Comparison of four β-glucuronidase and β-galactosidase-based commercial culture methods used to detect Escherichia coli and total coliforms in water

Methods based on the enzymatic properties of coliforms (β-galactosidase for total coliforms and β-glucuronidase enzymes for Escherichia coli detection) are used to assess drinking water quality. They were developed to diminish background effects of heterotrophic bacteria and circumvent the need for a confirmation stage required by both multiple-tube fermentation and membrane filter techniques (Clark 1980; Evans et al. 1981; Means & Olson 1981; Seidler et al. 1981; Burlingame et al. 1984; APHA 2005). The β-galactosidase enzyme was chosen because conventional coliform monitoring is based on the detection of the presence of β-galactosidase. The β-glucuronidase enzyme was also chosen because the gene encoding this enzyme (uidA) was found to be specific (Brenner et al. 1972) and present in more than 97% of E. coli isolates (Lupo & Halpern 1970; Martins et al. 1993).

The MI agar (MI; BD, Franklin Lakes, NJ, USA), Colilert^® (IDEXX Laboratories, Westbrook, ME, USA), Chromocult Coliform^® agar (Chromocult coliform^®; Merk KGaA, Darmstadt, Germany) are three commercial test methods based on the determination of β-galactosidase and β-glucuronidase enzyme activities used to detect, within 24 h, total coliforms and E. coli in water samples. These three tests are easy to use, require no additional confirmatory step and provide a more rapid estimate of indicators of the bacteriological contamination of water compared to classical techniques (Brenner et al. 1993, 1996; Edberg et al. 1988; Horman & Hanninen 2006; Olstadt et al. 2007; Pitkanen et al. 2007; Hallas et al. 2008; Mavridou et al. 2010; Boubetra et al. 2011). Different collections of strains were tested with each commercial β-galactosidase and β-glucuronisade-based test method to establish their ability to recover total coliforms and E. coli strains. All of these methods were found to be at least as efficient as classical reference methods in terms of specificity and sensitivity (Landre et al. 1998; Rice et al. 1990, 1991, 1993). However, the expression of the β-glucuronidase enzyme was found to be variable depending on the medium and technique used (Chang et al. 1989; Shadix & Rice 1991; Feng & Lampel 1994; Maheux et al. 2008).

Alternatively, DC with BCIG agar (DC + BCIG; Noegen corporation, Lansing, MI, USA) is formulated to differentiate E. coli from other coliforms. Similar to MI, Chromocult coliform^®, and Colilert^®, the DC + BCIG agar medium contains a chromogenic agent to detect β-glucuronidase enzyme activity. However, it does not contain a chromogenic agent to detect the β-galactosidase enzyme activity. Feng & Hartman (1982) showed that E. coli colonies could be distinguished from other coliforms on membrane filters and plates of violet red bile agar if MUG (4-methylumbelliferyl-beta-D-glucuronide) was incorporated into the culture media. According to this, total coliform colonies are pink on DC + BCIG agar with the exception of E. coli. Because of the low cost of this medium, DC + BCIG agar could be advantageous to assess drinking water quality. Unfortunately, the performance of DC + BCIG agar as compared to reference methods is not well documented.

In the Province of Québec, the Programme d'accréditation des laboratoires d'analyse (‘Accreditation program of analytic laboratories’; PALA), is administered by the Centre d'expertise en analyse environnementale du Québec (CEAEQ), which certifies private, municipal and institutional laboratories. In 2010, the CEAEQ proposed amending their guidelines and since 2013 requires the measurement of the presence of E. coli rather than thermotolerant coliforms, as recommended in the United States and many European countries (AWWA 2005; Government of Quebec 2013). To comply with this new guideline, water testing companies must validate a new procedure able to detect the presence of E. coli rather than thermotolerant coliforms in drinking water. Plenty of methods, including MI, Chromocult coliform^®, Colilert^®, and DC + BCIG agar, are available to detect the presence of E. coli in water with high variability in cost. Currently, there is no study comparing these four methods using both pure cultures of bacteria and water samples.

In this study, we first used a collection of fecal and environmental bacteria isolated from different geographical origins to compare four commercial chromogenic test methods (MI, Colilert^®, Chromocult coliform^®, and DC + BCIG agar). To our knowledge, this is the first report on the comparison of these test methods, using a pure culture panel of this size. Secondly, we compared the four methods in terms of sensitivity using residential well water samples. Their ability to limit the growth of atypical colonies, ease of use and affordability were also compared. The results of this study will help analytical laboratories to choose the best method according to their own needs.

