INFECTION AND IMMUNITY, Aug. 1994, p. 3282-32880019-9567/94/$04.00+0Copyright 0 1994, American Society for Microbiology

Escherichia coli Serogroup 0111 Includes Several Clones ofDiarrheagenic Strains with Different Virulence Properties

LEILA C. CAMPOS,lt THOMAS S. WHITTAM,2* TANIA A. T. GOMES,3JOAO R. C. ANDRADE,4 AND LUIZ R. TRABULSI'

Departamento de Microbiologia, Universidade de Sdo Paulo,' and Departamento de Microbiologia, Immunologia e

Parasitologia, Escola Paulista de Medicina,3 Saio Paulo, SP, and Servico de Microbiologia e Imunologia,Universidade do Estado do Rio de Janeiro, RJ,4 Brazil, and Institute of Molecular Evolutionary Genetics,

Department of Biology, The Pennsylvania State University, University Park, Pennsylvania 168022

Received 4 March 1994/Returned for modification 17 April 1994/Accepted 9 May 1994

Genetic variation among isolates of Escherichia coli 0111 obtained mostly from patients with diarrhea inBrazil was assessed by multilocus enzyme electrophoresis to characterize chromosomal genotypes and by geneprobes and adherence assays to characterize virulence properties. Among the 152 isolates, we resolved 16distinct electrophoretic types (ETs), which differed on average at 40% of the enzyme loci. We identified fourmajor bacterial 0111 clones of different disease classes: ET 12, which includes the bulk of the enteropathogenicE. coli strains, typically showing localized adherence and intimate attachment in tissue culture assays; ET 1,which includes strains with a different set ofvirulence markers; ET 9, which includes strains that show intimateattachment but lack localized adherence and Shiga-like toxin genes; and ET 8, which includes strains that areShiga-like toxin producers and have the corresponding traits of enterohemorrhagic E. coli. Enteroaggregativestrains constituted ET 10 and also occurred in ET 1. Isolates of the major clones were found in South andNorth America and matched in ET and virulence factors to previously described diarrheagenic clones that arewidely disseminated in the human population. Because the major clones are genetically distantly related andexhibit different combinations of virulence factors, we hypothesize that they have distinct mechanisms ofpathogenesis. The results indicate that genetic divergence of bacteria with the 0111 antigen, as measured byallelic variation in enzyme loci, is accompanied by divergence in virulence properties of clones so thatidentification and classification of pathogenic E. coli strains cannot be based solely on serotyping or a singlevirulence factor.

Escherichia coli strains of serogroup 0111 were the firststrains implicated as the main cause of outbreaks of severegastroenteritis in infant nurseries in the United Kingdom in the1940s (4, 10); the serogroup is now recognized as one of theclassic serogroups of enteropathogenic E. coli (EPEC) (15).Since this early work, E. coli 0111 strains have been implicatedin numerous epidemics of serious enteric disease, including28% of 50 outbreaks of infantile diarrhea in the United Statesfrom 1934 to 1987 (18). An 0111 strain was incriminated in anoutbreak of prolonged, relapsing diarrheal illness in a day carecenter in Seattle, Wash. (25), and more recently, an 0111strain caused an extensive community outbreak of diarrheainvolving more than 700 people in Finland (34). In addition toclear-cut epidemics, E. coli 0111 strains are prevalent inendemic (or sporadic) cases of diarrhea. In Canada, Gurwith etal. (8) recovered 0111 strains in 3.5% of 418 children withacute diarrhea, significantly more often than in matchingcontrols, and in Brazil, Gomes et al. (7) found 0111 strains in19% of 500 cases of diarrhea in infants, with only 0.4% inmatching controls.

Surveys of 0111 strains of diverse origin have shown thatthere is a substantial amount of genetic and phenotypicvariation among bacteria of this 0 serogroup. Although motile

* Corresponding author. Mailing address: IMEG, Department ofBiology, 208 Mueller Laboratory, The Pennsylvania State University,University Park, PA 16802. Phone: (814) 863-1970. Fax: (814) 865-9131. Electronic mail address: tswl@psuvm.psu.edu.

t Permanent address: Departamento de Bacteriologia, InstitutoOswaldo Cruz, FIOCRUZ, 21045-900 Rio de Janeiro, RJ, Brazil.

0111 strains from outbreaks of infantile diarrhea usually haveeither H2, H12, or H21 flagellar antigens, in many cases,nonmotile (H-) isolates are common and isolates of otherspecific H types are often cultured (29, 33). Strains withserotypes 0111:H2 and 0111:H- typically carry genes speci-fying the EPEC adherence factor (EAF) that mediates local-ized adherence (LA) of bacteria to cultured epithelial cells andis a characteristic of strains of the classic EPEC serotypes (7,21). In contrast, 0111:H8 isolates have been classified asenterohemorrhagic E. coli (EHEC) because they lack EAF andpossess a plasmid related to one found in 0157:H7 strains thatcause hemorrhagic colitis (16).To test the hypothesis that separate outbreaks of infantile

diarrhea were caused by the same bacterial strains, Stenderupand 0rskov (33) examined 0111 isolates of diverse origin andfound that many cases of infantile diarrhea were caused byorganisms with identical serotypes, biotypes, and outer mem-brane protein patterns, suggesting that together these charac-ters marked bacterial clones with widespread distributions.This finding gained support from population genetic analysisby multilocus enzyme electrophoresis (32), which showed that0111:H2 strains of independent origin were clonally related,with identical alleles at many enzyme-encoding loci (24). Morerecently, Whittam et al. (35) found that 0111 isolates are

genetically diverse but that most 0111 isolates belong to alimited number of bacterial clones, as marked by distinctelectrophoretic types.The objective of the study reported here was to identify the

major clonal types of the 0111 serogroup associated withdiarrheal disease in Brazil and assess the variation among

