Environmental Microbiology (1999) 1(3), 231-241

Microbiai characterization of a JP-4 fuei-contaminatedsite using a combined iipid biomarker/poiymerase chainreaction-denaturing gradient gei eiectrophoresis(PCR-DGGE)-based approach

John R. Stephen,' Yun-Juan Chang,^Ying Dong Gan/ Aaron Peacock,' Susan M. Pfiffner,'Michael J. Barcelona,^ David C. White'^ andSarah J. Macnaughton'̂ ^^Center for Environmental Biotechnoiogy. The Universityof Tennessee. 10515 Research Drive, Suite 300.Knoxviile, TN 37932, USA.^Nationai Center for integrated Bioremediation Researchand Development. Department of Civil and EnvironmentalEngineering. University of Michigan, Mi 48109, USA.'^Environmentai Sciences Division. Oak Ridge NationalLaboratory, Oak Ridge. TN 37831. USA.

Summary '

The impact of poiiution on soii microbjai communitiesand subsequent bioremediation can be measuredquantitatively in situ using direct, non-culture-dependent techniques. Such techniques have advan-tages over culture-based methods, which oftenaccount for less than 1% of the extant microbiai com-munity, in 1988, a JP-4 fuel spill contaminated thegiacio-fiuvial aquifer at Wurtsmith Air Force Base,Michigan, USA. In this study, lipid biomarker charac-terization ofthe bacteriai and eukaryotic communitieswas combined with poiymerase chain reaction-denaturing gradient gei electrophoresis (PCR-DGGE)analysis of the eubacterial community to evaiuatecorrelation between contaminant (JP-4 fuei) concen-tration and community structure shifts. Vadose, capil-lary fringe and saturated zone samples were takenfrom cores within and up- and down-gradient fromthe contaminant plume. Lipid biomarker analysis indi-cated that sampies from within the plume containedincreased biomass, with large proportions of typicallyGram-negative bacteria. Outside the piume, iipid pro-fiies indicated low-biomass microbial communitiescompared with those within the initial spiii site. 16SrDNA sequences derived from DGGE profiles fromwithin the initial spiii site suggested dominance of

Received 4 September, 1998; revised 8 January, 1999; accepted 1February. 1999, "For correspondence. E-mail sjmac@utk,edu; Tel.(+1) 423 974 8005; Fax (+1) 423 974 8027,

the eubacterial community by a limited number ofphylogenetlcally diverse organisms. Used in tandemwith poiiutant quantification, these moiecuiar tech-niques shouid facilitate significant improvementsover current assessment procedures for the deter-mination of remediation end-points.

introduction

Accidental spillage of JP-4 jet fuel has been a significantsource of hydrocarbon contamination in soils and ground-waters (Fang ef al., 1997). Microorganisms have beenshown to use many of the components of such fueis,including the 'BTEX' compounds (benzene, toiuene, ethyi-benzene and xylenes), as electron donors, thereby facili-tating contaminant remediation (Weidemeier et al., 1995).One of the major concerns in the fieid of environmentairemediation is in the establishment of protocols for thedetermination of defensible end-points. Current practicesoften monitor the disappearance of poilutants and theirby-products to regulatory levels or use single-species toxi-city tests (Rand and Petrocelll, 1985). However, shifts inthe microbial community can aiso provide a measure ofbioremediation effectiveness (Pfiffner et al.. 1997; Whiteet aL, 1998), Indeed, the microbial community responsemay prove to be a more comprehensive assessment indi-cator of residual toxicity, because it may provide a moresensitive measure of biologically available contaminantsthan current toxicity tests. In this role, microbial communityanalysis could serve as a useful compiement to currentmethods for measuring the disappearance or sequestrationof hazardous contaminants.

Traditionary, methods used to monitor microorganismsrequire ex situ culture anaiysis. iHowever, It has beendocumented repeatedly that such culture techniquesaccount for oniy between 0.1% and 10% ofthe communitynumericaliy detectabie by direct counting (Si^inner et al..1952; White, 1983; Bakken, 1985; Tunlid and White,1992; White et al., 1993). Equally, current culture-basedmethods have now been shown to overlook many of thein situ numerically dominant species completely (e.g.Amann et ai., 1995, and references therein). Signaturelipid biomarkers (SLBs) can be used to determine shifts

1999 Blackwell Science Ltd

232 J. R. Stephen et al.

in microbial biomass, nutritional or physiological stressresponses and community diversity in situ. Total cellularphospholipid fatty acids (PLFAs) are not stored and areturned over rapidly. They can therefore be used as indica-tors of viable biomass (Vestal and White, 1989). Increasesin the ratios of trans/cis monoenoic PLFAs in cells are indi-cative of the effects of toxic/sublethal stress on bacterialcommunities and of the growth phase of cells respectively(Guckert etal.. 1986; Heipieper etal.. 1992; White etal.,1996a,b). Also, bacteria make poly-^-hydroxyalkanoicacid (PHA) as endogenous storage lipids (Findlay andWhite, 1983; Doi, 1990), and the relative proportion ofthis lipid to PLFAs can provide a measure of cellular nutri-tional status. Specific PLFA biomarkers can be used asindicators of microbial community composition, e.g. eubac-teria, fungi, algae. Gram-negative and Gram-positiveorganisms, sphingomonads, actinomycetes and sulphate-reducing bacteria (Vestal and White, 1989; White et al..1996a,b). Plasmalogen-derived dimethylacetals (DMAs)are formed after mild acid methanolysis, and an increasedratio of DMA to PLFA indicates increased proportions offacultative and obligate anaerobic metabolism (Whiteet al.. 1996b). Despite its versatility, PLFA analysis doeshave limitations for the analysis of the Gram-negative bac-teria community structure. Generally, the PLFA profiles ofGram-negative bacteria are dominated by monoenoic,saturated and cyclopropane fatty acids (Wilkinson, 1988;Zelles. 1997), the vast majority of which are broadly dis-tributed and, as such, uninformative in subdividing theGram-negative community structure.

