Vol. 59, No. 11APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Nov. 1993, p. 3605-36170099-2240/93/113605-13$02.00/0Copyright X 1993, American Society for Microbiology

Phospholipid Fatty Acid Composition, Biomass, and Activityof Microbial Communities from Two Soil Types

Experimentally Exposed to Different Heavy MetalsA. FROSTEGARD,* A. TUNLID, AND E. BAATH

Department ofMicrobial Ecology, Lund University, Helgonavagen 5, S-223 62 Lund, Sweden

Received 26 April 1993/Accepted 23 August 1993

The phospholipid fatty acid (PLFA) pattern was analyzed in a forest humus and in an arable soilexperimentally polluted with Cd, Cu, Ni, Pb, or Zn at different concentrations. In both soil types, there weregradual changes in the PLFA patterns for the different levels of metal contamination. The changes in the forestsoil were similar irrespective of which metal was used, while in the arable soil the changes due to Cucontamination differed from those due to the other metals. Several PLFAs reacted similarly to the metalamendments in the two soil types, while others showed different responses. In both soils, the metal pollutionresulted in a decrease in the iso-branched PLFAs i15:0 and i17:0 and in the monounsaturated 16:1w5 and16:1w7c fatty acids, while increases were found for i16:0, the branched brl7:0 and brl8:0, and thecyclopropane cyl7:0 fatty acids. In the forest soil, the methyl branched PLFAs 1OMel6:0, 1OMel7:0, and1OMel8:0 increased in metal-polluted soils, indicating an increase in actinomycetes, while in the arable soil adecrease was found for 1OMel6:0 and 1OMel8:0 in response to most metals. The bacterial PLFAs 15:0 and 17:0increased in all metal-contaminated samples in the arable soil, while they were unaffected in the forest soil.Fatty acid 18:2w6, which is considered to be predominantly of fungal origin, increased in the arable soil, exceptin the Cu-amended samples, in which it decreased instead. Effects on the PLFA patterns were found at levelsof metal contamination similar to or lower than those at which effects on ATP content, soil respiration, or totalamount of PLFAs had occurred.

Numerous studies have demonstrated the adverse effect ofdifferent heavy metals on soil microbial biomass and activity(3, 13, 17, 40). One of the purposes of these studies has oftenbeen to determine at what level of metal contamination aneffect on the soil microbiota can be detected. However, it isseldom known whether the observed effects are due tochanges in species composition or to reduced physiologicalcapacities of the microbial community. In soil from a fieldexperiment in which plots had been treated with sewagesludge containing different levels of heavy metals, Chanderand Brookes (9, 10) found that the conversion of substrate Cto biomass C was less efficient in high-metal soils. Theysuggested that this could be due to differences either in thesize of the biomass or in the microbial community structure.In a later experiment, the same authors eliminated the effectof different sizes of the initial biomasses by repeated fumi-gations but still found a lower conversion of substrate C intobiomass in metal-amended soil (11).

Several studies that used plate count techniques havedemonstrated a shift in the composition of fungal speciestowards a more metal-tolerant community in metal-contam-inated soils (2, 24, 45). There have also been reports ofeffects on the bacterial community composition, generallyshowing an increase in gram-negative bacteria in metal-contaminated soils (8, 14). Investigations which involve thecultivation of microorganisms on agar plates have the draw-back that only a minor part of the community is studied,since the majority of soil microorganisms are nonculturablewith the techniques known today (7). It is therefore impor-tant to improve the methods used so that as large a propor-tion of the community as possible can be studied. One way

* Corresponding author.

to examine the entire microbial community structure is toanalyze the phospholipid fatty acid (PLFA) composition ofthe soil, since different subsets of a community have differ-ent PLFA patterns (39). It is usually not possible to detectindividual strains or species of microorganisms with thismethod, but changes in the overall composition of thecommunity can be detected instead. This will give moreinformation than methods such as plate counts, since theentire community may respond in a different way to metalstress than the culturable part would.

In environmental studies, PLFA analyses have hithertomostly been used to describe microbial communities inseawater and lake water and in studies of biofilms andsediments (39). Only a few investigations have used thisapproach to detect changes in the community structure dueto different environmental disturbances, especially in soil.Smith et al. (36) found a shift towards a more gram-negativebacterial community in a subsurface sediment contaminatedwith creosote wastes. An increase in gram-negative bacteria,as well as in actinomycetes, due to liming, ash fertilization,and alkaline deposition was indicated by altered PLFApatterns of different coniferous forest soils (4, 20). Not manyinvestigations have used PLFA profiles as a means to detectchanges in the soil microbial community structure caused bymetals. Ohya et al. (32) reported on a change in PLFAcomposition of bacteria isolated from soil after amendmentwith Zn, but to our knowledge no studies of effects of heavymetals have been carried out with patterns of PLFAs ex-tracted directly from soil.

In the present study, two soil types were contaminatedwith Cd, Cu, Ni, Pb, or Zn at different concentrations in along-term laboratory experiment. The PLFA patterns weredetermined after 6 months in order to ascertain whetherchanges due to metal contamination could be found, whether

3605

on Septem

ber 15, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

3606 FROSTEGARD ET AL.

TABLE 1. Microbial biomass measured as total amounts ofL-PO4, totPLFA, and ATP content and microbial respiration

in non-metal-amended soil samples after 6 months ofincubation at room temperature, expressed as

means ± standard error

L-P04 totPLFA ATP RespirationSample (ILg Of C02(nmol of P g1) (nmol g1) (ATgg1s)g-1)

Forest soil 1,885 ± 55 1,349 ± 99 4.02 ± 0.21 24.7 ± 0.2Arable soil 101 ± 5.0 65 ± 2.4 0.87 ± 0.07 NDa

I ND, not determined.

the changes were similar for different metals and for differentsoil types, and at what levels of contamination the effectscould be detected. The results were compared with thelevels at which changes in microbial biomass and activitycould be found.

MATERIALS AND METHODS

Soils and sampling sites. Two soils from the south ofSweden were used, a forest soil and an arable soil. The forestsoil was collected in a Scots pine forest (Pinus sylvestris L.).Samples were taken only from the AO/A02 layers. The soilhad a high organic-matter content (-80%) and a pH of -4.5.The arable soil was a sandy loam (4.4% organic matter, pH7.8). The ATP, lipid phosphate (L-PO4), and total phospho-lipid fatty acids (totPLFA) contents and the respiration ratefor the control samples after 6 months of incubation are

given in Table 1.After sampling, visible roots were removed, and the soils

were sieved (2.0-mm mesh size). Portions of forest soil (70 g,dry weight [dw]) or sandy loam (200 g, dw) were put inplastic jars. The metals were added to the soils at 10concentrations. Forest soil metal concentrations were asfollows: 0 to 64 mmol of Cd, 0 to 128 mmol of Cu, 0 to 128mmol of Ni, 0 to 256 mmol of Pb, and 0 to 256 mmol of Zn.Concentrations in arable soil were as follows: 0 to 32 mmolof Cd, 0 to 64 mmol of Cu, 0 to 64 mmol of Ni, 0 to 128 mmolof Pb, and 0 to 128 mmol of Zn. There were thus fivenon-metal-treated samples (controls) for each soil type. Allmetal concentrations are given per kilogram (dw) of soil. Cd,Cu, and Zn were added as solutions of sulfate salts; Ni andPb were added as solutions of nitrate salts. The jars weresealed with air-tight lids, and the samples were then incu-bated at room temperature (-22°C). The jars were aeratedregularly, and the moisture content was adjusted (whenneeded) to 65% (forest humus) and 6% (sandy loam) H20 g-1(wet weight) by adding distilled water. After 6 months, thePLFA composition of the different soil samples, as well asthe microbial biomass and activity of the soils, was deter-mined.

