Outer Membrane Targeting, Ultrastructure, And Single Molecule Localization of the Enteropathogenic Escherichia Coli Type IV Pilus Secretin BfpB

Outer Membrane Targeting, Ultrastructure, and Single MoleculeLocalization of the Enteropathogenic Escherichia coli Type IV PilusSecretin BfpB

Joshua A. Lieberman,a Nicholas A. Frost,b Michael Hoppert,c Paula J. Fernandes,a* Stefanie L. Vogt,d Tracy L. Raivio,d

Thomas A. Blanpied,b and Michael S. Donnenberga

Division of Infectious Diseases, Department of Medicine, University of Maryland School of Medicine, Baltimore, Maryland, USAa; Department of Physiology and Program inNeuroscience, University of Maryland School of Medicine, Baltimore, Maryland, USAb; Institut fuer Mikrobiologie und Genetik, Universitaet Goettingen, Goettingen,Germanyc; and Department of Biological Sciences, University of Alberta, Edmonton, Alberta, Canadad

Type IV pili (T4P) are filamentous surface appendages required for tissue adherence, motility, aggregation, and transformationin a wide array of bacteria and archaea. The bundle-forming pilus (BFP) of enteropathogenic Escherichia coli (EPEC) is a proto-typical T4P and confirmed virulence factor. T4P fibers are assembled by a complex biogenesis machine that extrudes pili throughan outer membrane (OM) pore formed by the secretin protein. Secretins constitute a superfamily of proteins that assemble intomultimers and support the transport of macromolecules by four evolutionarily ancient secretion systems: T4P, type II secretion,type III secretion, and phage assembly. Here, we determine that the lipoprotein transport pathway is not required for targetingthe BfpB secretin protein of the EPEC T4P to the OM and describe the ultrastructure of the single particle averaged structures ofthe assembled complex by transmission electronmicroscopy. Furthermore, we use photoactivated localizationmicroscopy todetermine the distribution of single BfpBmolecules fused to photoactivated mCherry. In contrast to findings in other T4P sys-tems, we found that BFP components predominantly have an uneven distribution through the cell envelope and are only foundat one or both poles in a minority of cells. In addition, we report that concurrent mutation of both the T4bP secretin and the re-traction ATPase can result in viable cells and found that these cells display paradoxically low levels of cell envelope stress re-sponse activity. These results imply that secretins can direct their own targeting, have complex distributions and provide feed-back information on the state of pilus biogenesis.

Enteropathogenic Escherichia coli (EPEC) is an important causeof pediatric infectious diarrhea throughout the developingworld (26, 32, 43). Typical EPEC strains carry a large EPEC adher-ence factor (EAF) plasmid that encodes a type IV pilus (T4P), thebundle-forming pilus (BFP) (73). The BFP is a confirmed viru-lence factor (8) and adhesin that mediates the initial stages ofadherence to the host intestinal epithelium (40). The expression ofBFP is associated with a distinctive pattern of three-dimensionalclusters when incubated with cells, termed localized adherence(67), and the formation of dynamic autoaggregates of bacteria inliquid culture (3). T4Ps mediate diverse cellular processes in abroad range of bacteria, including adhesion and colonization (15,25, 86), twitching and social motility (10, 41), and horizontal genetransfer (5, 36).

T4Ps are long, thin, flexible homopolymeric three-start helicalfibers approximately 85 in diameter (21, 60). The fibers arecomposed of the pilin structural protein (21) and are both assem-bled anddisassembled by complex biogenesismachines consistingof 10 to 18 proteins (57, 58). T4Ps are common to many Gram-negative pathogens, including Vibrio cholerae (34), Pseudomonasaeruginosa (10), Neisseria meningitidis (14), Francisella tularensis(15), Legionella pneumophila (72), and EPEC (25). T4Ps are sub-divided into T4aP and T4bP based on key differences in the pilinmonomer and genetic organization (51). T4b pilins generally havea largermore complex structure, a longer leader sequence, and areN methylated on small hydrophobic residues instead of the phe-nylalanine typical for T4a pilins. The genes encoding T4bP bio-genesis machines are generally contiguous, while those for T4aPmachine components are found in unlinked operons. Further-

more, T4bP predominate in enteric bacteria and mediate aggre-gation phenotypes more often than the T4aP systems (51).

T4P machinery proteins share significant sequence similarityand structural homology to components of type II secretion (T2S)systems, DNA uptake systems (5, 57), and filamentous phage as-sembly systems (45, 65) and even have orthologues in proteinsinvolved in archaeal flagellum assembly (57). Furthermore,T4Ps have recently been detected in Gram-positive pathogensof the genus Clostridium (79, 80) and in archaea (28, 53). Thesequence and structural similarities across such a wide range oforganisms strongly suggest an ancient and shared evolutionaryhistory (57, 58).

All T4Ps consist of a mature pilin that is cleaved and, in Gram-negative bacteria, Nmethylated by a dedicated pre-pilin peptidase(74). Gram-negative T4P biogenesis machines contain a core setof conserved proteins thought to assemble into a supramolecularbiogenesismachine that spans both the innermembrane (IM) andthe outer membrane (OM) (39). These proteins include a poly-topic IMprotein, at least one cytoplasmic nucleotide binding pro-

Received 6 October 2011 Accepted 5 January 2012

Published ahead of print 13 January 2012

Address correspondence to Michael S. Donnenberg, [email protected].

* Present address: Global Scientific Solutions for Health, Baltimore, Maryland, USA.

Supplemental material for this article may be found at http://jb.asm.org/.

Copyright 2012, American Society for Microbiology. All Rights Reserved.

doi:10.1128/JB.06330-11

1646 jb.asm.org 0021-9193/12/$12.00 Journal of Bacteriology p. 16461658

tein energizing pilus dynamics, pre-pilin-like proteins, and theOMsecretin protein (21, 57, 58). Thewell-conserved secretin pro-teins are not only essential for T4P biogenesis but also form theOM ring structure found in T2S systems (9, 64), type 3 secretion(T3S) systems (6), and the filamentous phage assembly system(45). Secretins are reported to form homomultimers of 12 to 15monomers (9, 18, 42, 55, 64).

Nonlipoprotein secretins require a small lipoprotein pilotin, oranother small lipoprotein, for targeting to the OM and multi-merization, while some secretins are lipoproteins themselves anddo not require a pilotin (42, 54). True pilotin proteins, such asPulS of theKlebsiella oxytocaT2Smachine, are acylated and trans-ported to the OM by the Lol-sorting pathway (17) and are re-quired formultimer stability (6). A recent report (76) showed thatlipidation of theN. meningitidis lipoprotein, PilW, is required forboth the stability of PilW and the efficient assembly of the PilQsecretin complex, although PilW does not appear to be a bona fidepilotin such as PulS. The Lol pathway acylates an N-terminal cys-teine of lipoproteins following signal peptide cleavage and trans-port from the cytosol by the Sec apparatus (77). By default, the Lolmachinery transports lipoproteins to the OM, unless an asparticacid, glycine, proline, or aromatic residue is present at the secondN-terminal residue constituting a Lol-avoidance signal (70, 85).Unlike nonlipoprotein secretins with cognate pilotins, the BfpBsecretin of EPEC is itself a lipoprotein and is palmitoylated in vivo(61), and no pilotin can be identified in the BFPmachine compo-nents (68). HxcQ is the T2S secretin of Pseudomonas aeruginosaand one of the few other secretins experimentally demonstrated tobe a lipoprotein (81). Lipidation of HxcQ is required for proteinfunction and transport to the OM, suggesting that lipoproteinsecretins can self-pilot (81). However, the role of Lol in transport-ing lipoprotein secretins has not been determined (61, 81).

Secretins play a crucial role in T4P biogenesis machines; with-out a functional secretin multimer T4P biogenesis fails, a pheno-type that manifests in EPEC as the loss of autoaggregation in thebfpB deletion mutant (4, 61). Secretins are thought to serve as theexit pore for pilus fibers as they extend and, in some systems suchas BFP, retract (6, 19, 42). To successfully transport substrates,secretins interact with other system components. In the case of theBFP system, BfpB interacts with two soluble proteins in theperiplasm, BfpU and BfpG, and recruits these proteins to the OM(23). Although the exact function of BfpU is unknown, it is re-quired for the function of BFP machines; BfpU is likely not aspecific chaperone since its absence does not alter expression, pro-cessing, or localization of the pilin (69).

In many T4P systems, including the BFP of EPEC (30), as wellas those of Neisseria gonorrhoeae (83) and N. meningitidis (14),pilus retraction is an important step in pilus function and is drivenby a dedicated hexameric ATPase (3, 62). In the case of the N.gonorrhoeae T4Ps, deletion of genes encoding both the secretin(pilQ) and the retraction ATPase (pilT) resulted in cell toxicityand membrane blebbing, which was attributed to pili assemblingwithout an exit pore (84). Curiously, a separate study found that apilQ pilT double mutant strain of N. meningitidis was viable, de-spite the presence of intraperiplasmic pilus fibers (14). It is unclearwhat accounts for these species-specific differences. The effects ofsimultaneous mutations in the secretin and retraction ATPase inT4bP systems have not been studied.

The subcellular localization of T4P and T2S components hasbeen the subject of some debate. Although T4Ps in Myxococcus

xanthus and Pseudomonas aeruginosa have been shown by trans-mission electronmicroscopy (TEM) andfluorescencemicroscopyto exit the cell primarily at the pole (13, 20, 36, 56), the case ismuch less clear for other T4Ps. Moreover, Cowles and Gitai (20)detectedT4Pswith nonpolar origins inP. aeruginosa. In the case ofBFP fibers, TEMhas not revealed an exit location perhaps becauseof the extensive, overlapping, and complex meshwork formed bythese interacting fibers. Studies with fluorescent fusion proteinshave been complicated by gene dosage effects. Lybarger et al. (47)described polar localization patterns when T2S proteins were ex-pressed in trans but peripheral foci of fluorescence when ex-pressed in their native stoichiometry from their wild-type locus.Buddelmeijer et al. (12) observed a similar effect of the PulS pilo-tin on the localization of the PulD secretin in the T2S of Klebsiellaoxytoca. Furthermore, all localization studies of T4P or T2S com-ponents with fluorescent fusion proteins to date have been re-stricted in resolution by the limits of diffraction, and no previousstudies of T2S systems and T4P protein localization studies haveimaged single fluorescent molecules. Given the small sizes of bac-teria, diffraction limited resolution and bulk excitation of fluores-cent probes dramatically limit the ability to localize the pilus bio-genesis machine and its components. Recent advances influorescent imaging have made subdiffraction limit imaging pos-sible (7, 35). These so-called super-resolution techniques revealunprecedented detail of bacterial cell biology.

