Vol. 57, No. 7APPLIED AND ENVIRONMENTAL MICROBIOLOGY, JUlY 1991, P. 1905-19130099-2240/91/071905-09$02.00/0Copyright © 1991, American Society for Microbiology

Use of a xylE Marker Gene To Monitor Survival of RecombinantPseudomonas putida Populations in Lake Water by Culture

on Nonselective MediaCRAIG WINSTANLEY,l* J. ALUN W. MORGAN,2 ROGER W. PICKUP,2 AND JON R. SAUNDERS1

Department of Genetics and Microbiology, University of Liverpool, Liverpool L69 3BX,l and Windermere Laboratories,Institute ofFreshwater Ecology, Ambleside, Cumbria LA22 OLP,2 United Kingdom

Received 26 November 1990/Accepted 20 March 1991

IncQ marker plasmids were previously constructed to enable the analysis of the survival of populations ofPseudomonas putida released into lake water (C. Winstanley, J. A. W. Morgan, R. W. Pickup, J. G. Jones, andJ. R. Saunders, Appl. Environ. Microbiol. 55:771-777, 1989). We constructed equivalent IncP plasmids,pLV1016 and pLV1017, to provide conjugative alternative systems. Detection of the xylE gene carried bymarker plasmids was found to be a valid indicator to use for studying the survival of released populations byculturing on nonselective media. These plasmids were used to study the survival of populations of Pseudomonasputida in both sterile and untreated lake water. The effects of inoculum size, the metabolic burden imposed on

the cell by the unregulated expression of xylE, and an auxotrophic mutation carried by the host strain were

studied. We also assessed the reproducibility and hence the predictability of the survival of releasedpopulations. Model systems with a single lake water sample and model systems with three different lake watersamples, taken from the same site in consecutive months, were used to analyze variability between replicatesand to assess differences caused by host strain or water sample. A large variability was found depending on

which water sample was used. These findings imply that it will be difficult to predict accurately the survival ofreleased populations in the natural environment.

Before predictions can be made concerning the survival ofgenetically engineered microorganisms released into thenatural environment, there is a need to carry out studies withlaboratory model systems to assess both the potential of theorganisms for carrying out a required task and the safetyimplications of the proposed release. Since the accidental ordeliberate release of genetically engineered microorganismsis likely to involve runoff into freshwater habitats, a consid-eration of survival in such environments is important.To assess the survival of genetically engineered microor-

ganisms, methods with which to detect the released strainsand distinguish them from indigenous bacteria are needed.We previously constructed a marker plasmid system de-signed to study the fate of genetically marked bacteria inmodel environments (26). This system has been coupled tovarious methods which enable the direct detection of re-leased strains in filtered water samples (17). The systemdepends on the detection of a marker gene, xylE, and itsproduct (catechol-2,3-dioxygenase; C230). On the markerplasmids, xylE is expressed from the bacteriophage lambdaPL promoter or from the lambda PR promoter under thecontrol of the temperature-sensitive lambda repressor cI857.Marker cassettes have been introduced into the broad-host-range, nonconjugative IncQ plasmid pKT230 (26). Tobroaden the potential of our system, the constituent partshave now also been introduced into the broad-host-range,conjugative IncP plasmid R68.45. The result is a markersystem that can be used to monitor both survival andplasmid transfer in model systems without selection forantibiotic resistance.

* Corresponding author.

The factors that affect the survival, persistence, andgrowth of bacteria in the aquatic environment are poorlyunderstood. Pseudomonas putida and other pseudomonadsare part of the indigenous microflora of most lake waters.Because of their wide range of novel catabolic activities,pseudomonads also have potential applications in the detox-ification of polluted soil or fresh water. The persistence ofthese microorganisms and the effects of their reintroductioninto the environment have not been fully examined. Usingsimple culture techniques, we have carried out a series ofsurvival studies with P. putida strains carrying IncQ andIncP marker plasmids. Studies with sterile lake water modelsystems have enabled an assessment of the stability of themarker system during growth. By releasing recombinantpopulations at different cell densities, we have been able tostudy the effect of inoculum size on survival. We have alsostudied the effects on survival of the unregulated expressionof xylE and of a host strain which carries an auxotrophicmutation. Our data indicate that xylE marker systems can beused to monitor strain survival in untreated lake water byculturing on nonselective media. A comparison of the sur-vival of populations of P. putida in both microcosms fromthe same lake water sample and microcosms from differentlake water samples taken at the same site in consecutivemonths has been carried out to assess the reproducibility andhence the predictability of survival patterns.

MATERIALS AND METHODS

Bacterial strains and plasmids. The release strains used inthis work were derived from the TOL plasmid-containingstrain P. putida PaWl. PaW8 is a plasmid-free derivative(27), and PaW340 is a plasmid-free, tryptophan-requiring,auxotrophic derivative (8). Escherichia coli AB1157 and

1905

on April 3, 2019 by guest

http://aem.asm

.org/D

ownloaded from

1906 WINSTANLEY ET AL.

AB2463 were obtained from P. Strike, University of Liver-pool, Liverpool, United Kingdom, and P. aeruginosaPA025(R68.45) was obtained from the Plasmid ReferenceCentre, Colindale, London, United Kingdom. PlasmidspLV1010 (Apr Smr PL xylE), pLV1012 (Kmr PR xylE c1857),pLV1013 (Smr Kmr PR xylE c1857), and pFBA10-1 (Apr PLxylE) have been described previously (26) (Apr, Smr, andKmr indicate resistance to ampicillin, streptomycin, andkanamycin, respectively). The construction of pLV1016 andpLV1017 is described in this work.Growth and maintenance of strains. The strains were

maintained by growth at 30°C on nutrient agar or in nutrientbroth (lab/m, Bury, Lancashire, United Kingdom) supple-mented with antibiotics (each at 50 jig/ml) when necessary(streptomycin and ampicillin for pLV1010; streptomycin andkanamycin for pLV1013; and tetracycline, ampicillin, andkanamycin for pLV1016 and pLV1017).

