APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Aug. 1982, p. 453-4600099-2240/82/080453-08$02.00/0

Vol. 44, No. 2

Evaluation of Factors Affecting the Membrane FilterTechnique for Testing Drinking Water

S. C. HSU AND T. J. WILLIAMS*Bureau ofDisease Control and Laboratory Services, Michigan Department of Public Health, Lansing,

Michigan 48909

Received 18 November 1981/Accepted 26 April 1982

The following studies were done in response to questions regarding theadoption and use of the membrane filter (MF) technique for testing drinking waterfor the total coliform indicator group. A comparison with the most-probable-number technique showed that MF procedures with m-Endo agar LES weresomewhat superior to the most-probable-number methods in terms of numbers ofcoliform organisms recovered. Medium preparation and storage studies indicatedthat rehydration of m-Endo agar LES should be done with boiling water for lessthan 15 min, that m-Endo agar LES should not be exposed to light for more than 4to 6 h, and that m-Endo agar LES plates may be used for up to 4 weeks and brothverification media for up to 3 weeks under given storage conditions. MF culturecolonies were commonly found which did not produce sheen as expected forcoliforms and yet were verified as coliforms. The occurrence and morphology ofthese atypical colonies were studied. Parallel inoculation of both lauryl tryptose(LT) and brilliant green bile (BGB) broth was found to be a better colonyverification approach than recommended LT preenrichment before transfer toBGB. Comparison of parallel verification results indicated very little justificationfor the use ofLT medium in MF verification procedures. In the case of overgrownor confluent cultures, the best coliform recoveries resulted from swabbing the MFplate and directly inoculating BGB medium with the swab. The occurrence ofovergrowth was defined and evidence was collected suggesting that overgrowth isa function of sample holding time. Evaluation of routine test data and bacterialpopulation reductions as a function of time indicated that nonquantitativerecovery of coliforms may not be significantly affected for at least a 72-h sampleholding time.

The membrane filter (MF) technique, which isnow familiar to all environmental and sanitarymicrobiologists, has been fully accepted as aprocedure for testing drinking water for the totalcoliform indicator group and is an approvedmethod of analysis under the federal Safe Drink-ing Water Act regulations (1, 7). However, wide-spread use of the MF technique is relatively newcompared with the long history of most-proba-ble-number (MPN) methods for indicating drink-ing water safety. The Michigan Department ofPublic Health adopted the MF technique for alldrinking water analysis during 1974.Because of continuing concern over the valid-

ity of the technique, various aspects of theprocedure have been studied and evaluated.Before adopting the MF technique as a routinetest procedure, extensive comparisons of MFand MPN procedures were done. Because theMF technique occasionally gave overgrown cul-tures which potentially interfere with coliformgrowth, new standard techniques for dealingwith overgrown cultures were needed. Verifica-tion studies of many types ofMF colony growth

revealed that a substantial portion of nonsheencolonies gave positive brilliant green bile (BGB)broth fermentation results. Since this defines theMPN-confirmed coliform indicator group butcontradicts usual MF colony interpretation, achange in verification procedures was indicated.Preliminary experience with parallel verificationresults with both lauryl tryptose (LT) and BGBmedia also suggested further study of currentlyrecommended verification procedures (1, 7).Concerns over MF Endo medium stability areimplied in federal laboratory certification guide-lines (1). For this reason, aspects of preparationand storage of media were systematically stud-ied. Finally, the undefined effects on test resultsof sample holding times before testing continueto be a significant problem for centralized watersupply-monitoring programs, and additionaldata were collected in this regard.

MATERIALS AND METHODSRoutine MF and MPN culturing. MF culturing was

done as described previously (1). All procedures used

453

on August 12, 2020 by guest

http://aem.asm

.org/D

ownloaded from

454 HSU AND WILLIAMS

for preparing and incubating cultures were certified byfederal inspection. An agar medium, m-Endo agarLES, was selected for routine use based on findings byMcCarthy and Delaney (3) and McCarthy et al. (4).Comparison of fecal coliform culturing with m-FCbroth (1) with and without agar also suggested that anagar-based medium yields more rapid growth andincreased recovery. For routine test procedures, 100-ml portions of sample were measured and filtered withcylindrical funnel assemblies graduated at 50 and 100ml. Filters were then rinsed with sterile phosphate-buffered dilution water. Cultures were incubated onscreens over water in a closed container for 18 to 24 hat 35°C. MPN testing was also done as describedpreviously (1) by the five-tube (10-ml sample) proce-dure and with LT broth as the initial enrichmentmedium. Routine confirmation of LT cultures givinggas at 24 or 48 h was done by transfer to BGB brothfermentation tubes. Gas production in BGB within 48h at 35°C was considered a positive result, with theMPN index determined as shown previously (1).

