APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Oct. 1985, p. 755762 99224/85/17558$2./ Copyright 1985, American Society for Microbiology Vol. 5, No. 4 Holding Effects on Coliform Enumeration in Drinking Water Samples AUDREY E. McDANIELS,* ROBERT H. BORDNER, PETER S. GARTSIDE, JOHN R. HAINES, KRISTEN P. BRENNER, AND CLIFFORD C. RANKIN Biological Methods Branch, Environmental Monitoring and Support Laboratory, U.S. Environmental Protection Agency, Cincinnati, Ohio 45268 Received 31 January 1985/Accepted 28 June 1985 Standard procedures for analying drinking water stress the need to adhere to the time and temperature conditions recommended for holding samples collected for microbiological testing. The National Drinking Water Laboratory Certification Program requires compliance with these holding limits, but some investigators have reported difficulties in meeting them. Research was conducted by standard analytical methods to determine if changes occurred when samples were held at 5 and 22 C and analyed at, 24, 3, and 48 h. Samples were analyed for coliforms by the membrane filter and fermentationtube procedures and for heterotrophs by the pour plate method. A total of 17 treated water samples were collected from a large municipal distribution system from August to December 1981, and 12 samples were collected from January to May 1983. The samples were dosed with coliforms previously isolated from the water system, Enterobacter cloacae in 1981 and Citrobacterfreundii in 1983. The coliform counts declined linearly over time, and the rates of decline were significant at both 5 and 22 C. Within 24 h at 22 C, levels of E. cloacae and C. freundii decreased by 47 and 26%, respectively, and at 5 C, E. cloacae numbers declined by 23%. Results from these representative laboratorygrown coliforms reinforced those previously obtained with naturally occurring coliforms under the same experimental conditions. Significantly, some samples with initially unacceptable counts (greater than 4/1 ml) met the safe drinking water limits after storage at 24 h at 5 and 22 C and would have been classified as satisfactory. In contrast to the coliform losses, heterotrophic plate counts of samples held at 22 C for 3 and 48 h increased by.5 to 2.5 orders of magnitude and often interfered with coliform counts on the membrane filter. Heterotrophic counts of the same samples held at 5 C for 3 and 48 h declined slightly. Based on the results of this study, it is recommended that drinking water samples be iced and that they be analyed as soon as possible on the day of collection to minimie changes in bacterial densities. Aquatic microbiologists and public health authorities have long recognied the importance of adhering to sample collection procedures and holding limits that will provide valid data from the microbiological analysis of water. The American Public Health Association (1) and the Environmental Protection Agency (3) stipulate that all water samples, potable, ambient, recreational, and wastewater, should be iced or refrigerated, analyed as soon as possible after collection, and allowed a maximum transit time of 6 h. However, current regulations permit drinking water samples to be transported or held at ambient temperature for up to 3 h when sent by mail or common camer. In conformance with the Safe Drinking Water Act (9), the Interim Primary Drinking Water Regulations (4) specify the sampling and analytical procedures for the microbiological analysis of public water supplies and state that only certified laboratories may perform these analyses. The nationwide Drinking Water Laboratory Certification Program is based on technical criteria and procedures (13) which include sample collection and holding conditions. State water laboratories that serve large geographic areas, receive samples from distant water supplies, and do not have efficient mail or other sample transport systems may not receive samples within the designated holding time. Some states have challenged the application of these holding limits to drinking water samples which are normally expected to be of high quality and low bacterial density. The purpose of this study was to obtain more definitive * Corresponding author. 