Commensal and pathogenic Escherichia coliuse a common ...pathogenic and commensal E. coli. pili enterohemorrhagic Escherichia coli normal flora B acterial adherence to host tissues

Commensal and pathogenic Escherichia coli usea common pilus adherence factor for epithelialcell colonizationMarıa A. Rendon*, Zeus Saldana*, Aysen L. Erdem*, Valerio Monteiro-Neto*†, Alejandra Vazquez‡, James B. Kaper§,Jose L. Puente‡, and Jorge A. Giron*¶

*Department of Immunobiology, University of Arizona, 1501 North Campbell Avenue, Tucson, AZ 85724; †Centro de Ciencias da Saude Centro UniversitarioMaranhao, Rua Josue Montelo No. 1, Sao Luıs 65075-120, Maranhao, Brazil; ‡Departamento de Microbiologıa Molecular, Instituto de Biotecnologıa,Universidad Nacional Autonoma de Mexico, Cuernavaca, Morelos 62210, Mexico; and §Department of Microbiology and Immunology, University ofMaryland School of Medicine, 685 West Baltimore Street, Baltimore, MD 21201

Communicated by Roy Curtiss, Arizona State University, Tempe, AZ, May 2, 2007 (received for review August 4, 2006)

Enterohemorrhagic Escherichia coli (EHEC) O157:H7 is a food-bornepathogen that causes hemorrhagic colitis and the hemolytic uremicsyndrome. Colonization of the human gut mucosa and productionof potent Shiga toxins are critical virulence traits of EHEC. AlthoughEHEC O157:H7 contains numerous putative pili operons, their rolein the colonization of the natural bovine or accidental human hostsremains largely unknown. We have identified in EHEC an adher-ence factor, herein called E. coli common pilus (ECP), composed ofa 21-kDa pilin subunit whose amino acid sequence corresponds tothe product of the yagZ (renamed ecpA) gene present in all E. coligenomes sequenced to date. ECP production was demonstrated in121 (71.6%) of a total of 169 ecpA� strains representing intestinaland extraintestinal pathogenic as well as normal flora E. coli.High-resolution ultrastructural and immunofluorescence studiesdemonstrated the presence of abundant peritrichous fibrillar struc-tures emanating from the bacterial surface forming physicalbridges between bacteria adhering to cultured epithelial cells.Isogenic ecpA mutants of EHEC O157:H7 or fecal commensal E. colishowed significant reduction in adherence to cultured epithelialcells. Our data suggest that ECP production is a common feature ofE. coli colonizing the human gut or other host tissues. ECP is a pilusof EHEC O157:H7 with a potential role in host epithelial cellcolonization and may represent a mechanism of adherence of bothpathogenic and commensal E. coli.

pili � enterohemorrhagic Escherichia coli � normal flora

Bacterial adherence to host tissues is a complex process that, inmany cases, involves the participation of several distinct ad-

hesins, all of which may act at the same time or at different stagesduring infection (1). Many pathogenic bacteria display polymericadhesive fibers termed ‘‘pili’’ or ‘‘fimbriae’’ that facilitate the initialattachment to epithelial cells and subsequent successful coloniza-tion of the host (1). Pili are virulence factors that mediate inter-bacterial aggregation and biofilm formation, or mediate specificrecognition of host-cell receptors (2). It is clear that pili play similarbiological roles for commensal bacteria because they also have tocolonize specific niches and overcome the host’s natural clearingmechanisms. It is thought that commensal and some pathogenicEscherichia coli strains use type I pili or curli to colonize human andanimal tissues (3, 4). However, analysis of the genome sequence ofE. coli K-12 and some pathogenic E. coli strains shows the presenceof multiple putative pili operons (5–8), suggesting that other pilimight be produced by pathogenic and normal flora E. coli (NFEC)in the human gut.

Enterohemorrhagic E. coli (EHEC) O157:H7 is a potentiallyfatal food-borne pathogen that can cause hemorrhagic colitis (9)and the hemolytic uremic syndrome (HUS) (10). Hallmarks ofEHEC pathogenicity are the production of intestinal attachingand effacing (AE) lesions (11) and the secretion of potent Shigatoxins, which are responsible for the HUS (10). Most of the

EHEC genetic elements responsible for the AE lesions arecontained within the pathogenicity island called the ‘‘locus ofenterocyte effacement’’ (LEE) (11). Whereas the LEE of en-teropathogenic E. coli can confer the ability to cause AE lesionswhen introduced into a laboratory E. coli K-12 strain, the clonedEHEC LEE cannot (12, 13). This variation is explained by recentreports that show that an EHEC-specific non-LEE effectormolecule is required for full development of AE lesions byEHEC (14, 15). EHEC strains are a subset of the Shiga-toxigenicE. coli (STEC) pathogroup. There are STEC strains that do notcontain the LEE region and that can cause severe disease inhumans, including HUS (16), indicating the existence of othernon-LEE virulence factors. Cattle, other farm animals, and wildanimals are important reservoirs of STEC strains of differentserotypes, including O157:H7 (17). However, the bacteria canonly cause disease in neonatal cattle (18).