Lipid extraction and fractionation. All solvents and chem-icals used were of analytical grade. Fatty acid standardswere obtained from Supelco Inc. (Bellefonte, Pa.) and Lar-odan Fine Chemicals (Malmo, Sweden). Glassware waswashed in Deconex phosphate-free detergent (1SP), rinsed10 times with tap water, 5 times with distilled water, and 5times with deionized water, and heated at 400°C overnight.The extraction procedure followed that described by

Frostegard et al. (21). One sample from each jar was used.Briefly, 1.00-g (wet weight) portions of forest humus or

3.00-g (wet weight) portions of sandy loam were extracted ina one-phase mixture consisting of chloroform, methanol, and

citrate buffer (1:2:0.8, vol/vol/vol). After splitting the ex-tracts into two phases by adding chloroform and buffer, thelipid-containing phase was dried under a stream of nitrogenand stored at -20°C. The lipid material was fractionated oncolumns containing silicic acid into neutral and glyco- andphospholipid-containing polar lipids (21). The phospholipidfraction was dried under a stream of nitrogen and saved forpreparations of fatty acid methyl esters.

Mild alkaline methanolysis and GC analysis. The phospho-lipids were subjected to a mild alkaline methanolysis (16),and the resulting fatty acid methyl esters were separated ona Hewlett-Packard 5890 gas chromatograph (GC) equippedwith a flame ionization detector. The column used was a50-m HP5 capillary column (phenylmethyl silicone; Hewlett-Packard Co., Palo Alto, Calif.). Hydrogen was used as acarrier gas, and injections were made in a splitless mode.The temperature program was as follows: initial temperatureof 80°C for 1 min, increasing at 20°C min-1 to 160°C and thenincreasing at 5°C min-1 to the final temperature of 270°C,which was kept for 5 min. Relative retention times ofsupposed fatty acid methyl esters were compared with thoseof standards.Known amounts of methyl tridecanoate (13:0) and methyl

nonadecanoate (19:0) were added before methanolysis. ThetotPLFA was calculated with 19:0 as the internal standard.When the recovery of 13:0 was considerably lower than thatof 19:0, PLFAs i14:0 and 14:0 were excluded from principalcomponent analyses (PCA). This only occurred in the sam-ples from the arable soil contaminated with Cu.

Fatty acid nomenclature. Fatty acids are designated interms of the total number of carbon atoms:number of doublebonds, followed by the position of the double bond from themethyl end of the molecule. cis and trans configurations areindicated by c and t, respectively. The prefixes a and iindicate anteiso and iso branching, br indicates an unknownmethyl branching position, lOMe indicates a methyl groupon the 10th carbon atom from the carboxyl end of themolecule, and cy refers to cyclopropane fatty acids.

Identifications by GC-MS. Mass spectrometric (MS) anal-yses were carried out with a Hewlett-Packard 5970 GC/MSsystem. The GC conditions were those described above, buthelium was used as a carrier gas. Electron energy in electronimpact was 70 eV. Identifications of fatty acid methyl esterswere based on comparison with spectra that were eitherobtained from standards or reported in the literature (34).The positions of double bonds were determined by analyzingdimethyldisulfide adducts (30).

Determinations of biomass and activity. The microbialbiomass was determined by analysis of the ATP content (1),the totPLFA, and the L-PO4 content after digestion ofextracted lipids with persulfate (21). The respiration of theforest humus was measured by GC at 20°C (31). Respirationwas not measured in the sandy loam since the high pH of thesoil made the CO2 measurements unreliable.

Statistical analyses. Individual PLFAs (expressed as log10moles percent) were subjected to PCA to elucidate majorvariation and covariation patterns. To evaluate the metalconcentrations at which a change in PLFA patterns could beconsidered statistically significant, a SIMCA analysis wasperformed (43). A principal-component model using cross-validation (44) was created for the nonamended controlsamples, and each of the metal-amended samples was thentested for similarity with the model, using P < 0.05 as arejection criterion. To compare the toxicity of the differentmetals, a partial least-squares (PLS) model was created,using cross-validation (29) (pro-PLS in the SIRIUS com-

APPL. ENvIRON. MICROBIOL.

on Septem

ber 15, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

PFLA COMPOSITION OF HEAVY-METAL-CONTAMINATED SOILS 3607

puter program) for each one of the metal series. With thismodel, the apparent toxicity of each metal-amended samplecould be estimated. For example, using a model for Zn, theapparent toxicity for Cu-treated soils could be expressed asan equivalent Zn concentration. PLS was also used tomaximize the variation due to metal contamination in thefirst component. By then regressing the loading values forthe PLFAs along the first component for each of the metalsversus those of the other metals, the similarity between thechanges in the different PLFA patterns could be determined.The PLS plots were very similar to the PCA plots and arethus not shown. All multivariate calculations were per-formed with a computer program called SIRIUS (27).The levels of contamination at which 10 and 50% reduc-

tions (ED10 and ED50) in ATP and respiration could be foundwere calculated from the slope of the decreasing linear partof the dose-response curves.

RESULTS

PLFAs in uncontaminated soils. Both the forest soil and thesandy loam contained a variety of PLFAs composed ofsaturated, unsaturated, methyl-branched, and cyclopropanefatty acids. In the forest soil, 37 PLFAs with a chain lengthup to C20 were detected by GC-MS (Fig. 1). Most of themwere identified on the basis of comparison with mass spectraobtained from standards or previously reported in the liter-ature. A number of minor peaks, designated xl to x6, couldnot be completely identified but were regarded as fatty acidsbecause of the presence of several ions in their mass spectracharacteristic of methyl esters of fatty acids. The relativeretention times (to the internal standard 19:0) of the PLFAsare shown in Table 2.The composition of PLFAs differed in the two soil types.

The proportion, expressed in moles percent, of severalbranched and monounsaturated PLFAs was higher in thesandy loam soil than in the forest soil, while the proportionof the polyunsaturated 18:2o6 was higher in the forest soil(Tables 2 and 3). The methyl-branched fatty acid brl6:0(retention time relative to 19:0, 0.699) appeared to be uniqueto the arable soil and was not detected in the forest soil.Several of the minor unidentified PLFAs (xl to x6) found inthe forest soil were not detected, or were found in very lowamounts, in the sandy loam soil and were not included in thesubsequent data analysis of that soil type.PLFA patterns in polluted forest soil. The PCA of the forest

soil treated with the different metals showed that the PLFAcomposition of the soils changed after treatment with heavymetals (Fig. 2a to e). The nonamended controls and samplestreated with low doses of metals were found to the left in theplots, and high-dose samples were found to the right. Therewas a gradual change in PLFA pattern for each level of metalcontamination; this could be seen in either the first or thesecond principal component. Even the lowest-dose samplesappeared to differ from the control, although this was notstatistically significant (see below). Most of the variation inPLFA patterns was due to the metal contamination and wasexplained by the first principal component. This componentexplained 78.1, 66.0, 67.2, 58.9, and 74.4% of the variation inPLFA patterns for samples treated with Cd, Cu, Ni, Pb, andZn, respectively. The PLFA patterns also separated in thesecond component, which explained 9.3 to 23.0% of thevariation.The changes in PLFA patterns in the forest soil were

similar irrespective of which heavy metal was added. Thiscould be shown by using PLS analysis to maximize the

effects of the metals on the PLFA pattern in the firstcomponent. The loading values of the individual PLFAsalong this first component for each of the different metalswere then regressed versus those from the other metals,showing positive relationships (r2 = 0.628 to 0.828 [P <0.001], except for Pb, which had somewhat lower values,between 0.542 and 0.750).The PLFA changes are exemplified by the PCA loading

plot for samples treated with Zn (Fig. 3). Specific identifiedPLFAs, including the branched brl7:0 and brl8:0, the me-thyl-branched 1OMel6:0, 1OMel7:0, and 10Mel8:0, the iso-branched i16:0, and the monounsaturated 16:li7t and 19:1b(with unknown double-bond position), as well as the uniden-tified PLFAs xl, x2, and x5, increased in most cases at highdoses of all metals. These PLFAs were found to the right inthe plot (Fig. 3). The cyclopropane fatty acid cyl7:0 alsoincreased in response to all metal treatments except Cu. Themonounsaturated PLFAs 16:lX9, 16:lw7c, 16:1w5, and 19:la, as well as the branched i14:0, i15:0, a15:0, and i17:0, alldecreased in the metal-amended samples and were thusfound to the left in the plot (Fig. 3). The largest decrease wasfound for the unidentified PLFAsx4 andx6 (not shown in theplot). Palmitic acid (16:0) was in all cases found very closeto, or on, the origin in the PCA plots, indicating that thisPLFA was not affected by metal pollution. In most cases,this was also true for linoleic acid (18:2X6).

