Jason Eccles, Thames Water, Development Microbiology, Spencer House, Manor Farm Road, Reading, Berkshire, UK (e-mail: email@example.com).
Aims: This study examined the suitability of three analytical methods for isolating and enumerating Escherichia coli from conventionally treated sewage sludge.
Methods and Results: Crude sewage, mesophilic anaerobic digested (MAD) sludge, and final product sludge samples were taken from six sewage treatment works for analysis. Two of the three methods tested were membrane filtration techniques, utilizing chromogenic E. coli/coliform (CEC) media and membrane-lactose glucuronide agar (MLGA); the third method was a most probable number (MPN) technique utilizing Colilert in Quantitray 2000 (Idexx). The methods were evaluated for variation, consistency, false-positive and false-negative results, as well as method correlation. The methods gave good and consistent recovery of E. coli for a range of conventionally treated sewage matrices. All of the methods had a false-positive rate of <3%, although MLGA had a high false-negative rate (35·5%) compared with Colilert (3·81%) and the CEC method (6·75%). This resulted in slightly lower presumptive counts but comparable numbers of confirmed counts.
Conclusions: The three detection methods tested, chromogenic, MLGA and Colilert gave comparable recoveries, and did not vary by greater than one order of magnitude (1 log).
Significance and Impact of the Study: Forthcoming revisions to the Use of Sludge in Agriculture Regulations (1989) will categorize sewage sludge as untreated, conventionally treated or enhanced treated in accordance to microbiological standards. The standard will be based upon numbers of E. coli removed through the sludge treatment process and the numbers remaining in the final product. It is recommended that the Colilert 2000 (Idexx, Westbrook, Maine) and CEC methods would be equally suitable to assess the reduction of indigenous E. coli in conventionally treated sludges, and that MLGA be used with follow-up confirmatory testing.
Revisions to the UK Code of Practice for Agricultural Use of Sewage Sludge (1989) that are soon to come into force will state that the sewage sludge processing industry must undertake ‘necessary measures to ensure that waste is recovered or disposed of without endangering human health.’ To this end proposals in the draft regulations for sludge disposal (Anon. 2001) state that the requirements for sewage sludge treatment processes will tighten in order to reduce the risk of pathogens entering the food chain. These draft regulations are associated with the European working document on sludge (Anon. 2000a), although differences between the documents are apparent. The proposals for England and Wales categorize sludge in accordance to the Safe Sludge Matrix (Anon. 1999a) as untreated, treated or enhanced treated. Sludge that is compliant to a treated standard, also referred to as conventionally treated, must pass through a sludge treatment process that has demonstrated a 2-log reduction in Escherichia coli numbers. Furthermore a final product standard of <105E. coli per gram of biosolid (dry weight) will be enforced. Sludge compliant to an enhanced treatment standard must pass through a treatment process that has demonstrated a 6-log reduction in E. coli. The final product standard will be set at 103E. coli per gram of biosolid (dry weight). Additionally, enhanced treated sludge must not contain any Salmonella species in 2 g of dry solids. The European document (Anon. 2000a) proposes that equivalent to enhanced treatment processes should be validated by measuring the reduction of a test organism such as Salmonella senftenberg. The proposed final product standard is set at no Salmonella species to be present in 50 g wet weight of sludge.
The pathogens of concern, that may be present in untreated sludge, include E. coli O157, Salmonella, Cryptosporidia, Listeria, Campylobacter and Enteroviruses. With the exception of Salmonella, which must be analysed in enhanced treatment processes, these organisms are not monitored directly because of the difficulties and expense associated with detection. As indigenous E. coli has been shown to have similar survival characteristics to pathogens (Rice et al. 1992; Fenlon et al. 2000; UKWIR 2002) it has been selected as a surrogate to determine pathogen reduction through the treatment process. One advantage of using indigenous E. coli is that it is present in raw sludge in high numbers and can therefore be used to demonstrate multiple log reductions through treatment processes. This paper relates to the issues associated with the detection of indigenous E. coli. Methods for the detection of E. coli O157 from biosolids have been reviewed (UKWIR 2000).