Verification of MF colonies. Portions of MF colonieswere transferred with sterile hardwood applicatorsticks to LT or BGB fermentation tubes. For routinetesting, transfer was made from each colony verifiedto both LT medium and BGB medium at the end ofMFincubation. Organisms producing gas in LT mediumwithin 48 h at 35°C, but not in BGB medium, wereagain transferred from the positive LT culture to freshBGB medium. Only organisms producing gas in BGBcultures within 48 h of incubation for the first orsecond transfer were reported as coliforms. MF colo-ny counts were adjusted according to the percentageof selected MF colonies which gave positive BGBresults.MF colonies were selected for verification by the

following criteria. All colonies producing sheen werecandidates for verification. During analyst training,examples of all types of nonsheen colonies were givenverification testing. Experience gained during thistraining provided the judgement needed to recognize"atypical noncoliform" colonies and "atypical" colo-nies which are likely to give positive verification.Atypical colonies may be described as nucleated,having a milky appearance, or producing a markeddarkening of the medium below the colony; however,the training in recognition needs to be based on actualtesting experience. A maximum of five sheen or atypi-cal colonies was verified for routine test cultures.Sheen colonies were selected if present. If not, anycolonies judged to be "atypical" were verified.Overgrowth verification. MF cultures were consid-

ered as overgrown when isolated colonies appeared tocover better than half of the filter area or whenconfluent growth occurred. Semiquantitative resultswere obtained with individual streaks of material fromacross the overgrown culture. These were treated asdescribed for verification of MF colonies, with MFcolony count reported as greater than the number ofpositive streak transfers. More commonly, the entirearea of the culture was lightly brushed with a sterilecotton-tipped swab which was used to inoculate verifi-cation media. In this case, overgrowth results werestated only as positive or negative for coliforms.Inoculation and incubation of LT or BGB cultureswere otherwise done as described for single MF colo-ny verification.

APPL. ENVIRON. MICROBIOL.

MF and MPN comparative studies. During the rou-tine processing of water samples, approximately 10 to20 samples were randomly selected for comparisononce or twice a week for several weeks. These sam-ples were shaken; then a 50-ml portion was tested byMPN procedures, and 50 ml of the remaining samplewas filtered for MF culturing since the total samplesize of 120 ml did not permit testing of the standard100-ml sample used for routine MF testing. Results bythe two procedures were compared. Comparison workwas done by three different analysts at two differentlaboratories. In addition, low levels (ca. 10 organismsper 100 ml) of a known Escherichia coli strain wereinoculated into 100 samples of buffered dilution water,and these were tested by both MF and MPN proce-dures.Medium preparation and storage studies. Systematic

studies of the following factors affecting medium per-formance were done. (i) Because of the heat sensitivityof m-Endo agar LES, the effects of heating by twodifferent methods for different lengths of time werestudied. Rehydration of m-Endo agar LES (DifcoLaboratories, Detroit, Mich.) was done by placing themedium container on a hot plate or in a boiling waterbath. After the medium reached the boiling tempera-ture, portions were withdrawn at 5 min and at subse-quent 10-min intervals up to 60 min and used toprepare MF plates. Membrane filters were placed onthe solid medium and allowed to become moist. Inocu-lation of the filter surface was then done by lightlytouching it with the tip of an inoculation needle. Fourdifferent coliform strains known to produce sheenwere used for inoculation. After standard incubationthe quality of sheen was observed for each preparationcondition.

(ii) For studying the effect of light exposure, pre-pared MF plates were exposed before use to ambientlaboratory light for various periods of time up to 72 h.Laboratory light consisted of both fluorescent lightingand indirect sunlight. Exposed plates were inoculatedand incubated as described for the heating studies.After incubation, sheen production for exposed plateswas compared with that for unexposed plates from thesame lot of medium.

