of Microbial Communities from Two Soil Types Experimentally Exposed to Different Heavy Metals

Transcription

1 APPLIED AND ENVIRONMENTAL MICROBIOLOGY, Nov. 1993, p /93/ $2./ Copyright X 1993, American Society for Microbiology Vol. 59, No. 11 Phospholipid Fatty Acid Composition, Biomass, and Activity of Microbial Communities from Two Soil Types Experimentally Exposed to Different Heavy Metals A. FROSTEGARD,* A. TUNLID, AND E. BAATH Department of Microbial Ecology, Lund University, Helgonavagen 5, S 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 soil experimentally polluted with Cd, Cu, Ni, Pb, or Zn at different concentrations. In both soil types, there were gradual changes in the PLFA patterns for the different levels of metal contamination. The changes in the forest soil were similar irrespective of which metal was used, while in the arable soil the changes due to Cu contamination differed from those due to the other metals. Several PLFAs reacted similarly to the metal amendments in the two soil types, while others showed different responses. In both soils, the metal pollution resulted in a decrease in the iso-branched PLFAs i15: and i17: and in the monounsaturated 16:1w5 and 16:1w7c fatty acids, while increases were found for i16:, the branched brl7: and brl8:, and the cyclopropane cyl7: fatty acids. In the forest soil, the methyl branched PLFAs 1OMel6:, 1OMel7:, and 1OMel8: increased in metal-polluted soils, indicating an increase in actinomycetes, while in the arable soil a decrease was found for 1OMel6: and 1OMel8: in response to most metals. The bacterial PLFAs 15: and 17: increased 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, except in the Cu-amended samples, in which it decreased instead. Effects on the PLFA patterns were found at levels of metal contamination similar to or lower than those at which effects on ATP content, soil respiration, or total amount of PLFAs had occurred. Numerous studies have demonstrated the adverse effect of different heavy metals on soil microbial biomass and activity (3, 13, 17, 4). One of the purposes of these studies has often been to determine at what level of metal contamination an effect on the soil microbiota can be detected. However, it is seldom known whether the observed effects are due to changes in species composition or to reduced physiological capacities of the microbial community. In soil from a field experiment in which plots had been treated with sewage sludge containing different levels of heavy metals, Chander and Brookes (9, 1) found that the conversion of substrate C to biomass C was less efficient in high-metal soils. They suggested that this could be due to differences either in the size of the biomass or in the microbial community structure. In a later experiment, the same authors eliminated the effect of different sizes of the initial biomasses by repeated fumigations but still found a lower conversion of substrate C into biomass in metal-amended soil (11). Several studies that used plate count techniques have demonstrated a shift in the composition of fungal species towards a more metal-tolerant community in metal-contaminated soils (2, 24, 45). There have also been reports of effects on the bacterial community composition, generally showing an increase in gram-negative bacteria in metalcontaminated soils (8, 14). Investigations which involve the cultivation of microorganisms on agar plates have the drawback that only a minor part of the community is studied, since the majority of soil microorganisms are nonculturable with the techniques known today (7). It is therefore important to improve the methods used so that as large a proportion of the community as possible can be studied. One way * Corresponding author. 365 to examine the entire microbial community structure is to analyze the phospholipid fatty acid (PLFA) composition of the soil, since different subsets of a community have different PLFA patterns (39). It is usually not possible to detect individual strains or species of microorganisms with this method, but changes in the overall composition of the community can be detected instead. This will give more information than methods such as plate counts, since the entire community may respond in a different way to metal stress than the culturable part would. In environmental studies, PLFA analyses have hitherto mostly been used to describe microbial communities in seawater and lake water and in studies of biofilms and sediments (39). Only a few investigations have used this approach to detect changes in the community structure due to different environmental disturbances, especially in soil. Smith et al. (36) found a shift towards a more gram-negative bacterial community in a subsurface sediment contaminated with 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 PLFA patterns of different coniferous forest soils (4, 2). Not many investigations have used PLFA profiles as a means to detect changes in the soil microbial community structure caused by metals. Ohya et al. (32) reported on a change in PLFA composition of bacteria isolated from soil after amendment with Zn, but to our knowledge no studies of effects of heavy metals have been carried out with patterns of PLFAs extracted directly from soil. In the present study, two soil types were contaminated with Cd, Cu, Ni, Pb, or Zn at different concentrations in a long-term laboratory experiment. The PLFA patterns were determined after 6 months in order to ascertain whether changes due to metal contamination could be found, whether

2 366 FROSTEGARD ET AL. TABLE 1. Microbial biomass measured as total amounts of L-PO4, totplfa, and ATP content and microbial respiration in non-metal-amended soil samples after 6 months of incubation at room temperature, expressed as means ± standard error Sample L-P4 totplfa ATP Respiration (ILg Of C2 (nmol of P g1) (nmol g1) (ATg1s) g-1) Forest soil 1,885 ± 55 1,349 ± ± ±.2 Arable soil 11 ± ± ±.7 NDa I ND, not determined. the changes were similar for different metals and for different soil types, and at what levels of contamination the effects could be detected. The results were compared with the levels at which changes in microbial biomass and activity could be found. MATERIALS AND METHODS Soils and sampling sites. Two soils from the south of Sweden were used, a forest soil and an arable soil. The forest soil was collected in a Scots pine forest (Pinus sylvestris L.). Samples were taken only from the AO/A2 layers. The soil had a high organic-matter content (-8%) and a ph of The arable soil was a sandy loam (4.4% organic matter, ph 7.8). The ATP, lipid phosphate (L-PO4), and total phospholipid fatty acids (totplfa) contents and the respiration rate for 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.