Detection of xylE-expressing colonies. Colonies expressingxylE (C230 positive [C230+]) were identified by sprayingplates with a 1% (wt/vol) solution of catechol. A positivereaction was indicated by the production of a yellow color-ation due to the formation of 2-hydroxymuconic semialde-hyde. Strains which contained the regulated plasmidspLV1012, pLV1013, and pLV1016 were incubated at 42°Cfor 2 h prior to being sprayed.

Plasmid DNA preparation. Bulk plasmid DNA was pre-pared by a method based on that described by Guerry et al.(10). Small-scale preparation of plasmid DNA was done bythe rapid screening method described by Close and Rod-riguez (4).

Restriction digestions and agarose gel electrophoresis. Re-striction endonucleases were obtained from Boehringer Cor-poration Ltd. (Lewes, Sussex, United Kingdom), and diges-tions were performed with O'Farrell buffer (18) under theconditions recommended by the supplier. Agarose gel elec-trophoresis was carried out by standard procedures.

Preparation of cell extracts and C230 assays. Cell extractswere prepared and C230 assays were carried out as de-scribed previously (26).Lake water samples and release systems. Water samples

were collected from the surface water of Lake Windermere,Cumbria, United Kingdom. Sterile lake water was preparedby filtratioh through 0.22-,um-pore-size membranes(GSWPO4700; Millipore, Bedford, United Kingdom) andautoclaving at 121°C for 30 min. Lake water (untreated) andsterile lake water were kept at 10°C for the duration of therelease experiments.

Release strains were grown for 18 h with shaking at 30°Cin 250-ml flasks each containing 50 ml of nutrient broth. Forplasmid-containing strains, relevant antibiotics were in-cluded in the medium. The cells were harvested by centrif-ugation at 5,000 x g for 10 min and washed three times insterile distilled water (10 ml). Cell concentrations wereestimated turbidimetrically at 600 nm, and release systemswere inocula'ted with approximately 102, 103, 104, or 105CFU of lake water ml-1. Lake water model systems rou-tinely consisted of 250-ml water samples in presterilized300-ml conical flasks. Experiments involving pLV1O13 werecarried out in 250-ml bottles containing 100-ml water sam-ples. Tryptophan was added at 0.05% when appropriate.Subsamples (0.5 ml) were serially diluted in sterile distilledwater (4.5 ml). Dilutions (100 ,ul) were plated in triplicate onnutrient agar plates, which were subsequently incubated for72 h at 20°C. Normally, plates with 30 to 300 colonies weresprayed with 1% (wt/vol) catechol and counted.

Analysis of triplicate releases with a single water sample.

The five release strains were Paw8(pLV1016) (strain 1),Paw340(pLV1016) (strain 2), Paw8(pLV1017) (strain 3),Paw340(pLV1O17) (strain 4), and Paw340(pLV1017) supple-mented with tryptophan (strain 5). Four target inocula wereused: inoculum 1 (105 CFU ml-1), inoculum 2 (104 CFUml-'), inoculum 3 (103 CFU ml-1), and inoculum 4 (102 CFUml-'). Model system experiments were performed in tripli-cate for inoculum 1. At inoculum 1, the model systems werenumbered as follows: for strain 1, Ml to M3; for strain 2, M4to M6; for strain 3, M7 to M9; for strain 4, M10 to M12; andfor strain 5, M13 to M15. At inocula 2, 3, and 4, the modelsystems for strains 1 to 5 were numbered M16 to M20, M21to M25, and M26 to M30, respectively.Each model system was monitored for 28 days with

samples taken on days 0, 4, 8, 14, 22, and 28. On day 0,dilutions of Ml, M4, M7, M10, and M13 were used toconstruct M16 to M30 and to estimate the cell density in theoriginal model systems. Thus, the estimated numbers (N) ofCFU per milliliter obtained from certain model systems arelinked on day 0 (N16 = N1110, N21 = N1/100, and N26 =N111,000). The same sequence is true for M4, M7, M10, andM13 at inocula 2 to 4. Counts were always performed onthree nutrient agar plates at recorded dilutions. Thus, thedata comprised three counts at a specified dilution for eachmodel system on every sampling day.

Analysis of triplicate releases with water samples taken fromthe same site in three consecutive months. The three releasestrains were PaW8(pLV1013) (strain 1), PaW340(pLV1O13)(strain 2), and PaW340(pLV1013) supplemented with tryp-tophan (strain 3). For each of the three strains, releases werecarried out with water sampled in each of the 3 months.Three replicate model systems were inoculated for each ofthe strains in each of the months. Dilutions were plated outfrom each replicate model system on days 3, 6, 10, 14, 20,and 28 after the release. Counts were always performed onthree nutrient agar plates for each model system on eachsampling day. In each month, inoculation tubes of strain 1and strain 2 were made up, and from each of these, threereplicate model systems were inoculated on day 0. Addi-tional replicate model systems were inoculated and supple-mented with tryptophan for strain 3 model systems. Theintended inoculum was 105 CFU ml-1. The inoculum in eachcase was sampled, and an average of triplicate plate countswas used to give an indication of the size of the populationactually released. The initial inocula were all within therange of 3.2 x 104 to 6.2 x 105 CFU ml-1. Dilution plateswere taken from the inoculation tubes and from the threereplicate tubes on day 0 for October and November samples.

