doi:10.1016/j.jmb.2011.12.006 J. Mol. Biol. (2012) 415, 918–928

Contents lists available at www.sciencedirect.com

Journal of Molecular Biologyj ourna l homepage: ht tp : / /ees .e lsev ie r.com. jmb

A Structural Basis for Sustained Bacterial Adhesion:Biomechanical Properties of CFA/I Pili

Magnus Andersson 1, 2⁎, Oscar Björnham1, 3, Mats Svantesson 1,Arwa Badahdah 4, Bernt Eric Uhlin 2, 5, 6 and Esther Bullitt 7

1Department of Physics, Umeå University, SE-901 87 Umeå, Sweden2Umeå Centre for Microbial Research (UCMR), Umeå University, SE-901 87 Umeå, Sweden3Swedish Defence Research Agency (FOI), SE-906 21 Umeå, Sweden4Department of Oral Biology, Boston University School of Dental Medicine, Boston MA 02118, USA5Department of Molecular Biology, Umeå University, SE-901 87 Umeå, Sweden6The Laboratory for Molecular Infection Medicine Sweden (MIMS), Umeå University, SE-901 87 Umeå, Sweden7Department of Physiology and Biophysics, Boston University School of Medicine, Boston MA 02118-2526, USA

Received 20 September 2011;received in revised form2 December 2011;accepted 5 December 2011Available online9 December 2011

Edited by W. Baumeister

Keywords:enterotoxigenic Escherichiacoli;unwinding;optical tweezers;fimbriae;force spectroscopy

*Corresponding author. Departmenmagnus.andersson@physics.umu.seAbbreviations used: ETEC, entero

Escherichia coli; FMOT, force-measurlayer-to-layer; WLC, worm-like chaData Bank.

0022-2836/$ - see front matter © 2011 E

Enterotoxigenic Escherichia coli (ETEC) are a major cause of diarrheal diseaseworldwide. Adhesion pili (or fimbriae), such as the CFA/I (colonizationfactor antigen I) organelles that enable ETEC to attach efficiently to the hostintestinal tract epithelium, are critical virulence factors for initiation ofinfection. We characterized the intrinsic biomechanical properties andkinetics of individual CFA/I pili at the single-organelle level, demonstratingthat weak external forces (7.5 pN) are sufficient to unwind the intact helicalfilament of this prototypical ETEC pilus and that it quickly regains itsoriginal structure when the force is removed. While the general relationshipbetween exertion of force and an increase in the filament length for CFA/Ipili associated with diarrheal disease is analogous to that of P pili and type 1pili, associated with urinary tract and other infections, the biomechanicalproperties of these different pili differ in key quantitative details. Uniquefeatures of CFA/I pili, including the significantly lower force required forunwinding, the higher extension speed at which the pili enter a dynamicrange of unwinding, and the appearance of sudden force drops duringunwinding, can be attributed to morphological features of CFA/I piliincluding weak layer-to-layer interactions between subunits on adjacentturns of the helix and the approximately horizontal orientation of pilinsubunits with respect to the filament axis. Our results indicate that ETECCFA/I pili are flexible organelles optimized to withstand harsh motionwithout breaking, resulting in continued attachment to the intestinalepithelium by the pathogenic bacteria that express these pili.

© 2011 Elsevier Ltd. All rights reserved.

t of Physics, Umeå University, SE-901 87 Umeå, Sweden. E-mail address:.toxigenic Escherichia coli; CFA/I, colonization factor antigen I; UPEC, uropathogenicing optical tweezer; EM, electron microscopy; AFM, atomic force microscopy; LL,in; DFS, dynamic force spectroscopy; PBS, phosphate buffer solution; PDB, Protein

lsevier Ltd. All rights reserved.

919Biomechanical Properties of CFA/I Pili

Introduction

Escherichia coli are Gram-negative bacteria that area normal constituent of the intestinal tract micro-biota of humans and warm-blooded animals. Somepathogenic strains of enterotoxigenic E. coli (ETEC)cause diarrheal disease with severe symptoms anddehydration, leading to annual mortality of hun-dreds of thousands of children below the age of fivein developing countries.1 Our studies investigatethe structural features of the archetype ETEC pilus,colonization factor antigen I (CFA/I), which pro-motes host/pathogen interactions, thereby initiatinginfection. The aim of this research is to revealstructural clues for the development of novel drugsagainst this critical ETEC virulence factor.Adhesion pili are specialized to sustain attach-

ment of bacterial cells under the environmentalconditions surrounding their preferred host targettissue. For ETEC, this environment is the smallintestine. Fluid in the small intestine contains amixture of solutes, for example, carbohydrates,peptides, and lipids from ingested food, as well assecretions from the biliary tree and pancreas thatcontain various digestive enzymes (e.g., electrolytesand pH-regulating substances). The intestinal wallhas longitudinal and circular smooth muscle layersthat provide mixing and propulsive movement ofthis chyme for efficient digestion. Contractile ringsinterspaced along the intestine segment the chymeinto compartments and create a circular motion ofthe fluid. The propulsive movement results inforward fluid movement, and simulations haveshown that the contraction phase of the peristalticreflex generates pressure, shear stress, and a reversevortex-like flow of the chyme.2 Bacteria are there-fore exposed to a harsh environment that requiressophisticated adhesive tools, that is, the pili, tomaintain stable adhesion.CFA/I pili that adhere to the intestinal epithelium

are approximately 1-μm-long helical filaments witha diameter of 7.4 nm and 3.17 pilin subunits per turnof the helix.3 The major pilin subunit, CfaB, is anapproximately cylindrical protein that comprisesseven β-strands in an IgG-like structure.4 The N-terminal strand of one subunit fills a hydrophobicgroove of the preceding subunit, thereby producingstrong non-covalent interactions along the filament.This helical filament architecture is also seen in P piliand type 1 pili expressed on uropathogenic E. coli(UPEC) that infect the bladder and may reach theupper urinary tract and kidneys,5,6 in type 3 piliexpressed by Klebsiella pneumoniae that cause respi-ratory tract infections,7 and in E. coli S pili, which arecorrelated to neonatal meningitis and urinary tractinfections.8 Adhesion pili assembled via this “donorstrand exchange” mechanism9 provide effectivedamping against external shear forces by unwind-ing of their quaternary structure while leaving the

tertiary structure intact. This property providesdistribution of the shear forces among severalattached pili and thereby increases the adhesionlifetime.10,11 Thus, extension of pili by unwinding ofthe helical filament allows for motion withoutbreaking the binding structure. Since CFA/I piliare similar in architecture to UPEC-expressed pili,3

we believed that CFA/I pili were also capable ofunwinding their quaternary structure as well asregaining their original structure after exposure toforce, in a similar way to what has previously beenfound for other pili.8,12,13

