Host-Specific Fimbrial Adhesins ofNoninvasive ... · coli. Theiradhesiveproperties canhelpto...

MICROBIOLOGICAL REVIEWS, June 1982, p. 129-161 Vol. 46, No. 20146-0749/82/020129-33$02.00/0

Host-Specific Fimbrial Adhesins of NoninvasiveEnterotoxigenic Escherichia coli Strains

WIM GAASTRA AND FRITS K. DE GRAAF*Department of Microbiology, Biological Laboratory, Vrije Universiteit, Amsterdam, The Netherlands

INTRODUCTION ...............................................................OCCURRENCE.................................................................

Association of Adhesins and 0 Serogroups......................................Association of Adhesin and Enterotoxin Production ................................

HEMAGGLUTINATION ........................................................PRODUCTION AND DETECTION ............................................CHARACTERIZATION .........................................................The K88 Adhesins .............................................................The 987P Adhesin .............................................................The K99 Adhesin ............................................................The F41 Adhesin ..............................................................The CFA/I Adhesin...........................................................Other Adhesins of Human Enterotoxigenic E. coli Strains...........................

ADHESION ....................................................................Adhesion to Intesinal Epithelial Cells ............................................Nature of the Receptor Sites for Adhesins.......................................

EMMUNITY ....................................................................GENETICS.....................................................................

Genetic Organization of the K88 Determinant .....................................Subcellular Location of Polypeptides Encoded by the K88ab Determinant .............Molecular Cloning of the K9 Determinant .......................................Characterization of Plasmids Encoding CFA/1 and CFA/II..........................

CONCLUDING REMARKS ......................................................LITERATURE CITED...........................................................

129130130132133135137137141141143143144145145148149151151153153154155156

INTRODUCTION

The ability of certain bacteria to adhere toeucaryotic cells is recognized as a fundamentalfeature for the colonization of host tissues invivo. For many pathogenic organisms, adher-ence is ofparamount importance since they haveto compete with the commensal microorganismsfor a successful colonization of the host epitheli-al cell surfaces or the skin. Specific adhesivemechanisms have been found to play an impor-tant role in the attachment of Escherichia coli,Neisseria, Streptococcus, and Mycoplasma spe-cies to various epithelial tissues.

Enteric diseases caused by enteropathogenicE. coli strains can be classified into at least twodifferent forms. Some strains are invasive andcause a dysentery-like syndrome (enterocolitis);others isolated from diarrheal illnesses are non-invasive enterotoxigenic strains. The latterstrains are characterized (i) by their ability toproliferate in the small intestine and (ii) by theproduction of one or both of two types ofenterotoxins. Colonization of the mucosal epi-thelium is mediated by specific adhesins withwhich the cells can resist the flushing action ofthe peristalsis of the gut. The enterotoxins stim-ulate the epithelial cells to secrete fluid into the

lumen of the gut, thereby causing the actualdiarrhea. Two types of enterotoxin can be distin-guished on the basis of their thermostability.The surface of E. coli cells is covered with

substances capable of provoking immunologicalreactions which are commonly used for theserological classification of pathogenic and non-pathogenic strains. These surface antigens in-clude substances referred to as adhesins whichare responsible for the attachment of the cells tothe epithelial mucosa. The most common anti-gens of E. coli are the 0, K, and H antigens. Theterminology 0 (from the German: Ohne Hauch)and H (from the German: Hauch), introduced byWeil and Felix (190, 191) is used for the somatic(0) and flagellar (H) antigens of E. coli. Thesomatic 0 antigens are lipopolysaccharide com-plexes and constitute part of the outer mem-brane. They are thermostable and not inactivat-ed by temperatures of 100 or 121°C. The flagellarH antigens are protein in nature and are inacti-vated at 100°C. The term K antigen (from theGerman word Kapsel) was introduced by Kauff-mann and Vahlne (86). They are usually acidicpolysaccharides which form an envelope or cap-sule around the cell wall.A new type ofK antigen, numbered K88, was

first described in 1961 in a paper about the

129

on July 18, 2020 by guesthttp://m

mbr.asm

.org/D

ownloaded from

http://mmbr.asm.org/

130 GAASTRA AND DE GRAAF

serological analysis ofa number ofE. coli strainsfrequently isolated from edema disease and en-

teritis in swine (133). Since the discovery of theK88 antigen, the surface antigen characteristicsof porcine enterotoxigenic E. coli strains haveattracted considerable attention.

It was recognized that porcine neonatal diar-rhea is characterized by proliferation of certainserotypes of E. coli in the small intestine, manyof which bear the K88 antigen on their surface(176). The ability of these E. coli strains toproliferate and cause disease has been attributedto their adherent properties for the piglets' intes-tinal epithelium (4, 8, 34, 195). More recently,Nagy et al. (123, 124) reported the colonizationof the porcine small intestine by enterotoxigenicE. coli strains that lack the K88 antigen. Thesestrains were found to possess another adhesin,designated as 987P.

In 1972, Smith and Linggood (166) reported a

discovery made by Sojka that many calf andlamb enteropathogenic strains, although havingdifferent 0 antigens, possess a closely related orcommon K antigen. This antigen was referred toas the "common K antigen," KCo, and was

later designated as K99 (132). K99-positivestrains proliferate in the small intestine whereasK99-negative strains do not. For this reason itwas conceivable that the K99 antigen functionsin the same manner as the K88 antigen byfacilitating adhesion of the bacteria to the intesti-nal epithelium (17).A surface antigen, with characteristics similar

to the K88 antigen, on enterotoxigenic E. coli ofhuman origin was first described by Evans et al.(51). This antigen was found on an E. coli strain(H-10407) isolated from the liquid stool of a

patient with severe cholera-like diarrhea in Dac-ca, Bangladesh, and was termed colonizationfactor antigen (CFA). This surface antigen was

called CFA/I when a second, immunologicallydistinct, surface antigen from human enterotoxi-genic E. coli was discovered. The second sur-

face antigen was consequently termed CFA/II(42).

All of these surface antigens which conferadhesive properties on the enterotoxigenicstrains have been identified as nonflagellar, fila-mentous, proteinaceous appendages on the bac-terial cell surface. The existence of this type ofsurface appendage on the surface of differentspecies of bacteria was already reported in 1949by Anderson (3) and by Houwink and van

Iterson in 1950 (70). The most common termi-nologies for these structures are "fimbriae,"introduced by Duguid et al. (38), and "pili,"introduced by Brinton (13). Following the rec-

ommendation of Ottow (136) and Jones (80), weused the term fimbriae. Fimbriae are thinner andmore numerous than flagella and confer on bac-

teria adhesive properties including hemaggluti-nating activity. Both Brinton (13) and Duguidand co-workers (36) distinguished several typesof fimbriae among Enterobacteriaceae mainlybased on their morphology and hemagglutinatingproperties. The most common type of fimbriaeare named as type 1 fimbriae.The majority of both pathogenic and non-

pathogenic E. coli strains produce type 1 fimbri-ae. These fimbriae enable E. coli to adhere to awide variety of eucaryotic cells, i.e., erythro-cytes of several animal species (150), leukocytes(10), epithelial cells (127), and most other ani-mal, plant, and fungal cells as far as tested (37).The adhesive properties of type 1 fimbriae areinhibited by D-mannose. It is clear that type 1fimbriae are the most universal adhesin of E.coli. Their adhesive properties can help to colo-nize epithelial surfaces and therefore contributeto the pathogenicity of enteropathogenic E. colistrains. Several reviews describe the character-istics of type 1 fimbriae (13, 14, 37, 136). Thisreview is restricted to the description of anotherclass of E. coli fimbriae characterized as host-specific adhesins. In contrast to type 1 fimbriae,these adhesins seem to occur exclusively on theenterotoxigenic E. coli strains. They adhere tothe intestinal epithelia of a very limited numberof animal species and specifically agglutinateonly certain species of erythrocytes. In contrastto type 1 fimbriae, the hemagglutinating activityof these adhesins is not inhibitable by mannose.

OCCURRENCE

Association of Adhesins and 0 SerogroupsIn contrast to type 1 fimbriae which are pres-

ent on the majority of both commensal andpathogenic E. coli strains, the presence of host-specific adhesins on enterotoxigenic strainsseems to be associated with a relatively smallnumber of 0 serogroups (Table 1). Frequentlyoccurring (classical) serotypes of E. coli associ-ated with diarrhea in newborn piglets are08:K87, 045, 0138:K81, 0141:K85, 0147:K89,0149:K91, and 0157 (175). Incidentally, K88-positive E. coli strains belonging to other 0serogroups have been isolated from diseasedpigs (60). Furthermore, the occurrence of K88-positive enterotoxigenic E. coli strains of vari-ous 0 groups is known to fluctuate with thegeographical area and time, as well as with thevaccination policy of a particular country. So-derlind and Mollby (172) examined E. colistrains isolated from 200 piglets with neonataldiarrhea in Sweden and found that one-third ofthe strains were enterotoxigenic. 0 group 149comprised 24% of all strains, and all but one ofthe strains belonging to 0 group 149 possessedthe K88 antigen. In another comparative study

MICROBIOL. REV.


mbr.asm

.org/D

ownloaded from


NONINVASIVE ENTEROTOXIGENIC E. COLI STRAINS 131

TABLE 1. Occurrence of adhesins in relation to 0serogroups

Adhesin Origin 0 serogroupsaK88 Piglet 08, 045, 0138, 0141,

0147, 0149, 0157987P Piglet 09, 020, 0141K99 Calf, lamb 08, 09, 020, 0101K99 Piglet 064, 0101F41 Calf 09,0101CFA/I Human 015, 025, 063, 078CFAIII Human 06,08

a Strains belonging toO serogroups which have onlyoccasionally been found to possess one of the adhesinsare not included.

on piglets with diarrhea and healthy piglets ofthe same age, Soderlind and Moilby (173) isolat-ed 810 intestinal E. coli strains from 81 piglets. Aclear difference was found between E. colistrains isolated from both groups of piglets withregard to the distribution of 0 groups, K88antigen production, and the frequency of entero-toxigenicity. The K88 antigen was found exclu-sively in enterotoxigenic strains of0 groups 149and 8. Guinde and Jansen (62) studied 101 enter-otoxigenic E. coli strains ofporcine origin isolat-ed in the Netherlands. The K88 antigen wasfound on 52 of these strains. Six strains hadanother porcine-specific adhesin (987P), andseven strains possessed the K99 antigen. Of theK88-positive strains, 92% belonged to the so-called classical serogroups, and the others be-longed to serogroups 09 and 020.Nagy et al. (124) investigated the prevalence

of the 987P antigen among porcine enterotoxi-genic E. coli strains that lack the K88 antigen. Of119 strains tested, all belonging to 0 groups 9,20, and 101, 50% of the 09 and 14% of the 020serogroup strains produced colonies that wereagglutinated by anti-987P serum, whereas noneof the strains of serogroup 0101 reacted with theantiserum. In another study, Moon et al. (109)tested 111 enterotoxigenic E. coli isolates fromthe intestines of neonatal pigs with diarrhea thatdid not react in preliminary tests with K88antiserum. Fifty-five of these strains were 987Ppositive, 9 isolates produced K99, and 4 pro-duced the K88 antigen. The 987P adhesin wasfound to be most common in strains of sero-groups 09, 020, and 0141, and occasionally in0 groups 101 and 149 (Table 1). Apparently, themajority of enterotoxigenic strains isolated fromneonatal pigs belong to a restricted number of 0serogroups and produce either the K88, 987P, orK99 adhesin. Porcine enterotoxigenic strainswhich lack these antigens may produce anotherhitherto unrecognized type of adhesin whichmediates their adherence to the intestinal epithe-lial cells. Recently, Awad-Masalmeh et al. (5)

studied three of such strains, designated as 3P-,in more detail. The strains adhered to and colo-nized the pig intestine and produced diarrhea.Furthermore, they exhibited mannose-resistanthemagglutinating activity. The authors could notdemonstrate that these activities were associat-ed with a particular type of fimbriae, but theywere able to exclude a role for the type 1fimbriae produced by these strains in mediatingadhesion to the intestinal epithelial cells of pigs.The occurrence of the K88 as well as the 987P

adhesin appears to be restricted to porcine en-terotoxigenic E. coli strains. K88- or 987P-posi-tive strains have not been reported among en-terotoxigenic strains isolated from other animalsor humans.

E. coli strains isolated from calves or lambs inepidemiologically unrelated cases of diarrheawere examined by 0rskov et al. (132) for thepresence of the K99 antigen. All K99-positivestrains belonged to serogroups 08, 09, and0101. Myers and Guinee (121) reported theserotyping of 35 enterotoxigenic E. coli strainsisolated from calves in the United States. Thefollowing serotypes were detected: 08:K25,08:K85, 09:K35, 020:K?, 0101:K28, and0101:K30. In another study, Guinde and Jansen(62) described that 74 unselected field isolates ofbovine enterotoxigenic E. coli strains from theNetherlands and the United States belonged toserogroups 08 (16%), 09 (27%), 020 (3%), and0101 (54%). Several other reports on the occur-rence of the K99 antigen among bovine entero-toxigenic strains (11, 112, 158) confirmed thefrequent association of K99 production with 0groups 8, 9, 20, and 101 (Table 1).Moon et al. (110) reported the isolation of

K99-positive enterotoxigenic E. coli strains frompigs. Several enterotoxigenic E. coli strains ofhuman origin were also tested for the presenceof the K99 antigen, but all were negative. Fur-thermore, K99-positive enterotoxigenic E. colistrains of both calf and pig origin produced K99in the pig ileum in vivo, adhered to the pigepithelium, and caused profuse diarrhea in new-born pigs. The K99 antigen appears to occurrather commonly among pig strains of 0 groups64 and 101 and occasionally also on strains of 0group 9. These data indicate that although K99also facilitates adhesion to and colonization ofthe pig intestine, not all K99-positive enterotoxi-genic E. coli are able to do so.The occurrence of the K99 antigen seems to

be restricted to bovine, ovine, and porcine en-terotoxigenic strains.The adhesins of enterotoxigenic E. coli strains

associated with diarrhea in adults and childrenwere discovered more recently and designatedas CFA/I and CFA/II (42, 51). The enterotoxi-genic strains possessing these adhesins are con-

VOL. 46, 1982


mbr.asm

.org/D

ownloaded from



sidered a major cause of traveller's diarrheaand diarrhea among young children in develop-ing countries. The occurrence of CFA/I andCFA/II on strains of human origin also appearedto be associated with a very restricted number ofserotypes. CFA/I is found predominantly onstrains of serogroup 078, and CFAIII is foundpredominantly on strains of serogroups 06 and08 (Table 1). Less frequently, both adhesinswere observed on strains of other serogroups.Table 2 summarizes some reports on the occur-rence of the CFA/I adhesin among human en-terotoxigenic E. coli strains of various serotypesisolated at different locations all over the world.Although the CFA/I adhesin was frequently en-countered among the human enterotoxigenic E.coli strains, an important percentage of thesestrains does not possess the CFA/I or the CFA/II adhesin (23, 42). CFAJI and CFA/II, there-fore, are not prerequisites of virulence for all E.coli strains that cause diarrhea in humans (149).In general, however, it should be stressed thatunderestimation of the frequency of occurrenceof a particular adhesin may easily occur if onedoes not take into account that the production ofthese adhesins is influenced by growth condi-tions.

Association of Adhesin and EnterotoxinProduction

Enterotoxigenic E. coli strains associated withacute diarrhea in man and domestic animals allproduce a heat-stable (ST) and/or a heat-labile(LT) enterotoxin (148). The production of bothtypes of enterotoxins is controlled by transmissi-ble plasmids (66, 163). The correlation betweenenterotoxigenicity and the presence of adhesins

TABLE 2. Occurrence of CFA/I on enterotoxigenicE. coli strains of human origina

No. ofNo. of CFA/I-strains posi- 0 serogroupb encetested tive

strains

26 5 078 (5) 5450 22 078 (22) 660 30 015 (3), 025 (11), 063 (6), 42

078 (10)76 9 062 (1), 063 (4), 0128ac 141

(3)77 16 078 (16) 13089 6 063 (1), 078 (4), 0153 (1) 59178 38 078 (34) 170187 31 025 (2), 063 (9), 078 (18), 22

0128 (1), 0153 (1)a Strains isolated in various geographic areas.b Numbers in parentheses represent the number of

CFA/I-positive strains belonging to that particular 0serogroup.

TABLE 3. Association of adhesin and enterotoxinproduction

EnterotoxinAdhesin

ST LT ST + LT

K88 + + +987P + - -K99 + - -

F41 + - -CFA/I + + +CFAIII - - +

has been the subject of many studies to establishthe enteropathogenicity of E. coli strains isolat-ed from stools.

Classical pig enterotoxigenic E. coli strainswere originally thought to fall into two catego-ries: strains producing only ST and strains pro-ducing both ST and LT. The K88 antigen wasfound to be associated with strains producingboth enterotoxins (65-67, 161). Guinee andJansen (62), however, reported that 37 out of 52K88-positive strains produced only LT whereasthe other 15 strains produced ST + LT. Soder-lind and Mollby (173) reported an almost com-plete correlation between K88 production instrains belonging to serogroup 0149 and theproduction of both ST and LT. Remarkably,however, they observed that the K88-positivestrains belonging to serogroup 08 produced onlyST. Obviously, the presence of the K88 antigenmay be associated with ST, LT, or both entero-toxins (Table 3).The second adhesin detected on pig entero-

toxigenic strains, 987P, is always found to beassociated with the production of ST (62, 109;Table 3). Also, for the K99 antigen, a positivecorrelation has been reported with ST produc-tion (62, 77, 112; Table 3). With bovine strains,Guinee and Jansen (62) detected a 100%o positivecorrelation between ST production and the pres-ence of the K99 antigen on 74 strains. None ofthe 43 non-enterotoxigenic strains of bovineorigin possessed the K99 antigen. Kaeckenbeeket al. (84) reported that 99o of the calf entero-toxigenic strains they studied possessed the K99antigen, but in addition the K99 antigen wasdetected on 21% of the 441 non-enteropathogen-ic strains. Moon et al. (112) as well as Isaacsonet al. (77) reported that only 77% and 87%,respectively, of the enterotoxigenic strains pos-sessed the K99 antigen. Differences in the fre-quency of association of the K99 antigen withenterotoxigenic strains are probably due to diffi-culties in the detection of the K99 antigen (63,132). However, the presence of unknown adhe-sins cannot be excluded. In most porcine andbovine enterotoxigenic E. coli strains, the pro-duction of adhesins and of enterotoxins ap-

MICROBIOL. REV.


mbr.asm

.org/D

ownloaded from



peared to be encoded on separate plasmids. Insome K99-positive strains, however, the abilityto produce K99 and ST appears to be encodedby a single plasmid (C. L. Gyles, personalcommunication).A close association exists between the pres-

ence of CFAJI and the production of ST + LT or

ST only (42, 45, 59, 95, 141, 160). CFAII-positivestrains producing only LT seem to be relativelyrare (95). The production of CFA/I and ST hasbeen shown to be controlled by a single non-

conjugative plasmid (23, 51, 99, 140, 160) ofabout 60 x 106 daltons. The size of this plasmidisolated from various sources is somewhat vari-able. Occasionally a plasmid encoding the pro-duction of all three properties (CFA/I, LT, andST) has been observed (99). However, in mostcases the production of LT is encoded by a

separate plasmid (100, 186). Compared with E.coli strains which produce only ST, the diarrheacaused by E. coli producing both enterotoxins ismore severe and oflonger duration (103). Analysisof CFA/II-positive enterotoxigenic E. coli strainsfrom different geographic locations indicatedthat the majority of these strains produce bothST and LT (23, 42). The genes encoding thethree properties were found to be located on a

single plasmid of about 60 x 106 daltons (138).The data indicate that, in contrast to enterotoxi-genic E. coli strains associated with diarrhea indomestic animals, human enterotoxigenicstrains harbor plasmids which encode for bothadhesin and enterotoxin production. Conse-quently, the occurrence of CFA/I and CFA/II ina very restricted number of E. coli serotypesmay be explained by the close association ofenterotoxigenicity with certain serotypes (6,102, 131). The reason for this association be-tween chromosomally encoded cell wall anti-gens and plasmid-determined enterotoxin and/oradhesin production is not yet understood. It hasbeen suggested that one of the factors that mightcontribut to this nonrandom association mightbe the ability of certain strains to accept and toharbor the virulence plasmids more stably thanothers (47). Enterotoxigenic strains of humanorigin can be divided into different groups de-pending on the tendency of strains with respec-tive 0 antigens to lose the ability to produce LTand/or ST (6, 47).

HEMAGGLUTINATIONHemagglutination was the first observed man-

ifestation of the adhesive properties of fimbriat-ed enteric bacteria (38). The majority of E. colistrains exhibit a similar pattern of hemagglutin-ating activity with erythrocytes of various ani-mal species. This hemagglutination is mannosesensitive, subject to phase variation, and cor-related with the presence of type 1 fimbriae (12,

38). The host-specific adhesins described in thisreview differ from the adhesins which exhibitmannose-sensitive hemagglutination in severalproperties: their hemagglutinating activity is notinhibitable by D-mannose and its analogs, no

adhesin production occurs on cells grown at18°C, agglutinated bacteria can be eluted fromthe erythrocytes at 37°C, and the hemagglutinat-ing activity is destroyed by heating the cells for30 min at 65°C. Although the terminology "man-nose-resistant hemagglutination" has little sig-nificance because the hemagglutination is also"resistant" to numerous other substances, we

prefer to hold on to it in view of its common use

in the literature (35). In Table 4, the hemagglu-tinating activity of strains carrying various adhe-sins is summarized.

