Presence of enterotoxigenic Escherichia coli in biofilms formed in water containers in poor households coincides with epidemic seasons in Dhaka



Åsa Sjöling, Department of Microbiology and Immunology, Institute of Biomedicine, Sahlgrenska Academy, University of Gothenburg, Box 435, 405 30 Göteborg, Sweden. E-mail:



The objective of this study was to investigate if biofilms may be potential reservoirs for the waterborne pathogen enterotoxigenic Escherichia coli (ETEC) in household water in Dhaka, Bangladesh.

Methods and Results

Biofilms formed on submerged glass slides. Mature biofilms were found significantly more often on glass slides collected in the monsoon period between the two annual ETEC peaks in Bangladesh, that is, between May and August than the rest of the year (P < 0·03). Sixty-four per cent (49/77) of all biofilms analysed by quantitative real-time PCR were positive for ETEC. Significantly more ETEC-PCR positive biofilms were found during the epidemic peaks and during flooding periods than the rest of the year (P < 0·008). Planktonic ETEC was present in the household water during all seasons, but there was no correlation between presence or numbers of ETEC in water and the epidemic peaks.


We conclude that ETEC is continuously present in water and biofilms in household water reservoirs in Dhaka, which has a high prevalence of ETEC diarrhoea. The frequency of biofilms with ETEC was significantly associated (P < 0·008) with seasonal epidemic peaks of ETEC diarrhoea.

Significance and impact of the study

We show for the first time that enterotoxigenic Escherichia coli (ETEC), the causative agent of acute watery diarrhoea and travellers' diarrhoea is present in biofilms in household water tanks in Dhaka, Bangladesh.


Enterotoxigenic Escherichia coli (ETEC) is one of the most common causes of acute diarrhoeal disease in developing countries, causing up to 400 million diarrhoeal cases annually and approximately 400 000 deaths among children under the age of five every year (Wennerås and Erling 2004; Qadri et al. 2005a; Walker et al. 2007). ETEC is also a cause of severe to mild disease in adults. ETEC contains two main groups of plasmid-encoded virulence factors: colonization factors (CFs), which are adhesion molecules on the bacterial surface that mediate colonization of the small intestine by binding to the enterocytes (Gaastra and Svennerholm 1996; Nataro and Kaper 1998) and the heat-stable (ST) and heat-labile (LT) enterotoxins. The enterotoxins ST and LT induce diarrhoea by binding to receptors on the human intestinal epithelium where both toxins activate signalling pathways that ultimately result in the secretion of large amounts of water and electrolytes into the intestinal lumen (Field et al. 1978; Sack 1980; Sixma et al. 1993). ETEC strains express either ST only, LT only or both toxins simultaneously. Two genotypes of ST are expressed by strains causing disease in humans; STh and STp (Bölin et al. 2006).

In Bangladesh, diarrhoeal diseases are a major health problem, and ETEC accounts for about 11–20% of all diarrhoeal cases and affects both the local population and travellers to the region (Qadri et al. 2005a; Harris et al. 2008; Chowdhury et al. 2011). Bangladesh is a subtropical country characterized by heavy rainfalls. ETEC is endemic in the country, but the incidence of ETEC infections is low during the cool dry winter season from October to February, then the infection rates typically rise during the warm summer season before the monsoon rains with a peak in April–May and a second peak when the rains abate in September–October every year (Qadri et al. 2005a,b). The reason for these epidemic peaks is still elusive. Bangladesh is sometimes affected by flooding during the monsoon period, and severe floods in recent years occurred in 1998, 2004 and 2007 of which the latter occurred during the period of this study. The incidence and severity of diarrhoeal disease increase during floods (Harris et al. 2008; Qadri et al. 2005b). We have previously shown that ETEC is present in drinking and environmental water in Dhaka, the capital of Bangladesh (Begum et al. 2005, 2007; Lothigius et al. 2008) and viable after long-term water incubation (Lothigius et al. 2010), suggesting that water may be a reservoir for ETEC and a possible route of transmission.

