Comparison of Nine Brands of Membrane Filter and the

APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Aug. 1980, p. 186-191 0099-2240/80/08-0186/06$02.00/0 Vol. 40, No. 2 Comparison of Nine Brands of Membrane Filter and the Most-Probable-Number Methods for Total Coliform Enumeration in Sewage-Contaminated Drinking Water R. S. TOBIN,' * P. LOMAX,2 AND D. J. KUSHNER' Health Protection Branch, Health and Welfare Canada, Ottawa, Ontario KlA OL2,1 and Biology Department, University of Ottawa, Ottawa, Ontario KIN 6N5,2 Canada Nine different brands of membrane filter were compared in the membrane filtration (MF) method, and those with the highest yields were compared against the most-probable-number (MPN) multiple-tube method for total coliform enumeration in simulated sewage-contaminated tap water. The water was chlorinated for 30 min to subject the organisms to stresses similar to those encountered during treatment and distribution of drinking water. Significant differences were observed among membranes in four of the six experiments, with two- to fourtimes-higher recoveries between the membranes at each extreme of recovery. When results from the membranes with the highest total coliform recovery rate were compared with the MPN results, the MF results were found significantly higher in one experiment and equivalent to the MPN results in the other five experiments. A comparison was made of the species enumerated by these methods; in general the two methods enumerated a similar spectrum of organisms, with some indication that the MF method was subject to greater interference by Aeromonas. Although total coliform bacteria are of limited value as indicators of pollution in recreational and raw drinking water (9, 16), they are still the primary indicators of drinking water quality in the United States (36) and Canada (15). Because the coliform bacteria as a group are excreted in large numbers in feces and are relatively resistant to environmental stress, they are a sensitive indicator of drinking water quality (12,36). Their relative abundance in many source waters and susceptibility to water treatment techniques makes them also well suited to monitor treatment adequacy; normally the presence of single coliform colonies per 100-ml sample of finished drinking water warrants resampling at that site (15). Higher levels and more frequent coliform contamination will initiate more intensive examination of the problem and may require that the public be notified to boil water before consuming it. At present both the most-probablenumber (MPN) multiple-tube fermentation method and the membrane filtration (MF) method for total coliform enumeration are permitted (1, 15, 36). Several studies have demonstrated the comparability of the MF method with the MPN method in unchlorinated and natural waters (6, 7, 18, 20, 22), whereas others have shown that chlorine-stressed coliforms are not as efficiently enumerated with the MF method as with the MPN method (6, 14, 18, 20, 22, 26). Geldreich et al. (13) recovered only 25% of the total coliforms by the MF method as compared with the MPN method when studying water in dead-end pipes in a distribution system. Bissonnette et al. (3) recovered three times as many total coliforms by means of MPN as by means of MF counts. Recently the question of comparability of the two methods for chlorinated drinking waters has again been raised, especially in relation to the variation in bacterial counts from different brands of membrane filters (12). Differences among membrane filters in the fecal coliform test have been well documented (4, 8, 11, 21, 28, 30, 31, 35) and have led to the development of new, asymmetric membranes; fewer reports have shown differences among brands of membranes in the total coliform test (4, 21, 28). The present study was designed to determine whether there are significant differences among membrane filters in their ability to enumerate chlorine-stressed coliforms in sewage-contaminated drinking water and whether the MF results are comparable to those obtained by the MPN method. Identification of some of the isolates was performed to determine any selectivity or bias in the two methods. 186 MATERIALS AND METHODS Test water. Aerated Ottawa tap water (no detectable chlorine residual) was used as the challenge water. To produce a sewage-contaminated water for test pur-

