Microbiology (2003), 149, 3575–3586
DOI 10.1099/mic.0.26486-0
The distribution and genetic structure of
Escherichia coli in Australian vertebrates:
host and geographic effects
David M. Gordon1 and Ann Cowling2
School of Botany and Zoology1 and Statistical Consulting Unit2, Australian National University,
Canberra, ACT 0200, Australia
Correspondence
David Gordon
[email protected]
Received 14 May 2003
Revised
16 September 2003
Accepted 18 September 2003
Escherichia coli was isolated from more than 2300 non-domesticated vertebrate hosts living in
Australia. E. coli was most prevalent in mammals, less prevalent in birds and uncommon in fish,
frogs and reptiles. Mammals were unlikely to harbour E. coli if they lived in regions with a desert
climate and less likely to have E. coli if they lived in the tropics than if they lived in semi-arid or
temperate regions. In mammals, the likelihood of isolating E. coli from an individual depended
on the diet of the host and E. coli was less prevalent in carnivores than in herbivores or omnivores.
In both birds and mammals, the probability of isolating E. coli increased with the body mass of
the host. Hosts living in close proximity to human habitation were more likely to harbour E. coli
than hosts living away from people. The relative abundance of E. coli groups A, B1, B2 and
D strains in mammals depended on climate, host diet and body mass. Group A strains were
uncommon, but were isolated from both ectothermic and endothermic vertebrates. Group B1
strains could also be isolated from any vertebrate group, but were predominant in ectothermic
vertebrates, birds and carnivorous mammals. Group B2 strains were unlikely to be isolated from
ectotherms and were most abundant in omnivorous and herbivorous mammals. Group D strains
were rare in ectotherms and uncommon in endotherms, but were equally abundant in birds and
mammals. The results of this study suggest that, at the species level, the ecological niche of E. coli
is mammals with hindgut modifications to enable microbial fermentation, or in the absence of a
modified hindgut, E. coli can only establish a population in ‘large-bodied’ hosts. The non-random
distribution of E. coli genotypes among the different host groups indicates that strains of the
four E. coli groups may differ in their ecological niches and life-history characteristics.
INTRODUCTION
There is an extensive empirical and theoretical framework
regarding the ecological niche and life history characteristics
of plant and animal species. However, there have been few
attempts to characterize the ecological niche or life history
traits of micro-organisms. Even for relatively well known
groups of bacteria, such as the Enterobacteriaceae, there is
little consensus as to whether a particular species is animalassociated, plant-associated or free-living (Leclerc et al.,
2001).
Escherichia coli is considered to be a commensal of the lower
gastro-intestinal tract of mammals (Hartl & Dykhuizen,
1984; Selander et al., 1987). However, in a study of E. coli
isolated from non-domesticated Australian mammals,
Gordon & FitzGibbon (1999) showed that the probability
of isolating E. coli from an individual depended on the
taxonomic family of the host and on the geographic locality from which the host was collected. In an investigation
of the genetic structure of E. coli isolated from Australian
mammals using the technique of multi-locus enzyme
0002-6486 G 2003 SGM
electrophoresis, Gordon & Lee (1999) demonstrated that
host taxonomic family and geographic locality explained
a small but significant amount of the observed genetic
variation. However, there have been few studies that have
examined the distribution and genetic structure of E. coli
isolated from non-mammalian vertebrates.
We present the results of a survey that examined nonmammalian vertebrates and expanded the database available for mammals. This study is based on the examination
of faecal samples collected from over 2300 individuals
representing more than 350 species that encompass all
major vertebrate groups occurring in Australia.
The results of analyses investigating the host and geographic
factors that influence the probability of isolating E. coli
from a host individual are presented. In addition, the E. coli
strains isolated were assigned to one of the four groups of
E. coli described in the E. coli Reference (ECOR) Collection
(Ochman & Selander, 1984) using the recently developed
PCR-based method of Clermont et al. (2000). We show that
the same host and geographic factors that are significant
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 15 Jun 2017 18:26:16
Printed in Great Britain
3575
D. M. Gordon and A. Cowling
predictors of the presence of E. coli in a host species also
predict the distribution of strains of the different E. coli
groups among host species.
METHODS
Bacterial isolation and identification. Depending on the host
species being sampled, bacteria were isolated from freshly deposited
faeces or directly from faecal material in the rectum or cloacae of
the host. Faecal collection was carried out using a swab/transport
tube system containing Aimes transport medium. Bacteria were isolated by dilution-streaking the sample onto a MacConkey agar plate.
After incubation, a single colony with the appropriate morphology
was dilution-streaked onto a MacConkey agar plate and, after a
further round of incubation, dilution-streaked onto another
MacConkey agar plate, and then onto an LB agar plate. All lactosepositive isolates were tested for growth on a Simmons citrate agar
plate and for indole production. Freezer cultures were produced by
incubating the isolate overnight in LB broth, after which 1 ml of
the overnight culture was mixed with 30 ml glycerol and stored at
280 uC. All incubations were carried out at 35 uC.
Logistic regression was used to account for the presence/absence of
E. coli in each faecal sample in a nested analysis using the taxonomic
classification of the host (Subclass, Order, Family, Genus, Species) as
explanatory variables. To test the significance of the four highest
taxonomic levels, the ratio of the mean deviance of that level to the
mean deviance of the level below were compared to the F distribution
with the appropriate degrees of freedom. To test the significance of
Species, the usual x2 test of deviance was used.
Logistic regressions were fitted to presence/absence data for E. coli
using host diet, log10(body mass), climate where the host was collected
and Species as explanatory variables for mammals. Whilst for birds
the explanatory variables were human association, diet, body mass
and Species. In logistic regression, the order of fitting explanatory
variables affects their significance in the model. Therefore, variables
other than Species were fitted using a forward stepwise procedure:
the explanatory variable with the highest mean deviance was fitted
first, followed by the variable with the second highest mean deviance,
and so on. Species was fitted last. Interactions of the explanatory
variables, except Species, were also tested. The statistical significance of
each variable was tested using a x2 test for the change in deviance.
Predicted values were obtained using the final models, which included
only significant explanatory variables.
Bacterial genotyping. All strains with a phenotype consistent with
that of E. coli (Lac+ Cit2 Ind+) were genotyped. The recently developed method of Clermont et al. (2000) was used to assign the E. coli
isolates to one of the four main groups of E. coli identified in the
ECOR collection (Ochman & Selander 1984; Herzer et al., 1990).
The groups are designated A, B1, B2 and D. The Clermont method
is based on a multiplex PCR protocol that determines the presence
or absence of two genes (chuA and yjaA) and an anonymous DNA
fragment (TSPE4.C2). The presence/absence of the three PCR
products is used in the manner of a dichotomous key to assign an
unknown isolate to one of the ECOR groups. A fraction of group A
strains are negative for the three DNA products (Clermont et al.,
2000). Strains failing to produce any PCR products were repeated
twice more using newly prepared template DNA. After three
attempts, any strain negative for all three PCR products was identified using the BBL Enteric/Nonfermenter ID kit in conjunction with
the BBL crystal System Electronic code book. If the strain was identified as E. coli it was considered to be an ECOR A strain. The
presence of one or more PCR products of the correct size was
considered to be further evidence of the identity of an isolate as
E. coli.
For the samples from lizards and frogs, to test for differences between
species in the prevalence of E. coli, binomial general linear models
were fitted and the significance of the Species effect was tested.
