Advances in Microbiology, 2012, 2, 252-262
http://dx.doi.org/10.4236/aim.2012.23030 Published Online September 2012 (http://www.SciRP.org/journal/aim)
Detection and Persistence of Clinical Escherichia coli
in Drinking Water Evaluated by a Rapid Enzyme Assay
and qPCR
Annette S. Bukh, Nina E. Hansen, Peter Roslev*
Department of Biotechnology, Chemistry and Environmental Engineering, Aalborg University, Aalborg, Denmark
Email: *pr@bio.aau.dk
Received February 28, 2012; revised April 9, 2012; accepted May 7, 2012
ABSTRACT
The aims of this study were to evaluate two methods, qPCR and a chemiluminescent assay (ColiLight II), for rapid de-
tection of E. coli in water, and to examine the survival and persistence of clinical E. coli in drinking water and biofilm
using qPCR and ColiLight II. qPCR and ColiLight II were compared with a cultivation-based method (MPN), and sur-
vival and persistence of four clinical E. coli strains in water and biofilms on stainless steel (SS) and polyethylene (PE)
surfaces were studied in a flow-through reactor with non-disinfected drinking water using ColiLight II, qPCR, ATP
bioluminescence, and MPN. ColiLight II and qPCR correlated well with MPN. In drinking water, some clinical E. coli
strains showed prolonged survival in drinking water flow-through systems, and persisted 3 - 3.4 times longer than the
theoretical washout due to incorporation into biofilms. Strain specific attributes can significantly affect detection and
persistence of E. coli in drinking water matrices.
Keywords: Biofilm; Clinical E. coli; Chemiluminescent Assay; Lab-Scale Drinking Water Reactor System; Drinking
Water; qPCR
1. Introduction
Escherichia coli is widely used as an indicator of fecal
pollution when monitoring the microbial quality of drink-
ing water [1]. Standard methods for detection of E. coli
in drinking water are based on cultivation which requires
an analysis time of 18 - 48 h or even longer for confirma-
tion steps. In case of contamination situations, rapid me-
thods are needed to redress the problem [2].
Various rapid, fluorescent enzyme-based methods for
detection of total and fecal coliforms based on β-D-ga-
lactosidase and β-D-glucuronidase activity, respectively,
have been developed [3-7]. Additional, chemilumines-
cent-based detection of E. coli on membrane filters has
been proposed [8], and later, the same type of substrate
was used to detect low concentrations of stressed E. coli
in non-contaminated drinking water [9]. Chemilumines-
cence is more sensitive compared to fluorescence and
absorbance, and the higher sensitivity is utilized to de-
crease the required analysis time of the detection method.
Another rapid method to detect E. coli is quantitative
real-time PCR (qPCR). qPCR is a highly sensitive and
very specific detection method. Due to this, a cultivation
step can be omitted when applying qPCR for detection of
microorganisms. This is a great advantage for the use in
microbial ecological studies where the majority of target
organisms cannot be cultivated in the laboratory. How-
ever, prior to successful amplification, sampling strate-
gies and DNA extraction procedures must be optimized
in order to achieve representative quantification. qPCR
has previously been applied to detect E. coli in different
water types such as groundwater, sea water, freshwater,
and roof-harvested rainwater [10-12]. Clinical strains of
E. coli are rarely included in drinking water studies, but
from an epidemiological and a health perspective point
of view, it is more relevant to study these strains since
they have caused diseases in humans.
In this study, we applied the ColiLight II method and a
qPCR assay for detection of clinical E. coli in two dif-
ferent non-contaminated drinking water matrices. The
ColiLight II method has been optimized and validated
previously [9], and primers for detection of a part of the
1-deoxy-D-xylulose 5-phosphate synthase gene (dxs) in
E. coli was described previously [13]. In our study, we
optimized and validated the dxs qPCR assay regarding
annealing temperature, primer concentration, assay rep-
resentativeness, and lower level of quantification (LLOQ).
The two rapid methods were applied to detect planktonic
and sessile clinical E. coli strains in a lab-scale drinking
*Corresponding author.
C
opyright © 2012 SciRes. AiM
A. S. BUKH ET AL. 253
water reactor system during a period of washout in order
to examine the survival and persistence of clinical E. coli
strains in non-contaminated drinking water.
2. Materials and Methods
2.1. E. coli Strains and Growth Conditions
Four clinical E. coli strains were included in this study: E.
coli ATCC 25922, a clinical isolate originated from
Seattle, USA, 1946 (American Type Culture Collection;
Rockville, MD, USA), E. coli UTI CAB, a clinical strain
isolated from blood from a Danish female hospital pa-
tient with community-acquired bacteremia (CAB) and
the urinary tract as the primary site of infection, Hjørring,
Denmark, 2001 (Department of Clinical Microbiology,
Aalborg Hospital, Aarhus University Hospital, Denmark),
E. coli Gall CAB, a clinical strain isolated from blood
from a Danish male hospital patient with CAB and the
gallbladder as the primary site of infection, Hjørring,
2005 (Aalborg Hospital, Denmark), and E. coli O177:H-,
a clinical strain classified as attaching and effacing E.
coli (A/EEC) isolated from feces from a Danish hospital
patient (Department of Clinical Microbiology, Skejby
Hospital, Aarhus University Hospital, Denmark). In ad-
dition, E. coli strain ED1a was included for method vali-
dation (Institut national de la santé et de la recherche
médicale, Paris, France) [14]. All E. coli strains were
grown in Fluorocult® LMX Broth (LMX; Merck, Darm-
stadt, Germany) for 18 h at 37˚C on a shaking table.
2.2. Drinking Water
The drinking water used in this study consisted of mu-
nicipal tap water collected in the cities of Aalborg and
Aarhus, Denmark (Table 1).
Table 1. Distribution and most recent bacteriological drink-
ing water analyses from two Danish waterworks (Anon
2010). At Danish waterworks, the threshold values of cul-
turable total flora [CFU·ml1] at 22 and 37˚C are 50 and 5,
respectively, and coliforms may not be detected in a 100 ml
water sample.