-mm mesh size). Portions of forest soil (7 g, dry weight [dw]) or sandy loam (2 g, dw) were put in plastic jars. The metals were added to the soils at 1 concentrations. Forest soil metal concentrations were as follows: to 64 mmol of Cd, to 128 mmol of Cu, to 128 mmol of Ni, to 256 mmol of Pb, and to 256 mmol of Zn. Concentrations in arable soil were as follows: to 32 mmol of Cd, to 64 mmol of Cu, to 64 mmol of Ni, to 128 mmol of Pb, and to 128 mmol of Zn. There were thus five non-metal-treated samples (controls) for each soil type. All metal concentrations are given per kilogram (dw) of soil. Cd, Cu, and Zn were added as solutions of sulfate salts; Ni and Pb were added as solutions of nitrate salts. The jars were sealed with air-tight lids, and the samples were then incubated at room temperature (-22 C). The jars were aerated regularly, and the moisture content was adjusted (when needed) to 65% (forest humus) and 6% (sandy loam) H2 g-1 (wet weight) by adding distilled water. After 6 months, the PLFA composition of the different soil samples, as well as the microbial biomass and activity of the soils, was determined. Lipid extraction and fractionation. All solvents and chemicals used were of analytical grade. Fatty acid standards were obtained from Supelco Inc. (Bellefonte, Pa.) and Larodan Fine Chemicals (Malmo, Sweden). Glassware was washed in Deconex phosphate-free detergent (1SP), rinsed 1 times with tap water, 5 times with distilled water, and 5 times with deionized water, and heated at 4 C overnight. The extraction procedure followed that described by Frostegard et al. (21). One sample from each jar was used. Briefly, 1.-g (wet weight) portions of forest humus or 3.-g (wet weight) portions of sandy loam were extracted in a one-phase mixture consisting of chloroform, methanol, and APPL. ENvIRON. MICROBIOL. citrate buffer (1:2:.8, vol/vol/vol). After splitting the extracts into two phases by adding chloroform and buffer, the lipid-containing phase was dried under a stream of nitrogen and stored at -2 C. The lipid material was fractionated on columns containing silicic acid into neutral and glyco- and phospholipid-containing polar lipids (21). The phospholipid fraction was dried under a stream of nitrogen and saved for preparations of fatty acid methyl esters. Mild alkaline methanolysis and GC analysis. The phospholipids were subjected to a mild alkaline methanolysis (16), and the resulting fatty acid methyl esters were separated on a Hewlett-Packard 589 gas chromatograph (GC) equipped with a flame ionization detector. The column used was a 5-m HP5 capillary column (phenylmethyl silicone; Hewlett- Packard Co., Palo Alto, Calif.). Hydrogen was used as a carrier gas, and injections were made in a splitless mode. The temperature program was as follows: initial temperature of 8 C for 1 min, increasing at 2 C min-1 to 16 C and then increasing at 5 C min-1 to the final temperature of 27 C, which was kept for 5 min. Relative retention times of supposed fatty acid methyl esters were compared with those of standards. Known amounts of methyl tridecanoate (13:) and methyl nonadecanoate (19:) were added before methanolysis. The totplfa was calculated with 19: as the internal standard. When the recovery of 13: was considerably lower than that of 19:, PLFAs i14: and 14: were excluded from principal component analyses (PCA). This only occurred in the samples from the arable soil contaminated with Cu. Fatty acid nomenclature. Fatty acids are designated in terms of the total number of carbon atoms:number of double bonds, followed by the position of the double bond from the methyl end of the molecule. cis and trans configurations are indicated by c and t, respectively. The prefixes a and i indicate anteiso and iso branching, br indicates an unknown methyl branching position, lome indicates a methyl group on the 1th carbon atom from the carboxyl end of the molecule, and cy refers to cyclopropane fatty acids. Identifications by GC-MS. Mass spectrometric (MS) analyses were carried out with a Hewlett-Packard 597 GC/MS system. The GC conditions were those described above, but helium was used as a carrier gas. Electron energy in electron impact was 7 ev. Identifications of fatty acid methyl esters were based on comparison with spectra that were either obtained from standards or reported in the literature (34). The positions of double bonds were determined by analyzing dimethyldisulfide adducts (3). Determinations of biomass and activity. The microbial biomass was determined by analysis of the ATP content (1), the totplfa, and the L-PO4 content after digestion of extracted lipids with persulfate (21). The respiration of the forest humus was measured by GC at 2 C (31). Respiration was not measured in the sandy loam since the high ph of the soil made the CO2 measurements unreliable. Statistical analyses. Individual PLFAs (expressed as log1 moles percent) were subjected to PCA to elucidate major variation and covariation patterns. To evaluate the metal concentrations at which a change in PLFA patterns could be considered statistically significant, a SIMCA analysis was performed (43). A principal-component model using crossvalidation (44) was created for the nonamended control samples, and each of the metal-amended samples was then tested for similarity with the model, using P <.5 as a rejection criterion. To compare the toxicity of the different metals, a partial least-squares (PLS) model was created, using cross-validation (29) (pro-pls in the SIRIUS com-