3282

Vol. 62, No. 8

on June 19, 2018 by guesthttp://iai.asm

.org/D

ownloaded from

VIRULENCE PROPERTIES OF E. COLI SEROGROUP 0111 CLONES 3283

clones with regard to specific factors incriminated in virulence.We characterized the multilocus genotypes of a collection of0111 strains by using enzyme electrophoresis to detect allelicvariation in chromosomal structural genes and determined theclonal relationships among isolates. With the clonal frame-work, we examined the distribution of adherence and toxinfactors that contribute to the virulence of diarrheagenic E. coli.

MATERIALS AND METHODS

Bacterial strains. The study is based on the analysis of 152E. coli strains of the 0111 serogroup, most of which were

originally isolated from children with diarrhea in Brazil (Table1). For comparative purposes, we included 14 strains isolatedin Chile, 14 strains originally isolated in the United States, and1 strain originally recovered in Guatemala. All strains were

tested against standard flagellar (H) antisera by conventionalserology (6). Nonmotile strains were designated H-.

Multilocus enzyme electrophoresis. To analyze enzyme elec-trophoretic variation, bacterial cell lysates were prepared andsubjected to multilocus enzyme electrophoresis (32). Twentyenzymes, listed in Table 2, were examined for allelic variationas described previously (35). For each enzyme, electromorphswere determined through comparisons to standard mobilityvariants and assigned numbers by their rate of anodal migra-tion. Isolates that lacked detectable enzyme activity were

designated as having a null allele at the locus in question.To estimate the genetic relationships among isolates, elec-

tromorphs were equated with alleles at the correspondingenzyme locus, so that each bacterial strain was characterized byits multilocus genotype (allele combination) for the enzyme-encoding loci assayed. Distinctive multilocus genotypes were

designated electrophoretic types (ETs) and were numbered bytheir inferred relationships from phylogenetic analysis. Ge-netic distance between ETs was calculated from the proportionof mismatched alleles under the assumption that a mobilitydifference results from at least one codon change at thenucleotide level. The genetic distances were used to generate a

dendrogram by the average linkage algorithm and tested bybootstrapping as described previously (35).Adherence properties. Each strain was characterized by the

pattern of attachment of bacteria to HEp-2 cells in thepresence of D-mannose as assayed by previously describedmethods (30). Monolayers were examined after 3 and 6 h ofincubation. Three distinct patterns of adherence were distin-guished: LA occurred when bacteria bound to localized areas

of the HEp-2 cells in culture, forming distinct microcolonies;diffuse adherence (DA) occurred when bacteria adhered on

the entire surface of the HEp-2 cells without formation ofdiscrete microcolonies; and aggregative adherence (AA) oc-

curred when clumps of bacteria adhered to the HEp-2 cells andon the glass slide in the absence of HEp-2 cells in a charac-teristic stacked-brick pattern (20). In some cases, the adher-ence pattern could not be resolved, so the adherence pattern ofthese strains was considered indefinite.FAS test. Each strain was examined by the fluorescent-actin

staining (FAS) test, an in vitro tissue culture assay that isdiagnostic for the attaching and effacing lesions characteristicof disease caused by EPEC infections (12). The FAS testdetects the high concentration of filamentous actin that accom-panies intimate attachment of bacteria to tissue culture cells(11). Strains were examined by the FAS test and classified as

either FAS+ (dense concentrations of actin present) or FAS-(no dense actin concentrations). Readings of the slides were

made after 3 and 6 h of incubation.E-Hly. Enterohemolysin (E-Hly) production was detected by

TABLE 1. Source and flagellar antigens (H types) of152 E. coli 0111 strains

Origin Period No. of Common H typesisolatesa H2 H12 H9 H25 H-

BrazilSao Paulo (SP) 1958-1965 27 17 6 4Sao Paulo (SP) 1980-1993 63 13 5 3 1 40Botucatu (SP) 1993 1 1Ribeirao Preto 1993 14 8 1 2

(SP)Rio de Janeiro 1986-1990 8 5 1 2Recife 1993 10 8 1 1

Chile 1983 14 10 2United States 1954-1980 14b 1 1 2 1 6

Total 152C 63 14 6 2 57

a Includes six strains for which the H type was not determined.h Includes two H21 strains and one H28 strain.c Total includes an 0111:H4 strain from Guatemala not shown in the table.