In these studies we attempted to reconcile some of the unre-solved questions related to the BfpB secretin of the EPEC T4bP,including its basic architecture, the role of the Lol sorting systemin targeting BfpB to the OM, the effect of simultaneousmutationseliminating the secretin, and the retraction ATPase and the sub-cellular localization of the protein. We used a combination ofgenetic and functional techniques and applied photoactivationlocalization microscopy (PALM) to capture the distribution ofsingle molecules of BfpB in fixed cells.

MATERIALS AND METHODSStrains, plasmids, and growth conditions.The strains and plasmids usedin the present study are listed in Table 1. Bacterial strains were cultured inLuria-Bertani broth (LB) at 37C, except for ALN92, which was grown at30C. In EPEC strains, BFP was expressed as previously described (69) bygrowing strains in Dulbecco modified eagle medium (DMEM) lackingphenol red. Antibiotics were added at the following concentrations toselect for or maintain plasmids: ampicillin, 200 g ml1; chlorampheni-col, 20 g ml1; and kanamycin, 50 g ml1.

UMD946 was constructed by deleting codons 121 to 128 (the WalkerA box) of bfpF, replacing these residues with a scar sequence using themethod of Datsenko and Wanner (24), PCR template pKD4, and theprimers Donne664 and 665 (Table 2). The mutation was confirmed bysequencing and complementation to confirm that the mutation is non-polar. UMD947 was created by electroporating into UMD946(pKD4) the5.6-kb BamHI fragment from the nonpolar bfpB insertion mutantUMD923, containing bfpB sequence flanking the kanamycin resistancegene and replacing the native bfpB gene by one-step inactivation. No FRTsites are present in this fragment, and there was no further recombination,leaving this strain resistant to kanamycin.

Electron microscopy. Carbon support films approximately 10 to 15nm thick were prepared by indirect sublimation of carbon onto freshlycleavedmica. Homogeneous preparations of BfpB were diluted in 10mMTris-HCl buffer (pH 7.0) to a final concentration between 10 and 30g/ml. The samples were then prepared for electronmicroscopy by using4% (wt/vol) uranyl acetate as a negative staining solution essentially asdescribed previously (78). The carbon film was partially floated off the

Fine Localization of BfpB

April 2012 Volume 194 Number 7 jb.asm.org 1647

mica by introduction into a sample drop, then transferred to a drop ofwashing solution (double distilled water), and then completely floated offon a drop of negative staining solution, where it was adsorbed onto a400-mesh specimen grid. The staining solution was completely removed,resulting in a shallowly stained specimen. Electron microscopic imagingwas performedwith a Philips EM301 transmission electronmicroscope atcalibrated magnifications. The resulting images were grouped into pro-jection forms essentially as described previously (11) and processed bymodified Markham rotational analysis (38, 50).

Selection of defined areas, trimming, and image rotation were per-formed with Scion Image (Scion Corp.). For image analysis, 1,300 singleparticles were evaluated. Class averaging was used to verify three obvi-ously predominant projection forms (46). Based upon the outlines of theprojection forms (cf. Fig. 1), three-dimensional models were designedwith Adobe Dimension 3.0 (Adobe Corp.).

Generation of fluorescent fusion proteins. Inverse PCR was per-formed on pWS15 to create a SacI restriction site immediately upstreamof the stop codon. The fluorescent protein mOrange was amplified frompmOrange (Clontech) with primers mOrange Fw 2 and mOrange Rv 2and cloned into the newly generated SacI site and the BstBI site alreadypresent in the vector backbone to create pJAL-B2. The pJAL-B2 plasmidwas confirmed to complement the bfpB-null EPEC strain, UMD923. Theplasmid pEM116 was created by using QuikChange site-directed mu-tagenesis on pKDS302 with the primers Donn-1312 and Donn-1308 tocreate an XhoI and SacII site between bfpB and bfpC. The mOrange gene

was cloned into these sites using the primers Donn-1305 and Donn-1311,yielding pEM119. To create fusions of bfpB-mOrange and bfpB-PAm-Cherry3 in the context of the native BFP operon, both proteins were am-plified with the primers mOrange Fusion Fw or mCherry Fusion Fw andmOrange Fusion Rv. The PCR products were spliced into pEM119 di-gested with XhoI and SacII by the In-Fusion method (Clontech), replac-ing themOrange gene fused 5= to bfpC in pEM119 so that the gene prod-ucts would produce a functional C-terminal fusionwith BfpB and containa stop codon before bfpC. The two resulting plasmids, pJAL-B5 carryingmOrange and pJAL-B8 carrying PAmCherry3, were then subjected toQuikChange site-directed mutagenesis to remove the bfpB stop codonand XhoI sites using the primers pJALB6&9 QC Fw and pJALB6&9 QCRv. The resulting plasmids, pJAL-B6 and pJAL-B9, carry an in-frametranslational fusion of BfpB with either mOrange or PAmCherry3 withthe same two amino acid linker as found in pJAL-B2, but in the native bfpBlocus in the BFP operon. These two plasmids were each found to be suf-ficient to build functional pili when expressed in ALN92, and free BfpBdegradation products were not detected by Western blotting. All threeBfpB fusion protein constructs were confirmed by sequencing.

Microscopy and image analysis. For epifluorescence microscopy,overnight cultures of DH5 (pJAL-B2) were diluted 1:250 into LB withampicillin and 2 g 100 ml1 anhydrotetracycline (AHT) and grown for3 h at 37C with shaking. Bacteria were spotted onto a poly-L-lysine-coated slide under cover glass and observed in an LSM510 Meta (Zeiss)confocal microscope. Fluorescent proteins were excited with a 543-nm

TABLE 1 Strains and plasmids used in this study

Strain or plasmid Description and/or genotypea Source or reference

StrainsE2348/69 Serotype O127:H6 EPEC strain isolated from an outbreak in the United Kingdom 44UMD901 E2348/69 bfpA(C129S) 87UMD923 E2348/69 bfpB::aphA3 4UMD946 E2348/69 bfpF(121128) This studyUMD947 UMD 946 bfpB::aphA3 This studyNH4 E2348/69 hsrD 37ALN92 MC4100 RS88 (spy-lacZ) cpxA101 zii::Tn10 52XL1Blue recA1 endA1 gyrA96 thi-1 hsdR17 supE44 relA1 lac [F= proAB lacIqZM15 Tn10 (Tetr)] StratageneBL21-AI slyD F ompT hsdSB(rB

mB) gal dcm araB::T7RNAP-tetA slyD::cat 23

DH5 supE44 lacU169(80dlacZM15) hsdR17 recA1 endA1 gyrA96 thi-1 relA1 66XL10gold Tetr (mcrA)183 (mcrCB-hsdSMR-mrr)173 endA1 supE44 thi-1 recA1 gyrA96 relA1 lacHte

[F= proAB lacIqZM15 Tn10 (Tetr) Amy Camr]Stratagene

PlasmidspKD4 Flip-recombinase vector for the one-step method 24pASK-IBA3 Strep tag expression vector IBApBAD24 Ampr; L-arabinose-inducible expression vector 33pmOrange Cloning vector containingmOrange ClontechpmCherry3 Cloning vector containing PAmCherry3 75pAD07 bfpU-His gene cloned into pBAD24 23pWS15 bfpB-Strep gene cloned into pASK-IBA3 23pJAL-B2 pASK-IBA3::bfpB-mOrange This studypKDS302 pTRC99a carrying IPTG-inducible BFP operon 73pEM116 pKDS302::XhoI-SacII This studypEM119 pKDS302::XhoI-mOrange-SacII-bfpC This studypJAL-B5 mOrange cloned into XhoI and SacII sites pEM119 This studypJAL-B6 pJAL-B5 with stop codon of bfpB and XhoI site removed by QuikChange mutagenesis This studypJAL-B8 PAmCherry cloned into XhoI and SacII sites pEM119 This studypJAL-B9 pJAL-B8 with stop codon of bfpB and XhoI site removed by QuikChange mutagenesis This studypJAL-F1 bfpF cloned into pBAD24 86pJW15 Lux reporter 49pACYC184 pSC101 derived cloning vector 16pNLP27 degP promoter cloned into pJW15 This studypNLP27-Cm Camr cloned into pNLP27; Kans This study

a Camr, chloramphenicol resistant; Tetr, tetracycline resistant; Ampr, ampicillin resistant; Kans, kanamycin sensitive.

Lieberman et al.

1648 jb.asm.org Journal of Bacteriology

HeNe laser with anHFT477/543 beam splitter, and emissionwas collectedwith a 560-605BP filter. ALN92(pJAL-B6) and UMD923(pJAL-B2) wereprepared in the samemanner, except thatUMD923(pJAL-B2)was dilutedinto DMEM and ALN92(pJAL-B6) was diluted 1:50 and grown at 30Cwith 0.1 M IPTG (isopropyl--D-thiogalactopyranoside).

For photoactivated localization microscopy (PALM), ALN92(pJAL-B9) was grown as described above. After 3 h of growth in the presence ofinducer, the cell culture was centrifuged for 5 min at 13,000 g and 25Cand then resuspended in one-fifth volume of filter-sterilized phosphate-buffered saline (PBS). A droplet of 75 l of concentrated cell culture wasplaced on a coverslip coated overnight with poly-L-lysine (0.05% [wt/vol]) and allowed to settle for 30 min in the dark at 25C. Excess fluid wasremoved, and 75 l of freshly prepared, filter-sterilized 4% paraformal-dehyde at pH 7.4 was added for 45 min to fix the bacteria to the coverslip.Adhered cells were washed in filtered PBS and imaged using an OlympusIX81 inverted microscope with a100/1.45 Plan Apo oil immersion ob-jective lens essentially as previously described (29). A total of 15,000frameswere collected per field of view at a rate of 100Hz during excitationat 561 nm. Molecules were activated by 405-nm excitation initially set at50Wand gradually increased over the course of the imaging to main-tain a low density of activated molecules (3 per frame). Molecules werelocalized by fitting a two-dimensional elliptical Gaussian function to a99 pixel array at the peak, and the distribution of localized moleculeswas analyzed in Matlab. The molecular density was calculated within 40nm square pixels, and the distribution of BfpB-PAmCherry was analyzedin the resulting plots.