Municipality Aalborg Aarhus
Waterworks Engkilden Truelsbjergvaerket
Distribution [m3·year–1] 1,300,000 2,300,000
Disinfection No No
Culturable microorganisms 22˚C
[CFU·ml–1] 1 1
Culturable microorganisms 37˚C
[CFU·ml–1] <1 <1
Coliform bacteria [CFU·100·ml–1] <1 <1
E. coli [CFU·100·ml–1] <1 <1
The tap water originates from non-contaminated ground-
water, and no disinfection is used before distribution. For
the persistence experiments, water from Aalborg was
used, whereas both water types were used for comparison
of detection methods.
2.3. Enumeration of Culturable Microorganisms
in Drinking Water
Enumeration of the total amount of culturable micro-
organisms in drinking water was performed according to
DS 6222-1 [15]. In brief, colonies formed after incuba-
tion in a nutrient agar culture medium were determined
after growth at 22˚C and 37˚C for 68 ± 4 and 44 ± 4 h,
respectively.
Enumeration of culturable E. coli was based on a most
probable number (MPN) method using 96-well Master-
Blocks (Greiner Bio-One GmbH, Frickenhausen, Ger-
many). Four to seven samples were serially diluted in
Colilert-18 broth (IDEXX Laboratories, Inc., Westbroo-
ke, ME, USA). One row was used as reference, and the
MasterBlocks were incubated for 18 - 22 h at 37˚C. Sub-
sequently, the number of yellow wells that fluoresced
during illumination of the MasterBlocks with UV-light
(365 nm) was considered positive. Based on these counts,
MPN was calculated using the Bacteriological Analytical
Manual-MPN spreadsheet based on Thomas’s approxi-
mation [16].
2.4. ATP Bioluminescence Assay
The total microbial biomass was determined by measur-
ing ATP bioluminescence using a R&D Biomass Test
Kit (Promicol BV, Nuth, The Netherlands) in white 24-
well microplates (CulturPlates; PerkinElmer, Inc., Wal-
tham, MA, USA). Briefly, microbial ATP was released
from all cells present in 0.5 ml of sample by adding 50 µl
of lysis reagent. 50 µl of a reagent containing luciferin
and luciferase was added. Luciferin was hydrolyzed by
luciferase in the presence of oxygen and microbial ATP.
The released bioluminescence was measured for 1 sec in
a Victor X2 Multilabel Plate Reader (PerkinElmer, Inc.).
Autoclaved ddH2O was used as reference.
2.5. ColiLight II
ColiLight II is a chemiluminescent assay used for detec-
tion of viable E. coli [9]. An outline of the ColiLight II
principle is given below (Figure 1). 3 - 5 ml of LMX
broth supplemented with 0.2 mg·ml–1 of methyl-β-D-
glucuronide (MetGlu; Sigma, St. Louis, MO, USA) to
induce expression of β-D-glucuronidase were added to
each sample [17], which were incubated for 6 h at 37˚C
on a shaking table. Subsequently, the samples were cen-
trifuged for 15 min at 5000 rpm. Pellets were dissolved
in 0.5 ml of supernatant, transferred to white 24-well
Copyright © 2012 SciRes. AiM
A. S. BUKH ET AL.
254
Figure 1. Outline of the rapid detection principles, ColiLight
II and qPCR, for detection of E. coli in drinking water
matrices.
CulturPlates, and incubated for 8 - 10 min at room tem-
perature in the presence of 0.5 mg·ml–1 of protamine
(from salmon; Sigma) to permeabilize cells [18]. Subse-
quently, 200 µl of sβ-Glucor 102 (Michigan Diagnostics
LLC, Royal Oak, MI, USA) diluted 100-fold in 0.1
mol· l –1 NaPO4 buffer, pH 7.2, containing 10 mmol·l–1
EDTA and 15 mmol·l–1 glycine was added to each sam-
ple and incubated for 30 min at 27˚C. 400 µl of Trigger-
ing Reagent (Michigan Diagnostics LLC) was added to
each sample to reduce quenching of chemiluminescence
intensity. The plate was shaken for 10 sec, and the
chemiluminescence was measured for 1 sec in a plate
reader. Drinking water without coliforms was used as
reference.
A standard curve was prepared using serial dilutions of
β-D-glucuronidase purified from E. coli (Sigma) in 0.1
mol· l –1 NaPO4 buffer, pH 7.2. S/N ratios were calculated
[19], and the detection threshold (DT) was determined as
5SD +arithmetic meanarithmetic meanusing seven
blanks [20].
2.6. qPCR
An outline of qPCR-based detection of E. coli based on
the amplification of a part of dxs is given (Figure 1).
2.6.1. D NA Extracti on
Pellet and filter samples were transferred to 2 ml bead
tubes containing 0.1 mm glass beads (MOBIO Laborato-
ries, Inc., Carlsbad, CA, USA). Drinking water without
coliforms was used as extraction controls. 1 ml of DNA
extraction buffer (50 mmol·l–1 NaCl, 50 mmol·l–1 Tris-
HCl, 50 mmol·l–1 EDTA, 5% SDS, pH 8.0) was added to
each sample together with 1 µl of 1 mol·l–1 of DTT. The
samples were bead beaten for 5 min on a Vortex Genie 2
with adapter (MOBIO), and centrifuged for 1 min at
10,000 × g. Filters were removed, and DNA was ex-
tracted using half a volume of Phenol:Chloroform:Isoa-
myl alcohol 25:24:1, saturated with 10 mmol·l–1 Tris-
HCl, pH 8.0, 1 mmol·l–1 EDTA (Sigma) followed by one
volume of chloroform.
DNA was precipitated on ice for at least 1 h at –20˚C
using 0.7 volume of isopropanol and a final concentra-
tion of 0.3 mol·l–1 sodium acetate in the presence of 1 µg
of glycogen. The samples were centrifuged for 30 min,
and DNA pellets were washed with 0.5 ml ice-cold 70%
ethanol. After ethanol removal, DNA was resuspended in
25 µl of nuclease free water (Sigma). Templates were
stored at –80˚C prior to use.