the method described by Beutin et al. (2) as follows. Afterovernight growth in tryptic soy broth (Difco Laboratories,Detroit, Mich.) at 37°C, bacterial cultures were plated ontryptose blood-agar base (Difco) supplemented with 10 mMCaCl2 and 5% defibrinated sheep blood washed three times inphosphate-buffered saline, pH 7.2. Cultures were incubated at37°C and examined for hemolysis after 3 and 24 h of incuba-tion. Typically alpha-hemolysin causes lysis of both washed andunwashed sheep blood cells that is detectable after 3 h ofincubation at 37°C, whereas E-Hly lyses only washed sheepblood cells after 24 h of incubation (2). The test was consideredpositive for E-Hly when lysis zones around colonies appearedonly after 24 h of incubation.Gene probes. Strains were tested by colony hybridization as

described by Maas (17). The following specific gene probeswere used: EAF, a 1.0-kb BamHI-SalI fragment from plasmidpMAR2 (19); EAE (E. coli attaching and effacing gene, eaeA),a 1-kb SalI-KpnI fragment from plasmid pCVD434 (9); DAfactor, a 350-bp PstI fragment of pSLM852 (3); AA factor, a1-kb EcoRI-PstI fragment of pCVD432 (1); SLT-I (Shiga-liketoxin gene I), a 1,142-bp BamHI fragment of pJN37-19 (22);SLT-II (Shiga-like toxin gene II), an 842-bp SmaI-PstI frag-ment from plasmid pNN111-19 (22); and EHEC factor, a3.4-kb Hindlll fragment of pCVD419 (16). The DNA probefragments were labeled with [ca-32P]dATP by nick translation(26).

RESULTS

Of the 152 0111 strains tested for the presence of flagellarantigens, 58.6% were typeable with standard H antisera, sixcould not be typed, and the remaining 37.5% of the strainswere nonmotile (H-). Seven different H types were distin-guished among the 0111 strains. Strains of the most commonserotype, O111:H2, were recovered from all localities andaccounted for more than two-thirds (63 of 89 strains) of theO:H typeable strains (Table 1). Bacteria of three other Htypes, H12, H9, and H25, were found in both the samples fromBrazil and those from the United States. In contrast, strainswith serotype H4, H21, or H28 were uncommon and wererecovered in only single localities (Table 1).Enzyme polymorphism and ETs. All 152 isolates were

examined for allelic variation at 20 enzyme loci by multilocusenzyme electrophoresis (32). Seventeen of the 20 enzyme loci

VOL. 62, 1994

on June 19, 2018 by guesthttp://iai.asm

.org/D

ownloaded from

3284 CAMPOS ET AL.

TABLE 2. Sixteen 0111 ETs defined by distinct allele combinations for 20 enzyme-encoding loci

T No. of Allele at enzyme locuseisolates PGI IDH ACO G3P PE2 AK MDH PGD MlP GOT BGA ADH MPI G6P IPO CAK NSP TDH SKD GLU

1 18 5 2 7 4 5 2 4 8 6 6 82 1 5 2 7 4 5 2 4 8 6 6 103 1 5 5 7 4 7 2 4 8 6 6 84 1 5 5 7 4 7 2 4 8 6 6 85 1 5 2 7 4 5 2 4 8 6 6 46 1 5 5 6 4 4 2 4 8 6 6 07 1 5 2 6 4 5 2 4 12 6 6 88 7 8 5 7 4 5 2 4 8 8 6 59 8 6 5 6 4 5 2 4 8 8 6 010 2 6 5 6 4 5 2 4 8 8 6 811 2 6 5 7 4 5 2 4 8 7 6 2.512 105 6 5 7 4 5 2 4 8 6 6 2.513 1 6 5 7 4 5 2 4 8 6 2 2.514 1 6 3 0 0 4 4 4 10 4 6 415 1 6 0 0 4 4 4 4 10 4 6 416 1 6 5 6 3 5 2 4 8 8 6 0

2 4 4 4 4 2 2 3 22 8 4 4 4 2 2 3 40 4 2 4 4 2 2 4 22 4 2 4 4 2 2 3 20 4 4 4 4 2 0 2.4 20 8 4 4 4 2 2 4 20 4 2 4 4 2 4 3 26 8 4 4 4 2 4 8 44 8 4 4 4 2 4 4 20 8 4 4 6 2 4 8 26 10 4 4 4 2 4 4 26 10 4 4 4 2 4 4 26 10 4 4 4 2 4 4 22 10 4 4 4 2 4 1.4 22 10 4 4 4 2 4 4 24 8 4 4 4 2 4 4 2

a Abbreviations: PGI, glucosephosphate isomerase; IDH, isocitrate dehydrogenase; ACO, aconitase; G3P, glyceraldehyde-phosphate dehydrogenase; PE2,phenylalanyl-leucine peptidase; AK, adenylate kinase; MDH, malate dehydrogenase; PGD, gluconate-6-phosphate dehydrogenase; M1P, mannitol-1-phosphatedehydrogenase; GOT, aspartate aminotransferase; BGA, P-galactosidase; ADH, alcohol dehydrogenase; MPI, mannosephosphate isomerase, G6P, glucose-6-phosphate dehydrogenase; IPO, indophenol oxidase; CAK, carbamate kinase; NSP, nucleoside phosphorylase; TDH, threonine dehydrogenase; SKD, shikimatedehydrogenase; GLU, glutamate dehydrogenase.

were polymorphic, with an average of three alleles resolved perlocus. Comparison of the allele profiles of strains revealed 16distinct ETs (Table 2). Ten ETs were represented by singleisolates, whereas six ETs were represented by multiple isolates.Nearly 70% of the 0111 strains were indistinguishable in alleleprofile and belonged to a single ET, ET 12 (Table 2). Isolatesof ET 12 were recovered about five times more often thanisolates of ET 1, the second most common allele combinationobserved in the sample.

