitate was determined. The results are presentedas percentage, in each fraction, of the total radioactivity of the fractions from each batch of cells. IM,inner membrane; OM, outer membrane.

Microbiology: Davis et al.

Proc. Natl. Acad. Sci. USA 83 (1986)

0min 20 min 30 min

P I O R P I O R P I O R

...

41;.*'W!EF

.L:~~ ~ ~ ~ ~ ~ ~

FIG. 1. Effect of Str on patterns of labeled proteins in subcellularfractions. The cells of Table 1A were labeled with [35S]methionineafter treatment with Str (20 pg/ml) for 20 or 30 min, resulting in 25%or 50% inhibition of protein synthesis, respectively. Labeling andfractionation of these cells, and of parallel untreated cells, were asdescribed in Table 1. The fractions were brought to equal volumesand equivalent quantities were analyzed by NaDodSO4 gel electro-phoresis, followed by autofluorography. The first lane containsmolecular size standards of 92, 68, 55, 43, 29, 25, and 18 kDa. LanesP, I, 0, and R represent periplasmic, inner membrane, outermembrane, and residual fractions, respectively. (The cytoplasmicfractions, which contain the largest amount and variety of proteins,were not informative, and these distracting regions, between P andI, have been cut out.) The arrows indicate the position of alkalinephosphatase.

there were no doubt also misread alkaline phosphatasemolecules of abnormal size, there was not enough of any

0 min 30 min

P R P R

FIG. 2. Effect of Str on distribution of alkaline phosphatase. Theperiplasmic (P) and the residual (R) fractions (0.1 ml) ofthe 0- and the30-min samples from Fig. 1 were treated with rabbit antiserum toalkaline phosphatase (20 Al) and then with Staphylococcus aureuscells as described (25). The precipitates were dissolved in NaDodSO4and aliquots of the various fractions were analyzed by gel electro-phoresis followed by autoradiography.

particular size to be detectable as a band from the immuno-precipitate.These findings show that the misreading induced by Str

prevents much of the secretory protein from reaching itsdestination in the periplasm. In addition, misreading lowersthe solubility ofmuch of the newly formed protein. But whilethe protein that was aborted in secretion was presumablystuck in the membrane, its identification in the residualfraction does not unequivocally localize it in the membrane,since the increment in the residual fraction could includeaggregates of misread cytoplasmic proteins as well as insol-uble membrane proteins. However, the molecular size of thealkaline phosphatase band (Fig. 1) shows that it has beenprocessed from its precursor, and so at least these misreadsecretory molecules have evidently reached the membrane.It therefore seems likely that much of the increment in theresidual fraction represents stuck secretory proteins (andalso altered membrane proteins).

DISCUSSIONEffect of Streptomycin-Induced Misreading on the Membrane.The results presented here show that when Str had reacheda sufficient intracellular concentration in E. coli to causemisreading and partial inhibition of protein synthesis, itimpaired protein secretion, as shown by the decreasedproportion ofthe newly synthesized protein secreted into theperiplasm, and also by inhibition ofthe secretion of a specificperiplasmic protein, alkaline phosphatase (detected by gelelectrophoresis and by immunoprecipitation). In addition, asmight be expected ofgarbled protein, an increased proportionof the new protein was found in a particulate (residual)fraction with low solubility in nonionic detergent, and thatfraction contained a small amount of alkaline phosphatase.When derived from normal cells this low-solubility fraction

has been assumed to consist of membrane fragments, but inthe Str-treated cells, with extensive misreading, it might wellinclude aggregated cytoplasmic proteins. However, since thealkaline phosphatase in the residual fraction from the Str-treated cells has the molecular size of the mature form ratherthan that of its precursor, and since processing takes place inthe membrane, some of the residual fractions must consist ofmisread secretory proteins, aborted in secretion and stuck inthe membrane. In addition, integral membrane proteins,similarly altered in solubility by the misreading, no doubt alsocontribute to the residual fraction.The incorporation of abnormal proteins in the membrane

finally provides a reasonable explanation for the early ob-servation that aminoglycoside action makes cells leaky tosmall molecules (5, 6). It seems unlikely that the membranedamage is the direct cause of cell death, since the degree ofinterference with protein secretion did not increase progres-sively with increasing inhibition by Str (Table 1). Alterna-tively, this evidence for limited membrane damage does fit arole that was suggested 25 years ago, and has been largelydiscarded: that this damage is required for substantial entryof the antibiotic into the cell (5). The misreading would thusplay an indirect but essential role in the bactericidal action ofaminoglycosides, in the following sequence:

(i) The antibiotic penetrates slightly into the cell, by anunknown mechanism (possibly through imperfections due tointrinsic misreading, or at zones of growth). Its contact withchain-elongating ribosomes (the predominant form in grow-ing cells) causes a small degree of misreading.

(it) Some of the misread protein is incorporated into themembrane, where its poor fit creates channels that permitinflux of antibiotic. A rapidly expanding cycle of increasingmisreading and increasing leakiness then ensues.

6166 Microbiology: Davis et al.

Proc. Natl. Acad. Sci. USA 83 (1986) 6167

(iii) The intracellular antibiotic eventually reaches a con-centration that blocks all the initiating ribosomes, thuspreventing further protein synthesis.

(iv) Lethality results from irreversibility of this blockade.The irreversibility of uptake (27), whose mechanism is notwell understood, might account for this property; but thebinding to the initiating ribosomes may also be effectivelyirreversible.

Supporting Evidence for the Model. While previous inter-pretations of aminoglycoside action have left certain aspectsunexplained or contradictory, this proposed mechanism fitsall the features of which we are aware. In particular,extensive studies of the uptake of Str (reviewed in refs. 28 and29) have presented several puzzles that can now be ex-plained. As we have already noted, (i) a lag (whose lengthdecreases with increasing antibiotic concentration) is fol-lowed by a rapid shift to a more or less linear rate of uptake(5); and (ii) reversible inhibition of protein synthesis bychloramphenicol during the lag prevents uptake (5). Inaddition, (iii) once cells have reached the stage of secondaryuptake chloramphenicol no longer prevents uptake (30, 31).(iv) The rate of uptake is roughly proportional to drugconcentration over a wide range (31), as would be expectedof diffusion, or of electrophoresis by the membrane potential(32-34), through aqueous channels. [The role of the mem-brane potential explains why the lag phase as well as thephase of rapid uptake is energy dependent, as Bryan and vanden Elzen (35) have emphasized.] (v) Ribosomal ambiguity(ram) mutations, which increase misreading, not only in-crease sensitivity to Str (36) but also shorten the lag in itsuptake (37). Our model readily explains this effect: themisread proteins formed in these mutants should increasemembrane permeability, thus facilitating the entry of Strduring the lag phase. (vi) In a particularly critical observation,cells resistant to Str but sensitive to another aminoglycoside,gentamicin, took up Str immediately after treatment bygentamicin (30). Findings iii and vi rule out an earlierexplanation for the antagonistic effect of chloramphenicol(35): that the entry of Str to the cytoplasm might be mediatedby binding to sensitive ribosomes engaged in protein synthe-sis at the membrane surface.But perhaps the strongest support for the proposed mech-

anism comes from an old, and particularly mystifying, ob-servation: simultaneous treatment with puromycin acceler-ates killing by Str (38, 39), but pretreatment has the oppositeeffect (38). Even more, Hurwitz et al. (40), comparingmoderate and high concentrations ofpuromycin, have shownthat Str uptake exhibits the same paradoxical response. Ourmodel provides a ready explanation. At high concentrationsof puromycin the prematurely released misread polypeptidechains would be too short to enter the membrane and createchannels, whereas at moderate concentrations many of thechains would be long enough to do so. Moreover, theobservation that moderate concentrations not only allow Struptake but even stimulate it can be explained in two ways: byan increase in the number of the misread chains, resultingfrom premature release; or by an additional effect ofpuromycin, as follows.Membrane Damage by Puromycin. The study of the effect

of puromycin on Str uptake revealed a further surprisingfinding: when exposed simultaneously to Str and to moderateconcentrations of puromycin, a Str-resistant mutant took upStr just as effectively as a sensitive strain. Since Str at theconcentrations used would not cause significant misreadingin the resistant strain, its uptake suggests, by extension ofthemodel proposed in this paper, a remarkable conclusion: thatpolypeptide chains prematurely released by puromycin,without misreading, can create membrane channels, muchlike the misread chains induced by Str. This action wouldreadily account for the stimulatory effect ofpuromycin on Str

uptake. We have obtained preliminary experimental supportfor this hypothesis: treatment of an E. coli culture withpuromycin (50 pkg/ml) resulted in substantial leakage ofA260-absorbing material, almost as much as treatment withStr (unpublished data).The ability of puromycin to produce much the same degree

of membrane damage as Str, without causing cell death, hasseveral interesting consequences. It supports our conclusion(see above) that the membrane damage caused by Str is notthe direct cause of its lethal action. In addition, it mightprovide a useful way of circumventing permeability barriersfor small molecules, in eukaryotic as well as in prokaryoticcells. Finally, puromycin has been widely used to block proteinsynthesis in a variety of cells, and some of the observed effectsthat have been ascribed to that block may in fact be conse-quences of the induction of leakiness of the cell membrane tosmall molecules.Some Implications. It is now clear that the long effort to

eliminate the epiphenomena (11) and to identify "the" keystep in the lethal action of aminoglycosides has been mis-leading: all four steps in the mechanism proposed here areequally essential. Moreover, the evolution of such an intri-cate bactericidal process is intriguing, whether it has beenselected for its direct advantage to the organisms producingaminoglycosides, or whether it is a coincidental effect of asecondary metabolite evolved for some other function.

In addition to shedding light on the problem of aminoglyco-side action, our findings also suggest structural requirementof membrane proteins: not only must they have a sufficientlyhydrophobic surface to be retained in the membrane, but theymust also have been selected in evolution for the ability tofold, and to fit into the membrane, tightly. However, thenotion of a protein fitting badly is vague, and we do not knowthe nature of the resulting channels. These might arise alonga polar region in the predominantly hydrophobic surface ofthe embedded protein, or as a pore within the protein; theymight be formed by many different proteins, or only (assuggested to us by G. Khorana) by transmembrane proteinsthat contain gated channels and lose their gates as a result ofmisreading; and various small molecules might leak throughthe same channel or through different ones. Clearly, althoughproteins carrying random errors revealed the problem, moredetailed analysis will have to be carried out with proteinscarrying known alterations.

We thank Karen Kozarski for expert technical assistance. Thiswork was supported in part by National Institutes of Health GrantsGM 16830 and GM 34766.

1. Davis, B. D., Tai, P. & Wallace, B. J. (1974) in Ribosomes,eds. Nomura, M., Tissieres, A. & Lengyel, P. (Cold SpringHarbor Laboratory, Cold Spring Harbor, NY), pp. 771-789.

2. Tai, P. C. & Davis, B. D. (1985) in Scientific Basis ofAntimi-crobial Chemotherapy, eds. Greenwood, D. & O'Grady, F.(Cambridge Univ. Press, New York), pp. 41-68.

3. Smith, W., Tai, P. C., Thompson, R. C. & Davis, B. D. (1977)Proc. Nati. Acad. Sci. USA 74, 2380-2384.

4. Davis, B. D. & Tai, P. C. (1980) Nature (London) 283,433-438.

5. Anand, N., Davis, B. D. & Armitage, A. K. (1960) Nature(London) 185, 23-24.

6. Anand, N. & Davis, B. D. (1960) Nature (London) 185, 22-23.7. Roth, H., Amos, H. & Davis, B. D. (1960) Biochim. Biophys.

Acta 37, 398-405.8. Dubin, D. T. & Davis, B. D. (1961) Biochim. Biophys. Acta

52, 400-402.9. Dubin, D. T., Hancock, R. & Davis, B. D. (1963) Biochim.

Biophys. Acta 74, 476-489.10. Erdos, T. & Ullmann, A. (1959) Nature (London) 183,

618-619.11. Spotts, C. R. & Stanier, R. Y. (1961) Nature (London) 192,

633-637.12. Speyer, J. F., Lengyel, P. & Basilio, C. (1962) Proc. Nati.

Microbiology: Davis et al.

Proc. Natl. Acad. Sci. USA 83 (1986)

Acad. Sci. USA 48, 684-686.13. Flaks, J., Cox, E. C., Witting, M. L. & White, J. R. (1962)

Biochem. Biophys. Res. Commun. 7, 390-393.14. Mager, J., Benedict, M. & Artman, M. (1962) Biochim.

Biophys. Acta 62, 202-204.15. Gorini, L. & Kataja, E. (1964) Proc. Nati. Acad. Sci. USA 51,

487-493.16. Davies, J., Gilbert, W. & Gorini, L. (1964) Proc. Natl. Acad.

Sci. USA 51, 883-887.17. Chang, F. N. & Flaks, J. G. (1972) Antimicrob. Agents Che-

mother. 2, 294-307.18. Tai, P. C., Wallace, B. J., Herzog, E. L. & Davis, B. D.

(1973) Biochemistry 12, 609-615.19. Wallace, B. J., Tai, P. C., Herzog, E. L. & Davis, B. D.

(1973) Proc. Natl. Acad. Sci. USA 70, 1234-1237.20. Tai, P. C., Wallace, B. J. & Davis, B. D. (1978) Proc. Natl.

Acad. Sci. USA 75, 275-279.21. Wallace, B. J. & Davis, B. D. (1973) J. Mol. Biol. 75, 377-390.22. Bassford, P. J., Silhavy, T. J. & Beckwith, J. R. (1979) J.

Bacteriol. 139, 19-31.23. Davis, B. D. & Mingioli, E. S. (1950) J. Bacteriol. 60, 17-28.24. Oliver, D. & Beckwith, J. (1982) Cell 30, 311-319.25. Horiuchi, S., Marty, D., Tai, P. C. & Davis, B. D. (1983) J.

Bacteriol. 154, 1215-1221.26. Lunn, C. A. & Pigiet, V. (1982) J. Biol. Chem. 257,

11424-11430.

27. Nichols, W. W. & Young, S. N. (1985) Biochem. J. 228,505-512.

28. Hancock, R. E. W. (1981) J. Antimicrob. Chemother. 8,249-276.

29. Hancock, R. E. W. (1981) J. Antimicrob. Chemother. 8,429-445.

30. Holtje, J.-V. (1979) Antimicrob. Agents Chemother. 15,177-181.

31. Muir, M. E., van Heeswyck, R. S. & Wallace, B. J. (1984) J.Gen. Microbiol. 130, 2015-2022.

32. Damper, P. D. & Epstein, W. (1981) Antibiotic Agents Che-mother. 20, 803-808.

33. Mates, S. M., Eisenberg, E. S., Mandel, L. J., Patel, L.,Kaback, H. R. & Miller, M. H. (1982) Proc. Nati. Acad. Sci.USA 79, 6693-6697.

34. Eisenberg, E. S., Mandel, L. J., Kaback, H. R. & Miller,M. H. (1984) J. Bacteriol. 157, 863-867.

35. Bryan, L. E. & van den Elzen, H. M. (1977) Antimicrob.Agents Chemother. 12, 163-177.

36. Rosset, R. & Gorini, L. (1969) J. Mol. Biol. 39, 95-112.37. Ahmad, W., Rechenmacher, A. & Bock, A. (1980) Antimicrob.

Agents Chemother. 18, 798-806.38. White, J. R. & White, H. L. (1964) Science 146, 772-774.39. Yamaki, H. & Tanaka, N. (1963) J. Antibiot. 16, 222-226.40. Hurwitz, C., Braun, C. B. & Rosano, C. L. (1981) Biochim.

Biophys. Acta 652, 168-176.

6168 Microbiology: Davis et al.