2.6.2. dxs Assay
A part of dxs was amplified using previously described
primers [13]. The qPCR assay was optimized regarding
concentration of Mg2+, primer concentration, and an-
nealing temperature. qPCRs were carried out in tripli-
cates in an Mx3000PTM Quantitative PCR system (Strata-
gene, Agilent Technologies, CA, USA).
The 20 µl reaction mixtures contained 10 µl of Bril-
liant® II SYBR QPCR Low ROX Master Mix (Strata-
gene), 0.2 µmol·l–1 of the forward primer, 0.3 µmol·l–1 of
the reverse primer, a final concentration of 3 mmol·l–1 of
MgCl2, and 4 µl of template. The PCR conditions were
enzyme activation at 95˚C for 10 min followed by 35
cycles of denaturation at 95˚C for 30 sec and amplifica-
tion/extension at 62˚C for 1 min (data collection). Sub-
sequently, a melt curve analysis was done from 55˚C to
95˚C. A standard curve was prepared using serial dilu-
tions of E. coli ED1a gDNA in nuclease free water
(Sigma). The DNA stock concentration was based on
quantification of cells using McFarland standards, viable
microscopy counts using a counting chamber, and plate
counts on Tryptone Bile X-glucuronide Agar (Merck),
and on DNA concentrations measured on a UV-Vis
nanodrop spectrophotometer ND-1000 (NanoDrop
Technologies, Inc.; Wilmington, Delaware, USA).
LLOQ was determined for the qPCR assay. Intraassay
precision (expressed as coefficient of variation, CV)
was calculated from standard deviations of genome
copy numbers obtained from five replicates of nine dif-
ferent template concentrations run simultaneously. CV
for interassay precision was calculated from standard
deviations of genome copy numbers obtained from
Copyright © 2012 SciRes. AiM
A. S. BUKH ET AL. 255
five replicates of one template concentration at three
different days.
2.7. Comparison of Detection Methods
Correlation analyses were performed in order to compare
the outcome of the rapid detection methods ColiLight II
and qPCR with more conventional MPN enumeration.
The analyses were done using two different types of
drinking water collected from Aalborg and Aarhus mu-
nicipalities in Jutland, Denmark (Table 1). For each lo-
cation, 100 ml of drinking water was transferred to sterile
250-ml BlueCap bottles and spiked with concentrations
between ~102 and 107 CFU 100 ml1 of E. coli ATCC
25922, E. coli UTI CAB, E. coli Gall CAB, and E. coli
O177:H-. The samples were incubated for 24 h at 10˚C
on a shaking table. Subsequently, 33.3 ml of each sample
was filtrated through a 0.45 µm nylon membrane filter
and 33.3 ml through a 0.45 µm mixed cellulose ester
(MCE) membrane filter (47 mm; Frisenette, Knebel,
Denmark). The nylon filter was used for qPCR analysis,
and the MCE filter was used for ColiLight II analysis. In
addition, 4 × 1 ml of each water sample was used for
MPN analysis.
2.8. Persistence of E. coli in a Flow-Through
Drinking Water Reactor System
Persistence and washout of E. coli ATCC 25922, E. coli
UTI CAB, E. coli Gall CAB, and E. coli O177:H- was
studied in a lab-scale drinking water reactor system
(Figure 2). The system was comprised of a water reser-
voir (5 - l BlueCap bottle) connected to a CDC Biofilm
reactor (Biosurface Technologies corp., Bozeman, MT,
USA) via a peristaltic pump (U4-8R Midi 2.5 - 50 rpm;
Alitea AB, Stockholm, Sweden) with PVC tubes (116-
0549-19; MikroLab Aarhus A/S, Aarhus, Denmark). A
PVC pump tube (Yellow-Blue, 1.52 mm; SEAL Ana-
lytical Ltd., Hampshire, UK) was used to ensure a uni-
form water flow of 0.4 ml·min–1. The water was pumped
from the reservoir to the reactor with a total water vol-
ume of 350 ml resulting in a dilution rate, D, of 1.646
d–1.
The rector was fitted with PE and SS coupons to allow
studies of cell adherence to different surface materials.
The bioreactor was comprised of glass, and the PE and
SS coupons were inserted into coupon holders made of
polypropylene (PP). The total surface area of the inner
surfaces below the water surface during each experiment
was 388.6 cm2 of which 196 cm2 was constituted by
glass, 142 cm2 by PP, 25.3 cm2 by PE, and 25.3 cm2 by
SS. From the reactor, the water flowed through to a
waste bottle. Seven removable PP coupon holders each
containing three removable coupons were inserted in the
reactor. Three holders contained three PE coupons, three
Figure 2. Experimental set-up of a flow-through drinking
water reactor system used in the persistence experiments.
holders contained three SS coupons, and one holder con-
tained one PE and one SS coupon. The coupons were
below the water surface throughout the experiment. The
reservoir and the reactor were placed on magnetic stirrers.
Four experiments were performed, one for each E. coli
strain. Prior to each experiment, clean drinking water
was run through the system for 72 h. Subsequently, the
reservoir and the reactor were inoculated with ~104 ml–1
E. coli cells. The reactor system was exposed to the E.
coli cells for 24 h, followed by nine days of washout
where clean drinking water without E. coli was run
through the system. All experiments were conducted in a
darkroom at 10˚C. All reactor parts and tubes were dis-
infected with 70% of ethanol between experiments.
Sampling of water and coupons were performed prior
to E. coli inoculation, 24 h after E. coli inoculation at the
time where washout was initiated (day 0), and after 3 and
9 days of washout (day 3 and day 9, respectively). Addi-
tional sampling of the water phase was also carried out
after 6 days of washout (day 6). A total volume of 10 ml
of water was collected at each sampling. Using a glass
pipe, increments of 2 ml of water representing the verti-
cal dimension of the water column were drawn from the
reactor and pooled. One holder with three PE coupons
and one holder with three SS coupons were also drawn
from the reactor. Each of the three coupons from one
holder were each aseptically placed in 3.33 ml of buff-
ered peptone water in sterile 50 ml Greiner tubes, and to
release attached cells from the coupons, the samples were
vortexed for 10 sec followed by sonication for 10 sec
(×3). The coupons were removed, and the eluates from
the same coupon material were pooled. 7 × 0.2 ml of wa-
ter and coupon eluates were used for MPN determina-
tions, 3 × 0.5 ml were used for both ATP biolumines-
cence, 3 × 0.5 ml were used for ColiLight II analyses,
Copyright © 2012 SciRes. AiM
A. S. BUKH ET AL.
256
and pellets from 3 × 0.8 ml of each sample were used for
DNA extraction followed by qPCR.
