OmpF changes and the complexity of Escherichia coli adaptationto prolonged lactose limitation

E Zhang, Thomas Ferenci *Department of Microbiology G08, University of Sydney, Sydney, N.S.W. 2006, Australia

Received 11 May 1999; received in revised form 19 May 1999; accepted 20 May 1999

Abstract

The disaccharide lactose has no specific diffusion pathway across the outer membrane of Escherichia coli. At least threeclasses of spontaneous mutation affecting outer membrane permeability arose with each of three independent E. colipopulations adapting to prolonged lactose limitation in chemostats. Both structural and regulatory mutations affecting OmpFporin predominated in isolates after 210^280 generations of culture. Six types of ompF mutation were found, including in-framedeletions and substitutions at Arg82 and Asp113, all affecting the channel constriction residues of OmpF. Isolates hadincreased susceptibility to antibiotics and were affected in the quantity of OmpF, LamB and OmpA proteins. A minimum ofthree or four mutations was evident in all isolates after 280 generations in a lactose-limited environment, in addition to lacmutations defined in previous studies. ß 1999 Federation of European Microbiological Societies. Published by ElsevierScience B.V. All rights reserved.

Keywords: Outer membrane; ompF mutant; Lactose transport; Continuous culture; Chemostat selection

1. Introduction

Bacterial adaptation studies using nutrient-limitedchemostats have a long history [1]. A classic exampleof Escherichia coli adaptation is the ¢nding that lac-tose limitation provides a strong selection for lac-constitutive mutants in a population [1,2]. Improvedgrowth of E. coli during lactose selection was as-cribed to increased expression of proteins such asthe permease encoded by lacY and L-galactosidaseencoded by lacZ [3,4]. As also seen with other sub-

strates, prolonged nutrient-limited culture of bacteri-al populations leads to mutational adaptation intransport systems and increased e¤ciency for captur-ing or processing the limiting nutrient [5^8].

For the outer membrane permeation of di- or oli-gosaccharides such as maltose, trehalose, sucrose orra¤nose, glycoporins such as LamB, ScrY and RafYare important at low sugar concentration [9]. In con-trast, the uptake of lactose is not particularly en-hanced by sugar-speci¢c porins [9]. It is thereforean open question what, if any, outer membraneadaptations contribute to lactose scavenging underlactose limitation. To test the signi¢cance of lactosepermeation across the outer membrane, we studiedthe changes in outer membrane proteins of E. coliduring prolonged growth under lactose-limited con-

0378-1097 / 99 / $20.00 ß 1999 Federation of European Microbiological Societies. Published by Elsevier Science B.V. All rights reserved.PII: S 0 3 7 8 - 1 0 9 7 ( 9 9 ) 0 0 2 5 7 - 8

* Corresponding author. Tel. : +61 (2) 9351 4277;Fax: +61 (2) 9351-4571;E-mail: t.ferenci@microbio.usyd.edu.au

FEMSLE 8850 2-7-99

FEMS Microbiology Letters 176 (1999) 395^401

ditions. In this paper, we present data on the multi-ple changes, particularly in OmpF structure and reg-ulation, that are consistently selected in E. coli underlactose limitation.

2. Materials and methods

2.1. Bacterial strains and culture conditions

Bacterial strains used in this study are listed inTable 1. Strain G6 was used as the wild-type to in-oculate three parallel long-term chemostats. P1transduction [10] with P1 cml clr100/1000 grown onBW2637 as donor was used to introduce theompR : :Tn10 mutation into various chemostat iso-lates. Detergent and antibiotic susceptibility testswere done using impregnated disks (Oxoid Ltd, Ba-singstoke, UK).

The basal salts medium used in all experimentswas minimal medium A (MMA) [10] supplementedwith amino acids (40 mg l31) when necessary, andsugar (lactose, glucose or maltose) at concentrationsspeci¢ed for each experiment. Luria-Bertani broth,nutrient broth or MMA containing 0.2% (w/v) sugarwas used in batch culture.