755 data by sampling a distribution system representative of a public drinking water supply, following a carefully planned experimental design and using two recent coliform isolates from a potable water supply (Citrobacter freundii and Enterobacter cloacae) to compare the results with those from naturally occurring coliforms. The use of coliform spikes made it possible to adjust the concentrations to correspond to the low ranges of coliforms usually found in drinking water (approximately 2/1 ml or less) when positive samples occur. This investigation was carried out concurrently with a study on the recovery of naturally occurring coliforms in the same distribution system (8), because the presence of natural coliforms in the samples could not be predicted and data from spiked samples could be obtained easily. It was expected that if the responses were similar, the results from laboratorygrown coliforms would reinforce those from the natural coliforms. The statistical evaluations were based on data obtained by the more precise membrane filter (MF) technique, although both the MF and the mostprobablenumber (MPN) methods were used so that comparisons of coliform results from the two methods could be made. Heterotrophic plate counts (HPCs) were performed throughout the study to determine sample holding effects on heterotrophic populations. HPCs are used to monitor the efficiency of water treatment, show possible interferences with the coliform detection, predict the potential presence of opportunistic pathogens, and indicate increased possibilities for taste, odor, and corrosion problems. Both the U.S. Environmental Protection Agency and the National Acad Downloaded from http://aem.asm.org/ on December 18, 218 by guest
756 McDANIELS ET AL. APPL. ENVIRON. MICROBIOL. Drinking Water Sample (1983) 6 liters I liter Coliform Addition (C. Freundii) I r Microbiological Parameters 3 Subsamples 3 Subsamples 5 C 22 C (See 5 C) Subsample 1,2,3 (8 liters) MF MPN HPC Volumes Dilutions (1, 5 ml) (3) Replicates Replicates Replicates (3 per volume) (3 sets (3 per dilution) of 1 tubes) emy of Sciences recommend that HPCs be included in the monitoring of drinking water systems (4, 1). MATERIALS AND METHODS Sampling sites. Finished drinking water in the distribution system of a large metropolitan city in the midwestern United States was sampled from taps in the laboratory, from a firehouse in a residential location, and from fire hydrants found near the ends of mains. Sample collection and processing. A total of 17 samples were collected weekly or biweekly from August to December 1981, and 12 were collected from January to May 1983. Before collection of the samples, water was allowed to run from the outlet until it was clear and the temperature was stabilied. A 5 to 6liter sample was collected in a sterile polypropylene bottle. A 1liter sample was removed for measurement of chemicalphysical parameters, and the remaining portion was dechlorinated by adding 1 mg of Na2S23 per liter of water. The sample was returned to the laboratory within 1 h of collection. Samples collected in 1981 and 1983 were inoculated with E. cloacae and C. freundii, respectively. These coliform cultures were recent isolates from one of the sampling sites. They were maintained on nutrient agar slants at 4 C before use. For sample inoculation, the organisms were grown in 1 ml of tryptic soy broth (Difco Laboratories, Detroit, Mich.) for 24 h at 35 C and centrifuged at 75 x g for 2 min. The pellets were washed and centrifuged three times in sterile, tripledistilled water. To obtain coliform counts of 4 to 2/1 ml, we added 25 to 35 ml of a 19 dilution of the pellets in tripledistilled water to 5 to 6 liters of the sample. Experimental procedure. The experimental design for the 1983 study is shown in Fig. 1. After inoculation with the test organisms, the samples were thoroughly shaken and divided into six subsamples of 4 and 8 liters for the 1981 and 1983 series, respectively. To simulate ambient temperatures, half of the subsamples were designated to be used for assays at 5 C and the other half were designated to be used at 22 C. All subsamples were stored in the dark. To provide the initial Labo ratory Metals Nonmetals Conductivity Turbidity Physical Chemical Parameters 1 FIG. 1. Schematic outline of sampling and analytical procedures. F Field Initial Temperature Chlorne (Free, Total) ph counts, we performed microbiological assays immediately after the subsamples were prepared. Two sets of initial counts were obtained, one from the three subsamples randomly selected for incubation at 5 C and one from the three subsamples incubated at 22 C. These analyses were repeated after 2, 6 or 12, 24, 3, and 48 h in 1981, and after 6, 24, 3, and 48 h during the 1983 study. For the MF procedure, 5 and 1ml volumes of each subsample were analyed in duplicate in 1981. In 1983, 1 and 5ml volumes were analyed in triplicate. Fivetube, threedilution MPNs were inoculated from one subsample in 1981. A 1tube MPN test was performed on