Direct detection methods. The direct detection, on filteredwater samples, of C230 activity by the C230 assay or anenzyme-linked immunosorbent assay (ELISA) and the de-tection of xylE by hybridization have been described previ-ously (17).AO direct counts. Acridine orange (AO) staining of cells

was performed essentially by the method of Jones and Simon(12). To 10-ml lake water subsamples, AO was added to afinal concentration of 5 ,ug ml-' and left for 5 min. Sampleswere filtered through 0.22-,um-pore-size black polycarbonatefilters (Nuclepore; Costar UK Ltd., High Wycombe, Buck-inghamshire, United Kingdom) and washed with 10 ml ofdistilled water. Wet filters were carefully mounted undercoverslips with immersion oil and observed at a magnifica-tion of x 100.DVC. Direct viable counting (DVC) was performed by a

modification of the method of Kogure et al. (13). To 10-mllake water samples, yeast extract was added to 0.025%

APPL. ENVIRON. MICROBIOL.

on April 3, 2019 by guest

http://aem.asm

.org/D

ownloaded from

SURVIVAL OF P. PUTIDA POPULATIONS IN LAKE WATER 1907

FIG. 1. Restriction maps of pLV1016 and pLV1017. pLV1016 is a cointegrate of plasmids pLV1012 and R68.45. pLV1017 is a cointegrateof plasmids pFBA10-1 and R68.45.

(wt/vol) and tryptophan was added to 0.002% (wt/vol).Nalidixic acid was added to replicate samples at 0.005,0.001, and 0.0005% (wt/vol). Following incubation at 20°Cfor 18 h, the samples were filtered (0.22-,um-pore-size blackNuclepore filters). The filters were stained in the filter unitwith AO (0.01% [wt/vol]) for 3 min, washed with 10 ml ofdistilled water, and mounted under coverslips for fluores-

cence microscopy as described above. By comparing thesizes of cells treated in this way with those of cells directlystained with AO, we determined the number that had in-creased in size (DVC).

RESULTSConstruction of conjugative marker plasmids. R68.45,

which contains a tandem duplication of insertion sequence

IS21, was originally isolated on the basis of its enhancedability to promote chromosomal transfer (11). R68.45 hasbeen shown to mobilize E. coli plasmids to recA+ and recAE. coli recipients (20, 21). The molecular basis for suchtransfer involves the formation of a cointegrate intermediateduring inverse transposition of IS21 (25). Resolution of thecointegrates in recA+ recipients occurs at a low frequency,with R68.45 sometimes losing one copy of IS21 and becom-ing indistinguishable from plasmid R68 (alias RP1, RP4, andRK2) (21).R68.45 (Kmr Apr Tcr) was transferred by conjugation from

P. aeruginosa PA025 to E. coli ED8654 harboring plasmidpLV1012 or pFBA10-1 (both C230+). Selection of E. colicontaining both R68.45 and pLV1012 or R68.45 andpFBA10-1 was done by spreading each mating mixture on

nutrient agar containing kanamycin, ampicillin, and tetracy-cline and incubating the agar at 42°C for 12 h. Coloniesshowing C230 activity were identified by spraying with 1%catechol solution and checked for the presence of bothplasmids by rapid plasmid minipreparation testing. E. colicontaining plasmids R68.45 and pLV1012 or R68.45 andpFBA10-1 was crossed into P. putida PaW340. Selection forKmr, Apr, and Tcr (antibiotic concentration, 100 p,g ml-')was carried out at 28°C on succinate minimal mediumcontaining tryptophan. Since plasmids pLV1012 andpFBA10-1 are ColEl replicons, they are unable to replicatein a Pseudomonas background unless integrated into R68.45.

Ten colonies from each cross were selected at random andsubjected to plasmid minipreparation testing. In no case wasthe presence of a small, individually replicating vectorplasmid apparent. Bulk plasmid preparations were made fortwo subcolonies of each group of 10 colonies. The plasmidcointegrates for each cross were found to have identicalrestriction maps (Fig. 1).

Plasmid pLV1016 contains pLV1012 integrated intoR68.45 at the position of the IS21 duplication. In pLV1016,the insert is flanked by one copy of IS21 (Fig. 1), consistentwith cointegrates described previously with E. coli recipi-ents (21). The point of insertion of pLV1012 appears to haveoccurred in or close to its origin of replication (oriV).Plasmid pLV1017 is a cointegrate of pFBA10-1 and R68.45 inwhich the insert is also flanked by one copy of IS21 and thepoint of insertion of pFBA10-1 is again in or close to its oriV(Fig. 1). Deletion of part of pLV1012, pFBA10-1, or R68.45could not be detected, and the cointegrates were stable at28°C for over 60 generations under nonselective conditions(>99%).

Transfer and stability of pLV1016 and pLV1017. BothpLV1016 and pLV1017 were transferred from P. putidaPaW340 (tryptophan auxotroph) by selection on minimal

TABLE 1. C230 activities of various hosts containing pLV1017C230 sp act

Host strain (U/mg ofprotein)

Escherichia coli AB1157 .......................................... 25.4Escherichia coli AB2463 .......................................... 24.2Pseudomonas putida PaW8 ...................................... 20.9Pseudomonas putida PaW340 ................................... 20.0Pseudomonas putida PRS2000 .................................. 20.8Pseudomonas putida FBA11 ..................................... 15.3Pseudomonas aeruginosa PAO1 ................................ 20.7Pseudomonas fluorescens FH1 .................................. 22.6Aeromonas hydrophila NCTC 8049 ............................ 22.5Klebsiella pneumoniae NC18418 ................................ 13.2Klebsiella pneumoniae K2819 ................................... 13.1Serratia rubidaea ............................................ 16.5Acinetobacter calcoaceticus ADP1 ............................ 9.1