Multinomial models were used to assess whether diet, climate and
body mass affected the distribution of E. coli groups (A, B1, B2, D) in
mammals. Significance was assessed using a x2 test for the change
in deviance.
RESULTS
Definition of prevalence
Statistical analyses. The study design was unbalanced for two
Prevalence is usually defined as the fraction of hosts in
which a particular species of parasite is present. The
isolation procedure used in this study has the ability to
detect E. coli only when it is abundant in a host. The
dilution streaking method used for the primary isolation
of bacteria in the faecal sample typically results in about
100 well isolated colonies growing on the MacConkey
agar plate. Consequently, E. coli will only be detected
if it represents, on average, more than 1 % of the total
number of bacterial cells capable of growth on this
medium. MacConkey medium is non-selective with respect
to members of the Enterobacteriaceae (Ewing, 1986), but it
inhibits the growth of Gram-positive and many Gramnegative species, to the extent that about 92 % of the
bacteria recovered from mammals were members of the
Enterobacteriaceae (Gordon & FitzGibbon, 1999). Relative
to other members of the Enterobacteriaceae, E. coli has a
distinct colony morphology and colour on MacConkey
agar and consequently is unlikely to be overlooked. Finally,
one person processed over 95 % of the samples, so that
variability among observers was not a factor.
reasons. First, several families and many genera do not occur in
every climate region of Australia (Strahan, 1995). Second, the study
samples were acquired in an ad hoc manner so that the number of
samples from each species varied considerably.
Thus, in this study prevalence of E. coli is defined as the
fraction of hosts in which E. coli was a dominant member
of the Enterobacteriaceae community in that host.
Host and climate data. The taxonomic classifications used in this
study are those presented by Cogger (1992), Strahan (1995) and
Simpson & Day (1996). Values for the ‘typical’ body mass (g) of
each host species were taken from the literature. The diet of a particular species can vary significantly with season and with habitat.
Consequently for mammals, each host species was assigned to one
broad diet category. The diet categories include carnivorous
(invertebrate and vertebrate prey), herbivorous (predominantly
vegetation) and omnivorous (invertebrates, vertebrates, fungi,
pollen, nectar, seeds and vegetation). For birds, the diet categories
were carnivorous (vertebrates), frugivorous (fruit), insectivorous,
frugivorous/insectivorous, nectarivorous, nectarivorous/insectivorous,
granivorous (seeds) and omnivorous. The climate [desert, semi-arid
(grassland), temperate and tropical] of each host locality was determined using the classification developed by the Australian Bureau of
Meteorology (www.bom.gov.au), which is based on long-term temperature and rainfall data.
3576
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 15 Jun 2017 18:26:16
Microbiology 149
E. coli distribution and genetic structure
The prevalence of E. coli in vertebrates
Prevalence in mammals. The prevalence of E. coli
among host species varied from 0 to 100 % (Table 1).
Overall, E. coli was detected in 56 % of the 1063 mammalian hosts examined. Taxonomic rank accounted for 42 %
of the total deviance in the presence/absence of E. coli
in each faecal sample. The presence of E. coli in a faecal
sample is largely explained by the taxonomic levels of
Order, Genus and Species, whilst Subclass and Family
were not significant (Table 2).
Mammal species live in different climates and differ in their
body mass and diet. Each of these factors is a statistically
significant predictor of the presence of E. coli in a host
Table 1. Prevalence of E. coli for selected Australian mammal species (n>6)
Order
Monotremata
Dasyuromorphia
Dasyuromorphia
Dasyuromorphia
Dasyuromorphia
Dasyuromorphia
Dasyuromorphia
Dasyuromorphia
Dasyuromorphia
Dasyuromorphia
Dasyuromorphia
Dasyuromorphia
Dasyuromorphia
Diprotodontia
Diprotodontia
Diprotodontia
Diprotodontia
Diprotodontia
Diprotodontia
Diprotodontia
Diprotodontia
Diprotodontia
Diprotodontia
Diprotodontia
Diprotodontia
Diprotodontia
Peramelemorphia
Peramelemorphia
Chiroptera
Chiroptera
Chiroptera
Chiroptera
Chiroptera
Rodentia
Rodentia
Rodentia
Rodentia
Rodentia
Rodentia
Rodentia
Rodentia
Rodentia
Rodentia
Rodentia
http://mic.sgmjournals.org
Family
Species
Sample size
Prevalence (%)
Tachyglossidae
Dasyuridae
Dasyuridae
Dasyuridae
Dasyuridae
Dasyuridae
Dasyuridae
Dasyuridae
Dasyuridae
Dasyuridae
Dasyuridae
Dasyuridae
Dasyuridae
Burramyidae
Macropodidae
Macropodidae
Macropodidae
Macropodidae
Petauridae
Phalangeridae
Phalangeridae
Phascolarctidae
Potoroidae
Potoroidae
Vombatidae
Vombatidae
Peramelidae
Peramelidae
Vespertilionidae
Vespertilionidae
Vespertilionidae
Vespertilionidae
Vespertilionidae
Muridae
Muridae
Muridae
Muridae
Muridae
Muridae
Muridae
Muridae
Muridae
Muridae
Muridae
Tachyglossus aculeatus
Antechinus flavipes
Antechinus minor
Antechinus stuartii
Antechinus swainsonii
Dasycercus cristicauda
Dasyurus geoffroii
Dasyurus hallucatus
Dasyurus maculatus
Dasyurus viverrinus
Ningaui ridei
Parantechinus bilarni
Sarcophilus harrisii
Burramys parvus
Macropus eugenii
Macropus giganteus
Macropus robustus
Onychogalea fraenata
Petaurus gracilis
Trichosurus caninus
Trichosurus vulpecula
Phascolarctos cinereus
Bettongia penicillata
Potorous tridactylus
Lasiorhinus krefftii
Lasiorhinus latifrons
Isoodon obesulus
Perameles nasuta
Chalinolobus gouldii
Chalinolobus morio
Nyctophilus geoffroyi
Vespadelus darlingtoni
Vespadelus vulturnus
Mastacomys fuscus
Mus musculus
Notomys alexis
Notomys mitchellii
Pseudomys desertor
Pseudomys hermannsburgensis
Pseudomys patrius
Rattus fuscipes
Rattus lutreolus
Rattus rattus
Zyzomys argurus
9
23
8
26
18
19
9
34
14
42
9
9
89
11
13
24
26
11
10
60
52
18
15
11
10
7
8
8
8
7
24
12
16
33
10
40
7
21
35
8
84
20
7
14
44
22
0
23
22
16
89
76
79
74
0
0
79
36
92
96
96
82
40
100
75
78
93
82
80
86
75
75
25
0
25
42
13
33
80
3
100
0
0
37
75
65
100
29
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 15 Jun 2017 18:26:16
3577
D. M. Gordon and A. Cowling
Source of variation
Subclass
Order|S
Family|O, S
Genus|F, O, S
Species|G, F, O, S
Residual
Total
d.f.