Stirm et al. (179) showed that in two K88-positive strains investigated, the presence of theK88 antigen was associated with the ability toagglutinate guinea pig erythrocytes. The hemag-glutination reaction was best performed in thecold. Unlike hemagglutination by common ortype 1 fimbriae, the hemagglutination by K88fimbriae was mannose resistant (13, 38, 150).Jones and Rutter (82) showed that all of the 108K88-positive E. coli strains tested caused man-

nose-resistant hemagglutination of guinea pigerythrocytes at 0 to 3°C. Hemagglutination was

negative in K88-positive strains grown at 18°C or

in K88-negative mutants of a K88-positivestrain. Extracts containing the K88 antigen pos-

sessed hemagglutinating activity which couldnot be separated from the K88 antigen by frac-tionation or serological procedures. The authorsconcluded that K88 was a hemagglutinin andsuggested that the mechanism of hemagglutina-tion and attachment to the mucosa of pigletsmay be similar. The observation, however, thathemagglutination was effective at 4°C but elutedat 37°C might indicate that the fit for the naturalreceptor of the K88 antigen on the intestinalepithelial cells is better than for guinea pig

TABLE 4. Hemagglutinating activity of strainscarrying various adhesive antigens

Adhesin Orgn Erythrocyteshemagglutinated

K88 Porcine Guinea pig, chicken987P Porcine a

K99 Bovine, ovine, Horse, sheepporcine

F41 Bovine, porcine Human, guinea pig,horse, sheep

CFA/I Human Human, bovine, chickenCFA/II Human Bovine, chicken

a The 987P adhesin fails to hemagglutinate erythro-cytes from horse, guinea pig, rabbit, sheep, pig, orcattle.

VOL . 46, 1982


mbr.asm

.org/D

ownloaded from



erythrocytes. At higher temperatures, the bind-ing of the K88 antigen to the erythrocyte recep-tor may be disrupted as a result of the increaseof molecular agitation. Parry and Porter (137)tested a large number of different types of eryth-rocytes for their agglutination by cell-free prepa-rations of two serological variants of the K88antigen (K88ab and K88ac) as well as by wholecells carrying these antigens. They observed apowerful and consistent affinity of K88ab forchicken erythrocytes. The adhesion was stableat room temperature and mannose resistant.K88ac-positive cells failed to adhere to thechicken erythrocyte membrane, and cell-freeK88ac preparations did not cause hemagglutina-tion. Guinea pig erythrocytes were agglutinatedby both K88ab- and K88ac-positive bacteria.They confirmed the previous observation thatthe reaction with guinea pig erythrocytes wasunstable at room temperature. The number ofcells adhering to erythrocytes was low com-pared to the adhesion observed on microvilli.Hemagglutination of chicken erythrocytes bycell-free K88ab and the adhesion of K88ab-positive bacteria to these erythrocytes wasreadily inhibited by using antisera to K88ab. Anti-sera against K88ac gave a very poor inhibition ofthese reactions. Furthermore, antisera to theK88b and K88c determinant were only capableof inhibiting the hemagglutinating activity oftheir homologous antigen. However, anti-K88acand anti-K88a sera showed extremely low inhi-bition of K88ab agglutination of chicken erythro-cytes compared to their corresponding inhibitionof K88ac agglutination of guinea pig erythro-cytes. Probably, the selectivity of K88ab for thechicken erythrocyte may involve adhesion viathe K88b determinant.The ability of the 987P antigen to hemaggluti-

nate various types of erythrocytes was testedwith purified 987P and whole cells carrying the987P antigen (79). However, no agglutinationreaction was observed either at 4°C or at roomtemperature. This observation does not excludethe possibility that 987P can agglutinate erythro-cytes of types not yet included in the tests.

In a first report on the establishment of theK99 adhesin of calf and lamb enterotoxigenic E.coli strains, 0rskov et al. (132) described thatsome K99-positive strains were able to hemag-glutinate guinea pig erythrocytes in the cold.The reaction was not inhibited by mannose.Tixier and Gouet (183) studied the hemagglutin-ating ability of various enterotoxigenic strains ofbovine, porcine, and human origin with horseand snake erythrocytes at 4°C. The calf entero-toxigenic strains exhibited a strong mannose-resistant hemagglutination of horse erythrocytesand possessed fimbriae which could be the K99adhesin. These strains as well as the other

enterotoxigenic and non-enterotoxigenic strainswere also able to hemagglutinate snake erythro-cytes, but this reaction was mannose sensitiveas indicative for the presence of type 1 fimbriae.Burrows et al. (17) demonstrated that enterotox-igenic strains of calf origin caused mannose-resistant hemagglutination of sheep erythrocytesand were able to demonstrate that the presenceof the K99 antigen on the cell surface wasresponsible for this effect. When grown at 37°C,the K99-positive strains attached to brush bor-ders prepared from the small intestine of calves.The hemagglutinating and adhesive propertieswere not exhibited by strains grown at 180C, atwhich temperature the K99 antigen is not ex-pressed (132). The paired nature of hemaggluti-nation and adhesive properties for the K99 anti-gen were further demonstrated by transfer of theK99 plasmid to a K99-negative recipient strain.Transfer of the K99 plasmid conferred bothproperties on the exconjugants. Cell-free K99antigen also possessed hemagglutinating activityand is able to inhibit the adhesion of K99-positive bacteria to brush borders. At 37r, thecell-free K99 antigen showed no hemagglutinat-ing activity.

Morris et al. (114, 116) confirmed the observa-tion that cell-free K99 antigen is able to hemag-glutinate sheep erythrocytes and also showedthat guinea pig and human type 0 erythrocyteswere agglutinated. Heating of the antigen for 15min at 100°C destroyed the hemagglutinatingactivity. In contrast with these data, Isaacson(71) reported that purified K99 antigen did notreact with guinea pig erythrocytes and suggested(72) that the K99 preparations used by Morrisand co-workers contained an additional adhesinnot identical with K99. The existence of anotheradhesin on certain calf enterotoxigenic strainswas recently established by Morris et al. (117,118). This adhesin, designated as F41, is pro-duced by strains belonging to serogroups 09 and0101. The hemagglutinating activities of bothadhesins were recently reported by De Graafand Roorda (29). Purified K99 antigen showed astrong hemagglutination of horse and a weakerreaction with sheep erythrocytes, but no activitywith guinea pig erythrocytes. On the other hand,purified F41 antigen showed a strong hemagglu-tination reaction with guinea pig erythrocytesand a weaker activity with sheep and horseerythrocytes.

Mannose-resistant hemautination ofhumangroup A erythrocytes by enterotoxigenic strainsof several different serogroups and isolated fromadults with diarrhea was found to be associatedwith the presence of CFA/I on those strains (48).Neither CFA/I nor hemagglutination were pro-duced when the strains were grown at 180C.Hemagglutination was completely inhibited by

MICROBIOL. REV.


mbr.asm

.org/D

ownloaded from



pretreatment of CFA/I-positive cells with anti-CFA/I serum. Cell-free purified CFA/I fimbriaedo not hemagglutinate erythrocytes (44). Thusthe CFA/I released from bacterial cells is mono-valent. Hemagglutination, however, can be ob-tained by sensitizing micro latex beads withpurified CFA/I or by aggregates of the purifiedCFA/I which were induced at low pH. PurifiedCFA/I also retains its affinity for the epithelialcells of the rabbit small intestine and blocks theadhesion of CFA/l-positive bacteria (44). CFA/Ireleased by heating cells at 60°C for 30 min doesno longer agglutinate human erythrocytes (48,54). A hemagglutination typing system for en-

terotoxigenic E. coli possessing the colonizationfactor antigens CFAII or CFA/II has been pro-posed by Evans et al. (46). CFA/I-positive en-

terotoxigenic E. coli exhibit mannose-resis-tant hemagglutination with human, bovine, andchicken erythrocytes with decreasing intensityand not with guinea pig erythrocytes. CFAIII-positive E. coli strains exhibit only mannose-resistant hemagglutination with bovine andchicken erythrocytes at 4°C. The association ofCFA/I and mannose-resistant hemagglutinationof human and bovine erythrocytes and of CFA/II and mannose-resistant hemagglutination ofbovine erythrocytes was also observed by anumber of other investigators (6, 23, 33, 54, 59,95, 130, 131, 170, 182). However, a correlationbetween mannose-resistant hemagglutinationand the presence of colonization factor antigenswas not always observed, i.e., enterotoxigenicE. coli strains producing hemagglutination ofhuman and/or bovine erythrocytes but lackingCFAII or CFA/II were frequently encountered.Furthermore, the use of hemagglutination typingsystems as proposed by Evans et al. (46) may

easily lead to erroneous results if one does notrealize that a particular strain may be able toproduce more than one type of adhesin.

PRODUCTION AND DETECTION

Production of adhesins is generally a functionof the cultivation conditions, i.e., temperatureand composition of the growth medium as wellas of the actual strain. All adhesins of enterotox-igenic E. coli strains studied so far are notproduced at 18°C except for type 1 fimbriaewhich appear to be produced at all tempera-tures. Commercially available nutrient agars,

designed for the isolation of Enterobacteriacae,are often not suitable for the detection of adhe-sins on E. coli isolates. The influence of growthconditions on the detectability and production ofadhesins has been studied extensively for type 1fimbriae (12, 13). Only a limited number ofstudies, mainly concerning the K99 and the 987Pantigen, have been published on the influence of

growth conditions on the expression of host-specific adhesins.As mentioned before, the initial detection of

the K99 antigen was derived from the observa-tion of Sojka that many calf and lamb entero-pathogenic strains, although having different 0antigens, reacted not only with homologous butalso with heterologous OK antisera in slideagglutination tests, suggesting that these strainsmight share one or more common antigens.Detection of this common antigen appeared tobe dependent on the presence of capsular Kantigens and on the medium used for cultivation(132). In strains without capsular K antigen or inmore transparent colonies of mucoid-growingstrains, which have the least amount of capsularantigen, this common antigen could be detectedmore readily. Guinee et al. (63) undertook astudy on the detection of the K99 antigen be-cause of its possible significance for the identifi-cation of calf enterotoxigenic E. coli strains.Identification of these strains by means of sero-typing was hampered by the fact that strainswith the same serotype are often also found inhealthy calves. They developed a buffered semi-synthetic medium (Minca medium), which con-tains less amino acids and carbohydrate thancommon nutrient media (63, 64) to suppress theproduction of capsular K antigens. The dimin-ished production of capsular antigen facilitatedthe detection of the K99 antigen by slide aggluti-nation as well as by immunoelectrophoresis.

In an attempt to explain the difficulties in thedetectability of the K99 antigen on commerciallyavailable nutrient agars, De Graaf et al. (30)determined the extent of K99 production bystrains grown under different conditions. Theyobserved that, instead of Minca medium, a mini-mal salt medium with glucose can be used toobtain a high K99 production. In complex medialike nutrient broth the production of K99 wasstrongly reduced, and slide agglutination testsare not sensitive enough to detect the lowamount of K99 produced under these condi-tions. Optimal amounts of K99 antigen wereproduced at 370C. At 30°C a weak production ofK99 antigen was detected by hydrophobic inter-action chromatography or enzyme-linked im-munosorbent assay (ELISA) (Fig. 1). TheELISA also appeared to be a reliable assay forthe detection of the K99 antigen in calf feces(40). Furthermore, the production of the K99antigen appeared to be related to the 0 antigencarried by the host strains, but seemed to beindependent of the absence or the presence ofvarious K polysaccharide antigens. Under allconditions tested, strains with antigen 0101 pro-duced more K99 antigen than did strains belong-ing to 0 antigen group 8, 9, or 20. De Graaf et al.(26) have been able to give an explanation for the

VOL. 46, 1982


mbr.asm

.org/D

ownloaded from



K88

-26

24

22

15 20 25 30 35 coTemperature (*C)

FIG. 1. Temperature-dependent production of theK88 and K99 adhesins. The adhesin production wasdetermined in sonicated cell suspensions, using anELISA. Titers are determined in standard ultrasonicextracts and expressed as the highest twofold dilutionwhich still showed coloring.

observed differences in K99 production and de-tectability between complex and minimal media.By addition of various compounds to the mini-mal medium, it appeared that alanine specifical-ly represses K99 biosynthesis. A 1 mM concen-tration of L-alanine was sufficient to reduce theK99 production by about 95%. These resultswere confirmed by Contrepois et al. (21). Ex-periments on the kinetics of alanine-inducedrepression indicated an immediate effect on K99production without a significant lag period (Fig.2). D-Alanine also inhibits K99 production whenthe bacteria are grown in a medium with D-alanine as sole carbon source, but the metabolicdegradation product pyruvate showed no effect(26). Another factor that might affect K99 pro-duction is glucose, as indicated by Guinde et al.(64). In a recent paper, Isaacson (73) describedthat high concentrations of glucose (0.5%) re-

pressed K99 production. This repression couldbe overcome by addition of 0.5 mM cyclicadenosine monophosphate, indicating that K99synthesis is subject to cyclic adenosine mono-

phosphate-dependent catabolite repression. Onthe other hand, Contrepois et al. (21) reportedthat two types of K99-producing strains can be

distinguished. In one type of strain K99 produc-tion is not affected by glucose, whereas in theother type K99 production is glucose dependent.The nature of this phenomenon remains to beexplained.The ELISA appeared to be a reliable assay for

the detection of the K99 antigen in calf feces(40).The production of the F41 adhesion antigen is

also dependent on the composition of the growthmedium and comparable to the production of theK99 antigen (29, 30). Like the production of theK99 antigen, the production of F41 antigen isrepressed by alanine, which indicates that theregulation of production of both antigens mayhave a similar molecular basis.

In contrast with the regulatory phenomenaobserved for K99 production, no such effectshave been described for K88 production exceptfor the influence of temperature (Fig. 1). Nosignificant differences in K88 production wereobserved for an E. coli K-12 strain harboring theK88 plasmid and cultivated in a variety of me-dia.

Relatively little is known about cultivationconditions that affect the detectability and pro-duction of CFA/I. Evans et al. (49) prepared ahyperimmune serum against purified CFA/I, andcell suspensions of different CFA/I-positivestrains were tested for CFA/I production bytitrations with this serum. The known CFA/I-positive strains were negative when grown onMacConkey agar. The most efficient medium fordetecting CFA/I on E. coli isolates was a pep-

2.5 2n

2.0

1. 2

lb 1.0

22QL0.5

2 4 6 8 10 12 14

hours

FIG. 2. Kinetics of the alanine-induced repressionof K99 biosynthesis. L-Alanine (10 mM) was added tothe culture at an optical density of 0.2. Symbols: 0,K99 production in the control culture without alanine;A, K99 production in the presence of L-alanine;

, optical density. Titers are expressed as de-scribed in the legend of Fig. 1.

MICROBIOL. REV.


mbr.asm

.org/D

ownloaded from



tone-agar medium composed of 2% peptone,0.5% NaCl, and 2% agar. Tergitol-grown cellswere either negative or gave a very low titer.Maximum CFA/I titers were obtained with anagar medium consisting of 1% Casamino Acids,0.15% yeast extract, 0.005% MgSO4, 0.0005%MnCl2, and 2% agar (pH 7.4), designated asCFA agar (48). The ELISA technique appearedto be a useful tool for the diagnosis of diarrheacaused by CFA/I-positive strains (43).

Brinton (13, 15) has shown that the productionof type 1 fimbriae is subject to a phenotypiccontrol mechanism called phase variation. With-in one bacterial strain, fimbriate and nonfim-briate cells were distinguishable from each otherby electron microscopy and electrophoresis.The electrophoretic mobility of the fimbriatedcells was about half that of the nonfimbriate cells(16). Colonies of fimbriated cells are smaller andsmoother than colonies of nonfimbriated cells(12). Phase variation describes the all-or-noneoscillation between the fimbriate and nonfim-briate state. The probability of switching isstrongly influenced by environmental condi-tions. The change appeared to be spontaneousand takes place in either direction at a rate ofabout once per 1,000 bacteria per generation.Phase variation of type 1 fimbriae was found tobe under transcriptional control (39). Indepen-dently of phase variation, an additional regula-tory mechanism (181) appears to determine thenumber and/or length of fimbriae per cell. Themechanism of this kind of regulation, namedquantitative regulation of fimbriation by Swaneyet al. (181), is not understood at present.With respect to the other adhesins of entero-

toxigenic E. coli strains, the phenomenon ofphase variation and associated changes in colo-ny morphology has been described for the 987Pantigen (55, 124). This adhesin, like type 1fimbriae, is probably also encoded by chromo-somal genes. Phase variation for CFA/I in E. coliH10407 has been demonstrated by Brinton (15).For the other adhesins this phenomenon has notbeen described so far.The possibility of phase variation and the

effect of environmental conditions thereon mustbe taken into account when selecting methodsfor the detection and production of a particulartype of adhesin, especially when cells are capa-ble of synthesizing more than one type offimbri-ae.

CHARACTERIZATION

The K88 AdhesinsIn a first report about the discovery of the K88

antigen, 0rskov et al. (133) described the adhe-sin as a thermolabile substance, numbered asK88 (L), the development of which was sup-

pressed at 18°C. An isolation procedure for theK88 antigen was given by Stirm et al. (178, 180)who released the antigen from the cells byheating a cell suspension for 20 min at 60°C or bytreating a suspension in a Waring blender. Purifi-cation was achieved by repeated isoelectric pre-cipitation and preparative ultracentifugation.Chemical analysis revealed that the purified K88antigen was a protein, in contrast to the polysac-charide nature of all other K antigens known atthat time. The morphology of the K88 antigenwas studied in the electron microscope (179,188). It was found that K88-positive bacteriawere covered with a material of filamentousappearance. The purified K88 antigen had thesame fimbria-like structure and was visualizedas thin flexible threads (Fig. 3). The diameter ofthe K88 fimbriae was estimated as 2.1 nm (188).With the use of specific antisera, 0rskov et al.

