Investigating Escherichia coli habitat transition from sediments to water in tropical urban lakes

Study sites and environmental variables

A total of 35 water and 21 sediment samples were collected regularly at five independent urban lakes (Tasik Varsiti, Tasik Taman Jaya, Tasik Aman, Tasik Kelana and Tasik Central Park Bandar Utama), located 2–7 km between each other in the Klang Valley, Peninsular Malaysia, from May 2022 until November 2022 (Fig. 1). The sampling dates with respective coordinates and experiments conducted for each sampling are listed in Table S1. To avoid effects of precipitation, sampling was carried out when there was no rain. Surface water samples (≈ 0.1 m) were collected using autoclaved bottles (121 °C at 15 psi for 15 min) whereas surface sediment samples (≈ three cm depth) were taken with a shovel and collected using UV sterilized (at 245 nm wavelength for 20 min, intensity 550 µW cm⁻²) plastic zip lock bags. All samples were transferred on ice to the laboratory within 3 h for further analysis.

A conductivity probe (YSI Pro 30, Yellow Springs, OH, USA) and a pH meter (Hach HQ11d, Loveland, CO, USA) were used to measure in-situ water temperature and pH, respectively. For dissolved oxygen (DO), water samples were collected with DO bottles in triplicates, and fixed with manganese chloride and alkaline iodide solution, before titration with sodium thiosulphate solution according to Winkler’s method (Grasshoff, Kremling & Ehrhardt, 1999). Total suspended solids (TSS) was determined by filtering a known volume of water sample through a pre-combusted glass fibre filter (GF/F) (Sartorius, Goettingen, Germany) and measuring the weight increase after drying at 70 °C for a week. Particulate organic matter (POM) was determined by the weight loss after combustion at 500 °C for 2 h (HYSC MF-05, Seoul, Korea). Chlorophyll a (Chl a) was also concentrated on the GF/F filter and extracted with 90% (v/v) ice-cold acetone at −20 °C overnight. Chl a concentration was then measured via a spectrofluorometer (PerkinElmer LS55, Waltham, MA, USA) (Parsons, Maita & Lalli, 1984). The filtrate from the filtration was kept frozen until the determination of ammonium (NH₄) and phosphate (PO₄). These dissolved inorganic nutrients were determined on a spectrophotometer (Hitachi U-1900, Tokyo, Japan) via methods described by Parsons, Maita & Lalli (1984).

The sediment sample collected was dried in a freeze dryer (Labconco FreeZone 6 Liter, Kansas City, MO, USA). For sediment particle sizing, about 10 cm³ of dried sediments were mixed with distilled water until a final volume of 40 mL. Then 10 mL of sodium hexametaphosphate (20% final concentration) was added to disperse the sediment particles (Mil-Homens et al., 2006). The prepared sample was then homogenized and left overnight before analysis with the Beckman Coulter LS230 Particle Size Analyzer (Brea, CA, USA). For sediment organic matter content, the freeze-dried sediment was combusted at 500 °C for 3 h, and the organic matter content was measured via the loss on ignition method (Heiri, Lotter & Lemcke, 2001).

Enumeration of coliform and E. coli in water and sediment samples

For water samples, both coliform and E. coli were measured whereas for sediment samples, only E. coli was measured. The additional coliform measurement in the water samples helped in the classification of the lake waters according to the National Water Quality Standards for Malaysia (Department of Environment, 2008). Membrane filter technique (MFT) was used to enumerate both coliform and E. coli in water where a known volume of water sample (0.01 mL to 10 mL) was filtered through a sterile 47 mm diameter, 0.45 µm pore-size nitrocellulose membrane filter (Millipore, Burlington, MA, USA). For volumes <1.0 mL, the filtration vessel was filled with 5 mL sterile saline (0.85% sodium chloride (NaCl) final concentration) before addition of sample. After filtration, the membrane filter was placed on the CHROMagar™ ECC agar (CHROMagar, Paris, France) and incubated at 37 °C for 24 h. All blue and mauve-coloured colonies were counted as total coliform, whereas only blue colonies were counted as E. coli (Chromagar, 2019).

