Phenotypic and genotypic characterization of encapsulated Escherichia coli isolated from blooms in two Australian lakes.

Abstract

Escherichia coli has long been used as an indicator organism for water quality assessment. Recently there has been an accumulation of evidence that suggests some strains of this organism are able to proliferate in the environment, a characteristic that would detract from its utility as an indicator of faecal pollution. Phenotypic and genotypic characterization of E. coli isolated from blooms in two Australian lakes, separated by a distance of approximately 200 km, identified that the blooms were dominated by three E. coli strains. A major phenotypic similarity among the three bloom strains was the presence of a group 1 capsule. Genetic characterization of a conserved region of the cps gene cluster, which encodes group 1 capsules, identified a high degree of genetic variation within the bloom isolates. This differs from previously described encapsulated E. coli strains which are highly conserved at the cps locus. The phenotypic or genotypic profiles of the bloom strains were not identified in 435 E. coli strains isolated from vertebrates. The occurrence of these encapsulated strains suggests that some E. coli have evolved a free-living lifestyle and do not require a host in order to proliferate. The presence of the same three strains in bloom events in different geographical regions of a temperate climate, and at different times, indicates that free-living E. coli strains are able to persist in these water reservoirs. This study provides further evidence of circumstances where caution is required in using E. coli as an indicator organism for water quality.

Full text

Page 1

Environmental Microbiology (2005) 7(5), 631–640 doi:10.1111/j.1462-2920.2005.00729.x
© 2005 Society for Applied Microbiology and Blackwell Publishing Ltd
Blackwell Science, LtdOxford, UKEMIEnvironmental Microbiology 1462-2912Society for Applied Microbiology and Blackwell Publishing Ltd, 20047 5631640Original ArticleCharacterization of encapasulated blooming Escherichia coliM. L. Power
et al.
Received 10 May, 2004; revised 13 August, 2004; accepted 19
August, 2004. *For correspondence. Email mpower@vetsci.usyd.
edu.au; Tel. (+61) 2 9351 7119; Fax (+61) 2 9351 7348. †Present
address: Faculty of Veterinary Science, University of Sydney,
McMaster Building B14, Sydney, NSW 2006, Australia.
Phenotypic and genotypic characterization of
encapsulated Escherichia coli isolated from blooms in
two Australian lakes
Michelle L. Power,1*† Jane Littlefield-Wyer,2
David M. Gordon,2 Duncan A. Veal1 and
Martin B. Slade1
1Department of Biological Sciences, Macquarie University,
Sydney, NSW 2109, Australia.
2School of Botany & Zoology, Australian National
University, Canberra, ACT 0200, Australia.
Summary
Escherichia coli has long been used as an indicator
organism for water quality assessment. Recently
there has been an accumulation of evidence that sug-
gests some strains of this organism are able to pro-
liferate in the environment, a characteristic that would
detract from its utility as an indicator of faecal pollu-
tion. Phenotypic and genotypic characterization of E.
coli isolated from blooms in two Australian lakes, sep-
arated by a distance of approximately 200 km, identi-
fied that the blooms were dominated by three E. coli
strains. A major phenotypic similarity among the