(134) could distinguish two variants of the K88antigen, K88ab and K88ac. The nomenclatureindicates that a common antigenic determinantis present on both variants. The K88ab variantwas originally found on E. coli strains isolatedfrom swine edema or pig enteritis in England.The K88ac variant was first described on E. colistrains isolated in Ireland and the German Dem-ocratic Republic (135). Additional minor serolo-gical variations of the K88ac antigen have beenobserved, but these have not been taken into thediagnostic scheme (134). Recently a new K88variant, termed K88ad, was described on por-cine enteropathogenic E. coli strains isolated inthe Netherlands (61). Similar minor variations asfound for the K88ac antigen were found for theK88ad antigen. The K88 variants could be dis-tinguished on the basis of their behavior indouble diffusion tests against their homologousantisera. All K88-positive strains had one deter-minant in common-K88a. Upon immunoelec-trophoresis in Noble agar (Difco) K88ab wascathodic, K88ac was cathodic or anodic depend-ing on the strain from which the antigen wasisolated, and the K88ad was not mobile oranodic. This immunoelectrophoretic behavior ofthe K88 variants did not alter upon transfer ofthe corresponding plasmid to E. coli K-12, indi-cating that this property was not host depen-dent. Anodic and cathodic K88ac antigens couldnot be distinguished serologically. The serologi-cal variations of the K88 antigen may representan effort of the bacterium to escape the immuno-logical pressure imposed upon the K88-positivebacterial population by large-scale vaccinationswith K88-containing vaccines. Another explana-tion for the observed serological K88 variantscould be that the occurrence of pigs "resistant"to a particular type of K88 adhesin, due to thepresence of altered "receptor sites" on theintestinal epithelial cells, has led to the selection

VOL. 46, 1982


mbr.asm

.org/D

ownloaded from



FIG. 3. Electron micrographs of purified adhesion preparations. (a) K88ab adhesin; (b) K99 adhesin; (c) F41adhesin; (d) CFA/I adhesin; (e) CFAIII adhesin; (f) 987P adhesin. The bar represents 64 nm. Photo kindlyprovided by J. van Breemen. CFA/I and CFA/II were a gift from P. Klemm; 987P was a gift from L. K. Nagy.

of other naturally occurring K88 variants thatare better adapted to these altered "receptors."This last possibility could very well explain theobservations of Bijlsma et al. (9) who observedspecific binding sites for the various serological

K88 variants. The idea that the serological vari-ants of the K88 antigen are adaptations to al-tered circumstances is supported by the fact thatthe adhesiveness was found to be associatedwith the K88c or K88b antigens and that this

MICROBIOL. REV.


mbr.asm

.org/D

ownloaded from



ability is only neutralizable by the homologousantisera (195). It is interesting that most K88-positive E. coli strains isolated from infectedpiglets nowadays have the antigen of the K88acor K88ad type (61, 195), and there seems to be aprevailing tendency towards the disappearanceof the K88ab antigen.Mooi and De Graaf (105) purified the K88

variants by gel filtration on a Sepharose CL-4Bcolumn after initial treatment of a cell suspen-sion in a Waring blender and precipitation of thecrude K88 preparation with ammonium sulfate.Analysis of purified K88 adhesins on sodiumdodecyl sulfate-polyacrylamide gels showed asingle protein band with an apparent molecularweight of 23,500 to 26,000 depending on the K88variant isolated. It is clear, therefore, that theK88 fimbriae consist of hundreds of identicalprotein subunits. Very little variation was ob-served in the amino acid composition of theserologically different K88 proteins (Table 5).The K88 preparations contain all the common

amino acids, with the exception of cysteine.This implies that the subunits in the K88 fimbri-ae are not held together by disulfide bridges butby hydrophobic or electrostatic interactions be-tween adjacent subunits. Carbohydrate analysisindicated the presence of trace amounts of sug-ars in the preparations of Mooi and De Graaf(105). Apparently, the K88 antigen is no glyco-protein. However, there are three regions in theamino acid sequence of the K88ab protein sub-unit where N-glycosidation could have oc-curred. The empirical rule of Eylar (52) that N-

glycosidation only occurs at asparagine residuesin Asn-X-Sertlhr sequences is fulfilled at aminoacid residue positions 7-9 (Asn-Gly-Ser), 127-129 (Asn-Ala-Ser), and 202-204 (Asn-Ile-Thr) inthe primary structure of the K88ab subunit pro-tein (Fig. 4). The N- and C-terminal amino acidsequence of the different serological K88 vari-ants has been determined by Gaastra et al. (56),together with the N-terminal amino acid se-quence of the cyanogen bromide fragments de-rived from the K88 antigens. No difference wasfound between the K88 variants in the first 23 N-terminal amino acids, nor in the 24 C-terminalamino acids. This probably indicates that thereis an evolutionary constraint on these parts ofthe K88 protein and implies a role of these partsof the protein structure in either the function ofthe fimbriae or in the subunit-subunit interactionwithin the fimbriae. The N-terminal sequence ofsome of the cyanogen bromide fragments wasdifferent for the different K88 antigens. Thenature of these differences and the positionwithin the molecule are discussed together withthe primary structure of the K88ab antigen.The primary structure of the K88ab antigen

was determined by two groups in different ways.Klemm (91) determined the amino acid sequenceof the K88ab protein subunit. Gaastra et al. (57)determined the base sequence of the deoxyribo-nucleic acid (DNA) of the gene encoding theK88ab protein. The amino acid sequence of theprotein is given in Fig. 4. It was found that theK88ab subunit is synthesized in a precursorform with an additional 21 amino acid residues at

TABLE 5. Amino acid composition of various adhesin subunitsNo. of residues

Amino acidK88ab K88ac K88ad 987P K99 F41 CFA/I CFAJII

Asx 30 34 30 34 23 27 12 14Tlhr 29 25 25 28 19 18 15 9Ser 18 16 19 23 15 33 17 6Glx 17 17 21 16 10 24 11 20Pro 5 8 9 8 5 11 7 3Gly 36 35 35 26 19 39 10 11Ala 28 25 26 26 18 16 19 8Cys 0 0 0 2 4 NDa 0 0Val 21 20 20 15 8 17 19 7Met 3 3 4 1 3 3 3 2Ile 13 13 13 12 9 10 5 4Leu 20 22 20 17 9 14 12 7Tyr 10 10 9 5 7 13 4 5Phe 11 9 10 3 9 9 2 4His 0 2 2 0 3 6 1 1Lys 11 10 10 10 9 12 8 5Arg 8 10 7 2 6 4 1 3Trp 4 ND ND 1 ND ND 1 NDHyLysb 0 0 0 0 0 10 0 0

a "ND, Not determined.b HyLys, Hydroxylysine.

VOL. 46, 1982


mbr.asm

.org/D

ownloaded from


140 GAASTRA AND DE GRAAF MICROBIOL. REV.

5' AAACGGAGCCGCGGGATGGTTTTACGGTAATTCCGGAAAAATAAGGGTTACCGATTTCAGTTTATT A T TT TOGA TAT C A A _ GTT A AT T T TT CGA A A A A GA CT CT GAT TG CA CT GG CA AT T G CT G CA-7*' -20 -15 -10K88ab fit Lya Lys Thr Leu Ile Ala Leu Ala Ile Ala Ala

TCTGCTGCATCTGGTATGGCACATGCCTGGATGACTGGTGATTTCAATGGTTCGGTCGATATCGG-5 1 5 10lWbb SW Ala Ala Sex Gly Met Ala Hio Ala Trp Met Thr Gly Lip Phe Ase Gly Ser Val Asp Ile Cly

KSld Trp Met Thr Oly isp Phe Aw Gly Ser Val Lip Ile GlyT 0G T A GTATCACT0CAGATGATTATCGTCAG AAA TG00AATGGAAAGTTGGTACAGGTCTTAATG15 20 25 30KSab Gly Ser Ile Thr Ala LAp Asp Tyr Arg Gln Lya Txp Glu Trp Lys Val Cly Thx Gly Leu Am

K88&d Gly Ser Ile Thr Ala Asp Asp Tyr

GATTTGGTAATGTATTGAATGACCTGACCAATGGTGGAACCAAACTGACCATTACTGTTACTGGT35 40 45 50 55K8ab Gly Ph. Gly Amn Val Lou Lin Lp Lou T'r Ln Gly Mly Thr Ly Lou Thr Ile Thr Vo Thr Gly

AATAAGCCAATTTTGTTGGGCCGAACCAAAGAAGCATTTGCTACGCCAGTAAGTGGTGCTGTAGA60 65 70 7518ab Aan Lys Pro Ile Leu Lou Gly Arg Thr Lys Glu Ala Phe Ala Thr Pro V1 Ser Gly Gly Val Asp

TGGAATTCCTCAGATTGCATTTACTGACTATGAAGGAGCTTCTGTAAAACTCAGAAACACTGATG80 85 90 95K8lb Gly Ile Pro Gln Ile Ala Phe Thr Asp Tyr Glu Gly Ala Ser Val Lys Leu Arg Am Thr ALp

K8bd

GTOAAACTAATAAAGGTTTAGCATATTTTGTTCTGCCGATGAAAAATGCAGAGAOCACTAAAGTT100 105 110 115 120K88ab Sry Glu Txr Amn Lya Gly Lou Ala Tyr Phe Val Lou Pro Met Lys ALi Ala Glu Gly Thr Lys ValiCB3ad Met Lys Asn Ala Glu Cly Thr Lys

GGTTCAGTOAAAOTG AATOCATCTTATGCCOGTGTGTTCCGOAAAGCTCCCCTTA CTTCTG CGCA125 130 135 140C8ab Gly Ser Val LYB Val Am Ala Ser Tyr Ala Gly Val Phe Cly Lys Gly Cly Val Th Ser Ala Asp

CG0GGAG CTTTTTCG CTTTTTOCGOACGG0 TTGCCCCCTATCTTTTATOGTOGTTTOACGACCA145 150 155 160813Sb Gly Glu Lu Phe Ser Leu Phe Ala Gly Leu Arg Ala Ile Phe Tyr Cly Gly Leu Thr ThrMet Ala Loeu Phe Ala ' Oly

CTOTTTCGCGCTCTGCACTCACGAOTGGGAGTGCCCGCACCGCGCGCACAGACTTGTTTCC AACT165 170 - 175 180 185K88ab Thr Val Ser Cly Ala Ala Leu Thr Se GOly Ser Ala Ala Ala Ala Axg Thr Clu Leu Phe Gly SerLeu Pro Ala Gly Ser Ala Ala Ala Ala Arg Thr Clu Leu Phe Gly Ser

C T A T C A A G A A A T G A T A T T C T C G G A C A G JA T T C A A A A G T A A A C G C A A A T A T T A C T T C T C T T G T T G A190 195 200 205KSBab Lou Ser Arg Am Lp Ile Lou Gly Gln Ile Gln Arg Val Asn Ala Ln Ile Thr Ser Leu Val LipK88od L Ser L AM Asp Ile Lou Gly Gln Ile Gln Arg Val Amn Ala Am I1l Thr Ser Leu Val

COICCCA,GTTCTTACAOGCAAGACATGGAGTACACTGATCCAACT.TTCTTTCTCCTCCCTP TO2, ir 215 220 225K88sb Vol Gly Ser Tyx Ax'g Gl met [ Tyr Thr Lp Cly m Val Val Sex Ala Tyr

K8 Val GOly Sexr [he n|Glu et Tyx Thr Lip Gly Val Val Sex Ala Tyr

CACTCGCTATTG CAAACCCTCACACTATTGAGGCAACTTTTAATC ACCCTCTA ACTACC AGCACT230 235 - 240 245 250K88abAla Lu Gly Ile Ala Li Gly Gln Thr Ile Glu Ala Thr Ph. Li Gln Ala Val Thr Txr Ser ThrK88ad Ali Lou Gly Ile Ala A ie Gln Thr Ile Glu Ala Thr Ph. Li Oln Ala Val Thr Thr Ser Thr

CAGTGGAGCCCTCCCCTGAACGTAGCAATAACTTATT A C T C A A A CT T G C T G G A255 260K8ab Gln Trp Ser Ala Pxo Leu ALi Val Aa Ile Thr Tyr Tyr

K88ad ln Trp Ser Ala Pro Leu Asn Val Ala Ile Thr Tyr Tyr

FIG. 4. The nucleotide sequence of the gene encoding the K88ab protein subunit. The amino acid sequence ofthe K88ab subunit derived from the nucleotide sequence is also given, together with the available data on theamino acid sequence of the K88ad protein subunit. Symbols: - ,, The ATG start codon of the K88ab structuralgene; ,-, the TGA stop codon of the K88ab structural gene. The nucleotides which are underlined encoderibonucleic acid bases which are complementary to the 3' end of 16S ribosomal ribonucleic acid of E. coli and apossible Pribnow box. Differences between the amino acid sequences of the K88ab and K88ad protein subunitsare enclosed in a box.


mbr.asm

.org/D

ownloaded from



the N-terminal part of the molecule. This signalpeptide is cleaved off during the transport of thesubunit from the ribosome to its place of destina-tion. It also became clear from the determinationof the base sequence of the gene encoding theK88ab antigen that there is probably a promoterpreceding this gene. The differences observed inthe N-terminal amino acid sequences of thecyanogen bromide fragments of the various K88antigens could be located around amino acidresidues 150 and 220. As compared to the pri-mary structure of the K88ab variant, 15 differ-ences in the partial amino acid sequence of theK88ad variant have been observed at the mo-ment (Fig. 4). These differences are spread alongthe polypeptide chain and are not confined toone region. Most differences involve chargedamino acid residues, and therefore most likelyoccur at the surface of the protein. It is not yetknown whether these differences in primarystructure also reflect the observed differences inimmunological behavior. No homology was ob-served between the primary structures of theK88 and CFA/I protein subunits and the partial-ly known sequences of other adhesion antigens(27, 68, 90, 92; Fig. 5). As can be seen in Fig. 4,there is a specific charge distribution along thepolypeptide chain of the K88 antigen. The C-terminal part of the molecule is rather hydropho-bic, and among the last 45 amino acid residues,there is only one charged residue. The centralpart of the molecule, on the contrary, is ratherhydrophilic. It is also this part of the moleculewhere most of the basic amino acid residues arefound.

The 987P AdhesinThe isolation and purification of the 987P

antigen by the method of Brinton (13) has beenreported (55). The diameter of 987P fimbriaeas observed in the electron microscope was 7 nm(Fig. 3), and the antigen appeared to be com-posed of protein subunits with an apparentmolecular weight of 18,900 (Table 6). Amore detailed report on the purification andpartial characterization of the 987P antigen of

TABLE 6. Some characteristics of adhesins ofenterotoxigenic E. coli strains

Diameter Mol wtAdhesin fimbnae of sub- pI

imbria unit

K88 2.1 27,540 4.2987P 7 20,000 3.7K99 4.8 18,500 9.7F41 3.2 29,500 4.6CFAII 3.2 15,058 4.8CFA/II 3.2 13,000

pig enterotoxigenic E. coli strains was recentlypublished by Isaacson and Richter (79). The987P antigen was removed from the cells byhomogenization and purified by repeated precip-itation with MgCI2. Chemical analysis showedthat the 987P adhesin is composed primarily ofprotein but contains also an unidentified aminosugar. Electrophoresis of the purified antigen onsodium dodecyl sulfate-polyacrylamide gelsshowed a single protein band with an apparentmolecular weight of about 20,000. In the elec-tron microscope, the 987P antigen appeared asrather rigid fimbriae with a diameter of 7 nm andan apparent axial hole. Morphologically the987P antigen could not be distinguished fromtype 1 fimbriae of E. coli. Both fimbriae have thesame N-terminal amino acid, alanine (90). Theisoelectric point of 987P protein is at pH 3.7.

The K99 AdhesinIn 1977, Isaacson (71) reported a procedure

for isolation and purification of the K99 antigenfrom an E. coli K-12 strain harboring the K99plasmid. The antigen was removed from cellsgrown in Trypticase soy broth (BL MicrobiologySystems) by salt extraction and homogenizationof a concentrated cell suspension in a SorvallOmnimixer. Subsequently the antigen was puri-fied by ammonium sulfate precipitation and ion-exchange chromatography on diethylamin-oethyl-Sephadex. The purified material wascomposed of two protein subunits: a major com-

1 5 10 1 5 20Type 1 Ala-Ala-Thr-Thr-Val-Asn-Gly-Gly-Thr-Val-His--Phe-Lys-Gly-Glu-Val-Val-Asn-Ala-Ala-

K88 Trp-Met-Thr-Gly-Asp-Phe-Asn-Gly-Ser-Val-Asp-I le-Gl y-Gly-Ser-I le-Thr-Ala-Asp-Asp-Tyr -Arg-

K99 Asn-Thr-Gly-Thr-Ile-Asn-Phe-Asn-Gly-Lys-Ile-Thr-Ser-Ala-Thr-Cys-Thr-Ile-Glu-Pro-Glu-Val-

CFA/I Val-Glu-Lys-Asn-Ile-Thr-Val-Thr-Ala-Ser-Val-Asp-Pro-Val-Ile-Asp-Leu-Leu-Gln-Ala-Asp-Gly-

F41 Ala-Asp-Trp-Thr-Glu-Gly-Gln-Pro-Gly-Asp-Ile-Leu-Ile-Gly-Gly-Glu-Ile-Thr- X -Pro-Ser-Val-

FIG. 5. The amino acid sequence of the N-terminal part of the K88 (56), K99 (27), CFAJI (90), and F41 (54)adhesin subunits. The amino acid sequence of the N-terminal part of the subunit of type 1 fimbriae (68) is givenfor comparison.

VoL. 46, 1982


mbr.asm

.org/D

ownloaded from



ponent (molecular weight, 22,500) and a minorcomponent (molecular weight, 29,500). The ratiobetween both constituents was 5:1. When puri-fied K99 was negatively stained and examinedunder the electron microscope, it appeared topossess a rod-like filamentous structure with astrong tendency to aggregate. Individual rodshad an average diameter of 8.4 nm and meanlength of 130 nm. The isoelectric point of theantigen was 10.1. In contrast to the K99-positivebacteria, purified K99 no longer hemagglutinat-ed guinea pig erythrocytes in the presence of D-mannose. The hemagglutinating activity towardsthese erythrocytes was separated from the K99antigen during purification.

Morris et al. (114) published another isolationprocedure for the K99 antigen based on a meth-od for isolation of the K88 antigen as describedby Stirm et al. (180). E. coli B41 (O101:K99:H-)was grown on blood agar, and the K99 wasextracted from the cells by heating and shakingthe bacterial suspension for 30 mmn at 60°C.Subsequently a partial purification of the antigenwas obtained by repeated isoelectric precipita-tion at pH 4.9, gel filtration of the precipitatedmaterial on Sepharose 4B, and finally ion-ex-change chromatography on quaternary amino-ethyl Sephadex. During this procedure the abili-ty to hemagglutinate sheep erythrocytes was notseparated from the K99-positive material. Ab-sorption of the antigen with monospecific antise-rum to K99 abolished the hemagglutination, andthe authors therefore identified the K99 antigenas a mannose-resistant hemagglutin. In anotherpaper, Morris et al. (116) reported that theisoelectric point of the K99 antigen isolated fromE. coli B41 was 4.2. Furthermore, the isolatedK99 antigen showed hemagglutination of guineapig as well as sheep erythrocytes. These resultsare in agreement with a study of Burrows et al.(17) who demonstrated that the determinants forK99 and hemagglutinin were located on thesame plasmid and that culturing of K99-positivebacteria at 18°C suppresses both ofthese charac-ters.

In an attempt to explain the anomalies in thecharacterization ofthe K99 antigen, Morris et al.(115) examined K99 antigen isolated from E. coliB41 and precipitated with ammonium sulfate byimmunoelectrophoresis with OK sera preparedfrom a series of K99-positive strains belongingto 0 serogroups 9 and 101. They observed thatthe isolated material contained an anionic aswell as a cationic component which could beseparated by ion-exchange chromatography.The anionic component exhibited a strong hem-agglutination reaction with sheep erythrocytes,whereas the cationic component showed only aweak hemagglutination reaction which was lostupon storage. These results suggested that the

K99 isolated from E. coli B41 is composed oftwo antigenically distinct components. Isaacson(72) confirmed this conclusion but demonstratedthat the discrepancy in the presence of only thecationic component or the presence of both theanionic and the cationic component was due todifferences in the sera used to detect K99 as wellas the strain from which the antigen was isolat-ed. K99 isolated from E. coli K-12 (K99) and E.coli B41 (0101:K99:H-) yielded a single precipi-tation line with antisera prepared against E. coliK-12 (K99) and E. coli (064:K99). Using anti-sera prepared against E. coli of serotypeO101:K30:K99, the same precipitation line wasobserved, but an additional line was producedwith K99 isolated from E. coli B41. Apparentlythe K99 isolated from this strain contained anadditional (anionic) component. K99 isolatedfrom E. coli K-12 hemagglutinated sheep eryth-rocytes in the cold and horse erythrocytes,whereas K99 isolated from E. coli B41 hemag-glutinated guinea pig, sheep, and horse erythro-cytes with no temperature effect. Inclusion ofstandard absorbed anti-K99 serum in the hemag-glutination reaction inhibits only the horseerythrocyte reaction. Recently Morris et al.(118) reported that antibodies to the cationiccomponent of a K99 antigen extract preparedfrom E. coli B41 were present in antisera to allthe K99-positive E. coli strains in serogroups 8, 9,20, 64, and 101, but antibodies to the anioniccomponent were only detected in antisera to E.coli strains from the 09 and 0101 serogroups.The anionic component isolated from the K99preparation hemagglutinated sheep, guinea pig,and horse erythrocytes, whereas the cationiccomponents only showed hemagglutination ofhorse erythrocytes. In conclusion, the K99 anti-gen was definitively identified as a cationichemagglutin. De Graaf et al. (27) recently puri-fied the K99 antigen from different E. coli strainsby salt extraction, gel filtration, and treatmentwith deoxycholate. The purified preparation washomogeneous upon sodium dodecyl sulfate-polyacrylamide electrophoresis and appeared tobe composed of identical protein subunits with amolecular weight of 18,400. Amino acid analysisshowed a preponderance of amino acids withapolar side chains as well as the presence ofcysteine. The isoelectric point of the purifiedmaterial was established as 9.75 (Table 6). Re-cently, Isaacson et al. (74) confirmed that puri-fied K99 antigen is composed of only a singlesubunit. Analysis of the N-terminal amino acidsequence showed that the K99 antigen bears noresemblance to the N-terminal primary structureof the K88 and CFA/I adhesins (Fig. 5). Whenobserved in the electron microscope, K99 ap-peared to be a fimbria with a helical structureand a diameter of 4.8 nm.

MICROBIOL. REV.


mbr.asm

.org/D

ownloaded from



As mentioned before, when studying adhesinsone should take into account the possibility thatone strain possesses more than one type ofadhesin. The confficting data on the identifica-tion of the K99 antigen clearly demonstrate thisissue.

The F41 AdhesinThe purification and characterization of the

F41 antigen has been described by De Graaf andRoorda (29). F41 was characterized as a filamen-tous structure composed of protein subunitswith a molecular weight of 29,500. The F41adhesion antigen has a diameter of 3.2 nm,which is slightly thinner than the K99 adhesionantigen (Table 6). The N-terminal amino acidsequence of the F41 protein subunit showed alimited homology with the K99 protein subunit(Fig. 5).

