Nitrogen fixation is often enhanced in the rhizosphere as compared with bulk soil, due to asymbiotic microorganisms utilizing root exudates as an energy source. We have studied the activity and composition of asymbiotic soil diazotrophs following pulse additions of artificial root exudates and single carbon sources, simulating the situation of bulk soil coming into contact with exudates from growing roots. Artificial root exudates and single sugars rapidly induced nitrogen fixation. The population of potential diazotrophs was studied using universal and group-specific nifH polymerase chain reaction (PCR) and denaturing gradient gel electrophoresis (DGGE) analysis. Reverse transcription PCR of nifH mRNA confirmed that phylotypes with an apparently increasing population size also expressed the nitrogenase system. According to our results, the actively nitrogen-fixing population represents only a fraction of the total diazotroph diversity, and the results of group-specific nifH PCR and phylogenetic analysis of cloned nifH and 16S rRNA gene fragments identified active species that belonged to the genus Azotobacter. Rapid changes of transcriptional activity over time were observed, indicating different growth and activation strategies in different Azotobacter strains. Only sugar-containing substrates were able to induce nitrogen fixation, but substrate concentration and the presence of organic acids may have additional selective effects on the active diazotroph population.
Microbial biomass and activity in the vicinity of plant roots are usually higher than in bulk soil, a phenomenon known as the rhizosphere effect. During the vegetation period, plants release up to 20% of photosynthetically assimilated carbon (C) as rhizodeposition consisting of mucilage, sloughed-off tissue and low-molecular-weight root exudates (Marschner, 1995; Hutsch et al., 2002). The root exudates generally consist of carbohydrates, organic acids, amino acids and amides, vitamins and other compounds (Lynch and Whipps, 1990; Marschner, 1995; Kuiper et al., 2002). The amount and composition of root exudates entering the soil is variable, depending on plant species (Hertenberger et al., 2002; Hutsch et al., 2002), plant age and nutrient status, as well as environmental conditions (Lynch and Whipps, 1990; Kuiper et al., 2002). Studies have shown that 64–86% of the carbon released into the rhizosphere are respired by microorganisms (Hutsch et al., 2002), leading to a 10- to 100-fold increase in microbial population size as compared with the surrounding bulk soil (Weller and Thomashow, 1994).
It is generally accepted that studying soil microorganisms by cultivation on laboratory media has selective effects and thus yields results that are not representative of the whole microbial community (Wagner et al., 1993). Diazotrophs represent a physiologically and phylogenetically highly diverse functional group, and consequently the functional gene nifH (nitrogenase reductase) is the prevailing marker gene for the detection and identification of potential diazotrophs in environmental samples. This approach has been applied to a variety of environmental systems (Zehr et al., 1998; Widmer et al., 1999; Bagwell et al., 2002; Hamelin et al., 2002; Rösch et al., 2002; Diallo et al., 2004). However, the presence of nifH genes is in itself not indicative of active nitrogen fixation. Nitrogen fixation is tightly regulated in all microorganisms that have been studied (Merrick, 1992), and its expression may depend on environmental conditions. Due to the stringent transcriptional regulation of the nifH gene, the detection of nifH mRNA is a suitable molecular indicator for nitrogenase activity (Merrick, 1992; Zehr et al., 1996; Bürgmann et al., 2003).
The detection of nifH mRNA in soil samples is a relatively new approach, but successful amplification of nifH mRNA from soil has been reported in a number of recent publications (Burke et al., 2002; Brown et al., 2003; Bürgmann et al., 2003). Despite these pioneer studies we still know very little about the actual nitrogen-fixing activity of the various asymbiotic diazotrophs in soil, and the key players among free-living diazotrophs have not been unambiguously identified in many environments.
It may be hypothesized that plants specifically stimulate beneficial diazotroph bacteria by root exudation. An increased or altered rhizodeposition may selectively stimulate and enrich certain groups of microorganisms, leading to new community structures (Griffiths et al., 1999). In this study we applied pulse additions of carbon sources commonly found in root exudates to soil in a laboratory microcosm experiment. We examined the impact of the substrate additions on the community structure and the activity of asymbiotic diazotrophs. Comparative analysis of the nifH gene pool and nifH mRNA transcripts was performed in order to determine potential key species that can take advantage of exudates released to the rhizosphere, and to study the dynamics of the adaptive processes in soil.
