INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The hygienic quality of water is of utmost importance to society, and efficient bacteriological control of water is essential for implementing a good management of this vital resource. To meet the need for a reliable and easy method for testing water microbiological quality, the coliform index was created (11), and in Europe the "guideline value" threshold for fecal coliforms has been set at 100 CFU/100 ml for bathing waters (19).

The most commonly used quantification reference methods for monitoring the microbial quality of water employ membrane filtration or most probable number approaches, which both as a rule require 24 to 48 h for completion (2). In order to improve public safety and to reduce operational costs, faster microbiological detection is desirable for shortening the time required to implement appropriate measures in the case of an unacceptable level of contamination. To be useful, such methods ought to produce results within a working day; be quantitative, sensitive, and specific; require less work than the current standard methods; have a high throughput; and be nondestructive to the target organisms so as to allow confirmation work (17).

For the identification of fecal coliform bacteria (FC), acid and gas production have been the basis of numerous methods. However, the finding that -galactosidase-positive fecal coliforms, in particular Escherichia coli, in some cases do not produce gas due to the lack or loss of the enzyme formate-hydrogen lyase (13) has led to the proposition of new definitions. The revised edition of Report 71 (14) no longer mentions the requirement for gas formation and requires instead the presence of -galactosidase for coliforms and -glucuronidase specifically for E. coli.

In compliance with these new definitions, various methods based on substrate specificity have been developed to demonstrate -galactosidase activity in an appropriate selective medium. These methods include the use of substrates, which will give rise to chromogenic, fluorogenic, and luminescent products. However, such methods still require 18 to 24 h for completion (1, 9, 24). Fluorometers and luminometers have been used to improve the sensitivity of the detection, (3, 10, 15, 22), but automation of such methods has seldom been reported in the literature, and the total time to obtain a result still exceeds a working day when targeting the detection of one CFU (16). In an effort to increase the number of samples processed and to reduce the time required to analyze the samples, a semiautomated instrument, CA-100, as well as a specific microbiological culture medium, has been developed by Colifast Systems ASA to assess the presence or absence of FC by monitoring -galactosidase activity. To our knowledge, this is the first time such a semiautomated method has been reported. The data presented in this study were obtained by the Environment Agency Laboratory in Wales using the CA-100 in parallel with their reference methods for assessing the microbiological quality of Welsh environmental waters, freshwater as well as seawater.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Sampling. A total of 206 locations in Wales were used routinely for these experiments in the period between January and October 1999. Of these, 144 were from freshwater, whereas the remaining 62 sites were from seawater of various origins (beaches, harbors, and estuaries). The samples were collected in 500-ml sterile containers (Medfor), refrigerated at 4°C, and analyzed within 24 h. All samples were processed on the CA-100 without dilution. In a total of 60 experiments, 608 seawater samples and 403 freshwater samples were analyzed. Unless otherwise stated, all results are expressed as a percentage of the 1,011 samples.

Reference methods. All reference methods were performed in compliance with the current United Kingdom standard (14). In brief, membrane filtration was performed using 0.45-µm-pore-size, 47-mm-diameter membrane filters (Gelman) subsequently placed on pads soaked with lauryl sulfate broth (MLSB) for FC presumptive determination. Confirmation was achieved by monitoring acidification and gas production during growth in lactose-peptone-water. The presence of E. coli was confirmed by demonstration of indole production from tryptophan in tryptone-water. Occasionally, API 20E, API Rapid 20E, and API 20NE strips (BioMérieux, Marcy l'Étoile, France) were used on regrown cells streaked from selected CA-100 vials on nutrient agar.

CA-100 instrument. The CA-100 (Colifast Systems ASA) is a semiautomated, thermostatically controlled instrument for liquid handling and the measurement of fluorescence. It was developed to monitor the activity of various enzymes for the specific detection of microbial target groups. The choice of enzyme, in addition to the appropriate specific selective medium and incubation temperature, ensures the microbiological specificity of the detection. The selective medium used for the detection of coliform bacteria contains 4-methylumbelliferone--D-galactoside, a nonfluorescent conjugate, which gives rise to galactose and the fluorogenic 4-methylumbelliferone (MU) after hydrolysis by -galactosidase (4). For the detection of MU, the fluorometer of the CA-100 is equipped with specific excitation and emission filters at ca. 365 and 445 nm, respectively (18), and the fluorescence readings are expressed as relative fluorescence units (RFU).