The CFA/I AdhesinSeveral isolation procedures for the CFA/I

protein have been described. Initially the CFA/Iwas isolated by shearing the cells followed bydifferential ultrafiltration of the supernatant, iso-electric precipitation, and ultracentrifugationinto sucrose (49). Later the isoelectric precipita-tion and ultracentrifugation steps were replacedby ammonium sulfate precipitation and ion-ex-change chromatography (44). The purified anti-gen could be seen in the electron microscope asthin threads 7 nm in diameter and had an averagemolecular weight of 1.6 x 106 as determined by

sedimentation equilibrium centrifugation. Theapparent molecular weight of the CFA/I subunitwas 23,800 as determined by sodium dodecylsulfate-polyacrylamide gel electrophoresis (44).Because of the work of other authors, however,these data have become questionable.Klemm (90) isolated the CFA/I adhesin by

heating the bacteria for 20 min at 60°C followedby shearing. This method has the disadvantagethat prolonged heating of the CFA/I destroys itsantigenicity, i.e., heat treatment of cells pre-vents agglutination of cells by anti-CFA serum.The CFA/I purified in the way described byEvans et al. (44) retained its morphology, anti-genicity, and biological function. Purification ofthe CFA/I was obtained by Klemm (90) by gelfiltration on Sepharose columns. Contrary to thedata reported by Evans et al. (44), the molecularweight of the CFA/I subunits was found to be14,500 (90), with valine as the N-terminal aminoacid. Wevers et al. (192) reported molecularweights of 12,000 and 13,000 for the proteinsubunits of CFAII and CFAJII, respectively.These authors suggested that the value for themolecular weight of the CFA/I protein subunitas reported by Evans et al. (44) is probably adimer because a faint band corresponding to amolecular weight of 24,000 was detectable ontheir gels. The amino acid composition of theCFA/I protein subunit is given in Table 5. Thecomplete amino acid sequence of the CFAIIprotein subunit (Fig. 6) was recently determinedby Klemm (92). It shows no homology with theamino acid sequence of the K88ab protein sub-

1 5 10 15 20 25VAL-GLU-LYS-ASN- ILE-TKR-VAL-TH*-ALA-SER-VAL-ASP-PRO-VAL- ILE-ASP-LEU-LE- ..- A-As6P'-&s LA-LEU-

26 30 35 40 45 50PRO-SER-ALA-VAL-Lys-LEu-ALA-TYR-SER- PRO-ALA-SER-Lys-TmR-PHE -GLU-SER-Tya-AR6-VAL1T-Tt-&.P-IAL-H I S-

51 55 60 65 70 75ThR-ASN-ASx-ALA-THR-LYS-LYS-VAL-ILE-VAL-LYS-LEU-ALA-ASP-THR-PRO-kN-LEU-Tmt-ASP-VAL-LE-ASw-ALATmt-

76 80 85 90 95 100VAL-GL,e-MET-PRO- ILE-SER-VAL-SER-TRP_GLY-GLY 6LN VAL-LEU-SER-Tme-Tmn _ALA- ys_GL 6LUA&AA-

101 105 110 115 120 125LEU-GLY-TYR-SER-ALA-SER-GLY-VAL-AS"-GLY-VAL-SER-SER-SER-GLN-GLU-LEU-VAL- IL-SEv-.A-A.A-tN*-Lys-Tma-

126 1w30 135 140 145ALA-GL Y-T"I-ALA-PRo-ThR-ALA-GLY-ASN-TYR-SER-GLYYAL-VAL-SER-LEu-VA.-MlET-THRA-LEu-LY-SER--GN.

FIG. 6. The amino acid sequence of the CFA/I protein subunit (92).

VOL. 46, 1982


mbr.asm

.org/D

ownloaded from



unit nor with the partial amino acid sequences ofother adhesin subunits (Fig. 5). The molecularweight of the subunit as calculated from thecomplete amino acid sequence is 15,058 (92).The CFAII subunit is rather hydrophobic andlike the K88 subunit contains no cysteine. Theisoelectric point of the CFA/I antigen is at pH4.8 (54). Isoelectric points at low pH values havealso been determined for the type 1 fimbriae andK88 antigen of E. coli (13, 105). The pl valuesfor these antigens are 3.92 and 4.2, respectively,in contrast to the K99 antigen which has a plvalue of 9.75 (27). Electron micrographs of aCFA/I preparation isolated by shearing showflexible fimbria-like structures with a diameterof approximately 3 nm and resembling the K88and K99 fimbriae (Fig. 3). A similar observationhas been published by Brinton (15) who studieda clone of E. coli H10407 that produced amannose-resistant but not the mannose-sensi-tive hemagglutinin. We have observed that, de-pending on the concentration, the CFA/I fimbri-ae tend to aggregate and form bundles with thediameter reported by Evans et al. (44). Thisconcentration dependency was also observed byBrinton (personal communication). The pres-ence of very thin fimbriae was also reported byEvans and Evans (42) on a CFA/II-positivestrain.

Other Adhesins of Human Enterotoxigenc E.col

In a survey of enterotoxigenic E. coli strainsisolated from humans, Thorne et al. (182) stud-ied the adhesive properties of these strains byusing human buccal mucosal epithelial cells. Of32 enterotoxigenic strains examined, 52% boundto the buccal cells. No correlation, however,was found between adhesion to buccal cells andmannose-resistant hemagglutination of humanand guinea pig erythrocytes. One group of ad-herent strains were able to agglutinate humanand/or guinea pig erythrocytes, but anothergroup of adherent strains showed no reactionwith these types of erythrocytes. In a furtherstudy, Deneke et al. (32) described the presenceoffimbriae on enterotoxigenic E. coli strains thatadhered to buccal cells and exhibited mannose-resistant hemagglutination of human group Aand B and/or guinea pig erythrocytes. For purifi-cation, the fimbriae were detached from the cellsby blending, absorbed onto guinea pig erythro-cytes at 0°C, and subsequently eluted from theerythrocytes at 37, 42, and 50°C. The fimbriae ofstrain 334 appeared to be composed of twoprotein subunits with molecular weights of13,100 and 12,500, respectively. Purified fimbri-ae bound to human buccal cells as did the wholecells. The binding could be inhibited by anti-fimbriae Fab fragments prepared from antiserum

against the purified adhesin. The fimbriae exhib-it a hemagglutination pattern that differed fromthe CFA/I and CFA/II adhesins. In a morerecent paper, Deneke et al. (33) described sero-logical studies of these fimbriae. Three distinctantigenic types were identified on the enterotox-igenic E. coli strains that adhered to buccal cells.Antisera against these three adhesin types react-ed with 60 of 106 enterotoxigenic E. coli strainstested. These strains all produced LT and/or ST.Only a limited cross-reactivity was observedwith the three types offimbriae and their respec-tive antisera, although each of the three types offimbriae is composed of protein subunits withmolecular weights of 13,100 and 12,500. Therewas no correlation between fimbrial serotype,enterotoxin production, 0-antigen type, geo-graphical source of isolation, and mannose-re-sistant hemagglutination of various types oferythrocytes.

It should be noted that E. coli strain H10407,which has been used for the isolation and purifi-cation of the CFA/I adhesin by Evans et al. (44)and by Klemm (90), and the CFA-negativeH10407-P derivative strain both reacted withantiserum against the fimbriae isolated by Den-eke et al. Furthermore, both strains adhered tobuccal cells, but strain H10407-P had lost theability to hemagglutinate human erythrocytes.Because it cannot be excluded that the antiseraprepared by Deneke et al. contain antibodiesagainst type 1 fimbriae, the observed cross-reactivity may be attributed to the presence oftype 1 fimbriae on E. coli H10407 as reported byBrinton (15).An in vitro binding assay for human entero-

pathogenic E. coli strains to human fetal smallintestine was developed by McNeish and co-workers (101). With a strain of serotype026:K60:H11 originally isolated from a babywith diarrhea, the adhesion to human intestinaltissue was shown to be mannose resistant, hostspecific, and plasmid mediated (193). The genet-ic determinant for this mucosal adherence isencoded by a transmissible plasmid, pLG11, 56megadaltons (Md) in size, that also carried thegenes encoding for the production of the bacte-riocin, colicin lb (194). The nature of this adher-ence factor is not known, but it appears to bedistinct from the CFA/I or CFA/II adhesins.A strain of E. coli serotype 018ac:H-, isolat-

ed from the stool of a patient with diarrhea, wasfound to possess fimbriae that were able toagglutinate human group A erythrocytes but notbovine, chicken, or guinea pig erythrocytes(192). The fimbriae were not produced at 20°C.Assays for ST and LT were negative, but thestrain produced hemolysin and colicin, which ischaracteristic of E. coli strains causing extrain-testinal infections (104). The fimbriae were de-

MICROBIOL. REV.


mbr.asm

.org/D

ownloaded from



tached from the cells by shearing and purified byisoelectric precipitation and gradient centrifuga-tion in cesium chloride. The fimbriae had adiameter of 5 nm and variable length. Also thesefimbriae were antigenically different from theCFAII and CFA/II adhesins. The molecularweight of the protein subunits as determined bysodium dodecyl sulfate-polyacrylamide gel elec-trophoresis was 21,000. Chemical analysisshowed the presence ofone phosphate group persubunit. The amino acid composition of thissubunit differed from those reported for CFA/I,K88, and K99, as well as from type 1 fimbriae.The fimbriae contained two cysteine residuesper subunit, but as in the case of the K99antigen, no disulfide bridges between subunitswere present. The isolated fimbriae were shownto retain their filamentous structure in the elec-tron microscope. Isoelectric focusing revealedthe presence of two protein bands having pIvalues of 5.1 and 5.6, respectively. It should bementioned that Cravioto et al. (22) found that 23out of 44 E. coli Ol8ac strains all exhibitedmannose-resistant hemagglutination of humanerythrocytes. All strains, however, were isolat-ed from extraintestinal sources. In addition,Wevers et al. (192) reported that antiserumagainst purified fimbriae of their E. coli018ac:H- strain also agglutinated E. coli strainsof different serotypes obtained from urinarytract infections. This finding raises the questionas to whether these fimbriae are also importantfor colonization of the intestine.

ADHESION

Adhesion to Intestinal Epithelial CellsAlthough there is no doubt that the adhesins

of enterotoxigenic E. coli strains confer adhe-sive properties to these bacteria, little is knownabout the nature of the mechanism of adhesion.General concepts and principles of the adhesionof bacteria to eucaryotic cells have been re-viewed (80, 128). Bacterial cell surface charac-teristics such as charge and hydrophobic proper-

ties are supposed to be involved in attachment.Physicochemical aspects of the adhesion ofmicroorganisms to surfaces have recently beenreviewed by Rutter and Vincent (147). The sur-

face of both procaryotic and eucaryotic cells is,in sum, negatively charged. The repulsive elec-trostatic force between these like-charged cellsurfaces can be overcome by long-range andshort-range attractive forces. Specific binding offimbriae to receptors on the surface of theepithelial cells can overcome the repulsiveforces between bacteria and epithelial cells. Therole of short-range hydrophobic interactions inthe adhesive properties of fimbriated enterotoxi-

genic strains has been demonstrated by hydro-phobic interaction chromatography (169, 187,188). Smyth et al. (169) suggested that reductionin the cell surface potential by the masking of thecharge contribution of polysaccharide K anti-gens and lipopolysaccharide by adhesins withhydrophobic characteristics probably promotesadhesion and that hydrophobic bonds may beinvolved in the interaction of these adhesinswith the intestinal mucosa. Fimbriae might havehydrophobic patches, as a consequence of a highcontent of apolar amino acids.Adherence to and colonization of the epitheli-

al cell surfaces of the small intestine, withoutactual tissue invasion, followed by proliferationto a large population is consistent with clinicaland experimental observations of diarrhea inanimals and humans, caused by enterotoxigenicE. coli. The role of the K88 adhesion antigen inpromoting colonization of the intestine was stud-ied by implanting the plasmid encoding the K88antigen and a plasmid encoding the productionof enterotoxin into nonpathogenic strains of E.coli or alternatively by removal of these plas-mids from pathogenic strains and subsequentlyfeeding such modified strains to the neonate (81,164, 165). The implantation of the K88 andenterotoxin plasmids into a nonpathogenic E.coli strain resulted in the colonization of thesmall intestine and the development of diarrhea.K88-negative E. coli derived from K88-positiveenteropathogenic E. coli failed to proliferate inthe anterior small intestine and did not causediarrhea in newborn pigs. Introduction of a K88plasmid from another E. coli strain restored theability to produce diarrhea. It should be notedthat transfer of Ent and K88 plasmids to an E.coli K-12 strain did not convert this strain toenteropathogenicity, which emphasizes thatpossession of an adhesin and production of anenterotoxin is as such not sufficient to producediarrhea (164).The K88 antigen enables the organisms to

proliferate in the anterior small intestine (Fig. 7),where normally very few bacteria are present,instead of being carried along with the normalmovement of the chyme. It is this area of theintestine that is most sensitive to enterotoxicactivity, and the gut becomes progressively lessresponsive after the first few feet (69, 162).Bertschinger et al. (8) compared two enterotoxi-genic strains and a non-enterotoxigenic E. colistrain for their association with the porcine smallintestinal epithelium. The enterotoxigenicstrains were found from tip to base of the villiand contiguous to the brush border. These bac-teria were not found in the crypts. In contrast,the non-enterotoxigenic strain characteristicallyremained in the gut lumen near the tips of thevilli. Comparable results have been described by

VOL . 46, 1982


mbr.asm

.org/D

ownloaded from



FIG. 7. Scanning electron micrograph of E. coli 0141:K85ab:K88ab on villi of an ileum sample 18 h afterinfecting a newborn colostrum-deprived germfree piglet. Magnification: x3,900. Photo generously provided byJ. A. Morris.

Arbuckle (4). Adhesion of K88-positive bacteriato tissue from the small intestine in vitro couldbe inhibited by K88 antibodies. Adhesion of cell-free K88 antigen was also demonstrated. Inseveral other reports the in vitro adhesion ofK88-positive strains to either isolated intestinalepithelial cells (195) or epithelial cell brush bor-ders (83) has been described. The importance ofthe K88 antigen and the role of adhesion in thediarrheal disease caused by enteropathogenic E.coli in piglets is shown clearly when the numbersof organisms present in the small intestinal con-tents and on the tissues in piglets from sowsvaccinated with a K88-positive or a K88-nega-tive vaccine are compared. The latter vaccinecontains only 0 antigens and stimulates theproduction of 0-specific antibodies, whereas theK88-positive vaccine stimulates the productionof both 0 and K88 antibodies. Compared tononvaccinated piglets challenged with 1010 K88-positive organisms, the K88-negative vaccinecaused a fivefold drop in the amount of organ-isms present in the lumen of the small intestineand a threefold drop in the number of organismspresent on the small intestinal wall. With the

K88-positive vaccine, the drop in the number oforganisms was 10- and 100-fold, respectively(189).Moon and Whipp (111) reported the develop-

ment of resistance with age by the swine intes-tine to some enteropathogenic E. coli strains,whereas the response to other enteropathogenicstrains was positive at all ages. Walker and Nagy(189) challenged piglets with 1010 K88-positivecells at various time intervals after birth andobserved a gradual decrease in the number ofadherent organisms over a period of 5 weeks.The adhesiveness of a pathogen for its particu-

lar host is highly specific. Enterotoxigenic E.coli strains isolated from pigs fail to producediarrhea in calves and lambs by oral inoculationdue to their inability to proliferate in the anteriorsmall intestine of calves and lambs (162), where-as all enterotoxigenic strains isolated from diar-rheic calves or lambs produced a severe diarrheawhen administered orally to colostrum-fed neo-natal calves and lambs younger than 20 h. Thestrains had no effect on 3-day-old calves orlambs whether or not fed with colostrum.

Several porcine enterotoxigenic E. coli strains

MICROBIOL. REV.


mbr.asm

.org/D

ownloaded from



have been isolated that do not possess the K88antigen. These strains colonize the pig intestine,and the presence of adhesive surface compo-nents was suggested (8, 69). Nagy et al. (123)reported that a number of K88-negative porcineenterotoxigenic E. coli strains all belonging toserogroups 09, 020, and 0101 were able tocause severe diarrhea when inoculated into new-born pigs. The strains colonize the ileum andadhere to the ileal epithelium of the inoculatedpiglets. The presence offimbriae on these strainswas subsequently described (78). Strains ofserogroup 0101 possessed fimbriae with a diam-eter identical to the K99 antigen, but strains ofserogroups 09 and 020, including strain 987,produced other fimbriae with a diameter of 7nm. The number of fimbriated cells per culturevaried from 1 to 100%o. However, when inoculat-ed into ligated intestinal loops, all of the cellsbecame heavily fimbriated. Antiserum againstthe fimbriae of strain 987 agglutinated all of thestrains of serogroups 09 and 020 that producedfimbriae morphologically identical to those pro-duced by strain 987. However, in contrast toK88-positive strains, the 987P-positive stockcultures were unable to adhere to isolated intes-tinal epithelial cells in vitro (124). The differ-ences between in vivo and in vitro adhesioncould be explained by a difference in the produc-tion of fimbriae. Growth in the pig intestine invivo appeared to select or promote the develop-ment of fimbriated cells. Fimbriated coloniesisolated from the intestine can be distinguishedfrom nonfimbriated colonies by colonial mor-phology on blood agar plates (124). Similar ob-servations on colonial differences for fimbriatedand nonfimbriated clones of E. coli possessingtype 1 fimbriae were made by Brinton (12).With cells grown under conditions that promoteadhesin production and selected fimbriatedclones, Isaacson et al. (76) were able to demon-strate that 987P-positive cells adhere to porcineintestinal epithelial cells in vitro. Purified 987Pfimbriae as well as Fab fragments specific for987P fimbriae inhibited this adhesion. The at-tachment of fimbriated cells to the epithelialcells was a saturable process with a maximum of30 to 40 bacteria per epithelial cell. Smith andHuggins (164) have confirmed the results ofMoon and co-workers (78, 124) in the establish-ment of the 987P adhesin as a colonization factorin pigs. They observed that, in contrast to K88-positive bacteria which proliferated to a muchgreater degree in the anterior small intestine,987P-positive strains are found in larger num-bers in the posterior small intestine.The importance of the K99 antigen in facilitat-

ing adhesion of K99-positive cells to calf andlamb intestinal epithelial cells in experimentalinfections was described by Smith and co-work-

ers (162, 166). K99-positive strains of someserotypes are equally able to proliferate in thesmall intestine of pigs and cause a profusediarrhea (164). Infection experiments withstrains of different serotype indicated that thepolysaccharide K antigen in a K99-positivestrain and, perhaps to a lesser extent, the 0antigen were important in determining whetheror not a K99-positive strain would colonize thesmall intestine of pigs. Moon and co-workers(110) reported the isolation of several entero-toxigenic E. coli strains from pigs that werefound to possess the K99 antigen. The highestnumber of K99-positive bacteria is always foundin the posterior small intestine, and the concen-tration of bacteria decreased progressively to-wards the anterior small intestine (77, 164).

K99-positive bacteria adhere to porcine intes-tinal epithelial cells in vitro with a maximum of30 to 40 bacteria per epithelial cell (76).The ability of K99-positive enterotoxigenic E.

coli strains to produce diarrhea appears to beage dependent. Calves and lambs become resis-tant to experimental challenge after about 2 daysof age (162), and the isolation of K99-positivestrains h4s not been reported from diarrhea inpigs after weaning. In contrast, K88-positivestrains have been isolated from diseased pigsduring both the neonatal and postweaning period(174). Suckling mice are susceptible to K88-positive and K99-positive enterotoxigenic E.coli strains, but because pathogenic strains rap-idly lose their K88- and K99-encoding plasmidsduring growth in adult mice, the older mice areresistant (89). Runnels et al. (142) studied theadhesion of K99-positive strains to isolated in-testinal epithelial cells taken from neonatal andolder pigs, calves, and mice. It appeared that theepithelium in the small intestine develops resist-ance to K99-mediated adhesion with increasinghost age. For the K88-positive cells, no suchage-dependent adhesion could be detected.

Morris et al. (117) studied the adhesive prop-erties of a calf enterotoxigenic E. coli straincarrying the recently discovered F41 adhesin.Germfree piglets infected with the E. coli B41Mdeveloped diarrhea within 16 h. The bacterialcounts of scrapings from the washed ileal muco-sa were greater than 108 bacteria per g, andscanning electron microscopy showed plaquesof adherent bacteria in the mucosal folds of thevilli. Indirect immunofluorescent staining dem-onstrated the presence of the F41 antigen invivo. Isolated F41 antigen competitively inhibit-ed the attachment of strain B41M to calf intesti-nal brush borders.

Volunteer studies and experiments in rabbitshave confirmed the hypothesis that the CFA/Iplays a significant role as a virulence factor inhuman diarrhea caused by enterotoxigenic E.

VOL. 46, 1982


mbr.asm

.org/D

ownloaded from



coli (31, 50). Volunteers were challenged orallywith 108 organisms of a CFAII-positive strain orits CFA/I-negative derivative. Diarrhea oc-curred only in those volunteers ingesting CFA/I-positive E. coli accompanied by the appearanceof the organism in the stool for more than 7 daysafter the beginning of the experiment, whereasthose ingesting the CFA/I-negative strain shedthe organism in the stool for a maximum of 3days (50).A rather unusual type of virulent E. coli strain

which causes diarrhea in young rabbits wasinitially described by Cantey and Blake (18).This strain, designated as RDEC-1, colonizesthe ileum, cecum, and colon, is noninvasive, anddoes not synthesize either LT or ST. E. coliRDEC-1 binds to rabbit brush borders in vitro(19). This adherence was not influenced by thepresence of mannose. In vivo infectivity of E.coli RDEC-1 was host specific. The strain heavi-ly colonized the ileum and cecum of rabbits (109cells per g of tissue), but minimal colonizationwas observed in guinea pigs and rats (20). Theexistence of fimbriae responsible for the man-nose-resistant adherence to rabbit brush bordershas not been described by Cantey or Cheney.The strain, however, does possess type 1-relatedfimbriae (55).