To overcome this, we have used a complementarynucleic acid-based analysis. The polymerase chainreaction-denaturing gradient gel electrophoresis (PCR-DGGE) approach used employed primers that recognizethe 16S rDNA targets of all known and inferred (from16S rDNA clone libraries) eubacterial species (Muyzeret al.. 1993). PCR products of approximately equal lengthare separated on the basis of their melting behaviour in anacrylamide gel matrix. The banding patterns and relativeintensities of the recovered bands can provide a measureof changes in the eubacterial community between sam-pling sites and over time. The major limitation of this tech-nique lies in low sensitivity; It is likely that any given 16SrDNA sequence type must comprise at least 1% of thetotal target organisms in a sample to remain above thebackground level of minor bacterial amplification products(Muyzer et al.. 1993; Heuer and Smalla, 1997; Stephenetal., 1998). Excision and sequence analysis of individualbands can be used to provide fine-scale biomarkers andloosely infer the identity of the source organisms usingdatabase searches and phylogenetic methods.

These two technologies applied in tandem should per-mit broad and specific observations of whole-communitysuccession and subpopulation dynamics in contaminated

soils, as well as enabling the partial characterization ofbioremediative communities in situ. The aim of this studywas to provide a detailed microbial community structurecharacterization using lipid/PGR-DGGE-based analysisof a hydrocarbon-contaminated site. We describe a multi-faceted microbiological analysis using SLB and PCR-DGGE of total eubacterial 16S rDNA to measure microbialbiomass, community diversity and nutritional/physiologi-cal status (community stress indicators) at a highly con-taminated and at uncontaminated locations in a JP-4spill site for which the organic contaminants have beencomprehensively documented previously (Fang ef a/., 1997).

Results

Geochemicai and volatile organic compound(VOC) analyses

Total VOCs from within and up-gradient of the crash sitewere detected in the saturated zone at 732 and46 (ig kg ^ respectively. The VOCs were below detectionlimits in all vadose and capillary fringe zones and at alllevels downgradient of the site. These findings were con-trary to those of Fang et al. (1997), who detected traceamounts of VOCs down-gradient rather than up-gradientof the crash site. The major compounds detected withinthe crash site included the m and p xylenes (203 iLg kg '),1,2,4-trimethylbenzine (9iM.gkg '), various methylben-zines (24-44jigkg ^), naphthalene (22|xgkg ') andmethylnaphthalene (20ti.gkg"').

Biomass content

Biomass contents varied considerably between boreholes(Table 1, Fig. 1). The biomass content of the samples fromthe crash site was significantly higher than that of samplestaken from either up- or down-gradient of the site(P<0.05), as determined using analysis of variance(ANOVA). Bacterial cell numbers were calculated based onPLFA recovery data (Balkwill et al., 1988). It is importantto remember that, with any conversion factor, the numberof cells can vary by up to an order of magnitude (Findlayand Dobbs, 1993). Bacterial cell numbers for these sam-ples perg wet weight ranged from 5.2±0.2x10^ In thesaturated zone from the up-gradient sample to 3.9 ±0.3 X10^ in the sample taken from the capillary fringe ofthe crash site sample.

Community structure

The microbial community structures of the samples dif-fered depending on both borehole location and depth(Fig. 2). Figure 2 shows different structural groups ofPLFA plotted as a percentage of total PLFA. The microbial

1999 Blackwell Science Ltd, Environmental Micrt^iology, 1, 231-241

ci0)o0)

S

oil

t

cn

nts

of t

h

5?cou.eg

o>co•oQ

i

ilphi

:̂jnlU.oo

naei

CD

CD

o

and

DM

A)

ra

X

'LF

A,

Par

ker

(F

E

bio

1. T

ota

l

Z

CO

6a

Li.

u6Q

G-V

Q

CO

O

CS

-CF

CO

o

CO

6D

LLO

UG

-

sCO

Q Q Uz z z

Q cy O

CO CO t-J d

^ O> CO— — T COo CO CO d

so_

tf) CM T- CDCNJ O ^— O• - d CO d

lUT)

ra•a

cegCO

o

raf

8

CM —

t : CM

q o — T-•^ d 3 d

CC) O - ^

d r'c^ ' - 00

CJ - ^ ^ -

stScvK^

SE^cqgcDf g ^ C M O d o

Q LL

o ir

ra F

'. 0)

.o o)E -£_

O — raD ^ tE

S 03 >

1 ^ 7

^ in

(I) A

CO mCO ci)<3) Q)

u .2- ca- ca

CO >O CD

.S rara c

ra

'^ ra

E E onj ra m

C LL0) _i

CT *— .x- — i -d W _ Q^ Q,^ O O Q CU3̂- z o CC a :3 n ^ u "d

Community cfiaracterizarion by lipids/PCR-DGGE 233

2500 .

•3 2000

^ 1500oES 1000

500

0

Up-gradient Crash site* Down-gradient

Sample site

Fig. 1. Total biomass measured as PLFAg ' dry weight soil, ( ^ )represents the vadose zone; (•) represents the capillary fringe;and (@) represents the saturated zone. n=2 , error bars representstandard deviation. Downgradient vadose represents a single datapoint. Crash site samples (•'•) contained significantly more totalbiomass than did samples from up- and downgradient.

communities from the crash site samples contained signif-icantly more monoenoic PLFA (P<0.05), indicative ofGram-negative bacteria (Wilkinson, 1988), than did sam-ples from up- and down-gradient of the site. Although con-taining significantly less biomass than samples taken fromwithin the crash site (P<0.05), samples from up-gradientand down-gradient contained significantly higher relativeproportions of PLFA (P< 0.05) indicative of sulphate-redu-cing bacteria (10me16:0, i17:1w7c; Dowling etai., 1986and Ediund etai., 1985 respectively). At all sites, the rela-tive proportions of the biomarkers indicative of sulphate-reducing bacteria were significantly higher (P<0.05) inthe capillary fringe and saturated zones (Fig. 2). In allsamples, the relative proportions of terminally branchedsaturated PLFAs, such as i15:0, i17:0 and cy17:0, whichwere taken in this case to be indicative of anaerobicGram-negative bacteria (Wilkinson, 1988), increasedwith zone depth (P<0.05). Conversely, relative propor-tions of biomarkers typical of eukaryote PLFAs (e.g.18:2u>6 and 18:3) decreased with depth,

A hierarchical cluster analysis (HCA) of the bacterialPLFA profiles (arcsine-transformed mol percentage data)showed the relatedness between samples, as indicatedin Fig. 3. The bacterial PLFAs comprised the totalPLFAs minus the polyenoic and normal saturated PLFAsabove 18 carbons in chain length, both of which are gen-erally associated with eukaryote biomass (White et al.,1996b), although polyenoic fatty acids have been detectedin large amounts in some Shewanelia spp. (Nichols et ai.,1997; Watanabe ef al., 1997). From the HCA, it wasapparent that the bacterial populations from within thecrash site were dissimilar to one another and from theup- and down-gradient samples, whereas the PLFAprofiles from up- and down-gradient showed a higher

