Pergamon Comp. lmmun. Microbiol. infect. Dis. Vol. 18, No. 3, pp. 179-188, 1995

Copyright © 1995 Elsevier Science Ltd 0147-9571(94)00026-3 Printed in Great Britain. All rights reserved

0147-9571/95 $9.50 + 0.00

V I R U L E N C E F A C T O R S A S S O C I A T E D W I T H

S T R A I N S O F E S C H E R I C H I A C O L I F R O M C A S E S O F

S U D D E N I N F A N T D E A T H S Y N D R O M E (SIDS)

K. A. BETTELHEIM I*, B. J. CHANG 2, S. J. ELLIOTT 2, S. T. GUNZBURG 2 and J. L. PEARCE 3

t School of Agriculture, La Trobe University, Bundoora 3083, Australia, 2Department of Micro- biology, University of Western Australia, Nedlands, Western Australia 6009, Australia and

3Schools of Agriculture and Microbiology, La Trobe University, Bundoora 3083, Australia

(Received for publication 20 December 1994)

Abstract--Strains of Escherichia coli isolated from cases of Sudden Infant Death Syndrome, healthy infants and infants that died of other causes were subjected to a series of tests with particular reference to serotyping, toxigenicity and adherence factors. E. coli from SIDS infants tended to have a low hydrophobicity and high toxigenicity, compared to those from healthy infants, while no notable differences in haemagglutination patterns were observed between these two groups of strains.

Key words: E. coli, Sudden Infant Death Syndrome, toxins, adhesins, hydrophobicity.

Rrsmnr---Des souches d'Escherichia coli isolres de cas de syndrome de morts subites du nourrisson, d'enfants morts d'autres causes, et d'enfants en bonne santr, ont 6t6 soumises fi une srrie de tests intrressant notamment, le serotypage, le pouvoir toxique, et les facteurs d'adhrrence. Les souches de E. coli isolres de syndrome de mort subite du nourrisson pr~sentaient un faible pouvoir hydrophobe et un pouvoir toxique ~levr, compar6 fi celles isolres d'enfants en bonne santr. En revanche, il n'a pas 6t6 observ6 de diffrrences dans le pouvoir hrmagglutinant des souches appartenant aux deux groupes.

Mots-clefs: E. coil, syndrome de mort subite du nourrisson, toxines, adhesine, hydrophobicit6.

INTRODUCTION

A number of Escherichia coli virulence factors have been described recently. These have included various toxins and adhesins. As part of the current investigations into the possible roles of toxigenic E. coli in Sudden Infant Death Syndrome (SIDS) [1-3] it was decided to examine selected strains for the presence of some of these virulence factors or virulence-associated factors. Specifically important for this study were the ones associated with the ability of E. coli to adhere to the intestinal epithelial cells. Such adherence mechanisms are important for diarrhoeagenic E. coli, and may also play a role in other toxin-mediated E. coli infections. The ability to adhere would give the organisms a selective advantage within the intestine and thus permit them to colonize. Should such organisms be toxigenic then a far greater amount of toxin would be produced in near proximity to

*Author for correspondence.

179

180 K . A . Bettelheim et al.

the mucosal surface of the intestine and may therefore be absorbed. Of the various adherence mechanisms which have been investigated, there are factors giving characteristic adherence to HEp-2 cells such as localized (LA), diffuse (DA) [4] and aggregative (EAgg) [5] adherence. In addition the agglutination of erythrocytes of different animal species has been associated with virulence [6]. Both mannose sensitive and resistant haemagglutination has been investigated and patterns of haemagglutination have been associated with the colonization factor antigens and other fimbrial types. Similarly the level of hydrophobicity [7] has been linked to the presence of fimbriae on the bacterial cell surfaces. In this study the prevalence of these factors was investigated for "O" and " H " serotyped strains of E. coli isolated from infants, who had died from SIDS, infants who had died from other causes and healthy infants. These strains have also been tested for their effect on Vero-cells, for the heat-labile enterotoxin (LT), and for the recently described cell detachment (CD) phenomenon [8], which has been associated with ~-haemolysin.