We observed monopolar, bipolar, envelope, and indeterminate pat-terns of BfpB-PAmCherry distributions by eye and identified four bacte-ria that represented each category. In order to classify the distributionsobjectively, we analyzed the images in ImageJ and developed an algorithmthat assigned the representative bacteria to their predicted classes withhigh fidelity. Only bacteria whose entire outline could be seen were in-cluded for analysis. We used linescan analysis to assign each bacteria ashaving a nonenvelope or envelope distribution of BfpB-PAmCherrymol-ecules and to further categorize these envelope distributions as bipolar,monopolar, or nonpolar. Four-pixel wide lines were drawn along the longaxis down the center of each bacterium and along the short axis at the

FIG 1 Electron micrograph of pure BfpB preparation after negative staining.(A) Examples of the prominent projection forms 1 (top view), 2 (sideview), and 3 (bottom view) are circled. Approximately 23% of all particlescould be assigned to projection form 1, 53% to form 2, and 16% to form 3. (B)Galleries of the three prominent projection forms. For forms 1 and 3, therotational symmetry appears to be obvious, whereas form 2 shows bilateralsymmetry. (C) Approximately 8% of all particles on the carbon film could notbe assigned to one of the three described forms but are interpretable as tran-sient projections of the complex. The gallery shows a schematic viewof a tiltedcomplex (right column) and respective views of the negatively stained BfpBparticles. Rotation axis and tilting direction (following the complexes from topto bottom) is indicated by the arrow symbol. (D) The projection forms afterimage enhancement by rotational and translational analysis (1, top view; 2,side view; 3, bottom view). (E) Processed original image (see panel D1); thedarkest and brightest areas in the original image are colored blue and red. (F)Model of the bisected complex illustrating the distribution of negative stainingsalt (blue). The protein masses identified as brightest areas in the originalimage (see panel E) are marked by red asterisks; grooves between these massesare highlighted by arrows. (G and H) Model of the complex from two views(panel G is tilted toward the bottom view; panel H is tilted toward the topview) after the removal of a quarter section of the whole complex.

TABLE 2 Primers used in this study

Primer Sequence (5=3=) Source or referenceDonn664 GAATACATTAAAATTGATGGGGAAGAAAAGAGGTTTGCTTTTAGTTAGTGGTGTAGGCTGGAGCTGCTT This studyDonn665 AATATCACCGTATTTCTGAACATAATATGTCAGCAAAGCATAGATTGTTGTCATATGAATATCCTCCTTA This studybfpB C18S Fw CGCTCCTGGCATCTTCGTCGGGTAATGGATTTATAAAGATAATCTTGG This studybfpB C18S Rv CCAAGATTATCTTTATAAATCCATTACCCGACGAAGATGCCAGGAGCG This studybfpB S19D Fw CCGCTCCTGGCATCTTGCGACGGTAATGGATTTTATAAAGATAATCTTGG This studybfpB S19D Rv CCAAGATTATCTTTATAAAATCCATTACCGTCGCAAGATGCCAGGAGCGG This studypWS15 iPCR Fw 2 TTAATTATGAGCTCGCTTGGAGTCACCCGCAG This studypWS15 iPCR Rv 2 TATATTTAGAGCTCGCCAGATGCCTTGAGATCAATAATTC This studymOrange Fw 2 GAGATGGGAGCTCGTGAGCAAGGGAGAGGAG This studymOrange Rv 2 CTATATATTCGAATTACTTGTACAGCTCCATGCC This studyDonn-1312 CTCAAGGCTTCTGGCGAATGATACTCGAGCTGCGCTCCCGCGGCATAAAGAATAATCTTGGCG This studyDonn-1308 CGCCAAGATTATTCTTTATGCCGCGGGAGCGCAGCTCGAGTATCATTCGCCAGAAGCCTTGAG This studyDonn-1305 GCGATGGCTCGAGGTGAGCAAGGGCGAGGAG This studyDonn-1311 CTATATACCGCGGGAGCGCAGGCCCGACTTGTACAGCTCGTCCATGCC This studymOrange Fusion Fw GCGAATGATACTCGAGCTCGTGAGCAAGGGCGAGGAGAATAACATGG This studymOrange Fusion Rv ATTATTCTTTATGCCGCGGATCATTCGCCAGAAGCCTTCTTGTACAGCTCGTCCAT This studymCherry Fusion Fw GCGAATGATACTCGAGCTCGTGAGCAAGGGCGAGGAGGATAACATGG This studypJALB6&9 QC Fw TCAAGGCTTCTGGCGAGCTCGTGAGCAAGG This studypJALB6&9 QC Rv CCTTGCTCACGAGCTCGCCAGAAGCCTTGA This studyDegP5=Eco GGAATTCCCGCCATCGGCTGGCCTATGT 59DegP3=Eco CGGATCCGAGAGCCAGTGCACTCAGTGCT 59CamRNcoIF TTTTCCATGGTAAATACCTGTGACGGAAGAT This studyCamRNcoIR TTTTCCATGGTATCACTTATTCAGGCGTAGC This study



approximate midpoint of the cell length. The proximal and distal 10% ofthe longitudinal linescan was taken to represent the cell poles and that ofthe short axis the nonpolar envelope, whereas the middle 80% of eachlinescan was used to represent the cell interior. The fluorescence intensityper pixel for each of these six regions was determined for each bacterium,representing the fluorescence intensity of the boundary pixels at the polesand longitudinal midline, as well as the cell interior. If the average fluo-rescence intensity per pixel of the four cell boundary regions was at least1.6-fold greater than the average fluorescence intensity of the cell interiorsections, the distribution of molecules in that bacterium was classified asenvelope and considered for further analysis. If the fluorescence intensityper pixel for either cell pole was at least 1.5-fold more than the averagefluorescence intensity of the longitudinal midline boundary regions, thatbacteriumwas classified as having amonopolar distribution. These cutoffvalues were selected because they assigned the reference bacteria to theirpredicted classes with the highest concordance. If both ends of the longaxis were greater than the short axis ends, the bacterium was classified asbipolar. Nonenvelope distributions were not considered in the final anal-ysis because BfpB is anOMprotein and never found in the cytoplasm (61,68), and therefore nonenvelope distributions are either artifacts of imag-ing or optical slices through the membrane of the bacteria likely to con-found the analysis.

Excel was used for multivariable correlation analysis. Cell length, cellwidth, and the ratio of length to width were treated as independent vari-ables. The measurements tested as dependent values were: (i) the ratio offluorescence intensities at one pole to the other, (ii) the ratio of the totalpolar intensity to the total nonpolar membrane intensity, (iii) the ratiopolar and nonpolar envelope fluorescence to the fluorescence of the cellinterior, (iv) the distribution class of nonpolar, monopolar, or bipolar, or(v) the distribution class of nonpolar or any polar.

Autoaggregation assays.The autoaggregation phenotype requires ex-pression of functional BFP (4) and serves as a proxy metric for functionalBFP machines. Autoaggregation assays were performed as previously de-scribed (22) with minor modifications: overnight cultures were diluted1:100 in DMEM with appropriate antibiotics and inducer (AHT at 2 g100 ml1; arabinose at 0.2%) and grown at 37C. The autoaggregationindex was determined as previously described (3) at 3, 4, and 5 h postin-oculation. Each experiment contained three biological replicates of indi-vidual colonies, and each experiment was performed three times. Differ-ences between strains at each time point were determined by pairwiseanalysis of variance (ANOVA) inMicrosoft Excel using data from all ninetime points. E. coli cpxA laboratory strain ALN92 carrying pKDS302 or itsderivatives, pJAL-B6 and pJAL-B9, was grown overnight at 30C in Luriabroth, diluted 1:100 inmedium containing antibiotics, and 0.1 mM IPTGand observed microscopically for autoaggregation at 4 to 6 h postinduc-tion.

Construction of DegP-luciferase reporter and reporter assays. TheDegP-Lux reporter plasmid, pNLP27-Cm, was made by cloning the degPpromoter sequence into pJW15 using the primers DegP5=Eco andDegP3=Bam, creating pNLP27 (N. L. Price and T. L. Raivio, unpublisheddata). The chloramphenicol resistance cassette was amplified frompACYC184 by using the primers CamRNcoIF and CamRNcoIR andcloned into the NcoI site of pNLP27 rendering this plasmid kanamycinsensitive (Kans) and chloramphenicol resistant (Camr), creating pNLP27-Cm. The final reporter plasmid, pNLP27-Cm, was passaged throughEPEC strain NH4 and then electroporated into the strains used in thepresent study.

Luciferase reporter assays were performed as previously described (49,82). In four independent experiments, quadruplicate colonies were grownovernight and diluted 1:100 in DMEMwith antibiotics and buffered with0.1 M Tris (pH 7.5). For each culture, 200 l was transferred to a clear96-well plate (Nunc, catalog no. 439454) and to a 96-well white-walled,opaque bottom luminometry plate (Dynex, catalog no. 7417), and thecells were grown at 37C, with shaking at 225 rpm. The A595 and lumines-cence (400 ms) were determined from these plates for samples and for

medium-only wells. The plates were then covered, placed at 37C with225-rpm orbital shaking. The final measurement was determined by thefollowing formula:

RLU(cps)2.5 [(400-ms luminescence) (luminescence blank)]

[(absorbance sample) (absorbance blank)]

Thus, giving the cell density-adjusted relative luminescence units incounts per second per culture. The relative luminescence intensity units(RLU) for the four cultures for each strain were averaged and plotted(error bars represent the standard error of themean [SEM] in the figures).Single-factor, pairwise ANOVAwas performed as described above for theautoaggregation assay except that 12 data points were used for each strainat each time point.

Western blotting.Western blotting for bundlin, BfpB, and BfpU wasperformed as previously described (23, 27, 69). SDS-PAGE gels were runaccording to themanufacturers instructions (Bio-Rad) and transferred at21 V for 80min at 4C to Immobilon polyvinylidene fluoride (Micropore,catalog no. IPFL0010) and blocked overnight in 5% milkPBSTween.The blots were then probed for 1 h at room temperature with rabbitanti-bundlin (1:2,000), rabbit anti-BfpB (1:15,000), or mouse anti-BfpU(1:15,000) in 5%milkPBSTween, washed three times for 5min in PBS-Tween, and probed for 1 h at room temperature with IRDye (680 or 800nm)-conjugated anti-rabbit or anti-mouse secondary antibodies (LicorBiosciences). The blots were washed again three times for 5 min in PBS-Tween at room temperature and scanned with an Odyssey Western sys-tem (Licor Biosciences).