VOL. 40, 1980 poses, 2 to 10 ml of the supernatant of settled raw sewage was added to 6 liters of tap water to produce a total coliform density of 15,000 to 20,000 colonies/ 100 ml before chlorination. A sodium hypochlorite solution (BDH Chemicals, 5% [wt/vol] available chlorine) was added to the flask to give a final concentration of 0.8 mg of available chlorine per liter. The flask was left at room temperature for 30 min with mixing by inversion every 5 min; a 200-ml portion was removed for testing for free available chlorine and total available chlorine residuals by N,N-diethyl-p-phenylenediamine titration (1). Three milliliters of a 10% solution of sodium thiosulfate was then added to neutralize the chlorine and prevent further killing of bacteria. A control experiment verified that the counts were stable for at least 1 h. A freshly prepared neutralized mixture was used in each experiment. Comparative testing. The MF procedure used was the one-step m-endo LES (BBL Microbiology Systems) agar technique (1). Five replicate 100-ml samples of test water were filtered through each brand of membrane and incubated at 37 C for 24 h. Presumptive total coliform counts were expressed as arithmetic means. In comparisons with the MPN method, the fivetube, three-dilution method with 10-ml portions was used (1). Lauryl tryptose broth (Difco Laboratories) was used for the presumptive test, and all positive tubes were confirmed in brilliant green bile broth (Difco). Membranes tested. The following 47-mm, 0.45-um retention (except, which has a 0.7-rum retention) membrane filters were tested: Nuflow (N47/45), lot 3149, ethylene oxide sterilized by manufacturer; (11406), lot 678, autoclave sterilized in the laboratory; (HCWG 047 S3), lot C8C44583C, ethylene oxide sterilized by manufacturer; (HA-special for water microbiology) (HAWG 047 00), lot CTM094138A, autoclave sterilized in the laboratory; (HAWG 047 00), lot C7M18065B, autoclave sterilized in the laboratory; (M 045 G A47 X), lot 822L459, autoclave sterilized in laboratory; (GN-6), lot 82143, ethylene oxide sterilized by manufacturer; and (M 045 G S47 C), lot 834F573. All membranes were purchased from regular suppliers, except the (our designation), which was provided by Millipore Corp. This membrane is a slightly modified version for water microbiology that has now replaced the standard HA membrane (also tested here). Identification ofmicroorganisms. Colonies from the MF method were transferred directly to Mac- Conkey agar plates. Lauryl tryptose tubes from the MPN method showing positive fermentation were first subcultured to m-endo LES plates and then transferred to MacConkey agar plates. Isolates were identified by the Analytical Profile Index (API 20E) system (Analtab Products). RESULTS Chlorination of the simulated contaminated tap water samples was performed with the addition of 0.8 mg (by weight) of available chlorine COMPARISON OF MPN AND MF FOR TOTAL COLIFORIMS 187 per liter to the sample and using a contact time of 30 min at room temperature. Bacterial devitalization varied somewhat from one experiment to another, but was not correlated with the free available chlorine residual or total available chlorine residual as measured at the end of the 30-min period (Table 1). The bacterial reduction by chlorine for experiments 1 through 6 (in Table 1) in log units was 1.3, 1.7, 0.88, 2.1, 0.82, and 2.1, respectively. This level of bacterial inactivation was chosen to induce sufficient stress in the surviving organisms while still retaining enough bacteria to give sufficient colonies. Although the total coliform counts were obtained to standardize the contaminated tap water before chlorination, the variation in the degree of bactericidal activity in some experiments produced counts that were not within the ideal counting range for MF determinations. Membrane filter comparisons. The