Nature of the Receptor Sites for AdhesinsKearns and Gibbons (87) reported that plasma

membranes prepared from brush borders sus-ceptible to adhesion by K88-positive bacteriahad lost 78% of their receptor activity. The K88binding ability of the brush-border membraneswas enhanced when the membranes were incu-bated with the supernatant fractions obtainedduring the preparation of these membranes. Acomparable result was obtained when brush-border membranes isolated from a K88-resistantpig were incubated with the supernatant fractionof K88-susceptible brush borders. No increasein binding was observed on incubating brush-border membranes from either adhesive or non-

TABLE 7. Carbohydrate chains in the receptors forthe various adhesins

Adhesin Carbohydrate0 Reference

K88 B-D-Gal or Glc, Gal, and Fuc 58, 87, 152or GM1?

K99 Ga1NacP(1-4)Ga$(1-4)GlcCer 53

2aNeuAcCFA/I GalNac3(1-4)Ga$(1-4)GlcCer 53

2aNeuAc

aGal, Galactose; Glc, glucose; Fuc, fucose; GalNac,N-acety4alactosamine; Cer, ceramide.

adhesive brush borders with the supernatantfraction obtained from nonadhesive brush bor-ders. Analysis of the supematant fractions fromadhesive and nonadhesive brush borders re-vealed a consistent difference in the glycolipidprofiles of these fractions. Since this was theonly chemical difference to be demonstratedbetween the different brush-border membranesand in view of the correspondence betweenchemical and genotypic differences, the differ-ence in glycolipid profile was supposed to be anindication for the possible nature of the receptorfor the K88 antigen. Attempts to inhibit thebinding of radioactively labeled K88 antigen tobrush borders with several mono-, di-, and oligo-saccharides (152) suggested that in the binding ofthe adhesin to brush borders, complex interac-tions are involved and that galactosyl residuesmay be important (Table 7). Inhibition of theability of the soluble K88 antigen to agglutinateguinea pig erythrocytes by some mucous glyco-proteins suggested that a terminal ,-D-galactosylresidue could be involved in the interaction ofK88 with the erythrocytes and therefore thatsuch a residue might be present in the receptorfor the K88 antigen (58). A different specificitywas reported by Anderson et al. (2) who used anin vitro binding assay employing differentialfiltration to show that purified 125I-labeled K88antigen formed complexes with isolated porcineintestinal brush-border membranes. The forma-tion of these complexes was inhibited by glyco-proteins with terminal N-acetylglucosamine andN-acetylgalactosamine residues and to a lesserextent by free N-acetylhexosamines. The differ-ence in conclusion may result from the differ-ence in the assay system. Gibbons et al. (58)studied the adhesion of K88 to guinea pig eryth-rocytes whereas Anderson et al. (2) used a directbinding assay with brush-border membranes.Sellwood (152) also tested several glycosidasesto see whether they could destroy the receptor.None of them inhibited adhesion of iodated K88preparations to adhesive brush borders. TritonX-100 extracts of adhesive brush borders werefound to inhibit an ELISA developed for theK88 antigen (F. R. Mooi and A. Olthof, unpub-lished results). It was therefore concluded thatextraction with Triton X-100 solubilized a sub-stance that could be the receptor for the K88antigen. The chemical nature of this substance,however, is not yet known.A possible involvement of glycolipids in the

binding of enterotoxigenic E. coli strains to theintestinal epithelium has become very likelysince it has been demonstrated that glycolipidsact as receptor for uropathogenic E. coli strainson human erythrocytes and uroepithelial cells(85, 93).As indicated above, K88-positive bacteria do

MICROBIOL. REV.


mbr.asm

.org/D

ownloaded from



TABLE 8. Adhesion of K88ab-, K88ac-, or K88ad-positive E. coli strains to isolated pig intestinal brush

borders

Brush Adhesion by K88-positive cellsborder

phenotype K8&ib K88ac K88ad

A + + +B + + _C + - +D - - +E - _ _

not adhere to the brush borders from all piglets.Two pig phenotypes were demonstrated anddesignated "adhesive" (signifying that K88-pos-itive bacteria attached to their brush borders) or"non-adhesive" (K88-positive bacteria do notattach to their brush borders) (144, 154). Thetwo phenotypes were found to be the productsof two alleles at a single locus and are inheritedin a simple Mendelian way (154). In a recentstudy that involved all three K88 variants knowntoday, at least five different phenotypes could bedistinguished in the brush-border adhesive test(9) (Table 8). One phenotype is susceptible to allthree variants, three phenotypes are susceptibleto only one or two variants, and one phenotypeis entirely resistant to K88-mediated adhesion.Experiments in which the adhesion of cellscarrying one of the three immunological K88variants to the different brush-border pheno-types was inhibited with the various purifiedK88 adhesins indicate that the receptor sites forthe K88ab and K88ac adhesins have a closersimilarity than the receptor site for the K88adadhesin. This observation fits very well with thefact that there is much more homology betweenthe primary structures of the K88ab and theK88ac adhesins than there is between K88aband K88ad. The chemical basis for the differentreceptor sites is not known.The receptor for CFA/I has not yet been

studied in human intestinal brush-border prepa-rations. Faris et al. (53) studied the erythrocytereceptor that is responsible for the mannose-resistant hemagglutination of CFA/I- and K99-positive E. coli strains. They showed that thehemagglutination of these strains could be inhib-ited by mono- and disialogangliosides, especiallythe GM2 at concentrations above the criticalmicellar concentration (Table 7). Pretreatmentof erythrocytes with neuraminidase inhibited thehemagglutination of human erythrocytes by theCFA/I-positive strains but not the hemagglutina-tion of sheep erythrocytes by K99-positivestrains. No influence of treatment of erythro-cytes with trypsin was observed, whereas pro-nase treatment resulted in negative hemaggluti-nation. These results suggest that both

glycolipids and glycoproteins may be involved inthe erythrocyte receptor for these two adhesins.

IMMUNITYAs adhesion of enteropathogenic E. coli to the

intestinal epithelium of the infected animal is anessential prerequisite for the development ofenteric disease, the antibody-mediated inhibi-tion of bacterial adhesion will be one of thepotential defense mechanisms of the host againstthe pathogen. In a young animal, such antibodywill be derived either from maternal sourcesthrough the colostrum and milk or later by localsynthesis via the intestinal secretory immunesystem. Recently Sellwood (153) showed thatcolostrum from susceptible dams effected moreefficient in vitro opsonic phagocytosis and kill-ing of K88-positive E. coli than did colostrumfrom resistant dams. Furthermore, colostrumfrom susceptible sows inhibited the binding ofpurified K88 antigen to brush borders signifi-cantly better than the colostrum from resistantdams. Fractionation of colostrum revealed thatfractions rich in immunoglobulin M had thehighest opsonic activity.

Vaccination of a dam with a purified K88preparation can confer passive immunity for aK88-positive pathogen on her offspring (145).Rutter et al. (146) used in vitro tests on serumand mammary secretions from vaccinated andnonvaccinated dams to investigate the nature ofthe protective factors and concluded that neu-tralization of the adhesive properties of the K88antigen by K88 antibodies in colostrum and milkcontribute significantly to the protection of pig-lets from vaccinated dams. The K88 antibodiesappear to be predominantly in the immunoglob-ulin G fraction. Antibodies against the entero-toxin appeared not to be essential for protection,but a possible role of antibodies against the 0antigen present in the colostrum could not beexcluded.

Smith (159) showed that antiserum preparedagainst a K88-positive E. coli strain markedlydelayed or prevented the proliferation of thestrain in the small intestine of piglets to which ithad been administered after challenge with thepathogenic strain. Comparable data were pub-lished by Rutter and Anderson (143). The effectof the antiserum was apparently due to anti-K88antibodies (159). The antiserum was most effec-tive if administrated orally. In vitro adhesion ofK88-positive strains to isolated porcine intesti-nal epithelial cells was also almost completelyinhibited by the addition of anti-K88 antiserumto the incubation mixture (81, 137, 154, 195).Antisera raised against K88ab and K88ac inhib-ited the adhesion of both K88ab- and K88ac-positive cells to the epithelial cells (137). Thesame result was obtained with antiserum specif-

VOL. 46, 1982


mbr.asm

.org/D

ownloaded from



ic for the K88a determinant, and antibodiesagainst K88b and K88c determinants showed theappropriate selectivity for homologous strains.A protective effect of antiserum against theK88a determinant was not detected by Wilsonand Hohmann (195). The effectiveness of theK88a antiserum prepared by these authors, how-ever, is questionable since their serum maycontain the intact K88 antigen.

Since transplacental immunity does not nor-mally occur in pigs, the correlation between hightiters of antibodies in maternal colostrum againstthe adhesin used for vaccination and protectionof offspring leads to the conclusion that protec-tion was a result of consumption of adhesin-specific antibodies in colostrum.Another mechanism for countering the specif-

ic adhesion of K88-positive enterotoxigenic E.coli strains was described by Linggood and co-workers (96-98, 139). The passaging of K88-positive strains through media containing sowcolostrum antibodies raised by vaccination ofpigs with heat-stable antigens results in the lossof the K88 plasmid in vitro (97). Cured strainslose their ability to adhere to and agglutinatechicken erythrocytes. These observations implythat vaccination of dams may result in the pres-ence of high concentrations of "curing antibod-ies" in the intestine of piglets fed with colostrumwhich subsequently may lead to a decrease invirulence of the enterotoxigenic bacteria. Inother words, one of the main effects of immuni-zation is thought to be the selection for adhesin-negative organisms in the intestine of neonatalanimals. More recently, Linggood and Porter(98) observed examples of in vivo plasmid cur-ing, giving rise to K88-negative cells which wereunable to establish infection when administeredto further animals. In this study sows werevaccinated by a combination of the oral andparenteral routes to provide an optimum level ofserum and colostrum antibodies at the time ofparturition and to convey passive immunity tothe newborn piglets. A sow was challenged withlarge numbers ofE. coli 0149:K91:K88ac beforeparturition so that the piglets could receive theinfection soon after birth. They observed a rapidloss of the K88 plasmid from the pathogen in thegut of the immunized sow. Shortly after parturi-tion, 70% of the 0149 cells present in the fecesof the sow were K88 negative. In the feces of 10piglets, a general trend was observed of initiallyhigher numbers of K88-positive cells followedby an increasing number of K88-negative cellsup to 100% of the 0149 population. In anotherpig the reverse happened, and this piglet devel-oped diarrhaea. Apparently, the curing phenom-enon occurs both in the sow and in her progeny.Davidson and Hirsh (24, 25) reported that

inoculation of mice or pigs with a non-enterotox-

in-producing K88-positive E. coli strain protectsthe animals against a subsequent challenge of anenterotoxin-producing K88-positive strain of thesame serotype. They supposed that bacterialcompetition is responsible for protection.

Vaccination of sows with purified 987P adhe-sin confers protection to neonatal suckling pigsagainst diarrheal disease caused by enterotoxi-genic E. coli strains possessing the same antigen(125). This protection did not extend to entero-toxigenic strains possessing other adhesins (75,113). Protection was correlated to the level of987P-specific antibodies in maternal serum andcolostrum (75). Presumably, immunoglobulin Gis the major protective antibody. Adhesion of987P-positive strains to isolated porcine intesti-nal epithelial cells could be prevented by Fabfragments specific for the 987P antigen (76). Inthe same paper it was described that a labora-tory strain of E. coli which possesses type 1fimbriae also adhered to porcine intestinal epi-thelial cells in vitro. This adhesion was prevent-ed by purified type 1 fimbriae and by Fabfragments specific for this antigen. The signifi-cance of this observation in relation to a possiblepathogenicity for pigs of E. coli strains bearingonly type 1 fimbriae remains to be clarified.

Several reports have been published about thepassive immunization of newborn calves againstexperimentally induced enteric colibacillosis byvaccination of dams (120, 126). Vaccines usedwere either whole-cell vaccines (prepared fromlive or Formalin-killed bacteria) or concentratedcrude toxin preparations. The first indicationthat immunity in neonates to enterotoxigenic E.coli strains might arise from the acquisition ofanti-K99 immunoglobulins via ingested colos-trum was reported by Morgan et al. (113) whowere able to protect suckling pigs from entero-toxigenic colibacillosis after oral challenge withE. coli strain 431 (0101:K30,K99:NM) by vacci-nation of their dams with a purified K99 prepara-tion. The incidence and duration of diarrhea andthe degree of intestinal colonization (either dura-tion or extent) were less than observed withother vaccines. In addition, Isaacson et al. (75)showed an increase of K99-specific antibodies inserum and colostrum of pregnant dams afterparenteral vaccination with K99 antigen. Acreset al. (1) vaccinated pregnant cattle subcuta-neously with (i) purified K99 adhesin, (ii) For-malin-killed whole cells of E. coli B44(09:K30,K99:H-), or (iii) a bacterin containingsix different strains of bovine enterotoxigenic E.coli. After birth, calves were allowed to nursetheir dams and were challenged with cells of E.coli B44. Colostral antibody titers were deter-mined against K99, K30, and 09 antigens ofstrain B44. In a nonvaccinated control group, 9of 10 calves developed diarrhea and died within

MICROBIOL. REV.


mbr.asm

.org/D

ownloaded from


NONINVASIVE ENTEROTOXIGENIC E. COLI STRAINS

24 to 72 h. Also, the calves vaccinated with themultiple-strain bacterin developed diarrhea anddied. In contrast, calves nursing cows vaccinat-ed with either purified K99 or the homologouswhole-cell bacterin were protected against lethaldiarrhea. The authors found a highly significantcorrelation (P < 0.0005) between protectionagainst diarrhea and K99, but not K30 or 09,colostral antibody titers. Comparable resultswere reported by Nagy (122). All data are con-sistent with the hypothesis that K99 is a viru-lence factor and immunity against this adhesinprevents diarrhea, probably by blocking adhe-sion of the K99-positive bacteria to the intestinalmucosa. However, the K99 antibodies mightalso act by agglutinating or opsonizing the bacte-ria or both. A comparable study about thepassive protection of lambs against enteropatho-genic E. coli B44 (09:K30,K99:NM) by vaccina-tion of ewes with K99 adhesin isolated from E.coli B41 (0101:K99:NM) was published by Mor-ris et al. (119, 177). After sucking their dams, thelambs were dosed orally within 4 to 21 h afterbirth. Their data suggested that the polysaccha-ride K antigen may contribute to the virulence ofthe enteropathogenic strains since with controllambs challenged with the nonmucoid form ofstrain B44 (09:K99:NM), only loose feces weredetected, whereas strains of the mucoid formproduced a severe, watery diarrhea. None of thevaccinated dams developed diarrhea. Vaccina-tion produced high levels of K99 antibodies inthe serum and colostrum. The antibodies werepredominantly immunoglobulin G and wereshown to be associated with antiadhesive activi-ty.

Unlike the adhesive antigens from E. colistrains producing diarrhea in animals, the colo-nization factor antigens of human enterotoxi-genic E. coli have not been used to produceantisera that were used for vaccination. Immuni-ty to enterotoxigenic E. coli was observed inhuman volunteers who were rechallenged 9weeks after a first infection with an enterotoxi-genic E. coli strain by Levine et al. (94). Howev-er, these investigators used CFA/I-negativestrains. Their studies demonstrate that priordisease due to infection with an enterotoxigenicE. coli strain confers homologous immunityagainst a subsequent challenge.

GENETICS

Genetic Organization of the K88 Determinant

The genetic determinant of the K88 adhesinwas found to be located on a plasmid that couldbe transferred to other bacteria by conjugation(129). Indications were obtained that two plas-

mids were involved in the transfer of the K88determinant: (i) a transfer factor responsible forthe transfer of the K88 genes and/or chromo-somal markers to recipient cells, and (ii) anotherDNA segment on which the K88 determinantwas located (129). Electron microscopic studiesof Bak and co-workers (7) on isolated plasmidDNA of four E. coli strains carrying differentvarieties of the K88 determinant suggested thata plasmid carrying the K88 adhesin and a plas-mid carrying transfer genes might associate to acomposite plasmid that could dissociate againinto a plasmid carrying the K88 determinant andthe transfer plasmid. However, these differentplasmids could not be observed as a singlemolecular species in separate cell lines, so defi-nite phenotypes could not be assigned to thedifferent plasmids observed. Smith and Parsell(167) found that the ability to utilize raffinose(Raf) as a sole carbon source is frequently co-transferred with the K88 determinant. Shipley etal. (157) determined that the K88 and Raf genesare located on a single nonconjugative plasmid,approximately 50 Md in size. In some strainslarger conjugative plasmids were observed thatwere apparently recombinants between the Raf-K88 plasmid and a transfer factor. Mooi et al.(106) located the K88ab and Rafgenetic determi-nants on a K88 plasmid. It appeared that the twodeterminants are not closely linked, but separat-ed by a stretch ofDNA of about 20 Md. Schmittet al. (151) have found that K88 and Raf areflanked by direct repeats of IS1 and can betranslocated by a recA-independent recombina-tion. They suggested that Raf determinants en-dow enteropathogenic strains with a selectiveadvantage, because they enable these strains toestablish resistance against phagocytosis by cellwall production from raffinose.The K88ab (88), K88ac (41, 155, 156), and

K88ad (J. H. Meijerink, unpublished results)genes have been isolated by molecular cloning. Inall three cases the K88 determinant appeared to belocated on a HindIII fragment of approximatelythe same size (7 to 8 Md). The K88ac HindIIIfragment, contained in the recombinant plasmidpPS002, expressed at least six polypeptides inminicells, ranging in size from 18,000 to 70,000daltons. A 23,500-dalton polypeptide was identi-fied as the K88ac subunit with anti-K88ac anti-bodies. This molecular weight is in good agree-ment with the size of purified K88 antigensubunit (105). Plasmid pPS002 was used to con-struct deletion mutants in vitro (88). One dele-tion derivate of pPS002 in which a K88ac DNAfragment of approximately 4.2 Md was retainedwas used to isolate Trn insertion mutants. Byanalyzing the deletion and insertion mutants inminicells, four genes involved in the expressionof the K88ac adhesin were identified and

1S1VOL. 46, 1982


mbr.asm

.org/D

ownloaded from



mapped. The genes appeared to be organizedinto two operons. The first operon contains thegenes for the 17,000-, 29,000-, and 70,000,daltonpolypeptides. The second operon contains thegene of the fimbrial subunit. Mutants containingan insertion in the gene of the 17,000-daltonpolypeptide do not express the fimbrial subunit,and it was suggested that this polypeptide acts asa positive regulator of the second operon. Mu-tants which contain-an insertion in the gene ofthe 70,00-dalton polypeptide appear to shed theK88ac adhesin into the culture supernatant, sug-gesting a role for this polypeptide in the anchor-age of the fimbriae. The deletion and insertionmutants were also analyzed for their ability toagglutinate in anti-K88ac sera, to cause man-nose-resistant hemagglutination of guinea pigerythrocytes, and to adhere to isolated porcineintestinal epithelial cell brush borders. Strainsharboring the 4.2-Md K88ac DNA fragmentwere positive for all three properties. Mostinsertion mutants were negative in these tests.One mutant, however, which contained an inser-tion in the gene ofthe 29,000-dalton polypeptide,was K88 positive, adhered to brush borders, butwas unable to agglutiate guinea pig erythro-cytes. A likely explanation for this observationcould be that the hemagglutination test has alower sensitivity than the adhesion test.Mooi et al. (106) have cloned a 7.7-Md HindIU

DNA fragment containing the K88ab determi-nant (Fig. 8). By excision of a 3.4-Md EcoRIfragment, a smaller plasmid, designated aspFM205, was constructed which still expressedthe K88ab adhesin. The 4.3-Md K88ab DNAfragment contained in pFM205 expressed atleast six polypeptides in minicells, with molecu-

AO4hAa if.

lar weights of 17,000, 26,000, 27,000, 27,500,30,000, and 81,000. The 26,000-dalton polypep-tide was identified as the fimbrial subunit withanti-K88ab antibodies. The 30,000-dalton poly-peptide appeared to be a precursor ofthe 27,000-dalton polypeptide (108).The genes of the five polypeptides were locat-

ed on the physical map ofpFM20S by analysis ofdeletion derivatives of pFM205 in minicells. Byinhibiting proteolytic processing in minicells, itcould be determined that all five genes ex-pressed by the K88ab DNA were translated intoprecursors that were approximately 2,000 dal-tons larger than the mature polypeptides. Thisindicates that all five polypeptides are transport-ed across the cytoplasic membrane by meansof a signal sequence (107, 108). The presence ofa signal sequence in the fimbrial subunit precur-sor was also derived from the base sequence ofthe gene encoding this fimbrial subunit (57) (Fig.4).To study the process of fimbria biosynthesis

and to obtain infonnation about the function ofthe gene products in this process, a set ofdeletion mutants of pFM205 was constructed,each containing a deletion in a different gene,and studied in minicells. Strains containing thedeletion derivatives of pFM205 were also testedfor their ability to adhere to porcine intestinalepithelial cell brush borders and to agglutinateguinea pig erythrocytes. An ELISA was used todetermne whether the K88 antigen was properlytransported to the outside of the ceSl (107, 108).A deletion in the gene of the 81,000-daltonpolypeptide had no influence on the expressionof the fimbrial subunit or the transport of sub-units to the outside of the cell. The fimbrial

0.5Md4

Is I~v

eI I A

io [elo4 es ex~xwqqp$W 14P 4,14 I I a

pBR322 pBR322

K88 r 275 A - 27 1

pBR325 pOR325

1 18.5 lK99

FIG. 8. Physical and genetic map of the K88 and K99 determinants. The numbers enclosed in the thick barsrefer to the molecula weight (in kilodaltons) of the polypeptide encoded by that part of the DNA. The length ofthe thick bars corresponds to the length of DNA required to encode these polypeptides (a length of 0.5 isindicated). The black boxes indicate the presence of a signal sequence. The arrows show the direction oftranscription.