Quantitative Western blotting was performed to determine the stoi-chiometry of BfpB to BfpU and the quantity of each protein per EPECCFU. BfpUwas selected for this experiment because it is known to interactwith BfpB in vivo (23), and therefore the relative stoichiometry may pro-vide insight into machine function. In addition, the BfpU antibody is amouse monoclonal antibody that does not recognize other His-taggedproteins, while the BfpB antibody is a rabbit polyclonal that does notrecognize other Strep-tagged proteins. Since these antibodies are fromdifferent species, they can be used simultaneously to probe for BfpU andBfpB in the same Western blot. Wild-type E2348/69 EPEC was grownovernight at 37C in Luria broth with shaking and then diluted 1:100 inDMEM and grown for 5 h at 37C. Several 1-ml aliquots were taken andcentrifuged at 4C at 13,000 g in a benchtop refrigerated centrifuge. Thepellet was resuspended in 100l of Laemmli buffer and boiled for 10min.Simultaneously, serial dilutions of the culture were prepared, 100l fromthe 105 and 106 dilutionswere plated onLBplates and grownovernightat 37C, and the numbers of CFU were counted the next day. Four vol-umes of whole-cell lysate (10, 5, 2.5, and 1 l) were loaded onto an SDS-PAGE gel, along with a range of purified BfpU (35 to 1,500 ng) and BfpB(40 to 2,500 ng) mixed together. BfpU and BfpB were purified as previ-ously described, and concentrations determined using the extinction co-efficients and themeasuredA280 (23, 69).Western blottingwas performedas described above, and blots were simultaneously probed with anti-BfpUand anti-BfpB, and then with both anti-mouse and anti-rabbit secondaryantibodies. For each BfpB- and each BfpU-reactive band the integratedintensity was measured using the Odyssey software. Using the integratedintensities, standard curves were generated, and the numbers of BfpB andBfpU molecules per CFU of EPEC were calculated for each protein.

RESULTSElectronmicroscopyand singleparticle averaging revealBfpB isa dodecameric gated pore.Wepurified BfpB-Strep to homogene-ity by affinity chromatography on a Strep-Tactin column as pre-viously described (23). Electronmicroscopy and image analysis ofnegatively stained preparations revealed two different BfpB pro-jection forms exhibiting rotational symmetry (Fig. 1A1, A3, 1B1,B3, 1D1, and D3) and a third projection form with bilateral sym-metry (Fig. 1A2, B2, and D2). These three forms may be inter-preted as top, bottom, and side views of the complex, with

Lieberman et al.


outer and inner diameters of approximately 20 and 18 nm, respec-tively. It is not clear which of the top or bottom views representsthe periplasmic or outer face of the complex. Markham rotationalanalysis of a top or bottom view of a single particle reveals thatrotation by angles of n 30 around the central axis of the particlemeets the periodicity of the structure, which accounts for a 12-foldrotational symmetry. Other angles result in a blurred image (datanot shown). Twelve bright spots on the ring-like projection formand 12 dark spots of staining salt clearly indicate the presence of 12proteinmasses. This is illustrated for the top view in Fig. 1E. Here,the darkest and lightest areas of the original image are colored blueand red, respectively. The dark stain is accumulated in smallgrooves, located between protein masses (see Fig. 1F; grooves aremarked by arrows, and protein masses are marked by asterisks).The distribution of negative staining salt may be easily explainedby a model as depicted in Fig. 1F. The H-shaped side view of thecomplex has a trapezoidal outline (Fig. 1D2). The long side of thetrapezium measures 20 nm, corresponding to the diameter ofthe ring-like top view, and the height of the complex measures14 nm. The ring-like projection forms (Fig. 1D1 andD3) exhibita central accumulation of dark staining solution, which is indica-tive of a central depression or pore in the complex as previouslysuggested (Fig. 1F) (68, 68). A bright mass located in the center ofthe top view is indicative of a shallow depression (Fig. 1D1),whereas from the bottom view, the depression appears to bedeeper, i.e., a bright central mass is not visible (Fig. 1D3). Thisinterpretation is confirmed by the appearance of the side view(Fig. 1D2), where the blockingmass is located close to the top sideof the complex. Besides these prominent views it is possible todetect transient forms showing views of molecules tilted to therespective forms 1, 2, and 3 (Fig. 1C). In addition, the bottom view(Fig. 1D3) reveals six triangular leaflets emanating from the insideof the ring structure and extending toward the center of the pore.A tentative model based upon these data shows two views of thewhole complex from the outside and from the inside after a quar-ter of the complex has been removed (Fig. 1G and H).

Quantitative Western blotting reveals the relative stoichi-ometry of BfpB and BfpU inwild-type E2348/69.The number ofmolecules of BfpB per CFU of wild-type EPEC was determinedthrough five independent trials of quantitative Western blotting(Fig. 2) using the Odyssey system (Licor Biosciences). As a stan-dard against which tomeasure the quantity of all Bfp components,we also measured the number of BfpU molecules per cell, since amouse monoclonal antibody against BfpU is available (69). BfpU

was quantified in quadruplicate. On average ( the standard errorof themean [SEM]), there were 6.3 104 (1.4 104)moleculesof BfpB and 5.9 104 (2.8 104) molecules of BfpU per CFU.The average ratio of BfpB molecules to BfpU molecules per CFUwas 1.3 0.43, with a range of 0.6 to 2.4. Assuming a dodecamericstructure for BfpB based on the EM structure reported above,these data suggest 5.2 103 (1.2 103) BfpB multimers perCFU, assuming all BfpBmolecules are in multimers. This calcula-tion yields an estimate of 0.1 0.04 intact BfpB multimers permolecule of BfpU monomer, or 12.9 3.9 BfpU molecules persecretin multimer. The ratio of BfpU molecule to BfpB multimerhad a range of 4.9 to 21.2. There was significantly more variabilityin the number of BfpU molecules detected than was the case forBfpB. The variations in BfpU concentrations did not appear tocorrelatewith the abundance of BfpB. It should be noted that theseestimates assume that the antibodies bind with equivalent affinityto cellular and purified recombinant protein after each is boiledand denatured in SDS.

The Lol-sorting pathway is required for BfpB stability. Totest the hypothesis that the Lol-sorting pathway is required forBfpB transport to the outer membrane, two separate site-directedpointmutations in the bfpB coding sequencewere generated in thecomplementing plasmid, pWS15. The first amino acid after thepredicted signal peptide cleavage site, Cys18, was mutated to aserine generating the plasmid pJAL-B10, which codes forBfpBC18S. In addition, a Lol avoidance signal was created by mu-tating Ser19 to an aspartic acid residue, generating pJAL-B11,which codes for BfpBS19D. Each of these mutated plasmids wasassayed for its ability to restore the autoaggregation defect in thebfpB deletion mutant.

As expected, the wild-type BfpB construct was able to restoreautoaggregation in this strain, while the empty vector could not(Fig. 3A andB). The plasmid coding for themutated Lol lipidationtarget, BfpBC18S, was unable to restore autoaggregation at any timepoint assayed (Fig. 3C). Surprisingly, the Lol avoidance signalmu-tant, BfpBS19D, was able to restore autoaggregation by 4 h postin-duction (Fig. 3D). Although the BfpBS19D mutant is able to com-plement the bfpB deletionmutant, autoaggregation occurred laterthan when this strain was complemented with wild-type BfpB(Fig. 3F). The BfpBS19D-expressing strain was not statistically dif-ferent from the negative control at 3 h postinduction, was differ-ent fromboth the positive and the negative controls at 4 h, andwasnot statistically different from thewild-type BfpB control at 5 h. Incontrast, the BfpBC18S-expressing construct was equivalent to thenegative control at all time points (Fig. 3F). To determine whetherthe altered proteins were expressed, we performed immunoblot-ting. We were unable to detect the BfpBC18S construct, suggestingthat it is unstable, whereas both wild-type BfpB and BfpBS19D arereadily detected at 4 h postinduction (Fig. 3E). Since BfpBS19D isfunctional, this protein must be present in the OM at 4 and 5 hpostinduction.

PALM reveals that BfpB is not found predominantly at thebacterial poles. T4Ps are polar in some, but perhaps not all spe-cies, and the location of these structures may influence their func-tion. To test the hypothesis that BFP components localize to thecell pole, we fused either the gene for photoactivatable mCherry(PAmCherry) ormOrange to the 3= end of bfpB and examined thecellular distribution of the fusion proteins using PALM or tradi-tional epifluorescence. These fluorescent proteinswere chosen be-cause mCherry was successfully fused to the PulD of Klebsiella

FIG 2 Representative quantitative Western blot for BfpB and BfpU stoichi-ometry. Lane 1, marker; lanes 2 to 5, EPEC whole-cell lysates, 10, 5, 2.5, and 1l; lane 6, empty; lanes 7 to 13, protein standards were loaded as 2-fold serialdilutions containing both purified recombinant BfpB from 621 to 9.7 ng andBfpU from 148 to 2.2 ng. BfpB-reactive bands appear in the top panel, andBfpU-reactive bands appear in the lower panel. The double asterisk indicatesthe 50-kDamarker band, while the single asterisk indicates the 15-kDamarkerband. The predicted molecular mass of purified recombinant BfpB is 52 kDa,and that of BfpU is 16.5 kDa.



oxytoca in a previous report (12) andmOrange, like mCherry, is aderivative of DsRed (71). Indeed, we found that the bfpB-mOr-ange fusion could complement the bfpB mutant in trans and re-store autoaggregation, while the bfpB-mOrange and bfpB-PAm-Cherry fusions in the context of the entire BFP operon couldconfer the capacity to autoaggregate on the ALN92 laboratorystrain of E. coli that carries a constitutively activated cell envelopesensor in the form of the cpx mutation. ALN92 was previouslydemonstrated to support BFP elaboration and autoaggregationfrom the cloned BFP operon due to the cpx mutation; the Cpxpathway is important for efficient BFP expression and adherenceto host cells (52).