VOL. 57, 1991

on April 3, 2019 by guest

http://aem.asm

.org/D

ownloaded from

1908 WINSTANLEY ET AL.

media containing kanamycin, ampicillin, and tetracycline at100 ,ug ml-' (or growth at 42°C for E. coli strains) to a rangeof gram-negative bacteria (listed in Table 1). Greater than99% stability of transfer of antibiotic resistance and C230activity occurred in all strains. The only exception wasKlebsiella pneumoniae, which showed 90% stability. Bothplasmids were stably maintained during nonselective growthin all of the host strains, with the exception of K. pneumo-niae, in which a degree of instability was apparent (85%)stable inheritance in 20 generations). No difference in thestability of either cointegrate in recA and recA+ E. coli wasobserved. When stability tests were performed at 37°C orhigher, cointegrates were lost from the cells at a greaterfrequency. Loss of the C230+ phenotype was always asso-ciated with the loss of Kmr, Apr, and Tcr.

Expression of xylE from marker plasmids. C230 activitywas assayed for marker plasmids pLV1016 and pLV1017 invarious host bacteria (listed in Table 1), including recentfreshwater isolates. To assess the regulation of xylE expres-sion from pLV1016, we carried out time course studies ofC230 activity in nutrient broth containing kanamycin, ampi-cillin, and tetracycline after incubation at elevated tempera-tures. Since incubation at 42°C has an adverse effect on someof the host strains (26), a temperature of 37°C was used.Cultures were grown overnight at 28°C, after which thetemperature of incubation was raised and samples wereassayed after 0, 2, 4, 6, 9, and 12 h.Good temperature regulation of the pLV1016 system was

observed in all of the hosts. No C230 activity was observedin any plasmid-free hosts. Uninduced C230 activity in strainscontaining pLV1016 was relatively constant, with only a10-fold difference between E. coli and Pseudomonas strains.Increased uninduced C230 activity in Pseudomonas andAcinetobacter hosts relative to enteric bacteria was ob-served. This result was probably due to the Pseudomonaspromoter regions in front of xylE on this construct. Lowerlevels of C230 activity were observed in both uninduced andinduced cells containing pLV1016 than in those containingpLV1013, the equivalent marked plasmid based on pKT230(26). This result may have been due to the lower copynumber of R68.45 (2 to 3 copies per chromosome) than ofpKT230 (15 to 25 copies per chromosome).High levels of C230 activity were recorded in all hosts

containing pLV1017 (Table 1). When compared withpLV1010, the equivalent plasmid based on pKT230, theoverexpressing construct pLV1017 gave rise to less variationin C230 activity (26). This result may have been due to anincrease in the stability of pLV1017 in these hosts. The lowerlevels of C230 activity in strains containing pLV1017 than inthose containing pLV1010 reflect differences in plasmid copynumber.

Survival of P. putida populations in sterile lake water. (i) P.putida PaW8. The recovery on nutrient agar of a populationof P. putida PaW8(pLV1017) following release into sterilelake water can be seen in Fig. 2a. At initial inocula ofbetween 102 and 105 CFU ml-1, all populations grew to adensity of between 5 x 105 and 1 x 106 CFU ml-1. Afterrelease into sterile lake water at an initial inoculum of 2 x 105CFU ml-1, only a small proportion of the total populationlost the C230+ phenotype after 28 days (<20%). Populationsreleased at inocula below 105 CFU ml-' lost the C230+phenotype at an increased frequency. This result may indi-cate a growth advantage for cells that have lost pLV1017.The greater the number of cell cycles required for thepopulation to reach 105 CFU ml- 1, the higher the proportionof C230-negative (C230-) cells within the final population.

10~

lem"-10E voa

E,

III.(U.

IU.(U

10.

"04

102101

I

I

"4id

-"4

o 2 4 6 6 10 12 14 18 16 20 22 24 26 28linnw (dlays)FIG. 2. Survival of strains containing pLV1016 or pLV1017 in

sterile lake water. The graphs show the survival of PaW8(pLV1017)(a), PaW340(pLV1017) (b), PaW340(pLV1017) plus tryptophan (c),PaW8(pLV1016) (d), and PaW340(pLV1016) (e). The strains werereleased at four different inocula. Open symbols indicate the totalpopulation, and closed symbols indicate the numbers of C230+colonies.

b

APPL. ENVIRON. MICROBIOL.

I

on April 3, 2019 by guest

http://aem.asm

.org/D

ownloaded from

SURVIVAL OF P. PUTIDA POPULATIONS IN LAKE WATER 1909

The C230 activity recovered from populations on days 0and 28 revealed that lower levels of C230 activity weredetected in 28-day-old populations which had shown greaterplasmid instability (i.e., populations released at a smallerinitial inoculum; data not shown). Similar results were foundfor the detection of C230 by the ELISA and the directdetection of the xylE gene (data not shown). These resultsindicated that colony-forming ability and direct detectionmethods were in agreement. These results also suggestedthat the lambda promoter was expressed in released cellsprovided that the plasmid was maintained.The recovery of P. putida PaW8(pLV1016) on nutrient

agar after release into sterile lake water is illustrated in Fig.2d. The results are very similar to those for P. putidaPaW8(pLV1017), except that greater stability of pLV1016was observed in all populations. Direct detection of C230was not possible on day 0 or 28. This result indicated that thelambda promoter remained repressed during release in ster-ile lake water. Maintenance of the repression of the xylEgene was probably responsible for the increased stability ofpLV1016 over pLV1017 under these conditions.pLV1017 proved to be much more stable in sterile model