(iii) m-Endo agar LES degradation with time wasstudied by storage of lots of medium for up to 5 weeks.The same lot of dehydrated m-Endo agar LES wasused for all medium preparations. Agar plates wereprepared with loose-fitting petri dishes (60 by 15 mm)and stored in lots of 200 dishes in a cardboard boxsealed in a plastic bag. Stored media were refrigeratedat 5°C. For each period of storage, plates were inocu-lated and incubated as for the heating studies andcompared with freshly prepared medium.

(iv) When stored in the dark, the main factor affect-ing verification broth medium is thought to be evapo-ration of water from the medium. This was studied byobserving weight loss as a function of the time stored.Fermentation culture tubes were prepared with metalclosure caps, plastic closures, and plastic closures ontubes sealed in plastic bags. The average weight loss of10 tubes from each type of preparation was deter-mined as a function of the storage time.

Routine testing statistics. Data resulting from routinetesting activities were compiled for various periods oftime. Except for additional recordkeeping, all othersample processing and testing were done according to

on August 12, 2020 by guest

http://aem.asm

.org/D

ownloaded from

MF TECHNIQUE FOR TESTING DRINKING WATER 455

TABLE 1. Comparison of routine total coliformtesting results for drinking water samples by the MF

and MPN techniques'No. (% of total)b of samples that indicated:

Samples fromMPN and MF MPN MFanalyst agreement preferencec preferencec

A 545 (94.6) 6 (1.0) 25 (4.4)B 333 (91.2) 5 (1.4) 27 (7.4)C 385 (95.1) 4 (1.0) 16 (3.9)Overall 1,263 (93.8) 15 (1.1) 68 (5.1)a Comparison used 50-ml samples and direct m-

Endo, agar LES culturing.b The total numbers of samples for analysts A, B,

and C and overall were 576, 365, 405, and 1,346,respectively.

c Preferred method gave higher recovery or waspositive when the other procedure was negative.

standard written procedures which were not changedfor the period of data compilation. Samples involved inthese studies originated from the eastern half of Michi-gan's lower peninsula. Sample sources were catego-rized as municipal treatment plant taps, distributionsystems of municipal supplies, private water supplies(usually single-family dwellings), and public swimmingpools. Public water supplies and swimming pools wereroutinely sampled as part of periodic monitoring.Private water supplies were sampled on request, andthe group sampled may have contained more than theaverage number of suspect installations. Reportedstudies are described as follows.

(i) The occurrence of overgrown cultures were fol-lowed for summer (1 August to 31 August 1975) andwinter (6 November to 10 December 1975) periods.Positive and negative overgrowth verification resultswere also recorded. This study involved a total of6,215 samples from all sources.

(ii) Statistics for parallel verification results werecompiled for April to July 1980. Total specimens fromall sources for this period were 15,654, with 2,114verification tests done. LT and BGB broth fermenta-tion results were recorded for the initial parallel inocu-lations. False-negative results, as would be interpretedfor direct BGB inoculation only, were recorded forinitial LT(+) and BGB(-) results, where subsequentLT to BGB transfer gave a second BGB(+) result.False negatives, as for direct LT inoculation only,were recorded for initial LT(-) and BGB(+) results.

(iii) Statistics for municipal distribution, privatesupply, and swimming pool samples were kept forcalendar year 1976. For each sample in these catego-ries a record was retained of the verified coliformresult, whether or not overgrowth occurred, andwhether the sample holding time was 0, 1, 2, or >2days from date of collection to date of testing. Thisstudy included about 26,000 samples with a minimumof 359 samples in each holding time category.

Population changes for coliform strains. Variousstrains of organisms giving positive BGB coliformverification results were inoculated into large volumesof phosphate-buffered dilution water used for MFrinsing. These were then apportioned to 120-mI sterilecontainers stored together at room temperature, which

were individually tested by routine MF procedures at

different times. Populations immediately after inocula-tion ranged from several hundred to about 2,000organisms per 100 ml. The timing of populationchanges was begun for all strains when periodic countswere first found to be less than 100/100 ml. Tennaturally occurring strains were tested in this manner.

These included four E. coli and two Enterobacteraerogenes strains as indicated by the Standard Meth-ods for the Examination of Water and Wastewater (1)indole, methyl red, Voges-Proskauer, citrate testing.These six strains produced normal sheen upon MFculturing. Four atypical strains which did not produceMF sheen but gave positive BGB verification testswere also tested.