1999 Blackwell Science Ltd, Environmental Microbiology, 1, 231-241

234 J. R. Stephen et al.

60

UG-V UG-CF UG-S CS-V CS-CF CS-S DG-V DG-CF DG-S

Sample siteFig. 2. A comparison of the relative percentages of total PLFA functional groups from boreholes located upgradienl, within the crash sile anddowngradient of the contaminani plume. Functional groups are defined according to PLFA molecular structure, which is related to fatty acidbiosynthesis. Terminally branched saturates (ESI) are attributed to Gram-positive and. in this case, anaerobic Gram-negative bacteria:monoenoic unsaturates (M) are attributed to Gram-negative bacteria; branched monoenoic unsaturates (•) and mid-chain branched saturates(Cl) are attributed to actinomycetes and sulphate-reducing bacteria; normal saturates (;;7j) are ubiquitous; and polyenoics (13) are attributed toeukaryotes. UG. upgradient; CS, crash site; DG, downgradient; V, vadose zone, CF, capillary fringe; S, saturated zone. n=2, error barsrepresent standard deviation, DG-V represents a single data point. Asterisks denote the significantly (P<0,05) higher mole percentage ofmonoenoic PLFAs present in the crash site samples compared with those from the upgradient and downgradient sites. The # sign denotes thesignificantly higher mole percentage of terminally branched saturated PLFAs present in the capillary fringe and saturated zones comparedwith the vadose zones.

DG-Saturatett

DG-Vsdose

UG-Vadose

DC-Cap. Fr.

UG-Satu rated

UG-Cap. Fr.

CS-Stttu rated

CS- Cap. Fr.

CS-Vadose

0,8L_

0.6I

0 4I

0.0 Fig, 3, A dendrogram representation of ahierarchical cluster analysis (single linkagebased on Euclidean distance) for the bacteria)PLFA profiles described in Fig. 1. DG,downgradient; CS, crash site; UG. upgradient.

1999 Btackwell Science Ltd, Environmenial Microbiology. 1, 231-241

level of relatedness, with the two vadose zones clusteringtogether. The crash site vadose zone contained the mostunique community, resulting in the main from the higherrelative proportion of 2me13:0, a fatty acid of unknown ori-gin that was also detected in the crash site capillary fringeand saturated zones. A principal component analysis ofthe same data gave similar results, with two principal com-ponents derived accounting for, sequentially, 76% and15% of the variance inherent in the data set (data notshown). Principal component 1 was most strongly influ-enced by 2me13:0, cy17 and cy19 and accounted forthe diffuse grouping of the crash site samples. Of thesePLFAs, cy17:0 and cy19:0 are common components ofGram-negative bacteria. Principal component 2 wasmost strongly influenced by 10me16:0, representative ofthe sulphate-reducing bacteria Desuifobacter (Dowlingetal., 1986), 16:0, 18:1w9c and a15:0. Representative ofthe presence of anaerobic bacteria, DMAs were detectedin all samples, but were present in significantly greaterquantities in the samples taken from the crash site. Inthese samples, the relative DMA concentrations (DMA/PLFA) followed the order, capillary fringe > saturated > vadosezone. The relative DMA concentration followed no discern-ible order in the remaining boreholes.

Physioiogicai status

The lipid composition of microorganisms is a product ofmetabolic pathways and thus reflects the phenotypicalresponse of the organism to the environment encounteredin a specific sample (White et at., 1996b). Poly-p-hydroxyalkanoic acid was detected in sediment from thecrash site core (the ratio of PHA/PLFA following thedecreasing order: saturated > capillary fringe > vadose;Table 1), indicative of unbalanced growth on excess car-bon. The physiological status of Gram-negative commu-nities can be assessed from the ratios of specific PLFAs.Gram-negative bacteria make trans fatty acids to modifytheir cell membranes as protection against environmentalstresses (Guckert etal., 1986; Keift etai., 1994; Sikkemaet al., 1995), with the higher ratio of translcis fatty acidsindicating greater levels of environmental stress. Ratiosof 0.05 or less are representative of healthy non-stressedcommunities (White ef a/., 1996b). The frans to c/s ratio for16:1to7 and 18:1to7 was higher within the crash site sam-ples than in the samples taken from up-gradient and down-gradient of the site, where VOCs were either undetectableor present only in trace amounts (Table 1).

DGGE analysis of microbiai diversity

DGGE analysis of triplicate subsamples from the threedepths at each of the three sampling sites demonstratedthat the low-biomass up-gradient and down-gradient

Community cfiaracterizarion by iipids/PCR-DGGE 235

samples carried complex communities in which few discern-ible major bands were observed (data not shown). How-ever, analysis of triplicate subsamples taken within theheavily JP-4-impacted zone showed strong and reproduci-ble banding and stratification at all depths (Fig. 4A). Thebanding pattern is schematized in Fig. 4B to simplify label-ling. Two sequences (A and G) were recovered from allthree depths and represented unknown organisms fromthe [3-subgroup of the class proteobacteria. Bands C, H,M and N were absent from the saturated zone; sequenceH showed affinity to the u-subgroup of proteobacteria,whereas the others were classified as p-proteobacteria.Sequences E. I and K were unique to the capillary fringeand represented a novel bacterium associated withthe Fiexibacter-Cytophaga-Bacteroides phylum, a fi-proteobacterium and an a-proteobacterium respectively.Sequence D was the only major sequence recoveredfrom both the capillary and the saturated zones, represent-ing a member of the Cytophaga subgroup. Sequences B,F, J, L and O were only recovered from the saturated zoneand represented members of the e-subgroup of proteo-bacteria, Fiexibacter-Cytophaga-Bacteroides phylum.Gram-positive phylum, 5-subgroup of proteobacteria andthe «-proteobacteria respectively. Less conservativeplacements of the derived sequences within the RDPscheme are given in Table 2. Bands E and F (both relatedto the Saprospira group) co-migrated, as did bands K andL (representing a- and p-subdivision proteobacteria respec-tively) under the DGGE conditions used, reflecting the dif-ficulty in interpreting DGGE banding patterns without theuse of a secondary analytical technique. Results fromsequence analysis of clones were confirmed by hybridiza-tion of digoxygenin (DIG)-labelled DNA to a transfer mem-brane prepared from a duplicate gel (results not shown),The nucieotide sequences for A to O were depositedsequentially into GenBank as accession numbers AFO87798to AFO87811 and AF107769.