The ability of the four culture-based methods to detect non-E. coli total coliforms and E. coli strains was verified by using 129 total coliform (representing 76 species) and 19 E. coli strains of fecal and environmental origin (Tables 1 and 2). Species identification was reconfirmed using an automated MicroScan Autoscan-4 system (Siemens Healthcare Diagnostic Inc., Newark, DE, USA) or a Vitek 32 system (bioMérieux SA, Marcy l'Étoile, France). Bacterial strains were grown from frozen stocks kept at −80 °C in Brucella medium (Beckton, Dickinson and Company, Mississauga, Ontario, Canada) containing 10% glycerol, and cultured on brain-heart infusion (BHI) agar. Three passages were performed prior to analysis of each strain with each culture-based method.

Non-E. coli total coliform and E. coli cells were grown to the logarithmic phase (0.5–0.6 optical density measured at 600 nm (OD₆₀₀)) in BHI broth and adjusted to a 0.5 McFarland standard (Fisher Scientific Company, Ottawa, Ontario, Canada), before being serially diluted ten-fold in phosphate-buffered saline (PBS; 137 mM NaCl, 6.4 mM Na₂HPO₄, 2.7 mM KCl, 0.88 mM KH₂PO₄, pH 7.4). An aliquot of the 10⁻⁵ dilution was spiked in sterile reverse osmosis-purified water (resistivity of 18 MΩ·cm min at 25 °C) to produce suspensions containing approximately 50 colony-forming units (CFU) per 100 mL of water. Bacterial counts were verified by filtering 100 mL of each spiked water sample through a Millipore membrane filter (47 mm diameter, 0.45 μm pore size; Millipore Corporation, Billerica, MA, USA) with a standard platform manifold (Millipore Corporation) followed by incubation on BHI agar for 24 ± 2 h at 35.0 ± 0.5 °C. Tests to confirm the sterility of filter membranes and buffer used for rinsing the filtration apparatus were also performed.

The membrane filtration method was performed according to Maheux et al. (2009). Three 100 mL volumes were filtered on Millipore filters with a standard platform manifold. The first filter was incubated on MI agar (MI; BD, Franklin Lakes, NJ, USA), the second filter was incubated on Chromocult coliform^® agar (Chromocult coliform^®, Merk KGaA, Darmstadt, Germany), and the third filter was incubated on DC with BCIG agar (DC + BCIG; Neogen Corporation, Lansing, MI, USA) for 24 ± 2 h at 35.0 ± 0.5 °C, before determining colony counts and colour. Each preparation of MI, Chromocult coliform^®, and DC + BCIG plates was tested for performance using positive and negative control strains (Enterobacter aerogenes ATCC 13048, E. coli ATCC 25922, and Pseudomonas aeruginosa ATCC 27853), as recommended by the manufacturer's labeled instructions and the USEPA microbiology methods manual. Tests to confirm the sterility of the filter membranes and buffer used for rinsing the filtration apparatus were also performed (APHA 2005).

For the detection of total coliform and E. coli strains with Colilert^® (Colilert^®; IDEXX Laboratories Canada Corp., Toronto, Ontario, Canada), all preparation, validation, storage and handling steps were performed according to the manufacturer's instructions. Briefly, one snap pack containing the Colilert^® reagent was dissolved in 100 mL of spiked water samples. The solution was then added to a Quanti-tray^®, sealed and incubated at 35.0 ± 0.5 °C for 24 ± 2 h prior to the identification of total coliform positive samples presenting yellow colouration and E. coli samples presenting both yellow colouration and fluorescence under UV light (λ = 365 nm).

During the summer of 2012, 635 1-L raw well water samples from individual households were collected in the Québec City region (Canada). Each well water sample was divided into 100 mL subsamples for simultaneous testing by standard microbiological methods using MI, Chromocult coliform^®, DC + BCIG, and Colilert^® (see the ‘Membrane filtration method’ and ‘Liquid culture method’ sections).

All individual results were recorded using Microsoft Excel 2010 software (Microsoft Corporation; Redmond, WA, USA) and the statistical analysis was performed using the SAS 9.3 program (SAS Institute Inc. 2011. Cary, NC).

To determine the ubiquity (ability to detect all or most total coliform strains) species identification by MicroScan Autoscan-4 system or the Vitek 32 system was used as a reference. Ubiquity was calculated by dividing the number of strains detected by the test by the number of total coliform strains tested.