If the ratios between the amounts of PLFAs (in molespercent) in the highest-dose treatment and the controls arecalculated, the magnitudes of changes in individual PLFAsdue to the amendments of different heavy metals can becompared (Table 2). This ratio is subsequently termed theheavy metal effect ratio. One must bear in mind that thetoxicity of the highest dose of a certain metal is not neces-sarily similar to that of the other metals. Thus, the ratios fordifferent metals are not directly comparable. The largestincrease in the forest soil treated with heavy metals wasfound for brl7:0, with ratios up to 4.42. Ratios between 1.3and 2.5 were found in most cases for xl, i16:0, 16:lo7t,1OMel6:0, x2, brl8:0, 1OMel7:0, 1OMel8:0, x5, and 20:0.Exceptions were found for the Ni- and Pb-amended samples,in which ratios for l0Mel6:0 (Ni sample) and i16:0 andl0Mel8:0 (Ni and Pb) were between 1.4 and 1.7 in thesecond-highest-dose treatments (not shown) but then de-creased to near control levels in the highest-dose treatments.The greatest decrease was found for the unidentified PLFAx6, which decreased from 3.00 mol% in the controls toalmost undetectable levels (ratio, 0.06) in the high-metaltreatments. The ratios for i14:0, i15:0, a15:0, 16:1w9, 16:lw7c, 16:1w5, i17:0, x4, and 19:1a were in most cases lowerthan 0.7 and in several cases lower than 0.5 (Table 2).The amount (moles percent) of some of the PLFAs that

changed most in relative abundance in the forest soilamended with Zn was plotted against the levels of contam-ination (Fig. 4a and b). For most of these PLFAs, theincrease or decrease in moles percent started at Zn levelsaround 4 to 8 mmol kg (dw) of soil-', and then there werecontinuous gradual changes up to 128 or 256 mmol kg-'.However, more sudden changes were found for 16:1w5 (Fig.4b), which decreased from 1.5 mol% in the 4-mmol kg-'treatment to 0.6 mol% in the 16-mmol kg-' treatment, andfor x6, which decreased from ca. 2.3 mol% in the 16-mmolkg-' treatment to 0.7 mol% in the 64-mmol kg-' treatment(Fig. 4a). A few PLFAs, e.g., cyl7:0 (Fig. 4a) and 18:2X6and 18:1X9 (not shown), increased in moderately contami-nated samples and then decreased in the more heavilypolluted ones. The reverse was found for 18:1X7 (not

VOL. 59, 1993

on Septem

ber 15, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

9c

£

'00A

CO

3608

on Septem

ber 15, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

PFLA COMPOSITION OF HEAVY-METAL-CONTAMINATED SOILS

TABLE 2. PLFAs in forest soilsa

Peak Proportion of PLFA Ratio with given metalPeak PLFA RRT (mol%) inno. unamended soil Cd Cu Ni Pb Zn

1 i14:0 0.586 0.55 ± 0.10 0.25 0.49 0.51 0.67 0.612 14:0 0.614 1.82 ± 0.12 0.84 0.90 0.90 1.10 1.223 ilS:0 0.664 9.46 ± 0.47 0.74 0.34 0.57 0.53 0.544 a15:0 0.670 3.33 + 0.21 0.64 0.45 0.59 0.66 0.495 xlb 0.676 1.27 + 0.08 1.43 0.81 2.15 1.28 2.226 15:0 0.692 1.31 + 0.10 1.20 1.17 1.01 1.39 1.117 i16:1 0.728 0.80 + 0.08 1.12 0.97 2.05 2.23 1.158 i16:0 0.742 4.40 ± 0.04 2.00 2.30 0.82 0.79 2.179 16:1X9 0.753 0.92 ± 0.14 0.93 0.64 0.64 1.10 0.5010 16:1w7c 0.757 4.57 ± 0.14 0.65 0.51 0.84 0.88 0.6211 16:1b7t 0.760 0.61 + 0.03 2.31 1.23 2.42 1.68 1.6512 16:1X5 0.764 1.92 ± 0.11 0.38 0.38 0.58 0.49 0.4013 16:0 0.771 14.72 + 0.24 0.99 1.02 0.99 1.16 0.9314 brl7:0 0.775 0.19 ± 0.03 4.42 NIC 2.67 1.51 3.1415 lOMel6:0 0.805 7.07 ± 0.14 1.53 2.18 1.12 1.31 1.7616 i17:0 0.821 1.67 + 0.03 0.86 0.51 0.45 0.46 0.4117 a17:0 0.828 1.45 + 0.05 1.05 1.22 0.89 0.85 0.8318 17:1w8 0.833 0.83 ± 0.08 0.89 0.73 1.71 1.49 0.6119 x2b 0.836 1.04 ± 0.08 1.60 0.83 2.15 1.17 2.2220 cy17:0 0.841 1.10 ± 0.03 1.76 0.68 1.38 1.13 0.9821 17:0 0.849 0.99 ± 0.02 0.94 0.83 0.89 0.88 0.9122 brl8:0 0.852 0.76 ± 0.05 1.90 1.93 1.75 1.74 1.2223 lOMel7:0 0.881 0.96 ± 0.03 1.71 2.54 1.55 1.80 2.0824 x3b 0.894 0.10 ± 0.04 1.26 1.26 1.67 1.30 1.3225 x4b 0.897 1.65 ± 0.25 0.50 0.35 0.18 0.44 0.3726 18:2X6 0.905 4.65 ± 0.26 0.79 1.11 0.99 1.37 1.2127 18:1w9 0.909 10.12 ± 0.59 0.80 0.79 1.15 1.22 0.7428 18:1M 7 0.913 3.31 ± 0.03 0.89 1.11 1.21 1.09 0.9229 18:1 0.920 0.89 ± 0.04 0.94 1.00 1.12 1.00 0.8630 18:0 0.925 2.78 + 0.27 0.76 1.00 0.96 1.11 0.7831 19:1ad 0.932 0.54 + 0.02 0.93 0.38 0.71 0.62 0.3932 lOMel8:0 0.955 0.78 ± 0.02 1.32 1.50 1.05 1.04 1.4233 19:1bd 0.990 2.62 + 0.16 1.42 1.30 1.94 1.12 2.0834 cyl9:0 0.993 4.82 + 0.17 0.95 0.83 1.00 0.72 0.6035 xSb 1.020 0.59 + 0.05 1.41 2.05 1.89 1.25 2.0236 x6 1.030 3.00 ± 0.13 0.06 0.02 0.03 0.03 0.0437 20:0 1.072 1.60 + 0.10 1.32 1.97 1.60 1.18 1.69a PLFA peak numbers refer to peaks in the chromatogram in Fig. 1. The relative retention time (RRT) is calculated in relationship to the retention time of the

internal standard (19:0). In the unamended soil samples (controls), the proportions of the PLFAs, expressed as means ± standard error (n = 5), are given. Toshow the effects of different heavy metals, the ratios between the amounts of individual PLFAs found in the most metal-polluted samples and the means of thecontrol are given.

b xl through x6 designate unidentified fatty acid methyl esters.c ND, not determined.d 19:1a and 19:1b designate two different 19:1 with unknown double-bond position.