Discussion

A microbiai community response to the JP-4 fuel contam-ination at the KC-135 crash site at Wurtsmith AirforceBase was detectable from the results of the combinedlipid biomarker/PCR-DGGE analysis. Within the contami-nated site, the increased microbiai biomass and the shift inthe community lipid profiles towards domination by Gram-negative communities were related to the increased VOCconcentration within the crash site compared with samplestaken up- and down-gradient of the site (Table 1; Fangetai., 1997). Moreover, the PLFA profiles of the samplestaken up- and down-gradient of the contamination indi-cated greater similarity between microbiai biomass atthese sites than with those obtained from samples takenfrom the crash site (Fig. 3).

© 1999 Blackwell Science Ltd, Environmental Microbiology. 1, 231-241

236 J. R. Stephen et al.

1

F

'

Of

o

2 3 4 5

^

-I'

6 7

— E

K

- ^ — ••

8•

—

—

9 10- A

*

• < ;

• #

. M

Fig. 4. A. DGGE eubacterial communityprofile of soil from within the crash site. Theportion of the gel shown represents the range30-52% denaturant. in which all visible bandswere found. Lanes 1-3. saturated zone.Lanes 4-6, capiliary fringe. Lanes 7-9.vadose zone. Lane 10. migration standards,from top: Pseudomonas aeruginosa FRD-1,Shewanella putrifaciens sp. 200, Sphlngomonasaromalicivorans B0695. Alcaligenes eutrophusCH34 and {diffuse double band) Desulfovibriovulgaris Hiidenborough.B. Schematic of (A). Colour scheme oflabelled bands: red. sequences common to allthree soil zones; blue, sequences recoveredfrom both vadose zone and capillary fringe;green, sequences common to capiliaryIringe and saturated zone; black, sequencesonly recovered from a singie zone. Note:co-migrating bands are not necessarilyidentical. Letters refer to sequencecomparisons given in Table 2. Thin blacklines refer to sequences that were notidentified. The blue band marked « containednumerous sequence types (as defined byMspl digestion patterns of clones), whichwere not sequenced. Migration standards areindicated by black lines on the right. Only thebase of the D. vulgaris product is marked.

The previous study (Fang etal.. 1997), which character-ized the organic contamination and the geochemical featuresat this site extensively, used PLFA analysis to a limitedextent by analysing samples from one borehole (withinthe crash site) but to a greater depth than was donehere. The results presented in the study by Fang et al.(1997) suggested a significant anaerobic response in thecommunity adaptation to the presence of hydrocarbons.In our study, however, microbial community responseswithin the crash site were shown to occur in all threezones, ail of which were principally centred on the Gram-negative population. Throughout all the capiliary fringeand saturated zones, there were increases In the relativeproportions ot PLFA indicative of anaerobic bacteria aswell as of DMAs. All these data provide further evidence

of increasing anaerobic activity with depth. Relative pro-portions of biomarkers indicative of sulphate reducers(10me16:0. i17:1(o7c) tended to increase with depth, par-ticularly in the cores taken from up- and down-gradientof the contamination. These biomarkers were recoveredat very low abundance from the borehole within the mosthighly contaminated sampling site (Fig. 2, CS), which isin agreement with the tact that no 16S rDNA sequencesclearly related to known sulphate-reducing lineages wererecovered from this site. The level of each sulphate-reducing species was therefore likely to have beenbelow the detection limit of PCR-DGGE (===1% ot theeubacterial population; Muyzer ef a/., 1993; Stephenef al.. 1998). In all boreholes, the relative proportion ofeukaryote-type PLFAs (18:2to6, 18:3) decreased with

1999 Blackwell Science Ltd, Environmental Microbiology. 1. 231-241

Community characterizarion by iipids/PCR~DGGE 237

Table 2. Comparison of sequences derived from DGGE bands with databases.

Band

A

ecDEFGH1JKLMN0

Closest match

Nevskia ramosaWoltinelia succinogenesNeisseria flavescensBacteroides spb-17B0Flexibacter elegansFlexibacter ferrugineumUnknown proteobacteriumRasbobacteriumTttiobacillus thioparusClavibacter xyiiHirschia balticaPolyangium sp.Thiobacillus thioparusThiobaciilus thioparusSphingomonas sp. BAL 48

Percentage identity

91

86"

eg9298

mstm'.»;9190

RDP affiliation

2.14.22.14.52.14.22.7.1.22.7.2.32.7.2.32,14.22,14.1.92,14.22.16.1.72.14.12.14.42.14.22.14.22.14.1

RDP group name

p-Subdivision proteobacteriae-Subdivision proteobacteriap-Subdtvision proteobacteriaCytophaga subgroupSaprospira groupSaprospira group3-Subdivision proteobacteriaRhizobium-Agrobacterium subgroupp-Subdivision proteobacteriaArthobacter groupw-Subdivision proteobacleria8-Subdivision proteobacteriap-Subdivision proteobacteriap-Subdivision proteobacteriaa-Subdivision proteobacteria

Closest matches were derived from searches of GenBank. RDP classification is based on 'similarity rank' analysis service provided by the RDP(Maidak et al., 1997). Accession numbers are provided for unpublished sequences.a. 90% to unknown e proteobacteria BD4-8,b. 99% to ctone sequence D84619.

increasing depth, indicating an increased sparseness ofeukaryote biomass. Little or no poiyenoic PLFAs above19 carbons in chain length were detected in the core sam-pies, a finding that has been found to be characteristic ofsuch subsurface sediments (Smith et ai., 1986).