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
-12 -11 -10-9-8-7-6-5
y=7.571+0.809x; R2=0.982
Log(S /N)
Log(Concentration of b-D-glucuronidase [mol l
-1])
2.9. Data Analysis
Sample coefficients of determination, R2, and Pearson’s
product moment correlation coefficients, rP, were calcu-
lated to examine the linearity between datasets. QQ-plots
were made to confirm that the data were normally dis-
tributed, and removal rates were compared using one-
way ANOVA tests at 95% confidence level using Bon-
ferroni corrections. SPSS Statistics 17.0 was used for
statistical analyses (SPSS Inc, Chicago, IL, USA). Re-
moval rates were determined as the slope of the regres-
sion line
[mol·l
-1
]

0
C Cktln
(1)
where C is the bacteria concentration at time t, C0 the
initial concentration on day 0, and k the removal rate in
d–1.
3. Results
3.1. Detection of E. coli Using ColiLight II and
qPCR
The suggested rapid detection methods showed good
linear responses (Fi gure 3).
For ColiLight II, a standard curve was prepared using
known concentrations of purified β-D-glucuronidase
from E. coli (Figure 3(a)). A positive linear correlation
between log transformed enzyme concentrations and log
transformed S/N ratios was observed in a ~3-log interval
from ca. 10–9 to ca. 10–6 mol·l–1 of β-D-glucuronidase (R2
= 0.982). For qPCR, a standard curve using known con-
centrations of E. coli ED1a showed a positive log linear
correlation of E. coli genome copy number and PCR
threshold cycle number over a ~6-log interval from ~2 ×
101 to ~6 × 10
6 genome copies (Figure 3(b)), hence
LLOQ was determined to ~23 genome copies. R2 was
0.998, and the PCR efficiency, E, was 109.6%. CV for
repeatability was found to be 13.9%, and 23.2% for re-
producibility. R2 and E for all runs ranged from 0.997 to
0.998 and 92.4% to 109.6%, respectively, which was
considered satisfactory [21].
3.2. Correlation between MPN and Rapid
Methods
In order to examine whether ColiLight II and qPCR can
be used as alternatives or supplements to traditional cul-
tivation dependent methods, linear regression analyses
were performed. All three methods were applied to detect
E. coli ATCC 25922, E. coli UTI CAB, E. coli Gall CAB,
and E. coli O177:H- after incubation for 24 h at 10˚C in
two different types of drinking water (Table 1). The total
(a)
0
5
10
15
20
25
30
35
10-2 100102104106108
y=35.08-3.111log(x); R
2= 0.998; E=109.6%
CT (dRn)
Initial genome copies
(b)
Figure 3. Standard curves for E. coli based on (a) ColiLight
II by emission of chemiluminescence where the dashed line
indicates the detection threshold (n = 3); and (b) qPCR by
on amplification of a part of dxs (n = 5).
flora measured as culturable microorganisms in both wa-
ter types were below 200 CFU ml–1 at 22˚C and below 50
CFU ml–1 at 37˚C.
Linear regression analyses between the MPN enu-
meration and ColiLight II and MPN and qPCR in two
different water types showed positive linear relationships
(Table 2).
From the coefficient R2, we see that most of the varia-
tion (71% - 98%) was described for each isolate and each
drinking water type using each of the two rapid methods
in comparison with the MPN method.
However, when the results from each strain were com-
piled, the explained variance decreased (65% - 82%)
indicating that the biological variation had a higher im-
pact on the results than both the applied method and the
type of drinking water used, especially for the ColiLight
II method based on metabolic activity. From the correla-
tion coefficients, we see a strong linear relationship (rP
0.74) in the results obtained from MPN and ColiLight II
Copyright © 2012 SciRes. AiM
A. S. BUKH ET AL. 257
Table 2. Sample coefficients of determination, R2, and Pea-
rson’s product moment correlation coefficients, rP, between
results obtained by MPN and ColiLight II and by MPN and
qPCR for two drinking water types using four different
strains of E. coli.
E. coli ColiLight II qPCR
Aalborg Aarhus Aalborg Aarhus
R
2 r
P R
2 r
P R
2 r
P R
2rP
ATCC 25922 0.82 0.91 0.88 0.96 0.87 0.93 - -
UTI CAB 0.93 0.95 0.74 0.90 0.94 0.97 0.91 0.95
Gall CAB 0.71 0.78 0.92 0.89 0.93 0.96 0.72 0.85
O177: H- 0.75 0.87 0.92 0.74 0.98 0.99 0.95 0.97
All strains 0.65 0.81 0.82 0.89 0.71 0.85 0.69 0.83
and from MPN and qPCR.
The degree of correlation between MPN and ColiLight
II and between MPN and qPCR are illustrated for drink-
ing water from both Aalborg and Aarhus municipalities
(Figures 4(a) and (b), respectively). Here we see that a
positive, linear relationship between MPN and the two
rapid methods also exists when results from both types of
drinking water are used. The correlation coefficients for
MPN and ColiLight II and MPN and qPCR were 0.72
and 0.82, respectively, indicating a strong relationship
between the outcomes of the two methods. The explained
variations were below 70% again indicating the impor-
tance of biological variation.
3.3. Persistence of E. coli in a Flow-Through
Drinking Water Reactor System
Persistence of E. coli ATCC 25922, E. coli UTI CAB, E.
coli Gall CAB, and E. coli O177:H- in a flow-through
drinking water reactor system was examined over a time
period of nine days. The system was initially operated for
72 h with clean drinking water to allow initiation of a
conditioning film on the inner surfaces, and then inocu-
lated with a specific E. coli strain for 24 h in a flow-
through mode. Subsequently, the system was run with
clean drinking water to facilitate E. coli washout. Changes
in the total microbial flora in water and biofilms were
followed using an ATP assay (Table 3).