3 VOL. 59, 1993 PFLA COMPOSITION OF HEAVY-METAL-CONTAMINATED SOILS 367 puter program) for each one of the metal series. With this model, the apparent toxicity of each metal-amended sample could be estimated. For example, using a model for Zn, the apparent toxicity for Cu-treated soils could be expressed as an equivalent Zn concentration. PLS was also used to maximize the variation due to metal contamination in the first component. By then regressing the loading values for the PLFAs along the first component for each of the metals versus those of the other metals, the similarity between the changes in the different PLFA patterns could be determined. The PLS plots were very similar to the PCA plots and are thus not shown. All multivariate calculations were performed with a computer program called SIRIUS (27). The levels of contamination at which 1 and 5% reductions (ED1 and ED5) in ATP and respiration could be found were calculated from the slope of the decreasing linear part of the dose-response curves. RESULTS PLFAs in uncontaminated soils. Both the forest soil and the sandy loam contained a variety of PLFAs composed of saturated, unsaturated, methyl-branched, and cyclopropane fatty acids. In the forest soil, 37 PLFAs with a chain length up to C2 were detected by GC-MS (Fig. 1). Most of them were identified on the basis of comparison with mass spectra obtained from standards or previously reported in the literature. A number of minor peaks, designated xl to x6, could not be completely identified but were regarded as fatty acids because of the presence of several ions in their mass spectra characteristic of methyl esters of fatty acids. The relative retention times (to the internal standard 19:) of the PLFAs are shown in Table 2. The composition of PLFAs differed in the two soil types. The proportion, expressed in moles percent, of several branched and monounsaturated PLFAs was higher in the sandy loam soil than in the forest soil, while the proportion of the polyunsaturated 18:2o6 was higher in the forest soil (Tables 2 and 3). The methyl-branched fatty acid brl6: (retention time relative to 19:,.699) appeared to be unique to the arable soil and was not detected in the forest soil. Several of the minor unidentified PLFAs (xl to x6) found in the forest soil were not detected, or were found in very low amounts, in the sandy loam soil and were not included in the subsequent 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 PLFA composition of the soils changed after treatment with heavy metals (Fig. 2a to e). The nonamended controls and samples treated with low doses of metals were found to the left in the plots, and high-dose samples were found to the right. There was a gradual change in PLFA pattern for each level of metal contamination; this could be seen in either the first or the second principal component. Even the lowest-dose samples appeared to differ from the control, although this was not statistically significant (see below). Most of the variation in PLFA patterns was due to the metal contamination and was explained by the first principal component. This component explained 78.1, 66., 67.2, 58.9, and 74.4% of the variation in PLFA patterns for samples treated with Cd, Cu, Ni, Pb, and Zn, respectively. The PLFA patterns also separated in the second component, which explained 9.3 to 23.% of the variation. The changes in PLFA patterns in the forest soil were similar irrespective of which heavy metal was added. This could be shown by using PLS analysis to maximize the effects of the metals on the PLFA pattern in the first component. The loading values of the individual PLFAs along this first component for each of the different metals were then regressed versus those from the other metals, showing positive relationships (r2 =.628 to.828 [P <.1], except for Pb, which had somewhat lower values, between.542 and.75). The PLFA changes are exemplified by the PCA loading plot for samples treated with Zn (Fig. 3). Specific identified PLFAs, including the branched brl7: and brl8:, the methyl-branched 1OMel6:, 1OMel7:, and 1Mel8:, the isobranched i16:, and the monounsaturated 16:li7t and 19:1b (with unknown double-bond position), as well as the unidentified PLFAs xl, x2, and x5, increased in most cases at high doses of all metals. These PLFAs were found to the right in the plot (Fig. 3). The cyclopropane fatty acid cyl7: also increased in response to all metal treatments except Cu. The monounsaturated PLFAs 16:lX9, 16:lw7c, 16:1w5, and 19: la, as well as the branched i14:, i15:, a15:, and i17:, all decreased in the metal-amended samples and were thus found to the left in the plot (Fig. 3). The largest decrease was found for the unidentified PLFAsx4 andx6 (not shown in the plot). Palmitic acid (16:) was in all cases found very close to, or on, the origin in the PCA plots, indicating that this PLFA 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 moles percent) in the highest-dose treatment and the controls are calculated, the magnitudes of changes in individual PLFAs due to the amendments of different heavy metals can be compared (Table 2). This ratio is subsequently termed the heavy metal effect ratio. One must bear in mind that the toxicity of the highest dose of a certain metal is not necessarily similar to that of the other metals. Thus, the ratios for different metals are not directly comparable. The largest increase in the forest soil treated with heavy metals was found for brl7:, with ratios up to Ratios between 1.3 and 2.5 were found in most cases for xl, i16:, 16:lo7t, 1OMel6:, x2, brl8:, 1OMel7:, 1OMel8:, x5, and 2:. Exceptions were found for the Ni- and Pb-amended samples, in which ratios for lmel6: (Ni sample) and i16: and lmel8: (Ni and Pb) were between 1.4 and 1.7 in the second-highest-dose treatments (not shown) but then decreased to near control levels in the highest-dose treatments. The greatest decrease was found for the unidentified PLFA x6, which decreased from 3. mol% in the controls to almost undetectable levels (ratio,.6) in the high-metal treatments. The ratios for i14:, i15:, a15:, 16:1w9, 16: lw7c, 16:1w5, i17:, x4, and 19:1a were in most cases lower than.7 and in several cases lower than.5 (Table 2). The amount (moles percent) of some of the PLFAs that changed most in relative abundance in the forest soil amended with Zn was plotted against the levels of contamination (Fig. 4a and b). For most of these PLFAs, the increase or decrease in moles percent started at Zn levels around 4 to 8 mmol kg (dw) of soil-', and then there were continuous 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.6 mol% in the 16-mmol kg-' treatment, and for x6, which decreased from ca. 2.3 mol% in the 16-mmol kg-' treatment to.7 mol% in the 64-mmol kg-' treatment (Fig. 4a). A few PLFAs, e.g., cyl7: (Fig. 4a) and 18:2X6 and 18:1X9 (not shown), increased in moderately contaminated samples and then decreased in the more heavily polluted ones. The reverse was found for 18:1X7 (not

4 ' A 9c CO 368