MATERIALS AND METHODS

The in]ants

This investigation is part of a continuing series of studies on toxigenic E. coli associated with SIDS [1-3]. Typical E. coli-like strains from post mortem intestinal contents isolated on MacConkey agar from cases of SIDS, as well as other dead infants, in Victoria, Australia were studied [2]. The cause of death was determined only after a full independent investigation including a complete necropsy, which followed a standard protocol. Faecal specimens were also collected from healthy infants attending an infant welfare centre in Victoria and were similarly cultured. The period in which these strains were collected ranged from early 1989 to mid 1990.

The mean ages of the infants, whose specimens were used in this investigation were 3.6, 3.9 and 9.3 months respectively for the healthy infants, SIDS cases and infants who died of other causes. While the age range for the majority of the SIDS cases and healthy infants was relatively narrow, those dying of other causes included both very young infants aged under 1 month as well as some over 1 year old.

The E. coli strains

From each MacConkey plate as many morphologically distinct E. coli colonies as possible were selected. They were characterized biochemically as E. coli, serotyped, toxin tested and further tested as described previously [3]. They were O and H serotyped according to the methods of Chandler and Bettelheim [9] and Bettelheim and Thompson [10]. They were tested for their ability to produce the heat labile enterotoxin (LT) by both the use of murine Y-I adrenal cells [11] and by the Phadebact ETEC-LT method [12]. For the determination of Vero-cell cytotoxins the strains of E. coli were grown overnight (approx 18 h) in 5 ml tryptone soy broth (TSB) (Oxoid) in 50 ml Erlenmeyer flasks, incubated at 37°C and shaken at 100 rpm. The cultures were centrifuged at 2500g for 20 min. Culture supernatants were collected, filtered through 0.22 #m membrane filter and stored at 4°C until assayed.

A 1-day-old monolayer of Vero-cells in Minimal Essential Medium (Eagles, MEM) with 5% foetal bovine serum was obtained in the wells of microtitre trays. The supernatant MEM was removed and replaced by 200 #t fresh MEM and 50/Jl of culture supernatant prepared as described above was added. The trays were incubated at 37°C in 5% CO2 and

Escherichia coli and SIDS 181

observed over 4 days by means of an inverted microscope for characteristic rounding of the cells. This assay is also able to detect the cytotoxic necrotizing factors.

A group of 262 random strains, which had been tested as described above were subjected to the adherence-related virulence tests under code. A number of isolates from the same infant, which had the same serotype and toxin characteristics, were deliberately included and generally gave the same results in the virulence tests described below. A total of 160 strains were selected from these such that, where strains of identical phenotype were isolated from the same infant, only one was studied further. Where infants carried a variety of phenotypically diverse strains, the one carrying the greatest number of the virulence factors studied in this investigation, was selected. This selection of 160 strains, made for this investigation comprised 76 from healthy infants, 64 from SIDS cases and 20 from infants who died from other causes. Hence each infant is represented by one strain.

Adherence-related virulence tests

Haemagglutination tests. These were performed following the techniques of Evans et al. [6, 13]. Each strain was tested for MRSA and MSHA of human, bovine, chicken and guinea pig erythrocytes.

Hydrophobicity (salt aggregation) tests. The method of Lindahl et al. [7] was used. Results were scored 0-6 as follows:

(NH4)2S04 concentration giving agglutination Score >4M (No agglutination) 0

4 M 1 2 M 2 1M 3 0 .5M 4 0.25 M 5 0.125 M 6

Adherence assays. These were performed as described by Levine et al. [14], based on the methods of Cravioto et al. [ 15]. E. coli were added to HEp-2 cell monolayers in the presence 0.5% methyl-~-D-mannoside, incubated for 3 h at 37°C and stained with Giemsa. Patterns of adherence to the cells were observed microscopically and classified into LA, DA and EAgg. In addition, where cell detachment was observed this was separately noted.