For sediment samples, 2 g of fresh sediment sample was mixed with 18 mL of sterile saline and then sonicated for 50 s with an ultrasonicator (220 W, 2 mm probe; SASTEC ST-JY98-IIIN, Subang Jaya, Malaysia) (Epstein & Rossel, 1995). After allowing the mixture to settle for 10 min, the suspension was pipetted and used as inoculum in the MFT described above. The membrane filter was then placed on m-TEC agar (Sigma-Aldrich, Burlington, MA, USA) and incubated at 44.5 °C for 24 h. Purple- or magenta-coloured colonies were counted as E. coli (Merck KGaA, 2018).

Measuring E. coli decay or growth rates

Using the size fractionation method, the water sample was divided into three size fractions: total or unfiltered, <20 µm and <0.2 µm fractions (Lee et al., 2011). The <20 µm fraction was collected after filtration through a nylon net with a 20 µm mesh opening size, whereas the <0.2 µm fraction was collected after filtration with a 0.2 µm pore-size membrane filter (Millipore GTTP filter, Burlington, MA, USA).

As E. coli counts in the water was sometimes too low, we used a laboratory strain of E. coli (isolated from Tasik Varsiti) for the decay or growth experiment. A fresh E. coli culture was adjusted to 0.5 McFarland standard (about 1.5 ×10⁸ cfu mL⁻¹) before further serial dilution to 10⁵ cfu mL⁻¹. About 198 mL of each size fraction was then inoculated with 2 mL of 10⁵ cfu mL⁻¹ E. coli culture for a final concentration of about 10³ cfu mL⁻¹. Inoculated size fractions were then incubated at 30 ° C for 24 h in the dark. The abundance of E. coli was determined as cfu mL⁻¹ every 6 h through MFT on m-TEC agar. The cfu data was then transformed via natural logarithm and plotted against incubation time. A positive gradient of the best-fit regression line indicates E. coli growth whereas a negative gradient shows decay rate (Lee et al., 2011).

As protists are the major bacterial predators (Enzinger & Cooper, 1976), we also enumerated protists (Caron, 1983). A 50 mL water sample was preserved with glutaraldehyde (1% final concentration) during each sampling. At the laboratory, 1 to 2 mL preserved sample was filtered onto a black 0.8 µm polycarbonate filter (Millipore ATTP filter, Burlington, MA, USA) with a GF/A filter (Whatman, Little Chalfont, UK) as a backing filter. Filters were then rinsed twice with 0.1 M pH 4.0 Trizma-hydrochloride before being flooded with two mL of primulin solution (250 mg L⁻¹) for 15 mins. After staining, the solution was removed gently by vacuum filtration. The black filter was then placed on one drop of immersion oil on a clean glass slide, and the prepared slide was observed under an epifluorescence microscope (Olympus BX60F-3, Tokyo, Japan) with U-MWU filter cassette (excitor 330–385 nm, dichroic mirror 400 nm, barrier 420 nm).

Habitat transition experiment for E. coli

We added 2 g of fresh sediment sample to the bottom of an autoclaved universal bottle (121 °C at 15 psi for 15 min), carefully avoiding any contact with the inner wall of the bottle. Then 0.4% (w/v) sterile soft agar (Difco, East Rutherford, NJ, USA) (kept at 45 °C) was added to the bottle until it covers approximately one cm level above the sediment. After the agar solidified, 18 mL of sterile saline was added slowly (Fig. 2). To check for contamination, a blank without addition of sediment sample was also carried out. The habitat transition experiment was then incubated at 30 °C for 24 h in dark. The abundance of E. coli in the overlying saline was enumerated every 6 h via MFT on m-TEC agar. The cfu data was then natural logarithm transformed and plotted against incubation time where the gradient of the best-fit regression line was determined as E. coli increase rate (µ_increase).