Genetic relationships of clones. Genetic distances betweenthe 16 ETs listed in Table 2 were estimated from the numberof allele differences and analyzed with the average linkagealgorithm and bootstrap method as described previously (35).The dendrogram illustrates the genetic similarity and degree ofdivergence between the 0111 clonal lineages and reveals threemajor clusters of strains at a genetic distance of 0.50 (Fig. 1).The top cluster (A) was observed in 53% of the bootstrappedtrees and comprises ETs 1 to 7. Cluster A includes the secondmost common 0111 clone (ET 1) and overall accounts for 16%of the 0111 isolates. Most of the 0111 isolates (83%) fall intocluster B, which is composed of six ETs (8 to 13 and 16) andincludes the most common 0111 clone (ET 12). Only 1% ofthe 0111 strains belong to the highly divergent cluster C,represented by ETs 14 and 15, which groups at a geneticdistance of more than 0.75 from the other clusters.

In four cases, more than 75% of the bootstrap trees sup-ported specific groupings of ETs. These significant groupsinclude ETs 3 and 4, ETs 9 and 16, ETs 11 to 13, and ETs 14and 15 (Fig. 1). For each of these four groups of strains, it canbe inferred that the ETs are most closely related genetically toother ETs within the cluster. In contrast, the branching orderfor the remaining ETs within a cluster cannot be determinedwith statistical confidence based on the bootstrap analysis. Forexample, we can infer that ETs 3 and 4 are the closest relativesto one another in the A cluster but we cannot conclude that ET1 is more closely allied to ET 5 than it is to ET 2.

Virulence properties of 0111 clones. The clones of cluster Amarked by ETs 1 to 7 exhibited a variety of virulence proper-ties (Table 3). In ET 1 none of the strains with the H12 antigenhybridized with the gene probes. The 14 strains adhered to

HEp-2 cells, 6 showing DA and the other showing an indefinitepattern of adherence. The H4 strain showed AA but wasnegative with the AA probe, and the three H nontypeablestrains reacted with the AA probe. Two of these strains showedthe typical AA pattern.The remaining six ETs of cluster A showed distinct combi-

nations of virulence properties. Strains of ETs 3 and 4 werenearly identical in enzyme allele profile (Table 2), and thisgenetic similarity was reflected in other traits of these strains:both expressed H25 antigens (Table 3), exhibited a positivereaction in the FAS assay, and showed strong hybridizationwith the EAE gene probe. None of the ET 5, 6, or 7 strainsshowed any of the adherence patterns characteristic of diar-rheagenic E. coli. It is noteworthy, however, that the ET 7isolate was unusual because it was positive with both EAF andEAE gene probes yet failed to exhibit the appropriate adher-ence phenotypes.The 0111 strains of cluster B typically had the eaeA gene

and showed intimate adherence as assayed by the FAS test(Table 3). The exception was the 0111:H21 strains of ET 10which showed aggregative adherence to HEp-2 cells and werepositive with the AA gene probe (Table 3). In general, thedifferent clones of the B cluster were distinct in their virulencetraits. Strains of the clone marked by ET 8 were nonmotile,and some had the properties that define EHEC strains. (ET 8isolates were probe positive for both EHEC and SLT-I se-quences, and many expressed E-Hly.) In contrast, most ET 9isolates produced E-Hly but lacked SLT genes or EHECsequences.More than 70% of the 0111 strains belong to the clonal

lineage within the B cluster marked by ETs 11 to 13. Virtuallyall of these strains showed localized (LA') and intimate(FAS+) adherence and were positive for the correspondingEAF and EAE gene probes. The bulk of the ET 11 to 13isolates had the H2 flagellar antigen, although many isolates ofthe common clone marked by ET 12 were nonmotile.The occurrence of the various virulence factors in the 0111

clones is summarized in the right column of Fig. 1. Most of thevirulence factors occur in distantly related lineages of the 0111serogroup and are found in a variety of combinations. For

INFECT. IMMUN.

on June 19, 2018 by guesthttp://iai.asm

.org/D

ownloaded from

VIRULENCE PROPERTIES OF E. COLI SEROGROUP 0111 CLONES 3285

Virulence factors1 Olllab:H12 DA+ AA+

5 0111ab EHIy+

2 Olllac:H-

3 0111 ab:H25 -FAS+ eae+

4 0111ab:H25 U

6 Olllac:H9

7 Olllab:H- EAF÷ eae+

8 0111ac:H- FAS+ eae+ SLT-I÷ EHEC+ EHly+

9 Olllab:H9 iFAS+ eae+ EHly÷

16 0111:H9

10 Olllab:H21 AA+

11 Olllab:H2

12 Olllab:H2

13 Olllab:H2 LA+ EAF. FAS+ eae+

14 Olllab:H-

15 Olllab:H-

0.8 0.6 0.4 0.2 0

Genetic distanceFIG. 1. Dendrogram of genetic relationships of 16 ETs of E. coli 0111 strains. Genetic distance is measured in terms of the number of

detectable codon differences per enzyme locus. The numbers on the branches of the dendrogram are the percentages of bootstrapped treessupporting the branches. The presence of virulence factors in the branches of the tree is given in the right column. See the footnotes to Table3 for the abbreviations for virulence factors.

example, AA factor (AA') is found in strains of ET 1 ofcluster A and ET 10 of cluster B, two clusters whose clonalgenotypes differ on average at 60% of the enzyme loci. E-Hly(E-Hly+) is found in both cluster A (ET 5) and cluster B (ETs8, 9, and 16) in combination with different factors: in cluster A,it is the only factor found in a strain, whereas in cluster B, it isfound in combination with other adherence and toxin factors.The intimate adherence phenotype (FAS+) and the corre-sponding intimin gene (eae+) occur in 10 different ETs repre-senting diverse lineages of all three major clusters (Fig. 1). Incontrast, SLT-1 and EHEC genes are found only in the clonallineage marked by ET 8, and the DA phenotype is found onlyin ET 1 strains. The LA phenotype also occurs in severalchromosomal backgrounds, with all but two LA' strains (ETs14 and 15) belonging to the 0111:H2 clonal group composedof ETs 11 to 13.