RESULTS

Comparisons of MF and MPN testing resultsare shown in Table 1. The methods were as-

sumed to be in agreement when the MF countfell within the span covered by the MPN index(i.e., from the midpoints between the next high-est and next lowest MPN index from the mea-sured index). Only a few samples gave MFcounts within the quantitation range covered bythe standard five-tube index, 2.2 to 16/100 ml.Samples were considered to be out of agreementwhenever one method gave a positive result andthe other gave a negative result, regardless ofthe counts. Where samples did not agree, a

preferred method was selected as the one whichgave the best coliform recovery. The compari-son of methods for low levels of E. coli inbuffered dilution water is shown in Table 2. Thiscomparison shows consistently higher values bythe MPN method, although exactly the samenumber of positives was obtained by eithermethod. This difference may reflect the differ-ence in the nature of the values (i.e., directcount versus MPN statistical estimate) deter-mined by the two methods. MPN procedureshave a statistical bias to the high side and showconsiderably less precision than direct platecounts (5, 6).

Results of studies of medium preparation andstorage are shown in Table 3 and Fig. 1. Theeffect on sheen production of different heating

TABLE 2. Comparison of routine MF and MPNtesting techniques for cultures inoculated with E. coli

MF count % of MF MPN % of MPNper 100 ml samplesa index persamplesm

0-1 8.0 <2.2 8.02-3 19.0 2.2 23.54-5 20.5 5.1 23.06-7 17.5 9.2 14.08-10 35.5 .16.0 31.5

a A total of 200 samples was tested for each tech-nique.

VOL. 44, 1982

on August 12, 2020 by guest

http://aem.asm

.org/D

ownloaded from

456 HSU AND WILLIAMS

TABLE 3. Times required for change in MF culturesheen formation for two methods of heating during

rehydration of m-Endo agar LES mediumTime (min) to first Time (min) to sheen

changea IossbStrain

Direct Boiling Direct Boilingheat water heat water

A 20 50 20 50B 20 20 50 >60C 50 >60 >60 >60D 40 40 50 >60

a First time that a reduction in the amount of sheenproduced was noted.

b First time that the complete absence of sheenformation was observed.

procedures during m-Endo agar LES rehydra-tion is shown in Table 3. Light exposure studiesshowed an adverse effect of light on m-Endoagar LES medium sheen production. A reduc-tion in the amount of sheen produced was firstobserved at 4 h of exposure time. In some cases,sheen was no longer produced after a 16-hexposure of medium to ambient lighting. Storageof m-Endo agar LES plates in plastic bags underrefrigeration had very little effect on colonysheen formation. Some diminished sheen reduc-tion seemed apparent after 5 weeks of storage,but no differences could be discerned for mediastored for 0 to 4 weeks. Rates of weight loss ordehydration of verification broth cultures underdifferent storage conditions are illustrated in Fig.1.The occurrence of MF culture overgrowth is

shown in Table 4. The percentages of totalvalues show the relative proportions of cultureswith and without overgrowth for each sampletype. The percentages of positive values are thepercentages of cultures giving verified coliformrecovery within each of the 10 sample catego-ries.

Statistics for verification studies are shown inTable 5. The first two columns of numericalvalues result from isolated-colony verification,while the values for overgrown cultures arebased on overgrowth verification procedures.Statistics regarding LT and BGB agreement arebased only on the initial verification inoculationresults. False-negative statistics assume verifi-cation procedures which involve initial inocula-tion to only one of the media (i.e., either LT orBGB).Table 6 shows the holding times for various

types of samples. The calculated percentagevalues shown in Fig. 2 and 3 are based on thetotal sample groupings for each category givenin Table 6. Sample holding times were deter-mined from dates of sample collection. Data forsample holding times for synthetically prepared

samples are shown in Table 7. Timing of countsis in multiples of 48 ± 3 h.