At day 0, accumulation of biomass was detected on the
PE and SS surfaces. After 9 days of washout, biomass on
both surface types had increased in terms of higher con-
centration of ATP compared to day 0 whereas the ATP
concentration in the water phase had decreased.
3.3.1. E. coli in Biofilm on PE and SS Surfaces
During the 24 h of contamination, all four E. coli strains
did accumulate on PE and SS surfaces within the biofilm
reactor (Figure 5). At day 0, higher levels of E. coli were
ATCC 25922
UTI CAB
Gall CAB
O177:H-
8
6
4
2
0
R
2
= 0.52
Log (RLU/mL)
0 2 4 6 8
Log (MPN/mL)
(a)
ATCC 25922
UTI CAB
Gall CAB
O177:H-
R
2
= 0.67
Log (MPN/mL)
8
6
4
2
0
Log (genome copies/mL)
0 2 4 6 8
(b)
Figure 4. Correlation curves between (a) MPN and Coli-
Light II and between (b) MPN and qPCR for two water
types. Black: Aalborg water. Grey: Aarhus water.
Table 3. Biofilm formation on PE and SS surfaces measured
as ATP bioluminescence at day 0 and day 9 in three lab-
scale drinking water reactor system experiments (mean
RLU ± SE; n = 3).
Sample Day 0 Day 9
Water [RLU·ml–1] 822 ± 513 191 ± 64
PE coupons [RLU·cm–2] 3404 ± 1288 10,917 ± 2828
SS coupons [RLU·cm–2] 2089 ± 769 5842 ± 1063
detected on PE surfaces compared to SS surfaces using
both the MPN method and qPCR where the latter method
detected ca. 10 fold higher levels compared to MPN. The
levels of total E. coli detected on PE and SS surfaces by
qPCR were similar during the period of washout whereas
a decrease in culturable E. coli was observed by MPN.
1.45% of the initial concentration of culturable E. coli
Copyright © 2012 SciRes. AiM
A. S. BUKH ET AL.
AiM
258
ATCC 25922 in the water phase (MPN·ml–1) had accu-
mulated in the indigenous biofilm on the PE surface
(MPN·cm–2) at day 0, and for the strains UTI CAB, Gall
CAB, and O177:H-, these numbers were 1.79%, 0.13%,
and 1.04%, respectively (Figure 5(a)). Less E. coli cells
attached to the SS surface compared to the PE surface
(Figures 5 (a) and (c)).
Copyright © 2012 SciRes.
0.4% of the initial concentration of culturable E. coli
ATCC 25922 in the water phase (MPN·ml–1) was found
on the SS surface (MPN·cm–2) at day 0, and for UTI
CAB, Gall CAB, and O177:H- these numbers were 0.3%,
0.03%, and 0.45%, respectively (Figure 5(c)).
On SS and PE surfaces, the number of culturable E.
coli ATCC 25922 decreased 10 to 20-folds and UTI
CAB decreased 10 to 40-folds during the first three days
of washout. The number of culturable O177:H- did not
change significantly on either PE or SS surfaces. On PE,
the number of culturable Gall CAB increased 4-fold in
the same period whereas a 200-fold decrease was ob-
served on SS. After 9 days of washout, low numbers of
all four E. coli strains could be detected on the PE sur-
faces using MPN whereas only E. coli UTI CAB and
O177:H- could be detected on SS.
β-D-Glucuronidase activity could be detected in all
strains at day 0 on both PE and SS surfaces. After three
days of washout, 16% of the initial enzyme activity was
detected in Gall CAB on PE and below 0.3% in UTI
CAB and O177:H-. On SS, only E. coli Gall CAB could
be detected at day 3. At day 9, no β-D-glucuronidase
activity could be detected in any of the four strains on
either PE or SS surfaces. In general, the measured activ-
ity was 2 to 8 fold higher on the PE surfaces compared to
the SS surfaces. The lowest β-D-glucuronidase activity
was detected in E. coli ATCC 25922, and the highest
ATCC 25922Gall CAB
ATCC 25922Gall CABUTI CAB
100
101
102
103
104
105
106
0369
UTI CABO177:H-
Concentration [MPN/cm2]
Time [d]
100
101
102
103
104
105
106
0369
O177:H-
Concentration [genome copies/cm2]
Time [d]
*
(a) (b)
100
101
102
103
104
105
106
0369
ATCC 25922
UTI CAB
Gall CABATCC 25922Gall CAB
UTI CAB
O17
Concentration [MPN/cm2]
Time
7:H-
100
101
102
103
104
105
106
0369
[
d
]
O177:H-
Concentration [genome copies/cm2]
Time
*
[
d
]
(c) (d)
Figure 5. Concentration of E. coli in biofilms on PE ((a) and (b)) and SS ((c) and (d)) coupons in a flow-through drinking
water reactor system during washout. Concentration of ((a) and (c)) culturable E. coli (MPN·cm–2 ± 95% CI; n = 7) and ((b)
nd (d)) E. coli measured by qPCR (genome·copies·cm–2 ± SE; n = 3). *: not sampled. a
A. S. BUKH ET AL. 259
activity was detected in E. coli O177:H-.
The total amount of E. coli quantified by qPCR did not
change significantly on the PE and SS surfaces during
the washout procedure except for a 25-fold and a 100-
fold increase, respectively, in the number of O177:H- at
day 3, and a 4-fold increase in the number of UTI CAB
at day 9 compared to day 0. At day 3 and day 9, qPCR
for E. coli Gall CAB and E. coli O177:H- (only day 9)
was conducted both with and without a 4-h enrichment
step in a 5:1 solution of mineral modified glutamate agar
and brain heart infusion broth at 37˚C on a shaking table.
This was done in an attempt to examine whether cells
detected by qPCR could still multiply or existed in a
dead or non-culturable state. At day 3, we saw a 1951 ±
1259 (mean ± SE; n = 3) fold increase in the total num-
ber of E. coli Gall CAB in water, a 381 ± 42 fold in-
crease on PE surfaces, and a 504 ± 47 fold increase on
SS surfaces. At day 9, the 4-h enrichment did not result
in any increase in amplifiable DNA for either E. coli Gall
CAB or E. coli O177:H-. Hence, the cells were either
more stressed and required a longer enrichment step or
the cells were non-culturable or had died.