Lactose-limited chemostats of 80 ml were set up aspreviously described [11]. The lactose concentrationin the feed medium was 0.02% (w/v). The chemostatswere run at a dilution rate of 0.3 h31. Samples weretaken at weekly intervals from the three parallel pop-ulations labeled A, B and G, by plating onto non-selective nutrient agar. Twelve to 20 randomly sepa-rated colonies were puri¢ed by streaking, numberedfor further testing and stored in glycerol stocks at370³C before further characterisations as describedbelow.

2.2. Outer membrane fractions and electrophoresis

Bacteria were grown in batch culture in 100 mlMMA containing 0.2% lactose to mid-exponentialphase. Membrane extraction and electrophoresiswere performed as previously described [12,13]. Toidentify the position of OmpF, OmpC and LamB ingels, protein pro¢les of MH513 (ompF-lacZ),MH225 (ompC-lacZ) and G6 grown in MMA con-taining 0.2% maltose were included as controls.

2.3. Estimation of LamB levels and assay ofL-galactosidase activity

The starch-Sepharose binding assay was used tomeasure changes in LamB expression and bindinga¤nity. Retention of bacteria in columns is depend-ent on the number of LamB channels on the surfaceas well as the intrinsic binding a¤nity [14]. Covalentcoupling of starch to oxirane-activated Sepharoseand the assay procedure were described previously[15]. Assay of L-galactosidase activities was per-formed according to published procedures [10] usingsodium dodecyl sulfate (SDS)- and chloroform-treated cells.

2.4. Mutation analysis

PCR ampli¢cation of the ompF gene used twoprimers: Omf4 (5P-TCCCTGCTCTGTTAGTAG-3P) and Omr5 (5P-CGTCGTACTTCAGACCAG-3P). The reaction pro¢le consisted of 34 cycles of:denaturation at 94³C for 30 s, followed by annealingat 50³C for 30 s, and extension at 72³C for 1.5 min ina DNA thermal cycler (Perkin-Elmer Cetus). PCRproducts were puri¢ed directly with the WizardPCR Preps DNA puri¢cation system (PromegaCorp., Sydney, Australia). The nucleotide sequencewas determined using the above primers and dye-terminator sequencing reactions (Applied Biosystems377 sequencer).

3. Results

3.1. Properties of lactose-adapted populations after280 generations of selection

Wild-type E. coli K-12 was grown in chemostatswith limiting lactose at a dilution rate of 0.3 h31 for4 weeks (280 generations). Twenty randomlystreaked colonies from the culture were tested eachweek for the properties shown in Fig. 1. To test ifnon-speci¢c outer membrane permeability increased,susceptibility of isolates to antibiotics and SDS wasfollowed (Fig. 1A). More than 80% of isolates after140 generations became more susceptible to penicil-lin. In addition, the number of isolates that weresusceptible to cloxacillin and SDS increased rapidly

FEMSLE 8850 2-7-99

E Zhang, T. Ferenci / FEMS Microbiology Letters 176 (1999) 395^401396

from week 2 to week 4. Similar trends were found intwo other independent populations, but with slightlyvaried time courses.

A second property analysed in Fig. 1B was an

assay for the LamB glycoporin activity of bacteria[14] and was used to test the role of LamB in adapt-ing to limiting lactose. There was a transient increasein LamB-dependent binding after 3 weeks of culture,but the increase was not maintained. This trend wasalso found in another independent population andsuggests some LamB change occurred but was notthe most permanent factor in adapting to limitinglactose. Analysis of the mutations responsible forthe activities in Fig. 1 are described below.

3.2. Analysis of outer membrane protein pro¢les fromchemostat isolates

To test for major changes in outer membrane pro-teins, the composition of 15 randomly sampled indi-vidual chemostat isolates was analysed from twopopulations after 210 generations (A and G). Elec-trophoresis in an SDS-polyacrylamide gel containing8 M urea resolved changed outer membrane proteinpatterns in all 15 isolates. Isolates G31 and G42provide examples of these changes (Fig. 2). In G31and in the majority of other isolates, OmpF becamemore prominent relative to OmpC and particularlyto OmpA. A two-fold increase in LamB was alsofound in four out of 15 analysed isolates, includingboth G31 and G42. This result correlated with theincrease in starch binding shown in Fig. 1B.