EHEC strains bind to many cultured cell types (19), and to theintestine of gnotobiotic piglets, newborn rabbits, and neonatalcalves (20). Despite efforts to identify putative adhesins, the onlyfactor of EHEC demonstrated to play a role in colonization invivo is intimin (21, 22). Intimin is an outer-membrane proteinencoded on the LEE that mediates close adherence to theenterocyte cell membrane (23) via its own translocated receptor(Tir) (24), integrin (25), or a host cell protein called nucleolin(26). Other less well characterized surface adhesins have beenreported (27–31), but their role in in vivo adherence and hostcolonization remains elusive. The genome of EHEC O157:H7strains contains 16 loci-encoding genes putatively involved in pilibiosynthesis (6, 8, 32); however, what role they play in thepathogenic scheme of these organisms remains largely unknown.In this article, we demonstrate the production of a pilus inEHEC, other E. coli pathogroups, and NFEC isolates that wehypothesize is involved in promoting bacterial adherence andhost colonization.

ResultsDetection of Unique Pili on EHEC Adhering to Cultured Epithelial Cells.An ultrastructural approach by high-resolution SEM was under-taken to investigate whether any of the 16 loci of EHEC

Author contributions: Z.S. and A.L.E. contributed equally to this work; M.A.R., J.B.K., J.L.P.,and J.A.G. designed research; M.A.R., Z.S., A.L.E., V.M.-N., and A.V. performed research;J.B.K. and J.L.P. contributed new reagents/analytic tools; M.A.R., Z.S., A.L.E., J.L.P., andJ.A.G. analyzed data; and J.L.P. and J.A.G. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.

Abbreviations: AE, attaching and effacing; EHEC, enterohemorrhagic E. coli; ECP, E. colicommon pilus; HUS, hemolytic uremic syndrome; IB, immunoblotting; IF, immunofluores-cence; NFEC, normal flora E. coli; STEC, Shiga-toxigenic E. coli.

¶To whom correspondence should be addressed. E-mail: [email protected].

This article contains supporting information online at www.pnas.org/cgi/content/full/0704104104/DC1.

© 2007 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0704104104 PNAS � June 19, 2007 � vol. 104 � no. 25 � 10637–10642

MIC

ROBI

OLO

GY

Dow

nloa

ded

by g

uest

on

Mar

ch 1

6, 2

020

EDL933-encoding putative piliation genes direct the expressionof functional pili. Examination of EDL933 adhering to culturedepithelial cells (HEp-2 and HeLa) for a total of 6 h revealed theproduction of thin (4-nm-wide), f lexible fibers resembling pilithat extended several micrometers away from the surface of thebacilli (Fig. 1). The peritrichous pili-like structures had a ten-dency to intertwine and to coil together forming thicker (12-nm-wide) structures that seemed to promote bacteria-to-bacteria interactions. These filamentous bridges were evident onbacteria adhering to the epithelial cell surface as early as 1.5 hafter infection (Fig. 1 A) and throughout the 6-h duration of theexperiment (Fig. 1 B–D). At 3 h after infection, the bacteriaformed AE lesions and appeared embedded in concavities

formed on the eukaryotic cell surface (Fig. 1B). TransmissionEM analysis of the bacteria present in the supernatant of infectedcells showed peritrichous long (4-nm-wide) pili (Fig. 2A), whichwe hypothesized corresponded to those observed by SEM(Fig. 1).

Purification and Identification of the Pilin Subunit. To elucidate thenature of these fibrillar structures, EHEC EDL933 incubatedwith HEp-2 epithelial cells was collected for isolation of the pili.The major component of these pili was resolved as a 21-kDaprotein (Fig. 2B, lane 1) whose size and N terminus amino acidsequence matched that of the predicted protein encoded by theyagZ gene found in the genome of EHEC O157:H7 (6, 8), E. coliK-12 (5), uropathogenic E. coli (7), and meningitis-associated E.coli (33). In light of the apparent wide distribution of yagZ amongthe E. coli, we propose the generic name ‘‘E. coli common pilus’’(ECP) for the pilus composed by the subunit protein product ofthis gene, herein proposed to be renamed ‘‘ecpA.’’

Sequence analysis of the predicted EcpA subunit revealed that60% of the protein is hydrophobic (33) and its C terminus doesnot contain the typical two-cysteine residues present in many pilitypes (34). Rabbit polyclonal antibodies produced against puri-fied ECP, specifically detected the 21-kDa protein by immuno-blotting (IB) (Fig. 2B, lane 2), and decorated the purified ECPfilaments by immuno-EM (Fig. 2C). No reactivity was seen withheterologous anti-type-I pili antibody [supporting information(SI) Fig. 5A], demonstrating the specificity of the anti-ECPantiserum.