2
3
10
28
48
971
1062
Deviance
82?77
75?65
51?93
227?50
175?56
842?51
1455?92
F
1?64
4?86
0?64
2?22
P value
0?333
0?025
0?767
0?007
0?000
Prevalence of E. coli
Table 2. Analysis of deviance for the presence/absence of
E. coli in each faecal sample from mammals with respect to
the taxonomic rank of the host
1.0
0.9
0.8
0.7
11
6
0.0
_0.1
0
4
18
1
9
2
5
16
3
8
7
14
10
1
2
log
individual (Table 3). The inclusion of any of the
possible second- and third-order interaction terms (e.g.
Diet*Climate) did not significantly improve the explanatory
power of the statistical model. Inclusion of host species
as a random effect in the analysis did not alter the significance of diet, climate and body mass as predictors of the
prevalence of E. coli. Climate, diet and body mass accounted
for 65 % of the among-species variation in prevalence.
The probability of isolating E. coli from a host individual
will depend on the climate in which the host lives, its
diet and body mass. Overall, E. coli is unlikely to be isolated
from hosts living in the desert, whilst hosts living in the
tropics are less likely to harbour E. coli than hosts living
in temperate or semi-arid environments. E. coli is less likely
to be isolated from carnivores than herbivores and most
likely to be isolated from omnivores.
The probability of isolating E. coli from a mammal increases
with the body mass of the host. The relationship between
the prevalence of E. coli and host body mass is illustrated
for the carnivores (Fig. 1). The observed prevalence of E. coli
in the echidna (Tachyglossus aculeatus) is clearly lower
than that predicted by its body mass and this discrepancy
probably reflects the echidna’s highly specialized diet of ants
and termites. Dasycercus cristicauda is a desert-dwelling
species and this explains the low incidence of E. coli in this
species.
Table 3. Analysis of deviance for the prevalence of E. coli
among mammalian host individuals
The data are unbalanced so terms are added to the model sequentially beginning with body mass.
Source of variation
d.f.
Deviance
P value
log10(Body mass)
Climate
Diet
Species
Residual
Total
1
3
2
88
968
1062
322?95
51?88
26?80
214?69
839?60
1455?92
0?000
0?000
0?000
0?000
3578
15
13
17
0.6
0.5
0.4
0.3
0.2
0.1
12
10
3
[Body Mass (g)]
4
5
Fig. 1. The relationship between the prevalence of E. coli in
carnivorous Australian mammals and the log10[body mass (g)]
of the host species. The solid line depicts the relationship
between body mass and prevalence as determined using the
data for 209 individuals belonging to the Order Dasyuromorphia. Points represent the observed prevalence for carnivorous
species where n>5. The circles denote species in the
Chiroptera: 1, Chalinolobus gouldii; 2, Nyctophilus geoffroyi;
3, Scotorepens balstoni; 4, Vespadelus darlingtoni; 5, Vespadelus
vulturnus. Squares denote species in the Dasyuromorpha:
6, Antechinus bellus; 7, Antechinus flavipes; 8, Antechinus
stuartii; 9, Antechinus swainsonii; 10, Dasycercus cristicauda;
11, Dasyurus geoffroii; 12, Dasyurus hallucatus; 13, Dasyurus
viverrinus; 14, Parantechninus bilarni; 15, Sarcophilus harrisii;
16, Sminthopsis murina. Triangles denote species in the Order
Monotremata: 17, Ornithorhynchus anatinus; 18, Tachyglossus
aculeatus.
Prevalence in birds. Prevalence among host species
varied from 0 to 100 %. Prevalence data for those species
for which more than six individuals were examined are
presented in Table 4. E. coli was detected in 23 % of the
634 bird hosts examined. Taxonomic rank accounted for
37 % of the total deviance in the presence/absence of
E. coli in each faecal sample (Table 5). The presence/
absence of E. coli is explained largely by the taxonomic
rank of Genus. Taxonomic Order, Family and Species
were not significant explanatory variables (Table 5).
A significant number of the birds examined were sampled
from localities where they lived in close association with
humans, whilst others were collected from localities away
from significant levels of human habitation. Diet varies
substantially among species and although not as extensive
as in mammals, among-species variation in body mass is
considerable. Most of the birds sampled were collected
from temperate regions and consequently climate could
not be included as a factor in the statistical analyses.
Human association, diet and body mass were found to be
significant predictors of the prevalence of E. coli in birds
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 15 Jun 2017 18:26:16
Microbiology 149
E. coli distribution and genetic structure
Table 4. Prevalence of E. coli for selected Australian bird species (n>6)
Order
Anseriformes
Ciconiiformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Passeriformes
Family
Species
Sample size
Prevalence (%)
Anatidae
Accipitridae
Artamidae
Dicruridae
Dicruridae
Fringillidae
Maluridae
Maluridae
Meliphagidae
Meliphagidae
Meliphagidae
Meliphagidae
Meliphagidae
Meliphagidae
Meliphagidae
Meliphagidae
Muscicapidae
Oriolidae
Pachycephalidae
Pardalotidae
Pardalotidae
Pardalotidae
Passeridae
Passeridae
Passeridae
Petroicidae
Ptilinorhynchidae
Chenonetta jubata
Falco berigora
Gymnorhina tibicen
Grallina cyanoleuca
Rhipidura fulginosa
Carduelis carduelis
Amytornis barbatus
Malurus cyaneus
Acanthagenys rufogularis
Lichenostomus chrysops
Lichenostomus fuscus
Lichenostomus penicillattus
Manorina melanophrys
Melithreptus brevirostris
Philemon corniculatus
Phylidonyris novaehollandiae
Turdus merula
Oriolus sagittatus
Pachycephala rufiventris
Acanthiza lineata
Acanthiza pusilla
Sericornis frontalis
Neochmia temporalis
Passer domesticus
Taeniopygia guttata
Eopsaltria australis
Ptilonorhynchus violaceus
10
37
22
8
15
7
7
38
7
7
27
19
12
10
10
15
7
8
13
7
17
23
12
7
7
10
32
70
16
50
25
47
0
0
34
14
14
19
32
8
0
20
7
71
0
0
0
6
26
8
86
14
8
6
(Table 6). Whilst human association, diet and body mass
were significant predictors of prevalence, these factors
accounted for only 21 % of the among-species variation in
prevalence. Birds living in close association with humans
were more likely to have E. coli than birds living away from
human habitation. The probability of a host harbouring
E. coli increased with its body mass. The prevalence of
E. coli was lowest in exclusively seed- or fruit-eating species
and higher in those species that include nectar, insects or
vertebrates in their diet.