A number of different methods are now available for the selective isolation of E. coli (Anon. 1983, 1994, 1996; Sartory and Howard 1992; Brenner et al. 1993; Sueiro et al. 2001). It is assumed that these methods can also be used for sludge analysis. However, all these methods were developed for the detection of E. coli in drinking or raw water samples (Anon. 2000b) and although current research is beginning to assess the performance of these methods for wastewater analysis (Yakub et al. 2002), their comparative performance for use with sewage sludge is unknown. The current drinking water regulations allow different water companies to use different methods of analysis, providing their chosen method has been validated against the referenced method. As the new sludge regulations (Anon. 2001) set specific levels of E. coli for compliance it is necessary for standardization of sludge methods. It is therefore important that a specific method, with a known recovery, is included in the new regulations in order to set the standard and allow direct comparison for new and alternative methods. No such referenced method is currently included in the draft regulations (Anon. 2000a, 2001).
It was the purpose of this study to compare methods that may be used to enumerate E. coli in order to pass the conventional sludge treatment standard. These methods may also be used for measuring E. coli from enhanced treated samples, but they may be restricted by their limit of detection and therefore enhanced treated samples were not included in this study. The three methods compared were membrane-lactose glucuronide agar (MLGA), chromogenic E. coli/coliform (CEC) media and Colilert. MLGA was the chosen E. coli method in the UK Water Industry Research (UKWIR) funded survey of treatment processes throughout the UK (UKWIR 1999). The Centre of Applied Microbiology and Research (CAMR) developed the CEC method during the UKWIR funded project ‘Pathogens in Biosolids and their significance in beneficial use programmes’ (UKWIR 2002). The Colilert method (Edberg et al. 1988; Edberg et al. 1989) has recently been validated by the Thames Water laboratory for use on sewage sludge, and this study has formed part of the validation process.
All three methods compared are based upon the expression of the enzyme β-gluruconidase. This enzyme has been reported to be present in over 94% of E. coli (Hansen and Yourassowsky 1984).
Materials and methods
Waste water treatment works
Sludge matrices differ considerably depending upon the sewage input into the treatment works and the type and level of treatment undertaken. Samples were collected from different sewage treatment works with conventional sludge processing facilities across the Thames Valley region. The treatment works varied in age, treatment process and sewage feed (urban and rural locations) to represent all conventional sludge types that may be encountered during routine analysis.
Final product samples were taken from six different sewage treatment works, including three sites that dewatered the sludge to produce a ‘cake’ final product (>15% dry weight) and three sites that did not dewater the sludge and had a liquid final product (<10% dry weight). Crude sewage and mesophillic anaerobic digested (MAD) sludge samples were also taken from three of these treatment works.
Sludge samples were taken in accordance to British Standard (1998; BS EN ISO 5667-13). The specification between treatment works differed considerably and sampling points were not always readily accessible, for example some treatment works had sampling taps whereas others required the use of sampling arms. Sampling taps or pipework were flushed prior to sample collection. The sampling arms were washed and disinfected prior to use. As sludge stratifies in the unmixed liquid storage tanks samples were taken, where possible, from the settled sludge at the base of the tank. Cake samples were taken from a depth of at least 20 cm from the surface of the pile using a spade or trowel. All the samples were split for microbiological and dry weight analysis.
Each identified sample point at each treatment works was sampled on at least five separate occasions. The samples were then analysed with at least five replicates for each method of analysis (Fig. 1).
Three methods of analysis were compared in the study, namely the membrane filtration method utilizing MLGA (Sartory and Howard 1992) modified for sludge analysis (UKWIR 1999); the membrane filtration method utilizing CEC agar, based on the technique employed in the UKWIR (2000) study; and the Colilert method as developed for sludge analysis at Thames Water's Laboratory (Reading).