The increased translcis ratios detected in the contami-nated samples indicated increased bacteriai adaptationto metabolic stress. Although soii is a complex system,and care should be taken when interpreting changes inthe trans/cis ratios (Frostegard et al., 1996), such shiftshave been detected previously in and correlated withorganic solvent contamination of hydrocarbon-contaminatedgroundwater and sediments (Cox et al., 1994; Pfiffneret al., 1997). hydrocarbon-contaminated soii in iaboratory-based microcosm studies (White et al., 1998) and metai-poiiuted soils (Frostegard et al., 1993; 1996), as weil asin pure culture studies using solvent treatments (Pinkartet at., 1996). For the contaminated soiis anaiysed here,the trans/cis ratios in soii taken from the crash site werehigher than those from samples taken from upgradientand downgradient of the site. Furthermore, the trans/cisratios from the crash site followed the decreasing order,saturated zone > capillary fringe > vadose zone, an orderthat corresponded with that of the PiHA:PLFA ratio(Tabie 1), indicative of unbalanced growth (Findiay andWhite, 1983; Doi, 1990). Increases in the PHA:PLFAratio have been detected previously in soils contaminatedwith organic solvents (Cox et al., 1994; White et al., 1998).Taken with the shift in the trans/cis ratio, these data indi-cate an adaptation of the microbial communities at thissite to growth in metabolicaliy 'stressful' conditions withabundant available carbon.

The soils up- and down-gradient of the crash site wereshown to contain similar PLFA profiles (Fig. 3) with low

biomass, about one order of magnitude less than compar-able samples taken from within the crash site. Similarly,PCR-DGGE analysis of samples from these sites gener-ated complex profiles indicative of an even distribution ofnumerous bacterial species. None of the major bandsfrom these sites co-migrated with strong bands from thecrash site (data not shown). PCR-DGGE of samplesfrom within the crash site generated strong amplificationproducts, and the patterns were indicative of communitiesdominated by a limited number of species. Given the prob-able heterogeneity of the geochemistry within the con-taminant plume, it is clear that a far more extensive siteanalysis is required before the contribution of the speciesdetected to natural bioremediation can be fully evaluated.

In conclusion, shifts in biomass content and communitystructure throughout the JP-4-contaminated soil sampleswere detected, which seemed to be clearly related to theincreased VOC concentration. Measured as PLFAs. thehighest viabie biomass levels, comprising bacteria thatwere actively responding to metabolicaliy stressful condi-tions and growing on excess carbon, were detected inthe most highly contaminated site. The PLFA and DGGEprofiles of samples taken from outside the plume (bothup- and down-gradient) indicated little or no ongoing bio-remediation, with only trace or non-detectable VOC con-tamination. DGGE analysis of the contaminated sitedemonstrated that the vadose, capillary fringe and satu-rated zones were dominated by distinct populations ofbacteria, few if any of which could be assigned to culturedgenera. The sequences of these bands are of limited usein species identification in the absence of cultured organ-isms from the same site. However, they will be of valuein the compilation of molecular biomarkers indicative ofactive bioremediation as more data become available

1999 Blackwell Science Ltd, Environmentai Microbiology, 1, 231-241

238 J. R. Stephen et al.

from this and similar sites. It is fair to speculate that thesesequences represent at least some of the major eubacter-jal species active in bioremediation at the site of sampling.

Experimental procedures

Fieid site and sampling

The contaminated area was iocated at the KC-135 crash siteat Wurtsmith Air Force Base (WAFB). Oscoda, Ml, USA(Fig. 5). Sediment samples were obtained using a Geoprobepiston corer (Geoprobe Systems). Cores were taken fromthree boreholes, up-gradient, within and down-gradient ofthe initial crash site (Fig. 5). Each core was sectioned intovadose (=2.13m), capiiiary fringe (~2.43m) and saturatedzones (= 2.74m). Sampies were split for (i) lipid biomarkeranalysis (75 g cone line sections); (ii) DNA extraction and sub-sequent PCR (= 1.5g; in sterile whirlpacks); and (iii) volatileorganic compound analysis (VOCs; 7.0 g; EPA vials). Sam-ples for VOC analysis were preserved with 5 ml 40%NaHSO^ aqueous solution. Lipid and DNA samples were pre-served on dry ice and shipped overnight to the University ofTennessee, Knoxville.

Voiatile organic compounds

Samples were analysed for VOCs on a HP-5890 series II gaschromatograph (GC) with an HP 5972 mass selective (MS)detector as described by Fang etal. (1997). Before analysis,samples were spiked with two internal standards (diO-xylene and d8-napthalene) and two spike compounds (1,2-dif tuorobenzene and d5-chlorobenzene) at 20 [ig I"'. Separationwas accomplished using an HP-624 GC column; 60 mX 0.25 mm internal diameter (film thickness d.f. = 1.8^Lm;Hewlett-Packard). For all 43 compounds detected, calibrationcurves were linear between I.Omgl ' and 200M.gl '. Com-pounds were identified based on relative retention time andverified by mass spectra. Concentrations of VOCs were cal-culated using the internal standard method and are reportedas M-gkg"̂ (Fang etai.. 1997).

50 100 m

Ftg. 5. Schematic map of the site at Wurtsmith AFB showing thearea covered by the plume and the locations of boreholes andmonitoring weils.

Lipid analysis

All solvents used were of GC grade and were obtained fromFisher Scientific. Duplicate subsamples of each zone fromeach soil core (six subsamples per core) were extractedusing the modified Bligh/Dyer technique as described pre-viously by White et al. (1979). The total lipids obtained werethen fractionated into glyco-, neutral and polar lipids (Guckertet at., 1985), The polar lipid was subjected to a sequentialsaponification/acid hydrolysis/esterification (Mayberry andLane. 1993). The PLFA and dimethylacetal (DMA) methylesters were recovered. The PLFA and DMAs were separated,quantified and identified by gas chromatography-mass spec-trometry (GC-MS; Ringelberg etai., 1994). Fatty acids wereidentified by relative retention times, comparison with authen-tic standards (Matreya) and by the mass spectra (collected atan electron energy of 70mV; Ringelberg et al.. 1989). Fattyacid nomenclature is in the form of 'A;BioC'. where "A' desig-nates the total number of carbons, 'B' the number of doublebonds and 'C the distance of the closest unsaturation fromthe aliphatic end of the molecule. The suffixes 'c' for cis and'f for trans refer to geometric isomers. The prefixes 'i ' , 'a'and 'me' refer to iso and anteiso methyl branching and mid-chain methyl branching, respectively, with cyclopropyl ringsindicated by 'cy' (Kates, 1986).

The glycolipid fraction was subjected to ethanolysis, andthe p-hydroxy acids from the PHA were extracted and ana-lysed by GC-MS (Findlay and White. 1983). The GC was pro-grammed from an initial temperature of SCC to 280-C at 10^Cper min and then held at this temperature for 3 min. The injec-tor and housing temperatures were maintained at 27O''C and29O''C respectively. Mass spectra were collected as describedabove. PHA was identified by relative retention time andmass spectra (Findlay and White, 1983).