DISCUSSION

Although E. coli 0111 has been recognized as a pathogenicserogroup for more than 40 years, there has been a growingbody of evidence indicating that strains of the 0 grouprepresent a diverse mixture of E. coli, with a variety ofvirulence traits and mechanisms of pathogenesis. Stenderupand 0rskov (33) clearly showed a striking correlation ofbiotype and outer membrane protein patterns with H serotypeamong 0111 strains of diverse origin, a result that gave strongsupport to the hypothesis that many cases of infantile diarrheaare caused by bacteria of special clones that have worldwidedistributions. Clonal analysis of 0111 isolates based on mul-tilocus enzyme electrophoresis disclosed extensive geneticvariation among 0111 strains and provided a genetic basis forclassifying isolates into homogeneous groups of closely relatedbacteria that mark individual lineages (35).

VOL. 62, 1994

on June 19, 2018 by guesthttp://iai.asm

.org/D

ownloaded from

3286 CAMPOS ET AL.

TABLE 3. Presence of virulence factors in E. coli 0111 strains of 16 ETs

EP No. of H type6 No. of isolates positive with gene probec: Adherenceisolates EAF EAE AA SLT-I EHEC FASe E-Hly pattern

1 18 H12 (14) DA, INH4 (1) AAH ND (3) 3 AA, IN

2 1 H- IN3 1 H25 1 if IN4 1 H25 1 if IN5 1 H ND 1 IN6 1 H9 IN7 1 H- 1 1 IN8 7 H- (6) 2 2 IN

1 if IN1 1 if IN2 2 2 2f 2 IN

H ND(1) 1 1 1 1f 1 IN9 8 H9 (4) 4 3/1f 4 IN

H- (3) 3 3f 3 INH ND 1 if IN

16 1 H9 1 1 IN10 2 H21(2) 2 1 AA11 2 H2 (2) 2 2 2 LA12 105 H2 (60) 58 59 58/2f LA

H- (44) 44 44 43 LAH28 1 1 1 LA

13 1 H2 1 1 1 LA14 1 H- 1 1 1 LA15 1 H- 1 1 1 LA

"See Table 2 for allele profiles defining ETs.b Number of isolates in parenthesis. ND, not determined.c All isolates tested were negative with probes for the DA factor and the SLT-II gene.d Pattern of adherence of bacteria to HEp-2 cells. IN, indefinite pattern.e Values were determined after 3 h of incubation unless otherwise indicated.f After 6 h of incubation.

The results presented here indicate that genetic divergenceof 0111 clones, as measured by allelic variation in enzyme loci,is also reflected in differences in virulence factors. We identi-fied four major bacterial clones that occur in South and NorthAmerica, and each one is distinct in the combination ofvirulence factors that characterize its members. Each one ofthe clones is discussed in detail below.ET 12 clone. The most frequently isolated 0111 clone in our

study was marked by ET 12 and accounted for nearly 70% ofall the Brazilian 0111 strains. Virtually all members of thisclone showed the characteristics of class I EPEC strains: theycarried sequences homologous to EAF and eaeA and showedboth localized (LA') and intimate (FAS+) adherence. Al-though most ET 12 isolates were motile, with the H2 flagellarantigen, 40% of the isolates of this clone were nonmotile andthus could not be identified by O:H serotyping alone. The factthat the LA phenotype was found nearly exclusively in the ET12 clonal group explains why many LA' 0111 strains areindistinguishable in biotypic characteristics from 0111:H2strains (29).

Direct comparison of the allelic variants of ET 12 withelectrophoretic profiles described in a previous study (35)indicates that ET 12 is a member of a common 0111:H2 clonerepresented by 43 isolates recovered from patients on fourcontinents and referred to as DEC 12, or diarrheagenic E. coliclone 12 (35). This close overall relatedness of the Brazilianisolates of ET 12 and the bacteria of DEC 12 is supported bythe similarity in the phenotypes of these strains: most DEC 12isolates exhibit LA and intimate attachment, and a highpercentage of strains were nonmotile. It is noteworthy that this

widespread clone includes Stoke W (0111:H-), a strainoriginally recovered from an epidemic of diarrhea in Scotlandin 1947 (35). Stoke W has the same adherence phenotype(LA' FAS+), hybridizes with both EAF and EAE geneprobes, and causes attaching and effacing lesions in experimen-tal infections in the rabbit ileum (27).ET 1 clone. The second most common clone identified in