DISCUSSIONThese studies involve various aspects of the

MF technique for drinking water analysis. Thefollowing discussion is divided into major areasin which the studies provided pertinent informa-tion. Some unconventional terminology is usedwhich has specific meaning as defined in previ-ous sections. These terms include typical, atypi-cal and typical noncoliform MF colonies, MFovergrowth, and direct BGB verification. It willalso be noted that primary emphasis has beenplaced on positive versus negative results ratherthan on actual MF colony count values, becausegenerally agreed upon Michigan water supplyprotection policy calls for investigation and cor-rection of water supply deficiencies wheneverMF counts are greater than or equal to 1/100 ml.This policy also includes all state monitoring ofpublic water supplies required by state andfederal law.MF technique preference. The standard MPN

test with BGB confirmation for the coliformindicator group has a long history of providingincreased public health protection. For this rea-son, introduction of the MF technique was usu-ally done by demonstrating equivalence betweenMF techniques and standard MPN methods.The MF technique is more precise as a direct-count method and more sensitive in being able toanalyze larger sample volumes and gives resultsmore rapidly. However, it does not necessarilydefine the same total group of organisms includ-

Storage Time (Days)FIG. 1. Average percent weight loss curves for LT

and BGB broth fermentation culture tubes (18 by 150mm) containing an initial average 9.35 ml of mediumstored at room temperature. Curves include culturetubes with (A) metal closures only, (B) plastic closuresonly, and (C) plastic closures with batches of tubessealed in plastic bags.

APPL. ENVIRON. MICROBIOL.

on August 12, 2020 by guest

http://aem.asm

.org/D

ownloaded from

MF TECHNIQUE FOR TESTING DRINKING WATER 457

TABLE 4. Relative occurrence of overgrown MF cultures

Sample type Total no. of Cultures not overgrown' Overgrown cultures"samples % of total % Positive % of total % Positive'

Municipal treatment 812 94.3 (766) 3.0 (23) 5.7 (46) 15.2 (7)plants

Municipal water 1,105 90.9 (1,005) 3.7 (37) 9.1 (100) 13.9 (14)distribution systems

Private water supplies 3,553 84.5 (3,002) 12.0 (360) 15.5 (551) 33.9 (187)Swimming pools 745 81.6 (608) 3.9 (24) 18.4 (137) 10.2 (14)Overall 6,215 86.6 (5,381) 8.3 (444) 13.4 (834) 26.6 (222)

a Number of samples is given in parenthesis.b Swabs of overgrown cultures producing positive verification results when used to inoculate BGB broth

fermentation tubes.

ed under MPN testing, and any MPN preenrich-ment benefits in recovery of attenuated orga-

nisms are lost.Because of fundamental differences between

direct MF counts and MPN statistical estimates,comparison of the two methods for coliformrecovery is not straightforward. Comparison oftest values for synthetic E. coli samples (Table2) illustrates the problem. The MF and MPNvalues appear to be similar and certainly agree

with respect to the number of positive andnegative results. However, 92% of the MFcounts are included in a fairly precise range of 2to 10/100 ml, while the range of MPN indices isnot defined, with more that 30% of the valuesbeing >16/100 ml. It is not known whether this isbecause the MPN procedure is more sensitive or

simply because of the statistical bias and lack ofprecision inherent in the MPN method.The comparison of the two methods for rou-

tine sample analysis (Table 1) was done with an

arbitrary definition of agreement. The limitsused are probably well within the combineddeviation for the two methods, but, even so,substantial agreement was observed (overall,93.8% of the samples tested). According to givencriteria regarding recovery, where the methodsdisagreed, the MF method was preferred in five

of six such instances. These results were inexcellent agreement among different analystsand with those of McCarthy et al. (4) who usedm-Endo agar LES and a preenrichment step.Using the same agreement and preference crite-ria, their work on 654 samples gave 607 (92.8%)in agreement, 4 (0.6%) MPN preferred, and 43(6.6%) MF preferred. Because of the agreementbetween the two studies, we concluded that theenrichment step was not worthwhile. This con-clusion has been reinforced by recent work ofEvans et al. (2) who also used m-Endo agar LESfor initial MF culturing and found little benefitfrom preenrichment culturing.MF culture preparations. In preparing m-Endo

agar LES plates, it was found that the medium isindeed heat and light sensitive. Both heat andlight cause rapid darkening of the medium; how-ever, the immediate effect on MF colony sheenproduction was not substantial. Based on report-ed studies (Table 3 and other observations),heating with boiling water during medium rehy-dration appears to be an advantage, and suchheating should not exceed 15 min. Also, expo-sure of prepared media to laboratory lightingshould not exceed 4 h. Both limits can be readilymet in standard preparation procedures, and anyexcess heat and light should be avoided.