shown), which decreased in moderately contaminated sam-ples and then increased in high-metal samples. The twoPLFAs 16:1w7c and 16:1w7t showed contrasting responses.While 16:1h7c decreased throughout the metal series (Fig.4a), 16:1w7t increased at Zn concentrations above 4 mmol kgof soil-1 (Fig. 4b). Since the total amount of PLFAs changedvery little due to the metal contamination (see below), theuse of absolute values (nanomoles per gram of soil) wouldgive results similar to the data expressed in moles percent. Inthe other metal series, the changes in individual PLFAs overthe metal gradients were similar to those found for Zn (datanot shown).PLFA patterns in polluted arable soil. PCA of the arable

soil samples treated with the different metals separated thecontrols from the metal-treated samples. The control andlow-metal samples were found to the left and high-metalsamples were found to the right in the scores plots (Fig. 5a toe). The change in PLFA composition due to the metalcontamination was mostly shown by the first principal com-

ponent, which explained 51.1, 48.4, 48.7, 49.4, and 50.5% ofthe variation in PLFA patterns for the samples treated withCd, Cu, Ni, Pb, and Zn, respectively. The samples alsoseparated in the second principal component, which ex-plained 20.6 to 30.3% of the variation in PLFA patterns. Thelowest-dose treatments generally did not separate the metal-treated samples from the controls, but at higher metal levelsthere were gradual changes in PLFA patterns.The changes in PLFA patterns due to contamination of the

arable soil with Cd, Ni, Pb, and Zn were similar in mostcases. When the loading values of the individual PLFAsalong the first PLS component for each of the differentmetals were regressed versus those from the other metals,positive relationships were found (rd = 0.355 to 0.557 [P <0.001], except for Ni, which when regressed versus Cd hadr2 = 0.298 [P < 0.01] and when regressed versus Zn had r2 =0.123 [not significant]). The r2 values for these regressionsfor the arable soil were generally lower than those for theforest soil, which shows that the similarity between the

VOL. 59, 1993 3609

on Septem

ber 15, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

3610 FROSTEGARD ET AL.

TABLE 3. PLFAs in arable soila

Proportion of Ratio with given metalPLFAPLFA (mol%) in

unamended Cd Cu Ni Pb Znsoil

i14:0 0.99 ± 0.07 1.20 0.63 0.88 1.11 1.0814:0 1.21 ± 0.13 2.13 0.61 0.75 1.10 1.11i15:0 9.21 ± 0.34 0.71 0.98 0.68 0.63 0.74a15:0 6.04 ± 0.22 0.93 0.79 0.91 0.80 1.0515:0 0.93 + 0.12 3.42 0.90 1.00 1.47 1.23brl6:0 0.51 ± 0.02 0.88 0.90 0.94 0.55 1.16i16:1 0.57 ± 0.05 0.71 0.96 1.12 0.98 1.04i16:0 3.49 ± 0.05 0.85 0.83 1.14 0.98 1.0116:1h9 1.06 ± 0.21 0.73 0.82 1.52 1.90 1.5016:1w7c 7.19 ± 0.20 0.87 1.06 0.82 0.91 0.8816:1w7t 0.61 ± 0.01 0.69 1.05 0.85 0.75 0.7816:1X5 4.42 ± 0.11 0.77 1.15 0.79 0.93 0.8016:0 14.67 ± 0.39 1.22 1.02 1.03 0.93 0.96brl7:0 0.54 ± 0.02 0.96 1.42 0.62 1.52 1.38lOMel6:0 8.82 ± 0.13 0.67 1.00 0.74 0.67 0.75i17:0 2.58 ± 0.07 0.64 0.83 0.88 0.66 0.77a17:0 2.31 ± 0.07 0.79 0.86 1.08 0.87 1.1517:1w8 1.08 ± 0.01 1.35 0.94 1.34 1.50 1.46cy17:0 2.96 ± 0.07 0.85 1.49 1.05 1.01 1.2717:0 0.63 ± 0.04 1.21 1.25 1.74 1.44 1.29brl8:0 1.97 ± 0.06 0.88 1.02 1.24 0.92 1.07lOMel7:0 0.86 ± 0.02 0.76 0.81 1.06 1.00 0.9918:2w6 1.84 ± 0.20 2.52 0.78 1.96 4.13 2.5918:1w9 6.68 ± 0.32 1.20 0.90 1.15 1.10 1.0818:1M 7 8.38 ± 0.51 1.15 1.23 1.24 1.30 1.0318:1 1.17 ± 0.06 0.97 1.09 1.35 1.28 0.9518:0 2.33 ± 0.08 1.25 0.94 1.09 1.05 0.9519:1a 0.89 ± 0.05 0.81 1.41 1.13 1.07 1.77lOMel8:0 2.18 ± 0.06 0.55 0.70 0.99 0.62 0.81cyl9:0 3.42 ± 0.25 0.82 1.12 1.06 1.08 1.1020:0 0.49 ± 0.07 0.89 1.00 1.22 0.64 0.86

a In the unamended soil samples (controls), the proportions of the PLFAs,expressed as means ± standard error (n = 5), are given. To show the effectsof different heavy metals, the ratios between the amounts of individual PLFAsfound in the most metal-polluted samples and the means of the control aregiven.

changes in PLFA patterns between the different metal treat-ments was less evident for the arable soil than for the forestsoil.The PLFA changes are exemplified by the PCA loading

plot for samples treated with Zn (Fig. 6). In most cases, 15:0,brl7:0, 17:0, lOMel7:0, and 18:2X6 increased in the high-metal samples (to the right in the plot), while i15:0, 16:1w7c,16:1X5, lOMel6:0, and i17:0 decreased (to the left). For allmetals except Zn, 16:1o7t decreased in the metal-amendedtreatments. Saturated 16:0 was found on or very close to theorigin in the PCA plots for all metals.The heavy metal effect ratios for individual PLFAs are

given in Table 3. The ratios for 18:2X)6 were between 1.96and 4.13 in samples polluted with Cd, Ni, Pb, and Zn. Ratioshigher than 1.2 were found in most cases for PLFAs 15:0,16:1X9, brl7:0, 17:1X8, 17:0, and 18:1Xo7. For 15:0, brl7:0,cyl7:0, and 17:0, a higher ratio was found in almost all caseswhen the heavy metal ratio was calculated for the nexthighest level of contamination. lOMel7:0 had values similarto or lower than those for the controls. However, whenheavy metal effect ratios were calculated for the next highestlevel, the ratios (1.31 to 3.00) indicated an increase due to allmetals. Ratios lower than 0.8 were found in most cases fori15:0, 16:1h7t, 16:1w5, lOMel6:0, i17:0, and lOMel8:0.

The effect of Cu on the PLFA pattern in the arable soildiffered from that of the other metals (Fig. 7). This was alsoshown when the PLS loading values of the individual PLFAsfrom the first component were regressed versus those fromthe other metal treatments. No relationship was found (r2close to zero). The most striking difference in PLFA profilecompared with those with the other metals was found for18:2)6, which decreased with Cu contamination but in-creased in response to all other metals. In the sample withthe highest dose of Cu, 18:2X6 had a ratio of 0.78 comparedwith the controls (Table 3). The greatest decrease was foundfor 10Mel7:0 and 10Mel8:0 (Fig. 7), although the effect wasnot as strong in the sample with the highest dose of Cu, asseen from the heavy metal effect ratios given in Table 3. Thestrongest increase in the Cu-contaminated soil was found forcyl7:0 and brl7:0. These PLFAs also increased in responseto other metals (at least in the next highest doses of contam-ination; see above), but for cyl7:0 the increase due to Cuwas more pronounced than that due to the other metals(Table 3). PLFAs 16:1X5 and 16:17c increased in responseto Cu, while they generally decreased somewhat in thepresence of other metals. PLFAs that responded to Cusimilarly to other metals were 15:0, 17:0, 18:1X7, and 19:1a,which generally increased in moderately and/or highly con-taminated samples, and a15:0, brl6:0, i16:0, i17:0, and a17:0,which generally decreased in metal-treated soil.