5 VOL. 59, 1993 PFLA COMPOSITION OF HEAVY-METAL-CONTAMINATED SOILS 369 TABLE 2. PLFAs in forest soilsa Peak Proportion of PLFA Ratio with given metal no. PLFA RRT (mol%) in unamended soil Cd Cu Ni Pb Zn 1 i14: ± : ± ils: ± a15: xlb : i16: i16: ± :1X ± :1w7c ± :1b7t :1X ± : brl7: ± NIC lomel6: ± i17: a17: :1w ± x2b ± cy17: ± : ± brl8: ± lomel7: ± x3b ± x4b ± :2X ± :1w ± :1M ± : ± : :1ad lomel8: ± :1bd cyl9: xsb x ± : a 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:). In the unamended soil samples (controls), the proportions of the PLFAs, expressed as means ± standard error (n = 5), are given. To show 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 the control 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 samples and then increased in high-metal samples. The two PLFAs 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 kg of soil-1 (Fig. 4b). Since the total amount of PLFAs changed very little due to the metal contamination (see below), the use of absolute values (nanomoles per gram of soil) would give results similar to the data expressed in moles percent. In the other metal series, the changes in individual PLFAs over the metal gradients were similar to those found for Zn (data not shown). PLFA patterns in polluted arable soil. PCA of the arable soil samples treated with the different metals separated the controls from the metal-treated samples. The control and low-metal samples were found to the left and high-metal samples were found to the right in the scores plots (Fig. 5a to e). The change in PLFA composition due to the metal contamination was mostly shown by the first principal component, which explained 51.1, 48.4, 48.7, 49.4, and 5.5% of the variation in PLFA patterns for the samples treated with Cd, Cu, Ni, Pb, and Zn, respectively. The samples also separated in the second principal component, which explained 2.6 to 3.3% of the variation in PLFA patterns. The lowest-dose treatments generally did not separate the metaltreated samples from the controls, but at higher metal levels there 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 most cases. When the loading values of the individual PLFAs along the first PLS component for each of the different metals were regressed versus those from the other metals, positive relationships were found (rd =.355 to.557 [P <.1], except for Ni, which when regressed versus Cd had r2 =.298 [P <.1] and when regressed versus Zn had r2 =.123 [not significant]). The r2 values for these regressions for the arable soil were generally lower than those for the forest soil, which shows that the similarity between the

6 361 FROSTEGARD ET AL. PLFA TABLE 3. PLFAs in arable soila Proportion of PLFA Ratio with given metal (mol%) in unamended Cd Cu Ni Pb Zn soil i14:.99 ± : 1.21 ± i15: 9.21 ± a15: 6.4 ± : brl6:.51 ± i16:1.57 ± i16: 3.49 ± :1h9 1.6 ± :1w7c 7.19 ± :1w7t.61 ± :1X ± : ± brl7:.54 ± lomel6: 8.82 ± i17: 2.58 ± a17: 2.31 ± :1w8 1.8 ± cy17: 2.96 ± :.63 ± brl8: 1.97 ± lomel7:.86 ± :2w ± :1w ± :1M ± : ± : 2.33 ± :1a.89 ± lomel8: 2.18 ± cyl9: 3.42 ± :.49 ± a In the unamended soil samples (controls), the proportions of the PLFAs, expressed as means ± standard error (n = 5), are given. To show 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 the control are given. changes in PLFA patterns between the different metal treatments was less evident for the arable soil than for the forest soil. The PLFA changes are exemplified by the PCA loading plot for samples treated with Zn (Fig. 6). In most cases, 15:, brl7:, 17:, lomel7:, and 18:2X6 increased in the highmetal samples (to the right in the plot), while i15:, 16:1w7c, 16:1X5, lomel6:, and i17: decreased (to the left). For all metals except Zn, 16:1o7t decreased in the metal-amended treatments. Saturated 16: was found on or very close to the origin 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 1.96 and 4.13 in samples polluted with Cd, Ni, Pb, and Zn. Ratios higher than 1.2 were found in most cases for PLFAs 15:, 16:1X9, brl7:, 17:1X8, 17:, and 18:1Xo7. For 15:, brl7:, cyl7:, and 17:, a higher ratio was found in almost all cases when the heavy metal ratio was calculated for the next highest level of contamination. lomel7: had values similar to or lower than those for the controls. However, when heavy metal effect ratios were calculated for the next highest level, the ratios (1.31 to 3.) indicated an increase due to all metals. Ratios lower than.8 were found in most cases for i15:, 16:1h7t, 16:1w5, lomel6:, i17:, and lomel8:. APPL. ENvIRON. MICROBIOL. The effect of Cu on the PLFA pattern in the arable soil differed from that of the other metals (Fig. 7). This was also shown when the PLS loading values of the individual PLFAs from the first component were regressed versus those from the other metal treatments. No relationship was found (r2 close to zero). The most striking difference in PLFA profile compared with those with the other metals was found for 18:2)6, which decreased with Cu contamination but increased in response to all other metals. In the sample with the highest dose of Cu, 18:2X6 had a ratio of.78 compared with the controls (Table 3). The greatest decrease was found for 1Mel7: and 1Mel8: (Fig. 7), although the effect was not as strong in the sample with the highest dose of Cu, as seen from the heavy metal effect ratios given in Table 3. The strongest increase in the Cu-contaminated soil was found for cyl7: and brl7:. These PLFAs also increased in response to other metals (at least in the next highest doses of contamination; see above), but for cyl7: the increase due to Cu was more pronounced than that due to the other metals (Table 3). PLFAs 16:1X5 and 16:17c increased in response to Cu, while they generally decreased somewhat in the presence of other metals. PLFAs that responded to Cu similarly to other metals were 15:, 17:, 18:1X7, and 19:1a, which generally increased in moderately and/or highly contaminated samples, and a15:, brl6:, i16:, i17:, and a17:, which generally decreased in metal-treated soil. Biomass and activity. The microbial biomass at different levels of metal contamination was measured as ATP content, L-PO4 content, and totplfa in both soil types. The respiration at different metal concentrations was measured only in the forest soil. The values for the metal-contaminated samples were calculated as a percentage of the values in the control samples. In both soil types, the effects of heavy metals, here exemplified by the Cd-amended samples (Fig. 8a and b), on totplfa and L-P4 content were minor even at the highest doses, while there was a decrease in ATP content and respiration due to the metal contamination. By using the linear part of the respective curves, the ED5 and ED1 in ATP content, respiration, L-PO4 content, and totplfa could be calculated. In none of the cases did the totplfa or L-PO4 content decrease below 5% of the control values, and for L-PO4 not even a 1% reduction