In this work, we used data from three methodol-ogies in order to characterize and elucidate thefunction of CFA/I pili under force exposure. Force-measuring optical tweezers (FMOTs) were used tomeasure the force required to unwind an individualpilus at a single-organelle level. Data were alsocollected to assess bond kinetics during unwinding.The interactions between adjacent layers responsiblefor pilus stability were modeled and analyzed in thisstudy using the quaternary structure and orientationof subunits determined previously by a hybridapproach that combined crystallographic data withresults from electron microscopy (EM).14 Finally, theunwinding and retraction biomechanics were mod-eled by Monte Carlo simulations and fitted to thedata.Our results strongly suggest that the force needed

for pilus unwinding is a function of both subunit–subunit interactions and the pitch of the subunits.The limited subunit–subunit interactions as well astheir horizontal subunit orientation relative to thenormal direction of the applied force lead to the lowforces required for CFA/I pili's filament unwinding.Thus, physical properties of the pilus filamentfacilitate the sustained adhesion of ETEC bacteriain the gut and thereby facilitate initiation ofdiarrheal disease.

Results

Unwinding an individual pilus

CFA/I pili are normally observed in both thewound and the unwound state, as seen in the atomicforce microscopy (AFM) and transmission EMimages of cells expressing CFA/I pili shown inFig. 1. The biomechanical properties of CFA/I piliwere investigated with FMOT, to provide informa-tion on the force–extension relationship, bondopening rates, and detailed properties of intersubu-nit bonds. In order to avoid any effect of the distaltip and the adhesin, we nonspecifically bound apilus to a bead. Tensile stress was then applied byseparating the bacterium–bead complex as de-scribed in Materials and Methods. A typicalunwinding response of an individual CFA/I pilus

Fig. 2. CFA/I pili unwind under force. (a) A data plot ofthe elongation of a CFA/I pilus. The force is constant at∼8 pN due to sequential unwinding of the helix-likestructure. At 2 μm, the force increases almost linearly, andat ∼2.1 μm, the pilus detaches from the bead. (b) Ahistogram of the distribution of unwinding forces for allinvestigated pili (n=88) at steady-state conditions. Themean of the Gaussian fitting is 7.5±1.5 pN.

Fig. 1. CFA/I pili exhibit helical and extended filament conformations. (a) An AFM image of HMG11/pNTP119 cellsexpressing CFA/I pili shows that under normal growth and sample preparation conditions, pili still attached to bacteriaare found in the wound, helical state (white arrow) and the unwound, extended fibrillar state (cyan arrow). The scale barrepresents 1 μm for the large image and 250 nm for the insets, which are data taken from the positions of the arrows. (b)Transmission EM data of CFA/I pili isolated from ETEC bacteria exhibit wound and unwoundmorphologies, as seen in ahelical rod form (white arrowhead) and in a thin, extended form (white arrow), respectively. A rare cross-section view of apilus rod is noted by the black arrow. The inset images are 2× magnifications of the structures marked in the micrograph;inset images are filtered and contrast enhanced to highlight the structural differences. The scale bar represents 50 nmon theEM and 25 nm on the insets.

920 Biomechanical Properties of CFA/I Pili

under steady-state conditions is shown in Fig. 2a.The extension response can be divided into threeregions that represent different geometrical config-urations of the pili. Region I represents elasticstretching of the quaternary structure constitutingonly a small fraction of the total elongation. At themicroscopic level, it represents straightening of thehelix-like filament and an increase, without disrup-tion of the layer-to-layer (LL) interactions, of thedistance between the layers. Region II is entered athigher force where the layers are extended further,and the interactions between subunits in adjacentlayers are sequentially broken at a qualitativelyconstant threshold force, unwinding the helicalfilament. The sequential nature of the unwindingis caused by the higher probability of breakingbonds adjacent to an already broken bond, incomparison to bonds that are surrounded by intactLL bonds.15 This unwinding occurs at a force of∼7.5 pN for CFA/I pili, during which a filamentwith a helical structural design has been trans-formed under force into a much longer open helicalstructure in which subunits are only connected headto tail via the N-terminal extension. In the experi-ment shown in Fig. 2a, the pilus detached from thebead at ∼2.1 μm, resulting in a zero force response.A histogram of the unwinding forces for allinvestigated CFA/I pili is shown in Fig. 2b. Theaverage unwinding force, representing the strengthof the LL interaction, was found to be 7.5±1.5 pN(n=88 distinct CFA/I pili). This is substantiallylower than the average unwinding forces observed

in the same experimental setup with S, P, and type 1pili expressed by UPEC, which unwind at 21±2 pN,28±2 pN, and 30±2 pN, respectively.8,12

Fig. 4. CFA/I pili unwinding is well fit to aWLCmodel.Force measurement of a single CFA/I pilus at a speed of0.1 μm/s. The gray curve corresponds to the unwinding,whereas the black curve represents the rewinding. Both theunwinding and the rewinding curves show adip in force atthe transition between regions III and II. A WLC model isfitted (broken line) to the rewinding curve, which is in itslinearized regime until ∼1.9 μm.