A more striking pattern change was evident in 10of the 15 isolates, as represented by G42 in Fig. 2.OmpF disappeared from its normal position and adi¡erent band was evident. This result raised thepossibility that OmpF changed mobility in G42, asdescribed previously for some OmpF structural mu-tants [17]. Indeed, when the ompR gene, which reg-ulates the expression of OmpC and OmpF porins[16], was rendered non-functional in G42, bothOmpC and the altered band disappeared. As withG31, the proportion of proteins in G42 was alteredrelative to wild-type and OmpA decreased.

Table 1Bacterial strains used in this study

Strain Genotype Origin or reference

G6 Hfr G6 his [22]BW2637 G6 trpE ompR : :Tn10 [14]MH225 MC4100 x(ompC-lacZ+)10-25 [23]MH513 MC4100 araD+x(ompF'-lacZ+)16-13 [23]

Fig. 1. Adaptive changes of chemostat bacteria in a lactose-lim-ited E. coli population. A: Weekly samples from week 1 to week4 in chemostat A were plated onto nutrient agar and 20 ran-domly separated colonies were tested for susceptibility to anti-biotic and SDS by a disc di¡usion assay with: cloxacillin (5 Wg);penicillin (10 units) and SDS (750 Wg). B: The LamB-dependentstarch binding activity from samples was assayed with starch-Se-pharose columns as previously described [15].

FEMSLE 8850 2-7-99

E Zhang, T. Ferenci / FEMS Microbiology Letters 176 (1999) 395^401 397

3.3. Analysis of ompF changes in lactose-limitedcultures

The ¢rst one-third of the ompF gene from 51 che-mostat isolates was sequenced, since all reported mu-tations involving the pore function were mapped tothis part of the gene [17]. Six types of mutations wereobtained in the 51 isolates sequenced from three in-dependent populations (Table 2). An in-frame dele-tion of 18 bp removed residues 117^122 of OmpF inisolates from one population (B). A di¡erent in-frame deletion of 12 bp (removing residues 116^119) was found in two other populations (A andG). Residues Arg82 and Asp113 were also a¡ectedin several populations. Population A also containedisolates with a double mutation involving Arg82 toSer as well as the 12-bp deletion. All these changesa¡ected residues at the channel constriction ofOmpF and potentially improve outer membrane

permeability through increased OmpF channel size[18].

Interestingly, the increased antibiotic susceptibilityin Fig. 1A was only partly associated with the OmpFchanges. Indeed, isolates like G50 without ompF mu-tations also had enhanced susceptibility to penicillinindependent of porin structure or regulation (Table3). The susceptibility of the ompF mutants was testedwith and without inactivated ompR leading to ab-sence of OmpF. The strains with Arg82 substitutionshad increased susceptibility to penicillin and chlor-amphenicol but no signi¢cant di¡erence in suscepti-bility to SDS and cloxacillin. But the penicillin sus-ceptibility increase was not due to OmpF, since thelack of OmpF in the ompR mutant derivative ofA422 and G31 did not modify susceptibility. In con-trast, the Asp113 mutation led to increased suscept-ibility to penicillin, chloramphenicol as well as SDSin an ompF-dependent manner. The deletions in

Fig. 2. Outer membrane protein pro¢les of isolates from lactose-limited chemostats. The pro¢le of outer membrane proteins from chemo-stat isolates was resolved in a SDS-polyacrylamide gel containing 8 M urea as previously described [13]. The protein pro¢le from thestarting strain, G6, was used as control. For comparison, the protein pro¢les from MH513 (ompF-lacZ) and MH225 (ompC-lacZ) whichwere grown in Luria-Bertani broth were included to show normal positions of OmpC and OmpF bands; the protein pro¢le labeled asG6(M) from G6 grown in MMA containing 0.2% maltose batch culture was included to show the maltose-induced LamB band. The lanewith G42(OmpR: :Tn10) is the pro¢le of isolate 42 after insertion of the ompR mutation. All other outer membrane proteins were ex-tracted from bacteria grown on MMA containing 0.2% lactose in batch culture.