Demonstration of ECP on EHEC O157:H7 Adhering to Host EpithelialCells. Next, we sought to confirm the identity of the structuresproduced by EHEC adhering to cultured epithelial cells (Fig. 1)and also of the pili seen on bacteria recovered from thesupernatants by immuno-EM, immunofluorescence (IF), andimmuno-SEM. Bacteria recovered from the supernatants ofinfected cells displayed ECP that were decorated by anti-ECPantibodies (Fig. 2D). Bacteria adhering to HEp-2 cells producedabundant ECP, which were seen by IF as a specific f luorescent

Fig. 1. SEMs showing production of fibers by adhering EDL933. EDL933adhering to HEp-2 cells for 1.5 (A), 3 (B), 4.5 (C), and 6 (D) h were visualized bySEM. Note the presence of tethered fibers that create thicker structures, whichappear to form physical bridges between bacteria. (Scale bars: 0.1 �m.)

A

F

B C D

E

Fig. 2. Identification of the pili produced by adhering EDL933. (A) Pili (indicated by the arrow) produced by bacteria obtained from the supernatant of infectedHEp-2 cells. (B) SDS/PAGE of pili purified; lane 1, Coomasie blue staining showing the pilin subunit with an apparent molecular mass of 21 kDa; lane 2, reactivityof the pilin with anti-ECP antibodies. (C) Immuno-EM of purified ECP (Inset). (D) Immuno-EM of ECP produced by EDL933 recovered from the supernatants ofinfected HEp-2 cells. (E) IF showing production of ECP (green) by EDL933 adhering to HEp-2 cells. Bacterial and nuclear DNA was stained with propidium iodide(red). (F) Immuno-SEM using anti-ECP antibodies and anti-rabbit IgG conjugated to 30-nm gold particles (arrows). (Scale bar: 0.5 �m.)

10638 � www.pnas.org�cgi�doi�10.1073�pnas.0704104104 Rendon et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

6, 2

020

fibrillar pattern associated with the bacteria (Fig. 2E). Lastly,using immuno-SEM we demonstrated that the fibers tetheringthe adhering bacteria are ECP (Fig. 2F). The fibers were notdetected with anti-type-I-pili antibody and anti-curli antibodies,used as negative controls (Fig. 5 B and C). Attempts to inhibitadherence of EHEC by using anti-ECP antibodies were unsuc-cessful (data not shown). The bases for this result are unknown,although we cannot rule out the possibility that our antibodieshave no access for those ECP epitopes or regions involved inadherence.

Expression of ecpA and Environmental Regulation. Consistent withthe ultrastructural (Fig. 1) and kinetics studies of ECP produc-tion followed by IF (Fig. 3A), RT-PCR experiments showed thatecpA expression occurs in bacteria interacting with culturedepithelial cells between 1.5 and 6 h after infection, showing anincrease of 1.34-fold over time (Fig. 3B). Production of ECP isapparently driven from a putative operon comprised of six genes,among which four are predicted to encode proteins with homol-ogy to proteins involved in pili bioassembly (SI Table 2).

Expression of bacterial pili is under the influence of environ-mental cues and, sometimes, host factors (35). Previously, it wasshown that meningitis-associated E. coli strains, but not anyother E. coli pathogroups, were able to assemble the EcpAprotein (formerly YagZ) into pili named ‘‘Mat’’ (meningitis-associated and temperature-regulated pilus) only after growth at20°C in LB broth (33). EHEC is indeed able to assemble theEcpA protein into pili upon infection of cultured epithelial cellsat host temperature, a phenotype that is biologically significantfrom the host–pathogen interaction standpoint. We sought todetermine whether eukaryotic cells triggered ECP productionand what environmental cues might be signals to activate the ecpoperon. Adherence experiments performed with formalin-killedHEp-2 cells or with cells separated from the bacteria by a 0.2-�mfilter showed similar levels of production of ECP (SI Table 3),suggesting that a host-cell product or direct contact of thebacteria with host cells are not required for induction.

We then compared ecpA transcription and production of ECPafter growth of EHEC in LB versus DMEM at 26°C versus 37°Cand with aeration versus 5% CO2 atmosphere versus anaerobi-osis. No ECP was observed in LB medium in any of theconditions examined (SI Table 3 and Fig. 3 C and D). In general,bacterial growth in DMEM at 26°C yielded higher levels ofexpression of the pili compared with growth at 37°C; thisproperty could be relevant during the life of the bacteria outsidethe bovine or human hosts. The presence of 5% CO2 was

generally a stimulator of ECP production at either temperature(SI Table 3 and Fig. 3 C and D). The level of ecpA transcriptionfound under the conditions examined was in line with thephenotypic data. Thus, like many other pili types, production ofECP is subjected to regulation by environmental cues such astemperature, oxygen tension, and growth media. The molecularmechanisms that modulate ECP expression are unknown andneed to be investigated.