three bloom strains was the presence of a group 1
capsule. Genetic characterization of a conserved
region of the cps gene cluster, which encodes group
1 capsules, identified a high degree of genetic varia-
tion within the bloom isolates. This differs from previ-
ously described encapsulated E. coli strains which
are highly conserved at the cps locus. The phenotypic
or genotypic profiles of the bloom strains were not
identified in 435 E. coli strains isolated from verte-
brates. The occurrence of these encapsulated strains
suggests that some E. coli have evolved a free-living
lifestyle and do not require a host in order to prolifer-
ate. The presence of the same three strains in bloom
events in different geographical regions of a temper-
ate climate, and at different times, indicates that free-
living E. coli strains are able to persist in these water
reservoirs. This study provides further evidence of
circumstances where caution is required in using E.
coli as an indicator organism for water quality.
Introduction
The faecal coliform Escherichia coli is commonly used to
assess recent faecal pollution of water bodies. When fae-
cal coliform bacteria are incapable of cell division in the
external environment, counts in a water body are propor-
tional to the amount of faecal input into that water body
(Barnes and Gordon, 2004). However, when cells are
capable of growth in the external environment, there is a
non-linear relationship between the rate of faecal inputs
into an environment and the faecal coliform cell densities
in that environment (Barnes and Gordon, 2004). In recent
years the characteristics which make E. coli the indicator
of choice have been questioned, as a number of studies
have demonstrated that E. coli may persist and multiply
in the external environment (Korhonen and Martikainen,
1991; Fujioka et al., 1999; Solo-Gabrielle et al., 2000;
Topp et al., 2003) in the apparent absence of faecal con-
tamination (Camper et al., 1991; Ashbolt et al., 1995; Gor-
don, 2001). The presence of E. coli in the external
environment in the absence of faecal contamination indi-
cates that at least some strains of E. coli may replicate in
environments external to a mammalian host.
Over the past 30 years, annual coliform blooms have
occurred in Lake Burragorang, the major water supply
reservoir in Sydney, Australia, and in Lake Burley Griffin,
a recreational lake in Canberra, Australia. During a bloom
event in 1983 in Lake Burragorang, E. coli serotype O8
was the predominant isolate detected (Ashbolt et al.,
1995). Characteristic of E. coli serotype O8 is the pres-
ence a group 1 capsule (Whitfield and Roberts, 1999).
Group 1 capsules are encoded by the cps gene cluster
that comprises four conserved loci wzi, wza, wzb and wzc
(Rahn et al., 1999). The cps cluster is located in the same
chromosomal region as the synthesis genes for gluconate
6-phospho dehydrogenase
homopolymer O-specific polysaccharides (rfb) (Nelson
and Selander, 1994; Whitfield and Roberts, 1999). The
cps, gnd and rfb gene organization in E. coli serotypes
O8 and O9 are similar to Klebsiella pneumoniae sero-
types O5 and O3 respectively (Whitfield and Roberts,
1999). DNA sequence data suggest that a recombination
event in which a block of Klebsiella spp. DNA, containing
the cps, gnd and rfb loci, was transferred to E. coli (Nelson
and Selander, 1994; Sugiyama et al., 1997; Rahn et al.,
(gnd) and mannose

