Helical Disposition of Proteins and Lipopolysaccharide in the Outer Membrane of Escherichia coli

In bacteria, several physiological processes once thought tobe the products of uniformly dispersed reactions are now knownto be highly asymmetric, with some exhibiting interesting geometriclocalizations. In particular, the cell envelope of Escherichiacoli displays a form of subcellular differentiation in whichpeptidoglycan and outer membrane proteins at the cell polesremain stable for generations while material in the lateralwalls is diluted by growth and turnover. To determine if materialin the side walls was organized in any way, we labeled outermembrane proteins with succinimidyl ester-linked fluorescentdyes and then grew the stained cells in the absence of dye.Labeled proteins were not evenly dispersed in the envelope butinstead appeared as helical ribbons that wrapped around theoutside of the cell. By staining the O8 surface antigen of E.coli 2443 with a fluorescent derivative of concanavalin A, weobserved a similar helical organization for the lipopolysaccharide(LPS) component of the outer membrane. Fluorescence recoveryafter photobleaching indicated that some of the outer membraneproteins remained freely diffusible in the side walls and couldalso diffuse into polar domains. On the other hand, the LPSO antigen was virtually immobile. Thus, the outer membrane ofE. coli has a defined in vivo organization in which a subfractionof proteins and LPS are embedded in stable domains at the polesand along one or more helical ribbons that span the length ofthis gram-negative rod.

INTRODUCTION

Top

Introduction

The idea that bacteria might be no more than small sacks ofenzymes and substrates mixed together in an amorphous jumblehas given way to the realization that many proteins, processes,and macromolecules have geometrically diverse locations andexhibit asymmetric behaviors. One example of this changing conceptis that in Escherichia coli the cell poles can be consideredto be distinctive organelles in which neither peptidoglycannor outer membrane (OM) proteins are recycled or diluted (9,10, 12, 26). Numerous proteins or physiological complexes congregateexclusively at one or both poles, with implications for proteinsecretion, environmental sensing, cellular development, virulence,and gene transfer (14, 19, 31, 54, 55). Furthermore, one ormore of these asymmetries may regulate bacterial morphologybecause inaccurate localization of extraneous pole-like domainsproduces cells with aberrant shapes (12, 44).

Another surprising facet of prokaryotic structure is that severalproteins assemble in long filaments that spiral across the innerface of the cytoplasmic membrane, reaching from one end of thecell to the other. The MreB and Mbl proteins form such fibrils,or cables, in Bacillus subtilis (22) and Caulobacter crescentus(17) and may serve as internal scaffolds that position peptidoglycan-synthesizingenzymes, thereby defining the shape of the cell wall (17, 29,33). The MinCD proteins oscillate from pole to pole along helicalpaths, inhibiting polymerization of the principal cell divisionprotein FtsZ and helping to define the cell's center (56). FtsZitself moves from completed septa to new division sites by travelingalong helical pathways on the inner membrane (1, 58), and evennonassembled FtsZ monomers oscillate back and forth along helicaltracks (60). DNA dispersal at cell division also relies on proteinswith helical distributions: the SetB protein is required forproper chromosomal segregation in E. coli (16), and the ParAprotein moves along spiral-shaped threads to segregate R1 plasmids(13). And finally, the SecA and SecY secretion complexes aredistributed in helical fashion in the inner membrane of B. subtilis(4), although this may not be true in E. coli (3). Except forthe Sec proteins, the other spiraling proteins seem to sharethe common function of partitioning some substance—shapeelements (MreB and Mbl), cell volume (MinCD and FtsZ), or DNA(SetB and ParA)—so that the materials are distributeduniformly to daughter cells after division.

One question that arises is whether one of the above proteinscreates the principal scaffold on which the others assembleor travel, or if there is an underlying helical structure inwhich proteins move independently. A recent observation hintsat the importance of the latter possibility. Gibbs et al. foundthat in E. coli newly synthesized LamB protein diffuses throughoutthe OM in a helical fashion (20). This behavior implies theexistence of long-lived spiral boundaries that restrict thediffusion of LamB, and perhaps other surface proteins, to definedOM "canals." However, the OM has been considered to be continuousand homogeneous, and no obvious structure has been observedthat might serve such a purpose. On the other hand, polar domainsconstitute an exception to the idea of the uniformity of theOM (9, 12). Since an unknown mechanism maintains this distinctionbetween the poles and the cell cylinder, another overlookedstructure(s) may exist in the cell envelope as well.

In this work we ask whether components of the OM have geometricdistributions different from that expected to occur by randomdispersal. When either protein or lipopolysaccharide (LPS) waslabeled with a fluorescent compound and then chased by growingcells in the absence of label, stained material was organizedin stable helical swaths. These data, combined with previousobservations, suggest that the gram-negative cell envelope hasa stable and innate helical property that includes all of itscomponents: peptidoglycan, proteins, and lipids.