was found. The ED1 values for ATP, respiration, and totplfa were compared with the doses at which changes in PLFA patterns were found to be statistically significant (P <.5), using SIMCA analysis (Tables 4 and 5). In the forest soil, the doses at which changes in PLFA pattern occurred corresponded fairly well to the ED1 values for ATP content, respiration, and totplfa for each individual metal, with the exception of Zn, for which no decrease in ATP could be detected over the metal range used, and Pb, for which the ED1 values of respiration and totplfa were higher than the dose at which a change in PLFA pattern was found. In the arable soil, the changes in PLFA pattern were found at the same or lower metal levels than the ED1 values for both ATP content and totplfa for all metals except Zn. DISCUSSION The two soil types used were chosen so that they would be contrasting in ph and organic matter content. The unamended samples of the two soil types also had somewhat different PLFA profiles (Tables 2 and 3). This was expected, since it is well known that different soil types have different microbial communities. Although differing in abundance, several PLFAs reacted similarly to the metal amendments in

7 VOL. 59, 1993 PFLA COMPOSITION OF HEAVY-METAL-CONTAMINATED SOILS 3611 a 64 b 1 P O. o Cd C d 16Cu FIG _2.PCAshoingvaratin_n_PFA_ attrn_n_ aforst_ oi ~~~~~~~~~~~~~~ Ni FIG. 2. PCA showing variation in PLFA pattern in a forest soil i due to heavy metal contamination. Scores are for samples contane mated with the indicated levels: (a),.125,.25,.5, 1, 2, 4, 8, 16, 8 Pb 32, and 64 mmol of Cd kg (dw) of soil-'; (b) to 128 mmolof Cu kg 26(dw) of soil-'; (c) to 128 mmol of Ni kg (dw) of soil-'; (d) to 256 mmol of Pb kg (dw) of soil-'; (e) to 256 mmol of Zn kg (dw) of soil'. The cross indicates origin of coordinates the two soil types. One example is provided by the iso- 4 branched PLFAs i15: and i17:, 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:, br17:, 32 q~~~~~~y17:, and br18: 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 is found in both soils might represent different groups of organisms, and thus changes in the abundance of a certain

8 3612 FROSTEGARD ET AL. APPL. ENvIRON. MICROBIOL. a, 16:15 16:ho9 6: 1~~~~~~~ ~~~~~ * 165:1w7c als: 19:la i14: is:1z7 ~~~~~~~~~~~~~~H. 16.lw7c18:1 14: lomel6: 1T: i16: il6:1 lomel7: 115l: 4 lmels: k il7: cid;{i;;e-18: al : 17: br 16:1, 7t brl7: 17:1w8 2: is:1.,9 cy17: 19:lb FIG. 3. PCA showing loading values for individual PLFAs in a forest soil contaminated with to 256 mmol of Zn kg (dw) of soil-'. PLFAs found to the right in the plot had increased in the high-metal samples, while PLFAs found to the left were more abundant in the control and low-metal samples. PLFAs x4 and x6 were included in the PCA, but they were excluded from the plot in order to be able to show the positions of the other PLFAs more clearly. x4 and x6 were found to the lower left outside the plot since they almost disappeared in the metal-treated samples. The cross indicates origin of coordinates. PLFA might represent changes in different organism groups in the two soil types. A predominance of gram-negative over gram-positive bacteria is often found in metal-contaminated soils (13, 17). In this study, some evidence for a similar shift was indicated by a decrease in several iso- and anteiso-branched PLFAs, all of which are commonly found in gram-positive bacteria (33). A further indication that such a shift had also occurred in our study was indicated by the increase in cyl7:, which is considered to be typical for gram-negative bacteria (41), in the metal-contaminated samples of both soil types (Fig. 3, 6, and 7; Tables 2 and 3). However, the interpretation must be tentative since, e.g., i16: and i16:1 increased because of some metals and 18:1o)7, which is commonly found in gram-negative bacteria (41), showed variable responses. This indicates that the separation between gram-positive and gram-negative bacteria in soil by PLFA patterns is not straightforward. This was also the case in a study of the effect of liming on the PLFA pattern of soil microorganisms (2). Several PLFAs, including the methyl-branched lomel6:, lomel7:, and lomel8:, reacted differently to the metals in the arable soil compared with the forest soil. The tuberculostearic acid lomel8: is found almost exclusively in actinomycetes (25, 28). lomel6: and lomel7: are also produced by several actinomycete genera (26, 37). All three lome-branched PLFAs increased in the forest soil (Fig. 3; Table 2), and the results thus suggest an increase in this group of organisms after metal contamination. The results for the arable soil are more difficult to interpret since not all xl x2 IL OMe17: o Amount of Zn (mmol) FIG. 4. Amounts of representative PLFAs in forest soil samples contaminated with to 256 mmol of Zn kg (dw) of soil-l, expressed as moles percent totplfa in each sample. lome-branched PLFAs responded similarly, and in some cases they also responded in different ways to different metals (Fig. 6 and 7; Table 3). If lomel8: is taken as an indicator of actinomycetes, the results indicated that this group of organisms either decreased or was unaffected in response to metals in the arable soil, while the opposite was found in the forest soil. In a study by Hiroki (23), the degree of tolerance in microorganisms isolated from a fallow paddy field contaminated by Cd, Cu, apd Zn was suggested to be fungi > bacteria > actinomycetes. Jordan and Lechevalier (24) reported that the Zn tolerance in actinomycetes isolated from a forest close to a smelter increased compared with isolates from noncontaminated soil but that actinomycetes were generally less tolerant than bacteria and fungi. Similar observations were made for isolates from lead mine waste,

9 VOL. 59, 1993 PFLA COMPOSITION OF HEAVY-METAL-CONTAMINATED SOILS 3613 a 16b32 is C e Cd 8 C FG5.CshwgvraoinPApteinaaalsi due to heavy metal cotamination. Scores aeforsamplescontam 12 Pb Ni FIG. 5. PCA showing variation in PLFA pattern in an arable soil due to heavy metal contamination. Scores are for samples contain- 128 mated with the indicated levels: (a),.6,.125,.25,.5, 1, 2, 4, 8, 16, and 32 mmol of Cd kg (dw) of soil'; (b) to 64 mmol of Cu kg (dw) of soil-; (c) to 64 mmol of Ni kg (dw) of soil'; (d) to 128 msmol of Pb kg (dw) of soil-; (e) to 128 mmol of Zn kg (dw) of.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 ~~~~~~~~~less sensitive to Cd than those of eubacteria. These results, 2 as well as those from the present study, thus indicate that different 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: and 17:, which have been considered to 8 be of predominantly bacterial origin (19, 38). They increased in all metal-contaminated samples in the arable soil (Fig. 6 and 7; Table 3), while they were found close to the origin in all PCA plots from the forest soil (Fig. 3; Table 2). This suggests that they were unaffected by the metal treatments in the latter soil. 16:1w7t also showed different responses in the