Fig. 3. CFA/I pili have many fewer LL contacts ascompared to P pili. Crystal structures of pilin subunits areshown modeled into their respective helical pilus rods;subunit interactions (residue–residue) are visualized withblue lines. (a) CfaB/CfaB has 7 residue–residue contacts(43 atom–atom interactions) when subunits are fitted intothe EMmap (see Supplementary Fig. 1 and Ref. 3 for the fitof CfaB into the CFA/I pilus helical reconstruction map),whereas (b), which represents a PapA/PapA-modeledstructure from Ref. 16, displays 28 residue–residuecontacts (124 atom–atom interactions).

921Biomechanical Properties of CFA/I Pili

LL interactions and subunit orientation

The force required for the unwinding of CFA/Ipili in region II correlates with weak LL interactionsbetween subunits. Our analyses of the crystalstructure of CfaB modeled into the helical recon-struction CFA/I pili from EM data (SupplementaryFig. 1) show that the surface contact area in CFA/Ipili is 710 Å2, as compared with 1150 Å2 for PapA inP pili (Fig. 3). If the force required to pull apartadjacent layers of the pilus had increased linearlywith increased surface area, our data would predictthat the difference in force to unwind CFA/I piliwould be only approximately 60% less than that of Ppili. Instead, the measured decrease is 3.7-fold. Tolook at the interactions in more detail, we examinedthe specific interactions between residues thatproduce LL bonds. As CFA/I pili and P pili arehelical filaments with 3.17 and 3.28 subunits perturn, respectively, LL interactions could occurbetween subunits that are 2, 3, and 4 subunitsfarther along the helical axis. As shown in Fig. 3,results calculated by van der Waals overlaps ≥−4 Åshow that CFA/I pili have 7 residue–residueinteractions between subunits, whereas P pili have28 interactions. As seen in Supplementary Table 1, inCFA/I pili, all 7 interactions are between the n and n+3 subunit, whereas in P pili, there are 11 in-teractions between n and n+3 and 17 interactionsbetween n and n+4.In addition to the specific residue–residue in-

teractions discussed above, the orientation of thepilin subunits in the helical filament also appears toplay a role in the force required for unwinding. Theforce available for unwinding the pilus filament isthe component of force normal to the subunit–subunit interaction between layers. Thus, when thepulling direction is along the filament axis, the

orientation of subunits with respect to this axisdefines the proportion of the exerted force, tensilestress, available for unwinding. For CFA/I pili, thesubunits are oriented only 4° from horizontal,resulting in almost all of the force contributing tounwinding, which requires 7.5 pN of force forsteady-state unwinding. P pili on the other hand areoriented 13° from horizontal and require a force of28 pN.15 In contrast to the orientations of thesesubunits, the subunits of Hib pili are oriented 82°from horizontal. While FMOT experiments have notbeen performed on Hib pili from Haemophilusinfluenzae bacteria, in EM data, these pili are neverseen to unwind.17 These data indicate that the forcesrequired for purification, including shearing of Hibpili from the bacteria using a blender, are notsufficient to unwind these filaments that havealmost vertically oriented subunits.

Linearized pili

The fluctuations of the constant force level ofCFA/Ipili during LL unwinding are larger than thefluctuations that have been seen for other pili. Asseen in Fig. 2, the force required for unwinding the LLinteractions of CFA/I fluctuated, on average, ∼2 pNabout the mean, in comparison to ∼1 pN for P and Spili. This fluctuation not only is larger but alsorepresents amuch larger fraction of the force requiredfor unwinding, 2 pN/7.5 pN=26% for CFA/I pili ascompared with 1 pN/28 pN=4% for P pili.The higher force response, denoted region III in Fig.

2a, corresponds to elastic stretching of the pilus once itis already in its linearized form. A linearized pilus cansuitably be described, since the units comprising along filament are much smaller than the filament'scontour length, as a semiflexible polymer. However,subunits in a pilus are connected head to tail with aconstraint regarding the degree of rotation, wherefore

922 Biomechanical Properties of CFA/I Pili

a free-jointed chain model does not acceptablyreproduce the force–extension response. Instead, apilus behaves more like a continuously flexibleuniform rod, in which the units act together as a“worm-like chain”, 18 as previous work hasshown.10,19 The almost linearly increasing force seenin region III originates from the extension of a worm-like chain (WLC) for a limited interval in proximity toits contour length. In Fig. 2a, the CFA/I pilus startedits elastic response at∼2.0 μmwith an almost linearlyincreasing force that suddenly dropped to zero at∼2.1 μm when the pilus detached from the bead.Force–extension curves in measurements with forcesup to the highest forces that can be achieved at highprecision with the current experimental method, 60–70 pN, always showed an approximately linearresponse in region III.

Fig. 5. Lack of a linker region may limit changes in thequaternary structure of CFA/I pili under force. Crystal-lographic structures of two adjacent CfaB subunits (PDBID: 3F84) and two adjacent PapA subunits (PDB ID: 2UY7)are shown with prolines, which can isomerize under thepresence of force (colored gold); black arrows point to theconnecting region between subunits. Adjacent CfaB sub-units (a) are closely spaced with no residues available toform a flexible hinge; thus, subunits have limited motionwithout steric clashes. No subsequent conformationchange is expected after the transition of CFA/I pilifrom kinked to linear filament, irrespective of whether thetransition arises from a force-induced proline isomeriza-tion within the N-terminal extension of CfaB. The longerlinker region between PapA subunits (b) allows greatflexibility via this hinge region, independent of anyproline isomerization.