Biomass and activity. The microbial biomass at differentlevels of metal contamination was measured as ATP content,L-PO4 content, and totPLFA in both soil types. The respi-ration at different metal concentrations was measured onlyin the forest soil. The values for the metal-contaminatedsamples were calculated as a percentage of the values in thecontrol samples. In both soil types, the effects of heavymetals, here exemplified by the Cd-amended samples (Fig.8a and b), on totPLFA and L-P04 content were minor evenat the highest doses, while there was a decrease in ATPcontent and respiration due to the metal contamination.By using the linear part of the respective curves, the ED50

and ED10 in ATP content, respiration, L-PO4 content, andtotPLFA could be calculated. In none of the cases did thetotPLFA or L-PO4 content decrease below 50% of thecontrol values, and for L-PO4 not even a 10% reduction wasfound. The ED10 values for ATP, respiration, and totPLFAwere compared with the doses at which changes in PLFApatterns were found to be statistically significant (P < 0.05),using SIMCA analysis (Tables 4 and 5). In the forest soil, thedoses at which changes in PLFA pattern occurred corre-sponded fairly well to the ED10 values for ATP content,respiration, and totPLFA for each individual metal, with theexception of Zn, for which no decrease in ATP could bedetected over the metal range used, and Pb, for which theED10 values of respiration and totPLFA were higher than thedose at which a change in PLFA pattern was found. In thearable soil, the changes in PLFA pattern were found at thesame or lower metal levels than the ED10 values for bothATP content and totPLFA for all metals except Zn.

DISCUSSION

The two soil types used were chosen so that they would becontrasting in pH and organic matter content. The un-amended samples of the two soil types also had somewhatdifferent PLFA profiles (Tables 2 and 3). This was expected,since it is well known that different soil types have differentmicrobial communities. Although differing in abundance,several PLFAs reacted similarly to the metal amendments in

APPL. ENvIRON. MICROBIOL.

on Septem

ber 15, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

PFLA COMPOSITION OF HEAVY-METAL-CONTAMINATED SOILS 3611

a b64

0

1 P11280.5

O. o

82

4

032

64

1 8

2 8 32

Cd 16Cu16

4

PC 1

PC 1 FIG _2.PCAshoingvaratin_n_PFA_ attrn_n_ aforst_ oi

d

C 0

128

64 2~~~~~~~~~~~~~~6

32

2

16 32

4 16

Pb

Ni8

8

PC 1

PC 1 FIG. 2. PCA showing variation in PLFA pattern in a forest soili due to heavy metal contamination. Scores are for samples contan-

e mated with the indicated levels: (a) 0, 0.125, 0.25, 0.5, 1, 2, 4, 8, 16,

32, and 64 mmol of Cd kg (dw) of soil-'; (b) 0 to 128 mmolof Cu kg0 26(dw) of soil-'; (c) 0 to 128 mmol of Ni kg (dw) of soil-'; (d) 0 to 256

mmol of Pb kg (dw) of soil-'; (e) 0 to 256 mmol of Zn kg (dw) of

soil'. The cross indicates origin of coordinates.