Prevalence in reptiles. E. coli was isolated from 33 % of
the crocodile hosts examined (n=33), 4 % of the turtles
(n=56), 2 % of the snakes (n=44) and 10 % of the
lizards (n=314). Prevalence data for individual species
where at least six hosts were examined are presented in
Table 7.
E. coli is unlikely to be isolated from most species of
lizard, although prevalence did vary significantly among
Table 6. Analysis of deviance for the prevalence of E. coli
among bird host individuals
Table 5. Analysis of deviance for the presence/absence of
E. coli in each faecal sample from birds with respect to the
taxonomic rank of the host
The data are unbalanced so terms are added to the model sequentially beginning with human association.
Source of variation
d.f.
Deviance
F
P value
Source of variation
d.f.
Deviance
P value
Order|S
Family|O, S
Genus|F, O, S
Species|G, F, O, S
Residual
Total
8
26
56
38
505
633
23?76
79?60
113?29
34?15
431?01
681?81
0?97
1?51
2?25
0?480
0?099
0?005
0?648
Human association
log10(Body mass)
Diet
Species
Residual
Total
1
1
7
119
504
633
32?96
7?64
13?99
201?40
425?20
681?81
0?000
0?006
0?051
0?000
http://mic.sgmjournals.org
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 15 Jun 2017 18:26:16
3579
D. M. Gordon and A. Cowling
Table 7. Prevalence of E. coli for selected Australian reptile species (n>6)
Order
Crocodilia
Crocodilia
Squamata
Squamata
Squamata
Squamata
Squamata
Squamata
Squamata
Squamata
Squamata
Squamata
Squamata
Squamata
Squamata
Squamata
Testudines
Testudines
Testudines
Family
Species
Sample size
Prevalence (%)
Crocodylidae
Crocodylidae
Agamidae
Scincidae
Scincidae
Scincidae
Scincidae
Scincidae
Scincidae
Scincidae
Scincidae
Scincidae
Scincidae
Gekkonidae
Elapidae
Elapidae
Chelidae
Chelidae
Chelidae
Crocodylus johnstoni
Crocodylus porosus
Ctenophorus isolepes
Ctenotus taeniolatus
Ctenotus pantherinus
Egernia saxatilis
Eulamprus heatwolei
Eulamprus tympanum
Lampropholis delicata
Lampropholis guichenoti
Niveoscincus microlepidotus
Pseudemoia entecasteauxii
Tiliqua rugosa
Oedura leseuri
Pseudechis porphyriacus
Rhinoplocephalus nigrescens
Chelodina sp
Elseya dentata
Emydura victoriea
23
10
8
9
7
12
29
20
8
23
7
11
44
11
16
10
9
21
21
26
50
0
0
0
17
3
0
12
22
0
0
32
0
0
0
0
5
0
species (x2=35?3, d.f.=11, P=0?000). E. coli was relatively
common in four species: Lampropholis delicata, Lampropholis guuichenoti, Egernia saxatilis and Tiliqua rugosa. The
two Lampropholis species are small skinks and all individuals examined of these two species were collected from a
cemetery located in central Sydney, New South Wales.
Egernia saxatilis is a moderately sized skink that includes
a significant amount of fruits and flowers in its diet, whilst
Tiliqua rugosa is a large skink that is unusual in that it feeds
almost exclusively on vegetation as an adult.
Prevalence in frogs. E. coli was isolated from 12 % of
the frogs examined (n=106). However, 10 of the 13 isolates of E. coli were recovered from two species of large
tree frog (Litoria infrafrenata and Litoria caerulea) collected in the suburbs of Cairns, Queensland. Both Litoria
species often live in or around houses and sheds.
Prevalence in fish. E. coli was isolated from 10 % of the
fish examined (n=138). No differences in the prevalence
of E. coli could be detected among the four species for
which there were reasonable samples (n>12) (x2=3?6,
d.f.=3, P=0?30).
The distribution of A, B1, B2 and D strains in
vertebrates
A supplementary data file for all strains examined for this
study can be obtained as part of the online version of this
paper at http://mic.sgmjournals.org. The supplementary
file contains the results of the Clermont et al. (2000)
method for the assignment of E. coli strains to one of
the four groups (A, B1, B2, D) identified in the ECOR
collection, together with data concerning the diet and
3580
body mass of the host of origin, as well as the climate in
which the host was collected.
Distribution of E. coli groups in mammals. Of the 497
E. coli isolates characterized for their group membership
15 % were members of group A, 33 % were group B1,
35 % were group B2 and 17 % were group D strains. Host
diet and the climate where the host was collected were
found to account for a significant amount of the variation in the distribution of groups among hosts (Table 8).
Host body mass also explained a portion of the variation, but the nature of the relationship between the distribution of the E. coli groups and body mass depended
upon host diet (Table 8). Climate, host diet and body
mass accounted for 40 % of the among-species variation
in the distributions of A, B1, B2 and D strains. However,
Table 8. Analysis of deviance for the distribution of strains
of the four groups (A, B1, B2, D) of E. coli among mammalian host individuals
The data are unbalanced so terms are added to the model sequentially beginning with diet.
Source of variation
d.f.
Deviance
P value
Diet
Climate
log10(Body mass)
Diet*log10(Body mass)
Species
Residual
Total
6
9
3
6
183
1320
1527
66?70
41?44
6?89
88?19
299?42
842?83
1345?47
0?000
0?000
0?075
0?001
0?000
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 15 Jun 2017 18:26:16
Microbiology 149
E. coli distribution and genetic structure
Relative frequency of
E. coli groups (%)
100
80
60
40
20
0
A
B1 B2 D
Desert
A B1 B2 D
Semi-arid
A B1 B2 D
Temperate
Fig. 2. The predicted probability of observing
E. coli groups A, B1, B2 and D in different
climates assuming a 150 g omnivorous host.
A B1 B2 D
Tropical
even after accounting for these effects, differences among
host species still accounted for a significant fraction of the
observed variation (Table 8).
The distribution of A, B1, B2 and D strains in hosts from
different climates is illustrated using the predictions of the
statistical model adjusted to reflect rodent hosts (omnivores
weighing 120 g), a group present in most Australian biomes
(Fig. 2). Group A strains are predicted to predominate in
hosts living in desert regions and strains from the other
groups are expected to be uncommon. Group A strains are
predicted to occur in hosts living in temperate and tropical
regions, but are anticipated to be uncommon in hosts living
in semi-arid environments. Group B2 strains are expected to
be less frequently isolated from hosts living in the tropics
than from hosts living in semi-arid or temperate climates.
The predicted relationship between the frequencies of A,
B1, B2 and D strains and body mass is illustrated for each
diet class for hosts living in a temperate climate (Fig. 3). In
omnivores B2 strains predominated. The frequency of B2
strains increased and the frequency of B1 strains declined
as host body mass increased. Group A and D strains were
relatively rare and their frequency showed little change
with host body mass. In herbivores, B2 strains were most
common in hosts weighing less than 6 kg, but as body mass
increased above 6 kg, B2 strains became much less common,
with B1 strains becoming correspondingly more common.