Page 2

632M. L. Power et al.
© 2005 Society for Applied Microbiology and Blackwell Publishing Ltd, Environmental Microbiology, 7, 631–640
1999) with the end of the translocation close to the histi-
dine biosynthesis operon which maps next to rfb (Sug-
iyama et al., 1997).
In this study we describe the phenotypic and genotypic
characteristics of strains responsible for recent faecal
coliform blooms in two temperate Australian water bodies
(Lake Burragorang, a protected drinking water reservoir,
and Lake Burley Griffin, a recreational lake), and demon-
strate that three encapsulated strains of E. coli were
responsible for these blooms.
Results
Characteristics of bloom events
A coliform bloom is defined as a presumptive coliform
count in excess of 10 000 cfu per 100 ml. Blooms are
generally preceded by presumptive counts of less than
150 cfu per 100 ml. In Lake Burragorang bloom events
usually occur when water temperatures exceed 18∞C
(Ashbolt et al., 1995). In Lake Burley Griffin bloom events
occur from mid-summer to early autumn when the water
temperature is 18∞C or higher.
Detailed phenotypic and genotypic characterization
(see below) were made of strains sampled during blooms
occurring in April 2002, February 2003, April 2003 and
February 2004 in Lake Burley Griffin and from a bloom
occurring in Lake Burragorang during February 2003. The
results indicate that, in any single bloom event, from one
to three strains were responsible for the elevated coliform
counts. Furthermore, during a bloom event the strains
responsible for the elevated counts were present at most
of the sampling locations in Lake Burragorang and Lake
Burley Griffin.
Phenotypic characteristics of bloom strains
All strains isolated from bloom events exhibited a mucoid
colony appearance on MacConkey agar. Microscopic
examination of the bloom isolates using negative staining
and differential interference microscopy (DIC) identified a
large capsule surrounding individual cells (Fig. 1). Among
the bloom isolates, three main phenotypic profiles were
observed (Table 1). One of the phenotypic profiles (A-000)
matched the profile of E. coli. The second profile (A-010)
also resembled that of E. coli, but differed in being adonitol
positive and MUG negative. The third biochemical profile
(B1-001) differed from that of E. coli in several respects:
indole, lysine, melibiose and MUG negative. The pheno-
typic profiles A-000 and A-010 (Table 1) resulted in an E.
coli identification (> 94% confidence) with either the
API20E or BBL Enteric/non-fermentor identification sys-
tems. The phenotypic profile B1-001 resulted in a Citro-
bacter youngii identification with the API20E system.
Using the BBL Enteric/non-fermentor system the B1-001
profile yielded an E. coli or Shigella sonnei identification
depending on whether a particular isolate exhibited a pos-
itive arabinose reaction.
Genotypic characteristics of bloom strains
The method of Clermont and colleagues (2000) assigned
the bloom strains to either ECOR group A or ECOR group
1.5 mM
Fig. 1. Negative staining of Escherichia coli
bloom strain A-000 demonstrates the surround-
ing capsule. Scale bar 1.5 mm.

Page 3

Characterization of encapasulated blooming Escherichia coli633