MATERIALS AND METHODS

Top

Materials and Methods

Bacteria, plasmids, and general techniques. The bacterial strains used in this study are listed in Table1. Plasmid pBMM1 (34) was constructed by ligating the 2.123-kbEcoRI-HindIII fragment containing the res-npt-res kanamycinresistance-encoding cassette from pCK155 (28) into the EcoRI-HindIIIsites of pBCSK (Stratagene, La Jolla, Calif.). Bacteria weregrown in Luria-Bertani (LB) broth, on LB agar plates, or inM9 minimal glucose medium (37) supplemented with 40 µgof amino acids (L-threonine, L-leucine, L-arginine, L-lysine,and L-histidine) per ml. Where appropriate, antibiotics wereadded to the following concentrations: ampicillin, 100 µg/ml;chloramphenicol, 20 µg/ml; kanamycin, 50 µg/ml;tetracycline, 25 µg/ml. When indicated, aztreonam, a specificinhibitor of penicillin binding protein 3, was added to culturesat a final concentration of 5 µg/ml. Unless otherwisespecified, chemicals and reagents were purchased from SigmaChemical Co. (St. Louis, Mo.).

View this table:
[in this window]
[in a new window]
TABLE 1. E. coli strains used in this study

Deletion of the hisI gene. The chromosomal hisI gene was deleted and replaced with a kanamycinresistance cassette from plasmid pBMM1. pBMM1 was digested withEcoRI and HindIII, and a 2.5-kb DNA fragment was purified. Withthis fragment as a template, the kanamycin resistance cassettewas amplified by using a pair of primers, the 3' ends of whichconsisted of a 20-nucleotide sequence complementary to eitherthe 5' or 3' end of the npt gene. To each of these was addeda 49-nucleotide sequence complementary to chromosomal sequencesadjacent to the hisI gene. The complete primer sequences were5'-GCA AAG TCA CCT TCT TCT CGC GCA CTA AAC AGC GAC TGT GGA CCAAAG GCG TGG GCG AAG AAC TCC AGC-3' and 5'-CGG CAG ATT TGC GCTCGG CGA GCA GTT GTT CCA GTT GAT ACA GGA ACA GCG TGC TAA TGTGGT TAC GTG-3'. PCR amplification was initiated by a singlemelting step of 94°C for 5 min, followed by 30 amplificationcycles (94°C for 1 min, 65°C for 1 min, 72°C for1 min), concluding with a single incubation at 72°C for7 min. The resulting 1.5-kb PCR fragment was purified by usinga gel extraction kit (QIAGEN Inc., Valencia, Calif.) and electroporatedinto E. coli KM32, where the chromosomal hisI gene was replacedwith the antibiotic cassette via ${lambda}$ Red recombination (41). Recombinantswere selected by plating on LB medium plus kanamycin and screenedfor the inability to grow in histidine-deficient minimal medium.The presence of the hisI deletion in strain KM32I-9 was confirmedby diagnostic PCR. This hisI::Kan^r allele and the neighboringrfb locus from E. coli K-12 were cotransduced into E. coli 2443by P1 transduction (36) to create strain AG430-2K, which formedrough colonies and lacked a complete O antigen.

LPS preparation. LPS was extracted by the method of Hitchcock and Brown (21),with minor modifications. Cells were incubated in 1.5 ml ofLB medium overnight at 37°C, harvested by centrifugation,washed with 1 ml of 0.1 M phosphate-buffered saline (PBS; pH7.4), and resuspended in 1 ml of PBS. To this suspension wasadded 50 µl of lysis buffer consisting of 1 M Tris-HCl(pH 6.8) containing 2% sodium dodecyl sulfate (SDS), 4% ß-mercaptoethanol,10% glycerol, and bromophenol blue. Cells were lysed by boilingfor 10 min, proteinase K was added to a final concentrationof 500 µg/ml, and the samples were incubated at 60°Cfor 60 min, mixed well, and centrifuged at 14,000 x g to removecell debris. Samples (10 µl) were loaded onto a 12% polyacrylamidegel (iGels; Gradipore Ltd., Frences Forest, Australia), andLPS was separated by electrophoresis at a 35-mA constant current.Gels were stained with the Pro-Q Emerald 300 LPS gel stain kit(Molecular Probes Inc., Eugene, Oreg.) in accordance with themanufacturer's instructions. A commercial LPS preparation fromE. coli serotype O55:B5 (Sigma Chemical Co.) was loaded ontoeach gel as a control.

Labeling of OM proteins. E. coli OM proteins were labeled with succinimidyl ester (NHS)-Texasred or NHS-Alexa-Fluor 488 (NHS-AF488) (Molecular Probes, Inc.)as previously described (9, 12). In brief, bacteria were incubatedat 37°C in LB broth to an optical density at 550 nm (OD₅₅₀)of 0.7, washed twice with sodium bicarbonate buffer (0.1 M,pH 8.0), and stained with NHS-Texas red (final concentration,100 µg/ml) or NHS-AF488 (50 µg/ml). Stained cellswere washed three times with bicarbonate buffer, diluted inLB broth, and incubated at 37°C with or without aztreonamin the absence of the dye. Samples were withdrawn after varioustime intervals, fixed with 0.5% formalin, and visualized byphase and fluorescence microscopy.