There was a slight decline in the frequency of group A strains
with increasing body mass. In carnivores, B1 strains were
http://mic.sgmjournals.org
0.8
0.6
0.4
0.2
0.0
0.5
Relative frequency of E. coli groups
The presence/absence of the three PCR products generated
using the Clermont method can be used to assign a strain to
one of the four E. coli groups. Alternatively, the Clermont
method gives rise to eight possible genotypes based on the
possible combinations of the presence or absence of the
three PCR products. (We note that the genotype chuA2
yjaA+ and TSPE4.C2+ was never observed.) When the
analysis presented in Table 8 was repeated, analysing the
frequencies of the seven observed genotypes rather than A,
B1, B2 and D frequencies, the same outcomes were observed.
All factors included in the model presented in Table 8 are
statistically significant and there are only small changes
in the probability values. This outcome demonstrates that
E. coli genotypes are non-randomly distributed with respect
to climate, host diet and host body mass.
(a)
1.0
1
1.5
2.5
2
3
3.5
4
(b)
1.0
0.8
0.6
0.4
0.2
0.0
2
2.5
3
3.5
4
4.5
5
3.5
4
(c)
1.0
0.8
0.6
0.4
0.2
0.0
1
1.5
3
2.5
2
log10 [Body mass (g)]
Fig. 3. Predicted probability of observing E. coli groups A, B1,
B2 and D for (a) carnivorous, (b) herbivorous and (c) omnivorous
mammalian hosts at different body masses (log10) in temperate
climates. Tick marks at the top of the graph show body masses
of host species used in the study. (a) A, solid line; B1, dashdot line; B2, dashed line; D, dotted line. (b, c) A, dashed line;
B1, dash-dot line; B2, solid line; D, dotted line.
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 15 Jun 2017 18:26:16
3581
D. M. Gordon and A. Cowling
Table 9. Frequency of occurrence of the four groups (A,
B1, B2, D) of E. coli in vertebrates
Host group
No. of
strains
Fish
Frogs
Turtles
Snakes and lizards
Crocodiles
Birds
Mammals
12
13
4
33
10
134
497
E. coli group (%)
A
B1
B2
D
0
7?7
25
15?1
20
8?2
15?5
91?7
84?6
50
69?7
70
49?2
32?7
8?3
0
25
6?1
10
22?3
35?3
0
7?7
0
9?1
0
20?1
16?5
the most common. There was an increase in the frequency
of group A strains with host body mass and a corresponding decline in the frequency of B1, B2 and D strains.
Distribution of E. coli groups in non-mammalian
vertebrates. The relative frequencies of strains of the four
groups recovered from fish, crocodiles, turtles, snakes and
lizards as well as birds and mammals are presented in
Table 9. Too few strains were isolated from the ectothermic vertebrates or birds to allow for any meaningful
statistical analysis.
Distribution of E. coli groups among hosts groups.
The frequency of strains in the four groups isolated from
the different vertebrate host groups is depicted in Fig. 4.
For this analysis, the ectothermic vertebrates were pooled
as a single host group. The relative frequencies of the four
groups were similar in mammalian herbivores and omnivores, but differed from that found in carnivores. The distributions of A, B1, B2 and D strains were similar in birds
and carnivorous mammals. Ectothermic vertebrates had a
very different distribution of the four groups compared to
endothermic vertebrates.
DISCUSSION
The repeatability of the results
The patterns observed in this study appear to be temporally
stable and independent of the method used to isolate the
bacteria. The majority of mammalian hosts were sampled
over two periods, 1993–1997 and 2000–2002. The same
basic patterns were observed for the prevalence of E. coli
in each set of samples. E. coli group A strains are overrepresented in Tasmanian devils (Sarcophilus harrisii) and
this pattern was observed in two sets of samples collected
8 years apart from two different localities (data not presented). Pupo et al. (2000b) reported on the results of a
study of E. coli isolated from mammals collected from one
locality in eastern Australia. The majority of their samples
were from a single species of native rat, but they did sample
a number of other species. Although the method they used
to isolate bacteria was quite different from the one used in
the present study, they also observed that E. coli was less
commonly isolated from bats and carnivorous marsupials
than from omnivores or herbivores.
Climate and the prevalence of E. coli
E. coli is less likely to be isolated from hosts living in desert
or tropical climates than hosts living in semi-arid or temperate climates. This pattern of variation was consistent
across host orders, and does not appear to be a host effect
as the prevalence of E. coli in species of the same genus
differed depending on climatic zone. Why E. coli is less
prevalent in mammals living in the tropics is unknown.
Maximum summer temperatures in the desert regions of
central Australia are typically in excess of 35 uC, the relative
humidity is less than 25 % and median rainfall is less than
20 mm a month. It seems likely these conditions will
adversely affect the survival of E. coli in the external
environment and may, in turn, limit the transmission of
E. coli among host individuals.
Alternatively, the near absence of E. coli in the faeces of
desert-dwelling mammals may be a consequence of host
biology. Mammals living in arid regions have adaptations
that enable them to conserve water. Whilst water loss
through defecation represents a small fraction of an animal’s
water balance, morphological modifications in the colon to
maximize water retention do occur (Stevens, 1988; Woodall
& Skinner, 1993; Murray et al., 1995). It may be that the
lower water content of material in the distal colon adversely
affects the survival of E. coli cells in this region and as
a result cell densities are reduced to such an extent that
E. coli is unlikely to be detected in faeces.
Relative frequency of
E. coli groups (%)
100
3582
80
60
40
20
0
A B1 B2 D A B1 B2 D A B1 B2 D A B1 B2 D A B1 B2 D
Birds
Mammalian Mammalian Mammalian
Fish, frogs,
carnivores omnivores herbivores
reptiles
Fig. 4. The observed frequency of E. coli
groups A, B1, B2 and D in different vertebrate host groups.
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 15 Jun 2017 18:26:16
Microbiology 149
E. coli distribution and genetic structure
Host effects and the prevalence of E. coli
The likelihood of a host harbouring E. coli will depend
primarily on three factors: first, the frequency with which a
host individual is exposed to E. coli. Second, the probability
that an exposure event will result in the establishment of
a population. Third, the mean length of time the E. coli
population can persist in the host, or alternatively the rate
at which the host loses its population of E. coli. If one
assumes a cohort of new-born hosts uninfected with E. coli,
then the change with age in the number of hosts with and
without E. coli is described by the following differential
equations:
If contact rate is not the sole factor responsible for the
variation in prevalence observed among species, then this
would suggest that the among-species variation may, in
part, be due to the ability of E. coli to establish a population
in the gastro-intestinal tract of the host.
dN
~{bNzdE{kN
da
dE
~zbN{dE{kE
da
where N is the number of individuals in the cohort without
E. coli and E the number of hosts harbouring E. coli. The
transmission rate, b, represents the rate at which a host is
exposed to E. coli and the probability that exposure will
result in the establishment of a population. The E. coli
population is lost from a host individual at the rate d.
