| Front page | | Contents | | Previous | | Next |

Coliform bacteria and E. coli in drinking water. Comparison of EU reference method with alternative methods

3 Danish equivalency study on TTC-Tergitol (EN ISO 9308-1:2000) compared to five other methods – spiked samples

• 3.1 Materials and Methods
  • 3.1.1 Water samples
  • 3.1.2 Membranfilters
  • 3.1.3 Methods
• 3.2 Results and discussion
• 3.3 Conclusion on spiked samples

This equivalency study included five different methods (four membrane filtrations and one MPN) to be compared with TTC-Tergitol when analysing samples of drinking water spiked with the coliform bacteria Enterobacter aerogenes and E. coli.

Besides the comparison of methods, two different brands of membrane filters were tested as it has been shown (Ossmer, Schmidt & Mende, 1999) that different types of filters may affect the results significantly. Furthermore the samples were analyzed right after spiking and again after 24 hours at 0 – 5°C which should show the possible effect of refrigerated storage as it might be used in normal sampling procedures.

The choice of methods is described in details in “Background” and in part 1.3.

This study aimed to verify for Danish drinking water the problems seen with the EU Reference Method when background flora interferes as well as the suitability of Colilert as demonstrated in the EU trial published by Niemela, Lee & Fricker (2003). At the same time the Danish study included other internationally tested methods to examine their performance for Danish drinking water.

3.1 Materials and Methods

3.1.1 Water samples

Two different levels of microbiological quality of drinking water were used:

1. drinking water from a public water supply (“Public”)
2. drinking water from a private water supply (well) with high heterotrophic counts at 15.000 cfu/ml (“Well”).

Both types of water were inoculated with Enterobacter aerogenes as well as E. coli. The cultures were grown overnight in Brain Heart Infusion broth (BHI; Merck 1.10493) at 37°C respectively 44°C after which they were stored in a water/ice-bath. Immediately after placing in water/ice the concentrations of the cultures were determined for each of the cultures by pour plating in Yeast Extract Glucose agar as double spreadings (YEA; EN/DS 6222:2002) incubated at 37°C overnight. The cultures were kept on ice (0°C) overnight and used for dilution and inoculation of water samples. Dilutions used for inoculation were calculated from the YEA counts.

Four litres of each type of water were inoculated with 15 ml diluted mixed culture to give a final concentration on 10 – 25 cfu/100 ml.

Exact concentrations of the used inoculums were determined as described above for the BHI-tubes, immediately after spiking the water samples.

Inoculation was done while the water was homogenized on a magnetic stirrer. After inoculation the water was left on the stirrer for 30 minutes to ensure a homogenous sample. Hereafter the water samples were distributed as follows:

1. 1300 ml were analyzed immediately in portions of 100 ml (six membrane filtrations on two different filters and one MPN technique)
2. 1300 ml were stored at 0 – 5°C for 24 hours before analysing in portions of 100 ml
   1400 ml were diluted with 1400 ml non-inoculated water of the same type (public or well). From these 2800 ml:
3. 1300 ml were analyzed immediately in portions of 100 ml
4. 1300 ml were stored at 0 – 5°C for 24 hours before analysing in portions of 100 ml

Table 2: Designation of the different samples

3.1.2 Membranfilters

Two different brands of membrane filters (0,45 µm cellulose-mixed-esters) were tested:

MF1 = Gelman 66278 – GN 6. White filter.
MF2 = Millipore HAWG04700. Black filter.

3.1.3 Methods

The tested methods are listed in table 3.

Table 3: Methods tested in the equivalency study.

All combinations of type of water, dilution, type of filter and methods were tested immediately as well as after 24 hours.

For every combination 100 ml of sample was analyzed.

3.2 Results and discussion

In table 4 is shown the results of the determination of concentrations of the inoculums used for spiking the water samples.

Table 4: Colony counts of inoculum used for spiking determined in YEA 37°C. Counts from both spreadings are given.

Based on the results from day -1 it was decided to spike the samples with 15 ml of 10^-7 of each culture. This resulted in (calculated from the Day 0 results):

[(51 + 58)/2] x 15 ml : 4000 ml x 100 ml = approx. 20 E. coli pr. 100 ml

[(9 + 11)/2] x 15 ml : 4000 ml x 100 ml = approx. 4 E. aerogenes pr. 100 ml, i.e. 20 E. coli + 4 E. aerogenes = 24 coliform bacteria pr. 100 ml.

These calculated concentrations of coliform bacteria and E. coli are given as “Expected counts” in table 5.

The results of the comparison of the seven methods are shown in tables 5 and 6. In table 5 results are given as the colony counts or as the most probable number and in table 6 the same results are shown as percentage of the expected counts.

As described in 3.1.2 two filter types were compared. The readings of the plates showed that the black filters (MF2) did not support the differentiation of the different colored colonies as well as the white filters. Furthermore the black filters did not allow a good assessment of the color of the agar underneath the filter. It was therefore decided to use only the counts on MF1 as the counts on MF2 should be treated with some caution. The data from MF2 are not shown in this report.

It should although be noted that the MF2 filters were overgrown in the same samples as MF1.

Table 5: Colony counts resp. most probable numbers from the equivalency study with spiked samples.
”C” = coliform bacteria; ”EC” = E. coli; “OG” = overgrown; “-“ = not tested.

a) Due to a mistake presumptive E. coli were not verified by oxidase test immediately after counting, but after prolonged incubation where the presumptive colonies were found to be oxidase positive, i.e. non-coliforms.