Persistence length of an unwound CFA/I pili andcomparison with P pili

It is possible to analyze the force-rewindingbehavior of an individual pilus by reversing thedirection of motion of the probing bead. An exampleof data from a pilus being unwound (gray) andcontracted (black) is shown in Fig. 4. The two curvesoverlap, indicating that the measurement is in asteady-state regime.15 However, the rewindingcurve shows a small dip in force, ∼4 pN, at∼1.9 μm. Such dips are always observed for P andtype 1 pili12 and originate from the lack of anucleation kernel; that is, there are no layers formedthat can wind the subunits into the helix-like shape.Therefore, linearized pili require a certain amount ofslack before an n subunit and an n+3 subunit are inclose enough proximity to form an LL interaction.This particular behavior suggests that the CFA/I piliare assembled using a mechanism similar to that ofP pili.To quantify the persistence length of the unwound

CFA/I pili, we fitted a WLC (red broken line in Fig.4) to experimental data with Eq. (1). The persistencelength, which is a measure of the resistance towardsbending, was assessed to 4.5±1.4 nm, which isslightly longer than for previously assessed valuesof P and type 1 pili, 3.2±0.6 and 3.3±1.6 nm,respectively.10,19We then sought a plausible explanation for the

lack of an S-shaped transition, and therefore lack of aconformational change, when additional force wasapplied to CFA/I pili that were in region III of thecurve. As seen in Fig. 5, in an extended conforma-tion, the n to n+1 CfaB subunits are closely apposed,correlating with less flexibility than between PapAsubunits The data support CFA/I pili adopting amore rigid structure than P, S, F1C, or type 1 pili bythe end of region II. This more rigid fiber is capableof resisting the conformational change that is

visualized as the S-shaped force–extension curve inregion III seen for P, S, F1C, and type 1 pili.We then compared the architecture and connec-

tions between adjacent units of the major subunitCfaB in CFA/I pili to PapA in P pili, to look moreclosely at the molecular details, and noticed bothsimilarities and differences. A crystallographicstructure of two adjacent CfaB subunits and twoadjacent PapA subunits revealed a possible expla-nation of the transition that is observed in region IIIof P pili unwinding but not in CFA/I pili. In Fig. 5,adjacent PapA subunits are moderately spaced withresidues between subunits that can form a hinge,allowing flexibility between PapA subunits. Con-versely, adjacent CfaB subunits are closely spaced,with no residues available to form a flexible hinge.This close apposition limits the ability of thesubunits to rotate with respect to each other withoutsteric clashes.We calculated the persistence length, R, of the

wound CFA/I helical rod morphology from

Fig. 6. CFA/I pili unwind under steady-state condi-tions. Experimental data (gray line) and a fit with a two-state force–extension model (black line) to a force curve ina steady-steady regime. The model excellently reproducesthe experimental data.

923Biomechanical Properties of CFA/I Pili

electron micrographs. The variable curvature of arod gives an indication of the rigidity of its structureand thus the strength of the LL interactions. Amicrograph of negatively stained CFA/I is shown inFig. 1b. It is seen that the filaments have an inherentflexibility along their length, rather than adopting along, straight morphology. It was found that theaverage persistence length for CFA/I was 1.4 μm,whereas previous work indicated that P pili arerather straight with an R of 8 μm.6 EM data alsoreveal the ease with which CFA/I pili unwind, asnoted by the presence of unwound filaments inessentially all fields of view of CFA/I micrographs,in contrast to their relatively rare appearance in Ppili6 and their total absence in Hib pili.17 Theseresults are reproduced in our AFM imaging analysisof bacteria expressing CFA/I pili, as unwound piliare observed on some imaged cells (e.g., Fig. 1a).

Fig. 7. DFS of the LL bonds. The horizontal broken linecorresponds to the steady-state unfolding force whereasthe tilted broken line is a fit to the data that corresponds tothe dynamic regime. The vertical broken line representsthe velocity (corner velocity) at which the elongation goesin to a dynamic regime. For CFA/I, that velocity is1400 nm/s.

Force extension model

The complete force–extension response was mod-eled by Monte Carlo methods, similar to the methodpresented in Ref. 19. The pili extension modelincluded elastic stretching, a WLC model, and atwo-state energy landscape where the kinetics weredescribed via a stochastic probability functionweighted by the Arrhenius factor. A fit (blackcurve) of the model to an experimental force curve(gray curve) is shown in Fig. 6. Clearly, the modelexcellently reproduces the force–extension behaviorof a CFA/I pilus.

Dynamic force spectroscopy

A series of dynamic force spectroscopy (DFS)measurements is shown in Fig. 7. Each data pointrepresents the unwinding force, that is, force ofregion II, for a given elongation speed,

:L. In this

study, we performed DFS at 0.1, 1, 2, 4, 8, 16, and32 μm/s. The horizontal broken line is a fit to thedata points representing steady state, defined as theregion for which the force required to unwind thefilament is not dependent on the elongation speed.The tilted broken line is a fit to F(

:L)=kBT/xAT ln

(:L/

:Lth), where xAT is the bond length and

:Lthis the

thermal elongation speed.15 In this equation, theforce required for unwinding increases as theelongation speed increases. As a result, the bondlength can be derived from the inclination (slope) ofthe fitted curve. The intersection of the broken linescorresponds to the corner velocity,

:L4, which is the

elongation speed that separates the steady-stateregime from the dynamic regime. This was assessedto 1400±200 nm/s.

Discussion

Comparison of the mechanical propertiesof CFA/I pili from ETEC with helix-likepili expressed by UPEC

It has been shown that pilus unwinding enhancesbacterial adhesion to host cells by distributing theforces acting on a bacterium in a flow to severalpili.11 It is also likely that this unwinding propertyhas coevolved with the adhesin to optimize theadhesion lifetime of the bacterium.20 From themeasurements presented here, it is concluded thatCFA/I pili can also unwind and that the biome-chanical model developed in Refs. 15 and 21describes the data well. Since the structure of thesepili is principally the same as that of type 1 and Ppili, this was to be expected a priori. However, therewere remarkable biophysical differences betweenCFA/I and the previously studied helix-like pili

Table 1. Mechanical parameter values of CFA/I, S, P, type 1, and type 3 pili

Strain/plasmid/pilus type

HMG11/pNTP119/CFA/I

HB101/pANN801-13a/SI(sfaXtruncated)

HB101/pPAP5b/P

HB101/pPKL4b/type 1

HB101/pBSN50c/F1C

JM109/pmrkABCDd/type 3

Fuw (pN) 7.5±1.5 21±2 28±2 30±2 26±1 66±4ΔxAT (nm) 1.1±0.1 0.56±0.14 0.76±0.11 0.59±0.06 0.38±0.04 —:L4 (nm/s) 1400±200 700±100 400±100 6±3 1400±160 —

Fuw is the force at which the pilus unwinds.a Taken from Ref. 8.b Taken from Ref. 12.c Taken from Ref. 13.d Taken from Ref. 7.