Distribution of ecpA Among E. coli Strains. To assess the distributionof ecpA among E. coli strains from different sources, we per-formed a PCR-based ecpA survey in a collection of 176 strainsrepresenting NFEC and the major E. coli pathogroups (EHEC,enteropathogenic, enterotoxigenic, enteroaggregative, enteroin-vasive, rabbit pathogenic, avian pathogenic, and uropathogenic).Using primers G84 and G85 (SI Table 4), which derive from the5�- and 3�-ends of ecpA, we found that this gene was present in169 (96%) of these strains (data not shown), supporting thenotion that this locus is highly common among intestinal andextraintestinal E. coli strains. We investigated whether theremaining 4% of the strains lacked ecpA or possessed geneticdifferences in the ecp operon that would account for theirnegativity with the primers used. We performed a multiplex PCRby using primers (SI Table 4) for internal sequences of ecpR,ecpA, ecpB, and ecpC, and found that these genes were absent inall cases (SI Fig. 6), indicating that the remaining 4% of thestrains actually lack the ecp operon.

Production of ECP by EHEC and non-EHEC E. coli. We investigated theproduction of ECP by flow cytometry in 169 ecpA� clinical andnatural E. coli isolates grown statically overnight at 26°C inDMEM. This collection included 43 EHEC strains belonging todifferent serotypes (O157:H7 and non-O157:H7), among which38 were LEE� and 5 were LEE�. ECP production was demon-strated in 35 of the 38 LEE� and 2 of the 5 LEE� EHEC (Table1). The level of production of ECP varied among the strainstested, as determined by IB and IF (SI Fig. 7). Furthermore, 84(66%) of the 126 non-EHEC E. coli strains tested produced ECP,albeit to different levels (Table 1). Representative strains of thiscollection showed ECP when adhering to HEp-2 cells (SI Fig.7B). We speculate that the remaining ECP� E. coli mightproduce ECP in vivo or under other in vitro growth conditions.These data suggest that most pathogenic and nonpathogenic E.coli strains are able to produce ECP.

Antibody Reactivity of Human and Bovine Sera to ECP. The presenceof antibodies against a particular antigen is a biological marker

A B

C

D

Fig. 3. Kinetics and environmental regulation of ecpA expression. (A) Time-dependent production of ECP by adhering bacteria demonstrated by IF. (B)Expression of ecpA mRNA analyzed during the course of cell infection (1.5–6 h). Influence of temperature and oxygen tension in environmental regulation ofECP demonstrated by IB (C) and RT-PCR (D). DnaK and amplification of 16S RNA (rrsB) were used as loading controls; RNA from uninfected cells was also usedas a negative control (�). M, mass standards.

Rendon et al. PNAS � June 19, 2007 � vol. 104 � no. 25 � 10639

MIC

ROBI

OLO

GY

Dow

nloa

ded

by g

uest

on

Mar

ch 1

6, 2

020

of the production of that antigen in a host. The presence ofanti-ECP antibodies in three pools of sera (five normal humansera, five sera from HUS patients, and five sera from bovines)was investigated. Regardless of the origin of the sera, IgGreactivity against EcpA was seen by IB (data not shown). Theseresults suggest that circulating anti-ECP antibodies are alreadypresent in healthy humans and bovines, which correlates with ourobservations that NFEC are also able to produce ECP.

ecpA Mutants of EHEC and NFEC Are Deficient in Adherence toEpithelial Cells. To provide genetic evidence of the involvement ofECP in bacterial adherence, the ecpA gene of EHEC EDL933was targeted for mutagenesis. The resulting EDL933�ecpA

strain lacked ECP production (Fig. 4 A, B, and D) and wassignificantly impaired in adherence compared with the wild-typestrain (P � 0.0023) or the ecpA mutant transcomplemented withplasmid pMR13 containing ecpA (Fig. 4 C and E).EDL933�ecpA(pMR13) produced abundant ECP, and this ob-servation correlated well with the level of adherence seen for thisstrain (Fig. 4). To our knowledge, this is the first report of amutation in an EHEC O157:H7 pilus gene that significantlyaffects adherence to human epithelial cells.

Next, we investigated the role of ECP for NFEC adherence bymutating ecpA in a HEp-2 cell-adherent E. coli isolate (Leo21)obtained from the stool of a healthy child. The resulting NFECLeo21�ecpA lacked ECP production (data not shown) and,consequently, became significantly deficient in adherence(�94% reduction, P � 0.0008) in comparison to the parentalstrain (SI Fig. 8). In support of this observation, similar resultswere obtained with a second NFEC (Leo6) ecpA mutant (datanot shown). In all, our data point to a significant role for ECPin cell adherence and perhaps in colonization of the host gutmucosa by EHEC and nonpathogenic E. coli.