RESULTS

The strains included all the common "O" serogroups which had been found in earlier investigations with groups O1, 02, 04, 06, 025 and 075 being more prominent as shown (Table 1). The most frequent serotypes are listed in Table 2, while Table 3 lists the HEp-2 cell reactions of the strains. The distribution of the haemagglutination reactions is given in Table 4. Only two strains were found to conform to the HA types associated with CFA I or CFA II described by Evans et al. [6]. They had both been isolated from the same infant, which had died of congenital heart disease and had the same serotype O90.H18. While only one strain per baby was generally included, as these two differed they are mentioned separately. One strain conformed to CFA type I by producing mannose resistant haemagglutination with human, bovine and chicken erythrocytes, while the other strain

182 K. A. Bettelheim et al.

Table 1. Common O serogroups of strains studied

No. of strains studied from

Healthy Infants who died Infants who All Common infants of other causes died of SIDS infants 0 serogroups (%) (%) (%) (%)

OI 5 (6.6) 0 (0) 5 (7.8) 10 (6.3) 0 2 7 (9.2) 1 (5) 6 (9.4) 14 (8.8) 0 4 10 (13.2) 0 (0) 3 (4.7) 13 (8.1) 0 6 8 (10.5) 1 (5) 5 (7.8) 14 (8.8) 0 2 5 3 (3.9) I (5) 7 (10.9) 11 (6.9) 0 7 5 2 (2.6) I (5) 2 (3.1) 5 (3.1) Other 41 (53.9) 16 (80) 36 (56.3) 93 (58.1) Total 76 20 64 160

Table 2. Common OH serotypes of strains studied

No. o f strains studied from

C o m m o n Healthy Infants who died Infants who All O H serotypes infants of other causes died of SIDS infants

O I . H 7 3 0 4 7 O2 .HI 2 I 1 4 O2 .H4 2 0 1 3 O4 .H5 4 0 2 6 O6.H- 4 0 3 7 O 6 H 3 1 4 I 0 5 O25.H I I I 6 8 O75 .H- I I 1 3 Total 21 4 18 43

only produced mannose resistant haemaggtutination with bovine and chicken erythrocytes (CFA II). Neither strain produced LT. The other haemagglutination patterns included those described by Evans et al. [6], as associated with enteropathogenic E. coli (EPEC), extraintestinal infections, and normal stool. As the scheme of Evans et al. [6] included monkey erythrocytes, which were not included in this study, further subtyping using their scheme cannot be done. The distribution of hydrophobicity among the strains is given in Table 5. As both haemagglutination and hydrophobicity examine bacterial surface characteristics, a comparison of the two sets of results appeared warranted. The mean hydrophobicity index of non-haemagglutinating strains from SIDS cases and healthy infants was 1.7 and 2.0 respectively. For those where the presence of type I pili was indicated it was 2.2 and 4.1 and for those giving mannose-resistant haemagglutination with

Table 3. HeLa cell adherence of strains studied

No. of strains studied from

Type of Healthy Infants who died Infants who All adherence or infants of other causes died of S1DS infants cell de tachment (%) (%) (%) (%)

A A 8 (10.5) I (5) 8 (12.5) 17 (10.6) DA 20 (26.3) 8 (40) 20 (31.3) 48 (30) LA 0 (0) 0 (0) 2 (3.1) 2 (1.3) ID* I (I.3) 0 (0) 3 (4.7) 4 (2.5) N A t 22 (28.9) 7 (35) 19 (29.7) 48 (30) C D 25 (32.9) 4 (20) 12 (18.8) 41 (25.6) Total 76 20 64 160

*ID + , indistinguishable type of adherence. tNA, no adherence.