When BfpB-mOrange was expressed in trans, either in a labo-ratory strain of E. coli or in the bfpB-null mutant, intense foci offluorescence were observed at one or both cell poles (see Fig. S1 inthe supplementalmaterial). A small amount of fluorescence couldbe detected around the cell periphery. However, when BfpB-mO-range was expressed in the context of the other BFP proteins fromthe bfp operon in ALN92, we observed a more diffuse pattern offluorescence around the cell envelope; in some bacteria, fluores-cence was exclusively observed at the poles.

The results from the epifluorescence imaging illustrate theconcept that expression of BfpB-mOrange in the context of thenative operon gives strikingly different results from expression ofthe protein in trans. For finer resolution and quantification of thedistribution of BfpB molecules, we used PALM to localize singleBfpB-PAmCherry molecules with high precision. We examinedALN92 expressing BfpB-PAmCherry from the BFP operon in

fixed cells. We imaged 427 bacteria in three independent experi-ments. Briefly, singlemolecules of BfpB-PAmCherry were imagedby oblique illumination of fixed cells, using low-intensityUVpho-toactivation to maintain a sparse density of activated molecules.Bacteria lacking the pmCherry3 plasmid displayed no photoacti-vatable fluorescence (results not shown). Individual moleculeswere identified and localized as described previously (29), and themolecular density was plotted in 40-nm pixels. We observed thatthe distribution of BfpB-PAmCherry molecules appeared to fallinto one of four categories, i.e., bipolar, monopolar, envelope, orindeterminate, and we selected four model bacteria for each class.Bacteria expressing soluble PAmCherry from pmCherry3 wereindistinguishable from the indeterminate class, as expected ofan exclusively cytosolic distribution of the free fluorophore (datanot shown).

We used a linescan analysis to estimate the average fluores-cence intensity at four regions of interest: the envelope at the lon-gitudinal midline of the cell; the first and last 10% of the celllength, which we defined as the cell poles; and the cell interior.Bacteria whose margins could not be clearly seen were excludedfrom the linescan analysis, leaving 276 bacteria for further classi-fication. Of note, bacteria involved in autoaggregates were ex-cluded from this analysis because themargins between aggregatedbacteria could not be precisely defined.

To objectively classify the distribution of molecules in a bacte-rium, we generated a set of rules that utilized these data to assignthe model bacteria to their predicted classes. We reasoned thatsince BfpB is an outermembrane protein that is not detected in the

FIG 3 BfpB Residue C18 is required for protein stability, while the S19D Lol avoidance signal can be overcome. Images of bfpBmutant strain cultures at 3.5 hpostinoculation complemented with null cloning vector (pASK-IBA3) (A) or vector encoding wild-type BfpB (pWS15) (B), BfpBC18S (pJAL-B10) (C), orBfpBS19D (pJAL-B11) (D). (E) Western blot of whole-cell lysates from cultures at 4 h using anti-BfpB antiserum. An asterisk indicates the 50-kDa marker. Anarrowhead indicates the BfpB monomer band. (F) Quantitative autoaggregation index (AI) determined at 3, 4, and 5 h postinoculation into DMEM for eachstrain. Diamonds indicate null vector control, squares correspond to wild-type BfpB, triangles indicate BfpBC18S, and crosses () indicate BfpBS19D.

Lieberman et al.


cytoplasmic fraction (61, 68), localization of molecules in the ap-parent interior of the cell occurred primarily when the portion ofthe cell membrane closest to the coverslip was in the focal plane.To select cells in which the focal plane crossed the center of thez-axis of the cell, we excluded any bacteria from further analysis ifthe average fluorescence at the boundary of the cells, including atthe longitudinalmidline of the bacterium and at the poles, did notexceed the average interior fluorescence by 1.6-fold. For theremaining 141 cells, we determined whether the intensity at eitherpole was at least 1.5-fold greater than the average fluorescence ofthe boundary pixels at the longitudinal midline of the cell. Thesetwo cutoff values gave the highest concordance between assign-ment of model bacteria by eye and by our algorithm. In this way,wewere able to assign bacteria tomonopolar, bipolar, or nonpolarclasses. In contrast to the results observed by fluorescencemicros-copy when BfpB-mOrange was expressed in trans, 80.1% (n 113) of all included bacteria displayed a nonpolar distribution ofBfpB-PAmCherry molecules characterized by a distribution ofmolecules along the boundary of the cell. A further 12.8% (n 18)had a concentration of molecules at a single pole, and 7.1% (n10) had a bipolar distribution of molecules (Fig. 4).

We reasoned that physical and temporal changes in the bacte-rial life cycle may modulate the location of BFP componentswithin the cell. Therefore, we performed a multivariable correla-tion analysis to determine whether particular distributions were

associated with physical parameters of the bacteria (Table 3). Ofthe relationships tested, only one approached statistical signifi-cance: the ratio of the total polar fluorescence to the total nonpolarmembrane fluorescence was positively associated with the ratio ofcell length to cell width (r 0.266, P 0.074).

In addition, we observed two novel distributions of BfpB-PAmCherry. In 8.5% (n 12) of the imaged bacteria, we notedfoci of fluorescence away from the poles. Although not includedfor polarity analysis, bacteria involved in autoaggregates com-monly formed nonpolar foci, often multiply in each cell envelope(data not shown). Furthermore, in 2.8% (n 4) of the imagedbacteria, we saw banding of BfpB-PAmCherry with lines of mol-ecules apparently oriented at an oblique angle to the long axis, asmight be expected of a helical distribution. Example images areseen in Fig. 4. Taken together, these results illustrate in unprece-dented detail a subcellular localization of the T4P secretin morecomplex than previously demonstrated. Furthermore, these stud-ies extend the previously reported importance of gene dosage ef-fects when studying the distribution of bacterial secretion systemswith fluorescent fusion proteins from type 2 secretion systems totype IV pili (47).

A bfpB bfpF double mutant is viable in BFP-inducing condi-tions and expresses bundlin at levels similar to those for wild-type E2348/69. In N. gonorrhoeae the absence of the secretin andthe retraction ATPase is cytotoxic (84), while the equivalent dou-ble mutation in N. meningitidis is viable (14). In both systems,intraperiplasmic pili were observed (14, 84). To test the hypothe-sis that the bfpB bfpF double mutation in EPEC is cytotoxic, strainUMD947 was created as described inMaterials andMethods. Im-portantly, complementation ofUMD947with a plasmid encodingBfpB resulted in a strain that could formautoaggregates, but couldnot disaggregate, a phenotype consistent with the bfpFmutation.In contrast, complementation of UMD947 with a plasmid encod-ing BfpF resulted in a strain that could not form autoaggregates, asexpected from a bfpBmutation. Thus, the phenotypes of the dou-ble bfpB bfpF mutant were consistent with its genetic defects (seeFig. S2 in the supplemental material). UMD947 was grown inBFP-inducing conditions alongwith wild-type E2348/69, the bfpBmutant (UMD923), and the bfpF deletion mutant (UMD946),and with each strain complemented for the relevant mutation(s)in trans (data not shown). These strains were assayed for theirgrowth rates in BFP-inducing conditions and growth curves wereconstructed from the optical density at 600 nm (OD600) valuesdetermined after vortexing to disrupt autoaggregates (Fig. 5). Thisresult was corroborated by CFU counts on serial dilutions of wildtype, single, and double mutants at 3, 4, and 5 h postinduction asdescribed in Materials and Methods for quantitative Westernblotting.

Surprisingly, no strain showed evidence of growth defects in

FIG 4 Distributions of single BfpB-PAmCherry molecules localized usingPALM. Singlemolecules of BfpB-PAmCherry expressed from the complete bfpoperon in ALN92 cells were imaged by PALM and classified as bipolar (7.1%)(A), monopolar (12.8%) (B), or nonpolar (80.1%) (C) distributions. (D) Bac-teria with a low ratio of fluorescence at the cell envelope relative to the cellinterior were classified as nonenvelope and were not included in the cell po-larity analysis. (E and F) Two unanticipated distributions of BfpB-PAmCherrymoleculeswere also observed: nonpolar foci of fluorescence (E) andbanding atoblique angles to the longitudinal axis, suggesting a helical distribution ofmolecules (F). Note that the bacterium pictured in panel C has bilateral pe-ripheral clusters suggestive of a helical distribution. For all images, the grayvalue indicates the number of molecules per pixel and was normalized to thebrightest pixel in each image.

TABLE 3Multi variable correlation of fluorescence distribution with cell parameters

Dependentvariable

Putative dependent variables (coefficients of correlation SEM)

Pole A/pole B ratio Pole B/pole A ratioPoles/nonpolarmembrane ratio

Envelope ratio(yes/no)

Total membraneinterior

Nonpolar vs anypolar

Length 0.0 0.0 0.0 0.0 0.0 0.0Width 0.0 0.0 0.0 0.0 0.0 0.0Length/width 0.50 1.5 0.90 1.1 16.32 9.1a 3.53 6.8 0.19 0.49 0.1 0.34a r 0.266, P 0.074.



DMEM when assayed by determining the OD600 (Fig. 5). Simi-larly, there was no evidence that the growth rate of the doublemutant was reduced in comparison to the other strains when as-sayed by viable counts (data not shown). Although all of the non-complemented strains displayed growth rates similar to wild-typeE2348/69, two strains did not grow nearly as well: the bfpB bfpFdoublemutant complemented for bfpB and the bfpFmutant com-plemented for bfpF (data not shown). These results strongly sug-gest that the bfpB bfpF double mutation is not cytotoxic to EPEC.The reduced growth rate in some of the complemented strainsmay be the result of Bfp protein expression at aberrant levels.

We performed immunoblotting to determine whether the vi-ability of the bfpB bfpF strain was due to the degradation of bund-lin, the major pilus subunit. Bundlin was detected in all of thestrains tested, except the negative control, bfpA mutant strainUMD901 (Fig. 6A). To determine whether any evidence of intra-cellular T4P expression by the bfpB bfpF double mutant could bedetected, cells from wild-type E2348/69 and the isogenic bfpB,bfpF, and bfpB bfpF doublemutant strains were observed by TEM.No membrane abnormalities were observed (data not shown). Inaddition, when cells harboring the bfpB bfpF mutation wereprobed with an antibody against bundlin, no fluorescence wasobserved by microscopy before or after osmotic shock (see Fig. S3in the supplemental material). Not only does the bfpB bfpF strainthrive in BFP-inducing conditions, but it is also free of membraneabnormalities and altered expression levels of bundlin do not ac-count for this survival. Despite the presence of bundlin detectedby immunoblotting, no intraperiplasmic pili were observed byeither fluorescence microscopy or TEM. These results suggest theexistence of a feedback mechanism to limit expression of BFPwhen OM egress is prevented.