As E. coli also undergoes intrinsic growth in the saline environment (Hrenović & Ivanković, 2009), we setup a microcosm experiment by replacing raw sediment sample with autoclaved sediment sample from Tasik Kelana (n = 2) and Tasik Central Park Bandar Utama (n = 2). The sediment was autoclaved to replicate possible nutrient contribution from the sediment but prevent adding bacteria to the microcosm. We then inoculated 10³ cfu mL⁻¹ of E. coli to the sterile saline. The microcosm was then incubated at 30 °C for 24 h in the dark and the abundance of E. coli in the saline was enumerated every 6 h via MFT on m-TEC. The best-fit linear slope was determined as E. coli intrinsic growth (µ_intrinsic), and the habitat transition rate was finally estimated by the following equation: µ_increase–µ_intrinsic.

Data analysis

All data were reported in this study as mean ± standard deviation (SD) unless stated otherwise. Values beyond mean ± 2 × SD were determined as outliers, and the coefficient of variation (CV) was used to measure the dispersion of data. Before statistical analysis, bacterial cfus were transformed by log (cfu + 1), whereas for growth or decay rate estimations, bacterial cfus were transformed via natural logarithm. Correlation analysis was carried out to identify relationships among variables, whereas linear regression was used for rate analysis. Student’s t-test was used to compare between groups, whereas one-way ANOVA (analysis of variance) with Tukey’s post-hoc analysis was used to determine the differences among lakes, and p ≤ 0.05 was considered significant. PAST (PAleontological STatistics) software (version 4.09) for Windows (Hammer, Harpper & Ryan, 2001) was used to perform the statistical analyses, whereas plots were made in GraphPad Prism (version 9.5.1.733) for Windows (Swift, 1997).

Environmental variables

Table 1 lists the size and land use of five lakes, and physico-chemical variables measured in the water samples collected at the five lakes in this study. Surface water temperature and pH varied little among the five lakes and ranged from 28.7 ± 0.8 °C to 29.9 ± 1.2 °C (CV = 3%) and 6.8 ± 0.4 to 7.2 ± 0.3 (CV = 5%), respectively. In contrast, DO levels varied among the five lakes, from 2.65 ± 1.78 mg L⁻¹ at Tasik Taman Jaya to 9.65 ± 2.00 mg L⁻¹ at Tasik Central Park Bandar Utama (ANOVA: n = 18, F (4,13) = 3.08, p = 0.05).

In contrast, TSS and POM concentrations were different among the lakes, and ranged from 21 ± 7 mg L⁻¹ to 65 ± 13 mg L⁻¹ (ANOVA: n = 18, F (4,13) = 14.06, p < 0.001), and from 13 ± 2 mg L⁻¹ to 43 ± 5 mg L⁻¹ (ANOVA: n = 18, F (4,13) = 34.8, p < 0.001), respectively. TSS and POM concentrations were highest at Tasik Central Park Bandar Utama (Tukey’s HSD: TSS: q > 6.65, p < 0.01; POM: q > 6.41, p < 0.01). Chl a concentration also varied among lakes (ANOVA: n = 18, F (4,13) = 14.14, p < 0.001), and was highest at Tasik Aman (90.63 ± 15.14 µg L⁻¹) (Tukey’s HSD: q > 5.06, p <0.03). For dissolved inorganic nutrients, NH₄ varied among five lakes (ANOVA: n = 18, F (4,13) = 16.85, p < 0.001), ranged from 0.30 to 98.83 µM and was highest at Tasik Taman Jaya (Tukey’s HSD: q > 4.70, p < 0.04), whereas PO₄was similar among the lakes (ANOVA: n = 18, F (4,13) = 1.80, p = 0.19), and varied from 0.15 to 0.60 µM.

For the physico-chemical properties of sediments (Table 2), average particle size ranged from 55.2 ± 26.2 to 613.4 ± 124.2 µm, and were different among the five lakes (ANOVA: n = 18, F (4,13) = 21.62, p < 0.001) with the largest average particle size at Tasik Kelana (Tukey’s HSD: q > 5.52, p < 0.02). The sediment texture at Tasik Varsiti was mainly loam, whereas in other lakes were mainly sand. Average sediment organic matter measured ranged from 3 to 72 mg g⁻¹ and was not different among the five lakes (ANOVA: n = 18, F (4,13) = 2.96, p = 0.06).