Individuals are lost from the cohort due to death, which
occurs at the rate m. The total number of individuals in the
cohort is given as N+E. The equilibrium prevalence of the
host harbouring E. coli is:
%E ~
bz kd
bz kd z1
Thus, in a cohort of hosts the prevalence of E. coli will
depend on both the transmission rate and the rate at
which a host loses its population of E. coli. Where, for a
given transmission rate, the prevalence will decline as the
residency time (1/d) of the E. coli population in an
individual declines.
The observation that birds are more likely to harbour E. coli
if they are commensally associated with humans than if
they are not indicates that exposure rates do vary among
hosts. These higher levels of exposure may be a consequence
of many species of birds, such as house sparrows (Passer
domesticus) and starlings (Turdus merula) feeding on
household food scraps or in compost heaps that are likely
to contain E. coli. In addition, these higher exposure
levels may also be due to elevated levels of environmental
contamination by E. coli due to the faeces of domestic pets
and livestock, hosts in which E. coli is prevalent. The
observation that E. coli was more prevalent in frogs and
lizards living in association with humans provides support
for elevated levels of E. coli in the environment, as these
species feed primarily on insects or other invertebrates. In
mammals, variation in prevalence due to exposure would
be expected; however, hosts in which E. coli is normally
uncommon (bats and small marsupial carnivores) and
which live in urban or suburban habitats were not sampled.
http://mic.sgmjournals.org
E. coli is not common in most species of birds or carnivorous mammals, many of which feed largely or exclusively
on insects. Why E. coli is rare in these hosts is unknown. Is
E. coli uncommon in insectivores because these hosts are
seldom exposed to E. coli? This seems to be unlikely for
two reasons. First, E. coli can be isolated from insects
(unpublished data) and insectivores eat many insects per
day. Second, it is not immediately obvious why grazers
such as kangaroos that feed predominantly on grasses or
browsers such as wallabies that feed on the foliage of trees
and shrubs would be more frequently exposed to E. coli in
their diet. Yet, E. coli is prevalent in kangaroos and wallabies.
The gut morphology of the host, diet, food retention times
and body size, are all inter-related factors that correlate with
host phylogeny (Hume, 1999). Animals represent an energyrich, easily digested food source. Consequently carnivorous
mammals have relatively simple guts. The digestive tract is
dominated by the small intestine, and the colon is short
and often poorly differentiated from the small intestine.
Amongst Australian carnivores, a caecum is usually lacking
or is poorly developed (Hume, 1999). Short intestines,
coupled with high quality food, result in rapid gut transit
times. Furthermore, gut transit times increase with body
size and vary from about 1 hour for the 18 g dasyurid
Sminthopsis crassicaudata to 13 hours for the 1000 g
dasyurid, Dasyurus viverrinus (Hume, 1999).
By contrast, plant tissue represents an energy-poor food
source that is difficult to digest. Microbial fermentation
enables mammals consuming plant tissue to increase the
nutritional value of the material they ingest, but it requires
the presence of a fermentation chamber. In kangaroos and
wallabies (Macropodidae), as well as the rat kangaroos
(Potoroidae), the foregut provides the primary site of
microbial fermentation, although these species also have
a caecum (Hume, 1999). The hindgut is the site of microbial fermentation in the balance of Australian herbivores
(Hume, 1999). In the majority of small (<10 kg) herbivores
the caecum forms the fermentation chamber, whilst in the
larger species fermentation takes place in the colon (Hume,
1999). Retention times are significantly longer in herbivores
as compared to carnivores and, as in carnivores, increase
with body mass. For example, food transit times are about
35 h in the 1 kg common ringtail possum (Pseudocheirus
peregrinus) and about 100 h in the 10 kg koala (Phascolarctos cinereus) (Hume, 1999).
The digestive tract of Australian omnivores is intermediate
between that of the carnivores and the herbivores (Hume,
1999). In the small, primarily insectivorous, mountain
pygmy possum the caecum is very small and the colon
short, whilst the sugar glider (Petaurus breviceps) has a
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 15 Jun 2017 18:26:16
3583
D. M. Gordon and A. Cowling
well-developed caecum. The bandicoots have a modestsized caecum and colon. All Australian rodents possess a
caecum (Chivers & Langer, 1994). Food transit rates in
omnivores fall between those of carnivores and herbivores
(Hume, 1999).
Several mathematical models have been developed with the
purpose of exploring the dynamics of bacterial populations
in the gastro-intestinal tract. These models have investigated
the dynamics of bacterial establishment using chemostat
models (Topiwala & Hamer, 1971; Baltzis & Fredrickson,
1983; Freter et al., 1986; Smith & Waltman, 1995; Stemmons
& Smith, 2000; Ballyk et al., 2001) and plug flow reactor
models (Ballyk & Smith, 1999; Ballyk et al., 2001; Jones et al.,
2002). The chemostat models describe a physical system
closely resembling that of a caecum, while the plug flow
reactor models describe a typical carnivore gut. These
models have shown that establishment is more likely when
the bacterial population exhibits wall growth than when
no wall growth can occur. Nutrient concentration is also a
factor and there is a threshold concentration of nutrients
entering the model gut, below which the bacterial population cannot establish and above which the population will
establish and persist. Further, the rate at which material
moves through the gut (transit time) can also determine
whether a bacterial population can establish, where decreasing transit times lead to populations being less likely to
establish. Consequently, the conditions for establishment
are predicted to be more restrictive for the small intestine
than for the hindgut, due to the faster transit times
experienced in the small intestine.
The quantity and type of nutrients entering the lower
gastro-intestinal tract will vary with host diet. Gut turnover rates and transit times will also vary with diet, gut
morphology and body mass. The theoretical models suggest
that these factors may well influence the likelihood of E. coli
being able to establish a population in a host and hence
determine the prevalence of E. coli in a particular host
species. Further experimental and theoretical work is
required to determine if one of the factors of host diet,
gut morphology or body mass (food transit rate) is likely to
dominate the dynamics of bacterial establishment in the
gastro-intestinal tract.
To what extent residence time of an E. coli population
varies with diet, gut morphology or body mass is unknown.
There have been few studies that have examined the
persistence time of an E. coli strain in a host and these
studies have largely been restricted to humans (Sears et al.,
1950, 1956; Sears & Brownlee, 1952; Shooter et al., 1977;
Caugant et al., 1981). The results of these studies indicate
that a strain may persist for months if not years. However,
the data also suggest that eventually the resident strain is
inevitably lost. The factors responsible for the turnover of
strains in a host are unknown. Long-term longitudinal
studies of the persistence of E. coli in different host species
are required.
3584
From the results of this survey of the distribution of E. coli
in Australian vertebrates suggest that, as a species, E. coli is
adapted to mammals with hindgut microbial fermentation
chambers, or in the absence of a hindgut fermentation
chamber the host must be ‘large’. The contrast between
rodents and carnivores highlights the apparent importance
of a hindgut fermentation chamber. E. coli is common in
most rodent species, all of which possess a caecum. E. coli is
uncommon in bats and many marsupial carnivores, all of
which lack a caecum. Yet, many species of rodent are at
least as small as bats and smaller than many of the species
of marsupial carnivore in which E. coli is uncommon. The
observation that the prevalence of E. coli in carnivores
increases with host body mass indicates that a ‘large’ body
mass can compensate for the lack of a hindgut fermentation
chamber.