MICROBIOL. REV.


mbr.asm

.org/D

ownloaded from



subunits, however, were not assembled intonormal fimbriae as indicated by the fact thatthey were more thermosensitive and could notbe released from the cells at 65°C. Moreover,strains harboring this deletion plasmid mutantdid not adhere to isolated brush borders oragglutinate guinea pig erythrocytes. Mutantswith a deletion in the gene of the 27,000-daltonpolypeptide also expressed the fimbrial subunit.However, in these mutants the subunit wasrapidly degraded. Apparently, the fimbrial sub-unit is present in a form susceptible to proteoly-sis in the absence of the 27,000-dalton polypep-tide. Although it is clear that there must be aninteraction between the 27,000-dalton polypep-tide and the fimbrial subunit, it is not clear atwhat stage of the biosynthesis of fimbriae thisinteraction occurs. Strains harboring a plasmidwith a deletion in the gene of the 17,000-daltonpolypeptide expressed the fimbrial subunit andexported it to the outside of the cell. Thesestrains did not adhere to brush borders or agglu-tinate erythrocytes, indicating that the fimbrialsubunits are not assembled into normal fimbriae.Possibly, the fimbrial subunit has to be modifiedby the 17,000-dalton polypeptide to be assem-bled in fimbriae. Strains harboring a plasmidwith a deletion in the gene of the 27,500-daltonpolypeptide produced very low amounts of theK88ab adhesin, but still adhered to brush bor-ders and agglutinated guinea pig erythrocytes.All mutations except the deletion in the gene ofthe 27,500-dalton polypeptide could be comple-mented in trans. Therefore, it is not yet clearwhether the 27,500-dalton polypeptide is in-volved in the biosynthesis of the K88ab fimbri-ae.

Although the overall picture is quite similar,there are some differences between the resultsobtained with the cloned K88ab and K88acdeterminants. Kehoe et al. (88) suggest that the17,000-dalton polypeptide is a positive regulatorof the operon containing the gene of the fimbrialsubunit. Mooi (108) has isolated several deletionderivatives of pFM205 which do not express the17,000-dalton polypeptide but still express thefimbrial subunit. Moreover, the location of the17,000-dalton polypeptide outside the cytoplasm(Table 9) excludes a positive regulatory role, atleast at the level of transcription. Kehoe et al.(88) isolated an insertion mutant in the gene ofthe 29,000-dalton polypeptide which producedthe K88ac adhesin, showed adhesion to epitheli-al cells, but did not agglutinate erythrocytes.This polypeptide is probably analogous to the27,000-dalton polypeptide described by Mooi etal. (107), but strains harboring plasmids with adeletion in the gene of the 27,000-dalton poly-peptide did not produce the K88ab adhesin(108).

TABLE 9. Cellular location and possible functionsof the polypeptides encoded by the K88ab

determinantMol wt

ofpoly- Cellular location Possible functionpep-tides

81,000 Outer membrane Assembly and/or anchor-age of fimbrial sub-units

27,500 Periplasm27,000 Periplasm Necessary for integra-

tion of the fimbrialsubunit into outermembrane

26,000 Outer membrane Subunit of fimbriae17,000 Periplasm Involved in posttransla-

tional modification offimbrial subunits

Subcellular Location of Polypeptides Encoded bythe K88ab Determinant

To obtain more information about a possiblerole of the polypeptides encoded by the K88abdeterminant in the biosynthesis of the K88abfimbriae, the subcellular location of these poly-peptides was determined (184). Since all fivepolypeptides encoded by the K88 DNA (81,27.5, 27, 26, and 17 kilodaltons) are synthesizedas precursors, they were expected to be locatedoutside the cytoplasm. Labeled minicells har-boring the chimeric plasmid composed ofpBR322 and the 4.3-Md fragment encoding theK88 antigen were fractionated by a passagethrough a French pressure cell into a membranefraction and a soluble protein fraction. Themembrane fraction was then separated into aninner membrane and an outer membrane frac-tion by sucrose gradient centrifugation. Theperiplasm and the cytoplasm fraction were iso-lated after conversion of minicells to sphero-plasts and disruption of the spheroplasts bysonication. The 81,000-dalton polypeptide wasfound predominantly in the outer membrane, the27,500- and 27,000-dalton polypeptides weremainly in the periplasm, and the 17,000-daltonpolypeptide was found exclusively in the peri-plasm (Table 9). The K88ab protein subunit wasfound in varying amounts in all subcellular frac-tions, due to the compartmentation proceduresused. Most K88ab protein was detected outsidethe cell.

Molecular Cloning of the K99 DeterminantIn 1972 Smith and Linggood (166) reported

that the K99 antigen is encoded by a conjugativeplasmid. The size of this plasmid is about 52 Md(171). Recently, DNA fragments of the conjuga-

153VOL. 46, 1982


mbr.asm

.org/D

ownloaded from



tive K99 plasmid, designated as pRI9901 andoriginating from E. coli B41, were cloned intothe HindIII site of vector pBR322 and subclonedinto the BamHl site ofpBR325 (185). The small-est fragment obtained that still expressed theK99 antigen was 4.5 Md in size. With regard tothe serological, adhesive, and morphologicalproperties, no difference in the nature of the K99antigen was observed between E. coli strainscarrying the recombinant plasmid and the straincarrying pRI9901. One of the clones harboringthe 4.5-Md fragment produced very little K99,although this fragment was indistinguishablefrom the fragments harbored by the K99-posi-tive recombinants as analyzed by multiple-cut-ting restriction endonucleases. SpontaneousK99-negative mutants of E. coli B41 have beendescribed (117) carrying plasmid DNA thatshowed no difference from their K99-positiveparental strain. These observations might indi-cate that the expression of the K99 genes can beswitched off without loss of the K99-encodingDNA as described for type 1 and 987P fimbriae.E. coli K-12 strains carrying K99 recombinantplasmids produced 16 to 32 times more K99 thandid the E. coli K-12 strain harboring the parentalplasmid pRI9901. The regulation of K99 expres-sion by recombinant plasmid DNA was similarto that of natural K99-positive isolates. By delet-ing various regions in the K99 DNA, the positionof the structural gene of the K99 subunit couldbe estimated. The precise location of this genewas determined by nucleotide sequencing (28).Like the K88 protein subunit, the K99 proteinsubunit is synthesized in a precursor form with asignal peptide of 22 amino acid residues (Fig. 8).Using F41 antiserum, no F41 antigen could be

detected on E. coli cells to which the K99-encoding plasmid was transferred by conjuga-tion. This indicates that the genetic determinantfor the production of F41 antigen is not locatedon the K99 plasmid (29). It is not known whetherthe F41 antigen determinant is located on thechromosome or on another plasmid carried byE. coli B41, i.e., the plasmid encoding for STproduction.

Characterization of lids Encoding CFA/Iand CFA/IU

E. coli strain H10407 (078:K80:H11) containsthree plasmids with molecular weights of 60, 42,and 3.7 Md, respectively (51). A laboratory-passed derivative strain H10407-P still possessesthe 42- and 3.7-Md plasmids. The loss of the 60-Md plasmid in H10407-P is accompanied by theloss of CFA/I and ST production (51). Consis-tent with this observation, Smith and co-work-ers (160) have shown that the loss of expressionof CFA/I and ST from five strains belonging tothe 078 serogroup and isolated in different coun-

tries corresponds to the loss of a 56- to 61-Mdplasmid. McConnell et al. (99) have proven thatCFA/I and ST production are encoded by asingle nonconjugative plasmid in strains belong-ing to serogroup 078. One strain of serotype063:H- contained a 65-Md plasmid coding forCFA/I and both ST and LT production (Table10). Reis et al. (140) described a nonconjugativeCFA/I-ST plasmid of 61 Md in strains of sero-type 0128ac:H12 isolated from children withdiarrhea in Brazil. Plasmids coding for CFAIIand ST have also been identified in E. colistrains of serogroup 026, 063, 0114, 0128, and0153 (R. L. Shipley, D. G. Evans, and D. J.Evans, Jr., unpublished results; 23). The strainswere isolated in different countries, and themajority also had a plasmid of approximately 60Md in size. Cravioto et al. (23) compared theCFA/I-ST plasmids of several of these strainswith the plasmids from the 078 strains by re-striction enzyme analysis. Cleavage with therestriction endonucleases EcoRI or HindIII indi-cated that all plasmids shared several fragments.Generally there appeared to be more fragmentsin common between plasmids identified instrains isolated in the same geographical area.One particular CFA/I-ST plasmid, NTP113,

was characterized in more detail in order toidentify the region(s) coding for CFA/I produc-tion. Different transposons were inserted intoNTP113, and E. coli strains carrying the plasmidderivatives were tested for CFA/I and for STproduction. CFA/I-negative, ST-positive mu-tants were obtained. These mutants were thenused to construct deletions. It appeared that tworegions of plasmid NTP113 are required forexpression of CFA/I. The two regions are sepa-rated by a length of DNA corresponding to amolecular weight of at least 25 Md. A plasmidwith a mutation mapping in one CFA/I regionwas tested for complementation with anotherplasmid derivative with a mutation in the otherCFA/I region. Strains carrying these two plas-mids produced CFA/I (23, 168).Some of the CFA/II-positive strains examined

by Cravioto et al. (23) lost the ability to producethe CFAIII simultaneously with the productionof ST and LT. Other strains, however, lost these

TABLE 10. Genetics of adhesin productionMol

Adhesin Genetic location wt Transmissibility(xl06)

K88 Plasmid 51 Nonconjugative90 Conjugative

987P Chromosome?K99 Plasmid 57 ConjugativeCFAII Plasmid 52-65 NonconjugativeCFA/II Plasmid 60 Nonconjugative

MICROBIOL. REV.


mbr.asm

.org/D

ownloaded from



properties separately. Some of the strains thathad lost the ability to produce the CFA/II, withor without enterotoxin production, had lost aplasmid. Other CFAlII-positive and -negativepairs, however, still had the same plasmid con-tent. Penaranda et al. (138) described that CFA/1I-positive strains from different geographic lo-cations contain a 60-Md large nonconjugativeplasmid coding for CFAIII, ST, and LT produc-tion.

CONCLUDING REMARKSThe etiological role of enterotoxigenic E. coli

strains in diarrheal diseases ofman and domesticanimals has been firmly established during thelast two decades. Besides the production ofenterotoxins, at least one other important viru-lence factor, the ability to adhere to the intesti-nal epithelial mucosa by fimbria-like adhesins,has been recognized. However, other cell sur-face antigens (i.e., K antigens) also contribute tothe capacity to provoke diarrhea in combinationwith adhesin and enterotoxin production. Themajority of enterotoxigenic E. coli strains isolat-ed from neonatal pigs, calves, or lambs possessone of the adhesive antigens-K88, K99, or987P. A limited number of enterotoxigenicstrains of human origin produce CFAII or CFA/II. Other host-specific adhesins, however, havebeen observed on these strains.

In contrast to the presence of type 1 fimbriaewhich occur on the majority of E. coli strains,striking correlations have been observed be-tween the presence of host-specific adhesins,the production of enterotoxins, and the serotypeof the pathogen. Both the production of entero-toxins and these adhesins appear to be restrictedto a relatively small number of serotypes. Thereason for this restriction is not known. A possi-ble explanation might be a difference in theability of strains belonging to various serotypesto stably maintain the plasmids encoding thesevirulence factors. The location of CFA/I orCFA/II and enterotoxin production on the sameplasmid obviously explains the close associationof both properties. The close association ofenterotoxin production and the presence of oth-er adhesins like K88 or K99 cannot be explainedin this way because their genetic determinantsare mostly located on different plasmids. Unfor-tunately, the interpretation of the data presentedin several studies on the occurrence and charac-teristics of the various adhesins is obscured bythe fact that the authors insufficiently realizedthe possibility that E. coli strains produce morethan one type of fimbriae at the same time.Cloning of cells producing only one type offimbriae, as described by Brinton, can help toavoid this confusion.

Production of adhesins is affected by environ-

mental conditions such as temperature and com-position of the growth medium. None of theadhesins is synthesized below 20°C, indicating aregulatory mechanism that suppresses adhesinproduction at non-physiological temperatures.The effect of medium composition on adhesinproduction varies for the respective antigens. Ingeneral, the synthesis of adhesins seems to beoptimal when the strains are grown in minimalmedia. Complex media can contain substanceswhich inhibit adhesin production as shown forK99 and F41. At the present no data are avail-able to indicate whether these factors regulatethe number of adhesin-positive cells in a popula-tion, the number of fimbriae per cell, or thelength of individual fimbriae. Another importantmechanism that controls the production of fim-briae is the phase variation. This phenomenonhas been extensively studied for the universaladhesin of E. coli strains, namely, type 1 fimbri-ae. Phase variation has also been described forsome host-specific adhesins, such as the 987Pantigen.Most of the host-specific adhesins can be

described as very thin and flexible filamentousproteinaceous appendages of the bacterial cellsurface. Some of the adhesins however (i.e.,987P) cannot be distinguished from type 1 fim-briae as observed in the electron microscope.

Despite their similar function, no similaritiesin the primary structure of the adhesin subunitshave been detected so far. This similar function,however, does not imply that adhesins are ableto mediate the interaction between bacteria andany type of epithelial cell because each adhesin,except for type 1 fimbriae, exhibits a strikinghost and tissue specificity in vivo. This specific-ity resides in the chemical composition of aparticular adhesin and the surface componentsof the host epithelial cells.

Relatively few studies have been done toresolve the identity of the intestinal epithelialcell surface components which function as "re-ceptor sites" for adhesins. Several indicationshave been obtained that carbohydrate chains areessential components of these receptors. Fur-thermore, data obtained with K88- and K99-positive strains denote a nonrandom distributionof receptor sites along the intestinal mucosa.Hopefully, the study of receptor substances andthe mechanism of fimbriae-mediated adhesionwill receive more attention in the near future.Immunity of neonates to infections has been

obtained by immunization of dams with purifiedadhesin preparations. Several mechanisms havebeen proposed for the way in which this immuni-ty is acquired, for instance antibody-mediatedloss of virulence plasmids and prevention ofadhesion.

Probably with the exception of 987P, the

VOL. 46, 1982 155


mbr.asm

.org/D

ownloaded from



genetic determinants encoding for the synthesisof host-specific adhesins are located on plas-mids. The necessary genetic information ap-pears to be organized in one or two operonsencoding for several polypeptides. The possiblerole of these polypeptides in the biogenesis ofadhesins will be a subject of future studies.

ACKNOWLEDGMENTSWe thank J. van Breemen, J. A. Morris, and L. K. Nagy for

supplying the electron micrographs, and P. A. M. Guin6e,J. D. A. van Embden, J. A. Morris, and F. R. Mooi for helpfulsuggestions and critical reading of the manuscript.

LITERATURE CITED1. Acre, S. D., R. E. Isaacson, L. A. Babluk, and R. A.Kapay. 1979. Immunization of calves against entero-toxigenic colibacillosis by vaccinating dams with purifiedK99 antigen and whole cell bacterins. Infect. Immun.25:121-126.

2. Andera, M. J., J. S. Whiebead, and Y. S. Kin. 1980.Interaction of Escherichia coli K88 antigen with porcineintestinal brush border membranes. Infect. Immun.29:897-901.

3. Anderson, T. F. 1949. On the mechanism of adsorption ofbacteriophages and host cells, p. 76-95. In A. A. Milesand N. W. Pirie (ed.), The nature of the bacterial surface.Oxford University Press, Oxford, England.

4. Arbukde, J. B. R. 1970. The location of Escherichia coliin the pig intestine. J. Med. Microbiol. 3:333-340.

5. Awad-Ma e, M., H. W. Moon, P. L. Rnnels, andR. A. S der. 1982. Pilus production, hemagglutina-tion, and adhesion by porcine strains of enterotoxigenicEscherichia coli lacking K88, K99, and 987P antigens.Infect. Immun. 35:305-313.

6. Bick, E., R. MOlby, B. KaJser, G. Stinatzlng, T. Wad-str6m, and D. Habte. 1980. Enterotoxigenic Escherichiacoli and other gram-negative bacteria of infantile diar-rhea: surface antigens, hemagglutinins, colonization fac-tor antigen, and loss of enterotoxigenicity. J. Infect. Dis.142:318-327.

7. Bak, A. L., G. Chrasn, C. Cnitanen, A. Sten-dermp, I. 0rskov, and F. 0rskov. 1972. Circular DNAmolecules controlling synthesis and transfer of the sur-face antigen (K88) in Escherichia coli. J. Gen. Microbiol.73:373-385.

8. Berbacgr, H. U., H. W. Moon, and S. C. WWbpp. 1972.Association of Escherichia coli with the small intestinalepithelium. I. Comparison of enteropathogenic and non-enteropathogenic porcine strains in pigs. Infect. Immun.5:595-605.

9. BijIsma, 1. G. W., A. de NUs, and J. F. FrIk. 1981.Adherence ofEscherichia coli to porcine intestinal brushborders by means of serological variants of the K88antigen. Antonie van Leeuwenhoek J. Microbiol. Serol.47:467-468.

10. Bletok, E., ad K. Jana. 1982. Adhesion of piliatedEscherichia coli strains to phagocytes: differences be-tween bacteria with mannose-sensitive pili and thosewith mannose-resistant pili. Infect. Immukn. 35:264-269.

11. Braaten, B. A., and L. L. Myers. 1977. Biochemicalcharacteristics of enterotoxigenic and nonenterotoxi-genic Escherichia coli isolated from calves with diarrhea.Am. J. Vet. Res. 38:1989-1991.

12. Brlaton, C. C. 1959. Non-flagellar appendages of bacte-ria. Nature (London) 183:782-786.

13. Brlnton, C. C,, Jr. 1965. The structure, function, synthe-sis and genetic control of bacterial pili and a molecularmodel for DNA and RNA transport in gram negativebacteria. Trans. N.Y. Acad. Sci. 27:1003-1054.

14. Brinton, C. C., Jr. 1967. Contributions of pili to the

specificity of the bacterial surface, and a unitary hypoth-esis of conjugal infections heredity, p. 37-70. In B. D.Davis and L. Warren (ed.), The specificity of cell sur-faces. Prentice-Hall Inc., Englewood Cliffs, N.J.

15. Briaton, C. C. 1978. Studies on the piliation of a humanenteropathogen, Escherichia coli, H-10407, p. 34-70. InProceedings of the 13th Joint Conference on Cholera.Publication number (NIH) 78-1590. Department ofHealth, Education and Welfare, Washington, D.C.

16. Brinton, C. C., Jr., A. BuzzeR, and M. A. Lauffer. 1954.Electrophoresis and phage susceptibility studies on afilament-producing variant of the E. coli B bacterium.Biochem. Biophys. Acta 15:533-542.

17. Burrows, M. R., R. Sellwood, and R. A. Gibbons. 1976.Haemagglutinating and adhesive properties associatedwith the K99 antigen of bovine strains of Escherichiacoli. J. Gen. Microbiol. 96:269-275.

18. Cantey, J. R., and R. K. Blake. 1977. Diarrhea due toEscherichia coli in the rabbit: a novel mechanism. J.Infect. Dis. 135:454-462.

19. Cbeney, C. P., E. C. Boedeker, and S. B. Formal. 1979.Quantitation of the adherence of an enteropathogenicEscherichia coli to isolated rabbit intestinal brush bor-ders. Infect. Immun. 26:736-743.

20. Cheney, C. P., P. A. Schad, S. B. Formal, and E. C.Boedeker. 1980. Species specificity of in vitro Escherich-ia coli adherence to host intestinal cell membranes andits correlation with in vivo colonization and infectivity.Infect. Immun. 28:1019-1027.

21. Contrepols, M., J. P. Ghradeau, H. C. Dubourguler, andP. Gonet. 1982. Mise en evidence de quelques facteursintervenant dans la biosynthese de l'ag K99, p. 206-227.In P. Pohl and J. Leunen (ed.), Resistance and pathogen-ic plasmids. C.E.C. Seminar 1981 Brussels, NationalInstitute for Veterinary Research.