10 3614 FROSTEGARD ET AL. APPL. ENvIRON. MICROBIOL. 18:2w6 brl8: PC1 19:1a 17:lwg a17: cyl7: 13:1c9 CY brl7: brl:brls: 1Me16: 16:1c7c : 15: 15:1 i19 a15:o 17: 15: lomel7: 16:1.6 brl6: 2: ils: lomels:o 114: 16:1l7t FIG. 6. PCA showing loading values for individual PLFAs in an arable soil contaminated with to 128 mmol of Zn kg (dw) of soil-'. PLFAs found to the right in the plot had increased in the high-metalcontaminated samples, while PLFAs found to the left were more abundant in the control and low-metal samples. PLFA 16:1X9 was not included in the PCA since it coeluted with the following peak in the GCs of several samples. The cross indicates origin of coordinates. two soils. It decreased in response to all metals (except Cu; Fig. 7) in the arable soil (Fig. 6; Table 3), while it generally increased in the forest soil following metal amendment (Fig. 3; Table 2). It is generally agreed that fungi are less sensitive to metal pollution than bacteria (13). In the present study, this would be the case in the arable soil, provided that PLFA 18:2w6 is regarded as a reliable indicator of fungal biomass. This PLFA increased with increasing contamination level for all metals except Cu. The decrease in this PLFA in samples contaminated with Cu could be explained by the fact that Cu is known to be toxic to fungi and is commonly used as a fungicide. In the forest soil, on the other hand, 18:2w6 seemed to be unaffected by the treatments. One explanation might be that 18:2w6 is also found in plant residues, and it has been reported to constitute a dominant fraction of the total fatty acids in Pinus sylvestris L. (35). The effect of metals on the portion of this PLFA that was derived from fungi might therefore be masked by a large amount of 18:2w6 derived from plant material. Metal toxicity depends on the physicochemical properties of the soil (12, 15) and is usually stronger in sandy loam soils than in highly organic soils. However, in our study there appeared to be little difference between metal toxicity in the two different soils, despite the difference in organic matter content (compare ED5 values for ATP and respiration for the two soils; Tables 4 and 5). One reason for this could be that the sandy loam used in the present investigation had a ph of 7.8, compared with ph 4.5 in the forest soil. Several studies, primarily with laboratory media, have shown that ph affects the toxicity of metals (6 and references cited 14: 17:16 17: 18:2w6 19:1a lomel7: 15:1.6 1OMel6: 18:1X7 brl7: lomelg: il7: cyl9: cyl7: a17: 16:lw7t 16:1w7c 2: 18: 18:1 16: 116: 16:lcoS 116:1 15: ils.- brl6: a15: FIG. 7. PCA showing loading values for individual PLFAs in an arable soil contaminated with to 64 mmol of Cu kg (dw) of soil-'. PLFAs found to the right in the plot had increased at high levels of metal contamination, while PLFAs found to the left were more abundant in the control and low-metal samples. The cross indicates origin of coordinates. therein), and it is generally argued that metal toxicity is lower at higher ph. Doelman (13) suggested that, for several processes that follow a sigmoid dose-response relationship to a toxic substance, a retardation of the entire process might be preceded by a qualitative shift in the microbial community. In the present study, the levels found for detectable changes, given as ED1, in ATP content, totplfa, and respiration were rather similar for each of the metals and also similar to levels at which changes in the PLFA pattern were evident by SIMCA analysis (Tables 4 and 5). However, the SIMCA analysis, which was used to investigate the metal levels at which changes in PLFA patterns could be considered statistically significant, was much more rigorous than the calculations to determine at what levels 1% decreases could be found in ATP, respiration, and totplfa. In the latter case, because of inherent variation, a 1% decrease would be very unlikely to be statistically significant. Also, for changes in the PLFA pattern, no exact dose could be calculated for each metal. Instead, the doses given are the lowest ones used in the experiment at which significant changes in the PLFA pattern could be found. The changes thus actually occurred at a metal level somewhere between the one given in Tables 4 and 5 and the one just below those used in the experiment. Considering this, it is possible that the changes in PLFA patterns occurred at metal levels lower than those which affected ATP content and respiration, and thus our results are not in conflict with the suggestion of Doelman (13) mentioned above. The effects of different heavy metals on the PLFA patterns, ATP content, totplfa, and respiration of the soils used in the present study were found at concentrations corresponding to the lower levels of the ranges in the studies summarized by Baath (3). The phospholipid content of both soils, measured either as

11 VOL. 59, 1993 PFLA COMPOSITION OF HEAVY-METAL-CONTAMINATED SOILS 3615 a 1 L-PO4 totplfa i* ATP * resp TABLE 4. Levels of heavy metal contamination of forest soil at which changes in PLFA patterns were found as calculated by SIMCA analysis and levels at which an ED5 or ED1 in ATP content, respiration, and totplfa were found, expressed as millimoles per kilogram (dw) of soil 8 1- o.