However, it is well known that bacteria in aqueous environments prefer to form biofilms (Wingender and Flemming 2011). Biofilms are surface-associated bacterial communities surrounded by an extracellular matrix that shelters the bacteria from external stresses such as antibiotic treatment and dehydration. Biofilms form readily on solid surfaces in contact with nonsterile water or in liquid/air interfaces (Costerton et al. 1995). In nature, biofilms are a dominating habitat for waterborne species and are often composed by multiple species including bacteria, protozoa and virus and the composition varies with environmental factors (Brümmer et al. 2000). Bacterial biofilm formation is regulated by complex regulatory systems that respond to environmental signals (Karatan and Watnick 2009; Monds and O'Toole 2009). The first step in biofilm formation is a reversible monolayer of bacteria attached to the surface; this step is initially dependent on flagellar motility and on expression of surface adhesion molecules (Sauer et al. 2002; Moorthy and Watnick 2004). Later stages of microcolonies and subsequently more developed biofilms are composed of complex multilayer structures with different bacterial populations surrounded by exopolysaccharides, external DNA and proteins (Karatan and Watnick 2009; Flemming and Wingender 2010). In the present study, we sought to determine whether ETEC form biofilms in household drinking water in Dhaka and if there is a correlation with seasonal ETEC diarrhoea incidence.

Materials and methods

Selection of sample sites

Households in Mirpur, which is a poor slum area in Dhaka, Bangladesh were selected based on results from a previous birth cohort where children in Mirpur that suffered from several incidences of ETEC diarrhoea 2002–2004 were identified (Qadri et al. 2007). Households with the highest incidence of ETEC infections among children in the previous cohort study, that is, most likely to have ETEC within the household, and with household water storage in water tanks, a requirement for this study, were selected for inclusion in the present study. The water tanks were large water reservoirs kept outside or inside the home for daily consumption. As the water supply and sewage system is not well established in Mirpur, each household collects water from municipal water taps open daily, and the water is carried to the homes and stored in the water tanks. The tanks were drained every 3–6 month, and a few tanks were drained every month. A total of 75 households were included in the study, and samples were taken once or up to three times from each household (usually one sample/household) with sampling occasions during all months of the year. The adults of each household were informed about the purpose of the study and gave oral consent to collect water and to perform biofilm screening. Bacterial samples were analysed mainly at the International Centre for Diarrhoeal Disease Research, Bangladesh (icddr,b) who has the premises for working with pathogens.

Household sampling procedures

Biofilm samples were collected by submerging glass slides, disinfected with 99·5% ethanol and mounted into Petri dishes, into the water tanks. Two holes were made in the lid of the Petri dishes so that water could pass freely through the dish. The lid and plate were tied with a rope, and a piece of cork was attached at the other end to keep the dish floating at a depth of 15–20 cm. The glass slides inside the Petri dishes were placed into household water tanks of 6–8 households each month. After 30 days, the Petri dishes with glass slides were collected and a water sample was taken from the water tank.

Staining and DNA extraction

After collection, the samples were brought to the laboratory at the iccdr,b hospital in Dhaka, and the slides were washed with 3 × 10 ml of phosphate-buffered saline to remove the nonadherent cells and other debris. One slide was air-dried and gram stained (Isenberg 1998) to analyse for adherent cells and exopolysaccharide that stains pink in gram staining indicating the formation of biofilms. The extent of biofilm formation was graded on a scale from 1 to 3 for biofilm-positive slides using a light microscope and 40× magnification. Adherent cells were collected from the second slide by applying 250 μl PrepMan®Ultra (Applied Biosystem, Warrington, UK) on top of the biofilm to dissociate the attached bacterial cells. Bacterial cells were extracted by scraping off the biofilms from the slides with the open end of a sterile 1000 μl pipette tip. The cells were collected in a sterile 1·5 ml micro-centrifuge tube (Eppendorf, Hamburg, Germany). Extraction of bacterial DNA was performed using the DNeasy Blood and tissue kit (Qiagen, Hilden, Germany) following the manufacturer's protocol for extraction of Gram negative bacterial DNA. Water samples were processed by filtering 500-ml aliquots through a 0·22-μm-pore filter. Two quarters of each filter were used for DNA extraction and E. coli isolation on MacConkey agar plates, respectively, as described previously (Lothigius et al. 2008) with the remaining two quarters frozen at −80°C.