128

2

the two soil types. One example is provided by the iso-

4 branched PLFAs i15:0 and i17:0, which decreased in all

64 metal-amended soil samples (Fig. 3, 6, and 7; Tables 2 and

3). Some other PLFAs also reacted similarly to metal

8 amendment in the two soil types. For example, i16:0, br17:0,32 q~~~~~~y17:0, and br18:0 generally increased, while 16:1w7c and

~~~~~~16:1w5 decreased in the high-metal samples, although the

proportional decreases were less pronounced in the arable

soil (Tables 2 and 3). This might be due to the lower levels of

~~Zn metal contamination used in the latter soil type a-nd not

necessarily to different metal toxicities in the two soil types.

Furthermore, it must be pointed out that a PLFA that isfound in both soils might represent different groups of

PC 1 organisms, and thus changes in the abundance of a certain

VOL. 59, 1993

on Septem

ber 15, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

3612 FROSTEGARD ET AL.

115l:0 40° l0MelS:0

k il7:0 ciD;{i;;e-18:al :0

17:0 br 016:1, 7t brl7:0

17:1w8 20:0iS:1.,9 cy17:0 19:lb

xl

x2

PC 1FIG. 3. PCA showing loading values for individual PLFAs in a

forest soil contaminated with 0 to 256 mmol of Zn kg (dw) of soil-'.PLFAs found to the right in the plot had increased in the high-metalsamples, while PLFAs found to the left were more abundant in thecontrol and low-metal samples. PLFAs x4 and x6 were included inthe PCA, but they were excluded from the plot in order to be able toshow the positions of the other PLFAs more clearly. x4 andx6 werefound to the lower left outside the plot since they almost disap-peared in the metal-treated samples. The cross indicates origin ofcoordinates.

PLFA might represent changes in different organism groupsin the two soil types.A predominance of gram-negative over gram-positive bac-

teria is often found in metal-contaminated soils (13, 17). Inthis study, some evidence for a similar shift was indicated bya decrease in several iso- and anteiso-branched PLFAs, allof which are commonly found in gram-positive bacteria (33).A further indication that such a shift had also occurred in ourstudy was indicated by the increase in cyl7:0, which isconsidered to be typical for gram-negative bacteria (41), inthe metal-contaminated samples of both soil types (Fig. 3, 6,and 7; Tables 2 and 3). However, the interpretation must betentative since, e.g., i16:0 and i16:1 increased because ofsome metals and 18:1o)7, which is commonly found ingram-negative bacteria (41), showed variable responses.This indicates that the separation between gram-positive andgram-negative bacteria in soil by PLFA patterns is notstraightforward. This was also the case in a study of theeffect of liming on the PLFA pattern of soil microorganisms(20).

Several PLFAs, including the methyl-branched lOMel6:0,lOMel7:0, and lOMel8:0, reacted differently to the metals inthe arable soil compared with the forest soil. The tubercu-lostearic acid lOMel8:0 is found almost exclusively in acti-nomycetes (25, 28). lOMel6:0 and lOMel7:0 are also pro-duced by several actinomycete genera (26, 37). All threelOMe-branched PLFAs increased in the forest soil (Fig. 3;Table 2), and the results thus suggest an increase in thisgroup of organisms after metal contamination. The resultsfor the arable soil are more difficult to interpret since not all

a,

0 6:0

0~~~~~~~~~~~~~~H.

10~~~~~~~ *~~~~~ 165:1w7c

10OMe17:0

2

IL

0

0

0--10

o 1 2 4 8 1 6 32 64 128 256Amount of Zn (mmol)

FIG. 4. Amounts of representative PLFAs in forest soil samplescontaminated with 0 to 256 mmol of Zn kg (dw) of soil-l, expressedas moles percent totPLFA in each sample.

lOMe-branched PLFAs responded similarly, and in somecases they also responded in different ways to differentmetals (Fig. 6 and 7; Table 3). If lOMel8:0 is taken as anindicator of actinomycetes, the results indicated that thisgroup of organisms either decreased or was unaffected inresponse to metals in the arable soil, while the opposite wasfound in the forest soil. In a study by Hiroki (23), the degreeof tolerance in microorganisms isolated from a fallow paddyfield contaminated by Cd, Cu, apd Zn was suggested to befungi > bacteria > actinomycetes. Jordan and Lechevalier(24) reported that the Zn tolerance in actinomycetes isolatedfrom a forest close to a smelter increased compared withisolates from noncontaminated soil but that actinomyceteswere generally less tolerant than bacteria and fungi. Similarobservations were made for isolates from lead mine waste,

16:15 16:ho9

alS:0 19:lai14:0 iS:1Z7

16.lw7c18:114:0 lOMel6:01T:0 i16:0

il6:1 lOMel7:0

APPL. ENvIRON. MICROBIOL.

on Septem

ber 15, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

PFLA COMPOSITION OF HEAVY-METAL-CONTAMINATED SOILS 3613

a 16b320 is

0

32 64

0.06

0.25 20.50.12524

16

84 Cd 8 CPC 1

PC 1 FG5.CshwgvraoinPApteinaaalsi

C due to heavy metal cotamination. Scores aeforsamplescontam

0.50.25

0

32

12

8 0 4 160.05

Pb 3216 Ni

_ _ _ _ _ _ _ _

PC 1PC 1 FIG. 5. PCA showing variation in PLFA pattern in an arable soil

e due to heavy metal contamination. Scores are for samples contain-128 mated with the indicated levels: (a) 0, 0.06, 0.125, 0.25, 0.5, 1, 2, 4,

8, 16, and 32 mmol of Cd kg (dw) of soil'; (b) 0 to 64 mmol of Cukg (dw) of soil-; (c) 0 to 64 mmol of Ni kg (dw) of soil'; (d) 0 to 128msmol of Pb kg (dw) of soil-; (e) 0 to 128 mmol of Zn kg (dw) of

0.5 soil-'. The cross indicates origin of coordinates.32

although no indication of increased lead tolerance was found(42). This is contradictory to a study by Babich and Stotzky(5) in which monocultures of actinomycetes were found to be

2~~~~~~~~~lesssensitive to Cd than those of eubacteria. These results,as well as those from the present study, thus indicate thatdifferent types of actinomycetes can respond differently to

4 Zn metal contamination.Other PLEAs that showed different responses in the two

16 soil types were 15:0 and 17:0, which have been considered to8 be of predominantly bacterial origin (19, 38). They increased

in all metal-contaminated samples in the arable soil (Fig. 6and 7; Table 3), while they were found close to the origin in

PC 1 all PCA plots from the forest soil (Fig. 3; Table 2). Thissuggests that they were unaffected by the metal treatments inthe latter soil. 16:1w7t also showed different responses in the

VOL. 59, 1993

on Septem

ber 15, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

3614 FROSTEGARD ET AL.

PC1FIG. 6. PCA showing loading values for individual PLFAs in an

arable soil contaminated with 0 to 128 mmol of Zn kg (dw) of soil-'.PLFAs found to the right in the plot had increased in the high-metal-contaminated samples, while PLFAs found to the left were more

abundant in the control and low-metal samples. PLFA 16:1X9 wasnot included in the PCA since it coeluted with the following peak inthe GCs of several samples. The cross indicates origin of coordi-nates.

two soils. It decreased in response to all metals (except Cu;Fig. 7) in the arable soil (Fig. 6; Table 3), while it generallyincreased in the forest soil following metal amendment (Fig.3; Table 2).

It is generally agreed that fungi are less sensitive to metalpollution than bacteria (13). In the present study, this wouldbe the case in the arable soil, provided that PLFA 18:2w6 isregarded as a reliable indicator of fungal biomass. ThisPLFA increased with increasing contamination level for allmetals except Cu. The decrease in this PLFA in samplescontaminated with Cu could be explained by the fact that Cuis known to be toxic to fungi and is commonly used as afungicide. In the forest soil, on the other hand, 18:2w6seemed to be unaffected by the treatments. One explanationmight be that 18:2w6 is also found in plant residues, and ithas been reported to constitute a dominant fraction of thetotal fatty acids in Pinus sylvestris L. (35). The effect ofmetals on the portion of this PLFA that was derived fromfungi might therefore be masked by a large amount of 18:2w6derived from plant material.

Metal toxicity depends on the physicochemical propertiesof the soil (12, 15) and is usually stronger in sandy loam soilsthan in highly organic soils. However, in our study thereappeared to be little difference between metal toxicity in thetwo different soils, despite the difference in organic mattercontent (compare ED50 values for ATP and respiration forthe two soils; Tables 4 and 5). One reason for this could bethat the sandy loam used in the present investigation had a

pH of 7.8, compared with pH 4.5 in the forest soil. Severalstudies, primarily with laboratory media, have shown thatpH affects the toxicity of metals (6 and references cited

PC 1FIG. 7. PCA showing loading values for individual PLFAs in an

arable soil contaminated with 0 to 64 mmol of Cu kg (dw) of soil-'.PLFAs found to the right in the plot had increased at high levels ofmetal contamination, while PLFAs found to the left were more

abundant in the control and low-metal samples. The cross indicatesorigin of coordinates.