The distribution of ECOR groups in vertebrates
The PCR-based Clermont method can produce eight genotypes, seven of which were observed. An analysis similar to
that presented in Table 8 showed that these genotypes are
non-randomly distributed with respect to climate, host
diet and body mass. Does this non-random distribution of
genotypes correspond to a non-random distribution of the
E. coli groups identified in the ECOR collection? For this
question to be answered in the affirmative requires two
assumptions to be valid. One is that the ECOR collection is
broadly representative of the genetic diversity to be found
in E. coli worldwide and that the great majority of strains
can be classified into one of the four ECOR groups (A,
B1, B2 and D). Recently, Thierry Wirth and colleagues
(personal communication) investigated the genetic structure of E. coli using the technique of multi-locus sequence
typing, MLST (Maiden et al., 1998). The MLST study was
based on over 460 clinical and faecal isolates from humans
and 41 species of animal predominantly collected from
Africa, Australia, Europe and North America. Analyses of
the MLST results demonstrated that the ECOR collection is
largely representative of the MLST diversity to be found in
faecal isolates of E. coli and confirmed the existence of
four genetic groups that correspond to the groups defined
in the ECOR collection. The MLST results revealed that
recombination among the four groups does occur. Strains
are observed which are a ‘mix’ of group A and B1 genetic
backgrounds. In addition, some strains are highly heterogeneous and appear to be a mix of A, B1, B2 and D
genetic backgrounds. However, three-quarters of the highly
heterogeneous strains were clinical isolates associated with
intestinal disease. Further details of the Wirth and colleagues
MLST study are available at www.mpiib-berlin.mpg.de.
Given that the great majority of E. coli strains can be assigned to one of four groups corresponding to those identified in the ECOR collection, then the other assumption is
that the Clermont method accurately assigns a strain to an
ECOR group. The Thierry Wirth MLST study included 15
of the Australian strains genotyped in the present study.
In all but one case there was agreement between group
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 15 Jun 2017 18:26:16
Microbiology 149
E. coli distribution and genetic structure
assignment based on the Clermont method and the phylogenic assignment based on the MLST data (Thierry Wirth,
personal communication). Johnson et al. (2001) demonstrated that a range of virulence factors associated with
extra-intestinal disease vary in their occurrence and their
frequency among the four groups of E. coli. All of the isolates examined in the present study have been screened for
30 virulence factors associated with intestinal and extraintestinal disease (unpublished data). The results of the
virulence factor screening are in close agreement with the
E. coli group assignment of the strains. Consequently, in
the discussion to follow we assume that E. coli consists of
four, largely distinct, genetic groups and that the group
assignments based on the Clermont method are accurate
for more than 90 % of the isolates.
E. coli groups A, B1, B2 and D are non-randomly distributed
among different mammalian hosts. Diet, gut morphology
and body mass are all significant predictors of the distribution of the groups among mammalian host. Birds
and carnivorous mammals have relatively simple tube-like
gastro-intestinal tracts and in both host groups B1 strains
predominate. The herbivorous and omnivorous animals
all possess a caecum and in these hosts, group B2 strains
predominate. The effects of gut morphology and body
mass on the distribution of the E. coli groups supports
the conclusion that the observed among-species variation
in prevalence is in part due to gut morphology and dynamics and is not solely a consequence of among-species
differences in exposure to E. coli.
The observed distribution of strains from the different
E. coli groups suggests that these groups may have different
life-history tactics and ecological niches. A and B1 strains
appear to be ‘generalists’ as they can be recovered from
any vertebrate group. Group B2 and D strains seem to be
‘specialists’ as they are largely restricted to endothermic
vertebrates. Group B2 strains appear to be the most specialized, as they predominate in mammals with a hindgut
fermentation chamber. Several factors may be responsible
for the different distributions of A, B1, B2 and D strains
among host groups. The observed distributions may reflect
among-group differences in the ability of strains to establish a population, as appears to be the case for B2 and
D strains in ectotherms. Competitive interactions among
strains within a host may also play a role. Are group
B1 strains primarily adapted to vertebrates with simple
hindguts or are they competitively inferior to group B2
strains in hosts with hindgut fermentation? Finally, these
patterns may reflect differences in the ‘typical’ transmission
dynamics of strains among host individuals. Strains of the
different E. coli groups differ substantially in the types of
disease they are capable of causing in humans. The majority
of extra-intestinal E. coli infections in humans are caused
by group B2 and D strains (Picard et al., 1999; Johnson
et al., 2001). Most strains responsible for intestinal E. coli
infections in mammals are derived from strains related to
groups A and B1 and to a lesser extent D (Pupo et al., 1997;
http://mic.sgmjournals.org
Rolland et al., 1998; Pupo et al., 2000a). The source of
strains responsible for extra-intestinal infections is thought
to be the faecal E. coli population of the infected individual,
whilst food and water are the main source of strains
responsible for intestinal infection.
Host habitat, diet, ‘typical’ body temperature and gut
morphology are all factors that influence the prevalence of
E. coli and the distribution of the four E. coli groups among
host individuals. However, the observation that even after
accounting for these factors, host species still explains a
significant amount of the among-host variation indicates
that other factors may be influencing the distribution of
E. coli. Variation in prevalence due to host species effects
appears to be particularly significant in birds, where the
factors of human association, host diet and body mass
account for only 21 % of the among-species variation. To
what extent the unexplained variation in birds is a consequence of stochastic factors as compared to unknown
ecological factors will require further study. By contrast, in
mammals, climate, host diet and body mass accounted for
65 % of the among-species variation in the occurrence of
E. coli. In mammals, the unexplained variation may be
due to the fact that even closely related species living in the
same climate will usually exhibit different microhabitat or
diet preferences. Similar species may also differ in their
behaviour, for example, in their social organization.
The results of this survey of the distribution of E. coli in
Australian vertebrates indicates that as a species, E. coli
is adapted to mammals with hindgut microbial fermentation chambers. In the absence of a hindgut fermentation
chamber, E. coli is more likely to establish a population
in ‘large’ hosts than in ‘small’ hosts. Within the species,
strains from the different genetic groups of E. coli appear
to have contrasting life-history tactics and ecological
niches, although the precise nature of these differences is
unknown, as is the nature of the adaptations that result in
these differences.
ACKNOWLEDGEMENTS
This study could not have been undertaken without the assistance
of the many people who took extra time and effort during their own
field studies to collect the samples on which this study is based. Their
help was essential to the success of this study and is very much
appreciated. The technical assistance of Joanne Allison was invaluable.
This study was funded in part by the Australian Research Council and
the Faculty Research Fund of the Australian National University.
REFERENCES
Ballyk, M. & Smith, H. L. (1999). A model of microbial growth in a
plug-flow reactor with wall attachment. Math Biosci 158, 95–126.