DNA analysis

Nucleic acid was extracted directly from triplicate 0.5 g sub-samples from each zone from each soil core (nine subsam-ples per core) according to the method of Borneman et al.(1996), with modifications described in Stephen et al.(1998). PCR amplification of the 16S rDNA fragments beforeDGGE was performed as described by Stephen etal. (1998).Briefly, thermocycling consisted of 35 cycles of 92°C for 45 s,55"C for 30 s and 6 8 ^ for 45 s. using 1.25 units of Expand HFpolymerase (Boehringer) and lOpmol each of tho primersdescribed by Muyzer etal. (1993) (the forward primer carriedthe 40 bp GC clamp) in a total volume of 25 ^1, Thermocyclingwas performed using a 'Robocyler' PCR block (Stratagene).The primers targeted eubacterial 16S regions correspondingto Esc/ier/c/raco/Zpositions 341-534 (Brosius ef a/.. 1981). Aportion (20%) of each PCR product was analysed by agarosegel eiectrophoresis (1.5% agarose, Ix TAE buffer) and ethi-dium bromide fluorescence. The amount of DNA used forDGGE analysis was standardized to 600 ng by comparisonwith molecular weight standards (1 kb+ ladder; Gibco BRL)using ALPHA-IMAGER software (Alpha Innotech).

DGGE was performed using a D-Code 16/16 cm gel sys-tem with 1.5 mm gel width (Bio-Rad) maintained at a constanttemperature of 6O'C in 61 of 0.5 x TAE buffer (20 mM Trisacetate, 0.5 mM EDTA, pHS.O). Gradients were formed

1999 Btackwell Science Ltd. Environmental Microbiotogy, 1, 231-241

Community characterizarion by lipids/PCR-DGGE 239

between 20% and 55% denaturant with 100% denaturantdefined as 7M urea plus 40% (v/v) formamide. Gels wererun at 35 V for 16h. Gets were stained in purified water(Milli-Ro; Miliipore) containing ethidium bromide (0.5mgl )̂and destained twice in 0.5xTAE for 15 min each. Imageswere captured using the ALPHA-IMAGER software (Alpha Innotech).

Extraction of DNA from acrylamide gels

The central 1 mm^ portion of strong DGGE bands wereexcised using a razor blade (American Safety Razor Company)and soaked in 50 (xl of purified water (Milli-Ro; Miliipore) over-night. A portion (15 |i.l) was removed and used as the templatein a PCR reaction as above. The products were purified byelectrophoresis through a 1.2% agarose/TAE gel followed byglass-milk extraction (Gene-Clean kit; BIO-101).

Cloning of PCR-amplifled products

Amplification products that failed to generate legible sequencedirectly were cloned into the PCR-TOPO 2.1 cloning vector(Invltrogen) according to the manufacturer's instructions.Recombinant (white) colonies were screened by a two-stageprocedure to ensure recovery of the DGGE band of interest.First, plasmid inserts (n = 12 for each band) were reamplifiedby PCR using vector-specific primers (M13 reverse and T7;Invitrogen). The products were digested with restriction endo-nuclease Mspl and analysed by agarose gel electrophoresis(2% agarose. Ix TAE buffer). Two products from each diges-tion pattern group were reamplified using the 16S-specificPCR primers described above (Muyzer ef a/., 1993) and sub-jected to DGGE analysis to select sequences that co-migratedwith the original band of interest. Sequences that were of highfrequency in clone libraries (as defined by digestion pattern)and co-migrated with the original environmental band wereselected for sequence analysis (two clones band"') andused as probes in confirmatory membrane hybridizationanalysis.

Membrane transfer and hybridization analysis

DNA was transferred electrophoretically from DGGE gels topositively charged nylon support hybridization membranes(Boehringer Mannheim) using a model SD electroblotter (Bio-Rad) at 40 mA for 1 h. The transfer buffer was 0.5 x TAE.Hybridization analysis used DIG-labelled PCR products(PCR DIG probe synthesis kit; Boehringer Mannheim) and aDIG nucleic acids detection system (Boehringer Mannheim)according to the manufacturer's instructions. Hybridizationwas at es^C overnight in a rotary oven (Personal Hyb; Strata-gene) in the manufacturer's 'standard buffer'. Washing andstaining were as recommended by the manufacturer.

Sequence analysis

PCR products from excised bands and cloned products weresequenced using the primer 516r (GWATTACCGCGGC-KGCTG; W ^ A orT, K = G orT; Lane etal.. 1985) and anABI-Prism model 373 automatic sequencer with dye termina-tors (Perkin-Elmer). Sequences were compared with the

GenBank database using the BLASTN facility of the NationalCenter for Biotechnology Information (http://ncbi.nlm.nih.gov).Sequences were classified using the RDP release of 31 July1998(Maidake/a/., 1997).

Statistical analysis

Analysis of variance (ANOVA) was used to determine whetherthere were significant differences between the lipid biomarkerdata obtained from the crash site (n=3) and that obtainedfrom up- and down-gradient of the site (n —6). ANOVA wasalso used to determine significant differences between lipidbiomarker data obtained from each zone ( n = 3 for eachzone). ANOVA was performed on the means of the duplicatesubsamples (pseudoreplicates) using STATISTICA version 5.1for Windows software (Statsoft). Groupings for ANOVA wereassigned a posteriori. Chromatographic peaks with zeroarea were replaced with half the minimum integrated areafor each injection before calculation of mole percentages.Plots of log (average) versus log (variance) were used todetermine the appropriate transformation of the variables(Downing, 1979). The square root transformation was cho-sen. A hierarchical cluster analysis (complete linkage method)based on Euclidean distance was performed on the trans-formed data. Hierarchical and principal component analyseswere performed using the statistical package EINSIGHT(Infometrix).

Acknowledgements

The authors thank Barry Kinsall of ORNL, and Mark Henryof the NCIBRD, University of f\/lichigan, for valuable-assistance in field sampling. This work was supportedby a grant from the Strategic Environmental Researchand Development Program (SERDP contract number11XSY887Y).

References

Amann, R.I., Ludwig, W., and Schleifer, K.-H. (1995) Phylo-genetic identification and in situ detection of individualmicrobial cells without cultivation. Microbiol Rev 59: 143-169.