ConA labeling of LPS O8 polysaccharide. Bacterial cultures were grown in LB medium to an OD₅₅₀ of 0.7,harvested by centrifugation, washed twice with 0.1 M PBS, andresuspended in one-third volume of PBS. Concanavalin A (ConA)-conjugatedAlexa-Fluor 488 (ConA-AF488; Molecular Probes, Inc.) was dissolvedin PBS (pH 7.4) to give a 5-mg/ml stock solution. ConA-AF488was added to the cell suspension to give a final concentrationof 200 µg/ml, and the mixture was incubated in the darkat room temperature for 30 min with gentle shaking. Stainedcells were washed twice and resuspended in PBS for microscopy.For chase experiments, a 250-µl volume of stained cellswas inoculated into 10 ml of LB medium and incubated at 37°Cin the dark and aliquots were withdrawn at intervals for microscopy.

Double labeling. Proteins and LPS in the same cells were labeled in a two-stepprocedure. E. coli 2443 was grown to an OD₆₀₀ of $~$ 0.7, and 3ml of cells was pelleted, washed twice with 0.1 M sodium bicarbonatebuffer (0.1 M, pH 8.0), and resuspended in 1 ml of the samebuffer. NHS-AF488 was added to a final concentration of 50 µg/ml,and the mixture was incubated for 30 min at room temperaturein the dark. The cells were then washed twice with 0.1 M PBS,pH 7.4, and either ConA-AF488 or unlabeled ConA (ICN BiomedicalInc., Irvine, Calif.) was added to a final concentration of200 µg/ml. The suspension was incubated at room temperaturein the dark for 30 min, washed twice with 0.1 M PBS (pH 7.4),and subjected to fluorescence microscopy or fluorescence recoveryafter photobleaching (FRAP).

Labeling of inner membrane. The inner membrane was stained with the dye FM4-64 as previouslydescribed (18). Briefly, cells were incubated at 37°C inLB broth to an OD₅₅₀ of 0.7, harvested by centrifugation, washedtwice with 0.1 M PBS (pH 7.4), and resuspended in one-thirdvolume of PBS. FM4-64 (Molecular Probes, Inc.) in dimethyl sulfoxidewas added to give a final concentration of 4 µg/ml, andthe mixture was incubated in the dark at room temperature for30 min with gentle shaking. Stained cells were washed twicewith PBS and resuspended in PBS for microscopy. For chase experiments,250 µl of stained cells was inoculated into 10 ml of LBmedium with or without aztreonam and incubated in the dark at37°C and aliquots were withdrawn at 30, 60, or 90 min formicroscopy.

Microscopy. Cell suspensions (5 µl) were placed on microscope slidescoated with 1% agarose and incubated at room temperature for10 min to immobilize the cells. Light microscopy was performedby viewing cells with a 100x oil immersion objective on a NikonOPTIPHOT-2. AF488-labeled and Texas red-labeled compounds weredetected by using either a fluorescein isothiocyanate filter(470-nm excitation, 530- and 550-nm emission) or a rhodaminefilter (595-nm excitation, 615-nm emission), respectively (ChromaTechnology Corp., Rockingham, Vt.). Images were captured witha black-and-white Pixera Penguin 600CLM cooled charge-coupleddevice camera controlled by Image-Pro Plus software (Media Cybernetics,Silver Spring, Md.).

Confocal microscopy and FRAP. FRAP was performed on cells labeled with either NHS-AF488 (forE. coli CS109 or 2443) or ConA-AF488 (for E. coli 2443). Overnightcultures in LB medium were diluted 1:100 into 10 ml of LB mediumand incubated for 1 h, after which aztreonam was added (to 5µg/ml) and the cells were grown for about four generationsto an OD₅₅₀ of 0.7. Cells to be labeled with ConA-AF488 wereharvested by centrifugation and washed with 0.1 M PBS (pH 7.4);those to be stained with NHS-AF488 were washed with 0.1 M sodiumbicarbonate buffer (pH 8.0). Each wash buffer contained aztreonam(5 µg/ml). At this point, the cells were labeled witheither ConA-AF488 (200 µg/ml) or NHS-AF488 (50 µg/ml)and incubated for 30 min at room temperature in the dark withgentle shaking. Labeled cells were washed twice with the respectivebuffers containing aztreonam, and 2 µl of cell suspensionwas spotted onto the surface of 0.3% agar squares of LB medium-aztreonamin Lab-Tek chamber slides (Nalge Nunc International, Naperville,Ill.) or, in some cases, onto poly-lysine-coated slides. A coverslipwas applied, and the cells were observed by confocal microscopy.In some experiments, labeled cells were chased by incubatingdilutions in LB medium or PBS (pH 7.2) and microscopy and FRAPanalysis were performed on samples harvested at 30-min intervalsover the course of 2.5 h.

Confocal microscopy was performed with a microscope (model LSM510Meta; Zeiss, Jena, Germany) fitted with a 100x 1.45 N.A. plan-fluarobjective and driven by Zeiss AIM software. The procedure wasa modified version of the method described by Cowan et al. (6).Two to five prebleach images of each microscopic field werephotographed, after which a small portion of a cell filamentwas circumscribed for bleaching. The size of the bleach areawas set so that its diameter was $~$ 30% greater than the widthof the filament. Each cell segment was bleached by exposingthe area for 20 to 50 iterations (2 to 4 s total) to the 488-nmline of an argon laser (30-mW maximum power, 6.1-A tube current).Fluorescence recovery was documented by photographing the cellevery 5 s for 5 to 10 min after photobleaching. Images werevisualized with the Zeiss LSM image browser, deconvolution wasperformed with AutoDeblur Gold CF 9.3.2 and the AutoVisualizeprogram (Autoquant, Watervliet, N.Y.), and image processingand analysis of fluorescence recovery were performed with theImage J program (National Institutes of Health, Bethesda, Md.).For the latter analyses, the average background value of a cell-freearea was subtracted from each image.