Culture, ELISA and PCR analyses of the samples

To determine the presence of culturable ETEC, all water samples were cultured on MacConkey agar followed by toxin ELISA or multiplex PCR analyses to distinguish STp from STh of lactose-fermenting colonies as described previously (Sjöling et al. 2007). The procedure for water samples has been described earlier (Lothigius et al. 2008). To determine and quantify the presence of ETEC in DNA extracted from water and biofilms, the enterotoxin genes, STh, STp and LT, were detected by real-time PCR using a PCR product standard curve as described previously (Lothigius et al. 2008). First DNA from ETEC reference strains (ST64111, Takeda 101 and 286 C2) was used in conventional PCR with primers against the genes of STh, STp and LT (Lothigius et al. 2008), and the PCR products were purified by QIAquick® PCR Purification Kit (Qiagen) according to the manufacturer's instructions. The concentration of the purified PCR products was measured in a nanodrop. The concentration (g PCR product μl−1) was measured, and the number of double-stranded PCR products per μl was calculated using the length of the fragment multiplied with the mole weight of dsDNA nucleotides (660 g mol−1). Using Avogadro's number; 6·022 × 1023 molecules per mole, the number of PCR fragments per μl was calculated (Sjöling et al. 2006). Serial tenfold dilutions were then used as a standard curve and compared with the Ct value of each sample to subsequently calculate the number of DNA gene copies per biofilm or water sample.

Statistical analysis

Statistical significance was tested by GraphPad Prism ver. 5 using the two-sided Fisher's exact test. A P-value < 0·05 was considered significant.


Biofilm formation in drinking water tanks

A total of 101 glass slides from 75 households in Mirpur (usually one sample/household) were collected over a period of 2 years (August 2006–August 2008). The samplings were distributed over all months of the year. Based on gram staining, 85% (86/101) of the submerged glass slides had adherent cells, that is, more or less developed biofilms. Such biofilms were found during all months of the year ranging from 60% to 100% positive samples per month (Fig. 1a). There was no significant difference in the number of slides positive for biofilm growth found during the different months or seasons of the year. The biofilm-positive glass slides were visually scored for appearance such as intensity and thickness of the formed biofilm using a 3-grade scale. The mean value of the intensity was plotted against the month of collection (Fig. 1b). The highest mean scores of intensity were recorded in May and October, and biofilms scored as 3 were significantly more often seen during May to August when precipitation gradually increases and reached a maximum than in the other months (Fig. 1c,d) (P < 0·03). To determine whether biofilms graded as 3 correlated with ETEC diarrhoeal peaks, we defined the epidemic peaks to April–May in 2007 and 2008 and July–September in 2007 during the period of flooding in Bangladesh in 2007 (Harris et al. 2008) and September–October in 2006. We found a trend but no significant correlation with biofilms graded as 3 and the epidemic periods (P < 0·063).

Figure 1.

Biofilm frequency and intensity over the year. (a) The percentage of biofilm-positive submerged glass slides recovered each month over a period of 2 years, August 2006–August 2008 (n = 4–15 slides per month). (b) The intensity of formed biofilms was scored as 1–3. The mean and SEM are plotted for all months of the year. (c) Examples of stained glass slides with a monolayer biofilm graded as 1. (d) A multilayered biofilm graded as 3.

Presence of ETEC in biofilms

The 86 samples with biofilm formations were subjected to DNA extraction. Good quality DNA was recovered from 76 samples, and these were further analysed for presence of ETEC DNA using real-time PCR quantification with a PCR product standard curve as described in detail previously (Lothigius et al. 2008). Biofilms positive for at least one of the ETEC toxin genes LT, STh and STp were identified by real-time PCR in a total of 49 (64%) samples with some positive samples recorded during all months. The percentage of ETEC positive glass slides ranged from 20% in November to 100% in May. An association of ETEC positive biofilms to any of the seasons was not found, but significantly more ETEC positive biofilms were collected during the epidemic peak seasons in April–May and July–September 2007 and September–October in 2006 relative to the rest of the year (P < 0·008) (Table 1). Real-time PCR quantification of the bacterial DNA expressed as the number of ETEC gene copies per biofilm present on one glass slide ranged from 40 to 65 000 (median 1015, 75th and 25th percentiles 340 and 1960), and the PCR mainly detected the genes for LT and STh. However, no significant differences were found in gene copy numbers per biofilm when comparing the seasons or when comparing the peak months to the rest of the year (Fig. 2).