three subsamples in 1983 for direct comparison with the volumes analyed by the MF method. HPCs were performed in triplicate on three dilutions of one representative subsample in 1981 and of three subsamples in 1983. The above analyses were performed on samples kept at both holding temperatures. Microbiological analyses. Total coliform and heterotrophic bacterial densities were determined by standard procedures (1, 3) for MF, MPN, and HPC analyses. The confirmed MPN analyses utilied lauryltryptose broth and brilliant greenlactosebile media (Difco). The MF tests were performed with.45,umporesie HA membrane filters (Millipore Corp., Bedford, Mass.) and mendo agar (Difco). Nonsheen colonies from mendo agar were verified biochemically with the API 2E system (Analytab Products, Plainview, N.Y.). The pour plate method was used to measure the heterotrophic bacteria on plate count agar (Difco) with incubation at 35 C for 48 h. Chemical and physical tests. Initial temperature, ph (Corning ph meter 4; Corning Glass Works, Medford, Mass.), and free and total chlorine residuals by the N,Ndiethylpphenylenediamine method (Hack Kit CN57; Hach Co., Ames, Iowa) were measured in the field. Specific conductivity (model 31 conductivity bridge; Yellow Springs Instrument Co., Yellow Springs, Ohio) and turbidity (model 31A turbidimeter; Hach) were determined in the laboratory. The 1983 water samples were analyed for the inorganic constituents NH3, N2, N2 plus NO3, NO3, total dissolved solids, and total phosphorus by the Physical and Chemical Downloaded from http://aem.asm.org/ on December 18, 218 by guest
VOL. 5, 1985 COLIFORM ENUMERATION 757 1. A F ti nu^ra mom O O 1 B r% 6 WAM m 24;xi 4~ EIEIiI I I I I I I I I I I 1 9F _ O.l _ 'P. op ~ 'O ''7 ' 9* c% t w ineij FIG. 2. MF coliform counts recovered from samples collected in 1981 (A and B) and 1983 (C and D). The maximum contaminant level is 4/1 ml. Containment Removal Branch of the Municipal Environmental Research Laboratory, U.S. Environmental Protection Agency, Cincinnati, Ohio. Atomic absorption measurements were made for Al, Be, Cd, Cr, Co, Cu, Fe, Pb, Mn, Hg, Ni, Se, V, and Zn by the Physical and Chemical Methods Branch of the Environmental Monitoring and Support Laboratory, U.S. Environmental Protection Agency, Cincinnati, Ohio. The methods used for inorganic and metals analyses have been described in the U.S. Environmental Protection Agency manual Methods for Chemical Analysis of Water and Wastes (12). Statistical analyses. Total coliform and HPC data were transformed for statistical analyses to achieve the best homogeneity of variance and normal distribution by the following formulas: log1 [(count per 1 ml) +.11 and log1 [(count per ml) +.1] for E. cloacae MF and HPC, respectively; and (count 6/.6) per 1 ml, logl [(count per 1 ml) +.1] and logl [(count per ml) + 1] for C. freundii MF, MPN, and HPC, respectively. C. freundii MF and MPN transformed counts were compared for precision and linearity of regression. The counts from each holding time interval were normalied against initial values by the formula log1 (N,INO), where No is the mean value of the initial counts and N, is the mean obtained for each of the holding times. Paired t tests were used to determine significant differences in recoveries between holding temperatures and holding times across all samples. Analysis of variance was used with the paired t test for C. freundii data. Pearson productmoment correlations and multiple regressions wete used to compare the initial HPCs and coliform MF densities with ph, initial temperature, free chlorine, specific conductivity, and turbidity. All identified statistical differences were significant at the 95% confidence level. Rates of coliform decline, adjusted to the means of ph values and date of collection, were determined by maximum likelihood estimates. Predicted initial hour counts were completed by the expression antilog (log1 Y + 1X) where Y is the count, X is the hour of the observed count, and,b is the maximum likelihood estimate of the rate of coliform decline. The time required for 5% loss in counts was also generated by the model. All statistical analyses were performed with computer programs in the Statistical Analysis System (SAS), SAS Institute, Inc., Cary, N.C. RESULTS Coliform MF recovery. The mean values of the MF coliform counts from the samples collected during the two sampling seasons, held at two temperatures, and analyed over time are shown in Fig. 2. The mean values from 17 samples collected in 1981, dosed with E. cloacae, and held at Downloaded from http://aem.asm.org/ on December 18, 218 by guest