Biotic variables

Total coliform and E. coli were detected in all five urban lakes (Fig. 3, Table S2). Total coliform ranged from 21 to 4,600 cfu mL⁻¹, and E. coli ranged from 1 to 2,300 cfu mL⁻¹. Total coliform in the water was different among the five lakes (ANOVA: n = 20, F (4,15) = 3.58, p = 0.03) but E. coli count was not different (ANOVA: n = 20, F (4,15) = 2.52, p = 0.09). The highest total coliform was detected at Tasik Kelana (Tukey’s HSD: q = 4.97, p = 0.02). For the urban lake sediments, E. coli was present in all five lake sediments (Fig. 4). Its abundance ranged from below detection to 12,000 cfu g⁻¹, and there was no difference among the five lakes (ANOVA: n = 20, F (4,15) = 2.69, p = 0.07).

E. coli decay or growth rates

Generally, the abundance of E. coli in the larger fractions (total and <20 µm fraction) decreased with incubation time, while the <0.2 µm fraction increased (Figs. 5 and 6, Table S3). Decay rates among the five lakes in the total fraction (ANOVA: n = 19, F (4,14) = 7.85, p < 0.01) and in the <20 µm fraction (ANOVA: n = 18, F (4,13) = 4.89, p = 0.01) were different (Fig. 7, Table S4). The highest decay rates in both the total fraction (Tukey’s HSD: q > 4.65, p < 0.04) and <20 µm fraction (Tukey’s HSD: q = 5.77, p = 0.01) were observed at Tasik Taman Jaya. The decay rates measured in the total fraction also did not differ from those in the <20 µm fraction (Student’s t-test: n = 37, t (35) = 0.43, p = 0.67). As decay was most likely attributed to protistan grazers (Lee et al., 2011), we measured protists abundance in the water samples, and observed that protists counts ranged from 3.04 ×10⁴ cells mL⁻¹ to 6.93 ×10⁴ cells mL⁻¹ but showed no differences among the five lakes (ANOVA: n = 15, F (4,10) = 2.18, p = 0.14). In contrast, E. coli grew in the <0.2 µm fraction, and the E. coli growth rates varied among five lakes (ANOVA: n = 18, F (4,13) = 3.65, p = 0.03).

Habitat transition experiment for E. coli

In the habitat transition experiment, E. coli abundance generally increased with time (Fig. 8, Table S5). E. coli increase rates (µ_increase) in the water column (p < 0.05) ranged from 0.40 to 0.59 h⁻¹ at Tasik Varsiti, 0.46 to 0.62 h⁻¹ at Tasik Taman Jaya, 0.41 to 0.74 h⁻¹ at Tasik Aman, 0.61 to 0.71 h⁻¹ at Tasik Kelana and 0.69 to 0.78 h⁻¹ at Tasik Central Park Bandar Utama (Table S6).

As the E. coli increase rate is a sum of both transition and intrinsic growth, we also measured E. coli intrinsic growth rates (Table S7). The intrinsic growth rates using sterile sediments from Tasik Kelana were 0.39 h⁻¹ and 0.32 h⁻¹, and were similar to Tasik Central Park Bandar Utama (0.41 h ⁻¹ and 0.36 h⁻¹). Although sediments at Tasik Kelana had the lowest organic matter content (3 to 8 mg g ⁻¹), whereas Tasik Central Park Bandar Utama had the highest (19 to 72 mg g⁻¹) among the five lakes, their E. coli intrinsic growth rates were not different (ANOVA: n = 4, F (1,2) = 0.48, p = 0.56). Therefore, for the calculation of habitat transition rates, we assumed the average intrinsic growth rate (0.37 ± 0.04 h⁻¹) for all five lakes (Fig. 9, Table S8). The habitat transition rates were different among five lakes (ANOVA: n = 18, F (4,13) = 4.01, p = 0.02), with the highest at Tasik Central Park Bandar Utama (Tukey’s HSD: q = 4.67, p = 0.04).