22. Cravkoto, A., R. J. Gros, S. M. Scotand, and B. Rowe.1979. Mannose-resistant haemagglutination of humanerythrocytes by strains of Escherichia coli from extrain-testinal sources: lack of correlation with colonisationfactor antigen (CFA/I). FEMS Microbiol. Lett. 6:41-44.

23. Cravoto, A., M. M. McConnell, B. Rowe, S. M. Scotand,H. R. Smith, and G. A. Willsbaw. 1980. Studies ofcolonisation factors in Escherichia coli strains isolatedfrom humans: identification and plasmid analysis, p.530-532. In R. C. W. Berkely, J. M. Lynch, J. Melling,P. R. Rutter, and B. Vincent (ed.), Microbial adhesion tosurfaces. Ellis Horwood Ltd., Chichester, England.

24. Davidson, J. N., and D. C. Hhrsh. 1975. Use of the K88antigen for in vivo bacterial competition with porcinestrains of enteropathogenic Escherichia coli. Infect. Im-mun. 12:134-136.

25. Davdson, J. N., and D. C. Hh-sh. 1976. Bacterial compe-tition as a means of preventing neonatal diarrhea in pigs.Infect. Immun. 13:1773-1774.

26. de Grad, F. K., P. Klaen-Boor, and J. E. van Hees.1980. Biosynthesis of the K99 surface antigen is re-pressed by alanine. Infect. Immun. 3t:125-128.

27. de Graaf, F. K., P. Klemm, and W. Gastra. 1981.Purification, characterization and partial covalent struc-ture of the adhesive antigen K99 of Escherichia coli.Infect. Immun. 33:877-3.

28. de Graaf, F. K., and F. R. Mooi. 1982. The K88 and K99plasmids of enterotoxigenic Escherichia coli strains, p.191-203. In P. Pohl and J. Leunen (ed.), Resistance andpathogenic plasmids, C.E.C. Seminar 1981 Brussels,National Institute for Veterinary Research.

29. de Graaf, F. K., and I. Roorda. 1982. Production, purifi-cation and characterization of the fimbrial adhesiveantigen F41 isolated from the calf enteropathdgenicEscherichia coli strain B41 M. Infect. Immun. 36:.7S1-753.

30. de Graaf, F. K., F. B. Wkentjes, and P. Klaen-Boor.1980. Production of K99 antigen by enterotoxigenicEscherichia coli strains of antigen groups 08, 09, 020, and0101 grown at different conditions. Infect. Immun.

MICROBIOL. REV.


mbr.asm

.org/D

ownloaded from



27:216-221.31. De a Cabada, F. J., D. G. Evans, and D. J. Evans, Jr.

1981. Immunoprotection against enterotoxigenic Esche-richia coli diarrhea in rabbits by peroral administration ofpurified colonization factor antigen I (CFA/I). FEMSMicrobiol. Lett. 11:303-307.

32. Deneke, C. F., G. M. Thorne, and S. L. Gorbach. 1979.Attachment pili from enterotoxigenic Escherichia colipathogenic for humans. Infect. Immun. 26:362-368.

33. Deneke, C. F., G. M. Thorne, and S. L. Gorbach. 1981.Serotypes of attachment pili of enterotoxigenic Esche-richia coli isolated from humans. Infect. Immun.32:1254-1260.

34. Drees, D. T., and G. L. Waxier. 1970. Enteric colibacillo-sis in gnotobiotic swine. An electron microscopic study.Am. J. Vet. Res. 31:1159-1171.

35. Duguid, J. P., and E. S. Anderson. 1967. Terminology ofbacterial fimbriae, or pili, and their types. Nature (Lon-don) 215:89-90.

36. Duguld, J. P., E. S. Anderson, and I. Campbell. 1966.Fimbriae and adhesive properties in Salmonellae. J.Pathol. Bacteriol. 92:107-137.

37. Duguld, J. P., and D. C. Old. 1980. Adhesive propertiesof Enterobacteriaceae, p. 185-217. In E. H. Beachey(ed.), Bacterial adherence, receptors and recognition,series B, vol. 6. Chapman and Hall, London.

38. Duguid, J. P., W. Smith, G. Dempster, and P. N. Ed-munds. 1955. Nonflageilar filamentous appendages("fimbriae") and haemmagglutinating activity in Bacteri-um coli. J. Pathol. Bacteriol. 70:335-348.

39. Eisenstein, B. I. 1981. Phase variation of type 1 fimbriaein Escherichia coli is under transcriptional control. Sci-ence 214:337-339.

40. EUens, D. T., P. W. de Leeuw, and H. Rozemond. 1979.Detection of the K99 antigen of Escherichia coli in calffaeces by enzyme-linked immunosorbent assay(ELISA). Vet. Q. 1:169-175.

41. Elwell, L. P., and P. L. Shipley. 1980. Plasmid-mediatedfactors associated with virulence of bacteria to animals.Annu. Rev. Microbiol. 34:465-496.

42. Evans, D. G., and D. J. Evans, Jr. 1978. New surface-associated heat-labile colonization factor antigen (CFA/II) produced by enterotoxigenic Escherichia coli of sero-groups 06 and 08. Infect. Immun. 21:638-647.

43. Evans, D. G., D. J. Evans, Jr., and S. Clegg. 1980.Detection of enterotoxigenic Escherichia coli coloniza-tion factor antigen I in stool specimens by an enzyme-linked immunosorbent assay. Infect. Immun. 12:738-743.

44. Evans, D. G., D. J. Evans, Jr., S. Clegg, and J. A. Pauley.1979. Purification and characterization of the CFA/Iantigen of enterotoxigenic Escherichia coli. Infect. Im-mun. 25:738-748.

45. Evans, D. G., D. J. Evans, Jr., and H. L. DuPont. 1977.Virulence factors of enterotoxigenic Escherichia coli. J.Infect. Dis. 136:S118-S123.

46. Evans, D. J., Jr., D. G. Evans, and H. L. DuPont. 1979.Haemagglutination pattems of enterotoxigenic and en-teropathogenic Escherichia coli detennined with human,bovine, chicken, and guinea pig erythrocytes in thepresence and absence of mannose. Infect. Immun.23:336-346.

47. Evans, D. J., Jr., D. G. Evans, H. L. DuPont, F. 0rskov,and I. 0rskov. 1977. Patterns of loss of enterotoxigen-icity by Escherichia coli isolated from adults with diar-rhea: suggestive evidence for an interrelationship withserotype. Infect. Immun. 17:105-111.

48. Evans, D. G., D. J. Evans, Jr., and W. Tjoa. 1977.Hemagglutination of human group A erythrocytes byenterotoxigenic and enteropathogenic Escherichia coliisolated from adults with diarrhea: correlation with colo-nization factor. Infect. Immun. 18:330-337.

49. Evans, D. G., D. J. Evans, Jr., W. S. Tjoa, and H. L.DuPont. 1978. Detection and characterization of coloni-zation factor of enterotoxigenic Escherichia coli isolated

from adults with diarrhea. Infect. Immun. 19:727-736.50. Evans, D. G., T. K. Satterwhite, D. J. Evans, Jr., and

H. L. DuPont. 1978. Differences in serological responsesand excretion patterns of volunteers challenged withenterotoxigenic Escherichia coli with and without thecolonization factor antigen. Infect. Immun. 19:883-888.

51. Evans, D. G., R. P. Silver, D. J. Evans, Jr., D. G. Chase,and S. L. Gorbach. 1975. Plasmid-controlled colonizationfactor associated with virulence in Escherichia coli en-terotoxigenic for humans. Infect. Immun. 12:656-667.

52. Eylar, E. H. 1965. On the biological role of glycopro-teins, J. Theor. Biol. 10:89-113.

53. Faris, A., M. Lindahl, and T. Wadstr6m. 1980. GM2-likeglycoconjugate as possible erythrocyte receptor for theCFA/I and K99 haemagglutinins of enterotoxigenic Esch-erichia coli. FEMS Microbiol. Lett. 7:265-269.

54. Freer, J. H., A. Ellis, T. Wadstrdm, and C. J. Sihyth.1978. Occurrence of fimbriae among enterotoxigenicintestinal bacteria isolated from cases of human infantilediarrhoea. FEMS Microbiol. Lett. 3:277-281.

55. Fusco, P., A. To, S. To, and C. Brinton, Jr. 1978. Thepurification and characterization of four types of E. colipili and the specificity of E. coli pili for immunity,colonization and adhesion, p. 60-70. In C. Miller (ed.),XIIP" U.S.-Japan Conference on Cholera, Atlanta, Ga.,September, 1977. National Institutes of Health, Bethes-da, Md.

56. Gaastra, W., P. Klemm, J. M. Walker, and F. K.de Graaf. 1979. K88 fimbrial protein: amino- and carbox-yl terminal sequences of intact proteins and cyanogenbromide fragments. FEMS Microbiol. Lett. 6:15-18.

57. Gaastra, W., F. R. Mooi, A. R. Stuitje, and F. K. deGraaf. 1981. The nucleotide sequence of the gene encod-ing the K88ab protein subunit of porcine enterotoxigenicEscherichia coli. FEMS Microbiol. Lett. 12:4146.

58. Gibbons, R. A., G. W. Jones, and R. Sellwood. 1975. Anattempt to identify the intestinal receptor for the K88adhesin by means of haemagglutination inhibition testusing glycoproteins and fractions from sow colostrum. J.Gen. Microbiol. 86:228-240.

59. Gross. R. J., A. Cravioto, S. M. Scotland, T. Cheasty, andB. Rowe. 1978. The occurrence of colonization factor(CF) in enterotoxigenic Escherichia coli. FEMS Micro-biol. Lett. 3:231-233.

60. Guinte, P. A. M., C. M. Agterberg, W. H. Jansen, andJ. F. Frik. 1977. Serological identification of pig entero-toxigenic Escherichia coli strains not belonging to theclassical serotypes. hnfect. Immun. 15:549-555.

61. Guin6e, P. A. M., and W. H. Jansen. 1979. Behavior ofEscherichia coli K antigens K88ab, K88ac, and K88ad inimmunoelectrophoresis, double diffusion, and hemagglu-tination. Infect. Immun. 23:700-705.

62. GuiSe, P. A. M., and W. H. Jansen. 1979. Detection ofenterotoxigenicity and attachment factors in Escherichiacoli strains of human, porcine and bovine origin; acomparative study. Zentralbl. Bacteriol. Parasitenkd.Infectionskr. Hyg. Abt. Orig. Reihe A. 243:245-257.

63. GulnEe, P. A. M., W. H. Jansen, and C. M. Agterberg.1976. Detection of the K99 antigen by means of aggluti-nation and immunoelectrophoresis in Escherichia coliisolates from calves and its correlation with enterotoxi-genicity. Infect. Immun. 13:1369-1377.

64. Guinte, P. A. M., J. Veltkamp, and W. H. Jansen. 1977.Improved Minca medium for the detection of K99 anti-gen in calf enterotoxigenic strains of Escherichia coli.Infect. Immun. 15:676-678.

65. Gyles, C. L. 1971. Heat-labile and heat-stable forms ofthe enterotoxin from E. coli strains enteropathogenic forpigs. Ann. N.Y. Acad. Sci. 176:314-322.

66. Gyles, C. L., and D. A. Barnum. 1969. A heat-labileenterotoxin from strains ofEscherichia coli enteropatho-genic for pigs. J. Infect. Dis. 120:419426.

67. Gyles, C. L., M. So, and S. Falkow. 1974. The enterotox-in plasmids of Escherichia coli. J. Infect. Dis. 130:40-49.

68. Hermodson, M. A., K. C. S. Chen, and T. M. Buchanan.

157VOL. 46, 1982


mbr.asm

.org/D

ownloaded from



1978. Neisseria pili proteins: amino-terminal amino acidsequences and identification of an unusual amino acid.Biochemistry 17:442-445.

69. Hohmann, A., and M. R. Wilson. 1975. Adherence ofenteropathogenic Escherichia coli to intestinal epitheli-um in vivo. Infect. Immun. 12:866-880.

70. Houwink, A. L., and W. van Iterson. 1950. Electronmicroscopical observations on bacterial cytology. II. Astudy on flagellation. Biochim. Biophys. Acta 5:10-44.

71. Isaacson, R. E. 1977. K99 surface antigen of Escherichiacoli: purification and partial characterization. Infect.Immun. 15:272-279.

72. Isaacson, R. E. 1978. K99 surface antigen of Escherichiacoli: antigenic characterization. Infect. Immun. 22:555-559.

73. Isaacson, R. E. 1980. Factors affecting expression of theEscherichia coli pilus K99. Infect. Immun. 28:190-194.

74. Isaacson, R. E., J. Colmenero, and P. Richter. 1981.Escherichia coli K99 pili are composed of one subunitspecies. FEMS Microbiol. Lett. 12:229-232.

75. Isaacson, R. E., E. A. Dean, R. L. Morgan, and H. W.Moon. 1980. Immunization of suckling pigs against en-terotoxigenic Escherichia coli-induced diarrheal diseaseby vaccinating dams with purified K99 or 987P pili:antibody production in response to vaccination. Infect.Immun. 29:824-826.

76. Isaacson, R. E., P. C. Fusco, C. C. Brinton, and H. W.Moon. 1978. In vitro adhesion of Escherichia coli toporcine small intestinal epithelial cells: pili as adhesivefactors. Infect. Immun. 21:392-397.

77. Isaacson, R. E., H. W. Moon, and R. A. Schneider. 1978.Distribution and virulence of Escherichia coli in thesmall intestines of calves with and without diarrhea. Am.J. Vet. Res. 39:1750-1755.

78. Isaacson, R. E., B. Nagy, and H. W. Moon. 1977.Colonization of porcine small intestine by Escherichiacoli: colonization and adhesion factors of pig entero-pathogens that lack K88. J. Infect. Dis. 135:531-539.

79. Isaacson, R. E., and P. Richter. 1981. Escherichia coli987P pilus: purification and partial characterization. J.Bacteriol. 146:784-789.

80. Jones, G. W. 1977. The attachment of bacteria to thesurfaces of animal cells, p. 139-176. In J. L. Reissig(ed.), Microbial interactions, receptors and recognition,series B, vol. 3. Chapman and Hall, London.

81. Jones, G. W., and J. M. Rutter. 1972. Role of the K88antigen in the pathogenesis of neonatal diarrhea causedby Escherichia coli in piglets. Infect. Immun. 6:918-927.

82. Jones, G. W., and J. M. Rutter. 1974. The association ofK88 antigen with haemagglutinating activity in porcinestrains of Escherichia coli. J. Gen. Microbiol. 84:135-144.

83. Jones, G. W., and J. M. Rutter. 1974. Contribution of theK88 antigen of Escherichia coli to enteropathogenicity;protection against disease by neutralizing the adhesiveproperties of K88 antigen. Am. J. Clin. Nutr. 27:1414-1449.

84. Kaeckenbeeck, A., M. Josse, and F. Schoenaers. 1977.Relations chez les Escherichia coli d'origine bovine entrele pouvoir enteropathogene, l'antigene K99 et l'hemag-glutination. Am. Med. Vet. 121:239-248.

85. Kaillenius, G., R. Mollby, S. B. Svenson, J. Winberg, A.Lundblad, S. Svensson, and B. Cedergren. 1980. The Pkantigen as receptor for the haemagglutinin of pyelone-phritic Escherichia coli. FEMS Microbiol. Lett. 7:297-302.

86. Kauffmann, F., and G. Vahlne. Ueber die Bedeutung desserologischen Formenwechsels fur die Bakteriophagen-wirkung in der Coli-Gruppe. Acta Pathol. Microbiol.Scand. Sect. B 22:119-137.

87. Kearns, M. J., and R. A. Gibbons. 1979. The possiblenature of the pig intestinal receptor for the K88 antigen ofEscherichia coli. FEMS Microbiol. Lett. 6:165-168.

88. Kehoe, M., R. SelHwood, P. Shipley, and G. Dougan. 1981.Genetic analysis of K88-mediated adhesion of enterotox-

igenic Escherichia coli. Nature (London) 291:122-126.89. Ketyi, I., B. Kuch, and A. Vertenyi. 1978. Stability of

Escherichia coli adhesive factors K88 and K99 in mice.Acta Microbiol. Acad. Sci. Hung. 25:77-86.

90. Klemm, P. 1979. Fimbrial colonization factor CFA/1protein from human enteropathogenic Escherichia colistrains. FEBS Lett. 108:107-110.

91. Klemm, P. 1981. The complete amino acid sequence ofthe K88 antigen, a fimbrial protein from Escherichia coli.Eur. J. Biochem. 117:617-627.

92. Klemm, P. 1982. Primary structure of the CFAI fimbrialprotein from human enterotoxigenic Escherichia colistrains. Eur. J. Biochem., in press.

93. Leffler, H., and C. Svanborg-Eden. 1981. Glycolipidreceptors for uropathogenic Escherichia coli on humanerythrocytes and uroepithelial cells. Infect. Immun.34:920-929.

94. Levine, M. M., D. R. Nalin, D. L. Hoover, E. J. Berg-quist, R. B. Hornick, and C. R. Young. 1979. Immunity toenterotoxigenic Escherichia coli. Infect. Immun. 23:729-736.

95. Levine, M. M., M. B. Rennels, V. Daya, and T. P.Hughes. 1980. Hemagglutination and colonization factorsin enterotoxigenic and enteropathogenic Escherichia colithat cause diarrhea. J. Infect. Dis. 141:733-737.

%. Linggood, M. A., M. L. Ellis, and P. Porter. 1979. Anexamination of the 0 and K specificity involved in theantibody-induced loss of the K88 plasmid from porcineenteropathogenic strains of Escherichia coli. Immunolo-gy 38:123-127.

97. Linggood, M. A., and P. Porter. 1978. Antibody inducedelimination of the plasmid controlled K88 adhesion fac-tor from a porcine enteropathogen. Immunology 35:125-127.

98. Linggood, M. A., and P. Porter. 1980. The antibody-mediated elimination of adhesion determinant from en-teropathogenic strains of Escherichia coli, p. 441-453. InR. C. W. Berkeley, J. M. Lynch, J. Melling, P. R.Rutter, and B. Vincent (ed.), Microbial adhesion tosurfaces. Ellis Horwood Ltd., Chichester, England.

99. McConnell, M. M., H. R. Smith, G. A. Willshaw, A. M.Field, and B. Rowe. 1981. Plasmids coding for coloniza-tion factor antigen I and heat-stable enterotoxin produc-tion isolated from enterotoxigenic Escherichia coli com-parison of their properties. Infect. Immun. 32:927-936.

100. McConnell, M. M., H. R. Smith, G. A. Wilishaw, S. M.Scotland, and B. Rowe. 1980. Plasmids coding for heat-labile enterotoxin production isolated from Escherichiacoli 078: comparison of properties. J. Bacteriol.143:158-167.

101. McNeish, A. S., P. Turner, J. Fleming, and N. Evans.1975. Mucosal adherence of human enteropathogenicEscherichia coli. Lancet ii:946-948.

102. Merson, M. H., F. 0rskov, I. 0rskov, R. B. Sack, I. Huq,and F. T. Koster. 1979. Relationship between enterotox-in production and serotype in enterotoxigenic Escherich-ia coli. Infect. Immun. 23:325-329.

103. Merson, M. H., R. B. Sack, S. Islam, G. Saklayen, N.Huda, I. Huq, A. W. Zulich, R. H. Yolken, and A. Z.Kapikian. 1980. Disease due to enterotoxigenic Esche-richia coli in Bangladeshi adults: clinical aspects and acontrolled trial of tetracycline. J. Infect. Dis. 141:702-711.

104. Minshew, B. H., J. Jorgensen, M. Swanstrum, G. A.Grootes-Reuvecamp, and S. Falkow. 1978. Some charac-teristics of Escherichia coli strains isolated from extrain-testinal infections of humans. J. Infect. Dis. 137:648-654.

105. Mooi, F. R., and F. K. de Graaf. 1979. Isolation andcharacterization of K88 antigens. FEMS Microbiol.Lett. 5:17-20.

106. Mooi, F. R., F. K. de Graaf, and J. D. A. van Embden.1979. Cloning, mapping, and expression of the geneticdeterminant that encodes for the K88ab antigen. NucleicAcids Res. 6:849-865.

MICROBIOL. REV.


mbr.asm

.org/D

ownloaded from



107. Mooi, F. R., N. Harms, D. Bakker, and F. K. de Graaf.1981. Organization and expression of genes involved inthe production of the K88ab antigen. Infect. Immun.32:1155-1163.

108. Mooi, F. R., C. Wouters, A. Wifjes, and F. K. de Graaf.1982. Construction and characterization of mutants im-paired in the biosynthesis of the K88ab antigen. J.Bacteriol. 150:512-521.

109. Moon, H. W., E. M. Kohler, R. A. Schneider, and S. C.Whipp. 1980. Prevalence of pilus antigens, enterotoxintypes and enteropathogenicity among K88-negative en-terotoxigenic Escherichia coli from neonatal pigs. Infect.Immun. 27:222-230.

110. Moon, H. W., B. Nagy, R. E. Isaacson, and I. 0rskov.1977. Occurrence of K99 antigen on Escherichia coliisolated from pigs and colonization of pig ileum by K99+enterotoxigenic E. coli from calves and pigs. Infect.Immun. 15:614-620.