systems than pLV1010. The C230+ phenotype of PRS2000(pLV1010) was previously shown to be unstable when thestrain was released into sterile lake water (17). At an initialinoculum of 105 CFU ml-1, C230+ colonies were undetect-able after 22 days. A subpopulation containing plasmidswhich had deletions of xylE but which retained the antibioticresistance genes of pLV1010 (Apr Smr) appeared (17). Stud-ies of PaW8(pLV1010) released into sterile lake wateryielded very similar results. Initial inocula of 102 or 103 CFUml-' led to the disappearance of the C230+ phenotype within6 days. Initial inocula of 104 to 105 CFU ml-' resulted in theloss of the C230+ phenotype after 20 to 28 days, although asubpopulation which was C230- but still resistant to strep-tomycin and ampicillin was apparent. After an initial decline(down to 102 to 103 CFU ml-'), total populations grew backto approximately 105 CFU ml-1 after 28 days (data notshown).

Figure 3a illustrates the release of P. putida PaW8(pLV1013) into sterile lake water. After a drop in populationover the first 6 days, the released strain grew back to levelsapproaching 105 CFU ml-1' over the 28-day period.

(ii) P. putda PaW340. Populations of P. putida PaW340(pLV1016), PaW340(pLV1017), and PaW340(pLV1013),tryptophan-requiring auxotrophs, declined in numbers afterrelease into sterile lake water at all inocula (Fig. 2b and e,and 3b). No loss of the C230+ phenotype was detectable inthe cultured populations. Direct detection of C230 activity inPaW340(pLV1017) populations released at 105 CFU ml-1revealed a rapid decline to undetectable levels within 3 days.For the PaW340(pLV1017) populations, on day 28 the de-tection of C230 by the ELISA was possible in two of thethree releases, but direct detection of the xylE gene was notsuccessful (data not shown). AO direct counts indicated apopulation equivalent in size to the initial P. putidaPaW340(pLV1017) inoculum. The length of 28-day-old P.putida PaW340 cells (4.45,um; standard deviation, +44%; n= 30) was larger than the lengths of released cells (3.25 ,um;standard deviation, ±+14%; n = 35) or 28-day-old P. putidaPaW8(pLV1017) cells (0.75,um; standard deviation, ±+12%;n = 30). In addition, the 28-day-old P. putidaPaW340(pLV1017) population did not take up the AO stainas effectively as the 28-day-old P. putida PaW8(pLV1017)population. Unlike previously reported nonculturable butviable (NCBV) Vibrio cholerae cells (5), P. putida

EILU 103101

E o3

1 '

10

_11.1'

10'los,

o 2 4 6 8 10 12 14 16 18 20 22 24 26 28rTie (days)

FIG. 3. Survival of strains containing pLV1013 in sterile lakewater. The graphs show the survival of PaW8(pLV1013) (a),PaW340(pLV1013) (b), and PaW340(pLV1013) plus tryptophan (c).The strains were released at four different inocula. Open symbolsindicate the total population, and closed symbols indicate thenumbers of C230+ colonies.

PaW340(pLV1017) did not respond to yeast extract-nalidixicacid treatment (DVC). P. putida PaW8(pLV1017) respondedto such treatment throughout the 28-day study. It was

concluded that the P. putida PaW340(pLV1017) populationhad died, probably as a result of its requirement for tryp-tophan.When tryptophan was added to sterile lake water, P.

putida PaW340(pLV1017) and PaW340(pLV1013) behavedin the same way as P. putida PaW8(pLV1017) and PaW8(pLV1013), respectively (Fig. 2c and 3c). When P. putidaPaW340 inocula larger than 105 CFU ml-1 were released intosterile lake water, the populations were maintained at ap-proximately the introduced level throughout the 28-daystudy. NCBV V. cholerae cells were produced from an

initial inoculum of 107 CFU ml-' (5). This result representsan additional difference between NCBV cells and P. putidaPaW340 cells.The release of P. putida PaW340(pLV1016) and PaW340

(pLV1013) into sterile lake water produced results verysimilar to those produced by the release of P. putida PaW340(pLV1017), although the C230+ phenotype could still bedetected after 28 days in two of four PaW340(pLV1013)releases (Fig. 2e and 3b). The xylE gene was not detected inthe 28-day-old P. putida PaW340(pLV1016) population.

VOL. 57, 1991

on April 3, 2019 by guest

http://aem.asm

.org/D

ownloaded from

1910 WINSTANLEY ET AL.

106

U. W

101

i06,105I

_ 104

E 03'1

IL 1021U 101

_ 104E 103'U.102

10

b Q

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28Time (days)

FIG. 4. Survival of plasmid-free strains in sterile lake water. Thegraphs show the survival ofPaW8 (a), PaW340 (b), and PaW340 plustryptophan (c). The strains were released at four different inocula.