Table 2ompF mutations in chemostat isolates

Mutation Amino acid replacement andremoved residues

Found in population(number of isolates)a

Found after howmany generations

Deletion of 18 bases (AGAATTTGGTGGTGATAC)

v(Glu117, Phe118, Gly119, Gly120,Asp121 and Thr122)

B (3) 140

Deletion of 12 bases (CCAGAATTTGGT) v(Pro116, Glu117, Phe118 and Gly119) A (8), G (9) 210Arg82 (CCA) Serine A (6), G (5) 210Arg82 (GCC) Proline G (8) 210Asp113 (ACG) Glycine A (1) 210Double mutation Arg82 (CCA) and deletionof 12 bases (CCAGAATTTGGT)

Serine v(Pro116, Glu117, Phe118and Gly119)

A (7) 210

aThe number in parentheses represents the number of times the mutation was found in each chemostat population.

FEMSLE 8850 2-7-99

E Zhang, T. Ferenci / FEMS Microbiology Letters 176 (1999) 395^401398

ompF increased susceptibility to all the compoundstested. As previously found with other ompF mutants[17], deletions within the selectivity loop L3 consti-tute a bigger channel than point substitution mutants[18].

The di¡erent types of ompF mutation present inthe populations provide an explanation for the tem-poral shift in antibiotic susceptibility patterns in Fig.1A. Since deletions in ompF resulted in susceptibilityto cloxacillin and SDS as shown in Table 3, thecloxacillin/SDS pattern in Fig. 1A can be explainedby the takeover of the population by ompF deletionmutants. Indeed, deletions became common only inthe fourth week of selection. Amongst week 3 iso-lates from population A, 6/12 sequenced were ompFpoint mutants and 6/12 were deletion mutants ordouble mutants. By week 4, 1/12 was a point mutantand 9/12 were deletion or double mutants. This se-quential sweep from point mutants to deletion mu-tants also appeared in population G where 11/12week 3 isolates sequenced were point mutants but8/12 week 4 isolates were deletion mutants.

As noted above, ompF-related mutations were notthe only source of penicillin susceptibility in chemo-stat isolates. The unidenti¢ed mutation present inG50 (as well as in isolates with ompF mutations inTable 3) was clearly prevalent in the populations.

The presence of this unidenti¢ed mutation helps toexplain the early increase in penicillin susceptibilityin Fig. 1A, since penicillin susceptibility was presentin more than 80% of isolates by week 2 but only 35%of these isolates contained ompF mutations. Indeed,the ompF mutation in isolates like A422 or G42probably arose in a clone already having the low-level penicillin susceptibility.

Other unidenti¢ed mutation(s) caused complexregulatory adaptations in the amount of outer mem-brane proteins noted above for G42 and other iso-lates (Fig. 2). The mutational change a¡ecting lamBunder lactose limitation may or may not be distinctfrom those causing changes in OmpF/C/A expres-sion, but not all mutants altered in OmpF expressionhad changed LamB levels. Indeed 5/15 isolates withincreased OmpF had reduced LamB expression and6/15 had no signi¢cant LamB change. Despite thiscomplexity, it was evident that lactose selectionchanged OmpF porin in at least two additive ways,by increasing the amount of OmpF as well as struc-tural change in the OmpF protein.

4. Discussion

At least three changes commonly occurred in lac-

Table 3Susceptibilitya of the ompF isolates to various compounds

Source strain OmpF change ompR : :Tn10 mutation Compound

Cloxa Penb Cmlc SDSd

G6 none (W.T.) 3 0 0 18 0A422 Arg82CSer 3 0 11 25 0A422 Arg82CSer + 0 9 23 0G31 Arg82CPro 3 0 8 26 0G31 Arg82CPro + 0 9 22 0A323 Asp113CGly 3 0 16 26 9A323 Asp113CGly + 0 8 26 9G42 v116C119 3 12 16 23 14G42 v116C119 + 0 16 23 0A411 v116C119, Arg82CSer 3 10 16 23 12A411 v116C119, Arg82CSer + 0 10 21 0B31 v117C122 3 14 16 23 7B31 v117C122 + 0 0 20 0G50 none 3 0 8 20 0G50 none + 0 9 21 0

aSusceptibility is expressed as the size (mm) of the zone of growth inhibition. The amount of antibiotic in the applied disks was: (a)Clox, cloxacillin, 5 Wg; (b) Pen, penicillin, 10 units ; (c) Cml, chloramphenicol, 30 Wg; and (d) SDS, sodium dodecyl sulfate, 750 Wg.