© 2005 Society for Applied Microbiology and Blackwell Publishing Ltd, Environmental Microbiology, 7, 631–640
B1. The gene yjaA was detected in strains with the phe-
notypic profile A-010, but yjaA was not detected in isolates
with the profile A-000 (Table 2). Strains with the profile B1-
001 were assigned to ECOR group B1. All strains isolated
from blooms yielded a polymerase chain reaction (PCR)
product for the rfb cluster.
Given the mucoid appearance of the colonies when
grown on selective agar and subsequent identification of
Table 1. The phenotypic profiles of coliform strains isolated during bloom events in Lake Burley Griffin and Lake Burragorang and the typical phe-
notypic profile of Escherichia coli faecal isolates (Ewing, 1986).
Test or substrate
Bloom profile
A-010
Bloom profile
A-000
Bloom profile
B1-001% of faecal E. coli positive
Oxidase
ONPG
MUG
Indole
Voges-Proskauer
H2S
Glucose
Citrate
Arginine
Lysine
Mannitol
Inositol
Adonitol
Sorbitol
Rhamnose
Sucrose
Melibiose
Arabinose
–
+
+
+
–
–
+
–
–
+
+
–
+
+
+
–
+
+
–
+
–
+
–
–
+
–
–
+
+
–
–
+
+
–
+
+
–
+
–
–
–
–
+
–
–
–
+
–
–
+
+
–
–
+/–
0
ND
ND
98
0
0
100
0
21
93
100
1
4
98
89
54
94
100
ND, not determined.
Table 2. Genotypes of Escherichia coli strains with a group 1 capsule isolated during bloom events and from soil, water and sediment samples
from Australia.
Type of sample Locality
Collection
date
ECOR
groupa
ECOR
genotypeb
wzi
genotypec
gnd
genotyped
Bloom sample
Bloom sample
Bloom sample
Bloom sample
Bloom sample
Bloom sample
Bloom sample
Bloom sample
Non-bloom sample
Non-bloom sample
Non-bloom sample
Non-bloom sample
Non-bloom sample
Non-bloom sample
Non-bloom sample
Non-bloom sample
Non-bloom sample
Non-bloom sample
Non-bloom sample
Non-bloom sample
Non-bloom sample
Non-bloom sample
Lake Burley Griffin, ACTe
Warragamba Dam, NSWe
Lake Burley Griffin, ACTe
Warragamba Dam, NSWe
Lake Burley Griffin, ACT
Lake Burley Griffin, ACT
Lake Burley Griffin, ACT
Lake Burley Griffin, ACT
Lake Ginnderra, ACT
Lake Burley Griffin, ACT
Centennial Park, NSW
Demondrille (2) NSW
Lake Burley Griffin, ACT
Lake Burley Griffin, ACT
Tallowa Dam, NSW
Lake Gindebine, NSW
Lake Burley Griffin, ACT
Lake Ginnderra, ACT
Lake Ginnderra, ACT
Lake Manger, WA
Tallowa Dam, NSW
Warrambucca Creek, NSW
15/4/02
18/2/03
15/4/02
18/2/03
17/4/03
16/2/03
17/4/03
3/2/04
1/12/03
17/7/03
17/12/03
7/6/03
24/2/03
7/5/03
2/12/03
28/7/03
10/4/03
15/12/03
6/6/03
10/12/03
2/12/03
5/10/03
A
A
A
A
A
B1
B1
B1
A
A
A
A
B1
B1
B1
B1
B1
B1
B1
B1
B1
B1
000
000
001
001
000
001
001
001
010
010
010
010
001
001
001
001
001
001
001
001
001
001
54
54
33
33
54
65
65
65
33
NP
87
15
65
65
65
NP
65
NP
NP
NP
65
NP
312
312
332
332
312
NP
NP
NP
332
125
NP
334
NP
NP
NP
231
334
121
111
111
032
141
a. ECOR group assignment based on the method and key presented by Clermont et al. (2000).
b. First digit, presence (1) or absence (0) of the chuA gene; second digit, yjaA; third, DNA fragment TSPE4.C2.
c. Fingerprint pattern resulting from the digestion of the wzi PCR product with: first digit, Hinf1 and second digit, Taq1. NP, no PCR product
detected.
d. Fingerprint pattern resulting from the digestion of the gnd PCR product with: first digit, kpn1, second digit, hinf1 and third digit, Rsa1. NP, no
PCR product detected.
e. Strains used for partial sequencing of the cps gene cluster.