therein), and it is generally argued that metal toxicity islower at higher pH.Doelman (13) suggested that, for several processes that

follow a sigmoid dose-response relationship to a toxic sub-stance, a retardation of the entire process might be precededby a qualitative shift in the microbial community. In thepresent study, the levels found for detectable changes, givenas ED10, in ATP content, totPLFA, and respiration wererather similar for each of the metals and also similar to levelsat which changes in the PLFA pattern were evident bySIMCA analysis (Tables 4 and 5). However, the SIMCAanalysis, which was used to investigate the metal levels atwhich changes in PLFA patterns could be considered statis-tically significant, was much more rigorous than the calcu-lations to determine at what levels 10% decreases could befound in ATP, respiration, and totPLFA. In the latter case,because of inherent variation, a 10% decrease would be veryunlikely to be statistically significant. Also, for changes inthe PLFA pattern, no exact dose could be calculated foreach metal. Instead, the doses given are the lowest onesused in the experiment at which significant changes in thePLFA pattern could be found. The changes thus actuallyoccurred at a metal level somewhere between the one givenin Tables 4 and 5 and the one just below those used in theexperiment. Considering this, it is possible that the changesin PLFA patterns occurred at metal levels lower than thosewhich affected ATP content and respiration, and thus ourresults are not in conflict with the suggestion of Doelman (13)mentioned above. The effects of different heavy metals onthe PLFA patterns, ATP content, totPLFA, and respirationof the soils used in the present study were found at concen-trations corresponding to the lower levels of the ranges in thestudies summarized by Baath (3).The phospholipid content of both soils, measured either as

18:2w6

19:1a

17:lwg

a17:0 cyl7:0

13:1c09 brl7:0CY brl:brlS:0

10Me16:0 16:1c7c. - 15:0 15:1117:0 i190 a15:O 17:0 15:0

lOMel7:0

16:1.6 brl6:0 20:0ilS:0

lOMels:O 114:0

14:0

16:1l7t

brl8:0

17:1617:0

18:2w6 19:1alOMel7:0 15:1.6 1OMel6:0

18:1X7 brl7:0lOMelg:0 il7:0 cyl9:0 cyl7:0

16:lw7t16:1w7c

a17:0 20:018:0 18:1

16:0116:0 16:lcoS

116:1 15:0ils.-

brl6:0

a15:0

APPL. ENvIRON. MICROBIOL.

on Septem

ber 15, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

PFLA COMPOSITION OF HEAVY-METAL-CONTAMINATED SOILS 3615

a

100

80

1-o.-

0o 60

0

40-

20-

0.125 0.25 0.5 1 2 4 8 1 6 32 64Amount of Cd added (mmol)

100-

80

0O 600

40

20

00.0660.125 0.25 0.5 1 2 4 16 3 2

Amount of Cd added (mmol)FIG. 8. ATP content, respiration rate, totPLFA, and L-P04

content in (a) a forest soil contaminated with 0 to 64 mmol of Cd kg(dw) of soil-' and (b) an arable soil contaminated with 0 to 32 mmolof Cd kg (dw) of soil-'.

L-P04 or as totPLFAs, was less affected by heavy metalcontamination than was the ATP content (Fig. 8a and b).Earlier results from different uncontaminated soils indicateda linear relationship between the ATP and L-P04 content insoil (21). There was thus a different relationship between thetwo biomass measurements in uncontaminated soils com-pared with contaminated ones. This might indicate that theturnover rate of phospholipids in soil is low, especially intoxified soils. However, this seems unlikely, since an in-crease was found for several PLFAs at high metal concen-trations. At the same time, no increase was found in L-P04content or totPLFA, indicating that some phospholipids hadbeen metabolized in the soil. One explanation might be thatorganisms persist in highly toxic soils with intact membranes

TABLE 4. Levels of heavy metal contamination of forest soil atwhich changes in PLFA patterns were found as calculated bySIMCA analysis and levels at which an ED50 or ED10 inATP content, respiration, and totPLFA were found,expressed as millimoles per kilogram (dw) of soil

Metal PLFA ATP Respiration totPLFApattern ED50 ED10 ED50 ED10 ED10

Cd 1 25 1.2 14 0.2 3.9Cu 8 68 5.3 > 128 12 15Ni 2 15 1.6 6 1.5 1.5Pb 2 15 3.8 33 11 11Zn 4 >256 >256 24 2.3 >256

but in a resting state, with lower ATP content and activitythan in uncontaminated soil.To compare toxicity of the different metals in different

soils, one can compare the ED50 values calculated for ATPcontent and respiration rate. Thus, in the forest soil themetal toxicity measured on the basis of ATP and soilrespiration indicated a higher toxicity for Cd and Ni than forCu and Zn (Tables 4 and 5). This could also be done withseparate PLFAs. However, only one PLFA, namely, x6,decreased in abundance by more than 50%, and manyPLFAs increased (Table 2). Furthermore, using just onePLFA at a time would increase random variation, decreasingthe precision in the estimated ED50 value. Since the effectsof the metals were similar in the forest soil, one way ofovercoming this problem would be to use PLS analysis (29)to regress the PLFA pattern by a single metal. By using Znand a PLS model with one component, a linear relationshipbetween added and predicted metal concentrations wasfound (r2 = 0.991) (Fig. 9). The PLFA patterns for the othermetals were then used to predict apparent toxicity expressedas Zn concentration, here exemplified by Cd and Cu. BothCd and Cu showed a linear relationship between the addedamount of the metal and predicted toxicity as micromoles ofZn (r2 = 0.949 and 0.908 for Cd and Cu, respectively). ForCu, approximately the same toxicity as Zn was found on amolar basis; e.g., 32 mmol of added Zn was estimated to givethe same effect on the PLFA pattern as 38 mmol of Cu. Cdwas more toxic, and addition of 32 mmol of Zn was esti-mated to have the same toxicity as 11 mmol of Cd. The datafor Ni and Pb fitted the Zn model less well (calculations notshown). For Ni, the correlation was less good (r2 = 0.827),while for Pb the slope of the regression differed from that ofthe other metals. This might be due to Pb affecting the PLFApattern slightly differently than the other metals did, al-though this was not evident from the PCA. Nevertheless, an

TABLE 5. Levels of heavy metal contamination of a sandy loamat which changes in PLFA patterns were found as calculated by

SIMCA analysis and levels at which an ED50 or ED10 inATP content and totPLFA were found,

expressed as millimoles per kilogram (dw) of soil

PLFA ATP totPLFAMetal patternEDm0 ED10 ED10

Cd 1 6.3 1.2 3.2Cu 2 14 3.1 >64Ni 2 23 9.7 4.3Pb 8 28 11 7.3Zn 8 17 1.9 2.9

L-PO4totPLFA

i* ATP* resp

*- L-P04totPLFAATP

VOL. 59, 1993

on Septem

ber 15, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

3616 FROSTEGARD ET A.

0

E

E

E2-

o~~~~~~~~~~~~~~~~~~o

C

a #1 2 Cu

a

0

0.2

Log added amount of metal (mmol)

FIG. 9. PLS model, created using cross-validation, of PLFAprofiles from forest soil contaminated with 0 to 256 mmol of Zn kg(dw) of soil-'. The PLFA profiles for soil contaminated with Cd, Cu,and Zn were then used to predict the apparent toxicity expressed asZn concentration. A linear regression for each metal was calculatedby using only the data points with filled symbols (r2 = 0.949, r2 =0.908, and r2 = 0.991 for Cd, Cu, and Zn, respectively). Open circlesdenote Cd-polluted samples not included in the regression, opentriangles denote Cu-polluted samples not included in the regression,and open squares denote Zn-polluted samples not included in theregression.

addition of 32 mmol of Zn could be calculated to be equiv-alent to 13 mmol of Ni or 55 mmol of Pb. Thus, the toxicityof the different metals towards the PLFA pattern decreasedin the order Cd = Ni > Zn = Cu > Pb. This hierarchy oftoxicity is normally found (see reference 3). Similar resultswere found when PLS models for metals other than Zn were

used, although the lower toxicity of Pb was less evident,especially when the PLS model used was from Pb-contami-nated soils.

In the arable soil, the use of PLS regression was lessstraightforward, since the different metals affected the PLFApatterns differently than in the forest soil (calculations notshown). However, the overall picture indicated that Cd wasthe most toxic metal. This was also seen for the ATPmeasurements (Table 5).

It is well-known that metal treatments affect the pH ofsoils. In a study of the combined effects of storage and heavymetals on the PLFA patterns (22), using the same soils as inthe present investigation, the pH was measured after 2years. In the arable soil, the highest levels of metal additiondecreased pH (H20) by 0.4 (Ni), 0.6 (Cu and Zn), and 1.3 (Niand Pb) units. In the forest soil, the pH decreased somewhatmore. For the two highest levels of metal addition, the pHscompared with that of the controls decreased 1.0 and 1.2(Cd), 1.4 and 1.6 (Cu), 1.1 and 1.3 (Ni), 1.3 and 1.9 (Pb), and1.4 and 1.5 (Zn) units. However, at the levels at whichchanges in PLFA patterns were found (Tables 4 and 5), thepH decrease was never more than 0.2 units, and for Ni andPb, the pH at those levels of metal amendment was the sameas that for the controls. Nevertheless, pH is a confoundingfactor in elucidation of the effects of heavy metals.Although the effects of metals on the PLFA pattern were