TABLE 5. Statistics for parallel routine MF verification procedures using LT and BGB broth fermentationNo. of:

Verification study Typical Atypical Overgrowncolonies colonies cultures

LT and BGB agreeing' 236 810 1,996LT(+) and BGB(-)' 1 23 1LT(-) and BGB(+)' 0 19 147False negatives by direct 0 8 0BGB inoculationb

False negatives by initial 0 19 147LT inoculation onlyb

a From results of initial parallel inoculation to both LT and BGB verification media.b False negatives as compared to initial parallel inoculation to both LT and BGB media with subsequent

transfer for LT(+) and BGB(-) results from LT to BGB for secondary confirmation.

VOL. 44, 1982

on August 12, 2020 by guest

http://aem.asm

.org/D

ownloaded from

458 HSU AND WILLIAMS

TABLE 6. Number of samples tested by MichiganDepartment of Public Health Lansing Laboratory at

various sample holding times

No. of samples from the followingHolding time sources:'

(days) Municipal Private Swimmingdistribution supplies pools

0 377 658 3591 3,511 7,385 1,5682 1,117 4,189 685

>2 1,268 2,752 333Unknownb 121 1,575 272

a Data include all samples for 1976 collected frommunicipal water supply distribution systems, privatewater supplies, and public swimming pools. The totalnumbers of samples from these sources were 6,394,16,559, and 3,217, respectively.

b Collection date not given in information submittedwith sample.

Some economy and certainly convenience canbe achieved by preparation of large batches ofculturing units. This in turn requires adequatelong-term storage. Although 1-week outdating ofa medium used in MF testing is recommended(7), our medium storage study observations sug-gest this to be overly cautious. Very little changein sheen production for natural coliform strainswas noted for m-Endo agar LES medium refrig-erated in sealed plastic bags for up to 5 weeks.BGB and LT verification media may also beused for longer than 1 week if plastic closuresare used to limit medium dehydration. Based ondehydration rates shown in Fig. 1 and an al-lowed 2% volume error in medium preparation,broth medium fermentation tubes with plastic

40-A B C

0

20

0

0 1 2 >2 0 1 2 >2 0 1 2 >2Sample Holding Time (Days)

FIG. 2. Percentage of test cultures showing over-

growth for different sample holding times betweensample collection and testing. Zero days correspondsto same-day sample collection and testing. Differentsample sources include (A) municipal water supplydistribution systems, (B) private water supplies, and(C) public swimming pools. Each percentage calcula-tion includes 330 to 7,380 total specimens.

01

0

00 1 2 '3 0 1 2 3 0 1 2 '3

Sample Holding Time (Days)

FIG. 3. Percentage of samples giving positive rou-tine coliform verification tests for different sampleholding times between sample collection and testing.Zero days corresponds to same-day sample collectionand testing. Darkened areas represent positive MFcultures not showing overgrowth, while crosshatchedareas represent additional coliform recovery fromovergrown MF cultures. Different sample sourcesinclude (A) municipal water supply distribution sys-tems, (B) private water supplies, and (C) public swim-ming pools. Each percentage calculation includes 330to 7,380 total specimens.

closures may be stored at room temperature forat least 2 weeks. Sealing of tubes in plastic bagsextends storage time to about 3 weeks under theconservative 2% volume change criteria.MF isolated-colony evaluation. For reasons

previously discussed, it is important that coli-forms recovered by the MF technique includethose strains previously recovered by the tradi-tional MPN methods. These coliforms have tra-ditionally been defined by the completed MPNtest as described in Standard Methods for theExamination of Water and Wastewater (1). Iso-lated MF colonies giving positive BGB verifica-tion tests come closer to the completed MPNdefinition than do confirmed MPN results. Infact, if MF culturing provides true isolation,verification of colonies then fulfills all aspects ofcoliform definition except Gram staining. Forthis work, only when selected colonies gavepositive BGB verification tests were culturesreported as positive for coliforms. This conven-tion is also supported by average 10% false-positive verification statistics for typical sheenMF colonies.Given this verification requirement, the judge-

ment used in selecting colonies for verificationbecomes critical. As previously discussed, suchjudgement can come only with experience incolony verification. Records have been kept forrates of positive verification for classes of colo-ny growth previously described. Overall averagemonthly positive verification rates for the periodJanuary 1977 to October 1979 were 90% for

APPL. ENVIRON. MICROBIOL.