Escherichia coli and SIDS

Table 4. Haemagglutination reactions of strains studied

Haemagglutination reaction cell-type

No. of strains studied from

Healthy Infants who died Infants who All infants of other causes died of SIDS infants

Human Mannose-R Mannose-S

Bovine Mannose-R Mannose-S

Chicken Mannose-R Mannose-S

Guinea Pig Mannose-R Mannose-S

24 3 15 42 2 0 1 3

0 3 I 4 0 0 0 0

0 1 0 1 40 11 28 79

0 0 0 0 40 12 28 80

Table 5. Hydrophobicity of strains studied

No. of strains studied from

Healthy Infants who died Infants who All Hydrophobicity infants of other causes died of SIDS infants index (%) (%) (%) (%)

0 12 (15.8) 5 (25) 27 (42.2) 44 (27.5) 1 7 (9.2) 0 8 (12.5) 15 (9.4) 2 3 (3.9) 2 (10) 6 (9.4) I1 (6.9) 3 7 (9.2) 5 (25) 3 (4.7) 15 (9.4) 4 16 (21.1) 2 (10) 8 (12.5) 26 (16.3) 5 3 (3.9) 0 4 (6.3) 7 (4.4) 6 28 (36.8) 6 (30) 8 (12.5) 42 (26.3) Total 76 20 64 160

183

human erythrocytes it was 1.3 and 3.9 respectively. No clear differences in hydrophobicity were observed with the other haemagglutination patterns.

The LT and Vero-toxin results are given in Table 6. Only two LT-producing strains were identified, both were SIDS isolates. A rounding of the Vero-cells with some disruption of the monolayer was observed with many strains (Table 6). A positive effect was considered to have occurred if /> 50% of the cells were affected.

As verocytotoxicity and hydrophobicity were more closely associated with SIDS, these properties were compared (Table 7).

DISCUSSION

The predominant serogroups were those most commonly associated with urinary tract infections (UTI) with the possible exception of O25 (Table 1). Such serogroups may have a greater potential for extra-intestinal infections and it is significant that they also seem

Table 6. Toxigenicity of E. coli strains studied

Heat-labile Vero-cell enterotoxin cytotoxin

Strain Total source No. (%) No. (%) no.

Healthy infant 0 (0) 3 (3.9) 76 Other infant 0 (0) 5 (25.0) 20 SIDS infant 2 (3.2) 33 (51.6) 64

184 K. A. Bettelheim et al.

Table 7. Comparison of the verocytoxicity and hydrophobicity of E. coli studied

Infant No. (%) E. coli No. (%) E. coli from which giving rounding not giving rounding E. coli isolated Hydrophobicity of Verocells of Verocells

SIDS Low (index 0-3) 26 (40.6) 18 (28.1) High (index 4-6) 8 (12.5) 12 (18.8) Total 34 (53.1) 30 (46.9)

Other infants Low (index 0 3) 6 (30.0) 6 (30.0) High (index 4 6) 2 (10.0) 6 (30.0) Total 8 (40.0) 12 (60.0)

Healthy infants Low (index 0-3) 5 (6.6) 24 (31.6) High (index 4-6) 4 (5.3) 43 (56.6) Total 9 (11.9) 67 (88.2)

All infants Low (index 0 3) 37 (23.1) 48 (30.0) High (index 4-61 14 (8.8) 61 (38.1) Total 51 (31.9) 109 (68.1)

to be more prominent in this collection. Of these, O1, 02 and 075 appear to be evenly distributed among both healthy infants and cases of SIDS. While 025 strains appear more prevalent among the SIDS isolates, the 04 strains, and to a lesser extent the 06, appear more prevalent amongst the healthy infants. This distribution may also be noted on closer inspection of the serotypes (Table 2), where O1.H7, O2.H1, O2.H4, O4.H5 and O75.H- are evenly distributed between healthy infants and SIDS cases. The bias towards healthy infants is maintained by strains of O6.H- and O6.H31, while the opposite appears to hold for O25.H1. Some virulence factors may be more commonly associated with certain clones, as identified by serotypes. The finding that O25.H1 isolates appear to be more strongly associated with SIDS cases, while the 025 serogroup in general is not amongst the most frequent UTI, may indicate that the O25.H1 strains may carry a separate set of virulence factors. This does not necessarily exclude other strains carrying some of these hypothetical virulence factors.