All water samples were recorded as positive (1) or negative (0) for total coliforms and E. coli. No method was used as a reference to determine the specificity and sensitivity of a particular test for the detection of total coliforms and E. coli. All the methods were compared to each other. Sensitivity (true positive rate) was calculated by dividing the number of positive samples by Method No. 1 plus positive samples by Method No. 2 by the number of positive samples in Method No. 2. Specificity (true negative rate) was calculated by dividing the number of negative samples by Method No. 1 plus negative samples by Method No. 2 by the number of negative samples in Method No. 2.

McNemar's test was used to compare paired proportions with a 95% confidence interval. When the (two-sided) p value was less than 0.05, it was concluded that there is a significant difference between both methods.

An overly conservative measure of agreement, Cohen's kappa coefficient, was also used to measure the inter-rater agreement. Fleiss (1981) magnitude guidelines were used to characterize the κ values (>0.75 = excellent, 0.40–0.75 = fair to good, and <0.40 = poor).

One hundred and twenty-nine total coliform strains (representing 76 species) from fecal and environmental settings were used to demonstrate the ability of MI agar (MI), Colilert^®, Chromocult coliform^® agar (Chromocult coliform^®), and DC with BCIG agar (DC + BCIG) culture-based methods to detect various total coliform strains (ubiquity; Table 1). The results obtained showed that 91 (70.5%), 68 (52.7%), 47 (36.4%) and 30 (23.3%) of the 129 non-E. coli total coliform strains tested yielded a positive signal with the MI, Colilert^®, Chromocult coliform^®, and DC + BCIG methods, respectively. No relationship was observed between isolate origin and false-negative results.

Nineteen E. coli strains from fecal and environmental settings as well as from different geographic origins were used to demonstrate the ability of the four culture methods to detect various E. coli strains (ubiquity; Table 2). For confirmation purposes, all strains that presented negative results were also tested a second time with a different lot of kit/media. The results obtained showed that 18 (94.7%), 18 (94.7%), 17 (89.5%) and 17 (89.5%) of the 19 E. coli strains tested yielded a positive signal with MI, Colilert^®, Chromocult coliform^®, and DC + BCIG methods, respectively.

Six hundred and thirty-five 1-L well water samples collected in the Québec City region during the summer of 2012 were divided into 100 mL subsamples for testing by all four methods to verify how these observations are transposed when real well water samples are tested. The MI, Colilert^®, Chromocult coliform^®, and DC + BCIG culture-based methods yielded a total coliform positive signal for 422 (66.5%), 328 (51.7%), 412 (63.9%) and 349 (55.0%) of 635 well water samples tested, respectively, while only 267 (42.0%), 161 (25.4%), 256 (40.3%) and 186 (29.3%) exceeded the concentration of 10 total coliform CFU/100 mL, respectively (Tables 3, 4 and 6). For each method, an E. coli positive signal was observed for 102 (16.1%), 94 (14.8%), 110 (17.3%) and 85 (13.4%), respectively, of the 635 well water samples tested (Tables 3, 5 and 7).

For the 635 1-L well water samples tested, 85 (13.3%) and 80 (12.6%) allowed the growth of more than 200 atypical colonies on the filter for Chromocult coliform^® and DC + BICG agar, respectively, whereas only six (0.9%) filters out of 635 contained more than 200 atypical colonies for the MI method.

In the present study, the ability of the four culture-based methods tested to detect total coliform strains was statistically different: MI agar (MI) presented the best detection level and DC with BCIG (DC + BCIG) agar the worst, with a difference of 47.2% between the two methods. Detection of total coliforms on the DC + BCIG method is not obvious. Indeed, contrary to the three other methods tested, the medium does not contain a chromogenic agent for β-galactosidase detection. It contains only a chromogenic agent for β-glucuronidase detection. Thus, on this medium, pink colonies are considered total coliforms. Therefore, contrary to the three other methods tested, identification tests of typical colonies should be conducted to confirm the results obtained. Similar to Maheux et al. (2008), the results of the present study lacked correlation between test methods based on the same enzymatic principle to recognize a strain as non-E. coli total coliform. Indeed, our results showed that there is no correlation between the four methods tested either within the same genera or the same species (Table 1).