(iii) Plasmid-free release hosts. The recovery of the releasehosts P. putida PaW8, PaW340, and PaW340 plus tryp-tophan on nutrient agar over 28 days is illustrated in Fig. 4.No obvious differences were observed between plasmid-containing and plasmid-free populations. Such differencescould well have been masked by the great variability in theresults between replicate model systems. This variabilitywas, in part, due to strains occasionally undergoing an initialdecline in number following release (3 to 6 days). After thisdecline, the populations resumed their characteristic sur-

vival pattern.Survival of P. putida populations in lake water. Figure 5

illustrates thte recovery of P. putida PaW8(pLV1017),PaW340(pLV1017), PaW8(pLV1016), and PaW340(pLV1016) in untreated lake water model systems. Whenhosts were introduced at approximately 105 CFU ml-', thepopulation initially declined within 20 days to 103 CFU ml-'and then became stable at approximately 103 CFU ml-' forup to 28 days. On day 28, no C230 or xylE DNA could bedirectly detected in any of the model systems. When lakewater was inoculated with smaller initial populations, thesealso declined over the 28-day period. At initial inocula of 102and 103 CFU ml-l, the decline continued toward the limit fordetection by culture methods. These results are in agreementwith those obtained from the release of P. putidaPRS2000(pLV1010), for which detection by any of the directmethods was not possible after the cells became uncul-turable (17). Similar survival patterns were obtained forPaW8(pLV1013) and PaW340(pLV1013) (data not shown).

10

102

101

io6

108

E

ILC)%

104103102101

d

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28Tirme (days)

FIG. 5. Survival of strains containing pLV1016 or pLV1017 inuntreated lake water. The graphs show the survival of PaW8(pLV1017) (a), PaW340(pLV1017) (b), PaW340(pLV1017) plus tryp-tophan (c), PaW8(pLV1016) (d), and PaW340(pLV1016) (e). Thestrains were released at four different inocula.

E 103ILl0%

101(

loo105

-O

E 1D3IL 102(

10B

102

_ 1iE 103,L 102,U.lo

C

APPL. ENVIRON. MICROBIOL.

I

on April 3, 2019 by guest

http://aem.asm

.org/D

ownloaded from

SURVIVAL OF P. PUTIDA POPULATIONS IN LAKE WATER 1911

00001.1 1 _

5 10 15 20 25 30 5 10 15 20 25 30 5 10 15 20 25 30DAYS DAYS DAYS

FIG. 6. Survival of strains containing pLV1013 in untreated lake water. The graphs show the survival of PaW8(pLV1013) (a), PaW340(pLV1013) (b), and PaW340(pLV1013) plus tryptophan (c) in three different water samples taken in consecutive months from the same site(indicated by squares [September], open circles [October], and closed circles [November]). The survival patterns for each of three replicatemicrocosms are presented.

Analysis of triplicate releases with a single water sample. Anexperiment was carried out to ascertain whether differencesin survival between strains containing pLV1016 andpLV1017 could be detected in the same water sample. A plotof survival of the strains in lake water showing averagevalues taken from three plates for inocula 2, 3, and 4 andnine plates (three for each replicate model system) forinoculum 1 is shown in Fig. 5.An analysis was developed for the data obtained by use of

a statistical model which assumes that the counts are dis-tributed in a way proportional to a Poisson distribution andthat the average count arises as the product of an initialinoculum and a survival proportion. Models of this form fallwithin the generalized linear model framework (16), andcomputer programs (2) were used to fit the models.To investigate strain and inoculum differences, we fitted

three statistical models to the data. The first assumed thatsurvival was identical in all model systems (the common-

proportions model). The second assumed that survival was

identical in all model systems with the same combinations ofstrain and inoculum but that there were different survivalpatterns for different combinations (the separate-proportionsmodel). The third assumed that each model system had a

unique survival pattern (the full model). The change in thegoodness-of-fit between the separate-proportions model andthe full model demonstrated the natural variability in sur-

vival patterns observed to occur between replicate modelsystems. The change in goodness-qf-fit between the com-mon- and the separate-proportions models demonstrated thevariability between the combinations of strain and inoculum.The latter variability was statistically tested to determinewhether it was significantly greater than the variabilitybetween model systems by use of the methodology outlinedby McCullagh and Nelder (16). The test was not significant(P > 0.05) and tentatively suggests that there is no differencebetween strains or inocula. To confirm this suggestion, weinvestigated separately the first inoculum, for which repli-cate model systems were available. The test, although bor-derline (0.05 < P < 0.054), was not significant. This analysistherefore suggests that there are no obvious differencesbetween the strains.

Analysis of triplicate releases with water samples taken fromthe same site in three consecutive months. The survival ofstrains containing the IncQ plasmid pLV1013 was alsoassessed. In this experiment, additional information wassought about the variability which may occur between watersamples. Thus, releases were carried out with samples takenfrom the same location in September, October, and Novem-ber 1989. A graph of the survival of each strain in the threedifferent months, with averages for each set of three plates,is shown in Fig. 6.Each strain was considered in turn (nine model systems)

VOL. 57, 1991

on April 3, 2019 by guest

http://aem.asm

.org/D

ownloaded from

1912 WINSTANLEY ET AL.

and tested with the methodology described above to deter-mine whether the variations among proportionate survival ineach month were significantly greater than those observedbetween replicate model systems. Subsequently, eachmonth was considered in turn (nine model systems) andtested to determine whether the variations among propor-tionate survival of each strain were significantly greater thanthose observed between replicate model systems.These tests indicated that proportionate survival differed

significantly between months for all three strains (P < 0.001)and between strains for all three months (P < 0.001 forSeptember and October and P < 0.05 for November). Therewere some problems with the analysis of the data. The actualinocula for September, October, and November, respec-tively, were 62 x 105, 5 x 105, and 1.6 x 105 for strain 1 and5.5 x 105, 3.2 x 105, and 15.7 x 105 for strains 2 and 3. Theapparent differences between months for each strain could tosome extent be a reflection of the different inocula. Also, theapparent differences between strains in September could beinfluenced by the different initial inocula.