Method using MLGA. A wet weight sample of 10 ± 0·1 g was measured into a 150-ml sterile container and 90 ml of sterile maximum recovery diluent (MRD; Oxoid CM733) was added to dilute the sample and the container capped and shaken thoroughly. The contents were then poured into a small stomacher bag and stomached, using a Colworth stomacher 400, for 1 min ± 5s. After stomaching the contents were poured back into the 150-ml container. A further 10- fold dilution was made by adding 1 ml of the stomached preparation to 9 ml of MRD. A 10-fold dilution series was then made to allow individual colonies of E. coli to be counted on agar plates (the numbers of bacteria present in the sludge were estimated using data from historical samples). The appropriate dilutions were then filtered through a 0·45-μm pore membrane (Pall Gelman Laboratories, Ann Arbour, MI, USA) and placed onto MLGA. The filter was incubated for 4 h ± 30 min at 30 ± 0·5°C and transferred to 44 ± 0·5°C for a further 14 h ± 30 min. A single experienced analyst studied the membranes under artificial lighting conditions within the laboratory. All green colonies were counted and considered as presumptive E. coli.
Method using CEC. A wet weight sample of 10 ± 0·1 g was measured into a 150-ml sterile container and 90 ml of sterile tryptone soy broth (TSB, Oxoid CM129) was added to dilute the sample and the container capped and shaken thoroughly. The contents were then poured into a small stomacher bag and stomached for 1 min ± 5 s. After stomaching the contents were poured back into the 150 ml container. A further 10 fold dilution was then made by adding 1ml of the stomached preparation to 9 ml of TSB. A 10-fold dilution series was then made to allow individual colonies of E. coli to be counted on agar plates. The appropriate dilution was then filtered through a 0·45-μm pore membrane (Gelman GN-6 Metricel) and placed onto the CEC media (Oxoid CM956). The filter was incubated for 4 h ± 30 m at 30 ± 0·5°C and then transferred to 44 ± 0·5°C for a further 14 h ± 30 min. All blue colonies were counted and considered as presumptive E. coli.
Method using Idexx ColilertTM Quantitray 2000. A wet weight sample of 10 ±0·1 g was measured into a 150-ml sterile container and 90 ml of sterile MRD was added to dilute the sample and the container capped and shaken thoroughly. The contents were then poured into a small stomacher bag and stomached for 1 min ± 5 s. After stomaching the contents were poured back into the 150 ml container. A further 10-fold dilution was made by adding 1 ml of the stomached preparation to 9 ml of MRD. A 10-fold dilution series was then made to the point where E. coli present could be counted in the Quantitray 2000. One millilitre of the final dilution was then added to 99 ml of sterile deionized water (SDIW) in a 100 ml Colilert vessel (Idexx). MRD was not used at this stage as Colilert is better reconstituted in SDIW. One sachet of Colilert 18 reagent (Idexx) was poured into the sample and shaken thoroughly until dissolved. This liquid was then added to a Quantitray 2000, sealed and incubated at 37 ± 0·5°C. The Quantitrays were stacked in the incubator at no >10 cm high with a space of no <1 cm between each stack. The Quantitrays were read after 19 ± 1 h incubation.
Confirmations. At least 15 positive colonies or wells or 30% of the total, whichever was the greater, from one replicate treatment at an appropriate dilution, were subcultured onto yeast extract agar (YEA) and MacConkey agar (MacA) to obtain single colonies of a pure culture. These colonies were confirmed as E. coli by the production of acid and gas at 44 ± 0·5°C in lactose peptone water (LPW); the production of indole at 44 ± 0·5°C in tryptone water (TW); the detection of galactosidase and glucuronidase activity at 37 ± 0·5°C in the Colilert detection system; and the detection of cytochrome oxidase activity. Organisms that did not demonstrate the typical E. coli characteristics were further identified using API 20E (bioMérieux, Hazelwood, MO, USA).
The MLGA medium was prepared according to Sartory and Howard (1992). CEC agar plates were prepared in accordance with manufacturers instructions (Brisdon 1998). The MLGA and chromogenic agar plates were prepared at the beginning of the week, stored in the dark at 4 ± 2°C and used within five days. Prior to use, plates were allowed to warm to room temperature.
Dry weight analysis
All samples for dry weight analysis were sent to Thames Water's UKAS accredited Millharbour laboratory and analysis carried out according to standard ‘Blue Book’ methodology (Anon. 1984).
Using the SPSS software package (Anon. 1999b) the E. coli detection methods were compared using analysis of variance.