In 2008, Maheux tested 33 reference and environmental non-E. coli total coliform strains (representing 26 species) to demonstrate the ability of MI, Colilert^®, Chromocult coliform^® agar (Chromocult coliform^®) and Readycult^® culture-based methods to detect various total coliform strains. They showed that the β-galactosidase of 15 (45.5%), 20 (60.6%), 19 (57.6%), and 19 (57.6%) of the total coliform strains tested was detected by the four methods, respectively. For confirmation purposes in this study, all strains that had presented negative results during testing by Maheux et al. (2008) were tested a second time with a different lot of kit/media. However, among the β-galactosidase-negative strains tested by Maheux et al. (2008; ATCC 43890, ATCC 43894, ATCC 43895, LSPQ 2127, LSPQ 3760, LSPQ 3761, and LSPQ 3762) on MI agar, β-galactosidase production was detected during this study. This observation seems to confirm the assumption suggesting that identification methods relying solely on the activity of a single enzyme are subject to a lack of robustness and may lead to misinterpretations since enzymatic activity can be transient and highly regulated by environmental factors (Maheux et al. 2008).

Based on the results obtained, the four culture-based methods tested are not statistically different using pure E. coli cultures. However, it should be noted that for each E. coli strain tested, β-glucuronidase production was detected with at least one of the four methods. Once again, this observation seems to confirm that enzymatic activity can be transient and regulated by environmental factors, including the composition of culture media.

The MI method detected significantly more total coliform-positive well water samples than Colilert^® and DC + BCIG agar (Table 6). For the detection of E. coli-positive water samples, all enzymatic culture-based methods tested were equivalent with the exception of DC + BCIG agar that detected statistically fewer E. coli-positive well water samples than the other three methods.

The MI method is more specific than the Chromocult coliform^® and DC + BCIG agar methods since fewer atypical colonies grew on MI compared to the other two methods. As a liquid culture method, the growth of atypical colonies could not be investigated for the Colilert^® culture-based method.

In terms of time to result, all four methods tested comparably since they required 24 hours for results. However, contrary to MI, Colilert, and Chromocult coliform^® methods, suspect total coliform colonies on DC + BCIG agar should be confirmed with additional tests.

In terms of ease of use, the Colilert^® method was the easiest to use. The unit-dosed packaging eliminates media preparation. Furthermore, there is no repeat testing due to clogged filters. Finally, contrary to other membrane filtration-based methods, its use does not require well-trained employees. The MI, Chromocult coliform^®, and DC + BCIG agar methods provided comparable ease of use in terms of membrane filtration methods. Media must also be prepared and quality control carried out for each batch. Employee training is also more important than for the Colilert^® method. However, employees already using membrane filtration equipment can easily use these methods (Table 8).

In terms of affordability, the Chromocult coliform^® and DC + BCIG agar are comparable. MI is approximately 30% more expensive than the two previous methods. Colilert^® reactants are more expensive (six to 10 times more expensive than chromogenic membrane filtration-based reactants per water sample; Table 8). However, the cost associated with employees is higher for chromogenic membrane filtration-based methods than for Colilert^® since the latter is much easier to use.

In water management, multiple parameters will influence the choice of an analytical method to assess drinking water. Despite the fact that MI, Colilert^®, and Chromocult coliform^® have been shown equivalent in terms of specificity and sensitivity, the ease of use and the cost will also influence the choice of a method. In this study, we addressed all these parameters to help authorities and analytical laboratories make a choice among all available methods for the purpose of their own needs.

We conducted a multiparametric comparison study of the MI agar, Colilert^®, Chromocult coliform agar and DC with BCIG agar methods in terms of ubiquity and sensitivity using both pure cultures of bacteria and residential well water samples. We also compared their ability to limit the growth of atypical colonies, ease of use and affordability. To our knowledge, this is the first report on the comparison of these test methods using a pure culture panel of this size. We showed that, since environmental laboratories already possess the equipment for membrane filtration methods, the use of the MI agar method seems to be the best option for the assessment of drinking water quality by total coliform and E. coli detection even if it costs 30% more than other chromogenic membrane filtration methods. Indeed, MI agar is more cost-effective than Colilert^® and more specific than the Chromocult coliform agar and DC with BCIG agar methods which showed more growth of atypical colonies. However, when no trained employee and/or no membrane filtration equipment are available, the Colilert^® method should be preferred. The results obtained in the present study are applicable solely to drinking water samples. Results could differ with other types of water.

We wish to thank Dr Steve Charette (IBIS; Institut de Biologie Intégrative et des Systèmes, Université Laval, Québec City (Quebec), Canada) for providing laboratory space and Dr Michel G. Bergeron (CRI; Centre de recherche en infectiologie, CHU de Québec, Québec City (Quebec), Canada) for providing the strains used in this study. This research project was funded in part by an Engage Grant (EG) 413660 from the Natural Sciences and Engineering Research Council of Canada (NSERC) and by the partners of the Research Chair on Drinking Water of Université Laval. The first author was supported by a postdoctoral fellowship from NSERC.