Table 1. Presence of ETEC DNA detected by real-time PCR in biofilms is associated with the epidemic seasons in Bangladesh
 Epidemic peak seasonsa, n = 35Endemic seasons, n = 41
  1. a

    Epidemic months were defined as April–May in 2007 and 2008 and July–September in 2007 (the flood epidemics) and September–October in 2006.

  2. b

    The association between ETEC positive biofilms and the epidemic peak seasons was considered to be statistically significant by the two-tailed Fisher's exact test. (P < 0·0082).

ETEC positive biofilms, n = 4828 (80%)b20 (49%)
ETEC negative biofilms, n = 287 (20%)21 (51%)
Figure 2.

Gene copy numbers determined by real-time PCR. The number of gene copies encoding the two major virulence genes encoding the toxins ST and LT recovered from one glass slide biofilm (filled symbols) or from 100 ml of drinking water (open symbols) collected at epidemic and endemic months during the study period. The mean value of the two PCR duplicates is shown for each sample, and the median value of all samples is indicated as a dash. No significant differences were found between epidemic and endemic periods (P > 0·05). (●) LT in biofilm epidemic period; (■) LT in biofilm endemic period; (▲) ST in biofilm epidemic period; (▼) ST in biofilm endemic period; (♢) LT in water epidemic period; (○) LT in water endemic period; (□) ST in water epidemic period; (△) ST in water endemic period.

Identification and bacterial load of ETEC in drinking water does not follow the epidemic peaks

Drinking water samples were collected at the same time as the biofilm samples from households in Mirpur during the course of the study. A total of 70 water samples were tested for growth on MacConkey agar followed by toxin GM1-ELISA, and 62 of these water samples were also filtered and subjected to DNA extraction and subsequently tested by real-time PCR. The frequency of ETEC positive samples detected by ELISA was 20% while 63% were real-time PCR positive for one or several of the toxin genes. Neither real-time PCR nor ELISA results detected any association between ETEC positive samples or the epidemic peaks/season. Thirty-one samples that were positive using real-time PCR were negative using ELISA. Real-time PCR quantification revealed that the 100-ml water samples contained a median of 3100 gene copies (Lothigius et al. 2008). No significant difference was found in gene copy numbers for LT and STh present in drinking water between the different seasons or in the epidemic peak seasons (Fig. 2).


Enterotoxigenic E. coli and other diarrhoeal bacterial pathogens are believed to spread by the faecal–oral route of transmission presumably by contaminated food and water. We and others have repeatedly reported presence of ETEC in environmental and drinking water in endemic areas (Begum et al. 2005, 2007; Lothigius et al. 2008; Patel et al. 2011) but as bacteria in aqueous environments tend to exist mainly in the form of biofilms (Wingender and Flemming 2011), the presence of ETEC in biofilms in drinking water sources needed to be established. This is to our knowledge the first study to confirm the presence of ETEC in biofilms formed in drinking water sources in an endemic area with high rates of acute watery diarrhoea.

Household drinking water biofilms were found during all months of the year. Biofilms scored as grade 3 were found significantly more often on glass slides collected between May and August than the rest of the year (P < 0·03) (Fig. 1b). These months have the highest average temperatures in Dhaka and also this is the peak period for the rainy season. As the households in this study were mainly from low-income groups and consisted mainly of sheds and one-room overcrowded homes, the home and drinking water temperatures were presumably also higher during these months. We did however not measure temperatures during the time of the sampling. Biomass and changed biodiversity of biofilms in estuarine waters have previously been found to be associated with higher temperatures and rainfall in the subtropics (Moss et al. 2006). Other studies have reported a link between higher temperature and earlier biofilm formation as well as higher bacterial biofilm biomass (Diaz Villanueva et al. 2011) Hence, formation of biofilm in drinking water tanks probably occurs faster during the warm and humid months in Bangladesh.