Experimental procedure. A general flow chart is presented in Fig. 1. Each sample was split and analyzed in duplicate by the CA-100 direct addition method, as well as by the MLSB method. Depending on the expected degree of contamination, either 0.1 and 1 ml or 1 and 10 ml were used for membrane filtration. For control, duplicates were occasionally included. Data were averaged and scored per 6 ml. The morning following an analysis, the instrument was restarted to perform a single cycle (rerun). This enables the identification of possible "false-negative late growers," defined as target organisms not detected during the CA-100 analysis but having enough activity to give a saturated fluorescence measurement the next morning. For direct confirmation of FC and E. coli, aliquots were removed from each vial and transferred to lactose-peptone-water and tryptone-water. This makes possible a direct comparison of the fluorometric method with other methods using the same sample. To assess carryover and possible contamination from one vial to another in the CA-100, control vials, containing sterile distilled water instead of sample, were used. The controls are placed after the last sample in the run, and their net fluorescence value is expected to remain below the pass level (negative).

View larger version (33K): [in this window] [in a new window]   FIG. 1.   Simplified flow chart of the experimental procedure.

Evaluation methodology. For evaluation, the two criteriapresence or absence on the one hand and the environmental threshold of 100 CFU/100 ml on the other handwere used in three different ways. When presence or absence was employed for the CA-100, the reference methods used for evaluation were direct confirmation (E1) or membrane filtration (E2). When 100 CFU/100 ml was employed for the CA-100, only the membrane filtration method was used (E3).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Reagent and instrument stability. The quality of the rinsing procedure was continuously monitored using controls in all experiments. The first 21 experiments showed 32.5% false-positive controls, whereas in the following 39 experiments all controls were negative after five cycles. Because they benefit from a longer rinsing period, high calibration and blank samples in positions 1 and 2, respectively, are less exposed to possible carryover. Their fluorescence value can therefore be used to assess the stability of both the medium and the instrument during a single experiment (from one cycle to the next), as well as between experiments. Fluorescence of the blank was found to increase by an average of 8.1 RFU during an experiment. Furthermore, the variation in the fluorescence reading of the blank was tenfold larger between than within experiments, whereas (high calibration  blank) fluorescence readings had a sixfold larger variation between than within experiments (Table 2). Similarly, the pass level showed a fivefold larger variation between experiments than within experiments (Table 2).

Pass level. The pass level was determined empirically by minimizing the false-positive and false-negative results using the direct confirmation evaluation (E1). This is illustrated in Fig. 2, where the net fluorescence of 444 samples has been converted to parts-per-billion values and is plotted against results obtained by MLSB and by direct confirmation. Only the 242 samples displaying <10 ppb in addition to those showing <6 CFU are shown. A pass level value of 3 ppb was found to be optimal and is used in Fig. 2 for the interpretation of the results according to E1.

View larger version (31K): [in this window] [in a new window]   FIG. 2.   Pass level optimized at 3 ppb based on the E1 evaluation method (presence-absence validated by direct confirmation). FC, absence of FC; FC+, presence of FC; ×, true negatives; +, true positives; , false positives; , false negatives.

Method evaluation. In this study, the analyzed samples had MLSB counts which ranged from 0 to 8 × 10⁴ CFU/100 ml. Of the samples, 80% had <10³ CFU/100 ml, and of these 55% contained <100 CFU/100 ml. This low contamination level is a challenging experimental situation with regard to the evaluation of the CA-100 method. Results from the three different evaluations, E1, E2, and E3, are presented in Table 3, and the complementary API identifications performed on selected CA-100 samples are shown in Table 4.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Reagent and instrument stability. One important result from this study was that the percentage of false-positive results was successfully limited. This, in part, is the result of implementing improved rinsing procedures. Independently, principles and limits of the method were evaluated by analyzing variations in data generated when measuring a fixed experimental parameter. The increase in fluorescence of the blank indicates that autohydrolysis of the medium is part of the reported blank fluctuation within an experiment. Together with the stronger variation observed between experiments, this illustrates the usefulness of formula 1 to compensate for the background associated with the medium.