generally similar, some PLFAs responded differently in thetwo soil types. Different metals could also affect a certainsoil in different ways, exemplified by Cu in the arable soil. Itis thus difficult to investigate metal-derived changes in a soilwithout knowing how that specific soil reacts to metalpollution. Therefore, laboratory experiments on soils withheavy metals must be regarded as a prerequisite to fieldstudies. Furthermore, it was difficult to elucidate changes inlarge groups such as gram-positive and gram-negative bac-teria, since several PLFAs specific for one group wereaffected differently by the heavy metal pollution. However,the analyses of PLFA patterns in soils still indicated changesin the microbial community, changes that could be quantifiedand compared, e.g., between different heavy metals by usingmultivariate statistics.

ACKNOWLEDGMENT

Financial support was obtained from the Swedish National Envi-ronment Protection Agency.

REFERENCES1. Arnebrant, K., and E. Baath. 1991. Measurements of ATP in

forest humus. Soil Biol. Biochem. 23:501-506.2. Arnebrant, K., E. Baath, and A. Nordgren. 1987. Copper

tolerance of microfungi isolated from polluted and unpollutedforest soil. Mycologia 79:890-895.

3. Baath, E. 1989. Effects of heavy metals in soil on microbialprocesses and populations (a review). Water Air Soil Pollut.47:335-379.

4. Baath, E., A. Frostegard, and H. Fritze. 1992. Soil bacterialbiomass, activity, phospholipid fatty acid pattern, and pHtolerance in an area polluted with alkaline dust deposition.Appl. Environ. Microbiol. 58:4026-4031.

5. Babich, H., and G. Stotzky. 1977. Sensitivity ofvarious bacteria,actinomycetes, and fungi to cadmium and the influence ofpH onsensitivity. Appl. Environ. Microbiol. 33:681-695.

6. Babich, H., and G. Stotzky. 1985. Heavy metal toxicity tomicrobe-mediated ecologic processes: a review and potentialapplication to regulatory policies. Environ. Res. 36:111-137.

7. Bakken, L. 1985. Separation and purification of bacteria fromsoil. Appl. Environ. Microbiol. 49:1482-1487.

8. Barkay, T., S. C. Tripp, and B. H. Olson. 1985. Effect ofmetal-rich sewage sludge application on the bacterial communi-ties of grasslands. Appl. Environ. Microbiol. 49:333-337.

9. Chander, K., and P. C. Brookes. 1991. Microbial biomassdynamics during the decomposition of glucose and maize inmetal-contaminated and non-contaminated soils. Soil Biol. Bio-chem. 23:917-925.

10. Chander, K., and P. C. Brookes. 1991. Plant inputs of carbon tometal-contaminated soil and effects on the soil microbial bio-mass. Soil Biol. Biochem. 23:1169-1177.

11. Chander, K., and P. C. Brookes. 1992. Synthesis of microbialbiomass from added glucose in metal-contaminated and non-contaminated soils following repeated fumigation. Soil Biol.Biochem. 24:613-614.

12. Collins, Y. E., and G. Stotzky. 1989. Factors affecting thetoxicity of heavy metals to microbes, p. 31-90. In T. J. Bever-idge and R. J. Doyle (ed.), Metal ions and bacteria. John Wiley& Sons, New York.

13. Doelman, P. 1985. Resistance of soil microbial communities toheavy metals, p. 369-384. In V. Jensen, A. Kjoller, and L. H.Sorensen (ed.), Microbial communities in soil. Elsevier, Lon-don.

14. Doelman, P., and L. Haanstra. 1979. Effects of lead on the soilbacterial microflora. Soil Biol. Biochem. 11:487-491.

15. Doelman, P., and L. Haanstra. 1986. Short- and long-termeffects of heavy metals on urease activity in soils. Biol. Fertil.Soils 2:213-218.

16. Dowling, N. J. E., F. Widdel, and D. C. White. 1986. Phospho-lipid ester-linked fatty acid biomarkers of acetate-oxidizing

APPL. ENvIRON. MICROBIOL.

on Septem

ber 15, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from

PFLA COMPOSITION OF HEAVY-METAL-CONTAMINATED SOILS 3617

sulphate-reducers and other sulphide-forming bacteria. J. Gen.Microbiol. 132:1815-1825.

17. Duxbury, T. 1985. Ecological aspects of heavy metal responsesin microorganisms, p. 185-236. In K. C. Marshall (ed.), Ad-vances in microbial ecology, vol. 8. Plenum Press, New York.

18. Duxbury, T., and B. Bicknell. 1983. Metal-tolerant bacterialpopulations from natural and metal-polluted soils. Soil Biol.Biochem. 15:243-250.

19. Federie, T. W. 1986. Microbial distribution in soil-new tech-niques, p. 493-498. In F. Megusar and M. Gantar (ed.), Per-spectives in microbial ecology. Slovene Society for Microbiol-ogy, Ljubljana, Slovenia.

20. Frostegard, A., E. Biath, and A. Tunlid. 1993. Shifts in thestructure of soil microbial communities in limed forests asrevealed by phospholipid fatty acid analysis. Soil Biol. Bio-chem. 25:723-730.

21. Frostegard, A., A. Tunlid, and E. Baath. 1991. Microbial bio-mass measured as total lipid phosphate in soils of differentorganic content. J. Microbiol. Methods 14:151-163.

22. Frostegard, A., A. Tunlid, and E. Biath. Unpublished data.23. HiroId, M. 1992. Effects of heavy metal contamination on soil

microbial population. Soil Sci. Plant Nutr. 38:141-147.24. Jordan, M. J., and M. P. Lechevalier. 1975. Effects of zinc-

smelter emissions on forest soil microflora. Can. J. Microbiol.21:1855-1865.

25. Kroppenstedt, R. M. 1985. Fatty acids and menaquinone anal-ysis of actinomycetes and related organisms, p. 173-199. In M.Goodfellow and D. E. Minnikin (ed.), Chemical methods inbacterial systematics. Academic Press Ltd., London.

26. Kroppenstedt, R M., and H. J. Kutzner. 1978. Biochemicaltaxonomy of some problem actinomycetes, p. 125-133. In M.Mordarski, W. Kurylowicz, and J. Jeljaszewicz (ed.), Proceed-ings of the International Symposium on Nocardia and Strepto-myces. Gustav Fischer Verlag, Stuttgart.

27. Kvalheim, 0. M., and T. V. Karstang. 1987. A general-purposeprogram for multivariate data analysis. Chemomet. IntelligentLab. Syst. 2:235-237.

28. Lechevalier, M. P. 1977. Lipids in bacterial taxonomy-a tax-onomist's view. Crit. Rev. Microbiol. 5:109-210.

29. Martens, M., and H. Martens. 1986. Partial least squares regres-sion, p. 1-92. In J. R. Piggott (ed.), Statistical procedures infood research. Elsevier Applied Science Publishers, London.

30. Nichols, P. D., J. B. Guckert, and D. C. White. 1986. Determi-nation of monounsaturated fatty acid double bond position andgeometry for microbial monocultures and complex consortia bycapillary GC-MS of their dimethyl disulphide adducts. J. Micro-biol. Methods 5:49-55.

31. Nordgren, A., T. Kauri, E. Baath, and B. Soderstrom. 1986. Soilmicrobial activity, mycelial lengths and physiological groups ofbacteria in a heavy metal polluted area. Environ. Pollut. 41:89-100.

32. Ohya, H., Y. Komai, and M. Yamaguchi. 1986. Zinc effects on asoil bacterial flora characterized by fatty acid composition of theisolates. Biol. Fertil. Soils 2:59-63.

33. O'Leary, W. M., and S. G. Wilkinson. 1988. Gram-positivebacteria, p. 117-185. In C. Ratledge and S. G. Wilkinson (ed.),Microbial lipids, vol. 1. Academic Press Ltd., London.

34. Ryhage, R., and E. Stenhagen. 1960. Mass spectrometry in lipidresearch. J. Lipid Res. 1:361-390.

35. Saranpil, P., and H. Nyberg. 1987. Lipids and sterols of Pinussylvestris L. sapwood and heartwood. Trees 1:82-87.

36. Smith, G. A., J. S. Nickels, B. D. Kerger, J. D. Davis, and S. P.Collins. 1986. Quantitative characterization of microbial bio-mass and community structure in subsurface material: a pro-karyotic consortium responsive to organic contamination. Can.J. Microbiol. 32:104-111.

37. Tisdall, P. A., G. D. Roberts, and J. P. Anhalt. 1979. Identifi-cation of clinical isolates of mycobacteria with gas-liquid chro-matography alone. J. Clin. Microbiol. 10:506-514.

38. Tunlid, A., H. A. J. Hoitink, C. Low, and D. C. White. 1989.Characterization of bacteria that suppress Rhizoctonia damp-ing-off in bark compost media by analysis of fatty acid biomar-kers. Appl. Environ. Microbiol. 55:1368-1374.

39. Tunlid, A., and D. C. White. 1992. Biochemical analysis ofbiomass, community structure, nutritional status, and metabolicactivity of microbial communities in soil, p. 229-262. In G.Stotzky and J.-M. Bollag (ed.), Soil biochemistry, vol. 7. MarcelDekker, New York.

40. Tyler, G. 1981. Heavy metals in soil biology and biochemistry,p. 371-414. In E. A. Paul and J. N. Ladd (ed.), Soil biochem-istry, vol. 5. Marcel Dekker, New York.

41. Willinson, S. G. 1988. Gram-negative bacteria, p. 299-457. InC. Ratledge and S. G. Wilkinson (ed.), Microbial lipids, vol. 1.Academic Press Ltd., London.

42. Williams, S. T., T. McNeilly, and E. M. Wellington. 1977. Thedecomposition of vegetation growing on metal mine waste. SoilBiol. Biochem. 9:271-275.

43. Wold, S. 1976. Pattern recognition by means of disjoint principalcomponents models. Pattern Recognition 8:127-139.

44. Wold, S. 1978. Cross-validatory estimation of the number ofcomponents in factor and principal component models. Tech-nometrics 20:397-405.

45. Yamamoto, H., T. Kadzunori, and T. Uchiwa. 1985. Fungal floraof soil polluted with copper. Soil Biol. Biochem. 17:785-790.

VOL. 59, 1993

on Septem

ber 15, 2020 by guesthttp://aem

.asm.org/

Dow

nloaded from