A secondary investigation was set up to determine the reason behind the observed differences between methods (see results). In this trial final product samples were taken from each treatment works and analysed in triplicate using the three methods. Colour differentiation between green (presumptive E. coli) and yellow (presumptive coliforms) colonies on MLGA had proved to be difficult and subjective in the earlier study. Therefore, in this secondary investigation two experienced analysts read the membranes. Any differences between the counts were reviewed and a final decision made by the senior analyst. Colonies showing even the slightest hint of green were classified as presumptive E. coli. To help colour differentiation the membranes were read under natural lighting conditions. Artificial fluorescent lighting within the laboratory made colony colour differentiation more problematic.
Both positive and negative colony morphologies and Colilert wells were sub-cultured and identification confirmed using the techniques previously described.
Escherichia coli was detected in all samples analysed with each method. A total of 84 samples were analysed, 29 crude sewage, 17 MAD sludge and 38 sludge final product. For all three methods, a total of 1794 tests were carried out, including all the replicate treatments. This included 615 tests on crude sewage, 435 tests on MAD sludge and 744 tests on final product sludge.
The dry weight of the samples ranged from 0·10 to 0·17% for crude sewage, 2·16 to 4·27% for MAD sludge, 1·23 to 6·11% for liquid final product and 15·69 to 29·17% for cake final product.
The three detection methods tested gave comparable recoveries and did not vary by greater than one order of magnitude (1 log), Fig. 2. The trend lines shown on the graphs indicate the relationship between the two methods. Deviation of the trend line from the line of equal recovery represents bias towards a particular method. The R2 value, or co-efficient of determination, indicates how well the trend line fits the data. R2 values range from zero to one, with values closer to one demonstrating good correlation between the two methods. In this study R2 values were greater than 0·98 for all E. coli detection methods compared.
The comparison between MLGA and Colilert (Fig. 2a) demonstrates a slight bias towards the Colilert method. Higher E. coli numbers within the sample appear to accentuate this bias. This is demonstrated in Fig. 2a by the distance of the trend line from the line of equal recovery. When the samples were categorized as crude sewage, MAD sludge and final product sludge the statistical analysis showed a marginal, but not statistically significant, difference between methods for final product samples (P = 0·127). Table 1 shows the statistical results of the E. coli method comparison. The difference between methods for crude and MAD samples was highly significant (P < 0·001 and P<0·001) Fig. 3. This difference (Fig. 3), shows individual sample results (±s.d.) from one of the treatment works, separated into the three sample types. Treatment works C was chosen as it gave a general representation of all the other treatment works. The graphs have been given a 2-log axis range in order to magnify detail. From these graphs the Colilert method clearly detects higher numbers of E. coli for both crude and MAD samples. It is more difficult to distinguish recovery levels between methods for the final product samples.
Table 1. Results from statistical analysis of data
The value shown is the probability of the difference between the means occurring by chance (probability values above 0·05 signify no significant difference at the 95% confidence limit).
The two membrane filtration methods showed a very strong relationship (R2 = 0·99) when the three sample types were compared (Fig. 2b). However, this relationship had a slight bias towards the chromogenic method, demonstrated by the trend line in Fig. 2b falling above the line of equal recovery. Analysis of these results after breakdown into their sample types shows a significant difference between methods for crude and MAD samples (P < 0·001 and P < 0·001) but no significant difference between final product samples (P = 0·618).
There was also a strong relationship between the chromogenic and Colilert methods (R2 = 0·987) when all data were compared (Fig. 2c). In this case there was a slight bias towards the Colilert method for samples with high numbers of E. coli. At the lower end of the E. coli range recovery was very similar, with the chromogenic medium giving slightly higher results. Statistical analysis of the data showed a significant difference between methods for crude sewage and MAD sludge at the 99% confidence level (P < 0·001 and P = 0·004 respectively). Recoveries from the final product samples were not significantly different (P = 0·923).
A total of 1492 presumptive E. coli isolates were analysed of which 31 were found to be true false positives (Table 2). The CEC agar method had the highest false-positive rate (2·95%) followed by the MLGA (1·75%) and Colilert (1·59%) methods. Including those isolates that did not produce gas when grown in lactose peptone water, 16·49% of the total colonies tested did not pass the approved confirmatory standard tests for E. coli. A total of 106 (7·1%) of the colonies tested were anaerogenic, but were subsequently identified as E. coli. Exclusion of these anaerogenic isolates reduced the false-positive level to 4·69%. Further confirmation using the API 20E test reduced the false-positive number to a true figure of 2·08%. Identification of the false-positive isolates included the species E. vulneris, Enterobacter cloacae, Ent. intermedius, Rahnella aquatilis and Citrobacter freundii.