In BFP-inducing conditions degP transcriptional activationin the bfpB bfpF double mutant is similar to that of wild type.Given that reduced expression of bundlin does not account for theability of EPEC cells to tolerate a double mutation of bfpB and

bfpF, we hypothesized that the envelope stress response pathwaywas upregulated inUMD947 and protected the cell envelope fromdamage. BFP proteins encounter periplasmic Cpx pathway effec-tors such as DegP and DsbA once they cross the IM (48); indeed,some low level of Cpx activity is required for pilus biogenesis (49,82). To assess the activity of the cell envelope stress response sys-tem in wild-type and various mutant strains, we constructed aluciferase reporter for degP transcriptional activation, pNLP27-Cm, that is compatible with the mutants.

Strains carrying the degP-lux reporter plasmid were grown inBFP-inducing conditions in quadruplicate independent culturesand assayed for cell density (OD595) and luciferase activity at 2, 4,6, and 8 h postinoculation. The luminescence was normalized bycell density, and the experiment was performed four times (Fig.6B). Surprisingly, the bfpB bfpF doublemutant exhibited degP-luxactivation that was indistinguishable from that of the wild-typestrain at 2 h. At 4 h, various strains showed differing levels ofdegP-lux activity, except for the bfpA and bfpBmutants. At 6 h andat 8 h, however, the bfpB bfpF strain had degP-lux levels that weresimilar to both bfpB and bfpF single mutants. Interestingly, thelevel of degP-lux activation in the bfpB single mutant was muchgreater than that of the other strains tested, peaking at 2 h, reflect-ing an early spike in cell envelope stress when this mutant is firstinoculated into BFP-inducing media. Thus, the viability and lack

FIG 5 The bfpB bfpF double mutant strain is viable in BFP-inducing condi-tions. Growth curves for wild-type EPEC (), bfpB mutant strain UMD923(), bfpFmutant strainUMD946 (), and the bfpB bfpFdoublemutant strainUMD947 () are shown.

FIG 6 Neither bundlin degradation nor degP activation account for the via-bility of the bfpB bfpF double mutant. (A) Representative Western blot forbundlin in whole-cell lysates of wild-type EPEC strain E2348/69 (lane 2), bfpAmutant strain UMD901 (lane 3), bfpB bfpF double mutant strain UMD947(lane 4), bfpB bfpF double mutant complemented for BfpB (lane 5), bfpB bfpFdouble mutant complemented for BfpF (lane 6), bfpBmutant strain UMD923(lane 7), and bfpFmutant strain UMD946 (lane 8). *, 20 kDa; **, 15 kDa. Thearrowhead indicates the bundlin band. (B) Luminescence data from degP-luxreporter assay from three independent experiments each performed in qua-druplicate. Error bars represent the SEM. Diamonds indicate wild-type EPECstrain E2348/69, squares indicate bfpAmutant strain UMD901, triangles indi-cate bfpB mutant strain UMD923, crosses () indicate bfpF mutant strainUMD946, and asterisks indicate bfpB bfpF double mutant strain UMD947.

Lieberman et al.


of detectable intracellular BFP formation by the bfpB bfpF doublemutant is not due to higher levels of compensatory Cpx envelopestress activation.

DISCUSSION

Secretins are outer membrane proteins common to and requiredfor the function of T4P, T2S, and T3S systems. BfpB is a lipopro-tein secretin required for biogenesis of the EPEC BFP. We ad-dressed several questions related to BfpB targeting to the OM,complex structure, subcellular localization, and the viability of asecretin and retraction ATPase double mutant, and our resultsillustrate the complexities of secretin function.

As predicted, we found that the N-terminal cysteine immedi-ately downstream of the proposed signal peptidase II cleavage sitein BfpB is necessary for protein stability since BfpBC18S does notcomplement a bfpB-null mutant strain and cannot be detected byWestern blotting. BfpB contains only two cysteine residues, Cys18and Cys540. Unlike BfpBC18S, mutation of Cys540 to a serine doesnot affect the stability or function of BfpB (J. A. Lieberman et al.,unpublished data). Therefore, these data strongly suggest thatCys18 is the palmitoylated cysteine residue and lipid anchor forBfpB (61) and show that acylation is required for stability. More-over, we found that a Lol avoidance signal delayed but did notblock complementation of the bfpB mutant with BfpBS19D. SinceBfpBS19D can support autoaggregation at 4 and 5 h postinocula-tion, this proteinmust be localized to theOMat these times. Theseresults suggest that while the Lol pathway likely lipidates BfpB atCys18, a Lol-avoidance signal retards but does not ablate BfpBfunction. Thus, BfpB likely contains dominant intrinsic OM tar-geting information, similar to theHxcQ lipoprotein secretin of theT2S system in P. aeruginosa (81). Our results complement andextend those of the previousHxcQ study inwhich the role of Lol insecretin transport was not tested (81).

Once inserted in the outer membrane, BfpB is presumed toserve as a pore for growing pili (19). Surprisingly, bfpB bfpFdoublemutant cells were viable, displayed no growth defect, and neitherperiplasmic pili normembrane blebs could be detected using elec-tronmicroscopy in such cells (not shown). Furthermore, we couldnot detect pili in these cells by immunofluorescence microscopyin either intact cells or after osmotic shock. Complementationwith each individual gene verified that the double bfpB bfpF mu-tant had the expected phenotypes. We considered the possibilitythat bfpB bfpFmutant cells are viable because they degrade bund-lin and thus avoid membrane damage. Surprisingly, bundlin wasdetected at similar levels in bfpB and bfpF single mutants, thedouble mutant, and the complemented forms of the double mu-tant. Since bundlin was expressed in BFP-inducing conditions inthe double mutant strain, we next tested the hypothesis that thesecells had an elevated cell envelope stress (Cpx) response as amech-anism to protect against damage from periplasmic pili. To oursurprise, a degP-lux gene reporter assay revealed that the Cpx ef-fector degP was transcribed in the bfpB bfpF double mutant at alevel similar to or lower than that observed in wild-type EPEC.Furthermore, all other mutants tested had significantly elevatedtranscription of degP over the wild type. The upregulation of degPtranscription was most pronounced in the bfpB mutant strain.Previous reports suggested that BfpB mediates extrusion ofperiplasmic EPEC proteins, including T3S system components(68). It may be that the inability to expel these components fromthe periplasm triggers an exaggerated stress response. Further-

more, the relative reduction in degP activation in the bfpB bfpFdoublemutant suggests that BfpFmay play a role in activating thisstress response and that BfpB may play a role in such signaling.Alternatively, it may be that when either the secretin or the retrac-tion ATPase is absent, the cells experience more stress on themembrane due to the resulting aberrant machines and activateCpx. However, when both BfpB and BfpF are lacking, either theBFP machine may not be able to assemble or there may be dis-rupted signaling for activation of Cpx.

The BfpB complex observed by electronmicroscopy representsthe first published single particle averaging model structure for aT4bP secretin but in many ways resembles other secretin com-plexes (42). From the 12-fold symmetry observed throughMarkham rotational analysis, we conclude that the complex is ahomododecamer; the six centripetal projections in Fig. 1D3 sug-gest a possible configuration of six dimers. The long bottompartof the trapezium structure seen in transverse views is most likelythe periplasmic N terminus, with a large vestibule for the assem-bling pilus, as has been reported for secretins and their substratesof the T2S, T3S, and T4aP machines (19, 42, 63). However, thissuggestion requires experimental evidence for confirmation. Wepropose that the contoured sides of the cylinder represent thestructured N0 and N3 subdomains of the N terminus (42, 64).The central density in the complex indicates that there is at leastone gate occluding the central channel. Our results underscore theconcept that each secretin family displays structural variations ona common theme and identify several key features of T4bP secre-tin complexes for the first time: a dodecameric structure, a chan-nel plug, and the presence of centripetal projections inside thechannel. It is apparent that significant conformational changeswould be required to accommodate the dynamic T4P fibers; fur-ther studies of secretin topology and architecturewill elucidate themechanics of pilus and substrate extrusion and the structural dif-ferences between secretin families.

Previous studies have demonstrated that the N terminus ofBfpB interacts with the soluble, periplasmic protein BfpU and thatthis interaction is required for BFP machine function (69). Wequantified the number of BfpB and BfpU molecules per CFU andestimate approximately 5.22 103 BfpB multimers and 5.91 104 BfpUmolecules per CFU. A previous study estimated 4 105

molecules of bundlin perCFU (69); thus, if all of the detectedBfpBis involved in forming multimers and each is associated with apilus, the ratio of bundlin to secretin complex is at least 76 to 1.The same study estimated 9 103 molecules of BfpU per EPECcell (69), approximately one-sixth of the current estimate. Thisdifference may reflect the variability of BfpU protein levels notedin the results. Although the antibodies used are highly specific forthe two BFP proteins they recognize, and they were raised againstaffinity-tagged, purified proteins that are confirmed by comple-mentation analysis to be functional, it has not been proven thatthey have identical affinity for the wild-type proteins, and thusthese estimates should be viewed with caution.

Once the dodecameric BfpB complex is transported to andassembled in the OM, it appears to undergo an organized distri-bution through the cell. When BfpB-mOrange was expressed intrans we observed polar foci; however, when all BFP machinecomponents were coordinately expressed from their clonedoperon, we saw a more diffuse pattern of membrane localization.We were able to observe the distribution of BfpB in unprece-dented detail through super-resolution microscopy. We used



PALM to localize single functional BfpB-PAmCherry moleculesand objectively quantified their distribution in a recombinant ex-pressing BFP from the bfp operon. We did not find that thesemolecules localized predominantly to the poles. Instead, the ma-jority of bacteria had an uneven distribution of BfpB molecules atthe cell periphery. Although infrequent, we also observed clustersof BfpB molecules away from the poles, particularly in cells in-volved in autoaggregates, and occasionally a banding pattern sug-gestive of a helical distribution. These observations indicate thatBfpB distribution is most likely not random. Cowles and Gitai(20) discovered that the bacterial actin homologue, MreB, was acrucial determinant for the initiation and maintenance of PilTretraction ATPase complexes at the poles of P. aeruginosa. Basedon the results of the present study and two recent studies demon-strating a helical organization for the type IV secretion system inAgrobacterium spp. (1, 2), we speculate that the bacterial cytoskel-eton plays a role in regulating BFPmachine protein localization inEPEC. MreB could coordinate the initiation of pilus biogenesismachines at the cell pole, possibly in a cell cycle-dependent man-ner. Furthermore, our studies extend earlier findings that proteinstoichiometry is a key determinant of machine localization in themembrane from T2S systems (47) to T4P systems.