DiscussionIt is well established that EHEC adheres to and colonizes theintestinal tracts of humans and farm animals and that, in vitro, itattaches to a variety of epithelial cell lines (19, 20). There are 16putative piliation operons present in the genome of EHECO157:H7 (6, 8), but what role pili play in the colonization of thehuman or bovine gastrointestinal tracts remains to be elucidated.Here, we found that EHEC expresses ecpA (yagZ), a highlyconserved gene present in the genomes of E. coli K-12 (5) andstrains with pathogenic attributes (6–8), and assembles theencoded protein EcpA into pili structures. In addition, we foundthat 96% of a collection of intestinal (NFEC, enteropathogenicE. coli, enterotoxigenic E. coli, enteroaggregative E. coli, EHEC,enteroinvasive, and rabbit pathogenic E. coli) and extraintestinal

Table 1. Production of ECP by ecpA � E. coli strains

N ECP� %

EHEC O157:H7 20 19 95EHEC non-O157:H7* 23† 18† 78EPEC 10 7 70ETEC 12 5 42EAEC 20 19 95EIEC 3 1 33UPEC 10 9 90APEC 10 7 70REPEC 1 0 0NFEC 60 36 60Total 169 121 71.6

ETEC, enterotoxigenic E. coli; EAEC, enteroaggregative E. coli; EIEC, en-teroinvasive E. coli; UPEC, uropathogenic E. coli; APEC, avian pathogenic;REPEC, rabbit pathogenic.*O113:H21, O55:H7, O26:H11, O15:H7, O15:NM, O111:H8, O111a:NM,O91:H21.

†Five strains are LEE� and two of those are EcpA�.

A B C

D

E

Fig. 4. EHEC EDL933 ecpA mutant is deficient in adherence to epithelial cells. Demonstration of production of ECP by EHEC strains by flow cytometry (A) andIB (B). (C) Quantification of adhering bacteria. The results shown represent the average of three separate experiments. (D) IF showing ECP (green) on bacteria(red). (E) Giemsa staining showing reduction of adherence in the ecpA mutant.

10640 � www.pnas.org�cgi�doi�10.1073�pnas.0704104104 Rendon et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

6, 2

020

(avian pathogenic and uropathogenic) E. coli strains containecpA. Further analysis by multiplex PCR of the ecpA� strainsrevealed that the remaining 4% of strains lacks the ecp operon.The importance of the biological role of ECP may expandbeyond E. coli because homologs of ecpA are found in thegenomes of Shigella boydii, Aeromonas hydrophila, and Yersiniamollaretii (data not shown).

To judge the biological significance of ECP in E. coli adher-ence, we chose EHEC EDL933 (O157:H7) as a model of studyover other pathogroups because of its clinical importance andbecause this organism does not routinely produce pili. We beganthis study by investigating the production of pili on EHECadhering to human cultured epithelial cells by high-resolutionSEM. Compelling data demonstrate that EHEC strains producea pilus structure composed of a 21-kDa pilin subunit whoseamino acid sequence corresponded to that of EcpA. Ultrastruc-tural EM and IF studies suggested that ECP contribute to EHECadherence by mediating direct binding of the bacteria to the cellmembrane through recognition of specific host cell receptors orby forming physical bridges between adhering bacteria. Ourresults provokingly suggest that the production of ECP may benecessary for stabilizing the adhering bacteria to the host cellmembrane favoring tissue colonization.

Expression of bacterial virulence factors is generally regulatedby complex molecular mechanisms that respond to host andenvironmental signals (35). The presence of host cells was notrequired for ECP production as initially thought, eliminating thehypothesis that eukaryotic cells or a soluble product weretriggering signals. In contrast to meningitis-associated E. coli(33), EHEC produced ECP after growth in DMEM at 26°C or37°C, but not in LB; suggesting that subtle differences in themechanisms regulating ECP production may exist betweendifferent strains. Production of ECP at temperatures �37°Cmight have important implications during the life of the organ-ism outside their bovine or human hosts, allowing for theirpersistence in the environment and contamination of produce.It is also possible that ECP mediate biofilm formation in someE. coli categories. The presence of 5% CO2, but not anaerobiosis,was favorable for ECP production. That ECP are produced at37°C under low oxygen tension is an indication that they mightbe produced by EHEC in the intestine.

The expression of the 16 putative pili operons present inEHEC O157:H7 was previously investigated through transcrip-tional analysis (32). Except for genes encoding curli (csgA), afimA-like gene, and ybgD and yehD (putative pilin genes), theexpression of the remaining 12 pilin subunit genes, includingecpA, was not demonstrated under the conditions examined (32).Likewise, genome-wide-scale efforts using transposon orsignature-tagged mutagenesis have not evidenced a role for ECPin adherence to Caco-2 cells or in colonization of the bovinegastrointestinal tract (36, 37). These apparently contradictoryresults may be attributed to differences in the experimentalassays and conditions used.