Bakken, L.R. (1985) Separation and purification of bacteriafrom soil. AppI Environ Microbiol 4^9: 1188-1195.

Balkwill, D.L, Leach. F.R.. Wilson, J.T., McNabb. J.F,, andWhite, D.C. (1988) Equivalence of microbial biomassmeasures based on membrane lipid and cell wall compo-nents, adenosine triphosphate, and direct counts in sub-surface sediments. Microbial Ecol 16; 73-84.

Borneman, J., Skroch, P.W., O'Sullivan, K.M., Palus, J.A.,Rumjanek, N.G., Jansen, J.L., et al. (1996) Molecularmicrobial diversity of an agricultural soil in Wisconsin.AppI Environ Microbioi %2: 1935-1943.

Brosius, J., Dull, T.L, Sleeter, D.D., and Noller. H.F. (1981)Gene organisation and primary structure of the ribosomalRNA operon in Escherichia coli. J Mol Biol 148: 107-127.

Cox, E.E., Major, D.W., Acton, D.W., Phelps, T.J., and White, D.C.(1994) Evaluating trichloroethylene biodegradation by

1999 Blackwell Science LXd, Environmental Microbiology, 1, 231-241

240 J. R. Stephen et al.

measuring the in situ status and activities of microbialpopulations. In Bioremediation of Chlorinated PolycyclicAromatic Compounds. Hinchee, R.E., Leeson, A,,Semprini, L., and Ong. S.K. (eds). Ann Arbor: Lewis Pub-lications, pp. 37-49.

Doi, Y. (1990) Microbial Polyesters. New York: VCH Publish-ers, pp. 1-8.

Dowling, N.J.E., Widdel, F.. and White, D.C. (1986) Phospho-lipid ester-linked fatty acid biomarkers of acetate-oxidizingsulphate-reducers and other sulphide-forming bacteria.J Gen Microbion32: 1815-1825.

Downing, J.A. (1979) Aggregation, transformation and thedesign of benthic sampling programs. J Fish Res BoardCan 36: 1454-1463.

Ediund, A., Nichols, P.D., Roffey, R., and White, D.C. (1985)Extractable and lipopolysaccharide fatty acid and hydroxyacid profiles from Desulfovobrio species. J Lipid Res 26:982-988.

Fang, J., Barcelona, M.J., and West, C. (1997) The use ofaromatic acids and phospholipid ester linked fatty acidsfor delineation of processes affecting an aquifer contami-nated with JP-4 fuel. In Molecular Markers in Environmen-tal Geochemistry. Egenhouse, R.E. (ed,). AmericanChemical Society Symposium 671. Washington DC:American Chemical Society, pp. 65-76.

Findlay, R.H., and Dobbs, F.C. (1993) Quantitative descrip-tion of microbial communities using lipid analysis. In Hand-book of Methods in Aquatic Microbial Ecology. Kemp, P.F.,Sherr, B.F., Sherr, E.B., and Cole, J.J. (eds), Boca Raton,FL: Lewis Publishers, pp. 271-284.

Findlay, R.H., and White, D.C. (1983) Polymeric beta hydro-xyalkanoates from environmental samples and Bacillusmegatarium. AppI Environ Microbiol 45: 71-78.

Frostegard, A., Tunlid, A., and Baath, E. (1993) Phospholipidfatty acid composition, biomass, and activity of microbialcommunities from two soil types experimentally exposedto different heavy metals. AppI Environ Microbiol 59:3605-3617.

Frostegard, A,, Tunlid, A., and Baath, E. (1996) Changes inmicrobial community structure during long-term incubationin two soils experimentally contaminated with metals. SoilBiol Biochem 28: 55-63.

Guckert, J.B., Antworth, C.P., Nichols, P.D., and White, D.C.(1985) Phospholipid ester-linked fatty acid profiles asreproducible assays for changes in prokaryotic communitystructure of estuarine sediments. FEMS Microbiol Ecol 31:147-158.

Guckert, J.B., Hood, M.A., and White, DC. (1986) Phospho-lipid, ester-linked fatty acid profile changes during nutrientdepletion of Vibrio cholerae: increases in the trans/cis andproportions of cyclopropyl fatty acids. AppI Environ Micro-biol 52: 794-801.

Heipieper, H.-J., Diefenbach, R., and Kewoloh, H. (1992)Conversion of cis unsaturated fatty adds to trans, a possi-ble mechanism for the protection of phenol-degradingPseudomonas putida P8 from substrate toxicity. AppIEnviron Microbiol 58: 1847-1852.

Heuer, H., and Smalla, K. (1997) Application of denaturinggel electrophoresis and temperature gradient gel electro-phoresis for studying soil microbial communities. InModern Soil Microbiology. Van Elsas, J.D., Trevors. J.T.,

and Wellington, E.M.H. (eds). New York: Marcel Dekker,pp. 353-374.

Kates, M, (1986) Techniques in Lipidology: Isolation, Analy-sis and Identification of Lipids. 2nd edn. Amsterdam:Elsevier Press.

Keift, T.L., Ringelberg, D.B., and White, D.C. (1994)Changes in ester-linked phospholipid fatty acid profilesof subsurface bacteria during starvation and desiccationof a porous medium. AppI Environ Microbiol 60: 3292-3299.

Lane, D., Pace, B., Olsen, G.J., Stahl, D.A., Sogin, M.L, andPace, NR. (1985) Rapid determination of 16S ribosomalRNA sequences for phylogenetic analyses. Proc NatIAcad Sci USA 82: 6955-6959.

Maidak, B.L,, Olsen, G.J., Larsen, N., Overbeek, R.,McCaughey, M.J., and Woese, C.R. (1997) The RDP(Ribosoma! Database Project). Nucleic Acids Res 25:109-111.

Mayberry, W.R., and Lane, J.R. (1993) Sequential alkalinesaponificatlon/acid hydrolysis/esterification: a one tubemethod with enhanced recovery of both cyclopropaneand hydroxylated fatty acids, J Microbiol Methods 18:21-32,

Muyzer, G., de Waal, E.C., and Uitterlinden, A.G. (1993) Pro-filing of microbial populations by denaturing gradient gelelectrophoresis analysis of polymerase chain reactionamplified genes coding for 16S rRNA, AppI Environ Micro-biol 59: 695-700.