The use of emerging super-resolution microscopy, coupledwith coexpression of all BFP components from the cloned operonprovide new and robust insight into the distribution of T4P bio-genesis machines. Our PALM results suggest a different distribu-tion for the secretin and, by inference, the BFP than for some otherT4P systems. The nonpolar localization of BFP may be a featurecommon to T4bP, and it may be related to function. The polarlocalization of T4aP in M. xanthus and P. aeruginosa are likelyessential for longitudinally directed movement, as observed intwitchingmotility and social gliding (31, 56).However, EPECmaynot use BFP formotility but rather for cell interactions. By distrib-uting pili around the cell envelope, EPEC cells may be better ableto form the BFP-dependent spherical autoaggregates characteris-tic of localized adherence. More efficient aggregation at the earlyinfection time points where BFP is most critical (86) may be aconsiderable advantage to the pathogen as it first colonizes thegastrointestinal tract, where it must compete with the commensalflora.

The results presented here identify several key features for amember of the T4bP family of secretins.We confirmed a previousreport (81) that lipoprotein secretins can direct themselves to theOM and determined that a Lol avoidance signal can delay but noteliminate this OM targeting. We report for the first time that si-multaneous mutation of the genes for a T4bP secretin and retrac-tion ATPase can result in viable apparently healthy cells that par-adoxically display less evidence of cell envelope stress than does asingle secretin mutant. This observation suggests a possible rolefor BfpB in sensing or suppressing cell envelope stress that may bemediated through BfpF. It is possible that differences in cell enve-lope stress response could account for the striking differences inviability of N. meningitidis and N. gonorrhoeae strains with theequivalent double mutations. The structure of the BfpB complexby TEM and single particle averaging displays similarities and in-triguing differences compared to other secretins, emphasizing theconcept that differences between secretin families are functionallyimportant. Finally, new insights into secretin cellular localizationgained through PALM suggest that pilus function may be influ-enced by subcellular distribution of pilus machines.

ACKNOWLEDGMENTS

We thank Ekaterina Milgotina for the construction of plasmids pEM116and pEM119, Wiebke Schreiber for BfpB purification for structure deter-mination, Brian Peters for assistance with epifluorescence imaging, Ru-ching Hsia and John Strong for assistance with TEM, andWensheng Luoand Courtney D. Sturey for helpful discussions of the manuscript.

Support was provided by National Institutes of Health grants T32GM008181 (J.A.L. and N.A.F.), F30 MH086185 (N.A.F.), R01 AI37606(J.A.L. and M.S.D.), and R01 MH080046 (T.A.B.).

REFERENCES1. Aguilar J, Cameron TA, Zupan J, Zambryski P. 2011. Membrane and

core periplasmic Agrobacterium tumefaciens virulence type IV secretionsystem components localize to multiple sites around the bacterial perim-eter during lateral attachment to plant cells. mBio 2(5):e00218-11. doi:10.1128/mBio.00218-11.

2. Aguilar J, Zupan J, Cameron TA, Zambryski PC. 2010. Agrobacteriumtype IV secretion system and its substrates form helical arrays around thecircumference of virulence-induced cells. Proc. Natl. Acad. Sci. U. S. A.107:37583763.

3. Anantha RP, Stone KD, Donnenberg MS. 1998. The role of BfpF, amember of the PilT family of putative nucleotide-binding proteins, in typeIV pilus biogenesis and in interactions between enteropathogenic Esche-richia coli and host cells. Infect. Immun. 66:122131.

4. Anantha RP, Stone KD, Donnenberg MS. 2000. Effects of bfp mutationson biogenesis of functional enteropathogenic Escherichia coli type IV pili.J. Bacteriol. 182:24982506.

5. Averhoff B, Friedrich A. 2003. Type IV pilus-related natural transforma-tion systems: DNA transport in mesophilic and thermophilic bacteria.Arch. Microbiol. 180:385393.

6. Bayan N, Guilvout I, Pugsley AP. 2006. Secretins take shape. Mol.Microbiol. 60:14.

7. Betzig E, et al. 2006. Imaging intracellular fluorescent proteins at nano-meter resolution. Science 313:16421645.

8. Bieber D, et al. 1998. Type IV pili, transient bacterial aggregates, andvirulence of enteropathogenic Escherichia coli. Science 280:21142118.

9. Bitter W, Koster M, Latijnhouwers M, De Cock H, Tommassen J. 1998.Formation of oligomeric rings by XcpQ and PilQ, which are involved inprotein transport across the outer membrane of Pseudomonas aeruginosa.Mol. Microbiol. 27:209219.

10. Bradley DE. 1980. A function of Pseudomonas aeruginosa PAO polar pili:twitching motility. Can. J. Microbiol. 26:146154.

11. Braks IJ, Hoppert M, Roge S, Mayer F. 1994. Structural aspects andimmunolocalization of the F420-reducing and non-F420-reducing hy-drogenases from Methanobacterium thermoautotrophicum Marburg. J.Bacteriol. 176:76777687.

12. Buddelmeijer N, Krehenbrink M, Pecorari F, Pugsley AP. 2009. Type IIsecretion system secretin PulD localizes in clusters in the Escherichia coliouter membrane. J. Bacteriol. 191:161168.

13. Bulyha I, et al. 2009. Regulation of the type IV pili molecular machine bydynamic localization of twomotor proteins. Mol. Microbiol. 74:691706.

14. Carbonnelle E, Helaine S, Nassif X, Pelicic V. 2006. A systematic geneticanalysis in Neisseria meningitidis defines the Pil proteins required for as-sembly, functionality, stabilization and export of type IV pili. Mol.Micro-biol. 61:15101522.

15. Chakraborty S, et al. 2008. Type IV pili in Francisella tularensis: roles ofpilF and pilT in fiber assembly, host cell adherence, and virulence. Infect.Immun. 76:28522861.

16. Chang ACY, Cohen SN. 1978. Construction and characterization ofamplifiable multicopy DNA cloning vehicles derived from the P15A cryp-tic miniplasmid. J. Bacteriol. 134:11411156.

17. Collin S, Guilvout I, Nickerson NN, Pugsley AP. 2011. Sorting of anintegral outer membrane protein via the lipoprotein-specific Lol pathwayand a dedicated lipoprotein pilotin. Mol. Microbiol. 80:655665.

18. Collins RF, et al. 2003. Three-dimensional structure of the Neisseriameningitidis secretin PilQ determined from negative-stain transmissionelectron microscopy. J. Bacteriol. 185:26112617.

19. Collins RF, et al. 2005. Interaction with type IV pili induces structuralchanges in the bacterial outer membrane secretin PilQ. J. Biol. Chem.280:1892318930.

20. Cowles KN, Gitai Z. 2010. Surface association and the MreB cytoskeleton

Lieberman et al.


regulate pilus production, localization and function in Pseudomonasaeruginosa.Mol. Microbiol. 76:14111426.

21. Craig L, Pique ME, Tainer JA. 2004. Type IV pilus structure and bacterialpathogenicity. Nat. Rev. Microbiol. 2:363378.

22. Crowther LJ, Anantha RP, Donnenberg MS. 2004. The inner membranesubassembly of the enteropathogenic Escherichia coli bundle-forming pi-lus machine. Mol. Microbiol. 52:6779.

23. Daniel A, et al. 2006. Interaction and localization studies of enteropatho-genic Escherichia coli type IV bundle-forming pilus outermembrane com-ponents. Microbiology 152:24052420.

24. Datsenko KA, Wanner BL. 2000. One-step inactivation of chromosomalgenes in Escherichia coli K-12 using PCR products. Proc. Natl. Acad. Sci.U. S. A. 97:66406645.

25. Donnenberg MS, Girn JA, Nataro JP, Kaper JB. 1992. A plasmid-encoded type IV fimbrial gene of enteropathogenic Escherichia coli asso-ciated with localized adherence. Mol. Microbiol. 6:34273437.

26. Donnenberg MS, Kaper JB. 1992. Minireview: enteropathogenic Esche-richia coli. Infect. Immun. 60:39533961.

27. Fernandes PJ, Guo Q, Donnenberg MS. 2007. Functional consequencesof sequence variation in bundlin, the enteropathogenic Escherichia colitype IV pilin protein. Infect. Immun. 75:46874696.

28. Frols S, et al. 2008. UV-inducible cellular aggregation of the hyperther-mophilic archaeon Sulfolobus solfataricus is mediated by pili formation.Mol. Microbiol. 70:938952.

29. Frost NA, Shroff H, Kong H, Betzig E, Blanpied TA. 2010. Single-molecule discrimination of discrete perisynaptic and distributed sites ofactin filament assembly within dendritic spines. Neuron 67:8699.

30. Gauthier NC. 2009. Bundle-forming pili from enteropathogenic Esche-richia coli generate moderate forces. Biophys. J. 96:519a.

31. Gibiansky ML, et al. 2010. Bacteria use type IV pili to walk upright anddetach from surfaces. Science 330:197.

32. Guerrant RL, et al. 2002. Magnitude and impact of diarrheal diseases.Arch. Med. Res. 33:351355.

33. Guzman LM, Belin D, Carson MJ, Beckwith J. 1995. Tight regulation,modulation, and high-level expression by vectors containing the arabi-nose PBAD promoter. J. Bacteriol. 177:41214130.

34. Herrington DA, et al. 1988. Toxin, toxin-coregulated pili, and the toxRregulon are essential for Vibrio cholerae pathogenesis in humans. J. Exp.Med. 168:14871492.

35. Hess ST, Girirajan TP, Mason MD. 2006. Ultra-high resolution imagingby fluorescence photoactivation localization microscopy. Biophys. J. 91:42584272.

36. Higashi DL, et al. 2011. N. elongata produces type IV pili that mediateinterspecies gene transfer withNeisseria gonorrhoeae. PLoS One 6:e21373.