Ballyk, M., Le, D., Jones, D. A. & Smith, H. L. (2001). Microbial
competition in reactors with wall attachment: a comparison of
chemostat and plug flow models. Microb Ecol 41, 210–221.
Baltzis, B. C. & Fredrickson, A. G. (1983). Competition of
two microbial populations for a single resource in a chemostat
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 15 Jun 2017 18:26:16
3585
D. M. Gordon and A. Cowling
when one of them exhibits wall attachment. Biotechnol Bioeng 25,
2419–2439.
Caugant, D. A., Levin, B. R. & Selander, R. K. (1981). Genetic
diversity and temporal variation in the E. coli populations of a
human host. Genetics 98, 467–490.
Chivers, D. J. & Langer, P. (1994). The Digestive System in Mammals:
Food, Form and Function. Cambridge: Cambridge University Press.
Clermont, O., Bonacorsi, S. & Bingen, E. (2000). Rapid and simple
determination of Escherichia coli phylogenetic group. Appl Environ
Microbiol 66, 4555–4558.
Cooger, H. G. (1992). Reptiles and Amphibians of Australia.
Chatswood, NSW: Reed Books.
Ewing, W. H. (1986). Edwards and Ewing’s Identification of
Pupo, G. M., Karaolis, D. K. R., Lan, R. & Reeves, P. R. (1997).
Evolutionary relationships among pathogenic and nonpathogenic
Escherichia coli inferred from multilocus enzyme electrophoresis and
mdh sequence studies. Infect Immun 65, 2685–2692.
Pupo, G. M., Lan, R. & Reeves, P. R. (2000a). Multiple independent
origins of Shigella clones of Escherichia coli and convergent evolution of their many characteristics. Proc Natl Acad Sci U S A 97,
10567–10572.
Pupo, G. M., Lan, R., Reeves, P. R. & Baverstock, P. R. (2000b).
Population genetics of Escherichia coli in a natural population of
native Australian rats. Environ Microbiol 2, 594–610.
Rolland, K., Lambert-Zechovsky, N., Picard, B. & Denamur, E.
(1998). Shigella and enteroinvasive Escherichia coli strains are derived
Enterobacteriaceae, 4th edn. New York: Elsevier.
from distinct ancestral strains of E. coli. Microbiology 144, 2667–2672.
Freter, R., Brickner, S. & Temme, S. (1986). An understanding of
Sears, H. J. & Brownlee, I. (1952). Further observations on the
colonization resistance of the mammalian large intestine requires
mathematical analysis. Microecol Ther 16, 147–155.
persistence of individual strains of Escherichia coli in the intestinal
tract of man. J Bacteriol 63, 47–57.
Gordon, D. M. & FitzGibbon, F. (1999). The distribution of enteric
Sears, H. J., Brownlee, I. & Uchiyama, J. K. (1950). Persistence of
bacteria among Australian mammals. Microbiology 145, 2663–2671.
Gordon, D. M. & Lee, J. (1999). The genetic structure of enterica
bacteria from Australian mammals. Microbiology 145, 2673–2682.
Hartl, D. L. & Dykhuizen, D. E. (1984). The population genetics of
Escherichia coli. Annu Rev Genet 18, 31–68.
Herzer, P. J., Inouye, S., Inouye, M. & Whittam, T. S. (1990).
Phylogenetic distribution of branched RNA-linked mulitcopy
single-stranded DNA among natural isolates of Escherichia coli.
J Bacteriol 172, 6175–6181.
Hume, I. D. (1999). Marsupial Nutrition. Cambridge: Cambridge
University Press.
Johnson, J. R., Delavari, P., Kuskowski, M. & Stell, A. L. (2001).
Phylogenetic distribution of extraintestinal virulence-associated traits
in Escherichia coli. J Infect Dis 183, 78–88.
Jones, D., Kojouharov, H. V., Le, D. & Smith, H. (2002). Bacterial wall
attachment in a flow reactor. SIAM J Appl Math 62, 1728–1771.
Leclerc, H., Mossel, A. A., Edberg, S. C. & Struijk, C. B. (2001).
Advances in the bacteriology of the coliform group: their suitability
as markers of microbial water safety. Annu Rev Microbiol 55, 201–234.
Murray, B. R., Hume, I. D. & Dickman, C. R. (1995). Digestive tract
individual strains of Escherichia coli in the intestinal tract of man.
J Bacteriol 59, 293–301.
Sears, H. J., Janes, H., Saloum, R., Brownlee, I. & Lamoreaux, L. F.
(1956). Persistence of individual strains of Escherichia coli in man
and dog under varying conditions. J Bacteriol 71, 370–372.
Selander, R. K., Caugant, D. A. & Whittam, T. S. (1987). Genetic
structure and variation in natural populations of Escherichia coli. In
Escherichia coli and Salmonella typhimurium: Cellular and Molecular
Biology, pp. 1625–1648. Edited by F. C. Neidhardt, J. L. Ingraham,
K. Brooks Low, B. Magasanik, M. Schaechter & H. E. Umbarger.
Washington, DC: American Society for Microbiology.
Shooter, R. A., Bettelheim, K. A., Lenox-King, S. M. J. & O’Farrell, S.
(1977). Escherichia coli serotypes in the faeces of healthy adults over
a period of several months. J Hyg 78, 95–98.
Simpson, K. & Day, N. (1996). Field Guide to the Birds of Australia.
Ringwood, VIC: Viking Penguin Books Australia.
Smith, H. L. & Waltman, P. (1995). The Theory of the Chemostat.
Cambridge: Cambridge University Press.
Stemmons, E. D. & Smith, H. L. (2000). Competition in a chemostat
with wall attachment. SIAM J Appl Math 61, 567–595.
Stevens, C. E. (1988). Comparative Physiology of the Vertebrate
characteristics of the spinifex hopping-mouse, Notomys alexis and
the sandy inland mouse, Pseudomys hermannsburgensis in relation to
diet. Aust Mammal 18, 93–97.
Digestive System. Cambridge: Cambridge University Press.
Ochman, H. & Selander, R. K. (1984). Standard reference strains of
Reed Books.
Strahan, R. (1995). The Mammals of Australia. Chatswood, NSW:
Escherichia coli from natural populations. J Bacteriol 157, 690–693.
Topiwala, H. & Hamer, G. (1971). Effect of wall growth in steady
Picard, B., Sevali Garcia, J., Gouriou, S., Duriez, P., Brahimi, N.,
Bingen, E., Elion, J. & Denamur, E. (1999). The link between
Woodall, P. F. & Skinner, J. D. (1993). Dimensions of the intestine,
phylogeny and virulence in Escherichia coli extra-intestinal infection.
Infect Immun 67, 546–553.
diet and faecal water loss in some African antelope. J Zool Lond 229,
457–471.
3586
state continuous culture. Biotechnol Bioeng 13, 919–922.
Downloaded from www.microbiologyresearch.org by
IP: 88.99.165.207
On: Thu, 15 Jun 2017 18:26:16
Microbiology 149