5~~~~~~~ 758 McDANIELS ET AL. Di1 1 T 4 1.5 I MF Ai 5eC NATURAL COLJFORMS 2 22C NATUFAL COUFORMS + s&c E. CLOACAE 22C E. CLOACAE 2 I_ 1 2 3 4 5 TIME (hours) FIG. 3. Coliform maximum likelihood estimates and 95% confidence interval based upon the mean loglo (N,INO) rates of decline for predicting count losses over time. T5, time required for 5% loss in counts. 5 and 22 C are shown in Fig. 2A and B, respectively. The results from 12 samples collected in 1983, dosed with C. freundii, and held at 5 and 22 C are depicted in Fig. 2C and D, respectively. Initial counts of samples inoculated with E. cloacae ranged from 1 to 2/1 ml, and those dosed with C. freundii ranged from less than 1 to 15/1 ml. The bar graphs show a general decline in coliforms throughout the 48h holding period at 5 and 22 C during both sample series. A lower rate of decline in counts was observed for C. freundii (Fig. 2C and D) than for E. cloacae (Fig. 2A and B) at both holding temperatures. Extensive variation in coliform recoveries can be seen among the individual samples (Fig. 2). In samples dosed with E. cloacae and held for 48 h at 5 C (Fig. 2A), the maximum loss was 8%; in contrast, one sample (Sept. 8a) showed no change. In samples held for 48 h at 22 C (Fig. 2B), the maximum loss was 91% and the minimum loss was 1%. The maximum decrease from initial counts for C. freundii in 48 h at 5 C (Fig. 2C) was 58% and at 22 C (Fig. 2D) was 87%. One sample dosed with C. freundii (March 22) showed minimal change at each temperature. It should be noted (Fig. 2) that initial counts for the majority of samples were greater than 4/1 ml. However, the counts for several of these samples declined to less than TABLE 1. Percent recoveries of coliforms by the MF method from samples held for 48 h at 5 and 22 C % Recovery Temp ('C) Time (h) E. cloacae C. freundii Mean Range Mean Range 5 6 81.3 3418 96.2 719 24 76.8 23124 94.1 79121 3 67.4 2418 94.3 68119 48 54.5 2296 8.9 3916 22 6 77.3 2815 86. 5212 24 52.7 1386 74.1 3718 3 37.8 169 68.4 2112 48 37.2 498 61.9 1299 APPL. ENVIRON. MICROBIOL. 4/1 ml by 3 and 48 h and would then have been recorded as satisfactory. After 3 h at 22 C, the counts in three E. cloacae samples (Fig. 2B; Sept. 21, Oct. 26, and Nov. 2) and two C. freundii samples (Fig. 2D; Feb. 7 and Apr. 11) that were initially unsatisfactory (slightly greater than 4) were less than 4/1 ml. The mean and range of percent recoveries of the dosed coliforms in samples held at both temperatures over the selected time intervals are listed in Table 1. Losses for samples dosed with E. cloacae were 33% at 5 C and 62% at 22 C after 3 h. The decreases for C. freundii counts were less striking (6% at 5 C and 33% at 22 C after 3 h). The ranges given for percent recoveries overlapped at all time intervals, illustrating the variation of results among individual samples. The paired t test indicated that the losses in coliform recoveries were significant within 24 h at 22 C for samples dosed with E. cloacae and C. freundii and at 5 C for those dosed with E. cloacae; however, at 5 C, losses were not statistically significant until 48 h for C. freundii. Mean rates of decline for E. cloacae, C. freundii, and natural coliforms expressed as log1 (N,/NO) per hour of holding time are shown in Table 2. The rate of decline at 22 C for untransformed natural coliform counts was 2.4 times that of E. cloacae and 4.5 times that of C. freundii. At 22 C, the rates of decline were 2.9, 4.9, and 2 times those at 5 C for natural coliforms, E. cloacae, and C. freundii, respectively. Models for estimating coliform rates of decline. A maximum TABLE 2. Mean rates of decline for coliform MF counts in samples held at 5 and 22C" Mean rates of decline at: Sample category 5 C 22 C E. cloacae.56 +.41.112 +.65 C. freundii.12 +.1.59 +.61 Natural coliforms.93 +.67.266 ±.83 " Hourly rate of change in log,) (N,IN, derived from the h to 48h values. Downloaded from http://aem.asm.org/ on December 18, 218 by guest