RESULTS

Top

Results

Some proteins are arranged in helical ribbons in the OM. Cell poles in wild-type E. coli and the poles and deformitiesin shape-defective mutants retain stable patches of OM proteinsafter prolonged growth (9, 12). Fluorescence light microscopyeasily detected labeled proteins in these relatively large regions,but there were hints that the cells also contained threadlikestrings of stable proteins in their lateral walls as well. Totest this possibility more rigorously, we used confocal microscopyto follow the fate of OM proteins.

Proteins in the OMs of E. coli CS109 and 2443 were labeled withNHS-Texas red or NHS-AF488, after which the cells were incubatedin the absence of the stain and visualized by confocal microscopy.As expected, proteins labeled by either dye remained in stablepatches at the poles after the chase period, but in addition,broad helical strings of stable material also appeared in thelateral walls (Fig. 1). In all of the cells examined, thesehelices were embedded in a light background of evenly distributedfluorescence (Fig. 1, upper row), implying that some proteinswere free to diffuse while others were immobile. The helicaldistribution was visible by simple inspection but was most easilydiscerned by assembling a series of z stack photographs takenin successive planes and then sharpening the pictures by deconvolution(Fig. 1, lower row). These steps effectively subtracted thehazy background contributed by uniformly distributed proteins.The stable structures appeared to consist of at least two intertwiningribbons, and although many helices seemed more prominent nearthe poles, most filaments exhibited helical distributions alongtheir entire length (Fig. 1). We conclude that at least someE. coli OM proteins are arranged in stable helical structuresthat persist in the face of diffusion and insertion of new material,indicating that these proteins are either immobile or poorlymobile.

View larger version (21K):
[in this window]
[in a new window]
FIG. 1. Pulse-chase visualization of OM protein helices in E. coli. OM proteins were labeled with NHS-Texas red or NHS-AF488. Stained cells were grown at 37°C in the absence of label, harvested, and visualized by confocal microscopy. In some cases, cells were forced to filament before and during the chase by adding aztreonam, which inhibits penicillin binding protein 3, an essential septation protein. Cells are presented as the original confocal picture or as processed by deconvolution. (A) E. coli CS109 elongated with aztreonam for 120 min, stained with NHS-Texas red, and chased for 120 min. (B) Nonelongated E. coli CS109 stained with NHS-Texas red and chased for 120 min. (C) E. coli CS109 elongated with aztreonam for 60 min, stained with NHS-Texas red, and chased for 120 min. (D) E. coli 2443 elongated with aztreonam, stained with NHS-AF488, and chased for 60 min.

There are no inherent diffusion barriers in the OM or inner membrane. One explanation for stable protein groupings in the OM is thatsuch patches arise because they are trapped in lipid domainsthat are so tightly sequestered from one another that diffusionbetween domains is restricted or prevented. We tested this propositionby labeling OM proteins with NHS-AF488 and using FRAP to measureprotein diffusion in the OMs of living cells. In every case,the bleached areas recovered fluorescence within 10 to 40 s(Fig. 2A to C and 3A to C), indicating that proteins in neighboringportions of the membrane were mobile. Fluorescence recoveryoccurred not only in the lateral walls but also at the poles(Fig. 2C and 3C), although the rate and extent of recovery mighthave been very slightly lower (compare Fig. 3C to Fig. 3A and B).The results indicate that no mechanical obstructions preventedprotein diffusion into broad areas of the wall or poles. Inparticular, proteins diffused uniformly into the bleached areasof elongated cells and no helical structures were reconstituted(Fig. 2A to C), suggesting that the diffusible proteins werenot part of stable helices.

View larger version (34K):
[in this window]
[in a new window]
FIG. 2. FRAP analysis of protein and LPS diffusion in the OM. E. coli was filamented by treatment with aztreonam for 1 to 2 h, after which OM proteins were stained with NHS-AF488 (A to C) or the LPS O8 antigen was stained with ConA-AF488 (D). A segment of each filament was photobleached by laser exposure and photographed every 5 s for several minutes. The results for three time points are shown here for individual experiments. For each set, the left image is just prior to photobleaching, the center image is just after photobleaching, and the right image is 1 to 5 min after bleaching. As shown in Fig. 3, the extent of fluorescence recovery occurred quickly and was approximately equivalent at any time from 0.5 to 10 min after photobleaching. (A to C) E. coli CS109 proteins stained with NHS-AF488. (D) E. coli 2443 O8 antigen stained with ConA-AF488.

View larger version (11K):
[in this window]
[in a new window]
FIG. 3. Rates of FRAP for OM proteins and LPS. Panels A to D represent the time course of fluorescence recovery for the photobleached segments of cells in Fig. 2A to D. The steep drop in fluorescence denotes the time of laser exposure and photobleaching. The extent of fluorescence recovery remained stable from 3 min to over 10 min (not shown).