Table 2. Details of the false-positive rate of each method tested
Number of isolates analysed
False-positives (including anaerogenous)
False positives (excluding anaerogenous)
Number true false positives (after API)
% False positives (including anaerogenous)
% False positives (excluding anaerogenous)
% True false positives (after API)
All samples tested positive for E. coli. A total of nine samples were analysed in triplicate from five waste water treatment works. All of the samples tested were either cake or liquid final product sludges. The dry solids were in the range of 1·54–6·15% for liquid final product and 20·77–35·08% for cake final product.
In this trial the plates from the membrane filtration methods were read under natural lighting conditions. The unconfirmed results were much closer in this secondary trial than the initial investigation, with the CEC method giving the highest overall recovery (Table 3). When the results were corrected for false-positive and false-negative reactions the results became even closer, with the MLGA method giving just a 0·1-log lower recovery than Colilert. The CEC method continued to give the highest recoveries after the false-positive and negative adjustments. The MLGA method had the largest increase in E. coli numbers between the presumptive and confirmed results, due to a high false-negative rate. From 262 presumptive negative colonies taken from MLGA plates a total of 35·5% confirmed as E. coli (Table 4). The false-negative rate for the CEC method was 6·73% and the Colilert false-negative rate was 3·81%.
Table 3. Secondary investigations
MLGA presumptive (green colonies)
MLGA confirmed (green + yellow colonies)
Chromogenic presumptive (blue colonies)
Chromogenic confirmed (blue + pink colonies)
Colilert presumptive (flourescent wells)
Colilert confirmed (flourescent + yellow wells)
NA, not analysed.
Mean Escherichia coli results from three replicates (log CFU g−1 dry weight). Presumptive followed by confirmed results are shown for each method, with the false-negative percentage rate for that sample given underneath (grey area).
% False negatives
% False negatives
% False negatives
% False negatives
% False negatives
% False negatives
% False negatives
% False negatives
% False negatives
Table 4. Numbers of false-negative isolates taken from each method tested
No. presumptive negatives analysed
No. presumptive negatives positive
% False negatives
In this study, the three methods demonstrated comparable E. coli recovery rates. In the main study Colilert, on average, recovered the highest numbers of E. coli followed by the CEC and MLGA methods. All of the methods had a false-positive rate of <3%. This was surprising as the methods were originally developed for drinking water analysis and it is unlikely that the challenges associated with sludge analysis were considered during the development of these methods. These additional challenges include the presence of heterotrophic bacteria within the sludge in much higher numbers and greater diversity than drinking water. It was also reasonable to assume that chemicals present in the sludge could have caused false-positive reactions. However, the only false-positive signals observed were from bacteria known to show this trait (Perez et al. 1986; Adams et al. 1990; Haines et al. 1993; Shadix et al. 1993; Ciebin et al. 1995; Venkateswaran et al. 1996). As these methods required considerable dilutions to achieve countable numbers of bacteria it is assumed that any chemical affects were diluted out. This study examined sludges from both urban and rural treatment works that would have varied chemical inputs. No false-positives were observed as a result of chemicals within the sample.
This study was used to compare the final results from already fully developed methods. It was not the intention to compare or recommend individual components of methods to produce a better overall analysis. It is therefore impossible to determine from this data which dilution medium (MRD or TSB) gave the highest recovery. However, it should be noted that the TSB media produced considerably larger volumes of froth during the stomaching procedure and this resulted in more residual particles left in the stomacher bag after emptying. Additional work to determine recoveries between MRD and TSB would therefore be appropriate for any further method development that may be required for standardization of methods.