VOL. 5, 1985 COLIFORM ENUMERATION 759 A 1 "q.. FIG. 4. of 9 5 I1 1.T,9 1 2 3 4 6 TIME (hours).5 B o I.6 MF" OO5C C. EUNDII U o 22 c 5 FREUNDII _U SOC NATURAL COLIFORMS * U 22'C NATURAL COUFORMS * 6 L. CLOACAE 22C E. CLOACAE MPN 5C C FREUNDII 22C C.FREUNDII 1 * S*C NATURAL COUFORMS * 22C NATURAL COUFORMS A c L CLOACAE * 22 C E. CQa&AE 1 2 3 4 K6 TIME (hours) Mean coliform recoveries normalied to ero hour in samples held over time for MF (A) and MPN (B) tests. likelihood estimate model, similar to a previously described model for natural coliforms (8), was prepared for the samples dosed with E. cloacae in this study. A graph based on the model is shown in Fig. 3. On the basis of the model, initial E. cloacae counts of 5 to 6/1 ml in samples held at 5 C and from 8 to 12/1 ml in samples held at 22 C could decline to acceptable levels (4/1 ml or less) by 3 h. The time required for 5% loss in counts at 22 C was 24 h for E. cloacae, and 9 h for natural coliforms. Comparison of MF and MPN recovery. Coliform recovery over all samples analyed by the MF and MPN methods followed similar trends of decline (Fig. 4). Regression analysis detected no significant differences in results from the two methods. Noncoliform growth on mendo agar. The organisms most frequently isolated on mendo agar in addition to coliforms were Aeromonas hydrophila and Acinetobacter spp. Also identified were Pseudomonas, Alcaligenes, and Flavobacterium spp. The growth of background colonies in some samples held for 3 h or longer interfered with coliform growth and made counting of coliform colonies extremely difficult (Fig. 5). Effects of physical and chemical parameters. The ranges of physical and chemical parameters found in water samples inoculated with coliforms were as follows: initial temperature, 7 to 2 C; free chlorine,. to 2.2 mg/liter; ph, 6.98 to 9.3; turbidity,.2 to 17 nephelometric turbidity units; and specific conductivity, 24 to 49 RS/cm. Regression analysis indicated that for samples held at 22 C, only ph was significantly correlated with rates of decline, with the greatest rate of decline corresponding to the highest ph. At 5 C, none of the parameters were significantly correlated with rates of decline of coliforms. The concentrations of metals, nonmetals, and other water quality parameters were either below the level of detection or within expected limits for drinking water. HPC responses. Initial HPCs for samples examined in 1981 ranged between 11 and 15/ml, with lower counts found during the fall and winter seasons. Initial HPCs for 1983 samples ranged from less than 1 to 15/ml. After 48 h at 5 C, half of the samples from both sampling periods declined in counts while the other half increased, but the changes were not significant. After 48 h at 22 C, HPCs from the dosed samples taken in 1981 increased by 1 to 3 orders of magnitude, and HPCs from the dosed samples taken in 1983 increased by 1 to 4 orders of magnitude. All of these changes were significant (Fig. 6 and 7). DISCUSSION In this investigation, coliform counts in dosed drinking water samples generally declined throughout the 48h storage period, following the trend previously reported for Downloaded from http://aem.asm.org/ on December 18, 218 by guest
76 McDANIELS ET AL. FIG. 5. MF mendo plates with coliforms (A) and coliforms and background colonies (B). naturally occurring coliforms (8). The lowest percent recoveries were obtained from distribution system samples that contained natural coliforms. These results were followed in increasing order of percent recovered by the results from samples dosed with E. cloacae and C. freundii. These results may be due to the fact that natural coliforms are subjected to greater stress in the distribution system than are laboratorygrown cultures. Stress factors include exposure to free chlorine residual, the lownutrient water environment, and reduced water temperatures during the winter months. Recovery of stressed or weakened coliforms was not attempted because the purpose of this investigation was to determine.appl. ENVIRON. MICROBIOL. sample holding effects when the analyses were conducted by standard methods. However, recent investigations on the survival of stressed coliforms have led to the development of special media for their recovery (2), and the new formulations may eventually replace those in current practice. It is often difficult to find natural coliforms in public water supplies because they should be eliminated by the treatment process. To obtain more definitive data on the response of coliforms held in finished waters at different temperatures, isolates representing the species frequently found in the system were added to the water samples and enumerated over time. These isolates were maintained in laboratory media that may have increased their survival potential beyond that of naturally occurring coliforms. However, both laboratorygrown and natural coliforms declined in number at significant rates when held in water over time at the two temperatures. Although rates of decline differed for the two groups, the trend was identical. These laboratorygrown coliforms served as indicators to predict the responses of natural coliforms. The same mathematical model was applicable for predicting both natural and dosed coliform