on August 12, 2020 by guest

http://aem.asm

.org/D

ownloaded from

MF TECHNIQUE FOR TESTING DRINKING WATER 459

TABLE 7. MF counts per 100 ml of sample at various sample holding times'

MF counts per 100 ml for:

Time E. coli strainsb E. aerogenes Atypical strainsd(days) strains"

1 2 3 4 1 2 1 2 3 4

0 186 79 183 105 142 89 49 142 142 1362 80 105 36 44 19 25 0 0 176 22 415 -6 4 10 0 08 1 0 91 78 14

10 012 19 7 133 202 30 18014 51 1 - 7116 - 13 4 9218 - 0 7 35

a MF coliform counts as a function of storage time at room temperature for isolated environmental strainsinoculated to buffered phosphate dilution water.

b Identified as E. coli strains by indole, methyl red, Voges-Proskauer, citrate testing.c Identified as E. aerogenes strains by indole, methyl red, Voges-Proskauer, citrate testing.d Atypical strains not producing typical MF sheen cultures but giving positive BGB broth verification test

results.e , Not done.

typical sheen colonies and 30% for atypicalnonsheen colonies. The range of monthly verifi-cation rates was 77 to 97% for sheen coloniesand 19 to 54% for atypical colonies. Table 5 dataindicate a proportion of 237 typical to 852 atypi-cal cultures for 15,654 specimens. Using thelower 19% rate for atypical colonies and theupper 97% rate for typical colonies, 162 positiveatypical cultures and 240 positive typical sheencultures would be predicted. Clearly, the num-ber of nonsheen atypical colonies conforming tothe MPN definition of coliforms is substantialrelative to the number of typical sheen coloniesobserved. If the MPN coliform definition is to beretained, MF colony interpretation should gobeyond simple sheen or no sheen considerationsin selecting colonies for verification.MF colony verification procedures. Routine

use of parallel inoculation of suspect MF growthto both LT and BGB broth fermentation culturesfor verification has indicated that changes inverification procedures are advisable. Datashown in Table 5 are typical of several years'experience. The need for an enrichment with LTmedium is questionable, and these data indicatevery little benefit from this step. Total agree-ment was found for typical sheen colonies forLT enrichment and direct BGB verification(eliminating initial LT enrichment). For atypicalcolonies, eight specimens (<1%) gave positiveLT results which were confirmed with second-ary transfer to BGB medium, while direct BGBverification was negative. On the other hand, 19specimens gave positive direct BGB verificationbut would have been reported as negative with apreliminary LT enrichment step as is currently

recommended in federally approved methodolo-gy. If a single medium for verification were to beselected, BGB would be preferred as recoveringthe greatest number of coliforms. With respectto additional cost, the benefit of preliminary LTenrichment is negligible. The advantage in re-covery is less than that of direct BGB verifica-tion, and what advantage there is amounts toonly 0.7% of samples verified. Similar conclu-sions were reached by Evans et al. (2) usingdirect verification inoculation to another selec-tive medium, m-Endo-based (m-LAC) broth.However, direct BGB verification perhaps cor-responds more closely to past definition of thecoliform group, and this approach uses an addi-tional gram-negative selective agent not em-ployed in the MF culturing.MF overgrowth. The occurrence of MF cul-

ture overgrowth during routine testing hasproved to be a substantial problem ranging fromabout 5 to 20% of specimens submitted fromvarious sample sources (Table 4). Recovery ofcoliforms [BGB(+) by definition] from over-growth is best done by direct inoculation ofovergrowth to BGB medium. As indicated inTable 5, direct BGB verification, relative to aninitial LT enrichment step, increases recoveryfrom overgrown cultures by 7% of total speci-mens.The meaning of overgrowth, where direct

BGB verification is negative, has not been de-fined. A number of factors may result in re-growth of organisms in a water system or in asample container which could cause MF over-growth but would not necessarily have sanitarysignificance. Bacterial populations of <5/ml