To provide genetic evidence of the role for ECP in EHEC celladherence, an isogenic ecpA mutant of EDL933 was constructed.This mutant was substantially reduced in adherence to HEp-2cells in comparison with the parental strain. The residualadherence observed in the ECP mutant was likely due to intimin.It is still an open question as to how commensal E. coli strainsadhere to the gut mucosa. The finding that two NFEC strainsthat were mutated in ecpA were also significantly reduced in theiradherence to HEp-2 cells further supports a role for ECP inEHEC and NFEC adherence. We attempted to block EHECadherence by using anti-ECP antibodies, but the results wereunsuccessful. Although the reason for this result is unknown, itis possible that the anti-ECP serum lacks antibodies directed toepitopes involved in cell adherence.

To determine whether production of ECP was a generalizedphenomenon among the E. coli, we surveyed a collection of 169ecpA� intestinal and extraintestinal E. coli strains. Production ofECP was demonstrated in 71.6% of the E. coli tested. AmongSTEC strains (including LEE� and LEE�), 86% produced ECP.This finding suggests that production of ECP is a commonproperty of STEC strains and perhaps a biological marker oftheir ability to colonize the intestinal epithelium. Notably, withinthe enteroaggregative E. coli group, 95% of the strains studiedproduced ECP. This observation is particularly significant giventhat only a minority of enteroaggregative E. coli strains produceany of the three AAF/I-III fimbriae reported in this diarrhea-genic E. coli group (11). Also of note is our finding that 9 of 10(90%) uropathogenic E. coli strains produced ECP, whichsuggested that, in addition to type I and Pap pili, ECP maycontribute to the adherence properties of this pathogroup. It ispossible that the ECP� E. coli strains found may have geneticalterations or, simply, may produce ECP under other experi-mental conditions. This idea is supported by the results showingthat not all ecpA� strains produced ECP under the conditionstested in this study. The remarkable high percentage of E. colistrains producing ECP is an indication that the pili must play asignificant biological role in the host–bacteria interplay. Thepresence of anti-ECP IgG in sera from healthy individuals andHUS patients may reflect the ability of NFEC and EHEC toproduce ECP in the intestine. It is unlikely that the bacteriawould expend a significant amount of energy in producing pilithat play no biological function. These observations have exten-sive implications regarding pathogenesis of disease caused by themajor diarrheagenic E. coli pathogroups and their evolution.

This article represents a reproducible demonstration of pilusproduction by EHEC O157:H7 and shows that a mutation in apilus gene of EHEC O157:H7 results in a substantial decrease inepithelial cell adherence. If ECP-mediated events are critical forEHEC to establish a successful intestinal infection, then it istempting to speculate that pathogenic E. coli strains use ECP tomimic commensal E. coli and provide themselves with anecological advantage for host colonization and evasion of theimmune system. This study supports our standing hypothesis thatECP is a common E. coli attribute that was inherited andconserved during the evolution of intestinal and extraintestinalE. coli, providing a widespread mechanism of host colonizationof different hosts and host tissues.

Materials and MethodsBacterial Strains and Plasmids. E. coli strains and plasmids aredescribed in SI Table 5. E. coli reference and diarrheagenic E.coli collections were kindly donated by Howard Ochman (Uni-versity of Arizona). Bacterial strains were propagated overnightin LB broth or DMEM (Invitrogen, Carlsbad, CA) at 26°C or37°C, statically, with aeration, under 5% CO2 atmosphere oranaerobiosis. Anaerobiosis was achieved by using the GasPakEZ anaerobe gas-generating pouch system (Beckton Dickinson,Franklin Lakes, NJ). For testing ECP production, bacterialcultures were normalized by spectrometry. Antibiotics wereadded, when necessary, at concentrations of 100 �g/ml (ampi-cillin) or 50 �g/ml (kanamycin). Arabinose (Sigma, St. Louis,MO) was used at a 100 mM concentration.

Interaction with Eukaryotic Cells. HEp-2 and HeLa epithelial cells(ATCC CCL-23 and CCL-2, respectively) were used in adher-ence assays carried out from 0 to 6 h of infection as described(38). The results obtained on ECP production by EHEC strainswere identical for both cell lines. Adherent bacteria were quan-tified by plating out 10-fold serial dilutions on LB agar. Replicasamples were used for IF and Giemsa staining. Inhibition ofadherence was performed by incubation of EDL933 and cultured

Rendon et al. PNAS � June 19, 2007 � vol. 104 � no. 25 � 10641

MIC

ROBI

OLO

GY

Dow

nloa

ded

by g

uest

on

Mar

ch 1

6, 2

020

cells with 1:10 and 1:100 dilutions of anti-ECP. Statisticalanalysis was performed by using Student’s t test.

Ultrastructural Studies. The presence of pili on bacterial cells wasvisualized by EM in a Phillips CM12 electron microscope at 80kV as described (38, 39). Immuno-EM was performed withanti-EHEC ECP antibodies and anti-rabbit IgG conjugated to10- or 30-nm gold particles as described (38). Initial studies usedrabbit anti-MatB antibodies (kindly provided by Timo K. Kor-honen, University of Helsinki, Helsinki, Finland) (33). Glasscoverslips containing fixed mammalian cells with adhering bac-teria were prepared for SEM as described (38) and visualized byusing a Hitachi S-4500 scanning electron microscope (Hitachi,Tokyo, Japan).