results of six experiments on nine different membrane filters, with five replicates each, are presented in Table 1. The membranes are listed in order of decreasing arithmetic mean of the five replicates, and the range of results is given. In these experiments, the best membrane yielded 1.7 to 4.4 times as many total coliform colonies as the worst membrane in the same experiment. The analyses of variance revealed that means were not significantly different in experiments 1 and 4. In all others, the means were significantly different (P < 0.05), and the membranes in each experiment were grouped such that those means enclosed by the same letter (Table 1) did not differ significantly (at P = 0.05) by Tukey's significant difference test (32). The same membranes were not always ranked highest in all of the experiments; however, average ranking of the membranes for all six experiments gave the following order with regard to total coliform rescuscitation: > > > > > > - > >. MPN versus MF comparisons. From the above studies, the membrane with the highest bacterial resuscitation rate was selected for comparison against the MPN method. Six experiments were performed with chlorinated sewagecontaminated tap water as in the previous experiments. In five of the six experiments (Table 2), the MF method gave values slightly higher than the MPN values. In only one of these experiments, however, was the difference significant (P < 0.01); in all the others, the analysis of variance proved that there was no significant difference between the means. Correction of MPN values for the bias inherent in the method (16, 18, 25, 34) did not alter these statistical

188 TOBIN, LOMAX, AND KUSHNER TABLE 1. Total coliform enumeration by the MFprocedure, using nine different membrane filters Meanb/100 ml and f Meanh/100 ml no. and membrane Rangea significant differ- no. and membrane Rangea and significant ences' differences" 1 (FAC: ND; TAC: 0.18 mg/liter)d 2 (FAC: ND; TAC: 0.32 mg/liter) 3 (FAC: ND; TAC: 0.33 mg/liter) Johns Manville GA 58-287 43-181 55-110 70-105 49-79 52-75 31-73 23-104 22-49 117 88 85 85 67 67 54 49 33 A 18-58 361 20-44 34 A 13-60 29 3-61 27 10-34 21 B 6-34 21 C 9-29 19 8-21 16 D 9-20 15 231-361 229-274 184-268 214-252 154-258 169-218 169-247 67-120 103 46-74 61 266 255 242 A 233 B 214 C 200 191 4 (FAC: 0.05 mg/liter, TAC: 0.43 mg/liter) 5 (FAC: ND, TAC: 0.6 mg/liter) 6 (FAC: ND, TAC: 0.41 mg/liter) 13-25 10-23 4-18 5-21 6-19 8-13 7-20 6-17 2-10 16 16 14 14 12 11 10 96 A 222-387 284 187-286 253 204-270 232 192-267 230 A 160-303 228 B 181-255 221 177-277 218 165-219 191 121-205 168 13-22 16 4-22 15 12-19 15 10-17 14 A 8-19 14 B 8-13 10 7-12 10 5-11 8 4-10 6 a Maximum and minimum (colonies/100 ml) of five replicates. 'Arithmetic mean (colonies/100 ml) of five replicates. 'Groups enclosed by the same letter are not significantly different (P = 0.05) by Tukey's significant difference technique (32). d FAC, Free available chlorine; TAC, total available chlorine; ND, not detectable. relationships (data not shown). Identification of organisms. Preliminary studies were made to identify the spectrum of coliforms enumerated by the MPN and MF methods from the sewage-contaminated tap water. Colonies were picked from plates from the MF method and from plates subcultured from the positive MPN tubes. Seven species of organisms (six coliforms plus Aeromonas hydrophila) were identified from 121 isolations. Overall, the most frequent isolates were A. hydrophila (43%), Citrobacter freundii (16.5%), and Enterobacter agglomerans (13%). Escherichia coli represented only 9% of the species isolated. Chlorination of water changed the spectrum of D APPL. ENVIRON. MICROBIOL. bacteria recovered by MF, presumably by selection of the more chlorine-resistant species. In the chlorinated waters, decreased recoveries of Klebsiella pneumoniae, E. coli, and C. freundii were somewhat compensated by higher recoveries of K. oxytoca as compared with the water before chlorination (Table 3). In general, the MPN and MF methods enumerated a similar spectrum of bacteria from the chlorinated water in both experiments. The main difference obtained was a higher incidence of A. hydrophila in the MF analysis in the two studies. DISCUSSION The MF method has gained wide acceptance