Page 4

634M. L. Power et al.
© 2005 Society for Applied Microbiology and Blackwell Publishing Ltd, Environmental Microbiology, 7, 631–640
capsules by microscopy, genetic characterization of the
capsule-encoding genes (cps gene cluster) was under-
taken. Nucleotide sequencing of the cps region of the
three bloom strains identified three genetic profiles. A
comparison of five known wzi loci from E. coli and K.
pneumoniae showed the genes are quite conserved at the
amino acid level with 1.3–5.5% variation (Table 3). In con-
trast, the wzi gene shows a large number of nucleotide
changes (Table 3). Restriction fragment length polymor-
phism (RFLP) analysis of the 1200 bp wzi PCR product
from strains recovered from blooms revealed three finger-
print patterns based on the digestion of the DNA product
with Hinf1 and Taq (Table 2). The primer pair designed to
amplify a 1200 bp fragment of the gnd produced a DNA
product for all ECOR group A bloom isolates, but not for
any of the ECOR group B1 bloom isolates. Digestion of
the gnd PCR product with Hinf1, Rsa1 and Kpn1 revealed
two fingerprint patterns among the ECOR group A iso-
lates (Table 2).
The virulence factor fimH was detected in the ECOR
group A bloom strains, but not in the ECOR group B1
bloom strains. None of the other virulence-associated
genes were detected in any of the bloom strains.
O antigen characteristics
The PCR primers designed to amplify a region of the wzm
gene of E. coli serotype O8 failed to produce a product
for any of the Lake Burragorang or Lake Burley Griffin
isolates. However, PCR products were obtained for three
serotype O8 isolates used as controls. The primers
designed for serotype O9-specific PCR amplified a
1000 bp product for isolates with the bloom profile A-000
and B1-001. The nucleotide sequences for these isolates
were 99% similar to K. pneumoniae serotype O3 (Gen-
Bank Accession No. AB010296) and 92% to 95% identity
to E. coli serotype O9 isolates (GenBank Accession No.
AB010293, AB010294 and ECORFBM). No PCR prod-
ucts were amplified for bloom profile A-010 isolates using
the wzm O8 and O9 primers. Nucleotide sequencing of a
3000 bp fragment, which spanned the manBC, rfb and
gnd loci, of one of these isolates, identified a large dele-
tion within the rfb cluster located 132 bp 5¢ from the start
of the wbdC gene. The 3000 bp fragment contained two
regions with close homology to E. coli strain E69 (Gen-
Bank Accession No. U90519). Immediately 5¢ to the dele-
tion site was a transposase gene (insB) and an IS1
transposable element (insA). The nucleotide sequence of
the gnd gene (1167 bp) was 96% identical to E. coli strain
E69. Relative to the sequence of E. coli E69, this gene
order is consistent with a 7000 bp deletion spanning the
duplicated manC2 and manB2 genes and in the rfb locus
wzm, wzt, wbdD, wbdA and wbdB genes.
The identity of bloom strains
A neighbour-joining tree was constructed by concatenat-
ing the sequence data for gapA (611 nucleotides) and pgi
(292 nucleotides) of strains identified as E. coli, C. freundii
and Salmonella enterica together with a subset of the
bloom strains, using a strain of K. pneumoniae as an
outgroup. The tree reveals that all of the bloom strains
included in the analysis clearly fall within the E. coli clade
(Fig. 2). Furthermore, there is high bootstrap support for
all of the nodes that define strains of the same species.
Frequency of translocated rfb locus in E. coli
A total of 535 strains of E. coli isolated from soil, water
and sediment samples as well as humans and animals
were screened for the presence of the rfb locus. Overall,
a PCR product was observed in 9.2% of the isolates. The
probability of detecting the rfb product depended on the
source of the isolates (environment, human or vertebrate)
and the isolate’s ECOR group membership (Fig. 3) (logis-
tic regression: source, P = 0.76; ECOR group, P < 0.0001;
source * ECOR group interaction, P < 0.011).
Characteristics of environmental isolates
The rfb PCR product was detected in 19 of the 90 E. coli
strains isolated from soil, water or sediments. Strains iso-
lated during bloom events were not included among the
90 environmental isolates. Restriction fragment length
polymorphism analysis of the wzi PCR products from
these environmental isolates revealed four fingerprint pat-
terns based on the digestion of the DNA product with
Hinf1 and Taq (Table 2). Digestion of the gnd PCR product
with Hinf1, Rsa1 and Kpn1 revealed eight fingerprints
among these non-bloom strains (Table 2). Isolates with
wzi and gnd fingerprints and biochemical profiles identical
to the bloom strains A-010 were detected in Lake
Ginnderra, another artificial lake in the Canberra area
(Table 2). Isolates with a wzi fingerprints and biochemical
profiles identical to the bloom strains B1-001 were
detected during non-bloom periods in Lake Burley Griffin
and in Tallowa Dam, NSW, a reservoir in the Sydney
region (Table 2).
Table 3. Per cent identity of nucleotide and per cent similarity of
amino acid sequences (highlighted grey) of the wzi region of the E.
coli bloom strains (bloom profile A-000, A-010 and B1-001), K. pneu-
moniae (KPNCPS) and E. coli (AF104912).
KPNCPSAF104912A-000A-010B1-001
KPNCPS
AF104912
A-000
A-010
B1-001
100
91.8
90.5
86.5
88.2
97.8
100
90.5
87.4
89.2
97.4
98.4
100
87
88.3
94.5
94.9
94.5
100
86.1
96.2
97.1
96.6
93.5
100