Environmental condition of the urban lakes

The surface water temperatures recorded at the five lakes were relatively high with low variability, and is typical of tropical waters (Lim et al., 2018). The DO concentrations measured at Tasik Varsiti, Tasik Aman, Tasik Kelana and Tasik Central Park Bandar Utama were at healthy levels, and within the range previously reported for tropical freshwater (Wong et al., 2022). However, for Tasik Taman Jaya, we observed hypoxic levels (2.65 ± 1.78 mg L⁻¹) (Farrell & Richards, 2009), which was not surprising as Wong et al. (2022) had previously classified Tasik Taman Jaya at Class III for total coliform and Class V for faecal coliform, indicative of extensive treatment required for the suitability of water supply (Department of Environment , 2008). All lakes were also observed with high Chl a, indicating varying levels of eutrophication (Lim et al., 2018).

Total coliform and E. coli in water and sediment of urban lakes

The total coliform abundance enumerated in five lake waters were within the range previously reported in Malaysia (Wong et al., 2022). According to National Water Quality Standards for Malaysia, Tasik Varsiti, Tasik Taman Jaya and Tasik Kelana were categorized as Class V for total coliform, whereas Tasik Aman and Tasik Central Park Bandar Utama as Class III. Relative to Wong et al. (2022), the water quality in these lakes had deteriorated over the last two years.

In our study, the abundance of E. coli in sediment and water were correlated (Pearson correlation: n = 40, r (38) = 0.53, p = 0.02) (Fig. 10), similar to Whitman, Nevers & Byappanahalli (2006) and Fluke, González-Pinzón & Thomson (2019). Although we did not use the same agar medium to enumerate E. coli abundance in water and sediment, we had previously shown that E. coli counts on m-TEC and CHROMagar ECC agar were strongly correlated (Regression: n = 18, R² = 0.99, F (1,7) = 561.35, p < 0.001) (Fig. 11, Table S9), and that the abundance of E. coli (log cfu mL⁻¹) obtained on m-TEC was on average 149% higher than that on CHROMagar ECC. In order to compare E. coli abundance in water and sediment, we corrected the abundance obtained, and found that the abundance of E. coli in the sediment was still higher than in the water column (Stephenson & Rychert, 1982; An, Kampbell & Peter Breidenbach, 2002; Garzio-Hadzick et al., 2010; Pandey et al., 2018; Fluke, González-Pinzón & Thomson, 2019).

Relative to the water column, sediment provides E. coli with more nutrients (Jamieson et al., 2005); lower UV intensity (Koirala et al., 2008); lesser bacterivore grazing (Wright et al., 1995) and lower oxygen (Lorke & MacIntyre, 2009). This host intestinal-like environment helps E. coli to survive better in sediments, even at varying climates (Ishii et al., 2006; Ishii & Sadowsky, 2008; Garzio-Hadzick et al., 2010; Rochelle-Newall et al., 2015; Tymensen et al., 2015; Fluke, González-Pinzón & Thomson, 2019).

As E. coli is dependent upon sediment organic matter for growth (Ishii et al., 2006), higher abundance of E. coli in sediments with higher organic matter has been reported (Lee et al., 2006). However in our study, sediment organic matter and sediment E. coli abundance were not correlated (Pearson correlation: n = 36, r (34) = −0.11, p = 0.66). One possible reason could be the organic matter replete state in our lakes. The sediment organic matter from this study was relatively higher (Organic matter content: Tasik Varsiti 3%, Tasik Taman Jaya 3.3%, Tasik Aman 2.7%, Tasik Kelana 0.5% and Tasik Central Park Bandar Utama 3.3%), than that reported by Lee et al. (2006) (i.e., 0.7%–1.1%). In this study, the sediment E. coli abundance also did not correlate with particle size (Pearson correlation: n = 36, r (34) = −0.07, p = 0.78), further substantiating that sediment texture may be less important for the survival of sediment E. coli (Lee et al., 2006).