Bangladesh experienced extensive flooding that affected several areas of Dhaka during July to September of 2007, which contributed to an earlier epidemic of ETEC cases than in other years with a peak in August (Harris et al. 2008). Despite this flood, slides were collected and analysed during this epidemic peak. Significantly (P < 0·008) more ETEC positive biofilms were detected by real-time PCR during the epidemic peaks of the entire study period indicating that ETEC in biofilms are linked to the epidemic peaks. This could indicate that more ETEC are circulating between the environment and community during epidemics and hence present in environmental water during the epidemic peaks, and we have also shown this in a previous study (Lothigius et al. 2008). However, the presence of ETEC in drinking water was not found to be higher during the epidemic peaks in this study in contrast to previous evidence of significantly higher numbers of ETEC positive drinking water samples during epidemics (Lothigius et al. 2008). This previous report was however not based on samples taken over the entire year and only covered the autumn peak and the winter season.

Similar to earlier studies, we detected culturable ETEC colonies in the water samples (Begum et al. 2007; Lothigius et al. 2008). However, as we had to detach the biofilms using harsh methods, we could not establish the presence of viable ETEC in the biofilms with confidence which made a comparison between viability in biofilms and water impossible. In addition, we could not establish any significant link between presence of culturable ETEC in water and the epidemic peak seasons or other seasons. The fact that the PCR method detects more positive samples and higher numbers of bacteria further emphasizes that conventional culture methods might underestimate the numbers of potentially infectious bacteria in environmental samples; hence, we choose to compare the results obtained with real-time PCR in this study.

Quantification of the bacterial load in household biofilm and water showed that the numbers varied over four orders of magnitude with up to 10 000 ETEC gene copies per biofilm or 100 ml of drinking water. We hypothesized that higher loads of bacteria would be present during the epidemic peaks, but the numbers of ETEC gene copies did not change significantly either in water or biofilms (Fig. 2). Biofilms may vary in biomass and composition and in the more developed stages of biofilms planktonic bacteria may be released from the biofilms or large clumps of biofilms can detach. Such clumps could contain high densities of pathogenic bacteria and might therefore be sources of infection (Wingender and Flemming 2011). Further studies of drinking water are needed to establish if shedding of large clumps of detached biofilm occurs in household waters.

The oral infectious dose for ETEC has been estimated at 107–109 cells (Qadri et al. 2005a), and according to our results, the biofilms in the households do not contain such high numbers of ETEC, and it does not seem possible that ingestion of a single cup of water could cause disease. However, the infectious dose decreases by several orders of magnitude in high-risk populations such as small children, elderly and immuno-compromised individuals. The population in the study area suffers from malnutrition and poverty and would probably be susceptible to lower levels of bacteria. In addition, Vibrio cholerae biofilms isolated from water sources have been reported to be hyperinfectious and able to out-compete planktonic bacteria as well as establish an infection at an oral dose several orders of magnitude lower in a mouse model (Tamayo et al. 2010). Hence, a few thousand ETEC cells from a biofilms may be able to cause disease in high-risk individuals. As mature biofilms formed more often in the warm rainy periods, there might be a link between onset of epidemics and detachment of biofilms. This however needs to be further investigated. In a previous study, we were able to show that ETEC of similar phenotypic characteristics were isolated both from water and patients with diarrhoea indicating water transmission (Begum et al. 2007). However, the constant presence of low levels of ETEC in biofilms and drinking water found in this study may contribute to the fact that ETEC are often recovered from the faeces of asymptomatic individuals. Although asymptomatic carriers of ETEC may be immune due to repeated infections of the same type of ETEC strains (Qadri et al. 2006), studies of ETEC numbers in faecal specimens of symptomatic and asymptomatic individuals in endemic areas would be interesting to perform.

In conclusion, this is the first study to our knowledge that has established presence of ETEC in household drinking water biofilms. ETEC is endemic in Bangladesh, and the results of this study indicate a constant presence of ETEC in the drinking water and biofilms of households in poor areas. However, the number of biofilms that contained ETEC increased significantly during the epidemic periods, which might suggest that biofilms could be a reservoir for ETEC.


The study was funded by the joint Formas Sida/SAREC funded programme for research on sustainable development in developing countries (209-2009-2018) and the Signe and Olof Wallenius foundation to Å.S. and the Swedish International Development Cooperation Agency (Sida, Grant-INT-icddr,b-HNO1-AV) and the International Centre for Diarrhoeal Disease Research, Bangladesh (icddr,b) to F.Q. The authors declare that they have no competing interests.