An examination of the adherence results (Table 3) shows that when the cell-detaching strains are removed from the analysis, there is an even distribution between HEp-2 adhering and non-adhering strains. When the distribution amongst the different categories of infants is examined, it is noted that the EAgg, DA and NA E. coli are relatively evenly distributed amongst the three groups of infants. It is not surprising that very few LA E. coli

were found as these have been generally associated with diarrhoeal disease [16]. The significance of DA and EAgg as virulence factors with respect to diarrhoeal disease are still under discussion so a possible role in extra-intestinal infection may be postulated and therefore a role for these virulence factors in SIDS cannot be excluded.

CD has been associated with diarrhoea [8]. In this study, CD was more common amongst isolates from healthy infants than SIDS infants and was not pursued further.

It is considered noteworthy that the majority of strains which gave mannose-sensitive agglutination of chicken erythrocytes gave similar agglutination with guinea pig erythro- cytes.

The ability of a bacterial pathogen to adhere to the surface of its potential host cell has been well recognized as a first stage in the infection process. With diarrhoeagenic E. coli a number of such factors have been described over the years, including LA, the K antigens K88 and K99 and the colonization factor antigens (CFA). These attachment processes are mediated by pili of various types. There have been a number of studies on the possible role of the adherence patterns. As an example Gomes et al. [16] found that the LA + E. coli,

Escherichia coli and SIDS 185

which were also EAF probe positive, were the most common enteropathogen detected in the faeces of children with diarrhoea in Brazil. In this study DA ÷ or EAgg + E. coli are also fairly commonly found. The roles, therefore, of these DA* or EAgg ÷ E. coil must be considered in relation to SIDS. While only two LA ÷ E. coli were found, which were from cases of SIDS is considered noteworthy, neither of these LA* E. coli belonged to EPEC serogroups. The O antigen of one was rough and the other was untypeable with the O sera available. It has been suggested that LA* E. coli should be considered EPEC regardless of serotype. However, there were some strains isolated, which belonged to O groups traditionally considered as belonging to the EPEC. The serotypes of these were O86.H27, O86.H30, Ol11.H25, Ol19.H2, O125.H6 and O125.H30. Three of those, O86.H27, O111 .H25 and O125.H30 were DA*. The O86.H30 strain was EAgg ÷ . A second strain of O 111 .H25 was NA as was the O 119.H2 strain. Most of these strains were isolated from SIDS cases except the O86.H30 and the O125.H6 strains which were isolated from healthy infants and the O119.H2 strain which was isolated from one of the other deaths. In addition two DA* E. coli of serotypes O78/O89.H16 and O78/O89.H51 and a NA strain of serotype O78.H- were isolated from SIDS cases. The 078 serogroup has been associated with pig scours. While a first impression may be that there is a slight tendency for the adherence characteristics to be more commonly associated with SIDS isolates (Table 3), this is not confirmed when the CD* strains, which would interfere with these studies, are removed from the calculation. The recalculated percentages of DA + strains from healthy infants and SIDS cases are now 30.0 and 34.9% respectively and for EAgg + strains they are 13.3 and 12.0% respectively. As EAggEC and DAEC are increasingly becoming associated with diarrhoeal disease [17, 18] it really is not surprising that no particular relationship with SIDS could be established.