3.3.2. C on ce ntrati on of E. coli in the Water Phase
The concentration of culturable E. coli in the water phase
of the flow-through drinking water reaction system de-
creased exponentially over time as a consequence of
washout, dead, and adhering to surfaces (Figure 6(a)).
At day 9, 0.0021% and 0.0069% of the initial concentra-
tions of the strains Gall CAB and O177:H-, respectively,
could be detected.
β-D-Glucuronidase activity of planktonic E. coli in the
water could be detected in all strains at day 0. After 3
days, 0.6% of the initial activity at day 0 could be de-
tected in Gall CAB. No activity was detected in the other
strains. The qPCR results showed that the amount of total
planktonic E. coli was rather constant over the 9 days
(Figure 6(b)). However, the levels were lower in the
water phase compared to on PE and SS surfaces.
3.3.3. Removal and Washout of E. coli from the
Flow-Through Drinking Water Reaction
System
To compare the removal of E. coli from the water phase
and the surfaces, removal rates, k, was calculated (Table
4). The theoretical washout was the dilution rate, D,
based on the flow rate and reactor volume. Culturable E.
coli was shown to adhere to biofilm on PE surfaces and
this interaction had an effect on the washout procedure as
removal of adhered cells were slower (k ranged from
–0.482 to –0.888) compared to the theoretical removal (k
= –1.646). On SS, the strains UTI CAB and O177:H- had
slower removal rates (k = –0.749 and k = –0.541, re-
spectively) compared to ATCC 25922 and Gall CAB (k
= –1.576 and k = –1.438, respectively). E. coli O177:H-
100
101
102
103
104
105
106
0369
ATCC 25922Gall CAB
UTI CABO177:H-
Concentration [MPN ml-1]
Time [d]
(a)
100
101
102
103
104
105
106
0369
ATCC 25922Gall CAB
UTI CABO177:H-
Concentration [genome copies/mL]
Time [d]
**
(b)
Figure 6. Concentration of E. coli in the water phase in a
flow-through drinking water reactor system during washout.
Concentration of a) culturable E. coli (MPN·ml–1 ± 95% CI;
n = 7) and b) E. coli measured by qPCR (genome·copies
ml–1 ± SE; n = 3). *: not sampled.
Table 4. Removal rates for decrease in E. coli culturable cell
numbers. R2 was >0.7 for all r a tes.
Water PE SS
[(MPN·ml–1) d–1] [(MPN·cm–2) d–1] [(MPN·cm–2) d–1]
Theoretical
washout –1.646
ATCC 25922–1.372* –0.805 –1.171*
UTI CAB –1.212* –0.901 –0.714
Gall CAB –0.892 –0.551 –0.950*
O177:H- –0.962 –0.568 –0.581
*No cultivable E. coli was detected at day 9; hence MPN was determined as
<0.05 MPN ml–1 which is just below the minimum detection limit of E. coli
in Colilert18 [22].
were capable of adhering to and persist in biofilms for up
to nine days on both SS and PE surfaces.
In the water phase, the removal rates of the two clini-
cal strains E. coli Gall CAB and O177:H- (k = –0.888
Copyright © 2012 SciRes. AiM
A. S. BUKH ET AL.
260
and k = –0.979, respectively) were slower than the re-
moval rates of E. coli UTI CAB and ATCC 25922 (k =
–1.393 and k = –1.662, respectively) which were similar
to the theoretical washout rate. In both water and biofilm,
the clinical E. coli strain O177:H- persisted three times
longer than the culture collection strain E. coli ATCC
25922.
4. Discussion
The results from this study showed that the two rapid
methods ColiLight II and qPCR could be applied to de-
tect E. coli in drinking water and to evaluate the survival
and persistence of E. coli in drinking water systems. We
found that clinical E. coli strains were able to persist in a
flow-through lab-scale drinking water reactor system for
at least nine days.
In the lab-scale reactor system, the relative amount of
accumulated biofilm on the PE and SS surfaces deter-
mined by measuring the ATP content was not signifi-
cantly different from each other during the time period.
This corresponds to the findings of Yu et al. who studied
biofilm formation on different water distribution pipe
materials in disinfected tap water [23]. Yu et al. found
similar biomass levels on PE and SS surfaces after 90
days of incubation in drinking water.
A relatively low metabolic activity by means of β-D-
glucuronidase activity was found on both PE and SS sur-
faces and in the water phase after 3 days. This can be
explained by the fact that the concentration of culturable
and active E. coli cells present on the surfaces and in the
water phase at day 3 were ca. or below that of the ana-
lytical limit of detection (102 - 103 cells) of the ColiLight
II method [9].
From the qPCR results we see, that the concentration
of nucleic acids was rather stabile over time both in the
water phase and on surfaces. This indicates that E. coli
was released from surfaces, mainly glass and PP surfaces
since the levels on PE and SS were rather constant, over
time to the water phase, and that the released cells were
dead, had entered a VBNC state, or a combination of
both. Furthermore, E. coli was not released from SS and
PE surfaces to an extensive degree, but had entered a
VBNC state or had died. These persistent levels of E.
coli DNA have also been found in other studies. Cell
integrity and respiratory activity together with loss of
culturability over time was found in a study by Cook and
Bolster, who tested the survival of Camp. jejuni and E.
coli in groundwater microcosms during starvation at 4˚C
[10]. Lothigius et al., who examined the survival and
gene expression of enterotoxigenic E. coli in sea water
and freshwater, also found high degree of cell integrity
for up to 12 weeks in both water types together with ex-
pression of genes involved in metabolic pathways and
genes encoding enterotoxins [12]. In our study, however,
the cellular integrity did not correlate with the metabolic
activity by means of β-D-glucuronidase activity as no
activity could be detected at day 9 where the DNA levels
were high. However, this may be explained by a very
low activity in the cells in combination with the sensitiv-
ity of the ColiLight II method.