DISCUSSION

R68.45 has been shown to participate in conjugal transferto aquatic and soil bacteria in laboratory-based matings andin situ (9, 19, 24). The broad-host-range nature of the markerplasmid system developed in this work should allow thedetection of many naturally occurring bacteria that couldreceive the marker plasmids from the released host. Thenonconjugative but mobilizable IncQ plasmids provide ameans of comparing the spread of genes from conjugativeand nonconjugative plasmids in natural populations. Themain advantage of these systems is that the presence of areleased host and plasmids transferred to other hosts can bedetected without the need for antibiotic selection, offeringthe advantage that positive selection for cells containing theplasmids need not be applied during their recovery frommodel environments. The effective expression of xylE in arange of gram-negative hosts also permits the construction ofa range of marked strains for release studies.These studies have shown that it is possible to monitor the

fate of xylE marker strains of P. putida in lake water andsterile lake water by determination of colony-forming ability,direct C230 assay, and direct detection of the xylE gene.Since no NCBV population was observed, colony formationproved a valuable source of information for the developmentof a model framework and enabled us to carry out a moredetailed statistical analysis of the survival of released P.putida strains by using both three replicate releases into asingle lake water sample and releases into three differentlake water samples (each with three replicates) in consecu-tive months. Additional information can now be easily addedto this data base, and the analysis of these results can bemodified as required.The test strain, P. putida PaW8, was able to survive and

maintain a concentration of 105 to 106 CFU ml-1 over 28days in sterile lake water. Only the auxotrophic test micro-organism, in the absence of growth supplements, declinedrapidly once released into sterile lake water. Wild-type P.putida can survive for up to 28 days without any reduction incell numbers in sterile lake water. Survival for this periodoccurs in the absence of any exogenous carbon source.Similar results have been found for Pseudomonas spp. andother genera (3, 7, 15, 22, 23). In P. putida PRS2000 and P.putida PaW8, recombinant plasmid pLV1010 was unstableduring growth in sterile lake water. The IncP regulated

plasmid pLV1017 was more stable than pLV1010. Thegreater stability of pLV1016 and pLV1013 may reflect areduction in metabolic burden because of the reducedexpression of xylE on these plasmids. In all cases, greaterinstability was observed in cells growing in sterile lake waterthan in nutrient broth cultures. Because of the variability insterile lake water microcosms, it was difficult to assesswhether any of the plasmids affected the survival of the host.In lake water, although the variability in the results wasreduced, differences in the survival of plasmid-free andplasmid-containing strains could not be determined becausethe presence of the xylE gene was essential for strainidentification in mixed microbial populations.When P. putida populations were released into untreated

lake water, the population initially declined. When theinoculum was larger than 104 CFU ml-1, after 15 days theinitial decline was normally followed by a period of little orno decline. This result indicates that the survival of P. putidawas enhanced by the introduction of a larger inoculum. Theincreased survival may have been due to a heavier dose ofpotential growth substrates derived from cell lysis of theoriginal inoculum. Models designed with flowthrough sys-tems may present a more realistic situation in which tomonitor the survival of such microorganisms. In addition, arapid initial decline in bacterial numbers followed by aperiod with little or no decline may be characteristic ofbacterial prey from protozoan predators (1). There was,however, an initial decline in numbers in populations re-leased into sterile systems, in which no predators arepresent, and in some cases, the initial decline in numbers inuntreated water continued. What dictates which few cellswill survive is unknown. It may be random chance or geneticselection. Similar periods of survival of pseudomonads innonsterile water, sewage, and soil have been observed (6,14, 23), but differences in microcosm design, environmentalsamples investigated, pretreatment of cells, and incubationconditions hinder a closer comparison with the results ob-tained here.When plasmids pLV1016 and pLV1017 were used as

markers for the release strains, no significant difference insurvival in lake water between strains was observed. Inthe three separate release studies with pLV1O13, however,there was an apparent difference in survival between strainsin each of the water samples. The survival patterns ofpopulations of PaW8 and PaW340 were similar, suggestingthat the auxotrophic nature of PaW340 was not a significantfactor. The addition of tryptophan led to an improvementin the survival of populations of PaW340 although, sincethe extra nutrient provided would be likely to cause adisruption of the natural population, this result is difficult tointerpret.The difference between the survival of strains depending

on the water sample was more clear cut, with large differ-ences in CFU obtained at the end of the 28 days. Althoughno attempt was made in this simple, preliminary study toidentify factors within the samples which could affect sur-vival, it would seem valid to conclude that the survival of astrain released into the natural environment could be highlyvariable depending on the time and place of the release. Itmay well be that the natural population varies to such anextent over short distances or short periods of time in naturalhabitats that it will be impossible to predict accurately thesurvival of a released organism.

APPL. ENVIRON. MICROBIOL.

on April 3, 2019 by guest

http://aem.asm

.org/D

ownloaded from

SURVIVAL OF P. PUTIDA POPULATIONS IN LAKE WATER

ACKNOWLEDGMENTS

Statistical analysis was carried out by Margaret Hurley at theInstitute of Freshwater Ecology, Cumbria, United Kingdom. Wethank Melanie Seasman for technical assistance.

This work was supported by a postgraduate studentship (toJ.A.W.M.) and a project grant (to J.R.S. and R.W.P.) from theNatural Environment Research Council, Swindon, United King-dom.

REFERENCES1. Alexander, M. 1981. Why microbial predators and parasites do

not eliminate their prey and hosts. Annu. Rev. Microbiol.35:113-133.

2. Baker, R. J., and J. A. Nelder. 1978. General linear interactivemodelling (GLIM). Release 3. Oxford Numerical AlgorithmsGroup, Oxford, United Kingdom.

3. Barcina, I., J. M. Gonzalez, J. Iriberri, and L. Egea. 1989. Effectof visible light on progressive dormancy ofEscherichia coli cellsduring the survival process in natural fresh water. Appl. Envi-ron. Microbiol. 55:246-251.