The haemagglutination originally described by Evans et al. [6] was designed to identify the CFA of ETEC. These are fimbrial mannose resistant haemagglutinins and it had originally been considered as a useful means of identifying presumptive ETEC. However, it was noted that E. coli, isolated both from control individuals as well as from extraintestinal sources, produced mannose resistant haemagglutination (MRHA), even though they were negative for the presence of CFA/I. On the basis of these studies Evans et al. [6] were able to subdivide E. coli into different groups, which were related to their origin. As only two strains of the same serotype, both isolated from an infant who died of congenital heart disease, produced patterns indicative of CFA/I or CFA/II respectively, it can be fairly strongly suggested that these colonization factors do not play a significant role in SIDS. The serotypes associated with the strains found by Evans et al. [6] to be associated with these higher haemagglutination types included O1, 02, 04, 06, 025 and 075 which have been associated with both facultatively enteropathogenic E. coil (FEEC) as well as non-FEEC. These O groups have most frequently been associated with SIDS isolates. These serogroups are also most commonly associated with UTIs [19, 20].

Only two infants were identified as carrying LT producing E. coli and both were SIDS cases. The 3.2% carriage rate in this study and the similarly low rate previously reported [21, 22] precludes a conclusion that LT is a possible cause of SIDS. Mild diarrhoeal disease has been observed as frequently occurring about a week before the death of an infant due to SIDS and this finding may be a reflection of these epidemiological observations.

This study, in which a high proportion (51.6%) of SIDS infants carried verocytotoxi- genie E. coli compared to only 3.9% of healthy infants, confirms earlier findings [2] and the recent report [23] from Tasmania, Australia. A role for these toxins in SIDS cannot

186 K .A . Benelheim et al.

be excluded and must be taken into consideration in any discussion on the causality of SIDS.

Antibodies to both toxins and these adherence factors have been shown to be present in human milk and there have been a number of studies showing that breast feeding protects infants from diarrhoeal disease, including those caused by E. coli [24]. Lack of breast feeding is an established risk factor for SIDS. It is very likely that antibodies to the various virulence factors produced by E. coli, which may be linked to SIDS, may also be present in human milk.

The basis of the salt precipitation test is that the hydrophobicity of the surface of the bacteria will determine the level of salt concentration at which these bacteria agglutinate. The most hydrophobic cells will precipitate out in low concentrations of a salt such as ammonium sulphate, while more hydrophilic cells require higher concentrations of salt. Heavily piliated E. coil cells tend to be relatively hydrophobic [7] and it was shown that the hydrophobicity of ETEC strains decreased according to which types of pili they carried in the following order CFA/I > CFA/II > K88 ~ K99 > type I. Thus a low score on the salt agglutination test means that the cells are less hydrophobic and more hydrophilic than cells with a high score. In order to simplify the review of the results from Table 5, the groups of strains of hydrophobicity score 0-3 inclusive were combined, as were those of score 4-6 inclusive. By this simplification 29 strains (38%) from healthy infants have a low hydrophobicity index and 47 (62%) have a high index. Alternatively the reverse situation is true for the SIDS isolates where 44 strains (69%) have a low index compared to 20 strains (31%) which have a high index. The strains from the other deaths, with 60% with a low index compared to 40% with a high index, are similar to those from the SIDS infants. While the healthy infant strains are unevenly grouped, the strains from the other infant deaths are evenly distributed along all the values. When the median hydrophobicity score was calculated it was found to be "4" for the healthy infants, "3" for those who died of other causes and "1" for the SIDS cases. This unexpected result suggests that SIDS E. coli isolates may have unique surface features, which warrant further study.

An examination of the possible linkage between hydrophobicity and verocytotoxigenic- ity (Table 7) shows that 40.6% of SIDS isolates had both a low hydrophobicity and were verocytotoxigenic compared to only 18.8% which were neither. The strains from healthy infants show a virtually opposite situation with only 6.6% being both of low hydropho- bicity and being verocytotoxigenic, 56.6% are neither. Using Fisher's Exact Test it was found that the probability that low hydrophobicity and verocytotoxigenicity not being linked was 0.00044 when the results from all infants were examined, while significance was not obtained with the separate groups.