VOL. 40, 1980 TABLE 2. Comparison ofmf and MPN total coliform enumeration methods MPN MPN/ 95% con- Mean MF Signifino. 100 ml fidence MFb range' canced limitsa 1 330 110-930 315 285-371 NS 2 170 43-490 140 89-176 NS 3 79 25-190 114 91-136 NS 4 23 7-170 91 73-102 P < 0.0l 5 5 5-13 7 3-9 NS 6 13 3-13 23 17-38 NS aas tabulated in reference 1. ' Arithmetic mean (colonies/100 ml) of five replicates obtained from the MF method, using the Johns- Manville GA membrane. 'Maximum and minimum (colonies/100 ml) from five replicates. d Statistical probability that the MPN and MF estimates differ from each other by chance (NS, estimates are not significantly different at P = 0.05). TABLE 3. Number of organisms identified from sewage-contaminated tap water before and after chlorination Before After chlorination chlorina- Species tion, MF MF MPN 1 2 1 2 1 2 Aeromonas hydro- 4 17 9 14 8 phila Klebsiella pneu- 3 2 1 1 6 moniae Klebsiella oxytoca 1 4 1 Enterobacter cloa- 1 1 1 cae Escherichia coli 2 3 2 4 Citrobacter freun- 4 1 13 2 diii Enterobacter agglomerans 2 2 1 3 8 as a means of enumerating total coliforms in drinking water and in environmental samples. Because of its speed, simplicity, precision, and reproducibility as well as the lower cost and space requirements, the method is almost universally preferred, but not universally applied, in analyzing drinking water. It has been demonstrated that physiological stress in general (2, 3, 17), and stress by chlorine in particular (20, 23, 29), can have lethal effects on some cells and leave others in a debilitated condition. The damaged cells may be recovered on nonselective medium but are not viable on more selective media (23, 29). It would be ideal to compare methods and membranes on authentic drinking water samples alone in order to obtain "naturally" stressed organisms, but due to the low COMPARISON OF MPN AND MF FOR TOTAL COLIFORIMS 189 coliform counts it would be impractical. Chlorination of water samples to obtain stressed organisms has been used in previous studies (19, 23). The sewage-fortified, chlorinated tap water used in this study can be considered to be representative of inadequately treated drinking water or water that has become recontaminated with sewage microorganisms in the distribution system. The presumptive MF counts were found to be equivalent in five of six experiments and superior to the MPN confirmed results in the other experiment. Membrane filter brand-to-brand variability was found to play a significant role in these results; the range of results in the comparative study showed that, in a given experiment, the membranes with best recoveries enumerated from two to four times more coliforms than the membranes with the lowest recoveries. This is a les significant difference than those previously reported for fecal coliform results (35), probably because of the high temperature stress in the latter during incubation. Nevertheless, the difference may well be worth considering by individual laboratories in their continued efforts in increasing sensitivity of the drinking water analyses. It could, for example, give an increase in sensitivity equivalent to doubling or quadrupling the sample volume. The differences seen among the membrane filters in their ability to resuscitate total coliforms do not appear to be influenced in a direct manner by the surface pore diameters as are fecal coliform recoveries (30, 35). The surface pore structure of most of these membrane filter brands has already been examined by Dutka and Tobin (11); after analysis of these data, no obvious relationship with total coliform resuscitation was found. They do, however, confirm the marked differences observed in structure between the (autoclave sterilized in the laboratory) and (ethylene oxide sterilized by the manufacturer); these differences may be reflected in the disparate total coliform recoveries by the two membranes. Some data were obtained that indicated some differences in selectivity of the MF and MPN methods; in particular, the MF method showed greater selectivity for Aeromonas, a major cause of inflated coliform counts. In examining these isolates further, none of them confirmed as coliforms when transferred to lauryl tryptose broth. Clark and Pagel (5) have stated that current methods have not routinely distinguished between Aeromonas and coliform bacteria. This has been considered a drawback in the use of MF results for presumptive total coliform counts in drinking water; however, Grabow and du