FEMSLE 8850 2-7-99

E Zhang, T. Ferenci / FEMS Microbiology Letters 176 (1999) 395^401 399

tose-limited cultures a¡ecting the outer membrane.In addition, it is expected that types of adaptationpreviously demonstrated, namely ampli¢cationevents [2,4] and changes in lacY [19], also occurunder these selection conditions. An example fromeach of two populations indicates the complexity ofthe changes after 280 generations of lactose-limita-tion selection. In each case, at least three mutationswere accumulated to provide the level of ¢tnessneeded to compete successfully in the chemostat en-vironment.

In an example from population G, isolate G42contained the following changes: (i) a mutation caus-ing low-level penicillin susceptibility (ompF-inde-pendent), (ii) an in-frame ompF deletion, (iii) a reg-ulatory mutation increasing OmpF levels and (iv) amutation increasing LamB levels (though this may ormay not be the same as (iii)). In another population,isolate A411 contained: (i) a mutation causing low-level penicillin susceptibility, (ii) an ompF point mu-tation and (iii) an ompF deletion, probably arisingafter the point mutation. It needs stressing thatboth isolates may contain additional mutations notde¢ned in this study. The polygenic changes illus-trate the mutational plasticity of the bacterial ge-nome in the face of strong selections. It also high-lights how complex reality is in contrast toevolutionary models in which one or two compo-nents are considered on the path to ¢tness [20].

The types of OmpF structural change that evolvedunder lactose limitation were similar to those ob-tained by Benson et al. [17], who isolated mutantsusing OmpF for transporting large maltodextrins.Both lactose selection and maltodextrin selectionyielded substitutions at the same residues (Arg82and Asp113) near the channel constriction ofOmpF [18]. The short deletions obtained in the che-mostat populations removed part of the selectivityloop L3, as was also observed in similar but notidentical mutants obtained by maltodextrin selection.The antibiotic and SDS susceptibility data presentedin Table 3 and the channel data presented in [21]indicate a broadly based increase in permeability ofthe mutated channels.

The increased proportion of OmpF protein in theouter membrane as a result of lactose-limitation se-lection was a common occurrence in most isolates.Hence both quantity and quality of OmpF were sub-

ject to selection. The nature of the complex changein OmpF, OmpC and OmpA levels was not resolved,and whether one mutation was the source of thesechanges is also unclear. Either regulatory mutationsor ampli¢cation events could increase ompF expres-sion relative to other outer membrane proteins.

These results with lactose limitation were quitedi¡erent from results with glucose limitation, inwhich the regulation of LamB expression was theprime target in increasing glucose permeability inseveral independent populations [8]. The currentstudy is consistent with the view that lactose is arelatively poor substrate for LamB as previouslyconcluded [9] and non-speci¢c pathways are the pri-mary target for selection to improve outer membranepermeability.

Acknowledgements

We thank Lucinda Notley-McRobb for expert ad-vice and assistance and the Australian ResearchCouncil for providing ¢nancial support.

References

[1] Dykhuizen, D. and Hartl, D. (1983) Selection in chemostats.Microbiol. Rev. 47, 150^168.

[2] Novick, A. and Horiuchi, T. (1961) Hyper-production of beta-galactosidase by Escherichia coli bacteria. Cold Spring HarborSymp. Quant. Biol. 26, 239^245.

[3] Dean, A.M. (1989) Selection and neutrality in lactose operonsof Escherichia coli. Genetics 123, 441^454.

[4] Dykhuizen, D.E., Dean, A.M. and Hartl, D.E. (1987) Meta-bolic £ux and ¢tness. Genetics 115, 25^31.

[5] Helling, R.B., Vargas, C.N. and Adams, J. (1987) Evolutionof Escherichia coli during growth in a constant environment.Genetics 116, 349^358.

[6] Sonti, R.V. and Roth, J.R. (1989) Role of gene duplications inthe adaptation of Salmonella typhimurium to growth on limit-ing carbon sources. Genetics 123, 19^28.