37. Hobson N, Price NL, Ward JD, Raivio TL. 2008. Generation of arestrictionminus enteropathogenic Escherichia coli E2348/69 strain that isefficiently transformed with large, low copy plasmids. BMC Microbiol.8:134.

38. Horne RW, Hobart JM, Markham R. 1976. Electron microscopy oftobacco mosaic virus prepared with the aid of negative staining-carbonfilm techniques. J. Gen. Virol. 31:265269.

39. Hwang J, Bieber D, Ramer SW, Wu CY, Schoolnik GK. 2003. Structuraland topographical studies of the type IV bundle-forming pilus assemblycomplex of enteropathogenicEscherichia coli. J. Bacteriol. 185:66956701.

40. Hyland RM, Beck P, Mulvey GL, Kitov PI, Armstrong GD. 2006.N-Acetyllactosamine conjugated to gold nanoparticles inhibits entero-pathogenic Escherichia coli colonization of the epithelium in human intes-tinal biopsy specimens. Infect. Immun. 74:54195421.

41. Kaiser D. 1979. Social gliding is correlated with the presence of pili inMyxococcus xanthus. Proc. Natl. Acad. Sci. U. S. A. 76:59525956.

42. Korotkov KV, Gonen T, Hol WG. 2011. Secretins: dynamic channels forprotein transport across membranes. Trends Biochem. Sci. 36:433443.

43. Kosek M, Bern C, Guerrant RL. 2003. The global burden of diarrhoealdisease, as estimated from studies published between 1992 and 2000. Bull.World Health Organ. 81:197204.

44. LevineMM, et al. 1978. Escherichia coli strains that cause diarrhoea but donot produce heat-labile or heat-stable enterotoxins and are non-invasive.Lancet i:11191122.

45. Linderoth NA, Model P, Russel M. 1996. Essential role of a sodiumdodecyl sulfate-resistant protein IV multimer in assembly-export of fila-mentous phage. J. Bacteriol. 178:19621970.

46. Ludtke SJ, Baldwin PR, Chiu W. 1999. EMAN: semiautomated soft-

ware for high-resolution single-particle reconstructions. J. Struct. Biol.128:8297.

47. Lybarger SR, Johnson TL, Gray MD, Sikora AE, Sandkvist M. 2009.Docking and assembly of the type II secretion complex of Vibrio cholerae.J. Bacteriol. 191:31493161.

48. MacRitchie DM, Buelow DR, Price NL, Raivio TL. 2008. Two-component signaling and gram negative envelope stress response systems.Adv. Exp. Med. Biol. 631:80110.

49. MacRitchie DM, Ward JD, Nevesinjac AZ, Raivio TL. 2008. Activationof the Cpx envelope stress response downregulates expression of severallocus of enterocyte effacement-encoded genes in enteropathogenic Esch-erichia coli. Infect. Immun. 76:14651475.

50. Markham R, Frey S, Hills GJ. 1963. Methods for the enhancement ofimage detail and accentuation of structure in electron microscopy. Virol-ogy 20:88102.

51. Milgotina E, Sonnenberg M. 2009. The bundle-forming pilus and othertype IVb pili, p 4157. In Jarrell K (ed), Pili and flagella: current researchand future trends. Caister Academic Press, Norfolk, United Kingdom.

52. Nevesinjac AZ, Raivio TL. 2005. The Cpx envelope stress response affectsexpression of the type IV bundle-forming pili of enteropathogenic Esche-richia coli. J. Bacteriol. 187:672686.

53. Ng SY, et al. 2011. Genetic andmass spectrometry analyses of the unusualtype IV-like pili of the archaeon Methanococcus maripaludis. J. Bacteriol.193:804814.

54. Nickerson NN, et al. 2011. Outer membrane targeting of secretin PulDrelies on disordered domain recognition by a dedicated chaperone. J. Biol.Chem. 286:14221432.

55. Nouwen N, et al. 1999. Secretin PulD: association with pilot PulS, struc-ture, and ion-conducting channel formation. Proc. Natl. Acad. Sci.U. S. A. 96:81738177.

56. Nudleman E, Wall D, Kaiser D. 2006. Polar assembly of the type IV pilussecretin inMyxococcus xanthus.Mol. Microbiol. 60:1629.

57. Peabody CR, et al. 2003. Type II protein secretion and its relationship tobacterial type IV pili and archaeal flagella. Microbiology 149:30513072.

58. Pelicic V. 2008. Type IV pili: e pluribus unum? Mol. Microbiol. 68:827837.

59. Price NL, Raivio TL. 2009. Characterization of the Cpx regulon in Esch-erichia coli strain MC4100. J. Bacteriol. 191:17981815.

60. Ramboarina S, et al. 2005. Structure of the bundle-forming pilus fromenteropathogenic Escherichia coli. J. Biol. Chem. 280:4025240260.

61. Ramer SW, Bieber D, Schoolnik GK. 1996. BfpB, an outer membranelipoprotein required for the biogenesis of bundle-forming pili in entero-pathogenic Escherichia coli. J. Bacteriol. 178:65556563.

62. Ramer SW, et al. 2002. The type IV pilus assembly complex: biogenicinteractions among the bundle-forming pilus proteins of enteropatho-genic Escherichia coli. J. Bacteriol. 184:34573465.

63. Reichow SL, et al. 2011. The binding of cholera toxin to the periplasmicvestibule of the type II secretion channel. Channels (Austin) 5:215218.

64. Reichow SL, Korotkov KV, Hol WG, Gonen T. 2010. Structure of thecholera toxin secretion channel in its closed state. Nat. Struct. Mol. Biol.17:12261232.

65. Russel M, Linderoth NA, Aali. 1997. Filamentous phage assembly:variation on a protein export theme. Gene 192:2332.

66. Sambrook J, Fritsch EF, Maniatis T. 1989. Molecular cloning: a laboratorymanual. Cold SpringHarbor Laboratory Press, Cold SpringHarbor,NY.

67. Scaletsky ICA, Silva MLM, Trabulsi LR. 1984. Distinctive patterns ofadherence of enteropathogenic Escherichia coli to HeLa cells. Infect. Im-mun. 45:534536.

68. Schmidt SA, et al. 2001. Structure-function analysis of BfpB, a secretin-like protein encoded by the bundle-forming-pilus operon of enteropatho-genic Escherichia coli. J. Bacteriol. 183:48484859.

69. SchreiberW, et al. 2002. BfpU, a soluble protein essential for type IV pilusbiogenesis in enteropathogenic Escherichia coli. Microbiology 148:25072518.

70. Seydel A, Gounon P, Pugsley AP. 1999. Testing the 2 rule for lipo-protein sorting in the Escherichia coli cell envelope with a new geneticselection. Mol. Microbiol. 34:810821.

71. Shaner NC, et al. 2004. Improved monomeric red, orange and yellowfluorescent proteins derived from Discosoma sp. red fluorescent protein.Nat. Biotechnol. 22:15671572.

72. Stone BJ, Abu Kwaik Y. 1998. Expression of multiple pili by Legionellapneumophila: identification and characterization of a type IV pilin gene



and its role in adherence to mammalian and protozoan cells. Infect. Im-mun. 66:17681775.

73. Stone KD, Zhang H-Z, Carlson LK, Donnenberg MS. 1996. A cluster offourteen genes from enteropathogenic Escherichia coli is sufficient for bio-genesis of a type IV pilus. Mol. Microbiol. 20:325337.

74. Strom MS, Nunn DN, Lory S. 1993. A single bifunctional enzyme, PilD,catalyzes cleavage and N-methylation of proteins belonging to the type IVpilin family. Proc. Natl. Acad. Sci. U. S. A. 90:24042408.

75. Subach FV, et al. 2009. Photoactivatable mCherry for high-resolutiontwo-color fluorescence microscopy. Nat. Methods 6:153159.

76. Szeto TH, Dessen A, Pelicic V. 2011. Structure/function analysis ofNeisseria meningitidis PilW, a conserved protein that plays multiple rolesin type IV pilus biology. Infect. Immun. 79:30283035.

77. Tokuda H. 2009. Biogenesis of outer membranes in Gram-negative bac-teria. Biosci. Biotechnol. Biochem. 73:465473.

78. Valentine RC, Shapiro BM, Stadtman ER. 1968. Regulation of glutaminesynthetase. XII. Electron microscopy of the enzyme from Escherichia coli.Biochemistry 7:21432152.

79. Varga JJ, et al. 2006. Type IV pili-dependent gliding motility in theGram-positive pathogenClostridiumperfringens andother clostridia.Mol.Microbiol. 62:680694.

80. Varga JJ, Therit B, Melville SB. 2008. Type IV pili and the CcpA protein

are needed formaximal biofilm formation by the gram-positive anaerobicpathogen Clostridium perfringens. Infect. Immun. 76:49444951.

81. Viarre V, et al. 2009. HxcQ liposecretin is self-piloted to the outer mem-brane by its N-terminal lipid anchor. J. Biol. Chem. 284:3381533823.

82. Vogt SL, et al. 2010. The Cpx envelope stress response both facilitates andinhibits elaboration of the enteropathogenic Escherichia coli bundle-forming pilus. Mol. Microbiol. 76:10951110.

83. Wolfgang M, et al. 1998. PilT mutations lead to simultaneous defects incompetence for natural transformation and twitching motility in piliatedNeisseria gonorrhoeae.Mol. Microbiol. 29:321330.

84. Wolfgang M, van Putten JP, Hayes SF, Dorward D, Koomey M. 2000.Components and dynamics of fiber formation define a ubiquitous biogen-esis pathway for bacterial pili. EMBO J. 19:64086418.

85. Yamaguchi K, Yu F, Inouye M. 1988. A single amino acid determinant ofthemembrane localization of lipoproteins in Escherichia coli.Cell 53:423432.

86. Zahavi EE, et al. 2011. Bundle forming pilus retraction enhances enter-opathogenic Escherichia coli infectivity. Mol. Biol. Cell 22:24362447.

87. Zhang H-Z, Donnenberg MS. 1996. DsbA is required for stability of thetype IV pilin of enteropathogenicEscherichia coli.Mol.Microbiol. 21:787797.

Lieberman et al.


Documents

Outer Membrane Targeting, Ultrastructure, And Single Molecule Localization of the Enteropathogenic Escherichia Coli Type IV Pilus Secretin BfpB