190 TOBIN, LOMAX, AND KUSHNER Preez (14) make an excellent case for retaining Aeromonas with total coliforms in indicating the overall quality of drinking water. They cite the potential pathogenicity and possible transfer of resistance factors from Aeromonas, as well as the ability of adequate water treatment methods to eliminate these organisms, as reasons not to exclude them from enumeration by the cytochrome oxidase test. Among membrane filter media, the m-endo LES has been selected on the basis of its superior sensitivity and selectivity (10, 14). The present study has demonstrated that the MF method gives results equivalent to those given by the MPN method even with chlorine-stressed coliforms. For other applications or types of water, different brands of membrane filters may be experimentally shown, in a manner similar to the present study, to give optimum results. This choice, based on experimental evidence, may help ensure that optimum results are obtained. ACKNOWLEDGMENTS We thank C. Breuil and B. J. Dutka for helpful discussions and P. St.-Onge of the Green Creek Sewage Treatment plant for providing water samples. This work was supported, in part, by a contract from Health and Welfare Canada to the University of Ottawa. LITERATURE CITED 1. American Public Health Association. 1975. Standard methods for the examination of water and wastewater, 14th ed. American Public Health Association, Washington, D.C. 2. Bissonnette, G. K., J. J. Jezeski, G. A. McFeters, and D. G. Stuart. 1975. Influence of environmental stress on enumeration of indicator bacteria from natural waters. Appl. Microbiol. 29:186-194. 3. Bissonnette, G. K., J. J. Jezeski, G. A. McFeters, and D. G. Stuart. 1977. Evaluation of recovery methods to detect coliforms in water. Appl. Environ. Microbiol. 33: 590-595. 4. Brodsky, M. H., and D. A. Schiemann. 1975. Influence of coliform source on evaluation of membrane filters. Appl. Microbiol. 30:727-730. 5. Clark, J. A., and J. E. Pagel. 1977. Pollution indicator bacteria associated with municipal raw and drinking water supplies. Can. J. Microbiol. 23:465-470. 6. Committee. 1972. Comparison of membrane filtration and multiple tube methods for the enumeration of coliformn organisms in water. J. Hyg. 70:691-705. 7. Coutant, C. C., and R. F. Shapple. 1966. Bacteria in an ixnpounding reservoir. J. Am. Water Works Assoc. 58: 1275-1277. 8. Dufour, A. P., and V. J. Cabelli. 1977. Comparison of membrane filter brands for recovery of the coliform group, p. 73-80. In R. H. Bordner, C. F. Frith, and J. A. Winter (ed.), Proceedings of the Symposium on the Recovery of Indicator Organisms Employing Membrane Filters. Report no. EPA-600/9-77-024. U.S. Environmental Protection Agency, Washington, D.C. 9. Dutka, B. J. 1973. Coliforms are an inadequate index of water quality. J. Environ. Health 36:39-46. 10. Dutka, B. J., and S. E. Tobin. 1976. Study on the efficiency of four procedures for enumerating coliforms in water. Can. J. Microbiol. 22:630-635. 11. Dutka, B. J., and R. S. Tobin. 1980. Ultrastructural and APPL. ENVIRON. MICROBIOL. other factors influencing the suitability of membrane filters for enumerating fecal coliforms. In B. J. Dutka (ed.), Membrane filtration: techniques, applications and problems. Marcel Dekker, New York. 12. Environmental Health Directorate. 1977. Microbiological quality of drinking water. Health and Welfare Canada, Ottawa, Canada. 13. Geldreich, E. E., M. J. Allen, and R. H. Taylor. 