- o O Amount of Cd added (mmol) *- L-P4 totplfa ATP Amount of Cd added (mmol) FIG. 8. ATP content, respiration rate, totplfa, and L-P4 content in (a) a forest soil contaminated with to 64 mmol of Cd kg (dw) of soil-' and (b) an arable soil contaminated with to 32 mmol of Cd kg (dw) of soil-'. L-P4 or as totplfas, was less affected by heavy metal contamination than was the ATP content (Fig. 8a and b). Earlier results from different uncontaminated soils indicated a linear relationship between the ATP and L-P4 content in soil (21). There was thus a different relationship between the two biomass measurements in uncontaminated soils compared with contaminated ones. This might indicate that the turnover rate of phospholipids in soil is low, especially in toxified soils. However, this seems unlikely, since an increase was found for several PLFAs at high metal concentrations. At the same time, no increase was found in L-P4 content or totplfa, indicating that some phospholipids had been metabolized in the soil. One explanation might be that organisms persist in highly toxic soils with intact membranes Metal PLFA ATP Respiration totplfa pattern ED5 ED1 ED5 ED1 ED1 Cd Cu > Ni Pb Zn 4 >256 > >256 but in a resting state, with lower ATP content and activity than in uncontaminated soil. To compare toxicity of the different metals in different soils, one can compare the ED5 values calculated for ATP content and respiration rate. Thus, in the forest soil the metal toxicity measured on the basis of ATP and soil respiration indicated a higher toxicity for Cd and Ni than for Cu and Zn (Tables 4 and 5). This could also be done with separate PLFAs. However, only one PLFA, namely, x6, decreased in abundance by more than 5%, and many PLFAs increased (Table 2). Furthermore, using just one PLFA at a time would increase random variation, decreasing the precision in the estimated ED5 value. Since the effects of the metals were similar in the forest soil, one way of overcoming this problem would be to use PLS analysis (29) to regress the PLFA pattern by a single metal. By using Zn and a PLS model with one component, a linear relationship between added and predicted metal concentrations was found (r2 =.991) (Fig. 9). The PLFA patterns for the other metals were then used to predict apparent toxicity expressed as Zn concentration, here exemplified by Cd and Cu. Both Cd and Cu showed a linear relationship between the added amount of the metal and predicted toxicity as micromoles of Zn (r2 =.949 and.98 for Cd and Cu, respectively). For Cu, approximately the same toxicity as Zn was found on a molar basis; e.g., 32 mmol of added Zn was estimated to give the same effect on the PLFA pattern as 38 mmol of Cu. Cd was more toxic, and addition of 32 mmol of Zn was estimated to have the same toxicity as 11 mmol of Cd. The data for Ni and Pb fitted the Zn model less well (calculations not shown). For Ni, the correlation was less good (r2 =.827), while for Pb the slope of the regression differed from that of the other metals. This might be due to Pb affecting the PLFA pattern slightly differently than the other metals did, although this was not evident from the PCA. Nevertheless, an TABLE 5. Levels of heavy metal contamination of a sandy loam at which changes in PLFA patterns were found as calculated by SIMCA analysis and levels at which an ED5 or ED1 in ATP content and totplfa were found, expressed as millimoles per kilogram (dw) of soil PLFA ATP totplfa Metal patternedm ED1 ED1 Cd Cu >64 Ni Pb Zn

12 3616 FROSTEGARD ET A. APPL. ENvIRON. MICROBIOL. E E 2- a o~~~~~~~~~~~~~~~~~~o a #1 2 Cu.2 Log added amount of metal (mmol) FIG. 9. PLS model, created using cross-validation, of PLFA profiles from forest soil contaminated with 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 as Zn concentration. A linear regression for each metal was calculated by using only the data points with filled symbols (r2 =.949, r2 =.98, and r2 =.991 for Cd, Cu, and Zn, respectively). Open circles denote Cd-polluted samples not included in the regression, open triangles denote Cu-polluted samples not included in the regression, and open squares denote Zn-polluted samples not included in the regression. addition of 32 mmol of Zn could be calculated to be equivalent to 13 mmol of Ni or 55 mmol of Pb. Thus, the toxicity of the different metals towards the PLFA pattern decreased in the order Cd = Ni > Zn = Cu > Pb. This hierarchy of toxicity is normally found (see reference 3). Similar results were 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-contaminated soils. In the arable soil, the use of PLS regression was less straightforward, since the different metals affected the PLFA patterns differently than in the forest soil (calculations not shown). However, the overall picture indicated that Cd was the most toxic metal. This was also seen for the ATP measurements (Table 5). It is well-known that metal treatments affect the ph of soils. In a study of the combined effects of storage and heavy metals on the PLFA patterns (22), using the same soils as in the present investigation, the ph was measured after 2 years. In the arable soil, the highest levels of metal addition decreased ph (H2) by.4 (Ni),.6 (Cu and Zn), and 1.3 (Ni and Pb) units. In the forest soil, the ph decreased somewhat more. For the two highest levels of metal addition, the phs compared with that of the controls decreased 1. and 1.2 (Cd), 1.4 and 1.6 (Cu), 1.1 and 1.3 (Ni), 1.3 and 1.9 (Pb), and 1.4 and 1.5 (Zn) units. However, at the levels at which changes in PLFA patterns were found (Tables 4 and 5), the ph decrease was never more than.2 units, and for Ni and Pb, the ph at those levels of metal amendment was the same as that for the controls. Nevertheless, ph is a confounding factor in elucidation of the effects of heavy metals. Although the effects of metals on the PLFA pattern were C generally similar, some PLFAs responded differently in the two soil types. Different metals could also affect a certain soil in different ways, exemplified by Cu in the arable soil. It is thus difficult to investigate metal-derived changes in a soil without knowing how that specific soil reacts to metal pollution. Therefore, laboratory experiments on soils with heavy metals must be regarded as a prerequisite to field studies. Furthermore, it was difficult to elucidate changes in large groups such as gram-positive and gram-negative bacteria, since several PLFAs specific for one group were affected differently by the heavy metal pollution. However, the analyses of PLFA patterns in soils still indicated changes in the microbial community, changes that could be quantified and compared, e.g., between different heavy metals by using multivariate statistics. ACKNOWLEDGMENT Financial support was obtained from the Swedish National Environment Protection Agency. REFERENCES 1. Arnebrant, K., and E. Baath Measurements of ATP in forest humus. Soil Biol. Biochem. 23: Arnebrant, K., E. Baath, and A. Nordgren Copper tolerance of microfungi isolated from polluted and unpolluted forest soil. Mycologia 79: Baath, E Effects of heavy metals in soil on microbial processes and populations (a review). Water Air Soil Pollut. 47: Baath, E., A. Frostegard, and H. Fritze Soil bacterial biomass, activity, phospholipid fatty acid pattern, and ph tolerance in an area polluted with alkaline dust deposition. Appl. Environ. Microbiol. 