Although sediments could act as reservoirs of E. coli, and contribute to the water column E. coli, E. coli dynamics in the water column is also dependent upon E. coli decay or growth rates in the water (Lee et al., 2011). The decay rate in total and <20 µm fractions were higher than the smaller fraction and consistent with previous reports (Lee et al., 2011; Wong et al., 2022). In the total fraction, the E. coli decay rates ranged from 0.02 to 0.16 h⁻¹ (or 0.50 to 3.94 d⁻¹). These decay rates were generally within the range reported by Flint (1987) measured at 37  °C, and higher than in subtropical water (Bitton et al., 1983). Temperature may explain the generally higher rates observed in this study, as microbial activity is at its optimum in tropical aquatic habitats (White et al., 1991).

In our study, the decay of E. coli in the larger fraction is mainly due to protistan bacterivory (Enzinger & Cooper, 1976; Lee et al., 2011; Wong et al., 2022). However, we found no correlation between decay rate and protist counts (Pearson correlation: n = 10, r (8) = −0.3, p = 0.62). As E. coli only accounts for small fraction of the total bacterial community, at about 4.48% of total culturable gram-negative rod in freshwater (Goñi Urriza et al., 1999), this could explain the uncoupling between protists and E. coli decay rates. Lee et al. (2011) have also reported that E. coli decay rates have a relatively small impact on the overall bacterivory rate. Although viral lysis could also cause E. coli mortality, previous studies have shown that its role is generally minimal (Lee et al., 2011). Moreover in the <0.2 µm fraction, where protists were removed, E. coli did not decrease but increased against incubation time, suggesting that viral lysis was not significant (Lee et al., 2011).

The role of habitat transition rates for E. coli persistence in the water column

Habitat transition experiments in this study showed that E. coli in sediments could transition from the sediment to the overlying water column without mechanical effects i.e., turbulence and resuspension. Although seasonal change in precipitation and turbulence can cause sediment resuspension and an increase in E. coli abundance in the upper water column (Li, Filippelli & Wang, 2023), the effect should be minimal in lakes as E. coli abundance quickly return to pre-resuspension level (Whitman, Nevers & Byappanahalli, 2006; Abia et al., 2017). Moreover, we have shown net transition rates in laboratory experiments without mechanical effects. Therefore, any precipitation will only increase the impact of habitat transition, and not affect the conclusion from this study.

The habitat transition rates fluctuated among lakes (CV = 53%) and was not correlated with sediment particle size (Pearson correlation: n = 32, r (30) = 0.38, p = 0.14) and organic matter (Pearson correlation: n = 32, r (30) = −0.08, p = 0.78). Although the habitat transition of E. coli from sediment to water may be associated with biofilm sloughing (Lee et al., 2006), the mechanisms that drive dispersal in E. coli biofilms are complicated, and some are still unknown (McDougald et al., 2012).

In this study, we observed the presence of E. coli in all five tropical urban lake waters. Although previous studies have also reported on the survival of E. coli in the water column, they did not include the effects of sediment (Lee et al., 2011; Wong et al., 2022). Given the higher abundance of E. coli in the sediments (Garzio-Hadzick et al., 2010; Fluke, González-Pinzón & Thomson, 2019), and that these E. coli can transition to the water column, the effects of sediment on the abundance of E. coli in the water column could be important.

In order to evaluate the effect of habitat transition on the abundance of E. coli in the water column, we compared E. coli habitat transition rates with total fraction decay rates. We found that in most cases (>80%), the habitat transition rates were higher than the total fraction decay rates (Fig. 12). Thus, there was a net increase rate of E. coli in the water column, calculated by the following equation: µ_{habitattransition}–µ_{totalfractiondecay}. The E. coli net increase rates ranged up to 0.36 h⁻¹ (0.16 ± 0.13 h⁻¹) in our study. When the habitat transition rate exceeds the total fraction decay rate, using E. coli as a faecal indicator could overestimate the faecal contamination level.

However, these rates were from microcosm-based experiments that did not include other biotic (e.g., biofilm and competition) and abiotic (sunlight) factors that can also affect E. coli survivability in-situ (Korajkic et al., 2014; Stocker et al., 2019; Petersen & Hubbart, 2020; Moon et al., 2023). Further studies are therefore needed to understand the role of these factors on E. coli habitat transitions. In this study, we showed the role of sediment as reservoirs and habitat transition as a possible explanation for the persistence of E. coli in tropical aquatic habitats (Wong et al., 2022).