924 Biomechanical Properties of CFA/I Pili

expressed. Mechanical parameters are summarizedin Table 1. For example, the unwinding force of S, P,F1C, and type 1 pili are within 21–30 pN, that is,∼3–4 times higher than that of CFA/I, whereas type 3pili expressed by K. pneumoniae unwind at ∼67 pN,that is, almost 10-fold higher than that of CFA/I.

Relation of the unwinding force to the LL contactarea and subunit orientation

As reported earlier,3 and also shown in Fig. 3, then-to-n+3 interaction of CFA/I compared to P pili issignificantly weaker, which correlates well with thelower unwinding forces measured with FMOT,since the unwinding force is directly connected tothe magnitude of the LL interactions that stabilizethe quaternary helix-like structure.3 However, thefindings in this work indicate that the limitedsurface area for the n-to-n+3 interaction is notsufficient to account for the significantly lowerunwinding force in comparison with that of P pili.Comparison of the contact area data of CFA/I and Ppili LL interactions predicted that∼60% of force wasneeded to unwind a CFA/I pilus as compared tounwinding of a P pilus. Nevertheless, the measuredunwinding force in situ, with FMOT, was ∼3.7-foldsmaller. To account for the large apparent discrep-ancy, the results suggest that both the number ofsubunit–subunit interactions and the pitch of thesubunits have a large influence on how much axialforce a pilus can withstand.We therefore propose that the very number of

residue–residue interactions and the subunit orien-tation relative to the tensile stress are contributingfactors for the reduced force needed to unwindCFA/I pilus filaments, as compared to P pili andHib pili. That is, the P pilus withstands strongerexternal forces before unwinding, due to itsincreased LL interactions (surface contact area of∼1150 Å2, including 28 residue–residue interac-tions) and, to a lesser degree, the 13° pitch of thesubunits from horizontal, as compared to the 4°subunit pitch and very limited contact area(∼710 Å2, including 7 residue–residue interactions)

for CFA/I pili. These data are consistent with allpresently available results from unwinding experi-ments performed with optical tweezers. This studyalso shows the benefit of using single-molecule forcespectroscopy techniques, such as FMOT, as acomplement to the static data obtained from, forexample, crystallographic and EM images, to gainvaluable information otherwise not accessible.Not only is the unwinding force for CFA/I much

lower than that for helix-like pili expressed by UPECP pili and S pili, the peak-to-peak value of the forceis also higher. There are two possible explanationsfor the high fluctuations in the qualitatively constantforce measured throughout region II. First, when asubunit–subunit bond opens/closes, the instanta-neous force change measured by the FMOT dependson the increase in length, ΔxAB, from the bondopening, and the force subsequently returns to itssteady-state value. That is, the measured peak-to-steady-state force will increase with increasedelongation distance per bond opening. Therefore, ifthe opening of a CFA/I subunit results in a largermomentary elongation change than ΔxAB seen for P,S, and type 1 pili, the average fluctuations in forcewould consequently be larger. Second, the opening/closing rates of the subunit–subunit bonds arecorrelated with the extent of LL interactions, withweaker interactions giving rise to faster opening/closure rates. If several subunits open during thetime scale of the experiment, they would beregistered as a single event, which would result inlarger fluctuations in the measured force. Data fromEM show weak LL interactions for CFA/I pili andsmaller contact surface area between layers, so thatthe second explanation, increased frequency ofsubunit–subunit bonds opening, is expected to bethe major mechanism underlying the large forcefluctuations.

Linearized pili

FMOT allows free rotation of the tethered objectduring force measurements. The multiple peptidebonds in the hinge region between subunits of P

925Biomechanical Properties of CFA/I Pili

pili could, therefore, rotate during the extension inregion II. CFA/I pili, with no linker regionbetween subunits, is more sterically constrainedand may require a proline isomerization forrotation to a linear fiber.3 Such an isomerizationcan be conformationally induced in the absence ofan enzyme to catalyze this reaction,22 but the forcerequired is larger than that exerted by FMOT inregion II. This inability to rotate individual sub-units would account for the extended, kinkedhelical form of CFA/I observed by EM. In thismodel, CFA/I pili would remain a kinked helix asit enters region III, with the filament direction notalways co-axial with the fibril axis. The FMOTforce exerted along the helix axis is distributed intoa component that follows the direction of thefilament and an axial component. It is only theforce component in the direction of the filamentthat is available to break the H-bonds of the β-strand/β-strand interaction and thereby slide theN-terminal extension of the n+1 subunit out ofthe groove of the n subunit. Therefore, anextended but kinked fibril such as CFA/I as itenters region III would require a stronger force tobreak apart a CFA/I pilus than would berequired to break a P pilus.At the end of region II, there is sometimes a dip of

2 pN in the force prior to its linear increase in regionIII (data not shown). We propose that this diprepresents a rotation of the tethering bead in thetrap. That is, in some cases, a twist is accumulated inthe CFA/I pilus as the LL bonds break but thesubunits do not turn. At the end of the unwindingregion, the tethering bead rotates in the trap, toreduce the stored energy just as the pilus is enteringregion III of the force curve.The increasing force seen in region III is an effect

of the inherent resistance of the open helicalstructure to be linearized. In this region of theforce–extension curves of P, S, and type 1 pili, allshowed an S-shaped curve. This pseudo-elastic,nonlinear force–extension response has been attrib-uted to a conformational change of individualsubunits at a force interval of 45–70 pN, and thisresponse perfectly matches the sticky-chain theoryfor linearized helix-like pili.21 Polymers that meetthe criteria of a “sticky chain” are those composed ofunits weakly connected by joints that are altered byforce in an individual or cooperative manner.23 Apilus under stress satisfies these criteria. For P pili inregion III, a conformational change of individualsubunits is clearly apparent from the S shape of thecurve in this region.12 CFA/I pili, however, do notshow such a conformational change for forces up to70 pN, providing evidence that CFA/I pili maintaina stable structure at high forces, with both apersistence length that correlates with the largestmeasured resistance to bending and a short N-terminal extension of CfaB pilins. This stability of

the subunit structure occurs despite the muchweaker LL bonds that hold together the helix-likestructure of the wound pilus filament.