Although there were no impenetrable barriers to diffusion withinthe OM, it was possible that such barriers exist in the innermembrane. To test this, we stained cells with the lipophilicdye FM4-64, which preferentially stains the inner membrane (18),and chased the label by growing the cells in the absence ofdye. If diffusion was mechanically restricted by variation inlipid composition or behavior, then a portion of the dye shouldbe trapped in domains limited to the poles or helical formationsin the side walls, whereas fluorescence in the rest of the cellwould be diluted by growth. A uniform distribution of FM4-64would indicate that diffusion was unimpeded by natural lipidbarriers. The latter result was observed (Fig. 4). All cellsdisplayed uniform fluorescence around their peripheries, indicatingthat there were no barriers to diffusion among inner membranedomains over the poles or in the lateral walls.

View larger version (25K):
[in this window]
[in a new window]
FIG. 4. Pulse-chase visualization of inner membrane stained with the lipophilic dye FM4-64. The inner membrane of E. coli CS109 was labeled with FM4-64, and stained cells were grown in the presence of aztreonam at 37°C for 60 min in the absence of label, harvested, and visualized by confocal microscopy. Panels: A, confocal image; B, image after processing by deconvolution.

Specific labeling of LPS O antigen. The outer leaflet of the OM of gram-negative bacteria is composedof LPS (48, 49). Even though proteins diffused freely in thisenvironment, the LPS itself could be distributed in one of twoways. LPS might simply be an inert, randomly diffusing matrixin which the immobile protein helices were embedded, or elsea portion of the LPS might also be stable and immobile. To distinguishbetween these possibilities, we labeled the O-antigen side chainof E. coli 2443 with ConA-AF488, a fluorescent lectin that bindsspecifically to ${alpha}$ -glucose and ${alpha}$ -mannose residues of the O8 antigenpresent in this strain (5, 8).

E. coli 2443 labeled with ConA-AF488 exhibited an evenly distributedfluorescence around the cell periphery (Fig. 5A). The interactionbetween dye and cells was quite stable: ConA was not washedaway by repeated rinsing with 0.1 M PBS (pH 7.4), and cellsretained their fluorescence after being incubated for 90 minin PBS in the absence of free ConA-AF488 (not shown). Stainingintensity decreased only when labeled cells were stored for7 days in the dark in the presence of 0.5% formaldehyde. Incontrast, the E. coli K-12 derivatives CS109, JM105, and KM32did not retain ConA-AF488 after washing (Fig. 5B and Table 1).These results were consistent with the fact that K-12 strainsdo not express a complete O antigen (30), which we confirmedby observing O chains in the LPS profile of E. coli 2443 (Fig.5C, lane 2) but not in E. coli CS109 (Fig. 5C, lane 3) or KM32(Table 1).

View larger version (52K):
[in this window]
[in a new window]
FIG. 5. Labeling and detection of LPS O8 antigen with ConA-AF488. (A and B) E. coli was stained with ConA-AF488, washed once in buffer lacking the dye, and examined by phase-contrast or fluorescence microscopy. Panels: A, E. coli 2443 (O8 antigen positive); B, E. coli CS109 (O-antigen-negative K-12 strain); C, LPS profiles of O-antigen-positive and -negative strains. LPS from E. coli was separated by SDS-PAGE as described in Materials and Methods and visualized with the Pro-Q Emerald LPS stain. Lanes: 1, control LPS preparation from E. coli serotype O55:B5; 2, LPS from E. coli 2443 (O8 antigen positive); 3, LPS from E. coli CS109 (O-antigen-negative K-12 strain). The series of evenly spaced bands represents variously sized O-antigen side chains in E. coli 2443 (lane 2) that are absent in E. coli CS109 (lane 3).

To assure ourselves that the ConA-directed staining characteristicsof E. coli 2443 reflected only the presence or absence of theO8 antigen, we moved the rfb genes of E. coli KM32 (a K-12 strain)into E. coli 2443 by phage P1 cotransduction with a closelylinked kanamycin resistance marker, creating strain AG430-2K.To be certain that no additional mutations had been created,the rfb_K_-₁₂ genes were moved by a second cotransduction fromAG430-2K into E. coli 2443T, creating strain AGTO2-1K. BothAG430-2K and AGTO2-1K lost the ability to synthesize O8-antigenside chains, as determined by SDS-polyacrylamide gel electrophoresis(PAGE) (not shown), and neither was stained by ConA-AF488 (Table1), indicating that replacing the rfb_O8 genes with their K-12variants eliminated ConA binding.

The inability of K-12 strains to bind ConA seems at odds withobservations that ConA agglutinates K-12 strains (45, 46) andimmobilizes E. coli JM105 to lectin-derivatized membranes (15).However, a small number of molecules may agglutinate or immobilizewhole cells, so such results do not prove that ConA binds uniformlyto the OM. Consistent with the results presented here, Stoitsovaet al. did not observe binding of gold-labeled ConA to E. coliK-12 (57). Thus, ConA may bind with low avidity to carbohydrateresidues other than those of the O side chain, as is suggestedby the lectin's ability to agglutinate E. coli serotypes havingdifferent O antigens (32). In any event, the ability to retaina substantial, visibly fluorescent ConA-linked dye was limitedto cells expressing the O8 antigen, which contains repeatedunits of ${alpha}$ -mannose (5).