Pass level and background -galactosidase activity. Previously, multiplying the blank's fluctuation by 5 has been used to establish the pass level for other fluorometric methods (22). In the present study, this approach gives a value of 62.5 RFU (n = 21), which is equivalent to 5.1 ppb when using values within an experiment. The values thus obtained indicate variations between cycles within one experiment. Therefore, within a cycle, the fluctuation is expected to be smaller. Despite the previously mentioned possible lower stability of controls compared to blanks, controls can be used to roughly assess blank stability within a cycle. The difference |blankcontrol| of all five cycles in 21 experiments was 7.3 RFU, which would give a new pass level of 36.5 RFU equivalent to 2.95 ppb. When the calculated RFU value for the 3-ppb pass level is used instead of the blank, a value close to 1 ppb (n = 60) is obtained. Thus, at least 2 ppb of the pass level is not needed to compensate for the instruments variations. In two similar studies by Davies and Apte (5, 6), the pass level was determined empirically by using experimental data. In their 1-h, 5-ml sample, -galactoside assay set to detect 300 CFU FC/100 ml of seawater sample, a pass level value of 60.3 nM was obtained which was later reduced to 22 nM. This is comparable with the 17 nM (3 ppb) pass level value determined in the present study.

7

7

1

Method evaluation. The species Vibrio spp. found in a false-positive vial was also detected four times in vials scored as true negatives. This finding indicates that this organism did not interfere with the CA-100 assay and supports previous findings by Davies et al. (8). The organism Proteus mirabilis was also identified in a true-negative vial. In one particular experiment, nine seawater samples all had ca. 0.5 CFU of FC/6 ml on the MLSB plates but, in addition, all plates contained a multitude of atypical pink colonies which were identified as Aeromonas hydrophila by API 20NE. Of 18 vials (duplicates of all nine samples), 5 were confirmed as true positives, 9 were confirmed as true negatives, 1 was confirmed as a false-negative late grower, and 3 were confirmed as false positives. Since the majority were true negatives, one can conclude that A. hydrophila is unlikely to interfere with the CA-100 and generate a false-positive result. However, because a few false positives were indeed found, this organism, when above a certain concentration, might be a source of interference. It should also be noted that of the total of 29 false-positive samples studied here, 20 were from seawater, indicating a tendency for such samples to have more false positives (3.3%) than do freshwater samples (2.2%).

Implications for rapid microbiological detection methods. The E1 results differ significantly from those reported by Van Poucke and Nelis (23), who evaluated a similar rapid enzymatic method with regard to presence-absence detection of total coliforms using a direct confirmation method similar to the one used for E1. As mentioned earlier, in the present study the background activity level was found to be an important interference factor with regard to the enzymatic assay. However, we report a false-positive rate of 2.9% in contrast to that of 56.7% reported by Van Poucke and Nelis (23). The difference can be explained in at least two ways. First, it can be argued that the detection of total coliforms, at a lower and less-selective temperature (35°C) than that used for the detection of FC, will be more prone to high levels of nontarget enzyme activity. Nontarget -galactosidase activity has been reported to decrease when the temperature increases from 35 to 44.5°C, whereas the activity due to E. coli -galactosidase increased (20). Second, the CA-100 is a flexible "dynamic" method, using multiple measurements able to take into account, to a large extent, the static contribution of nontarget enzymes, i.e., inhibited nontarget bacteria, VNC, or free enzymes. Three species, V. alginolyticus, Pseudomonas spp., and probably A. hydrophila, have been found to give false-positive results, whereas Vibrio spp. and Vibrio metschnikovii, on the other hand, were found in true-negative vials, thus showing their noninterference in the CA-100 testing. Despite the fact that "minimum interference concentration" was not determined in either case, false-positive interference from nontarget organisms is not expected to be significant since the study had very diversified sampling sources collected throughout the year. The 7.8% false negatives obtained have been shown to result from too short an incubation period. However, because of the challenging nature of the samples used in this study (average low FC contamination and very diversified sample origin associated with a fluctuating enzyme background), it is likely that the method can also be improved when using less-extreme experimental conditions. More homogeneity in the origin and/or nature of the collected samples could give a more stable -galactosidase background activity and therefore a better-customized pass level, enabling a faster detection. Nevertheless, a sudden increase in -galactosidase activity of a monitored water source could also by itself be perceived as relevant information regardless of the immediate FC recoverability. In our opinion, the results of this study show the potential of this enzymatic method as a rapid warning system of FC contamination for recreational waters.