4. Close, J. C., and R. L. Rodriguez. 1982. Construction andcharacterization of the chloramphenicol resistance gene car-tridge: a new approach to the transcriptional mapping of extra-chromasomal elements. Gene 20:305-316.

5. Colwell, R. R., P. R. Brayton, D. J. Grimes, D. B. Roszak, S. A.Huq, and L. M. Palmer. 1985. Viable but non-culturable Vibriocholerae and related pathogens in the environment: implicationsfor release of genetically engineered microorganisms. Bio/Tech-nology 3:817-820.

6. Compeau, G., B. J. Al-Achi, E. Platsouka, and S. B. Levy. 1988.Survival of rifampin-resistant mutants of Pseudomonas fluo-rescens and Pseudomonas putida in soil systems. Appl. Envi-ron. Microbiol. 54:2432-2438.

7. Cruz-Cruz, N. E., G. A. Toranzos, D. G. Ahearn, and T. C.Hazen. 1988. In situ survival of plasmid-bearing and plasmidlessPseudomonas aeruginosa in pristine tropical waters. Appl.Environ. Microbiol. 54:2574-2577.

8. Franklin, F. C. H., and P. A. Williams. 1980. Construction of apartial diploid for the degradative pathway encoded by the TOLplasmid (pWW0) from Pseudomonas putida mt-2: evidence forthe positive nature of the regulation by the xylR gene. Mol. Gen.Genet. 177:321-328.

9. Genthner, F. J., P. Chatterjee, T. Barkay, and A. W. Bourquin.1988. Capacity of aquatic bacteria to act as recipients of plasmidDNA. Appl. Environ. Microbiol. 54:115-117.

10. Guerry, P., J. LeBlanc, and S. Falkow. 1973. General method forisolation of plasmid deoxyribonucleic acid. J. Bacteriol. 116:1064-1066.

11. Haas, D., and B. W. Holloway. 1976. R factor variants withenhanced sex factor activity in Pseudomonas aeruginosa. Mol.Gen. Genet. 144:243-251.

12. Jones, J. G., and B. M. Simon. 1975. An investigation of errors

in direct counts of aquatic bacteria by epifluorescence micros-copy, with respect to a new method for dyeing membrane filters.J. Appl. Bacteriol. 39:317-329.

13. Kogure, K., U. Simidu, and N. Taga. 1979. A tentative directmicroscopic method for counting living marine bacteria. Can. J.Microbiol. 25:415-420.

14. Kurath, G., and R. Y. Morita. 1983. Starvation-survival physi-ological studies of a marine Pseudomonas sp. Appl. Environ.Microbiol. 45:1206-1211.

15. Lim, C. H., and K. P. Flint. 1989. The effects of nutrients on thesurvival of Escherichia coli in lake water. J. Appl. Bacteriol.66:559-569.

16. McCuliagh, P., and J. A. Nelder. 1983. Generalized linearmodels, p. 127-148. Chapman & Hall, Ltd., London.

17. Morgan, J. A. W., C. Winstanley, R. W. Pickup, J. G. Jones,and J. R. Saunders. 1989. Direct phenotypic and genotypicdetection of a recombinant pseudomonad released into lakewater. Appl. Environ. Microbiol. 55:2537-2544.

18. O'Farrell, P. J., E. Klutter, and M. Nakanishi. 1980. A restric-tion map of bacteriophage T4 genome. Mol. Gen. Genet.179:421-435.

19. O'Morchoe, S. B., 0. Ogunseitan, G. S. Sayler, and R. V. Miller.1988. Conjugal transfer of R68.45 and FP5 between Pseudomo-nas aeruginosa strains in a freshwater environment. Appl.Environ. Microbiol. 54:1923-1929.

20. Reimmann, C., and D. Haas. 1987. Mode of replicon fusionmediated by the duplicated insertion sequence IS21 in Esche-richia coli. Genetics 115:619-625.

21. Riess, G., B. Masepohl, and A. Puhler. 1983. Analysis ofIS21-mediated mobilization of plasmid pACYC184 by R68.45 inEscherichia coli. Plasmid 10:111-118.

22. Scheuerman, P. R., J. P. Schmidt, and M. Alexander. 1988.Factors affecting the growth of bacteria introduced into lakewa-ter. Arch. Microbiol. 150:320-325.

23. Sinclair, J. L., and M. Alexander. 1984. Role of resistance tostarvation in bacterial survival in sewage and lake water. Appl.Environ. Microbiol. 48:410-415.

24. Van Elsas, J. D., J. T. Trevors, and M. E. Starodub. 1988.Bacterial conjugation between pseudomonads in the rhizo-sphere of wheat. FEMS Microbiol. Ecol. 53:299-306.

25. Willetts, N. S., C. Crowther, and B. W. Holloway. 1981. Theinsertion sequence IS21 of R68.45 and the molecular basis formobilization of the bacterial chromosome. Plasmid 6:30-52.

26. Winstanley, C., J. A. W. Morgan, R. W. Pickup, J. G. Jones,and J. R. Saunders. 1989. Differential regulation of lambda PLand PR promoters by a cI repressor in a broad-host-rangethermoregulated plasmid marker system. Appl. Environ. Micro-biol. 55:771-777.

27. Worsey, M. J., and P. A. Williams. 1975. Metabolism of tolueneand xylenes by Pseudomonas ([sic]putida (arvilla) mt-2: evi-dence for a new function of the TOL plasmid. J. Bacteriol.124:7-13.

VOL. 57, 1991 1913

on April 3, 2019 by guest

http://aem.asm

.org/D

ownloaded from