The hydrophobicity and haemagglutination tests were applied in order to provide an indication of the presence and type of pili. It was therefore surprising that the E. coil from SIDS infants tended to have a low hydrophobicity compared to those from healthy infants (Table 5), while no notable differences in haemagglutination patterns were observed between these two groups of strains (Table 4). As pili are generally considered to confer increased hydrophobicity on bacteria, non-haemagglutinating bacteria would be expected to have a relatively low hydrophobicity index. This was observed for E. coil from both SIDS and healthy infants. The haemagglutinating and therefore probably piliated E. coli from the healthy infants did indeed have a raised mean hydrophobicity index of around 4. However, the two most common groups of haemagglutinating SIDS isolates did not have such a high mean hydrophobicity index indicating that the hydrophilic nature of the

Escherichia coli and SIDS 187

SIDS E. coil may be due to some factor or factors other than the pili, which are involved in the haemagglutination. The effect of this factor (or these factors) may override the effect of the pili in the hydrophobicity test. It could be suggested that these SIDS isolates may have characteristics similar to the EPEC, which are hydrophilic due to the presence of neutral LPS. Law [25] has suggested that these hydrophilic properties of EPEC may be relevant in their pathogenicity. These studies further suggest that E. coli isolated from cases of SIDS may represent unique clones.

Acknowledgements--We wish to thank the parents who gave access to specimens, the Fairfield Maternal and Child Health Centre and the Victorian Institute of Forensic Pathology. We acknowledge the technical assistance of H. Evangelidis and D. Smith. Financial assistance from the Australian National SIDS Foundation and the South Australian SIDS Research Foundation are gratefully acknowledged.

R E F E R E N C E S

1. Bettelheim K. A., Dwyer B. W., Smith D. L., Goldwater P. N. and Bourne A. J. Toxigenic Escherichia coli associated with Sudden Infant Death Syndrome. Med. J. Aust. 151, 538 0989).

2. Bettelheim K. A., Goldwater P. N., Dwyer B. W., Bourne A. J. and Smith D. L. Toxigenic Escherichia coil associated with Sudden Infant Death Syndrome. Scand. J. infect. Dis. 22, 467-476 (1990).

3. Bettelheim K. A., Evangelidis H., Pearce J. L., Goldwater P. N. and Luke R. K. J. The isolation of cytotoxic necrotizing factor (CNF)-producing Escherichia coli from the intestinal contents of babies who died of Sudden Infant Death Syndrome (SIDS) and other causes as well as from the faeces of healthy babies. Comp. Immun. Microbiol. infect. Dis. 16, 87-90 (1993).

4. Scaletsky I. C. A., Silva M. L. M. and Trabulsi L. R. Distinctive patterns of adherence of enteropathogenic Escherichia coli to HeLa cells. Infect. lmmun. 45, 534-536 0984).

5. Nataro J. P., Kaper J. B., Robins-Browne R., Prado V., Vial P. A. and Levine M. M. Patterns of adherence of diarrheagenic Escherichia coil to HEp-2 cells. J. Pediatr. infect. Dis. 6, 892-931 (1987).

6. Evans D. J. Jr, Evans D. G., Young L. S. and Pitt J. Haemagglutination typing ofEscherichia coli: Definition of seven haemagglutination types. J. din. Microbiol. 12, 235-242 0980).

7. Lindahl M., Fails A., Wastr6m T. and Hjert6n S. A new test based on 'salting out' to measure relative surface hydrophobicity of bacterial cells. Biochim. Biophys. Acta 677, 471-476 (1981).

8. Gunzburg S. T., Chang B. J., Elliot S. J., Burke V. and Gracey M. Diffuse and enteroaggregative patterns of adherence of enteric Escherichia coli isolated from aboriginal children from the Kimberley Region of Western Australia. J. infect. Dis. 167, 755-758 (1993).

9. Chandler M. E. and Bettelheim K. A. A rapid method of identifying Escherichia coli "H" antigens. Zbl. Bakt. Hyg., 1. Abt. Orig. A 129, 74-79 (1974).