We studied the impact of nine treatments with sugars, organic acids and artificial root exudates (treatments I–IX in Table 1). The measured sum of concentrations of all determined compounds in the first samples taken after 6 h were 14.1 ± 1.8 mg g−1 dry soil in treatments of the first series (I–V) and 15.1 ± 1.2 mg g−1 dry soil in treatments of the second series (VI–IX). Except for acetic acid (treatment I) all tested substrates were degraded during the course of the experiment (Fig. 1). In the sampling period between 6 h and 33 h after substrate addition, 85–94% of sugars (treatments II, III, VII), up to 75% of acids (treatments I, IV, VI) and 72–84% of the artificial root exudates (treatments V, VIII, IX) were degraded (Table 1). The maximal rates determined for substrate degradation in the glucose and sucrose treatments were slightly lower (0.62 and 0.64 mg g−1 dry soil h−1) than for fructose and the artificial root exudates (0.73–0.78 mg g−1 dry soil h−1). Citric acid and malic acid degraded with similar rates (0.44 and 0.36 mg g−1 dry soil h−1) that were lower as compared with sugars and artificial root exudates (Table 1). Glucose and fructose showed similar degradation characteristics (Fig. 1B and C). Both components were degraded rapidly after a lag-phase of approximately 15–18 h. Sucrose (Fig. 2G) was rapidly hydrolysed, resulting in increasing glucose and fructose concentrations. In artificial root exudates (Fig. 1E, H and I) all individual components decreased over time, sugars more rapidly than organic acids. However, the hydrolysis of sucrose in the mixtures led to an initial increase in the concentrations of glucose and fructose.
Table 1. Substrates used in the study and apparent substrate degradation.
Substrate added (dry mass g−1 dry soil)
Apparent substrate degradation
Total (%, 6–33 h)
Maximum rate (mg g−1 soil h−1)
. V: 0.326 g of fructose, 0.326 g of glucose, 0.616 g of sucrose, 0.109 g of succinic acid, 0.123 g of malic acid; adapted from Griffiths et al. (1999), modified by omitting N-containing compounds.
. VIII: 0.1928 g of fructose, 0.1928 g of glucose, 0.3643 g of sucrose, 0.352 g of succinic acid, 0.3975 g of malic acid.
. IX: 0.326 g of fructose, 0.326 g of glucose, 0.616 g of sucrose, 0.109 g of citric acid, 0.123 g of malic acid.
. Calculated from the sum of substrates in the treatment (Fig. 1).
Nitrate was the dominant inorganic nitrogen species in the soil microcosms. After 6 h, 0.148 ± 0.003 mg NO3-N g−1 dry soil was detected in treatments of the first series and 0.169 ± 0.005 mg NO3-N g−1 dry soil in the second series (Fig. 2A and B). Nitrite was below the detection limit in all extracts, and ammonium contributed less than 4% to total mineral nitrogen at the beginning of the experiment and showed no clear trends throughout the experiment (Fig. 2C and D). With exception of the acetic acid treatment, nitrate decreased rapidly in all treatments from 15 h onwards (Fig. 2A and B). The preliminary experiment showed that nitrate concentrations in control treatments (H2O) did not change significantly over 48 h.
In preliminary tests, nitrogenase activity showed acceptable reproducibility (coefficient of variation < 20%) between microcosms. In the main experiment, significant (P < 0.01) nitrogenase activity was induced in treatments with the sugars (treatments II, III, VII) and by all artificial root exudates (treatments V, VIII, IX), but not by any of the organic acid treatments (treatments I, IV, VI) (Fig. 2E and F). The control treatment with water showed no detectable nitrogenase activity throughout the experiment (data not shown). In the active treatments, significant (P < 0.01) nitrogenase activity could first be detected 15–18 h after substrate addition and increased until the end of the experiment (33 h), with the exception of treatment VIII (Fig. 2F), which showed a reduction in activity between 30 and 33 h. In treatments III, V and VII activity appeared to approach a plateau after 33 h, but rates were still increasing linearly in treatments II and IX (Fig. 2E and F). The fastest increase and highest rates of nitrogenase activity were measured in treatments with glucose, fructose and the artificial root exudate (treatments II, III, V). The rates determined for sucrose and the modified artificial root exudates (treatments VII, VIII and IX) in the second series were lower (Fig. 2F).
nifH polymerase chain reaction (PCR) analysis of the diazotroph populations
nifH genes were amplified using both universal and group-specific primer sets (Table 2). The reproducibility of the molecular analysis was tested on DNA extracted from three treatments with three replicate microcosm each in a preliminary experiment. Analysis of the denaturing gradient gel electrophoresis (DGGE) patterns of the nifH-g1-GC-amplified polymerase chain reaction (PCR) products showed reproducible pattern changes in the sucrose treatment, and highly similar patterns from all time points and replicates from samples treated with either 20 mg g−1 dry soil malic acid or water only. This indicated that both the molecular methods as well as the induced community changes in the system were reproducible. Analyses based on other PCR protocols for nifH (nifH-uni, nifH-b1 and nifH-a2) and for bacterial rDNA yielded highly reproducible fingerprints as well (data not shown).
. For DGGE, these primers had a GC-clamp attached to the 5′ end: GC CCG CCG CGC GCG GCG GGC GGG GCG GGG GCA CGG GGG G (Muyzer et al., 1993). Use of the GC-clamped primers is indicated in the text by ‘-GC’ appended to the name of the primer set.
All soil DNA extracts from the main experiment contained nifH genes, as shown by PCR amplification with the universal primer set. Polymerase chain reaction products were also obtained using each of the group-specific primer sets nifH-g1, nifH-b1, nifH-c1, nifH-f1 and nifH-a2. The primer set nifH-a2, which was designed for Azospirillum nifH sequences reproducibly yielded a product only in the 24 h sample of treatment VII (malic acid). No PCR products were observed in negative control reactions with any primer set (data not shown).
Restriction fragment length polymorphism (RFLP) analysis of nifH genes amplified using nifH-uni primers resulted in complex fragment patterns, which showed some apparently random variations in series 1, e.g. the strong ∼170 bp band in treatment I, 12 h (Fig. 3A). Some shifts in the banding patterns appeared to be associated with active nitrogen fixation; most prominently, samples with active N fixation in series 2 (treatments VII, VIII, IX) revealed a double band of approximately 300 bp that first appeared in the 15 h samples and became more pronounced over time (Fig. 3A, band c). In both series, bands a and b tended to increase in intensity in samples with nitrogenase activity; the strongest shift was observed in treatment V after 24 h (Fig. 3A). In all treatments except V, the initially dominant bands of the patterns remained unaffected over time. To improve the resolution of our analysis we then used PCR with group-specific primers and DGGE.
Of the diazotroph subpopulations targeted by the available primer sets, only the population amplified by the nifH-g1 primer set showed significant DGGE pattern changes related to nitrogenase activity (Fig. 3B). In active treatments, bands 3 and 4 tended to decrease in intensity over time, while a number of bands (Fig. 3B, bands 1, 2, 5–11) increased or newly appeared. The apparent population shift was quantified by plotting the relative contribution of bands with increasing intensity (Fig. 3B, bands 1, 2, 5–9, termed ‘potentially active phylotypes’) to the total intensity over time (Fig. 3C). The increase of potentially active phylotypes with time was significant (P < 0.05) in the nitrogen-fixing treatments II, III, V, VII and IX but not in treatment VIII. Among organic acids a significant increase of potentially active phylotypes was only observed for the citric acid treatment (IV), and this was based on a single band (Fig. 3B, band 9). The nifH-g1 DGGE pattern obtained from treatment V after 24 h contained two unique bands (Fig. 3B, bands 10 and 11) that distinguished it from the other nitrogen-fixing treatments.
In contrast to nifH-g1, the nifH-b1 primer set amplified nifH from a population that apparently changed little over time (Fig. 4). Similar results were observed with the nifH-c1 primer set, while nifH-f1 RFLP pattern analysis revealed that random phylotypes were amplified (data not shown).
Analysis of active diazotrophs with reverse transcription PCR
Of the potential diazotroph subpopulations tested with group-specific primers, the group amplified by the nifH-g1 primer set appeared to be the only one that reacted with population changes to the applied treatments. Reverse transcription PCR (RT-PCR) was performed using the nifH-g1-GC and nifH-b1-GC primer sets. Reverse transcription PCR with nifH-g1-GC yielded amplification products in all treatments exhibiting nitrogenase activity, and in none of the inactive treatments (Fig. 5). Transcription was in some cases detectable shortly before significant (P < 0.01) nitrogenase activity was first detected (treatment V at 12 h, and treatments VIII, IX at 15 h). All control reactions testing for DNA contamination of the RNA extracts were negative (data not shown). Reverse transcription PCR using the nifH-b1 primer set did not result in visible amplification products except in positive control reactions containing DNA (data not shown).
The DGGE analysis of nifH-g1-GC-amplified nifH mRNA showed a number of bands (Fig. 6, bands 1, 2, 5–9) that corresponded to the potentially active phylotypes observed in the community fingerprints (Figs 3B, C and 6). In contrast, bands that were absent from the RT-PCR pattern (Fig. 6, bands 3 and 4) decreased in rel tive intensity or even vanished from the population fingerprint over time (Fig. 3C). The transcription patterns also indicated changes in the composition of the nifH mRNA pool between the studied time points in each treatment. For example, band 9 was a dominant band in most transcription patterns from the early (12 h and 15 h) samples, but was weak or absent from later samples, which were in turn dominated by bands 5–8 (Fig. 6).
Phylogenetic analysis of nifH clones
For most of the bands of potentially active phylotypes (Fig. 6, bands 1, 2, 5–9) from the nifH-g1 amplification and some examples of potentially inactive phylotypes from nifH-g1 (Fig. 6, bands 3 and 4) and nifH-b1 (Fig. 4, bands A and B) amplification, one or more clones with identical DGGE migration behaviour were obtained and sequenced. The results confirmed our previous findings that the nifH-g1 primer set amplifies mostly, but not exclusively nifH from Azotobacter. All clones that co-migrated with bands of potentially active phylotypes (Fig. 6) were found to cluster close to published Azotobacter chroococcum and A. vinelandii nifH sequences (Fig. 7). Six of the clones, representing bands 6–9, coded for an amino acid sequence identical to a published NifH sequence of A. chroococcum (GenBank Accession No. AAA22140). Clones representing bands 10 and 11, which appeared only in treatment V, coded for an identical amino acid sequence that was most similar to an A. vinelandii NifH sequence (M11579) (Fig. 7). A clone from treatment VIII that yielded a band migrating between band 10 and 11 (labelled band 10b), and which could not be matched to a visible band in the original pattern, clustered closest to a different A. chroococcum strain (X03916). Three clones representing the non-transcribed bands 3 and 4 clustered with nifH sequences of Paenibacillus and environmental clones or with Methylobacter species (Fig. 7). Two clones sequenced from the nifH-b1 amplification (A and B) were most closely related to a group formed by environmental clones and Geobacter metallireducens.
To supplement the nifH-based data we performed an analysis of the soil bacterial community in 6 h and 24 h samples using 16S rDNA PCR and DGGE (Muyzer et al., 1993). The patterns obtained from 6 h samples and 24 h samples without active nitrogen fixation were complex, containing many poorly resolved bands (not shown). In contrast, the patterns of 24 h samples from treatments with nitrogenase activity were dominated by two distinct bands that were cut from the gel and sequenced. The sequence (160 bp) obtained from one dominant band (AY795564) had 100% sequence similarity to various Bacillus 16S rDNA sequences (e.g. Bacillus litoralis, AY608605), the other band co-migrated with the band from the A. vinelandii control strain; however, the obtained sequence (AY795565) showed 100% similarity to Paenibacillus sp. DSM 1352 (AJ345017), indicating co-migration of bands.
Model system and methods used
The experimental set-up in this study was designed to simulate the conditions that influence activity and establishment of bulk soil diazotroph communities when they come into direct contact with root exudates of a growing plant. In this study we used high substrate loads (15 mg g−1 dry soil ). Quantitative measures of root exudation are usually given as fluxes or as a percentage of plant photosynthetic carbon fixation (Marschner, 1995) and concentrations of dissolved organic matter within millimetres from the root surface are about 25–30 mg l−1 (Wenzel et al., 2001), considerably lower than the concentrations we used. However, little is know about the substrate concentrations very close to the root surface. Considering the high concentrations of solutes, e.g. in phloem sap (generally 15–20% dry matter, up to 90% of this as sucrose; Marschner, 1995), it is not unrealistic that high substrate concentrations occur naturally in the submillimetre zone around the root surface. Previous studies have employed continuous loading with dilute artificial root exudate mixtures to simulate the effect of continuous exposition on the microbial community (Griffiths et al., 1999). However, exudation is highest in apical root zones (Marschner, 1995) and thus maximum exudate concentrations are only experienced for a short time in a given volume of soil. Thus, a pulse addition of substrate may be an adequate model to describe the nutrient conditions a microbial community experiences near the surface of a growing root.
In the present study we have for the first time applied methods for simultaneous DNA and RNA extraction, nifH reverse transcription and PCR amplification developed previously (Bürgmann et al., 2001; 2003; 2004) to study activation of a complex soil diazotroph community. By directly comparing changes of the nifH gene pool (Fig. 3) with the analysis of nifH mRNA (Figs 5 and 6) and nitrogenase activity (Fig. 2C), we were able to identify potentially active and potentially inactive diazotrophic phylotypes. Successful amplification of nifH mRNA from soil has been reported in a few recent publications (Burke et al., 2002; Brown et al., 2003), but the interpretation of this type of data remains difficult because studies that directly relate activity, gene pool and transcription have so far been limited to simplified one-species model systems (Bürgmann et al., 2003). In the present study we could directly relate nifH transcription and nitrogenase activity under conditions that simulate those found in rhizosphere soil.
Activity of potential diazotroph microorganisms in soil
A shift in a community fingerprint as seen for the nifH-g1-amplified phylotypes in Fig. 3B has to be interpreted with caution due to the possibility of PCR and DGGE bias (Sekiguchi et al., 2001). Nevertheless, the apparent identity between the DGGE bands of nifH-g1 RT-PCR products (transcription fingerprint) and the bands of potentially active phylotypes identified in the nifH-g1 PCR product (community fingerprint) is intriguing (Fig. 6). This observation is most easily explained by assuming that expression of nitrogen fixation (as indicated by the detection of nifH transcripts) allowed these organisms to increase their population relative to those of the potentially inactive phylotypes. A population increase of adapted diazotroph species can be expected in the nitrogen-limited environments created by the substrate additions, but independent quantitative methods would be necessary to confirm the DGGE results. In all other tested subgroups (Table 2) we found no evidence for active diazotrophs, indicating that under the experimental conditions only a fraction of the entire potential diazotroph community appears to have participated in nitrogen fixation. It should be stressed that the analysis of community change based on nifH PCR alone would not have provided conclusive evidence to link a phylotype to activity, as community shifts within potential diazotrophs were also observed in treatments that did not exhibit nitrogenase activity (e.g. in treatments IV and VI; Fig. 3B).
The nifH-g1 primer set was designed to amplify nifH genes of Azotobacter species (Bürgmann et al., 2004). Phylogenetic analysis of the actively nifH-transcribing phylotypes detected with nifH-g1 RT-PCR confirmed that these nifH sequences were all similar to published nifH sequences of Azotobacter species (Fig. 7). This finding is in agreement with the known ecology and physiology of Azotobacter strains, which are frequently found in soil, prefer a pH above 6, elevated O2 partial pressures and grow well on media containing 1–2% of various sugars (Becking, 1992).
As our analysis of the diazotroph community is limited by the available nifH primer sets, we cannot assume that Azotobacter strains were the only active nitrogen-fixing microorganisms. The analysis of bacterial 16S rRNA with DGGE showed increasing dominance over time of a band co-migrating with that of the A. vinelandii reference strain in the nitrogen-fixing treatments. However, sequencing of dominant bands also indicated the increase of members of the genera Bacillus and Paenibacillus in these treatments. Both genera contain various nitrogen-fixing species, but group-specific nifH primers for these groups are currently not available.
That only a fraction of the detected diazotroph community was involved in nitrogen fixation in our experiment may be an indication that many of the potential diazotrophs that are frequently detected in soil and rhizosphere by nifH gene analysis (Ueda et al., 1995; Piceno et al., 1999; Widmer et al., 1999; Hamelin et al., 2002) might not actively participate in nitrogen fixation. Pioneer studies employing nifH mRNA detection in termite gut (Noda et al., 1999), lake and ocean water (Zani et al., 2000; Zehr et al., 2001) and rhizosphere soil (Burke et al., 2002) have also indicated that a varying portion of the potential diazotrophs in a given system may not be active. The literature contains plenty of evidence that the composition of the potential diazotroph community (Widmer et al., 1999; Poly et al., 2001; Hamelin et al., 2002) and its physiological traits (Limmer and Drake, 1996) vary greatly with soil properties. It would therefore be interesting to study if different environmental conditions, e.g. low pH or reduced oxygen partial pressure, which would be less amenable for Azotobacter, would result in different species participating in asymbiotic nitrogen fixation.
We further noticed that different individual nifH-g1 pylotypes in the community fingerprint and also in the transcription fingerprint consistently emerged or dominated at different time points. This may indicate that phylotypes were reacting to nutrient addition with different lag times or maybe preferred different substrate concentrations. An understanding of these dynamics may be of importance for future attempts to make use of rhizosphere-competent asymbiotic diazotrophs as plant growth-promoting organisms (Gaskins et al., 1985; Okon et al., 1995; Dobbelaere et al., 2003).
Our results are also relevant with regard to the use of glucose or other substrates to measure ‘potential’ nitrogenase activities in soils (Limmer and Drake, 1996; Kravchenko and Doroshenko, 2003). The fact that under such conditions the active diazotrophs may be originally minor components of the diazotroph community, and that community changes can occur rapidly, stresses that such measurements should indeed be interpreted with caution (Alef, 1995).
Selective effects of substrates and artificial root exudates
While the DGGE analysis of 16S rRNA fragments does not by itself allow for a full analysis of bacterial diversity, the results strongly suggested that within 24 h a shift occurred from a highly diverse, even population (many light bands, intense background ‘smear’) to a population dominated by two main DGGE phylotypes. This is in agreement with previous findings showing the selective effect of rhizospheres (Marilley et al., 1998) and the observation that high exudate concentrations result in strong shifts of the microbial community (Griffiths et al., 1999).
Remarkably, none of the organic acid treatments (I, IV, VI) induced nitrogen fixation, despite the fact that both citric and malic acid were consumed and nitrate was depleted in a similar fashion as compared with sugar-containing treatments. This was not expected as, e.g. malate and acetate are known to be oxidized by Azotobacter species (Repaske et al., 1960). The DGGE analysis of nifH-g1 PCR products of the citric and malic acid treatments could indicate that some potential diazotrophs increased their population. The somewhat slower depletion of nitrate and the aerobic experimental set-up therefore only partially explains the lack of nitrogen fixation in these treatments (Christiansen Weniger and Van Veen, 1991; Merrick, 1992).
Overall, the population change observed by nifH-g1 and nifH-uni fingerprinting was remarkably similar in all treatments with sugars or sugar-containing artificial root exudate mixtures. The differences in the 24 h nifH-g1 DGGE patterns between the three acid treatments (I, IV, VI) and between the three artificial root exudate treatments (V, VIII, IX), respectively, may indicate that the population shift was not exclusively determined by the sugars that were present. The influence of root exudate composition and concentration on the diazotroph community deserves further investigation, as an understanding of the dynamic processes of activation and community establishment may be relevant for the question whether certain plant growth-promoting diazotrophs are specifically selected for by plants through alteration of the amount and composition of root exudates.
Soil and sampling
All experiments were carried out using ‘Pappelacker’ soil (sieved < 2 mm; texture: sandy loam, pH: 7.5, total carbon: 16 mg g−1, total nitrogen: 1.3 mg g−1; Bürgmann et al., 2003). This soil originated from grassland, an environment with documented nitrogen fixation by asymbiotic diazotrophs (Nelson et al., 1976; Boddey and Dobereiner, 1995; Hamelin et al., 2002). The soil had been stored moist at 10°C in the dark for several months and did no longer exhibit measurable nitrogenase activity before substrate addition.
In all experiments, substrates were dissolved in sterile distilled water and adjusted to pH 7.5 if necessary. The volume of solution added to soil was calculated to result in a soil water content equivalent to 60% of water holding capacity. A preliminary experiment with various sucrose concentrations showed that nitrogenase activity after 24 h increased with increasing sucrose concentrations up to 15 mg g−1 dry soil. Higher concentrations resulted in reduced activity after 24, but high activity after 48 h, which was negligible in samples with lower sucrose concentration. Based on this, all further experiments were carried out with substrate concentrations of 15 or 20 mg g−1 dry soil. Substrates were added to microcosms containing 100 g dry mass of soil, fitted with cotton plugs and incubated at 25°C in the dark after substrate addition. The reproducibility of the microcosms and methods was studied in a preliminary experiment. Treatments with 20 mg g−1 dry soil sucrose or malic acid, and a control with distilled water were performed in triplicate. Nitrogenase activity, nitrate concentration and diazotroph community structure were analysed 6 h, 24 h and 48 h after substrate addition. At sampling times, soil samples were taken with a surface-sterilized spatula. Ten grams were used in the nitrogenase activity assay, 10 g were frozen at −80°C for chemical analyses and 0.5 g were weighed into tubes pre-filled with glass beads and buffer for nucleic acid extraction. The main experiment was split into two series that were carried out 2 months apart and analysed separately. Due to time and handling considerations, microcosms could not be replicated. Treatments included three artificial root exudates, three sugars and three carboxylic acids (Table 1). Sampling and activity measurements as described above were performed 6 h after substrate addition and then every 3 h until 33 h.
Nitrogenase activity was determined on 10 g of moist soil using the acetylene reduction assay as described previously (Alef, 1995; Bürgmann et al., 2003). Linear regression analysis for calculation of ethylene production rates was performed using Microsoft Excel. The significance of the calculated rate was tested with the t-statistic of Pearson's r :
with r = Pearson's r and N = number of observations.
Available nitrogen and substrates
Available nitrate, carboxylic acids and sugars were determined after extraction of 1 g of soil with 10 ml of 0.01 M CaCl2 (Houba et al., 2000) for 1 h on an overhead shaker. Extracts were centrifuged at 4500 g for 15 min and the supernatant was transferred into a fresh tube. To inhibit microbial growth, 100 µl of chloroform was added and extracts were stored at −20°C until measurement. Nitrate and carboxylic acids were determined in 0.45 µm of filtered extracts on a DIONEX IC20 Ion Chromatograph equipped with a AS11-HC 4 mm column (Dionex, Sunnyvale, CA) as described previously (Kleikemper et al., 2002). Malic and succinic acid could not be separated, and are reported as cumulative values if both were present. Sugars were determined using a Sucrose/D-Glucose/D-Fructose enzymatic kit (r-biopharm, Darmstadt, Germany) according to manufacturer's instructions. Pseudo zero-order maximal rates of substrate degradation were calculated from the slopes of linear portions of substrate utilization curves by linear regression based on at least three consecutive measurements and were tested for significance as described for acetylene reduction. Exchangeable ammonium was extracted from 1 g of soil by shaking for 1 h with 5 ml of 2 M KCl solution (Alef, 1995). Ammonium was determined colorimetrically using a modified salicylate method (Wagner, 1969; ISO, 1984). Briefly, 1 ml of extract, 4 ml of distilled water, 120 µl of reagent A (0.2% sodium nitroprusside, 17% Sodium salicylate in water) and 120 µl of reagent B (16% trisodium citrate, 2.25% sodium hydroxide and 0.2% dichloroisocyanuric acid sodium salt in water) were mixed and incubated for 20 min at 60°C. Absorption was determined on a Carey 50 spectrophotometer (Varian, Zug, Switzerland) at 690 nm.
Nucleic acid extraction and molecular fingerprinting
Nucleic acid samples were extracted from samples taken 6 h after the beginning of the beginning of the experiment (no nitrogenase activity in any sample), from the sampling just before and after the onset of nitrogenase activity in active samples (after 12 h and 15 h for series 1 and after 15 h and 18 h for series 2), and after 24 h (high nitrogenase activity in all active samples). Total nucleic acids were extracted from 0.5 g of soil as described previously (Bürgmann et al., 2001). One aliquot of each extract was used for DNA amplification. RNA was purified from a second aliquot using DNase digestion and acid phenol extraction in an RNA extraction buffer (Cheung et al., 1994; Bürgmann et al., 2003).
nifH-specific PCR was performed using primers (Table 2) and PCR protocols described previously (Bürgmann et al., 2004) with minor changes. For analysis by DGGE the forward (forB) primers of primer sets nifH-g1, nifH-b1 and nifH-c1 contained an additional GC-rich sequence at the 5′ end (Muyzer et al., 1993) (indicated by the ‘-GC’ suffix). As amplifications with the GC-clamped primer sets were less sensitive, reaction conditions were adjusted to improve sensitivity. The nifH-b1-GC amplification was performed as a booster-PCR with increased MgCl2 concentration (1 mM) on PCR product pre-amplified using nifH-b1 primer set and the stringent protocol described previously (Bürgmann et al., 2004). Amplifications with nifH-g1, nifH-c1, nifH-a2 and corresponding GC-clamped primer sets were performed directly on soil nucleic acid extracts. Due to the use of a different PCR buffer (Invitrogen AG, Basel, Switzerland), PCR conditions for nifH-g1 were re-optimized as shown previously (Bürgmann et al., 2004) to an annealing temperature of 66°C and 1.2 mM MgCl2. Amplifications with nifH-uni and nifH-f1 primer sets were performed as semi-nested PCR reactions. Reverse transcription with AMV reverse transcriptase (Promega, Madison, WI) was optimized from the method described previously (Bürgmann et al., 2003) by adding 0.1 mg ml−1 bovine serum albumin to the reaction and extending the 42°C incubation step to 90 min. These changes eliminated the need for dilution or additional purification of RNA extracts before reverse transcription. Fragments of the bacterial 16S rRNA genes were PCR-amplified for DGGE analysis using primers 341F GC and 534R as described previously (Muyzer et al., 1993).
Polymerase chain reaction products amplified with GC-clamped primers were subjected to DGGE (Muyzer et al., 1993) in 10% acrylamide gels with a denaturant gradient of 35–60%. Equal amounts of PCR product (estimated from band intensities) were loaded in each lane. Gels were run at 50 V for 15 min and at 200 V for 5 h at 60°C in a DCode electrophoresis system (Bio-Rad Laboratories, Hercules, CA) and stained with GelStar (Cambrex, East Rutherford, NJ). Gels were documented and analysed using a GelDoc gel documentation system and QuantityOne software (Bio-Rad, Hercules, CA). Only relative band intensities were considered to deduce potential population changes from DGGE patterns. Identical migration behaviour of bands from PCR and RT-PCR DGGE was verified by running samples from both experiments side by side.
Restriction fragment length polymorphism (RFLP) analysis was performed by overnight digestion of PCR product with HaeIII followed by electrophoresis in 12% bis-acrylamide gels as described previously (Widmer et al., 1999; Bürgmann et al., 2004). Staining, documentation and analysis were performed as described for DGGE.
Cloning and sequencing
Cloning of nifH fragments. Non-GC-clamped PCR products from the following treatments and sampling times were selected and cleaned using the QIAquick PCR purification kit (Qiagen AG, Basel, Switzerland): nifH-g1 PCR product: V 6 h, V 24 h, and VIII 24 h nifH-b1 PCR product: VIII 24 h. Polymerase chain reaction products were cloned into the pGEM-T easy cloning vector (Promega) according to the manufacturer's instructions. Clone libraries were screened by transferring cells from white colonies to PCR tubes containing the appropriate GC-clamped nifH primers and PCR mix as described above. Polymerase chain reaction products from nifH-positive clones were subjected to DGGE profiling. At least one clone representing each DGGE phylotype was selected for sequencing. Co-migration of clone bands with bands in community fingerprints was further verified by performing DGGE with mixtures of clone amplicons side by side with original community fingerprints. Plasmids were prepared with the Wizard SV miniprep kit (Promega) according to manufacturer's instructions.
Cloning of bacterial 16S rDNA fragments. For cloning, bands were cut from the DGGE gel, and DNA was eluted in sterile water at 80°C for 1 h. Eluted DNA was PCR-amplified from 10 µl of supernatant as above, but using primer 341F without a GC-clamp. Cloning and plasmid isolation were performed as described above. All sequencing was performed by Microsynth (Microsynth GmbH, Balgach, Switzerland).
Cloned nifH sequences were translated into their corresponding amino acid sequences and aligned with our existing database of published NifH sequences of cultured organisms and environmental clones (Bürgmann et al., 2004). Phylogenetic analyses were performed as described previously (Bürgmann et al., 2004) on a 96-amino-acid residue fragment using the phylip package (version 3.5c) (Felsenstein, 1989). The derived trees were edited in arb (Ludwig et al., 2004). Bacterial 16S rDNA fragment sequences were subjected to blast searches to find the closest matching sequences in the public database.
Nucleotide sequence accession numbers
The partial nifH sequences have been submitted to the GenBank database under Accession No. AY684103–AY684121, the partial 16S rRNA gene sequences under Accession No. AY795564 and AY795565.
This research was supported by grants from ETH Zurich, which is gratefully acknowledged.