Biological relevance

Persistence length measurements show that CFA/I pili are, of those pili measured to date, the mosteasily bent when in their helical rod form, and yetare the most rigid when they are unwound into theirextended, fibrillar conformation. The high level offlexibility of the helical structure could be ofadvantage for two reasons. First, a flexible piluscan “search”, due to thermal fluctuations, forreceptors in a larger volume than a stiff pilus, thusincreasing the probability for adherence. Secondly,high level of flexibility allows for a more uniformdistribution of the shear forces exposed to abacterium than if the pilus were stiff. That couldbe of advantage in an environment where the flowtends to include back-and-forth motion, whichexposes a bacterium to normal forces that willapply a pressure in the axial direction of a pilus. Itwas shown in a study by Jeffrey et al. that theperistaltic motion in the ileum creates both shearforces, including reversal wall shear forces, andnormal forces.2 A pilus with a high degree offlexibility will thereby be able to bend with the flowand absorb the axial pressure by bending thestructure without breaking an LL interaction.The adhesins expressed by CFA/I and type 1 pili

are similar in tertiary structure and show shear-enhanced binding.24 In a recent study, it was shownthat a CFA/I pilus manifests shear-enhanced bind-ing to its erythrocyte receptors in a stepwise increasein shear up to 5.0 dynes/cm2.24 This value is in themiddle of the 1–10 dynes/cm2 range of shear forcesoriginating from the peristaltic activity in thegastrointestinal tract.2 The unwinding and kineticdata determined for both CFA/I and type 1 piliopen up an interesting study and discussion of whythe adhesins of these pili show shear-enhancedbinding when, at the same time, the rod exhibits a4-fold difference in unwinding force and ∼250 timesdifference in kinetics.Findings in this article, through assessment of

data from multiple experimental techniques, allpoint to CFA/I pili having the absolutely weakestLL bonds of all helix-like pili investigated so far.Compared to the P, S, and type 1 pili, the extensionspeed at which CFA/I pili transition from steady-state to dynamic unwinding is ∼4 times higher thanthat of P pili, ∼2–7 times higher than that of S pili,and ∼250 times higher than that of type 1 pili (seeRef. 8 for a comprehensive table). Although type 1pili are found on bacteria that colonize manydifferent environments, P pili are more commonlyfound on isolates of the upper urinary tract, andtype 1 pili are more commonly found on bacteria

926 Biomechanical Properties of CFA/I Pili

from the lower urinary tract. The environmentalconditions in these regions are different; in the upperurinary tract, the urine is transported in boluses viaa peristaltic activity, whereas in the lower tract,urine is expelled from the bladder via the urethra ina more continuous manner.25,26 Thus, differentenvironmental conditions in the urinary tract alsorequire dissimilar types of adhesins and rods foroptimized bacterial adherence. Since the environ-mental conditions in the gut are very differentcompared to any portion of the urinary tract, it isplausible to state that the reduced unwinding forceand the fast kinetics of the CFA/I pili have evolvedin an optimal fashion for adhesion of the ETECbacteria.

Materials and Methods

Biological assay and FMOTs

The E. coli strain HMG11/pNTP11927 expresses 0.5–1 μm CFA/I pili on its cell surface as recorded by AFMshown in Fig. 1. Bacteria were grown on CFA agar platesat 37 °C overnight. Prior to an experiment, a colony fromthe agar plate was harvested and suspended in 500 μl offiltered phosphate buffer solution (PBS; 10 mM phosphateand 130 mM NaCl, pH 7.4, Sigma-Aldrich).Substrates were prepared by adding a 250-μl solution of

9.7-μm polystyrene beads (Duke Scientific Corp., Palo Alto,CA) diluted with filtered MilliQ H2O onto coverslips thatwere then placed in an oven for 60 min at 60 °C. The beadswere immobilized to the surface and later functionalizedwith2 μg/slide of poly-L-lysine (Sigma-Aldrich) and workedthereby as mounts for bacteria during an FMOT measure-ment to prevent optical damage of the bacteria by directcontact with the laser beam. This functionalization createsstrong electrostatic bonds with the bacterium and ensuresthat the bacterium is immobilized during a pilus force–extension experiment.A25-μl droplet containingbacteria and3.0 μm polystyrene beads (Duke Scientific Corp.) used asforce transducers was added on top of the immobilizedbeads. Finally, the chamberwas closed by adding a top coverslide and placed in a custom-made sample holder in themicroscope. The experiments were conducted at 25 °C.The optical tweezers system and measurement procedure

have previously been described in Refs. 15 and 28, and theFMOT studies were performed by the samemethod as thosecarried out on P, S, and type 1 pili.8,12 In a typical experiment,a bacteriumwas trapped andmounted on the 9.7-μmbead atlow laser power to prevent optical damage.29 A 3.0-μmbeadwas subsequently trapped at high laser power, and thetrapping constant, derived by the power spectrummethod,30

was in general ∼120 pN/μm. The extension/retractionvelocity for steady-state measurements was kept at0.1 μm/s, whereas dynamic measurements of region IIwere conducted at 0.1, 1, 2, 4, 8, 16, and 32 μm/s. All datawere compensated for Stokes drag force, and Faxen's lawwas used for adjustment of trapping near a surface (∼5 μm).In this work, we focus on exploring the biomechanical

properties of the rod. Therefore, in order to avoid anyeffect of the distal tip and the adhesin, we nonspecifically

bound a pilus to a bead by the method described inRef. 15. The presented force–extension data originate froma total of 88 measurements from 88 distinct pili. Theintervals for the values of the unwinding forces and thepersistence lengths correspond to the standard deviationsof the statistical data sets, respectively.