10. Bettelheim K. A. and Thompson C. J. New method of serotyping Escherichia coli: Implementation and verification. J. clin. Microbiol. 25, 781-786 (1987).

11. Bettelheim K. A., Wilson M. W., Shooter R. A. and O'Farrell S. M. Studies on the enterotoxigenicity of environmental Escherichia coli, belonging to serotypes normally considered enterotoxigenic. J. Hyg. Carnb. 84, 41 i-414 0980).

12. Bettelheim K. A., Gracey M. and Wadstr6m T. The use of the coagglutination test to determine whether Australian and New Zealand isolates of Escherichia coli produce the heat-labile enterotoxin. Zbl. Bakt. Hyg. A 260, 293-296 (1985).

13. Evans D. G., Evans D. J. Jr and Tjoa W. Hemagglutination of human group A erythrocytes by enterotoxigenic Escherichia coli isolated from adults with diarrhea: correlation with colonization factor. Infect. Immun. 18, 330-337 (1977).

14. Levine M. M., Prado V., Robins-Browne R., Lior H., Kaper J. B., Moseley S. L, Gicquelais K., Nataro J. P., Vial P. and Tall B. Use of DNA probes and HEp-2 cell adherence assay to detect diarrheagenic Escherichia coli. J. infect Dis. 158, 224-228 (1988).

15. Cravioto A., Gross R. J., Scotland S. M. and Rowe B. Adhesive factor found in strains of Escherichia coli belonging to the traditional infantile enteropathogenic serotypes. Curt. Microbiol. 3, 95-99 (1979).

16. Gomes T. A. T., Blake P. A. and Trabulsi L. R. Prevalence of Escherichia coli strains with localized, diffuse, and aggregative adherence to HeLa cells in infants with diarrhoea and matched controls. J. clin. Microbiol. 27, 266-269 (1989).

17. Brook M. G., Smith H. R., Bannister B. A., McConnell M., Chart H., Scotland S. M., Sawer A., Smith M. and Rowe B. Prospective study of verocytotoxin-producing, enteroaggregative and diffusely adherent Escherichia coli in different diarrhoeal states. Epidem. Infect. 112, 63~7 (1994).

188 K.A. Bettelheim et al.

18. Chan K. N., Phillips A. D., Knutton S., Smith H. R. and Walker-Smith J. A. Enteroaggregative Escherichia coli: Another cause of acute and chronic diarrhoea in England. J. Pediat. Gastroent. Nutrit. 18, 87-91 (1994).

19. Grueneberg R. N. and Bettelheim K. A. Geographical variation in serological types of urinary Escherichia coli. J. Med. Microbiol. 2, 219-224 (1969).

20. Brooks H. J. L., Benseman B. A., Peck J. and Bettelheim K. A. Correlation between uropathogenic properties of Escherichia coli from urinary tract infections and the antibody-coated bacteria test and comparison with faecal strains. J. Hyg. Camb. 87, 53~1 (1981).

21. Gurwith M. J., Langston C. and Citron D. M. Toxin-producing bacteria in infants. Am. J. Dis. Child. 135, 1104-1106 (1981).

22. Brunton J., Hinde D., Langston C., Gross R., Rowe B. and Gurwith M. Enterotoxigenic Escherichia coil in central Canada. J. clin. Microbiol. 11, 343 347 (1980).

23. Bettiol S. S., Radcliff F. J., Hunt A. L. C. and Goldsmid J. M. Bacterial flora of Tasmanian S1DS infants with special reference to pathogenic strains of Escherichia coli. Epidem. Infect. 112, 275-284 (1994).

24. Picketing L. K. and Morrow A. L. Factors in human milk that protect against diarrheal disease. Infection 21, 355 357 (1993).

25. Law D. Adhesion and its role in the virulence of enteropathogenic Escherichia coli. Clin. Microbiol. Re~,. 7, 152--173 (1994).