111. Moon, H. W., and S. C. Whipp. 1970. Development ofresistance with age by swine intestine to effects ofenteropathogenic Escherichia coli. J. Infect. Dis.122:220-223.

112. Moon, H. W., S. C. Whipp, and S. M. Skarvedt. 1976.Etiologic diagnosis of diarrheal diseases of calves: fre-quency and methods for detecting enterotoxin and K99antigen production by Escherichia coli. Am. J. Vet. Res.37:1025-1029.

113. Morgan, R. L., R. E. Isaacson, H. W. Moon, C. C.Brinton, and C.-C. To. 1978. Immunization of sucklingpigs against enterotoxigenic Escherichia coli-induceddiarrheal disease by vaccinating dams with purified 987or K99 pili: protection correlates with pilus homology ofvaccine and challenge. Infect. Immun. 22:771-777.

114. Morris, J. A., A. E. Stevens, and W. J. Sojka. 1977.Preliminary characterization of cell-free K99 antigenisolated from Escherichia coli B41 J. Gen. Microbiol.99:353-357.

115. Morris, J. A., A. E. Stevens, and W. J. Sojka. 1978.Anionic and cationic components of the K99 surfaceantigen from Escherichia coli B41. J. Gen. Microbiol.107:173-175.

116. Morris, J. A., A. E. Stevens, and W. J. Sojka. 1978.Isoelectric point of cell-free K99 antigen exhibiting hem-agglutinating properties. Infect. Immun. 19:1097-1098.

117. Morris, J. A., C. Thorns, A. C. Scott, W. J. Sojka, andG. A. Wells. 1982. Adhesion in vitro and in vivo associat-ed with an adhesive antigen (F41) produced by a K99-mutant of the reference strain Escherichia coli B41.Infect. Immun. 36:1146-1153.

118. Morris, J. A., C. J. Thorns, and W. J. Sojka. 1980.Evidence for two adhesive antigens on the K99 referencestrain Escherichia coli B41. J. Gen. Microbiol. 118:107-113.

119. Morris, J. A., C. Wray, and W. J. Sojka. 1980. Passiveprotection of lambs against enteropathogenic Escherich-ia coli: role of antibodies in serum and colostrum of damsvaccinated with K99 antigen. J. Med. Microbiol. 13:265-271.

120. Myers, L. L. 1978. Enteric colibacillosis in calves: immu-nogenicity and antigenicity of Escherichia coli antigens.Am. J. Vet. Res. 39:761-765.

121. Myers, L. L., and P. A. M. Guinee. 1976. Occurrence andcharacteristics of enterotoxigenic Escherichia coli isolat-ed from calves with diarrhea. Infect. Immun. 13:1117-1119.

122. Nagy, B. 1980. Vaccination of cows with a K99 extract toprotect newborn calves against experimental enterotoxiccolibacillosis. Infect. Immun. 27:21-24.

123. Nagy, B., H. W. Moon, and R. E. Isaacson. 1976.Colonization of porcine small intestine by Escherichiacoli: ileal colonization and adhesion by pig enteropatho-gens that lack K88 antigen and by some acapsularmutants. Infect. Immun. 13:1214-1220.

124. Nagy, B., H. W. Moon, and R. E. Isaacson. 1977.Colonization of porcine intestine by enterotoxigenic

Escherichia coli: selection of piliated forms in vivo,adhesion of piliated forms to epithelial cells in vitro, andincidence of a pilus antigen among porcine enteropatho-genic E. coli. Infect. Immun. 16:344-352.

125. Nagy, B., H. W. Moon, R. E. Isaacson, C.-C. To, andC. C. Brinton. 1978. Immunization of suckling pigsagainst enteric enterotoxigenic Escherichia coli infectionby vaccinating dams with purified pili. Infect. Immun.21:269-274.

126. Newman, F S., L. L. Myers, B. D. Firehammer, and J. E.Catlin. 1973. Prevention of experimentally induced en-teric colibacillosis in newborn calves. Infect. Immun.8:540-543.

127. Ofek, I., and E. H. Beachey. 1978. Mannose binding andepithelial cell adherence of Escherichia coli. Infect.Immun. 22:247-254.

128. Ofek, I., and E. H. Beachey. 1980. General concepts andprinciples of bacterial adherence in animal and man, p.1-29. In E. H. Beachey (ed.), Bacterial adherence,receptors and recognition, series B, vol. 6. Chapman andHall, London.

129. 0rskov, I., and F. 0rskov. 1966. Episome-carried sur-face antigen K88 of Escherichia coli. I. Transmission ofthe determinant of the K88 antigen and influence on thetransfer of chromosomal markers. J. Bacteriol. 91:69-75.

130. 0rskov, I., and F. 0rskov. 1977. Special O:K:H sero-types among enterotoxigenic E. coli strains from diar-rhea in adults and children. Occurrence of the CF(colonization factor) antigen and of hemagglutinatingabilities. Med. Microbiol. Immunol. 163:99-110.

131. 0rskov, F., I. 0rskov, D. J. Evans, Jr., R. B. Sack, D. A.Sack, and T. Wadstrom. 1976. Special Escherichia coliserotypes among enterotoxigenic strains from diarrhoeain adults and children. Med. Microbiol. Immunol.162:73-80.

132. 0rskov, I., F. 0rskov, H. W. Smith, and W. J. Sojka.1975. The establishment of K99, a thermolabile, trans-missible Escherichia coli K antigen, previously called"Kco", possessed by calf and lamb enteropathogenicstrains. Acta Pathol. Microbiol. Scand. Sect. B 83:31-36.

133. 0rskov, I., F. 0rskov, W. J. Sojka, and J. M. Leach.1%1. Simultaneous occurrence of E. coli B and Lantigens in strains from diseased swine. Acta Pathol.Microbiol. Scand. Sect. B. 53:404-422.

134. 0rskov, I., F. 0rskov, W. J. Sojka, and W. Wittig. 1964.K. antigens K88ab (L) and K88ac (L) in E. coli. A new 0antigen: 0147 and a new K antigen: K89 (B). ActaPathol. Microbiol. Scand. Sect. B. 62:439-447.

135. 0rskov, I., F. 0rskov, W. Wittig, and E. J. Sweeney.1969. A new E. coli serotype 0149:K91(B),K88ac(L):H1O isolated from diseased swine. ActaPathol. Microbiol. Scand. 75:491-498.

136. Ottow, J. C. G. 1975. Ecology, physiology, and geneticsof fimbriae and pili. Annu. Rev. Microbiol. 29:79-108.

137. Parry, S. H., and P. Porter. 1978. Immunological aspectsof cell membrane adhesion demonstrated by porcineenteropatogenic Escherichia coli. Immunology 34:41-49.

138. Penaranda, M. E., M. B. Mann, D. G. Evans, and D. J.Evans, Jr. 1980. Transfer of an ST:LT:CFAIII plasmidinto Escherichia coli K-12 strain RR1 by cotransforma-tion with PSC301 plasmid DNA. FEMS Microbiol. Lett.8:251-254.

139. Porter, P., M. A. Linggood, and J. Chidlow. 1978.Elimination of Escherichia coli K88 adhesion determi-nant by antibody in porcine gut and mammary secretionsfollowing oral immunization. Adv. Exp. Med. Biol.107:133-141.

140. Reis, M. H. L., M. Helioza, T. Affonso, L. R. Trabulsi,A. J. Mazaitis, R. Maas, and W. K. Maas. 1980. Transferof a CFA/I-ST plasmid promoted by a conjugative plas-mid in a strain of Escherichia coli of serotype 0128ac:H12. Infect. Immun. 29:140-143.

141. Reis, M. H. L., D. P. Matos, A. F. Pestana de Castro,M. R. F. Toledo, and L. R. Trabulsi. 1980. Relationshipamong enterotoxigenic phenotypes, serotypes and

159VOL. 46, 1982


mbr.asm

.org/D

ownloaded from



sources of strains in enterotoxigenic Escherichia coli.Infect. Immun. 28:24-27.

142. Runnels, P. L., H. W. Moon, and R. A. Schneider. 1980.Development of resistance with host age to adhesion ofK99+ Escherichia coli to isolated intestinal epithelialcells. Infect. Immun. 28:298-300.

143. Rutter, J. M., and J. C. Anderson. 1972. Experimentalneonatal diarrhoea caused by an enteropathogenic strainof Escherichia coli in piglets: a study of the disease andthe effect of vaccinating the dam. J. Med. Microbiol.5:197-210.

144. Rutter, J. M., M. R. Burrows, R. Sellwood, and R. A.Gibbons. 1975. A genetic basis for resistance to entericdisease caused by E. coli. Nature (London) 257:135-136.

145. Rutter, J. M., and G. W. Jones. 1973. Protection againstenteric disease caused by Escherichia coli-a model forvaccination with a virulence determinant? Nature (Lon-don) 242:531-533.

146. Rutter, J. M., G. W. Jones, G. T. H. Brown, M. R.Burrows, and P. D. Luther. 1976. Antibacterial activityin colostrum and milk associated with protection ofpiglets against enteric disease caused by K88-positiveEscherichia coli. Infect. Immun. 13:667-676.

147. Rutter, P. R., and B. Vincent. 1980. The adhesion ofmicro-organisms to surfaces physico-chemical aspects,p. 79-92.In R. C. W. Berkeley, J. M. Lynch, J. Melling,P. R. Rutter, and B. Vincent (ed.), Microbial adhesion tosurfaces. Ellis Horwood Ltd., Chichester, England.

148. Sack, R. B. 1975. Human diarrheal disease caused byenterotoxigenic Escherichia coli. Annu. Rev. Microbiol.29:333-353.

149. Sack, R. B. 1980. Enterotoxigenic Escherichia coli:identification and characterization. J. Infect. Dis.142:279-286.

150. Salit,I. E., and E. C. Gotschlich. 1977. Hemagglutinationby purified type 1 Escherichia coli pili. J. Exp. Med.146:1169-1181.

151. Schmitt, R., R. Mattes, K. Schmid, and J. Altenbuchner.1979. Raf. plasmids in strains of Escherichia coli andtheir possible role in enteropathogeny, p. 199-210. InK. N. Timmis and A. Puhler (ed.), Plasmids of medical,environmental and commercial importance. Elsevier/North-Holland Biomedical Press, Amsterdam.

152. Sellwood, R. 1980. The interaction of the K88 antigenwith porcine epithelial cell brush borders. Biochim.Biophys. Acta 632:326-335.

153. Sellwood, R. 1982. Escherichia coli-associated porcineneonatal diarrhea: antibacterial activities of colostrumfrom genetically susceptible and resistant sows. Infect.Immun. 35:396-401.

154. Sellwood, R., R. A. Gibbons, G. W. Jones, and J. M.Rutter. 1975. Adhesion of enteropathogenic Escherichiacoli to pig intestinal brushborders: the existence of twopig phenotypes. J. Med. Microbiol. 8:405-411.

155. Shipley, P. L., W. S. Dallas, G. Dougan, and S. Falkow.1979. Expression of plasmid genes in pathogenic bacte-ria, p. 176-180. In D. Schlessinger (ed.), Microbiology-1979. American Society for Microbiology, Washington,D.C.

156. Shipley, P. L., G. Dougan, and S. Falkow. 1981. Identifi-cation and cloning of the genetic determinant that en-codes for the K88ac adherence antigen. J. Bacteriol.145:920-925.

157. Shipley, P. L., C. L. Gyles, and S. Falkow. 1978. Charac-terization of plasmids that encode for the K88 coloniza-tion antigen. Infect. Immun. 20:559-566.

158. Sivaswamy, G., and C. L. Gyles. 1976. Characterizationof enterotoxigenic bovine Escherichia coli. Can J.Comp. Med. 40:247-251.

159. Smith, H. W. 1972. The nature of the protective effect ofantisera against Escherichia coli diarrhoea in piglets. J.Med. Microbiol. 5:345-353.

160. Smith, H. R., A. Cravioto, G. A. Wilishaw, M. M.McConnell, S. M. Scotland, R. J. Gross, and B. Rowe.1979. A plasmid coding for the production of colonisa-

tion factor antigen I and heat-stable enterotoxin instrains of Escherichia coli of serogroup 078. FEMSMicrobiol. Lett. 6:255-260.

161. Smith, H. W., and C. L. Gyles. 1970. The relationshipbetween two apparently different enterotoxins producedby enteropathogenic strains of Escherichia coli of por-cine origin. J. Med. Microbiol. 3:387-401.

162. Smith, H. W., and S. Halls. 1967. Observations by theligated intestinal segment and oral inoculation methodsin Escherichia coli infections in pigs, calves, lambs andrabbits. J. Pathol. Bacteriol. 93:499-529.

163. Smith, H. W., and S. Halls. 1968. The transmissiblenature of the genetic factor in Escherichia coli thatcontrols enterotoxin production. J. Gen. Microbiol.52:319-334.

164. Smith, H. W., and M. B. Huggins. 1978. The influence ofplasmid-determined and other characteristics of entero-pathogenic Escherichia coli on their ability to proliferatein the alimentary tracts of piglets, calves and lambs. J.Med. Microbiol. 11:471-492.

165. Smith, H. W., and M. A. Linggood. 1971. Observationson the pathogenic properties of the K88, Hly and Entplasmids of Escherichia coli with particular reference toporcine diarrhoea. J. Med. Microbiol. 4:467-485.

166. Smith, H. W., and M. A. Linggood. 1972. Furtherobservations on Escherichia coli enterotoxins with par-ticular regard to those produced by atypical piglet strainsand by calf and lamb strains: the transmissible nature ofthese enterotoxins and of a K antigen possessed by calfand lamb strains. J. Med. Microbiol. 5:243-250.

167. Smith, H. W., and Z. Parsell. 1975. Transmissible sub-strate-utilizing ability in enterobacteria. J. Gen. Micro-biol. 87:129-140.

168. Smith, H. R., G. A. Willshaw, and B. Rowe. 1982.Mapping of a plasmid coding for colonization factorantigen I and heat-stable enterotoxin production isolatedfrom an enterotoxigenic strain of Escherichia coli. J.Bacteriol. 149:264-275.

169. Smyth, C. J., P. Jonsson, E. Olsson, 0. So6derlind, J.Rosengren, S. Hjertin, and T. Wadstrom. 1978. Differ-ences in hydrophobic surface characteristics of porcineenteropathogenic Escherichia coli with or without K88antigen as revealed by hydrophobic interaction chroma-tography. Infect. Immun. 22:462-472.

170. Smyth, C. J., B. Kaijzer, E. Back, A. Faris, R. Mollby, 0.Soderlind, G. Stintzing, T. Wadstrom, and D. Habte.1979. Occurrence of adhesins causing mannose-resistanthaemagglutination of bovine erythrocytes in enterotoxi-genic Escherichia coli. FEMS Microbiol. Lett. 5:85-90.

171. So, M., H. W. Boyer, M. Betlach, and S. Falkow. 1976.Molecular cloning of an Escherichia coli plasmid deter-minant that encodes for the production of heat-stableenterotoxin. J. Bacteriol. 128:463-472.

172. Soderlind, O., and R. Mollby. 1978. Studies on Esche-richia coli in pigs. V. Determination of enterotoxicityand frequency of 0 group and K88 antigen in strainsfrom 200 piglets with neonatal diarrhoea. Zentralbl.Veterinaermed. Reihe B. 25:719-728.

173. So6derlind, O., and R. Mollby. 1979. Enterotoxins, 0-groups and K88 antigen in Escherichia coli from neonatalpigs with and without diarrhoea. Infect. Immun. 24:611-616.

174. Sojka, W. J. 1965. Escherichia coli in domestic animalsand poultry. Commonwealth Agricultural Bureau, Wey-bridge, England.

175. Sojka, W. J. 1971. Enteric diseases in new-born piglets,calves and lambs due to Escherichia coli infection. Vet.Bull. 41:509-522.

176. Sojka, W. J. 1973. Enteropathogenic Escherichia coli inman and farm animals. Can. Inst. Food Sci. Technol. J.6:52-63.

177. Sojka, W. J., C. Wray, and J. A. Morris. 1978. Passiveprotection of lambs against experimental enteric colibac-cillosis by colostral transfer of antibodies from K99-vaccinated ewes. J. Med. Microbiol. 11:493-499.

MICROBIOL. REV.


mbr.asm

.org/D

ownloaded from



178. Stirm, S., I. 0rskov, and F. 0rskov. 1966. K88, anepisome-determined protein antigen of Escherichia coli.Nature (London) 209:507-508.

179. Stirm, S., F. 0rskov, I. 0rskov, and A. Birch-Andersen.1967. Episome-carried surface antigen K88 of Escherich-ia coli. III. Morphology. J. Bacteriol. 93:740-748.

180. Stirm, S., F. 0rskov, I. 0rskov, and B. Mansa. 1967.Episome-carried surface antigen K88 of Escherichia coli.II. Isolation and chemical analysis. J. Bacteriol. 93:731-739.

181. Swaney, L. M., Y.-P. Liu, C.-M. To, C.-C. To, K. Ippen-Ihier, and C. C. Brinton, Jr. 1977. Isolation and charac-terization of Escherichia coli phase variants and mutantsdeficient in type 1 pilus production. J. Bacteriol.130:495-505.

182. Thorne, G. M., C. F. Deneke, and S. L. Gorbach. 1979.Hemagglutination and adhesiveness of toxigenic Esche-richia coli isolated from humans. Infect. Immun. 23:690-699.

183. Tixier, G., and P. Gouet. 1975. Mise en evidence d'unestructure agglutinant les hematies de cheval en presencede mannose et specifique des souches d' E. coli entero-toxiques d'origine bovine. C.R. Acad. Sci. Paris281:1641-1644.

184. Van Doorn, J., B. Oudega, F. R. Mooi, and F. K.de Graaf. 1982. Subcellular localization of polypeptidesinvolved in the biosynthesis of K88 ab fimbriae. FEMSMicrobiol. Lett. 13:99-104.

185. Van Embden, J. D. A., F. K. de Graaf, L. M. Schouls, andJ. S. Teppema. 1980. Cloning and expression of a deoxy-ribonucleic acid fragment that encodes for the adhesiveantigen K99. Infect. Immun. 29:1125-1133.

186. Wachsmuth, K., J. Wells, P. Shipley, and S. Falkow.1979. Heat-labile enterotoxin production in isolates froma shipboard outbreak of human diarrheal illness. Infect.Immun. 24:793-797.

187. Wadstrom, T., A. Faris, J. Freer, D. Habte, D. Hallberg,and A. Ljungh. 1980. Hydrophobic surface properties ofenterotoxigenic E. coli (ETEC) with different coloniza-

tion factors (CFA/I, CFA/II, K88 and K99) and attach-ment to intestinal epithelial cells. Scand. J. Infect. Dis(Suppl.) 24:148-153.

188. Wadstrom, T., C. J. Smyth, A. Faris, P. Johnsson, andJ. H. Freer. 1979. Hydrophobic adsorptive and hemag-glutinating properties of enterotoxigenic Escherichia coliwith different colonizing factors: K88, K99, and coloni-zation factor antigens and adherence factor, p. 29-55. InProceedings of the International Symposium on Neona-tal Diarrhea. Stuart Brandle Publishing Service, Edmon-ton, Canada.

189. Walker, P. D., and L. K. Nagy. 1980. Adhesion oforganisms of animal tissues, p. 473-494. In R. C. W.Berkely, J. M. Lynch, J. Melling, P. R. Rutter, and B.Vincent (ed.), Microbial adhesion to surfaces. Ellis Hor-wood Ltd., Chichester, England.

190. Weil, E., and A. Felix. 1918. Untersuchungen uber diegewohnlichen Proteusstamme und ihre Beziehungen zuden X-Stamme. Wien. Klin. Wochenschr. 23:637-639.

191. Weil, E., and A. Felix. 1918. Ueber die Doppelnatur derRezeptoren beim Paratyphus 1. Wien, Klin. Wo-chenschr. 36:986-988.

192. Wevers, P., R. Picken, G. Schmidt, B. Jann, K. Jann,J. R. Golecki, and M. Kist. 1980. Characterization of piliassociated with Escherichia coli 018ac. Infect. Immun.29:685-691.

193. Williams, P. H., N. Evans, P. Turner, R. H. George, andA. S. McNeish. 1977. Plasmid mediating mucosal adher-ence in human enteropathogenic Escherichia coli. Lan-cet i:1151.

194. Williams, P. H., M. I. Sedgwick, N. Evans, P. J. Turner,R. H. George, and A. S. McNeish. 1978. Adherence of anenteropathogenic strain of Escherichia coli to humanintestinal mucosa is mediated by a colicinogenic conjuga-tive plasmid. Infect. Immun. 22:393-402.

195. Wilson, M. R., and A. W. Hohmann. 1974. Immunity toEscherichia coli in pigs: adhesion of enteropathogenicEscherichia coli to isolated intestinal epithelial cells.Infect. Immun. 10:776-782.

161VOL. 46, 1982


mbr.asm

.org/D

ownloaded from


Host-Specific Fimbrial Adhesins ofNoninvasive ... · coli. Theiradhesiveproperties canhelpto...

Documents

Transcript of Host-Specific Fimbrial Adhesins ofNoninvasive ... · coli. Theiradhesiveproperties canhelpto...