Helical distribution of stable LPS. Having shown that ConA-AF488 specifically stains the O8 antigenof E. coli 2443, we used this reagent to determine how LPS wasorganized in the OM. When cells were labeled with ConA-AF488and grown for two to four generations in the absence of thelectin, stable patches of fluorescently labeled O8 antigen wereobserved at the cell poles and in helical ribbons along thelateral walls (Fig. 6). Although helical strings could be observedin single cells by confocal microscopy (Fig. 6A), the LPS heliceswere visualized more sharply in filamentous cells by deconvolutinga set of z stack images (Fig. 6B and C). The results indicatethat a subset of LPS molecules in the OM were not diluted byinsertion of new material. Instead, they were immobilized ina specific geometric arrangement on the outer surface of thecell.

View larger version (14K):
[in this window]
[in a new window]
FIG. 6. Pulse-chase visualization of OM LPS helices in E. coli 2443. The O8 antigen of LPS in the OM was labeled with ConA-AF488. Stained cells were grown at 37°C in the absence of label, harvested, and visualized by confocal microscopy. Panels: A, original confocal image of cells stained with ConA-AF488 and then chased for 60 min in the absence of label; B, confocal and deconvolved images of cells fila-mented for 30 min with aztreonam, stained with ConA-AF488, and then chased for 60 min in the absence of label; C, deconvolved confocal image of a cell filamented by expression of SulA for 60 min, stained with ConA-AF488, and then chased for 60 min in the absence of label. The two images represent the full length of a single filamentous cell.

Previous experiments established that a relationship existsbetween the cell division initiator FtsZ and the synthesis ofinert patches of peptidoglycan (10, 62). To determine if FtsZpolymerization contributed to the formation of the long-livedLPS domains, cells were elongated by expressing the FtsZ inhibitorSulA (40, 61). When SulA-induced cell filaments of E. coli 2443were labeled with ConA-AF488 and chased, the same helical ribbonsof stable O8 antigen were observed (Fig. 6C). Thus, generationof LPS helices did not require FtsZ polymerization.

The LPS helices appeared similar in extent and pitch to thestable protein helices in cells labeled with NHS-linked dyes.Although it seems reasonable that the two helices representa single structure, the bleed-over emission of the bright Alexa-Fluordye into the red emission channel prevented us from determiningdirectly if the LPS and protein helices were coincident.

ConA-labeled LPS does not diffuse in the OM. One explanation for the appearance of LPS helices might be thatthe OM is composed of discrete "canals" that are kept separatefrom one another. LPS might freely diffuse in any one canalbut be unable to cross into neighboring domains, thereby givingthe illusion of stable swaths of LPS. We tested this idea bysubjecting ConA-AF488-stained filaments to FRAP analysis. In23 cases in which cells filaments were labeled but not chased,the bleached portion of the uniformly stained filament did notregain fluorescence even after 30 to 40 min of incubation (Fig.2D and 3D and data not shown). When FRAP analysis was performedon cell filaments that had been labeled and chased so that LPShelices were visible, the same result was observed—LPSfrom the helices did not diffuse into the bleached areas (notshown). Thus, in contrast to OM proteins, ConA-labeled LPS didnot diffuse freely within the OM.

ConA binding does not inhibit protein diffusion. It was possible that ConA binding might alter the diffusionproperties of the OM. We tested this by labeling OM proteinsof E. coli 2443 with NHS-AF488, binding unlabeled ConA to theO8 antigen on the surface of these same cells, and observingprotein diffusion by FRAP. Labeled proteins returned to bleachedareas at the same rate regardless of the presence or absenceof ConA bound to LPS (not shown), indicating that ConA labelingdid not significantly alter protein diffusion in the OM.

DISCUSSION

Top

Discussion

Bacteria are highly organized and architecturally complex. Severalcytoplasmic proteins orient themselves and move along intricatehelical pathways in the cytoplasm or along the inner face ofthe cytoplasmic membrane, and the cell poles appear to be organelle-likedomains consisting of long-lived patches of peptidoglycan andmembrane proteins. To these patterns we now add the observationthat LPS and proteins form stable domains not only at the polesbut also along one or more helical ribbons within the OM ofE. coli. Thus, this part of the envelope is not homogeneousbut instead exhibits subtle structural variations. The cell'semphasis on creating privileged sites and populating them withspecific proteins suggests that these elaborate structures servefundamental purposes, so it is important to understand the mechanismsresponsible for their existence. The present results have implicationsfor the in vivo behavior of LPS, the distribution of OM proteins,and envelope synthesis.

LPS stability versus diffusion. The inability of ConA-labeled LPS to diffuse into bleached areasof the OM raises the question of why LPS should exhibit suchstability. An extensive literature supports the existence ofstrong bonds between neighboring LPS molecules (42, 43). Forexample, in artificial membrane vesicles stable islands of LPSpersist for days in a "sea of phospholipid," indicating theexistence of extremely durable lateral interactions betweenLPS molecules (59). Also, hexagonal arrays of tightly packedLPS subunits are observed in electron micrographs of crystallizedLPS (24, 64) and by X-ray crystallography and atomic force microscopy(25). Four forces generate this high degree of organization:van der Waals attractions between fatty acid chains, bridgingof LPS molecules by divalent cations, hydrogen bonding, andinteractions between the O polysaccharides of neighboring molecules(2, 23, 27, 42).