Pili Purification. Tissue culture bottles containing monolayers ofHEp-2 cells were infected with EHEC EDL933 for 6 h at 37°Cunder 5% CO2 atmosphere. The bacteria were collected and thepili were purified as described (38). The pili were denatured withHCl (40) and resolved by SDS/PAGE (41). A 21-kDa protein wasexcised and subjected to Edman degradation (Stanford Univer-sity, Stanford, CA). For IB, bacteria were normalized to equalamounts and HCl-treated before SDS/PAGE and then reactedwith primary anti-ECP-antibodies, followed by the secondaryperoxidase conjugate (Sigma, St. Louis, MO). The substrate wasa chemoluminescent reagent (GE Healthcare, Chalfont St. Giles,U.K.). Anti-DnaK (Stressgen Bioreagents, Victoria, BC, Can-ada) was used as control for the amount of protein loaded intothe gel.

Human and Animal Sera. Sera from HUS patients were a kind giftof Phillip Tarr (Washington University School of Medicine, St.Louis, MO) to J.B.K. Normal human and bovine sera were

obtained from the collection of J.B.K. These sera were tested forreactivity against ECP by IB.

Flow Cytometry. Flow cytometry was used to detect the produc-tion of ECP by all E. coli strains studied, as described (42).Briefly, bacteria grown overnight in DMEM were incubated withanti-ECP antibodies (1:1,000) followed by goat anti-rabbit IgGAlexa fluor conjugate (Invitrogen). The Alexa Fluor fluores-cence emission was collected through a 30-nm band pass filtercentered at 530 nm in which 50,000 events were measured.Bacteria were labeled with propidium iodide (Sigma) and de-tected through a 42-nm band pass centered at 585 nm. Theseexperiments were repeated three times in triplicate. The sampleswere analyzed in a Becton Dickinson FACScan.

Mutagenesis, Multiplex PCR, and RT-PCR. Detailed methods andprotocols for the generation of nonpolar mutants, plasmidconstruction, multiplex PCR and RT-PCR can be found in theSI Text.

We thank Diana R. Hernandez and Fabiola Avelino for technicalassistance; Dr. Timo K. Korhonen for the kind gift of anti-MatBantibodies; Dr. Iruka Okeke for critical discussions; Dr. Harry L. Mobley(University of Michigan, Ann Arbor, MI) for uropathogenic E. coliCTF073; Drs. Melha Mellata (Arizona State University, Phoenix, AZ)and Roy Curtiss for avian pathogenic E. coli strains; Dr. ElizabethHartland (Monash University, Melbourne, Australia) for the rabbitpathogenic E. coli strain; and Dr. Howard Ochman for the ECOR andDEC strains. This study was supported by National Institutes of HealthGrants AI66012 (to J.A.G.), DK58957 (to J.B.K.), and AI21657 (toJ.B.K.); Direccion General de Asuntos del Personal Academico GrantIN201703-3; Consejo Nacional de Ciencia y Technologıa 42918Q; andHoward Hughes Medical Institute Grant 75301-565101 (to J.L.P.).J.A.G. thanks the Arizona Hispanic Center of Excellence.

1. Ofek I, Hasty DL, Doyle RJ (2002) in Bacterial Adhesion (Am Soc Microbiol,Washington, DC), pp 63–96.

2. Jonson A-B, Normak S, Rhen M (2005) in Concepts in Bacterial Virulence, edsRussell W, Herwarld H (Karger, Basel), Vol 12, pp 67–89.

3. Schilling JD, Mulvey MA, Hultgren SJ (2001) J Infect Dis 183(Suppl 1):S36–S40.4. Olsen A, Jonsson A, Normark S (1989) Nature 338:652–655.5. Blattner FR, Plunkett G, III, Bloch CA, Perna NT, Burland V, Riley M,

Collado-Vides J, Glasner JD, Rode CK, Mayhew GF, et al. (1997) Science277:1453–1474.

6. Perna NT, Plunkett G, III, Burland V, Mau B, Glasner JD, Rose DJ, MayhewGF, Evans PS, Gregor J, Kirkpatrick HA, et al. (2001) Nature 409:529–533.

7. Welch RA, Burland V, Plunkett G, III, Redford P, Roesch P, Rasko D, BucklesEL, Liou SR, Boutin A, Hackett J, et al. (2002) Proc Natl Acad Sci USA99:17020–17024.

8. Hayashi T, Makino K, Ohnishi M, Kurokawa K, Ishii K, Yokoyama K, Han CG,Ohtsubo E, Nakayama K, Murata T, et al. (2001) DNA Res 8:11–22.

9. Riley LW, Remis RS, Helgerson SD, McGee HB, Wells JG, Davis BR, HebertRJ, Olcott ES, Johnson LM, Hargrett NT, et al. (1983) N Engl J Med308:681–685.