Our results indicated the existence of strain-differen-
tiated removal of E. coli from both water and surfaces in
a lab-scale drinking water reactor system due to incorpo-
ration into biofilms. The initial concentrations of E. coli
did not vary significantly between experiments and can-
not explain the differences in the measured concentra-
tions over time. These differences indicated a differenti-
ated washout where the strains Gall CAB and O177:H-
could persist longer in the system compared to ATCC
25922 and UTI CAB. Especially the strain O177:H-
showed high persistence in the system. It is known that
curli production by non-pathogenic and pathogenic E.
coli enhances the attachment of cells to SS surfaces [24]
and glass surfaces [25,26]. As an A/EEC strain, E. coli
O177:H- used in this study are likely to be able to pro-
duce curli which increases the ability of the strain to
form biofilm [26].
Our findings point towards differentiated strain-spe-
cific survival of E. coli in drinking water matrices. The
survival of different E. coli strains has been examined in
other types of freshwaters such as river water [27] and
well water [28], and both studies suggest strain differen-
tial survival of E. coli.
The ColiLight II method measures β-D-glucuronidase
activity in viable cells with a greater sensitivity than
methods based on fluorescent and chromogenic sub-
strates [9]. This method can be applied to detect meta-
bolic active E. coli in drinking water when present in
adequate concentrations. Using qPCR, the total amount
of extracted nucleic acids in a sample is measured with a
high specificity and sensitivity.
Overall, the two rapid detection methods, qPCR and
ColiLight II, are suitable for detection of E. coli in
drinking water and biofilms. Both methods can comple-
ment standard cultivation-based methods for detection of
E. coli in drinking water and can provide results within
one work day.
Clinical E. coli strains persisted longer in drinking
water than a culture collection strain, and strain specific
attributes can significantly affect detection and persis-
tence of E. coli in drinking water matrices.
5. Acknowledgements
We thank MD DMSc Professor Henrik C. Schønheyder,
Aalborg Hospital, MD PhD Brian Kristensen, Skejby
Hospital, and Professor MD PhD Erick Denamur, Institut
Copyright © 2012 SciRes. AiM
A. S. BUKH ET AL. 261
national de la santé et de la recherche médicale, Paris,
France for providing E. coli strains used in this study.
We also thank Margit Paulsen for providing technical
assistance. This work was supported by the Obel Family
Foundation and the Danish Council for Strategic Re-
search, the project SENSOWAQ—Sensors for Monitor-
ing and Control of Water Quality.
REFERENCES
[1] A. Rompré, P. Servais, J. Baudart, M. de-Roubin and P.
Laurent, “Detection and Enumeration of Coliforms in
Drinking Water: Current Methods and Emerging Ap-
proaches,” Journal of Microbiological Methods, Vol. 49,
No. 1, 2002, pp. 31-54.
doi:10.1016/S0167-7012(01)00351-7
[2] L. Fiksdal and I. Tryland, “Application of Rapid Enzyme
Assay Techniques for Monitoring of Microbial Water
Quality,” Current Opinion in Biotechnology, Vol. 19, No.
3, 2008, pp. 289-294. doi:10.1016/j.copbio.2008.03.004
[3] J. D. Berg and L. Fiksdal, “Rapid Detection of Total and
Fecal Coliforms in Water by Enzymatic Hydrolysis of
4-Methylumbelliferone-beta-D-galactoside,” Applied and
Environmental Microbiology, Vol. 54, 1988, pp. 2118-
2122.
[4] L. Fiksdal, M. Pommepuy, M. P. Caprais and I. Midttun,
“Monitoring of Fecal Pollution in Coastal Waters by Use
of Rapid Enzymatic Techniques,” Applied and Environ-
mental Microbiology, Vol. 60, 1994, pp. 1581-1584.
[5] I. George, M. Petit and P. Servais, “Use of Enzymatic
Methods for Rapid Enumeration of Coliforms in Fresh-
waters,” Journal of Applied Microbiology, Vol. 88, No. 3,
2000, pp. 404-413.
doi:10.1046/j.1365-2672.2000.00977.x
[6] T. Garcia-Armisen, P. Lebaron and P. Servais, “β-D-
Glucuronidase Activity Assay to Assess Viable Es-
cherichia coli Abundance in Freshwaters,” Letters in Ap-
plied Microbiology, Vol. 40, No. 4, 2005, pp. 278-282.
doi:10.1111/j.1472-765X.2005.01670.x
[7] P. Lebaron, A. Henry, A. Lepeuple, G. Pena and P. Ser-
vais, “An Operational Method for the Real-Time Moni-
toring of E. coli Numbers in Bathing Waters,” Marine
Pollution Bulletin, Vol. 50, No. 6, 2005, pp. 652-659.
doi:10.1016/j.marpolbul.2005.01.016
[8] S. O. Van Poucke and H. J. Nelis, “Rapid Detection of
Fluorescent and Chemiluminescent Total Coliforms and
Escherichia coli on Membrane Filters,” Journal of Micro-
biological Methods, Vol. 42, No. 3, 2000, pp. 233-244.
doi:10.1016/S0167-7012(00)00193-7
[9] A. S. Bukh and P. Roslev, “Characterization and Valida-
tion of a Chemiluminescent Assay Based on 1,2-Diox-
etanes for Rapid Detection of Viable Escherichia coli,”
Applied Microbiology and Biotechnology, Vol. 86, No. 6,
2010, pp. 1947-1957. doi:10.1007/s00253-010-2514-6
[10] K. L. Cook and C. H. Bolster, “Survival of Campylobac-
ter jejuni and Escherichia coli in Groundwater during
Prolonged Starvation at Low Temperatures,” Journal of
Applied Microbiology, Vol. 103, No. 3, 2007, pp. 573-
583. doi:10.1111/j.1365-2672.2006.03285.x
[11] W. Ahmed, F. Huygens, A. Goonetilleke and T. Gardner,
“Real-Time PCR Detection of Pathogenic Microorgan-
isms in Roof-Harvested Rainwater in Southeast Queen-
sland, Australia,” Applied and Environmental Microbiol-
ogy, Vol. 74, No. 17, 2008, pp. 5490-5496.
doi:10.1128/AEM.00331-08
[12] Å. Lothigius, Å. Sjöling, A. Svennerholm and I. Bölin,
“Survival and Gene Expression of Enterotoxigenic Es-
cherichia coli during Long-Term Incubation in Sea Water
and Freshwater,” Journal of Applied Microbiology, Vol.