Several groups have rationalized these observations by three-dimensionalmolecular modeling. In the model of Kastowsky et al., the structureof the LPS inner and outer cores are rigid and well definedbut the O side chain is flexible and can adopt many conformations(23). The most energetically favorable predicts that the O polysaccharideis bent into an "L" shape, folded over and lying parallel tothe membrane surface, and in contact with the head groups ofneighboring LPS molecules or membrane-embedded proteins (23).The authors concluded that the variety of intermolecular contactswould allow the O polysaccharide to self-assemble as "a mechanicallystable, feltlike network" (23). A similar conclusion is supportedby modeling that indicates that a single LPS molecule, viewedfrom above, is shaped like a triangular wedge or pie slice (64).Six such molecules can self-associate with their points facinginward to form a circular hexamer, and multiple hexamers canform a regular hexagonal lattice held together by divalent cations(27, 64). Brandenburg and Wiese reemphasize the importance ofthese interactions, although they believe that the aqueous gapsbetween the LPS hexamers might not be favored in vivo (2). Overall,several experimental and computational reasons support the ideathat the bonds between neighboring LPS molecules are strongand directional. It should be remembered, however, that thesestructural studies and modelings were performed on membranesof pure LPS when, in fact, the OM is full of proteins whosepresence would interrupt any strictly regular geometric arrangementof an LPS-only leaflet.

None of the above considerations indicates whether LPS moleculesdiffuse freely among themselves. Schindler et al. addressedthis question directly by incorporating exogenous rhodamine-labeledLPS into the OM of Salmonella enterica serovar Typhimurium (53).FRAP analysis in cell filaments indicated that LPS diffusedfreely with no immobile components (53). Mühlradt et al.reached essentially the same conclusion by using antibodiesto determine the fate of newly inserted LPS (39). Both of theseresults are the opposite of what we report here, in that wesaw little or no LPS diffusion into bleached areas in FRAP experiments.However, significant differences in the experimental approachesused may explain the disparities. Schindler et al. (53) labeledLPS with rhodamine isothiocyanate, which cross-links to amineslocated in the inner core of LPS. Addition of the bulky aromaticdye probably disrupts important hydrogen bonds, and the accompanyingsteric hindrance may prevent bridging by divalent cations (23).These alterations would seriously impair LPS-LPS interactionsso that the labeled compound would diffuse more readily thannormal. In fact, Schindler et al. (53) calculated an LPS diffusionrate that is $~$ 1,000 times faster than that determined by Mühlradtet al., who performed their experiments with wild-type cellsexpressing an LPS with longer O-antigen side chains (38, 39).The latter group labeled LPS with antibody and inferred theextent of LPS dispersal by using electron microscopy. They concludedthat LPS diffuses at least 10,000 times less rapidly than doother membrane phospholipids (38). Even so, Gibbs et al. recentlyproduced evidence that the OM protein LamB is distributed differentlyin live cells than in fixed cells, which they ascribe to redistributioncaused by chemical fixation (20). Susceptibility to this artifactsuggests that LPS may diffuse even more slowly in vivo thanthe already low rate reported by Mühlradt et al.

The preceding rate estimates may be sufficient to explain whywe observed little or no LPS diffusion. However, an additionalconsideration may apply to our results. Because ConA existsas tetramers at pH >7.0 (8), the multivalent lectin mightcross-link neighboring molecules so that native LPS-LPS interactionsare enhanced by the labeling system itself. The strength ofthese combined stabilizing forces could reduce or prevent normalLPS diffusion. On the other hand, even after labeling with ConA,LPS-LPS interactions were not so strong as to inhibit insertionof new material during pulse-chase experiments or to inhibitprotein diffusion. Thus, any ConA-enhanced LPS interaction wasneither overpowering nor permanent. Therefore, although we cannotconclude that LPS is completely immobile, its natural diffusionrate must be quite small.

Diffusion of OM proteins and polar stability. Unlike LPS, at least some proteins diffused freely in the OMto repopulate bleached areas. This diffusion was not affectedappreciably by binding of unlabeled ConA to the O8 antigen,so even if the lectin did cross-link neighboring LPS moleculesit did not impact protein movement. Because proteins in helicalstructures were stable over several generations, it seems thatOM proteins must be grouped into mobile and immobile populations.We cannot tell if the populations are composed of differentproteins or if the immobile group is a subpopulation of thewhole.

One place where this mobile-immobile dichotomy is very clearand potentially informative is at the cell poles. Peptidoglycanand the overlying LPS and OM proteins are retained at the poleseven after several generations of growth (9, 12). Nonetheless,OM proteins diffused into and out of these polar regions ata rate equal to or only very slightly less than the rate atwhich they diffused throughout the side wall. This is consistentwith previous observations that the IcsA protein, which is secretedto the OM specifically at the poles, moves from there to coverthe entire bacterial surface (50, 52). The conclusion must bethat there is no impenetrable barrier blocking protein diffusionbetween the poles and side walls. The most likely alternativefor explaining protein stability at the poles is that some proteinsare tethered to one or more polar components, with inert peptidoglycanbeing the most prominent candidate. Noncovalent peptidoglycanbinding domains, such as that found in OmpA (63), and covalentattachments, as exemplified by Braun's lipoprotein (47), mightcontribute the necessary anchors to immobilize one or more proteins.