Brazilian 0111 strains, marked by ET 1, is only distantlyrelated in chromosomal genotype to the ET 12 (0111:H2)clone as indicated by the observation that the multilocusenzyme genotypes differ in 35% of the enzyme loci. Thisdivergence in genetic background is accompanied by strikingdifferences in pathogenesis between these 0111 clones: ET 1isolates do not have the characteristics of class I EPEC butinstead show either AA or DA properties. Because the adher-ence properties of some isolates of the clone are not clear-cutin standard adherence assays, more work is required to im-prove assays for determining adhesion and other aspects ofvirulence of these organisms.The hypothesis that ET 1 distinguishes a pathogenic clone

that is genetically divergent from the classical EPEC 0111:H2clone gains support from early studies indicating the diversityand clonal nature of 0111 strains. In 1950 Kauffmann andDupont (10) demonstrated that 0111 strains isolated fromgeographically separate outbreaks of infantile diarrhea com-prised two serologically distinct (H2 and H12) types thatdiffered in many biochemical traits. Stenderup and 0rskov (33)found that both 0111:H2 and 0111:H12 serotypes wereassociated with distinct biotypes and major outer membraneprotein patterns and thus defined two stable widespread clones

INFECT. IMMUN.

on June 19, 2018 by guesthttp://iai.asm

.org/D

ownloaded from

VIRULENCE PROPERTIES OF E. COLI SEROGROUP 0111 CLONES 3287

common in infantile diarrhea. Whittam and colleagues identi-fied a clone (labeled DEC 6 in reference 35) based onmultilocus enzyme electrophoresis whose serotype was 0111:H12 and which belonged to an evolutionary lineage distinctfrom other classical EPEC strains. Comparison of the enzymeallele profile of ET 1 described here and DEC 6 shows thatthese ETs are identical at 20 enzyme loci (data not shown).Together these results indicate that ET 1 delineates a bacterialclone that has spread worldwide and is divergent in manygenotypic and phenotypic traits, including adherence proper-ties, from the common EPEC 0111:H2 clone.ET 8 clone. A third pathogenic clone of the 0111 serogroup

was defined by ET 8 and had the virulence factors of EHECstrains; that is, most ET 8 isolates were SLT+ EAE+ andcarried sequences homologous to the EHEC plasmid. This0111 clone cannot be identified by H type or specific virulencefactors because all strains were nonmotile and some strains didnot harbor either SLT or EHEC sequences, an observationsuggesting that these factors are frequently lost in culturestorage. However, we hypothesize that ET 8 marks a stableEHEC clone because organisms of the same ET were isolatedin Brazil and North America and ET 8 matches the profile ofa previously described bacterial clone, referred to as DEC 8,whose members typically have serotype 0111:H8 and produceSLT (35). SLT-producing 0111:H8 strains have been classifiedas EHEC by Levine (14). It is interesting that the seven strainsof this clone had the ac combination of 0 antigens (6), acharacteristic that we found only in one ET 1 strain, in the ET2 and ET 6 strains, and in three ET 9 strains.The idea that ET 8 has diverged from other 0111 clones and

has a different mechanism of pathogenesis is supported by theresults of experimental infections in animal models. An SLT-producing 0111 nonmotile strain (strain 10049 in reference 5),which is identical to ET 8 in its enzyme allele profile (data notshown), causes an immunocompromising effect in gnotobioticpigs, including loss of lymphogranular complexes in the sub-mucosa of the large intestine and reduction in the number oflymphocytes in peripheral blood (5). In contrast, experimentalinfections with strain 2430 (0111:NM), which has the sameallele profile as ET 12 (data not shown), does not result in thesame immunocomprising effect (5). The close relatedness ofstrain 10049 to ET 8 isolates from Brazil suggests that mem-bers of the 0111 EHEC clone marked by ET 8 may havesimilar immunocompromising effects in animal models and inhuman disease.ET 9 clone. This clone includes H9 and nonmotile strains as

well as a strain whose H antigen could not be determined. AllET 9 isolates were EAE+, and most of them were FAS+ andE-Hly+. This combination of virulence factors differs from theother major 0111 clones, but strains with similar characteris-tics were found in ET 1 and ET 8. This kind of virulencepattern seems to be relatively common among strains of otherE. coli serogroups, such as 055 and 026, and may reflect anovel mode of pathogenesis.

In general, we found good agreement between the results ofthe hybridization with gene probes and the expression of theappropriate adherence phenotype. For example, over all iso-lates, there were 109 probe-positive EAF strains and all butone strain had the LA phenotype. The single exception was anisolate of ET 7 (Table 3) which exhibited an indefinite patternin the cell adherence assay. Other studies (7, 31) have reportedsimilar results involving EAF+ LA- strains. The reason un-derlying this discrepancy is unknown but presumably involvesmutations at loci that influence the full expression of the LAphenotype. Another discrepancy was observed for an isolate ofET 1 which exhibited the AA adherence pattern but was

negative with the AA probe. The basis for this unexpectedresult is problematic. Finally, there were several cases in whichstrains were EAE probe positive but failed to show positiveresults in the FAS assay (Table 3). It is known that mutationsin the eae gene cluster may affect the function of intimin (13),and there are mutants with abolished FAS activity that are ableto adhere intimately to cells (28). We hypothesize that theEAE+ FAS- strains observed here are natural variants thathave accumulated mutations in one or more of the genes of theeae cluster and have lost full FAS activity.The clonal framework provided by multilocus enzyme elec-

trophoresis is clearly useful for identifying homogeneousgroups of related bacteria, especially for the nonmotile 0111strains that are recovered in high frequency. The comparisonof the genetic distances to the distribution of virulence factorssuggests that even relatively closely related strains, separatedat a genetic distance of only 0.10, can strongly differ invirulence attributes. This observation suggests that the rate ofacquisition and loss of virulence factors, such as the EAFplasmid or SLT genes, is rapid relative to the rate of accumu-lation of mutations at housekeeping genes in the evolution of0111 lineages. The finding that the eaeA gene also occurs inhighly divergent 0111 lineages suggests that it has undergonelateral transfers in the past via plasmids or phages. Furtherstudy of the sequence variation of specific virulence determi-nants will yield insights into the history of acquisition and lossand help elucidate the evolution of pathogenic clones. Finally,the question remains as to what extent the 0111 antigen itselfcontributes to the virulence of diarrheagenic E. coli (23).