108, No. 4, 2010, pp. 1441-1449.
doi:10.1111/j.1365-2672.2009.04548.x
[13] C. Lee, J. Kim, S. G. Shin and S. Hwang, “Absolute and
Relative QPCR Quantification of Plasmid Copy Number
in Escherichia coli,” Journal of Biotechnology, Vol. 123,
No. 3, 2006, pp. 273-280.
doi:10.1016/j.jbiotec.2005.11.014
[14] O. Clermont, M. Lescat, C. L. O’Brien, D. M. Gordon, O.
Tenaillon and E. Denamur, “Evidence for a Human-Spe-
cific Escherichia coli Clone,” Environmental Microbiol-
ogy, Vol. 10, No. 4, 2008, pp. 1000-1006.
doi:10.1111/j.1462-2920.2007.01520.x
[15] DS 6222-1, “Water Quality—Enumeration of Culturable
Micro-Organisms—Colony Count by Inoculation in a
Nutrient Agar Culture Medium,” Danish Standards Asso-
ciation, Copenhagen, 2002.
[16] W. E. Garthright and R. J. Blodgett, “FDA’s Preferred
MPN Methods for Standard, Large or Unusual Tests, with
a Spreadsheet,” Food Microbiology, Vol. 20, No. 4, 2003,
pp. 439-445. doi:10.1016/S0740-0020(02)00144-2
[17] I. Tryland and L. Fiksdal, “Enzyme Characteristics of β-
D-Galactosidase- and β-D-Glucuronidase-Positive Bacte-
ria and Their Interference in Rapid Methods for Detection
of Waterborne Coliforms and Escherichia coli,” Applied
and Environmental Microbiology, Vol. 64, 1998, pp. 1018-
1023.
[18] C. Johansen, A. Verheul, L. Gram, T. Gill and T. Abee,
“Protamine-Induced Permeabilization of Cell Envelopes
of Gram-Positive and Gram-Negative Bacteria,” Applied
and Environmental Microbiology, Vol. 63, 1997, pp. 1155-
1159.
[19] I. Bronstein, B. Edwards and J. C. Voyta, “1,2-Dioxe-
tanes: Novel Chemiluminescent Enzyme Substrates. Ap-
plications to Immunoassays,” Journal of Bioluminescence
and Chemiluminescence, Vol. 4, No. 1, 1989, pp. 99-111.
doi:10.1002/bio.1170040116
[20] S. O. Van Poucke and H. J. Nelis, “Development of a
Sensitive Chemiluminometric Assay for the Detection of
β-Galactosidase in Permeabilized Coliform Bacteria and
Comparison with Fluorometry and Colorimetry,” Applied
and Environmental Microbiology, Vol. 61, 1995, pp. 4505-
4509.
[21] M. Raymaekers, R. Smets, B. Maes and R. Cartuyvels,
“Checklist for Optimization and Validation of Real-Time
PCR Assays,” Journal of Clinical Laboratory Analysis,
Vol. 23, No. 3, 2009, pp. 145-151. doi:10.1002/jcla.20307
[22] A. Al-Turki and M. G. El-Ziney, “Evaluation of Com-
Copyright © 2012 SciRes. AiM
A. S. BUKH ET AL.
Copyright © 2012 SciRes. AiM
262
mercial Colilert18-Quantitray® Method by ISO Tech-
niques for Enumeration and Quantification of Total Coli-
forms and Escherichia coli in Drinking-Water of Burai-
dah, Saudi Arabia,” Journal of Applied Sciences, Vol. 9,
No. 18, 2009, pp. 3357-3363.
doi:10.3923/jas.2009.3357.3363
[23] J. Yu, D. Kim and T. Lee, “Microbial Diversity in
Biofilms on Water Distribution Pipes of Different Mate-
rials,” Water Science and Technology, Vol. 61, No. 1,
2010, pp. 163-171. doi:10.2166/wst.2010.813
[24] J. Ryu and L. R. Beuchat, “Biofilm Formation by Es-
cherichia coli O157:H7 on Stainless Steel: Effect of
Exopolysaccharide and Curli Production on Its Resistance
to Chlorine,” Applied and Environmental Microbiology,
Vol. 71, No. 1, 2005, 247-254.
doi:10.1128/AEM.71.1.247-254.2005
[25] O. Vidal, R. Longin, C. Prigent-Combaret, C. Dorel, M.
Hooreman and P. Lejeune, “Isolation of an Escherichia
coli K-12 Mutant Strain Able to Form Biofilms on Inert
Surfaces: Involvement of a New ompR Allele That In-
creases Curli Expression,” Journal of Bacteriology, Vol.
180, 1998, pp. 2442-2449.
[26] Z. Saldaña, J. Xicohtencatl-Cortes, F. Avelino, A. D.
Phillips, J. B. Kaper, J. L. Puente and J. A. Girón, “Syn-
ergistic Role of Curli and Cellulose in Cell Adherence
and Biofilm Formation of Attaching and Effacing Es-
cherichia coli and Identification of Fis as a Negative
Regulator of Curli,” Environmental Microbiology, Vol.
11, No. 4, 2009, pp. 992-1006.
doi:10.1111/j.1462-2920.2008.01824.x
[27] K. L. Anderson, J. E. Whitlock and V. J. Harwood, “Per-
sistence and Differential Survival of Fecal Indicator Bac-
teria in Subtropical Waters and Sediments,” Applied and
Environmental Microbiology, Vol. 71, No. 6, 2005, pp.
3041-3048. doi:10.1128/AEM.71.6.3041-3048.2005
[28] L. Watterworth, B. Rosa, H. Schraft, E. Topp and K. T.
Leung, “Survival of Various ERIC-Genotypes of Shiga
Toxin-Producing Escherichia coli in Well Water,” Water,
Air, & Soil Pollution, Vol. 177, No. 1-4, 2006, pp. 367-
382. doi:10.1007/s11270-006-9179-x