The initial trial showed that the MLGA method recovered slightly lower numbers of E. coli than the other methods. Possible explanations for these observations were examined in closer detail in the secondary trial. It became apparent that predominantly yellow colonies with even the faintest hint of green confirmed as E. coli. Reading these membranes was therefore subjective and should be undertaken by experienced analysts. It was found that examination of the membranes under natural lighting conditions aided reading. However, this only accounted for some of the discrepancies between the methods as MLGA continued to give lower recoveries in the secondary trial. In this trial, two analysts read the membranes under natural lighting conditions and all colonies showing even the slightest hint of green were classified as presumptive E. coli. All of those tested confirmed as E. coli. A high proportion (35·5%) of presumptive negative (yellow) colonies were found to be E. coli, which when added to the presumptive positive counts gave a confirmed result similar to those of the other methods.
It is therefore clear from this data that MLGA gives slightly lower presumptive results than the other methods tested in this trial. This is unsurprising as Sartory and Howard (1992) state in their findings that expression of green colouration by E. coli can be suppressed when the membrane has high numbers of nontarget organisms present. For statistical representation it is recommended that the sample dilution used should contain between 20 to 70 target colonies. Within this range the background numbers of heterotrophs and coliforms on the MLGA membranes were high despite growth at 44°C. It is highly likely that analysts will experience different levels of false negatives; this may be dependent upon the type of sludge analysed. It is, therefore, strongly recommended that users of this method undertake preliminary trials to determine the level of false negatives for each sludge type they will analyse. If high levels of false negatives are experienced, as they were in this trial, it is recommended that an alternative method be selected. Alternatively confirmations could be carried out on a percentage of the total colonies isolated and a statistically probable final result then calculated.
It is interesting to note that from the confirmations carried out 7·1% were found to be anaerogenic E. coli, which do not produce gas when grown in lactose peptone water. We conclude that gas production is not a reliable test for the identification of E. coli isolated from sewage sludge and therefore do not recommend it's use so as not to exclude anaerogenic strains.
The draft regulations (Anon. 2001) currently state a maximum level of E. coli at 105 CFU g−1 dry solids for compliance of conventionally treated sludge. The methods evaluated in this study are those likely to be adopted for analysis of sewage sludge and all three gave comparable results. It was however apparent that Colilert recovered slightly more E. coli at the high range, from the raw sewage and MAD sludges. At the lower range, mainly final product samples, the recoveries for each method were not significantly different.
Data from this study suggest that either the Colilert or the CEC method would be acceptable standard methods without routine confirmations of isolated colonies or positive wells (Colilert). In accordance to recommendations by Thompson (2001) a comprehensive interlaboratory trial must be carried out to ensure results from E. coli methods are consistent and fit for purpose.
It should be noted that during the course of these trials the draft regulations, or at least the interpretation of them, were changed for categorizing conventionally treated sludge. Initial interpretations were for a 3-log reduction throughout the entire treatment process, hence this study examined crude sewage through to a sludge final product. Revisions to the working document changed this stipulation to a 2-log reduction across the ‘sludge’ treatment process. Future method standardization trials should therefore include raw sludge instead of raw sewage.
This study examined sludge that had been passed through conventional treatment processes. The methods compared here may not be appropriate for enhanced treated sludges that should contain lower numbers of E. coli. Personal observations have shown that stomaching compost or soil samples often ruptures the stomaching bag and therefore alternative methods of bacteria dispersal should be examined in future studies. Sample types such as thermally dried sludge with very low water content should be soaked for a period to hydrate the sample before stomaching. This would allow better break up of the sample and better dispersal of the bacteria. Samples that have been passed through high alkaline or acid conditions should also be pH adjusted to prevent interference with the analysis (UKWIR 1999).
In conclusion, this study was undertaken because there is a paucity of published data comparing methods of recovery for indigenous E. coli from sewage derived sludges. The trials showed that E. coli recovery methods developed for drinking water analysis gave good and consistent recovery of E. coli for a range of conventionally treated sewage matrices. These methods can therefore be used reliably to assess the reduction of indigenous E. coli in treated sludges as required by current working documents for the introduction of new sludge regulations (Anon. 2000a; Anon. 2001)
We gratefully acknowledge the contributions of our Thames Water colleagues who analysed the sludges for dry weight. We are thankful to Dr Hillary Tillet for her statistical input in setting up of the experiments and Richard Cocks for his efforts in analysing the final data. The views expressed in the paper are those of the authors and not necessarily those of Thames Water.