Structural comparison

Subunit orientations were determined from the three-dimensional structures of CFA/I pili, P pili, andHib pili,withtheir respective major subunits (CfaB and PapA) or homol-ogous subunit (HifA) fitted as published previously4,16,17 intothemaps; for convenience, the fit of CfaB into theCFA/Ipilusstructure is shown in Supplementary Fig. 1. Subunit–subunitcontacts were determined after fitting of the CfaB subunit[ProteinData Bank (PDB) ID: 3F84] into theCFA/I pili helicalreconstruction map or the PapA homology-modeled andoptimized subunit16 into thePpili helical reconstructionmap.Contact criteria were −0.4 Å van der Waals overlap, andcontact area and specific residue–residue interactions werecalculated using UCSF Chimera software.31

CFA/I pili purification and EM imaging

Bacteria expressing CFA/I pili were pelleted and resus-pended in 3 ml of PBS per gram of wet cell pellet weight. Piliwere heat extracted at 65 °C for 25 min, and cells wereremovedby centrifugation at 10,000g for 30min.Ammoniumsulfate was added to the supernatant, 0.24 g per ml, androcked for a minimum of 2 h at room temperature.Precipitated pili were pelleted by centrifugation at 12,000gand resuspended in PBS, and the ammonium sulfateprecipitation procedure was repeated. The resuspended piliwere then dialyzed against TE (10 mM Tris, pH 7.4, with0.1 mM ethylenediaminetetraacetic acid).For imaging, CFA/I pili were placed on glow-dis-

charged, carbon-coated grids; washed with TE; andstained with 1% uranyl acetate. Images were recorded at120 kV in a Philips CM12 electron microscope on KodakSO163 film and scanned on a Nikon 9000 scanner.

Two-state model for a helix-like polymer

We used the Monte Carlo method to stochasticallysimulate the force–extension behavior of a single pilus.The method is built on the Bell–Evans theory for bondkinetics in combination with a WLC model, similar tothe one described in Refs. 19 and 32. In short, the Bell–Evans theory states that the probability for a bondtransition from state A to state B is described by the off-rate, koff, as

koff = kthoffefxbkBT ð1Þ

where koffth is the thermal off-rate, f is the force applied to

the bond, kB is Boltzmann's constant, and T is theabsolute temperature. The applied force thus changesthe bond transition rate that is equivalent, in our case, tothe unwinding rate of the individual subunits. Theforce–extension relation is described by an elastic model

927Biomechanical Properties of CFA/I Pili

in combination with a WLC that is commonly used todescribe macromolecules in thermodynamic equilibrium.In the WLC, the force is coupled to the end-to-endlength of the unwound part of the pilus as

f =kBTp

14

1−LLc

� �−2

−14+

LLc

" #ð2Þ

where p is the persistence length and Lc is the contourlength of the unwound part of the pilus.

Atomic force microscopy

AFM imaging of bacteria with pili was done essen-tially as described earlier with some modifications.33

Bacterial cells were suspended in 50 μl of filtered waterbefore 10 μl was placed onto freshly cleaved ruby redmica (Goodfellow Cambridge Ltd., Cambridge). Thecells were incubated for 5 min at room temperature andblotted dry before they were placed into a desiccator fora minimum of 2 h. Images were collected with aNanoscope V Multimode AFM equipment (Veecosoftware) using TappingMode™ with standard siliconcantilevers oscillated at resonant frequency (270–305 kHz). Images were collected in air at a scan rateof approximately 0.5–1.5 Hz.Supplementary materials related to this article can be

found online at doi:10.1016/j.jmb.2011.12.006

Acknowledgements

This work was supported by a Young ResearcherAward (Karriärbidrag) from Umeå University (toM.A.) and by grants from the Swedish ResearchCouncil (to B.E.U.) and the National Institutes ofHealth (GM055722 to E.B.). We are grateful to Mrs.Monica Persson for excellent technical assistance andtoDr. Stephen Savarino andMs. AnnetteMcVeigh fordevelopment of the CFA/I pili isolation protocol.

References

1. Nataro, J. P. & Kaper, J. B. (1998). DiarrheagenicEscherichia coli. Clin. Microbiol. Rev. 11, 142–201.

2. Jeffrey, B.,Udaykumar,H. S.&Schulze,K. S. (2003). Flowfields generated byperistaltic reflex in isolated guineapigileum: impact of contraction depth and shoulders.Am. J.Physiol.: Gastrointest. Liver Physiol. 285, G907–G918.

3. Mu, X. Q., Savarino, S. J. & Bullitt, E. (2008). The three-dimensional structure of CFA/I adhesion pili: trave-ler's diarrhea bacteria hang on by a spring. J. Mol. Biol.376, 614–620.

4. Li, Y. F., Poole, S., Nishio, K., Jang, K., Rasulova, F.,McVeigh, A., Savarino, S. J. et al. (2009). Structure ofCFA/I fimbriae from enterotoxigenic Escherichia coli.Proc. Natl Acad. Sci. USA, 106, 10793–10798.

5. Hahn, E., Wild, P., Hermanns, U., Sebbel, P.,Glockshuber, R., Haner, M. et al. (2002). Exploring

the 3D molecular architecture of Escherichia coli type 1pili. J. Mol. Biol. 323, 845–857.

6. Bullitt, E. & Makowski, L. (1998). Bacterial adhesionpili are heterologous assemblies of similar subunits.Biophys. J. 74, 623–632.

7. Chen, F. J., Chan, C. H., Huang, Y. J., Liu, K. L., Peng,H. L., Chang, H. Y. et al. (2011). Structural andmechanical properties of Klebsiella pneumoniae type 3fimbriae. J. Bacteriol. 193, 1718–1725.

8. Castelain, M., Sjöstrom, A. E., Fällman, E., Uhlin, B. E.& Andersson, M. (2010). Unfolding and refoldingproperties of S pili on extraintestinal pathogenicEscherichia coli. Eur. Biophys. J. Biophys. Lett. 39,1105–1115.

9. Sauer, F. G., Pinkner, J. S., Waksman, G. & Hultgren,S. J. (2002). Chaperone priming of pilus subunitsfacilitates a topological transition that drives fiberformation. Cell, 111, 543–551.