Nichols, D.S., Nichols, P.D,, Russell, N,J., Davies, N,W,, andMcMeekin, T.A. (1997) Polyunsaturated fatty acids in thepsychrophilic bacterium Shewanella gelidimarina ACAM456T: molecular species analysis of major phospholipidsand biosythesis of eicosapentaenoic acid. Biochim BiophysActan47: 164-176.

Pfiffner, S.M., Palumbo, A.V., Gibon, T., Ringelberg, D.8.,and McCarthy, J.F. (1997) Relating groundwater andsediment chemistry to microbial Characterisation at aBTEX-contaminated site. Appt Biochem Biotech 63-65:775-788,

Pinkart, H.C., Wolfram, J.W., Rogers, R., and White, D.C.(1996) Cell envelope changes in solvent-tolerant and sol-vent-sensitive Pseudomonas puitida strains followingexposure to O-xylene. AppI Environ Microbiol 62: 1127-1131.

Rand, G,M., and Petrocelli, S.R. (1985) Fundamentals ofAquatic Toxicology: Methods and Applications. New York:Hemisphere Publishing Coiporation.

Ringelberg, D.B., Davis. J.D., Smith, G.A., Pfiffner, S.M.,Nichols, P.D., Nickels, J.B., et al (1989) Validation of sig-nature polarlipid fatty acid biomarkers for alkane-utilizingbacteria in soils and subsurface aquifer materials, FEMSMicrobiol Ecol S2: 39-50.

Ringeiberg, D.B., Townsend, G.T., DeWeerd, K.A.,Sulita, J.M., and White, D.C. (1994) Detection of theanaerobic dechlorinating microorganism Desulfomoniletiedjei in environmental matrices by its signature lipopoty-saccharide branch-long-chain hydroxy fatty acids. FEMSMicrobiol Ecol 14: 9-18.

Sikkema, J., deBont, J.A.M., and Poolman, B. (1995)Mechanisms of membrane toxicity of hydrocarbons. Micro-biol Rev 59: 201-222.

1999 Blackwell Science Ltd, Environmental Microbiology. 1, 231-241

Skinner, F.A.. Jones, P.C.T., and Mollison, J.E. (1952) Acomparison of a direct and a plate counting technique forthe quantitative estimation of soii microorganisms. J GenfVlicrobioi 6: 261-271.

Smith, G.A.. Nickels, J.S., Kerger, B,D,, Davis, J,D,,Coiiins, S.P., Wilson, J.T., etai. (1986) Quantitative charac-terisation of microbiai biomass and community structure insubsurface materiai: a prokaryotic consortium responsiveto organic contamination. Can J fvlicrobiol 32: 104-111.

Stephen, J.R,, Chang, Y.-J,, Macnaughton. S.J., Whitaker, S.L..Hicks, C, Leung, K.T., et al. (1999) Fate of a metal-resis-tant inoculum in contaminated and pristine soils assessedby denaturing gradient gel electrophoresis. Environ ToxicolChem 18 (in press),

Tunlid, A., and White, D.C. (1992) Biochemical analysis ofbiomass, community structure, nutritional status and meta-bolic activity of the microbial community in soil. In Soii Bio-chemistry, Vol, 7. Boltag, J.-M., and Stotzky, G. (eds). NewYork: Marcel Dekker, pp. 229-262.

Vestal, R.J.. and White, D.C. (1989) Lipid analysis in micro-bial ecology. Siosc/ence 39: 535-541.

Watanabe, K.. Yazawa, K., Kondo, K., and Kawagushi, A.(1997) Fatty acid synthesis of an eicosapentaenoic acid-producing bacterium: de novo synthesis, chain elongation,and desaturation systems, J Biochem 122: 467-473.

Weidemeier, T.H., Swanson. M.A., Wilson, J,T,, Kampbell. D,H.,fuliller, R,N., and Hansen, J.E. (1995) Patterns of intrinsicbioremediation at two US Airforce bases. In IntrinsicBioremediation. Hinchee, R.H,, Wilson, J.T., andDowney. D.C, (eds). Columbus: Batelle Press, pp. 31-51.

White, D,C. (1983) Analysis of microorganisms in terms ofquantity and activity in natural environments. In Societyfor General fVlicrobiology, Vol, 34. f\/licrobes in their

Community characterizarion by Iipids/PCR~DGGE 241

Natural Environments. Slater, J,H., Whittenbury, R.,and Wimpenny, J.W.T, (eds). London: Cambridge Uni-versity Press, pp. 37-66.

White, D.C, Davis, W.fvl,, Nickels. J.S., King, J.D., andBobbie, R.J. (1979) Determination of the sedimentarymicrobial biomass by extractable lipid phosphate, Oecolo-giaAO: 51-62,

White, D.C, Ringeiberg, D,B., Hedrtck, D.B., and Nivens. D.E.(1993) Rapid identification of microbes from clinical andenvironmental matrices by mass spectrometry. In Identifi-cation of Microorganisms by Mass Spectrometry. Fenselau,C.(ed.). American Chemical Society Symposium series 541.Washington, DC: American Chemical Society, pp. 8-17.

White, D.C, Pinkart, H.C, and Ringeiberg, D.B, (1996a) Bio-mass measurements: biochemical approaches. In Manualof Environmental Microbiology. 1st edn. Hurst, C.J.,Knudsen, G.R,, Mclnerney, fvl.J,. Stetzenbach, LD., andWalter, M.V. (eds), Washington DC: American Societyfor Microbiology Press, pp. 91-101,

White, D.C, Stair, J.O., and Ringeiberg, D.B. (1996b) Quan-titative comparisons of in situ microbial biodiversity by sig-nature biomarker analysis. J Ind Microbioi 17: 185-196.

White, D.C, Flemming, C.A., Leung, K.T,, and Macnaughton,S.J, (1998) In situ microbial ecology for quantitative apprai-sal, monitoring, and risk assessment of pollution remedia-tion in soils, the subsurface, the rhizosphere, and inbiofilms. J PJIicrob Methods 32: 93-105.

Wilkinson, S.G. (1988) Gram-negative bacteria. In MicrobialLipids. Ratledge, C, and Wilkinson, S.G, (eds), London:Academic Press, pp, 299-488.

Zelles, L. (1997) Phospholipid fatty acid profiles in selectedmembers of soil microbial communities Chemosphere 35:275-294.

© 1999 Blackwell Science Ltd, Environmental Microbiology, 1, 231-241