Page 5

Characterization of encapasulated blooming Escherichia coli635
© 2005 Society for Applied Microbiology and Blackwell Publishing Ltd, Environmental Microbiology, 7, 631–640
Characteristics of vertebrate isolates
The rfb PCR product was detected in 33 of the 445 E. coli
strains isolated from vertebrates. No wzi PCR product was
detected in 23 of the strains producing a rfb PCR product.
For the balance of the strains, RFLP analysis of the wzi
DNA fragment from these vertebrate isolates revealed
seven fingerprint patterns based on the digestion of the
DNA product with Hinf1 and Taq. Of the 33 rfb positive
strains, no gnd PCR product was detected for one strain.
Of the 34 strains producing a gnd PCR product, digestion
with Hinf1, Rsa1 and Kpn1 revealed 24 different finger-
prints among the E. coli strains isolated from vertebrates.
Two of the strains isolated from vertebrates, one from
a human and one from a Tasmanian Devil, Sarcophilus
harrisii, had wzi and gnd fingerprints identical to the A-
010 bloom strains. Although the human isolate was an
ECOR group A strain, the isolate from a Tasmanian Devil
belonged to ECOR group B1. Neither of the strains from
vertebrates were adonitol positive, unlike the A-010 bloom
strains. Furthermore, virulence factor profiling revealed
that the group A strain isolated from the human was papA,
fimH, traT, fyuA, iron, ompT, iss and iutA positive. The
group B1 strain from a Tasmanian Devil was found to be
fimH, traT, fyuA, iroN, cvaC, iss and iutA positive. By
comparison, the A-010 bloom strains were found to be
fimH positive only.
Discussion
Escherichia coli is considered a commensal inhabitant of
the lower intestinal tract of mammals (Gordon and Cowl-
ing, 2003). The E. coli bloom strains described in this
study were isolated from lakes in the absence of faecal
input and appear to be essentially free living. The most
significant phenotypic characteristic identified in the three
E. coli strains associated with bloom events in Lake Burley
Griffin and Lake Burragorang is that they are encapsu-
lated. During previous blooms in Lake Burragorang (1978/
1979 and 1981) the predominant E. coli strains were also
Fig. 2. A neighbour-joining tree was constructed by concatenating
the sequence data for gapA (611 nucleotides) and pgi (292 nucle-
otides) of Australian strains identified as Escherichia coli, Citrobacter
freundii and Salmonella enterica together with a subset of the bloom
strains, using a strain of Klebsiella pneumoniae as an outgroup.
Numbers above the nodes define the level of bootstrap support (%)
for the nodes defining species. Strain numbers with prefix starting
with B represented isolates taken from birds; R, from reptiles; M, from
mammals; and E, from soil, water or sediments.
CIFR B1119
CIFR B960
CIFR B392
CIFR M309
CIFR MISC250
ESCO B108
ESCO M109
ESCO M237
ESCO B25
ESCO E259
ESCO E704
ESCO M767
ESCO B30
ESCO B91
ESCO R304
ESCO E256
ESCO E703
ESCO E258
ESCO E482
ESCO R345
ESCO E267
ESCO E270
ESCO E323
ESCO E361
ESCO E459
ESCO R1
ESCO R209
ESCO R87
ESCO M733
ESCO B90
ESCO M86
SAEN MISC309
SAEN R159
SAEN MISC384
SAEN R31
SAEN R291
KLPN B178
Bloom A-000
100
100
99
Bloom A-010
Bloom B1-001
0.005 substitutions/site
Fig. 3. Frequency with which the Klebsiella translocated rfb PCR
product was detected in Escherichia coli strains isolated from the
environment (soil, sediments and water), from humans (Canberra,
ACT residents) and Australian non-human vertebrates (Gordon and
Cowling, 2003) with respect to the ECOR group membership. Num-
bers above each bar denote the number of strains tested.
0.00
0.05
0.10
0.15
0.20
0.25
0.30
A B1 B2DA B1B2DAB1B2D
Proportion of strains with a Type 1 capsule
Environment Human Vertebrate
54
51
97
63
44
47
45
44
9
62
5 14