ACKNOWLEDGMENTSWe thank M. Magalhaes (Recife), A. C. Montelli (Botucato), and

C. E. Levy (Ribeirao Preto) for supplying bacterial cultures.This work was supported by Fundacao de Amparo a Pesquisa do

Estado de Sao Paulo (FAPESP) and by Public Health Service grantsAl 24566 and Al 00964 from the National Institutes of Health.

REFERENCES1. Baudry, B., S. J. Savarino, P. Vial, J. B. Kaper, and M. M. Levine.

1990. A sensitive and specific DNA probe to identify enteroaggre-gative E. coli: a recently discovered diarrheal pathogen. J. Infect.Dis. 161:1249-1251.

2. Beutin, L., J. Prada, S. Zimmermann, R. Stephan, I. 0rskov, andF. 0rskov. 1988. Enterohemolysin, a new type of hemolysinproduced by some strains of enteropathogenic E. coli (EPEC).Zentralbl. Bakteriol. Hyg. A 267:576-588.

3. Bilge, S. S., C. R. Clausen, W. Sau, and S. L. Moseley. 1989.Molecular characterization of a fimbrial adhesin, F1845, mediatingdiffuse adherence of diarrhea-associated Escherichia coli to HEp-2cells. J. Bacteriol. 171:4281-4289.

4. Bray, J. 1945. Isolation of antigenically homogeneous strains ofBact. coli neapolitanum from summer diarrhoea of infants. J.Pathol. Bacteriol. 57:239-247.

5. Christopher-Hennings, J., J. A. Willgohs, D. H. Francis, U. A. K.Raman, R. A. Moxley, and D. J. Hurley. 1993. Immunocompro-mise in gnotobiotic pigs induced by verotoxin-producing Esche-richia coli (O111:NM). Infect. Immun. 61:2304-2308.

6. Edwards, P. R., and W. H. Ewing. 1972. Identification of Entero-bacteriaceae, 3rd ed. Burgess Publishing Company, Minneapolis.

7. Gomes, T. A. T., M. A. M. Vieira, I. K. Wachsmuth, P. A. Blake,and L. R. Trabulsi. 1989. Serotype-specific prevalence of Esche-richia coli strains with EPEC adherence factor genes in infantswith and without diarrhea in Sao Paulo, Brazil. J. Infect. Dis.160:131-135.

8. Gurwith, M., D. Hinde, R. Gross, and B. Rowe. 1978. A prospec-tive study of enteropathogenic Escherichia coli in endemic diar-rheal disease. J. Infect. Dis. 137:292-297.

9. Jerse, A. E., J. Yu, B. D. Tall, and J. B. Kaper. 1990. A geneticlocus of enteropathogenic Escherichia coli necessary for the pro-

VOL. 62, 1994

on June 19, 2018 by guesthttp://iai.asm

.org/D

ownloaded from

3288 CAMPOS ET AL.

duction of attaching and effacing lesions on tissue culture cells.Proc. Natl. Acad. Sci. USA 87:7839-7943.

10. Kauffmann, F., and A. Dupont. 1950. Escherichia strains frominfantile epidemic gastro-enteritis. Acta Pathol. Microbiol. Scand.27:552-563.

11. Knutton, S., T. Baldwin, P. H. Williams, and A. S. McNeish. 1989.Actin accumulation at sites of bacterial adhesion to tissue culturecells: basis of a new diagnostic test for enteropathogenic andenterohemorrhagic Escherichia coli. Infect. Immun. 57:1290-1298.

12. Knutton, S., A. D. Phillips, H. R. Smith, R. J. Gross, R. Shaw, P.Watson, and E. Price. 1991. Screening for enteropathogenicEscherichia coli in infants with diarrhea by the fluorescent-actinstaining test. Infect. Immun. 59:365-371.

13. Law, D. 1994. Adhesion and its role in the virulence of entero-pathogenic Escherichia coli. Clin. Microbiol. Rev. 7:152-173.

14. Levine, M. M. 1987. Escherichia coli that cause diarrhea: entero-toxigenic, enteropathogenic, enteroinvasive, enterohemorrhagic,and enteroadherent. J. Infect. Dis. 155:377-389.

15. Levine, M. M., and R. Edelman. 1984. Enteropathogenic Esche-richia coli of classic serotypes associated with infant diarrhea:epidemiology and pathogenesis. Epidemiol. Rev. 6:31-51.

16. Levine, M. M., J. Xu, J. B. Kaper, H. Lior, V. Prado, B. Tall, J.Nataro, H. Karch, and K. Wachsmuth. 1987. A DNA probe toidentify enterohemorrhagic Escherichia coli of 0157:H7 and otherserotypes that cause hemorrhagic colitis and hemolytic uremicsyndrome. J. Infect. Dis. 156:175-182.