Helical structure of the OM. The pulse-chase experiments with labeled proteins and LPS revealeda new structural element in the gram-negative envelope. Despitethe free diffusion of bulk OM proteins, chased cells accumulatedlong-lived ribbons of protein wrapped around the cell in a helicalmanner. Similarly, swaths of ConA-labeled LPS maintained long-livedassociations over several generations of growth, indicatingthat the molecules were neither diluted by insertion of newmaterial nor removed or recycled.

A simple way to approach the question of how such protein andLPS helices survive is to envision an "insertion barrier" thatprevents addition of new molecules into these ribbons. One possibilityis that during the chase period LPS is inserted between ConA-labeledmolecules but then, as sufficient numbers of unlabeled moleculesaccumulate, new material is inserted preferentially betweenunlabeled LPS molecules. If ConA-enhanced LPS-LPS interactionstilt the balance of insertion toward non-ConA-linked domains,then the surviving helices would be artifacts generated by amechanism that indirectly preserves swaths of ConA-labeled material.If this is true, then the ConA-labeled ribbons would reflectthe geometry of the initial stages of OM synthesis, implyingthat insertion of new envelope material runs along one or morehelical tracks.

An alternate explanation is that the helical ribbons of proteinor LPS represent fixed domains that are naturally impervious,so that insertion of new material occurs only between ribbons.Addition of newly synthesized proteins or LPS would dilute labeledmolecules in the insertion zones, creating darker areas as thecell grows, but would not dilute label in the impervious zones,leaving the observed ribbons. If this interpretation is correct,then the helical structures would represent permanent componentsof the OM. In this case, synthesis of most of the OM might ormight not proceed via a helical mechanism; only synthesis ofthe inert portion of the envelope would have to be helical.For the rest of the envelope, LPS insertion into the OM couldbe diffuse and random within broad zones between the inert boundaries,so that any departure from complete randomness might be so slightas to be overlooked by previous techniques. If these zones werebounded by persistent helical ribbons, diffusion of lipids orproteins could be confined to spiral canals that loop aroundthe OM, possibly accounting for the helical diffusion of LamB(20).

As discussed above in relation to cell poles, a possible mechanicalexplanation for the stability of OM helices is that one or morecomponents are riveted to an underlying spiral arrangement ofinert cell wall, producing a tightly linked sandwich of peptidoglycan,LPS, and proteins that resists attempts to insert new material.In this scenario, the geometry of the ConA- or NHS-labeled ribbonswould reflect the underlying distribution of inert peptidoglycan.By evaluating the distribution of newly inserted peptidoglycanprecursors, de Pedro et al. produced images suggesting thatolder peptidoglycan is distributed as helical ribbons in theside wall of E. coli (11). Also, peptidoglycan is incorporatedinto the cell wall of B. subtilis in helical tracks (7), althoughit is not known whether these are stable over time. Althoughthese findings are consistent with the idea that the three envelopecomponents could be interconnected, it is not technically feasibleto visualize whether the inert peptidoglycan helices are coincidentwith the LPS and protein helices we observe. In summary, theobserved distribution of stable LPS ribbons could be the productof an underlying helical mechanism that governs either LPS insertionor peptidoglycan synthesis and envelope attachment.

O-antigen labeling. As a final technical note, the use of fluorescently labeledConA to screen for the presence, absence, or distribution ofthe O8 antigen should be useful for other applications as well.ConA labeling represents a simple in vivo screen for the presencein E. coli and Salmonella species of O antigens that contain ${alpha}$ -mannose (51), and lectins with different specificities shouldbe useful for monitoring the presence of alternate O-polysaccharideside chains.

ADDENDUM IN PROOF

Top

Addendum in proof

After this paper was accepted, we were able to disentangle thefluorescent emissions of the two dyes in double-labeled cellsby spectral imaging of samples observed by confocal microscopy.The results demonstrated that the stable outer membrane helicesof Texas red-stained proteins and Alexa-Fluor-stained LPS werecoincident with one another, implying that the proteins andLPS are components of a common helical structure in the outermembrane (B. M. Meberg and K. D. Young, unpublished data).

ACKNOWLEDGMENTS

We especially thank Tamara J. Casavan for instruction and helpin operating the confocal microscope and John Lee for assemblingthe graphics.

This work was supported by grant GM061019 from the NationalInstitutes of Health.

FOOTNOTES

* Corresponding author. Mailing address: Department of Microbiology and Immunology, University of North Dakota School of Medicine and Health Sciences, Grand Forks, ND 58202-9037. Phone: (701) 777-2624. Fax: (701) 777-2054. E-mail: kyoung{at}medicine.nodak.edu.

Present address: Department of Biotechnology, Indian Instituteof Technology, Kharagpur 721302, West Bengal, India.

REFERENCES

Top