10. Miller, E., Garcia, T. I., Hultgren, S. & Oberhauser, A(2006). The mechanical properties of E. coli type 1 pilimeasured by atomic force microscopy techniques.Biophys. J. 91, 3848–3856.

11. Björnham, O. & Axner, O. (2009). Multipili attachmentof bacteria with helixlike pili exposed to stress. J.Chem. Phys. 130.

12. Andersson, M., Uhlin, B. E. & Fällman, E. (2007). Thebiomechanical properties of E. coli pili for urinary tractattachment reflect the host environment. Biophys. J. 93,3008–3014.

13. Castelain, M., Ehlers, S., Klinth, J., Lindberg, S.,Andersson, M., Uhlin, B. E. & Axner, O. (2011). Fastuncoiling kinetics of F1C pili expressed by uropatho-genic Escherichia coli are revealed on a single piluslevel using force-measuring optical tweezers. Eur.Biophys. J. Biophys. Lett. 40, 305–316.

14. Li, Y. F., Poole, S., Nishio, K., Jang, K., Rasulova, F.,McVeigh, A. et al. (2009). Structure of CFA/I fimbriaefrom enterotoxigenic Escherichia coli. Proc. Natl Acad.Sci. USA, 106, 10793–10798.

15. Andersson, M., Fällman, E., Uhlin, B. E. & Axner, O.(2006). Dynamic force spectroscopy of the unfoldingof P pili. Biophys. J. 91, 2717–2725.

16. Mu, X. Q. & Bullitt, E. (2006). Structure and assembly ofP-pili: a protruding hinge region used for assembly of abacterial adhesion filament. Proc. Natl Acad. Sci. USA,103, 9861–9866.

17. Mu, X. Q., Egelman, E. H. & Bullitt, E. (2002).Structure and function of Hib pili from Haemophilusinfluenzae type b. J. Bacteriol. 184, 4868–4874.

18. Bustamante, C., Marko, J. F., Siggia, E. D. & Smith, S.(1994). Entropic elasticity of lambda-phage DNA.Science, 265, 1599–1600.

19. Björnham, O., Axner, O. & Andersson, M. (2008).Modeling of the elongation and retraction of Escher-ichia coli P pili under strain by Monte Carlo simula-tions. Eur. Biophys. J. Biophys. Lett. 37, 381–391.

20. Forero, M., Yakovenko, O., Sokurenko, E. V., Thomas,W. E. & Vogel, V. (2006). Uncoiling mechanics ofEscherichia coli type I fimbriae are optimized for catchbonds. PLoS Biol. 4, 1509–1516.

21. Andersson, M., Fällman, E., Uhlin, B. E. & Axner, O.(2006). A sticky chain model of the elongation ofEscherichia coli P pili under strain. Biophys. J. 90,1521–1534.

928 Biomechanical Properties of CFA/I Pili

22. Valiaev, A., Lim, D. W., Oas, T. G., Chilkoti, A. &Zauscher, S. (2007). Force-induced prolyl cis–transisomerization in elastin-like polypeptides. J. Am.Chem. Soc. 129, 6491–6497.

23. Jager, I. L. (2001). The “sticky chain”: a kinetic modelfor the deformation of biological macromolecules.Biophys. J. 81, 1897–1906.

24. Tchesnokova, V.,McVeigh,A. L., Kidd, B., Yakovenko,O., Thomas, W. E., Sokurenko, E. V. & Savarino, S. J.(2010). Shear-enhanced binding of intestinal coloniza-tion factor antigen I of enterotoxigenic Escherichia coli.Mol. Microbiol. 76, 489–502.

25. Griffiths, D. J., Constantinou, C. E., Mortensen, J. &Djurhuus, J. C. (1987). Dynamics of the upper urinarytract. 2. The effect of variations of peristaltic frequencyand bladder pressure on pyeloureteral pressure/flowrelations. Phys. Med. Biol. 32, 823–833.

26. Ozawa, H., Kumon, H., Yokoyama, T., Watanabe,T. & Chancellor, M. B. (1998). Development ofnoninvasive velocity flow video urodynamics usingDoppler sonography. Part II: clinical application inbladder outlet obstruction. J. Urol. 160, 1792–1796.

27. Jordi, B. J. A. M., Dagberg, B., de Haan, L. A. M.,Hamers, A. M., van der Zeijst, B. A. M., Gaastra, W. &Uhlin, B. E. (1992). The positive regulator CfaDovercomes the repression mediated by histone-like

protein H-NS (H1) in the Cfa/I fimbrial operon ofEscherichia coli. EMBO J. 11, 2627–2632.

28. Fällman, E., Schedin, S., Jass, J., Andersson, M., Uhlin,B. E. & Axner, O. (2004). Optical tweezers based forcemeasurement system for quantitating binding interac-tions: system design and application for the study ofbacterial adhesion. Biosens. Bioelectron. 19, 1429–1437.

29. Rasmussen, M. B., Oddershede, L. B. & Siegumfeldt,H. (2008). Optical tweezers cause physiologicaldamage to Escherichia coli and Listeria bacteria. Appl.Environ. Microbiol. 74, 2441–2446.

30. Berg-Sorensen, K. & Flyvbjerg, H. (2004). Powerspectrum analysis for optical tweezers. Rev. Sci.Instrum. 75, 594–612.

31. Goddard, T. D., Huang, C. C. & Ferrin, T. E. (2007).Visualizing density maps with UCSF Chimera. J.Struct. Biol. 157, 281–287.

32. Björnham, O., Nilsson, H., Andersson, M. & Schedin,S. (2008). Physical properties of the specific PapG–galabiose binding in E. coli P pili-mediated adhesion.Eur. Biophys. J. Biophys. Lett. 38, 245–254.

33. Balsalobre, C., Morschhäuser, J., Jass, J., Hacker, J. &Uhlin, B. E. (2003). Transcriptional analysis of the sfadeterminant revealing multiple mRNA processingevents in the biogenesis of S fimbriae in pathogenicEscherichia coli. J. Bacteriol. 185, 620–629.