10. Karmali MA, Petric M, Lim C, Fleming PC, Steele BT (1983) Lancet2:1299–1300.

11. Kaper JB, Nataro JP, Mobley HL (2004) Nat Rev Microbiol 2:123–140.12. Elliott SJ, Yu J, Kaper JB (1999) Infect Immun 67:4260–4263.13. McDaniel TK, Kaper JB (1997) Mol Microbiol 23:399–407.14. Campellone KG, Robbins D, Leong JM (2004) Dev Cell 7:217–228.15. Garmendia J, Phillips AD, Carlier MF, Chong Y, Schuller S, Marches O,

Dahan S, Oswald E, Shaw RK, Knutton S, et al. (2004) Cell Microbiol6:1167–1183.

16. Paton JC, Paton AW (1998) Clin Microbiol Rev 11:450–479.17. Caprioli A, Morabito S, Brugere H, Oswald E (2005) Vet Res 36:289–311.18. Dean-Nystrom EA, Bosworth BT, Moon HW (1997) Adv Exp Med Biol

412:47–51.19. Tarr PI, Bilge SS (1998) in Escherichia coli O157:H7 and Other Shiga Toxin-

Producing E. coli Strains, eds Kaper JB, O’Brien AD (Am Soc Microbiol,Washington, DC), pp 157–162.

20. Moxley RA, Francis DH (1998) in Escherichia coli O157:H7 and Other ShigaToxin-Producing E. coli Strains, eds Kaper JB, O’Brien AD (Am Soc Microbiol,Washington, DC), pp 249–260.

21. McKee ML, Melton-Celsa AR, Moxley RA, Francis DH, O’Brien AD (1995)Infect Immun 63:3739–3744.

22. Donnenberg MS, Tzipori S, McKee ML, O’Brien AD, Alroy J, Kaper JB (1993)J Clin Invest 92:1418–1424.

23. Jerse AE, Yu J, Tall BD, Kaper JB (1990) Proc Natl Acad Sci USA 87:7839–7843.

24. Kenny B, DeVinney R, Stein M, Reinscheid DJ, Frey EA, Finlay BB (1997) Cell91:511–520.

25. Frankel G, Lider O, Hershkoviz R, Mould AP, Kachalsky SG, Candy DC,Cahalon L, Humphries MJ, Dougan G (1996) J Biol Chem 271:20359–20364.

26. Sinclair JF, O’Brien AD (2002) J Biol Chem 277:2876–2885.27. Low AS, Dziva F, Torres AG, Martinez JL, Rosser T, Naylor S, Spears K,

Holden N, Mahajan A, Findlay J, et al. (2006) Infect Immun 74:2233–2244.28. Nicholls L, Grant TH, Robins-Browne RM (2000) Mol Microbiol 35:275–288.29. Paton AW, Srimanote P, Woodrow MC, Paton JC (2001) Infect Immun

69:6999–7009.30. Tarr PI, Bilge SS, Vary JC, Jr, Jelacic S, Habeeb RL, Ward TR, Baylor MR,

Besser TE (2000) Infect Immun 68:1400–1407.31. Torres AG, Giron JA, Perna NT, Burland V, Blattner FR, Avelino-Flores F,

Kaper JB (2002) Infect Immun 70:5416–5427.32. Low AS, Holden N, Rosser T, Roe AJ, Constantinidou C, Hobman JL, Smith

DG, Low JC, Gally DL (2006) Environ Microbiol 8:1033–1047.33. Pouttu R, Westerlund-Wikstrom B, Lang H, Alsti K, Virkola R, Saarela U,

Siitonen A, Kalkkinen N, Korhonen TK (2001) J Bacteriol 183:4727–4736.34. Simons BL, Rathman P, Malij CR, Oudega B, de Graaf FK (1990) FEMS

Microbiol Lett 55:107–112.35. Edwards RA, Puente JL (1998) Trends Microbiol 6:282–287.36. Dziva F, van Diemen PM, Stevens MP, Smith AJ, Wallis TS (2004) Microbiology

150:3631–3645.37. Tatsuno I, Kimura H, Okutani A, Kanamaru K, Abe H, Nagai S, Makino K,

Shinagawa H, Yoshida M, Sato K, et al. (2000) Infect Immun 68:5943–5952.38. Giron JA, Torres AG, Freer E, Kaper JB (2002) Mol Microbiol 44:361–379.39. Giron JA, Ho AS, Schoolnik GK (1991) Science 254:710–713.40. McMichael JC, Ou JT (1979) J Bacteriol 138:969–975.41. Laemmli UK (1970) Nature 227:680–685.42. Humphries AD, Raffatellu M, Winter S, Weening EH, Kingsley RA, Droleskey

R, Zhang S, Figueiredo J, Khare S, Nunes J, et al. (2003) Mol Microbiol48:1357–1376.

10642 � www.pnas.org�cgi�doi�10.1073�pnas.0704104104 Rendon et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

6, 2

020