17. Maas, R. 1983. An improved colony hybridization method withsignificantly increased sensitivity for detection of single genes.Plasmid 10:296-298.

18. Moyenuddin, M., I. K. Wachsmuth, S. L. Moseley, C. A. Bopp, andP. A. Blake. 1989. Serotype, antimicrobial resistance, and adher-ence properties of Escherichia coli strains associated with out-breaks of diarrheal illness in children in the United States. J. Clin.Microbiol. 27:2234-2239.

19. Nataro, J. P., M. M. Baldini, J. B. Kaper, R. E. Black, N. Bravo,and M. M. Levine. 1985. Detection of an adherence factor ofenteropathogenic Escherichia coli with a DNA probe. J. Infect.Dis. 152:560-565.

20. Nataro, J. P., J. B. Kaper, R. Robins-Browne, V. Prado, P. Vial,and M. M. Levine. 1987. Patterns of adherence of diarrheagenicEscherichia coli to HEp-2 cells. Pediatr. Infect. Dis. 6:829-831.

21. Nataro, J. P., I. C. A. Scaletsky, J. B. Kaper, M. M. Levine, andL. R. Trabulsi. 1985. Plasmid-mediated factors conferring diffuseand localized adherence of enteropathogenic Escherichia coli.Infect. Immun. 48:378-383.

22. Newland, J. W., and R. J. Neill. 1988. DNA probes for Shiga-liketoxins I and II and for toxin-converting bacteriophages. J. Clin.Microbiol. 26:1292-1297.

23. 0rskov, F., and I. 0rskov. 1992. Escherichia coli serotyping and

disease in man and animals. Can. J. Microbiol. 38:699-704.24. 0rskov, F., T. S. Whittam, A. Cravioto, and I. 0rskov. 1990.

Clonal relationships among classic enteropathogenic Escherichiacoli (EPEC) belonging to different 0 groups. J. Infect. Dis. 162:76-81.

25. Paulozzi, L. J., K. E. Johnson, L. M. Kamahele, C. R. Clausen,L. W. Riley, and S. D. Helgerson. 1986. Diarrhea associated withadherent enteropathogenic Escherichia coli in an infant andtoddler center, Seattle, Washington. Pediatrics 77:296-300.

26. Rigby, P. W. J., M. Dieckmnan, C. Rhodes, and P. Berg. 1977.Labeling deoxyribonucleic acid to high specific activity in vitro bynick translation with DNA polymerase. J. Mol. Biol. 113:237-251.

27. Robins-Browne, R. M., W. C. Yam, L. E. O'Gorman, and K. A.Bettleheim. 1993. Examination of archetypal strains of entero-pathogenic Escherichia coli for properties associated with bacterialvirulence. J. Med. Microbiol. 38:222-226.

28. Rosenshine, I., M. S. Donnenberg, J. B. Kaper, and B. B. Finlay.1992. Signal transduction between enteropathogenic Escherichiacoli (EPEC) and epithelial cells: EPEC induces tyrosine phosphor-ylation of host cell proteins to initiate cytoskeletal rearrangementand bacterial uptake. EMBO J. 11:3551-3560.

29. Scaletsky, I. C. A., M. L. M. Silva, M. R. F. Toledo, B. R. Davis,P. A. Blake, and L. R. Trabulsi. 1985. Correlation betweenadherence to HeLa cells and serogroups, serotypes, and biosero-types of Escherichia coli. Infect. Immun. 49:528-532.

30. Scaletsky, I. C. A., M. L. M. Silva, and L. R. Trabulsi. 1984.Distinctive patterns of adherence of enteropathogenic Escherichiacoli to HeLa cells. Infect. Immun. 45:534-536.

31. Scotland, S. M., H. R. Smith, and B. Rowe. 1991. Escherichia coli0128 strains from infants with diarrhea commonly show localizedadhesion and positivity in the fluorescent-actin staining test but donot hybridize with an enteropathogenic E. coli adherence factorprobe. Infect. Immun. 59:1569-1571.

32. Selander, R. K., D. A. Caugant, H. Ochman, J. M. Musser, M. H.Gilmour, and T. S. Whittam. 1986. Methods of multilocus enzymeelectrophoresis for bacterial population genetics and systematics.Appl. Environ. Microbiol. 51:873-884.

33. Stenderup, J., and F. 0rskov. 1983. The clonal nature of entero-pathogenic Escherichia coli strains. J. Infect. Dis. 148:1019-1024.

34. Viljanen, M. K., T. Peltola, S. Y. T. Junnila, L. Olkkonen, H.Jairvinen, M. Kuistila, and P. Huovinen. 1990. Outbreak ofdiarrhoea due to Escherichia coli O111:B4 in schoolchildren andadults: association of Vi antigen-like reactivity. Lancet 336:831-834.

35. Whittam, T. S., M. L. Wolfe, I. K. Wachsmuth, F. 0rskov, I.0rskov, and R. A. Wilson. 1993. Clonal relationships amongEscherichia coli strains that cause hemorrhagic colitis and infantilediarrhea. Infect. Immun. 61:1619-1629.

INFEC-F. IMMUN.

on June 19, 2018 by guesthttp://iai.asm

.org/D

ownloaded from