losses from similar waters and seasons of collection (Fig. 3). Similarly constructed models adapted for the water source might be useful in identifying and solving sample holding problems associated with particular geographical areas. Both groups of coliforms showed extensive variation among and between samples, and for both groups, initial counts of more than 4/1 ml declined to less than 4/1 ml by 24, 3, and 48 h. It is not clear why a lower percentage of coliforms was recovered for E. cloacae than for C. freundii over the 48h sampleholding period. However, the two coliform species would not be expected to respond ih exactly the same way to the same environmental conditions because their genetic composition is not identical. Since the two studies were separated by a period of nearly 2 years and were conducted with samples collected at different sites in the distribution system, there may have been sufficient changes in the chemical characteristics of the distribution system water and in the bacterial cultures to result in differences in the responses of the two species. Another important factor in coliform survival is the effect of noncoliform populations in the samples. When noncoliforms exceed 1,/ml, the detection of coliforms may be masked (5). With the exception of one outlier sample, initial HPCs of samples spiked with C. freundii were lower than those spiked with E. cloacae. For the same time period, the initial mean HPC (14/ml) of E. cloacaedosed samples increased by only 1 order of magnitude. The longer exposure to high backgrounnd counts which E. cloacae sustained could have adversely affected detection and survival of this coliform. As might be expected, the highest HPC rates of growth at ambient temperatures over 48 h occurred in the samples with the lowest initial counts, in which competition for nutrients and the concentrations of toxic metabolic products were less. Species reported to be antagonistic to coliforms (6) include Pseudomonas and Flavobacterium spp., both of which were isolated from the treated waters used in this study. One problem encountered was the prolific growth of organisms such as Acinetobacter and Aeromonas spp. on mendo plates. Enumeration of coliforms on MFs from samples with high background populations of heterotrophic bacteria became increasingly difficult with holding times from 24 to 48 h. Similar findings have been reported previously (7). Downloaded from http://aem.asm.org/ on December 18, 218 by guest
VOL. 5, 1985 COLIFORM ENUMERATION 761 i i U) i FIG. 6. SAMPLE COLLECTION DATES 1981 SAMPLE COLLECTION DATES 1981 j 2N.I I u SAMPLE COLLECTION DATES 1983 SAMPLE COLLECTION DATES 183 Heterotrophic bacteria counts recovered from samples held over time and dosed with E. cloacae at 5 C (A) and 22 C (B) and with C. freundii at 5C (C) and 22C (D). 4 2 1 HPC O{1 5C l 22C *@22C.... 22 C a 5C of <, Downloaded from http://aem.asm.org/ on December 18, 218 by guest Ot *~ ~~ ; ;.. ~ ~ ~ _: ; w 1 2 3 4 TIME (hours) FIG. 7. Mean heterotrophic bacteria recoveries normalied to erohour counts from samples held over time. Symbols: El, C. freundii (1983); *, natural coliforms (1981);, E. cloacae (1981). 5
762 McDANIELS ET AL. Rates of coliform decline were steepest during the months in which initial sample temperatures were lowest and when ph and specific conductivity were at the highest levels. The basic conclusion from regression analysis of the E. cloacae data was that rates of decline demonstrated a linear seasonal trend. This correlation was strongly evident in the data from samples stored at 5 C but occurred to a lesser degree at 22 C. The cause of this seasonal variation is unknown. Only ph had a statistically significant effect on rates of coliform decline at 22 C for both E. cloacae and C. freundii. The mean ph values for the 1981 and 1983 studies were 7.9 and 8.3, respectively, and the maximum values were 9. and 9.3, respectively. All of them were higher than the optimum ph range of 6.8 to 7.2. Large numbers of heterotrophic bacteria were found in the samples inoculated with E. cloacae. These samples contained the opportunistic pathogens Pseudomonas spp. and A. hydrophila. HPCs, therefore, may be a useful and easily performed indicator of potential human health haards and a measure of the efficiency of treatment plants. Standridge and Delfino (11) proposed that the present 3h limit for holding drinking water samples before analysis be extended to 48 h. The basis for their proposal was that little change occurred in MF and MPN coliform counts from samples analyed after holding for 24 and 48 h at ambient temperature. However, initial coliform counts were not recorded, so changes that might have occurred between and 24 h were not known. Therefore, there was no basis for this conclusion. Furthermore, in their study, 24h counts were not predictive of 48h counts by the x2 test. The data they presented showed that for samples that were coliform positive at 24 h, 42% (15 of 36) of MF coliformpositive samples and 31% (2 of 65) of MPN coliformpositive samples were coliform negative by 48 h. Such extensive losses cannot be considered insignificant changes. Also, background overgrowth was present on mendo plates for 2% (49 of 2,481) of the samples they examined, which may have masked the presence of coliforms. The conclusion drawn by Standridge and Delfino (11) is not supported by their own data or by the results of our study. Therefore, holding samples at ambient temperature for 48 h before analyses cannot be recommended. The following conclusions were made from this study. (i) Significant losses in coliform counts occur in some samples held for 24 h and may be even greater when held for 3 h. Coliforms in samples held for 24 h at 22 and 5 C showed losses as great as 47 and 23%, respectively. These losses increased to 62 and 33%, respectively, after 3 h. Rates of decline followed a linear rather than a curvilinear regression model. (ii) Because of these losses, coliform counts of approximately 18% of the samples which initially were unacceptable (greater than 4/1 ml) were acceptable after 3 h. (iii) Enumeration of coliforms by the MF method on mendo plates became more difficult as the HPCs increased APPL. ENVIRON. MICROBIOL. and interfered with accurate counting. (iv) Rates of decline for dosed coliforms reinforced similar but more rapid losses of natural coliforms. (v) Although it is useful and often necessary to spike samples in research studies, we do not recommend that an investigation of the effects of sample holding be based entirely on laboratorygrown cultures. Such research findings must be confirmed with field studies involving naturally occurring coliforms. Based on the results of this study, it is recommended that drinking water samples be iced and that they be analyed as soon as possible on the day of collection to minimie changes in bacterial densities. ACKNOWLEDGMENTS We thank C. I. Weber for reviewing the manuscript and offering valuable comments and D. W. O'Neill for excellent technical assistance. LITERATURE CITED 1. American Public Health Association. 1985. Standard methods for the examination of water and wastewater, 16th ed. American Public Health Association, Washington, D.C. 2. Andrew, M., and A. Russell (ed.). 1984. The revival of injured microbes, p. 372382. Academic Press, Inc., Orlando, Fla. 3. Bordner, R., J. Winter, and P. Scarpino (ed.). 1978. Microbiological methods for monitoring the environment, water and wastes. EPA 6/87817. U.S. Environmental Protection Agency, Cincinnati, Ohio. 4. Federal Register. 1975. National interim primary drinking water regulations. Fed. Regist. 4:595759572. 5. Geldreich, E. E., H. D. Nash, D. J. Reasoner, and R. H. Taylor. 1972. The necessity of controlling bacterial populations in potable waters: community water supply. J. Am. Water Works Assoc. 64:59662. 6. Herson, D. S., and H. T. Victoreen. 198. Hindrance of coliform recovery by turbidity and noncoliforms. EPA 6/2897. U.S. Environmental Protection Agency, Cincinnati, Ohio. 7. Hsu, S. C., and T. J. Williams. 1982. Evaluation of factors affecting the membrane filter technique for testing drinking water. Appl. Environ. Microbiol. 44:45346. 8. McDaniels, A. E., and R. H. Bordner. 1983. Effects of holding time and temperature on coliform numbers in drinking water. J. Am. Water Works Assoc. 75:458463. 9. Safe Drinking Water Act. Public Law 93523, December 16, 1974, 88 Stat. 166, 42 United States Code (USC) 3f. 1. Safe Drinking Water Committee. 1977. Drinking water and health: a report of the Safe Drinking Water Committee. National Academy of Sciences, Washington, D.C. 11. Standridge, J. H., and J. J. Delfino. 1983. Effect of ambient temperature storage on potable water coliform population estimations. Appl. Environ. Microbiol. 46:11131117. 12. U.S. Environmental Protection Agency. 1979. Methods for chemical analysis of water and wastes. EPA 6 4792. U.S. Environmental Protection Agency, Cincinnati, Ohio. 13. U.S. Environmental Protection Agency. 1982. Manual for the certification of laboratories analying drinking water: criteria and procedures quality assurance. EPA 57/9822. U.S. Environmental Protection Agency, Washington, D.C. Downloaded from http://aem.asm.org/ on December 18, 218 by guest