VOL. 44, 1982

on August 12, 2020 by guest

http://aem.asm

.org/D

ownloaded from

460 HSU AND WILLIAMS

could result in overgrowth if a large proportionof the population were capable of growth on MFmedia. Taking these factors into account, it stillseems advisable to consider overgrowth as anindicator of potential water supply problemsbased on the following considerations. (i) Someovergrown organisms (particularly pseudomo-nads) may be antagonistic to coliform growth.(ii) A comparison study of standard plate countswith overgrowth occurrence for 500 routine wa-ter supply samples showed 75% of overgrownMF cultures also gave plate counts of >2,000/ml(unpublished work). (iii) As indicated by directBGB verification, coliform recovery from over-grown cultures is about three times more fre-quent than for cultures without overgrowth (Ta-ble 4).Sample holding time effects. Although there

are no known correlations between holdingtimes and water supply construction or opera-tional aspects which could affect testing results,relatively large numbers of samples (Table 6)were included in sample holding time statisticsin hopes that the size of the sampling would tendto average out any such correlation within eachholding time group. Nearly all samples includedin this data were transported to the laboratorywithout special preservation or temperaturecontrol.

Figure 2 illustrates some definite trends inovergrowth occurrence with respect to increas-ing sample holding time. There were consistentsubstantial increases in percent overgrowth oc-currence for each longer holding time. Thistrend was similar for each type of sample sourceshown, and very few exceptions to this type oftrend were found in individual monthly statisticsfor the 12 months included in the total group.Overall, the percentage of drinking water speci-mens showing overgrowth increased an averageof about 5% of total samples for each 24 h ofsample holding time. A more rapid increase isapparent for chlorine-neutralized swimmingpool samples. Obviously, any interpretation ofthe sanitary significance of an overgrowth testresult must also consider the sample holdingtime.MF counts are expected to change with in-

creasing sample holding times, and federalguidelines (7) recommend that holding times belimited to less than 30 h. In fact, for any individ-ual sample, counts may be continually increas-ing or decreasing at rates dependent on a com-

plex combination of factors includingtemperature, pH, chemical content, and coli-form strains present. Although it indicates thattesting should be done as soon as possible, thisexpected behavior does not provide a basis forany specific sample holding time limit.

Preliminary work on coliform population de-creases in minimal nutrient media at near neutralpH indicates that the time to complete die-off(i.e., MF count of <1/100 ml) for strains investi-gated is more a matter of days than hours.Changes in counts with time shown in Table 7illustrate this point and suggest that cyclical die-off and regrowth patterns may occur over peri-ods of several days for some members of thecoliform group. The percentage of positive testresults for routine testing did not exhibit regularincreases or decreases with increasing sampleholding times. Figure 3 shows a lack of anyconsistent trends for up to 72 h of sampleholding time. The overall percentages of positivesamples for all holding times remain relativelyconstant for each sample source included: mu-nicipal distribution, 4.7 + 2.0%; private watersupplies, 15.9 + 1.8%; and swimming pools, 4.5+ 0.6%. For monitoring programs where investi-gation and corrective action are prompted byany positive MF count per 100 ml of sample,these data do not support any specific limit. Forthis type of response to any positive test results,the preceding data imply that the coliform indi-cator system is generally effective to at least 72 hof sample holding time.

LITERATURE CITED

1. American Public Health Association. 1975. Standard meth-ods for the examination of water and wastewater. Ameri-can Public Health Association, Inc., New York.

2. Evans, T. M., R. J. Seidler, and M. W. LeChevallier. 1981.Impact of verification media and resuscitation on accuracyof the membrane filter total coliform enumeration tech-nique. Appl. Environ. Microbiol. 41:1144-1151.

3. McCarthy, J. A., and J. E. Delaney. 1958. Membrane filterstudies. Water Sewage Works 105:292-296.

4. McCarthy, J. A., J. E. Delaney, and R. J. Grasso. 1961.Measuring coliforms in water. Water Sewage Works108:238-243.

5. McCarthy, J. A., H. A. Thomas, Jr., and J. E. Delaney.1958. Evaluation of the reliability of coliform density tests.Am. J. Public Health 48:1628-1635.

6. Thomas, H. A., Jr. 1955. Statistical analysis of coliformdata. Sewage Ind. Wastes 27:212-222.

7. U.S. Environmental Protection Agency. 1977. Manual forthe interim certification of laboratories involved in analyz-ing public drinking water supplies. Environmental Protec-tion Agency publication no. 600/8-78-008. Office of Moni-toring and Technical Support, U.S. EnvironmentalProtection Agency, Washington, D.C.

APPL. ENVIRON. MICROBIOL.

on August 12, 2020 by guest

http://aem.asm

.org/D

ownloaded from