1978. Interferences to coliform detection in potable water supplies, p. 13-19. In C. W. Hendricks (ed.), Evaluation of the microbiology standards for drinking water. Report no. EPA-570/9-78-OOC. U.S. Environmental Protection Agency, Washington, D.C. 14. Grabow, W. 0. K., and M. du Preez. Comparison of m- Endo LES, MacConkey and Teepol media for membrane filtration counting of total coliform bacteria in water. Appl. Environ. Microbiol. 38:351-358. 15. Health and Welfare Canada. 1979. Guidelines for Canadian drinking water quality, 1978. Supply and Services Canada, Ottawa, Canada. 16. Hendricks, C. W. 1978. Exceptions to the coliform and the fecal coliform tests, p. 99-145. In G. Berg (ed.), Indicators of viruses in water and food. Ann Arbor Science, Ann Arbor, Mich. 17. Hoadley, A. W., and C. M. Cheng. 1974. The recovery of indicator bacteria on selective media. J. Appl. Bacteriol. 37:45-50. 18. Hoffmann, D. A., J. H. Kuhns, R. C. Stewart, and E. I. Crossley. 1964. A comparison of membrane filter counts and most probable numbers of coliform in San Diego's sewage and receiving waters. J. Water Pollut. Control Fed. 36:109-117. 19. Kinney, E. C., D. W. Drummond, and N. B. Hanes. 1978. Effects of chlorination on differentiated coliform groups. J. Water Pollut. Control Fed. 50:2307-2312. 20. Lin, S. 1973. Evaluation of coliform tests for chlorinated secondary effluents. J. Water Pollut. Control Fed. 45: 498-506. 21. LUn, S. D. 1976. Evaluation of and HC membrane filters for the enumeration of indicator bacteria. Appl. Environ. Microbiol. 32:300-302. 22. LUn, S. 1977. Evaluation of methods for detecting coliforms and fecal streptococci in chlorinated secondary sewage effluents, p. 113-131. In R. H. Bordner, C. F. Frith, and J. A. Winter (ed.), Proceedings of the Symposium on the Recovery of Indicator Organisms Employing Membrane Filters. Report no. EPA-600/9-77- 024. U.S. Environmental Protection Agency, Washington, D.C. 23. Maxcy, R. B. 1970. Non-lethal injury and limitations of recovery of coliform organisms on selective media. J. Milk Food Technol. 33:445-448. 24. McCarthy, J. A., J. E. Delaney, and R. J. Grasso. 1961. Measuring coliforms in water. Water Sewage Works 108:238-242. 25. McKee, J. E., and R. T. McLaughlin. 1958. Application of molecular filter techniques to the bacterial assay of sewage. II. Experimental results for settled sewage. Sewage Ind. Wastes 30:129-137. 26. McKee, J. E., R. T. McLaughlin, and P. Lesgourgues. 1958. Application of molecular filter techniques to the bacterial assay of sewage. III. Effects of physical and chemical disinfection. Sewage Ind. Wastes 30:245-252. 27. Presswood, W. G., and D. K. Strong. 1978. Modification of M-FC medium by eliminating rosolic acid. Appl. Environ. Microbiol. 36:90-94. 28. Schaffer, D. J., M. C. Long, and K. G. Janardan. 1974. Statistical analysis of the recovery of coliform organisms on and Millipore membrane filters. Appl. Microbiol. 28:605-607. 29. Scheusner, D. L., F. F. Busta, and M. L Speck. 1971. Injury of bacteria by sanitizers. Appl. Microbiol. 21:41-45.

VOL. 40, 1980 COMPARISON OF MPN AND MF FOR TOTAL COLIFORIMS 191 30. Sladek, K. J., R. V. Suslavich, B. I. Sohn, and F. W. Dawson. 1975. Optimum membrane structures for growth of coliform and fecal coliform organisms. Appl. Microbiol. 30:658-691. 31. Standridge, J. H. 1976. Comparison of surface pore morphology of two brands of membrane filters. Appl. Environ. Microbiol. 31:316-319. 32. Steel, R. G. D., and J. H. Torrie. 1960. Principles anid procedures of statistics. McGraw-Hill Book Co., New York. 33. Stuart, D. G., G. A. McFeters, and J. E. Schillinge r. 1977. Membrane filter technique for the quantification of stressed fecal coliforms in the aquatic environment. Appl. Environ. Microbiol. 34:42-46. 34. Thomas, H. A., Jr. 1955. Statistical analysis of coliform data. Sewage Ind. Wastes 27:212-222. 35. Tobin, R. S., and B. J. Dutka. 1977. Comparison of the surface structure, metal binding and fecal coliform recoveries of nine membrane filters. Appl. Environ. Microbiol. 34:69-79. 36. U.S. Environmental Protection Agency. 1976. National interim primary drinking water regulations. Report no. EPA-570/9-76-003. U.S. Environmental Protection Agency, Washington, D.C.