Table 6: Percentage (%) of colony counts resp. most probable numbers compared to the expected counts calculated from the concentration of inoculum. This table converts the results in table 5 to percentage.
”C” = coliform bacteria; ”EC” = E. coli; “-“ = not tested or cannot be calculated due to overgrowth of the plates.

The results in table 5 showed that none of the media TTC-Tergitol, LSA 37, Chromogenic or Chromocult were suitable for water with high heterotrophic counts as the plates were overgrown so reading of typical colonies was not possible.

In the Dutch study (RIVM & KIWA , 2001) it was estimated that much more of the membrane filter was covered with background flora on TTC-Tergitol compared to LSA, which underlines the problem with using TTC-Tergitol for water with high counts of heterotrophic count. In the present Danish study LSA37 was also overgrown when analysing water with high heterotrophic count.

As both LSA37 and LSA44 shall be conducted on each water sample if coliform bacteria as well as E. coli are to be determined, the performance of LSA44 is of minor interest when LSA37 is found not to be suitable because of overgrowth. Furthermore the results from LSA44 itself were not very convincing as E. coli spiked in the water samples were not recovered in most of the samples.

MLGA showed reasonable results with a good coherence between undiluted samples and 1:1-diluted samples. The recoveries were from approximately 20 – 70% in spiked public water and 50 – 140% in spiked private water. The higher recoveries in the spiked private water compared to the spiked public water are expected to be caused partly by a natural content of coliform bacteria in the private water and therefore detection of these coliform bacteria as well.

The method using MLGA includes a technical problem as it is incubated at 30°C for 4 hours prior to 37°C for 14 hours. This means that samples analysed in the morning have to be read in the night. This implies that the samples have to be moved to 37°C late in the afternoon/early evening so that reading can be made the next morning without deviations. Contrary the risk is that the reading will not be done at the prescribed time but as the first thing in the morning even if this is after more than 14 hours at 37°C. Both situations may imply logistic problems causing more expensive analyses due to less flexibility in the planning of sampling as well as in the work in the laboratories. Alternatively it might result in unwanted modifications of the method.

Colilert was the other method able to detect coliform bacteria and E. coli in the spiked water without being overgrown. It is remarkable that the Colilert detects very high numbers of coliform bacteria in the spiked water from private wells (>> 100% recoveries). This may be due to:

• A high content of coliform bacteria in the well water before spiking. This might also explain the overgrowth of the other media but would not change the conclusion that the overgrown media are not suitable for counting coliform bacteria and E. coli in this water. Nevertheless the used well water is known to normally having a high heterotrophic count but no coliform bacteria as determined by the Danish national reference method (DS 2255) which is a MPN technique in MacConkey broth. It is possible that the normal findings are false-negative compared to Colilert as the methods uses two different detection principles that may allow detection with Colilert and ß-D-galactosidase activity, but not with lactose fermentation as principle.
• A false positive reaction in Colilert. This cannot be finally concluded and documented as the growth in the positive wells in Colilert was not identified. Nevertheless literature has shown Colilert to be rather specific for coliform bacteria (Niemela, Lee & Fricker, 2003).

Colilert was assessed to be the most easy and robust technique to handle including all aspects from preparing to reading and without any verification needed.

3.3 Conclusion on spiked samples

The equivalency study with spiked samples confirmed the problems with using TTC-Tergitol for detection of coliform bacteria and E. coli in waters with high heterotrophic counts as also stated in the scope of EN ISO 9308-1:2000 (“Due to the low selectivity, background growth can interfere with the reliable enumeration of coliform bacteria and E. coli, for example in some drinking waters, like shallow well waters, that have not been disinfected and yield a high background growth. This part of ISO 9308 is therefore especially suitable for disinfected water and other drinking waters of low bacterial numbers”).

The study demonstrated that the same problem was seen for three other membrane filtration methods: LSA37, Chromogenic agar and Chromocult which were all overgrown as well.

The only membrane filtration method not overgrown was MLGA which showed relatively good recoveries. The negative experience with this method was due to technical facts with incubation periods which means that the use of the method requires quite good logistics if the analyses should not be handled during the night as discussed above. Therefore this method was excluded from the further studies.

Colilert showed varying results for spiked public water compared to TTC-Tergitol with some results higher and some lower than TTC-Tergitol. A Student’s t-Test on log 10 converted results for the eight paired set of data for TTC-Tergitol/Colilert in spiked public water showed a non-significant difference (31% probability) which means that the two methods can be considered equivalent. The differences were also found to be non-significant if Students t-test was made on the four paired set of data for coliform bacteria and for E. coli separately.

It was decided to make a further equivalency study with TTC-Tergitol and Colilert with natural non-spiked water samples of varying microbiological quality. Two othmer methods were included as explained below. The four methods for the equivalency study with natural drinking water were:

• TTC-Tergitol as EU-reference method (EN ISO 9308-1:2000)
• Colilert as the most easy and robust technique to handle, apparently not affected by high heterotrophic counts (at least not for detection of E. coli) and with promising results from extensive international trials.
• Chromogenic agar as the chromogene agar giving the highest recoveries and not requiring verification.
• The Danish reference method DS 2255 (MPN technique using MacConkey broth) to compare the results with the method used today as national method.

| Front page | | Contents | | Previous | | Next | | Top |

Version 1.0 February 2007, © Danish Environmental Protection Agency