[7] Notley-Mcrobb, L. and Ferenci, T. (1999) Adaptive mgl-reg-ulatory mutations and genetic diversity evolving in glucose-limited E. coli populations. Environ. Microbiol. 1, 33^44.

[8] Notley-Mcrobb, L. and Ferenci, T. (1999) The generation ofmultiple coexisting mal-regulatory mutations through poly-genic evolution in glucose-limited populations of E. coli. En-viron. Microbiol. 1, 45^52.

[9] Ulmke, C., Lengeler, J.W. and Schmid, K. (1997) Identi¢ca-tion of a new porin, RafY, encoded by ra¤nose plasmidpRSD2 of Escherichia coli. J. Bacteriol. 179, 5783^5788.

FEMSLE 8850 2-7-99

E Zhang, T. Ferenci / FEMS Microbiology Letters 176 (1999) 395^401400

[10] Miller, J. (1972) Experiments in Molecular Genetics. ColdSpring Harbor Laboratory, Cold Spring Harbor, NY.

[11] Death, A., Notley, L. and Ferenci, T. (1993) Derepression ofLamB protein facilitates outer membrane permeation of car-bohydrates into Escherichia coli under conditions of nutrientstress. J. Bacteriol. 175, 1475^1483.

[12] Lugtenberg, B., Meijers, J., Peters, R., van der Hoek, P. andvan Alphen, L. (1975) Electrophoretic resolution of the `majorouter membrane protein' of Escherichia coli K12 into fourbands. FEBS Lett. 58, 254^258.

[13] Liu, X.Q. and Ferenci, T. (1998) Regulation of porin-medi-ated outer membrane permeability by nutrient limitation inEscherichia coli. J. Bacteriol. 180, 3917^3922.

[14] Heine, H.G., Francis, G., Lee, K.S. and Ferenci, T. (1988)Genetic analysis of sequences in maltoporin that contributeto binding domains and pore structure. J. Bacteriol. 170,1730^1738.

[15] Ferenci, T. and Lee, K.S. (1982) Directed evolution of thelambda receptor of Escherichia coli through a¤nity chromato-graphic selection. J. Mol. Biol. 160, 431^444.

[16] Pratt, L.A., Hsing, W.H., Gibson, K.E. and Silhavy, T.J.(1996) From acids to OsmZ ^ multiple factors in£uence thesynthesis of the OmpF and OmpC porins in Escherichia coli.Mol. Microbiol. 20, 911^917.

[17] Benson, S.A., Occi, J.L. and Sampson, B.A. (1988) Mutations

that alter the pore function of the OmpF porin of Escherichiacoli K12. J. Mol. Biol. 203, 961^970.

[18] Lou, K.L., Saint, N., Prilipov, A., Rummel, G., Benson, S.A.,Rosenbusch, J.P. and Schirmer, T. (1996) Structural and func-tional characterization of OmpF porin mutants selected forlarger pore size. 1. Crystallographic analysis. J. Biol. Chem.271, 20669^20675.

[19] Tsen, S.D., Lai, S.C., Pang, C.P., Lee, J.I. and Wilson, T.H.(1996) Chemostat selection of an Escherichia coli mutant con-taining permease with enhanced lactose a¤nity. Biochem.Biophys. Res. Commun. 224, 351^357.

[20] Dykhuizen, D.E. and Dean, A.M. (1990) Enzyme activity and¢tness: evolution in solution. Trends Ecol. Evol. 5, 257^264.

[21] Saint, N., Lou, K.L., Widmer, C., Luckey, M., Schirmer, T.and Rosenbusch, J.P. (1996) Structural and functional char-acterization of OmpF porin mutants selected for larger poresize. 2. Functional characterization. J. Biol. Chem. 271,20676^20680.

[22] Hat¢eld, D., Hofnung, M. and Schwartz, M. (1969) Geneticanalysis of the maltose A region in Escherichia coli. J. Bacter-iol. 98, 559^567.

[23] Hall, M.N. and Silhavy, T.J. (1981) The ompB locus and theregulation of the major outer membrane porin proteins ofEscherichia coli K12. J. Mol. Biol. 146, 23^43.

FEMSLE 8850 2-7-99

E Zhang, T. Ferenci / FEMS Microbiology Letters 176 (1999) 395^401 401