58: Babich, H., and G. Stotzky Sensitivity ofvarious bacteria, actinomycetes, and fungi to cadmium and the influence of ph on sensitivity. Appl. Environ. Microbiol. 33: Babich, H., and G. Stotzky Heavy metal toxicity to microbe-mediated ecologic processes: a review and potential application to regulatory policies. Environ. Res. 36: Bakken, L Separation and purification of bacteria from soil. Appl. Environ. Microbiol. 49: Barkay, T., S. C. Tripp, and B. H. Olson Effect of metal-rich sewage sludge application on the bacterial communities of grasslands. Appl. Environ. Microbiol. 49: Chander, K., and P. C. Brookes Microbial biomass dynamics during the decomposition of glucose and maize in metal-contaminated and non-contaminated soils. Soil Biol. Biochem. 23: Chander, K., and P. C. Brookes Plant inputs of carbon to metal-contaminated soil and effects on the soil microbial biomass. Soil Biol. Biochem. 23: Chander, K., and P. C. Brookes Synthesis of microbial biomass from added glucose in metal-contaminated and noncontaminated soils following repeated fumigation. Soil Biol. Biochem. 24: Collins, Y. E., and G. Stotzky Factors affecting the toxicity of heavy metals to microbes, p In T. J. Beveridge and R. J. Doyle (ed.), Metal ions and bacteria. John Wiley & Sons, New York. 13. Doelman, P Resistance of soil microbial communities to heavy metals, p In V. Jensen, A. Kjoller, and L. H. Sorensen (ed.), Microbial communities in soil. Elsevier, London. 14. Doelman, P., and L. Haanstra Effects of lead on the soil bacterial microflora. Soil Biol. Biochem. 11: Doelman, P., and L. Haanstra Short- and long-term effects of heavy metals on urease activity in soils. Biol. Fertil. Soils 2: Dowling, N. J. E., F. Widdel, and D. C. White Phospholipid ester-linked fatty acid biomarkers of acetate-oxidizing

13 VOL. 59, 1993 PFLA COMPOSITION OF HEAVY-METAL-CONTAMINATED SOILS 3617 sulphate-reducers and other sulphide-forming bacteria. J. Gen. Microbiol. 132: Duxbury, T Ecological aspects of heavy metal responses in microorganisms, p In K. C. Marshall (ed.), Advances in microbial ecology, vol. 8. Plenum Press, New York. 18. Duxbury, T., and B. Bicknell Metal-tolerant bacterial populations from natural and metal-polluted soils. Soil Biol. Biochem. 15: Federie, T. W Microbial distribution in soil-new techniques, p In F. Megusar and M. Gantar (ed.), Perspectives in microbial ecology. Slovene Society for Microbiology, Ljubljana, Slovenia. 2. Frostegard, A., E. Biath, and A. Tunlid Shifts in the structure of soil microbial communities in limed forests as revealed by phospholipid fatty acid analysis. Soil Biol. Biochem. 25: Frostegard, A., A. Tunlid, and E. Baath Microbial biomass measured as total lipid phosphate in soils of different organic content. J. Microbiol. Methods 14: Frostegard, A., A. Tunlid, and E. Biath. Unpublished data. 23. HiroId, M Effects of heavy metal contamination on soil microbial population. Soil Sci. Plant Nutr. 38: Jordan, M. J., and M. P. Lechevalier Effects of zincsmelter emissions on forest soil microflora. Can. J. Microbiol. 21: Kroppenstedt, R. M Fatty acids and menaquinone analysis of actinomycetes and related organisms, p In M. Goodfellow and D. E. Minnikin (ed.), Chemical methods in bacterial systematics. Academic Press Ltd., London. 26. Kroppenstedt, R M., and H. J. Kutzner Biochemical taxonomy of some problem actinomycetes, p In M. Mordarski, W. Kurylowicz, and J. Jeljaszewicz (ed.), Proceedings of the International Symposium on Nocardia and Streptomyces. Gustav Fischer Verlag, Stuttgart. 27. Kvalheim,. M., and T. V. Karstang A general-purpose program for multivariate data analysis. Chemomet. Intelligent Lab. Syst. 2: Lechevalier, M. P Lipids in bacterial taxonomy-a taxonomist's view. Crit. Rev. Microbiol. 5: Martens, M., and H. Martens Partial least squares regression, p In J. R. Piggott (ed.), Statistical procedures in food research. Elsevier Applied Science Publishers, London. 3. Nichols, P. D., J. B. Guckert, and D. C. White Determination of monounsaturated fatty acid double bond position and geometry for microbial monocultures and complex consortia by capillary GC-MS of their dimethyl disulphide adducts. J. Microbiol. Methods 5: Nordgren, A., T. Kauri, E. Baath, and B. Soderstrom Soil microbial activity, mycelial lengths and physiological groups of bacteria in a heavy metal polluted area. Environ. Pollut. 41: Ohya, H., Y. Komai, and M. Yamaguchi Zinc effects on a soil bacterial flora characterized by fatty acid composition of the isolates. Biol. Fertil. Soils 2: O'Leary, W. M., and S. G. Wilkinson Gram-positive bacteria, p In C. Ratledge and S. G. Wilkinson (ed.), Microbial lipids, vol. 1. Academic Press Ltd., London. 34. Ryhage, R., and E. Stenhagen Mass spectrometry in lipid research. J. Lipid Res. 1: Saranpil, P., and H. Nyberg Lipids and sterols of Pinus sylvestris L. sapwood and heartwood. Trees 1: Smith, G. A., J. S. Nickels, B. D. Kerger, J. D. Davis, and S. P. Collins Quantitative characterization of microbial biomass and community structure in subsurface material: a prokaryotic consortium responsive to organic contamination. Can. J. Microbiol. 32: Tisdall, P. A., G. D. Roberts, and J. P. Anhalt Identification of clinical isolates of mycobacteria with gas-liquid chromatography alone. J. Clin. Microbiol. 1: Tunlid, A., H. A. J. Hoitink, C. Low, and D. C. White Characterization of bacteria that suppress Rhizoctonia damping-off in bark compost media by analysis of fatty acid biomarkers. Appl. Environ. Microbiol. 55: Tunlid, A., and D. C. White Biochemical analysis of biomass, community structure, nutritional status, and metabolic activity of microbial communities in soil, p In G. Stotzky and J.-M. Bollag (ed.), Soil biochemistry, vol. 7. Marcel Dekker, New York. 4. Tyler, G Heavy metals in soil biology and biochemistry, p In E. A. Paul and J. N. Ladd (ed.), Soil biochemistry, vol. 5. Marcel Dekker, New York. 41. Willinson, S. G Gram-negative bacteria, p In C. Ratledge and S. G. Wilkinson (ed.), Microbial lipids, vol. 1. Academic Press Ltd., London. 42. Williams, S. T., T. McNeilly, and E. M. Wellington The decomposition of vegetation growing on metal mine waste. Soil Biol. Biochem. 9: Wold, S Pattern recognition by means of disjoint principal components models. Pattern Recognition 8: Wold, S Cross-validatory estimation of the number of components in factor and principal component models. Technometrics 2: Yamamoto, H., T. Kadzunori, and T. Uchiwa Fungal flora of soil polluted with copper. Soil Biol. Biochem. 17: