The protocol used in the present study was a long-term mesocosm experiment where the microbial succession around live barley roots and subsequent decomposing roots was assessed after seed coating with either the antagonistic strain Pseudomonas fluorescens DR54-BN14 or the fungicide imazalil. Four diversity measures were used: community level physiological profiles (CLPP), Bacteria-specific polymerase chain reaction-denaturing gradient gel electrophoresis (PCR-DGGE), actinomycete-specific PCR-DGGE and phospholipid fatty acid (PLFA), as well as total cell counts, colony-forming units (CFU) and culturable spore formers, and spore counts of the Bacillus cereus group. Analysis of non-treated plants provided a baseline description of the natural microbial succession from which effects of the treatments could be evaluated. A microbial succession occurred both in the rhizosphere and around decomposing roots, shown with all three diversity measures. A clear response to root death was found, and a clear distinction between root tip and root base samples. Using the recommended concentration of imazalil and a realistic number of DR54-BN14 for seed coating, transient, initial effects of both treatments on the microbial communities were observed at the root base with the PLFA analysis only. The lack of lasting significant side effects of DR54-BN14 is in agreement with an initial fast reduction in culturable DR54-BN14.
Microbial inoculants with antagonistic properties towards soil- and seedborne plant pathogens of agricultural crops have a potential to replace chemical pesticides that can be harmful to the environment and human health. Ongoing research is concerned with the potential risks of introducing microorganisms with antagonistic properties into the environment , and one of the major concerns is the effect of the antagonist on microbial diversity and function in soil. The microbial community is responsible for biological processes that are necessary for maintaining a healthy and fertile soil.
Effects from seed coatings with either a bacterial strain or a fungicide are probably restricted to certain microhabitats around the plant root. The rhizodeposition differs between young and older parts of the root, resulting in differences in the microbial diversity and activity in different microhabitats along the root [2–4]. In addition, during ageing of the plant, a natural succession of the rhizosphere microbial community also takes place . When the root dies, the rhizosphere is changed into a zone of decomposition, potentially changing the rhizosphere microbial community structure dramatically. Furthermore, it is possible that an added antagonistic strain can benefit and proliferate because of this change in habitat conditions.
Several Pseudomonas strains have been found to be promising biocontrol agents against phytopathogenic microorganisms, and studies on non-target effects have been conducted. Pseudomonas fluorescens CHA0 did not affect the population structure of bacterial isolates that were characterised by ARDRA on 10 days old cucumber roots in microcosms . Gagliardi et al.  found a long-term effect on microbial community diversity by a genetically modified Pseudomonas chlororaphis in a microcosm experiment, using the community level physiological profile (CLPP) and fatty acid methyl ester (FAME) methods. On the other hand, no effects were found with a Pseudomonas putida strain in potato rhizosphere using Bacteria-specific DGGE , or with a Pseudomonas aureofaciens strain measured by CLPP and cellulose decomposition . Most studies showed only limited transient non-target effects of antagonists but it might be organism dependent, and therefore so far all antagonistic strains must be evaluated individually before being released into the environment.
P. fluorescens DR54-BN14 is an antagonistic bacterium, originally isolated in Denmark from a sugar beet rhizosphere, and effective towards preemergence damping-off disease caused by Pythium ultimum and Rhizoctonia solani in laboratory experiments [10,11]. An antifungal compound viscosinamide, isolated from DR54-BN14 cultures, is believed to be the agent responsible for biocontrol properties of this strain [11,12]. Possible non-target effects of DR54-BN14 on the indigenous Pseudomonas and actinomycete populations in barley rhizosphere have been investigated , but possible effects on the whole microbial community have not yet been published.
Imazalil is the most commonly used seed-coat fungicide in Denmark to protect seedlings of winter and spring barley against preemergence damping-off disease. To our knowledge, there are no published reports on effects of this fungicide on the soil microbial diversity. Thus we chose to compare the effect of the antagonist with the effect of this chemical counterpart.
In this study the succession of the microbial community at the root base and root tip of barley grown under controlled environmental conditions in mesocosms was followed for 50 days. After removal of the above-ground plant material, the microbial community in the vicinity of the decomposing roots was followed for an additional 62 days. Three treatments were employed: seed coating with DR54-BN14 or imazalil and non-treated seeds.
We used several different diversity measures combined with cell counts and analysis of important groups of soil bacteria to detect succession and eventual side effects of the treatments. Among the numerous methods available, microbial diversity was measured as community level physiological profiles (CLPP), polymerase chain reaction-denaturing gradient gel electrophoresis (PCR-DGGE) of Bacteria and phospholipid fatty acid (PLFA) profiles. Besides these broad diversity measurements, total colony-forming units (CFU) and direct cell counts were performed, and measurements on two important groups of soil bacteria: spore formers, including the Bacillus cereus group, detected by CFU counts, and diversity of actinomycetes detected by specific PCR-DGGE. The counting of B. cereus group members was included as preliminary experiments had indicated this group to be sensitive of DR54-BN14. Actinomycetes were included as they are important members of soil microbial communities, especially around decomposing roots . The measures reflected the culturable as well as the non-culturable parts of the microbial communities. This combination of methods, in addition to previously published reports, allowed us to thoroughly evaluate any side effects and evaluate the methods for future studies of side effects of antagonistic bacteria.
2Materials and methods
The mesocosm set-up was as described in Thirup et al. . In brief, the soil used was a sieved sandy loam soil from the Royal Veterinary and Agricultural University experimental fields in Høje Taastrup, Denmark, with the following characteristics: coarse sand: 69%; fine sand: 20%; silt: 3.3%; clay: 4.0%; organic matter: 4.4% (dry wt.), and water holding capacity: 18.9% (dry wt.). The soil was distributed in non-transparent plastic tubes, 6 cm in diameter and with a length of 35 cm for the first four samplings and 70 cm for subsequent samplings. The soil water content was adjusted to 80% water holding capacity, and regularly watered to weight to ensure stable water content.
The mesocosms were incubated in a climate chamber with a cycle of 16 h light at 15°C, and 8 h dark at 10°C. Barley seeds were sown, either non-coated or coated with imazalil or P. fluorescens DR54-BN14 (5×107 CFU per seed). DR54-BN14 was a mutant of the wild-type DR54, chromosomally marked with a constitutively expressed green fluorescent protein (GFP) gene and kanamycin resistance . The strain had the same root colonisation ability as the wild-type . Imazalil in the form of Fungazil A (Cillus, Herlev, Denmark) was sprayed on seeds at the recommended dose level (1 ml kg−1 seeds) by using a type HEGE 11 rotator (Hans Ulrich Hege Maschinenbau, Waldenburg, Germany). Two plants were allowed to grow per mesocosm, except for the first sampling event at day 4, where six plants were used to ensure enough rhizosphere soil for the assays. After sampling at day 50, plant shoots were removed to mimic harvest, while the roots were left and allowed to decompose in the soil.
Sampling was performed 4, 7, 10, 14, 21, 35, 50, 63, 77, 91 and 112 days after sowing, but not all analyses were done on every sampling occasion (see below). On each sampling event three replicate mesocosms from each of the treatments were destructively sampled. At all sampling days rhizosphere samples from the upper 5 cm of the root system (root base) were collected, and at days 14, 35, 50, 63 and 91 rhizosphere samples from the root tips were also taken. Root tips were here defined as the lower 5 cm of the root system and in the present experiment this was between 35 and 70 cm below surface, depending on the plant age. Rhizosphere samples consisted of ca. 6×5 cm root with adhering soil (approximately 0.5 g dry soil per 30 cm root). Bulk soil from mesocosms with untreated seeds were sampled (ca. 0.5 g dry wt.) on days 0, 10, 21, 50 and 112. As roots were present in the mesocosm, the bulk soil sampled may have been increasingly affected by the rhizosphere. Therefore, only plate counts and acridine orange direct cell (AODC) measures were done on these samples at day 10 and onwards. Both rhizosphere and bulk soil samples were prepared in 6 ml phosphate buffer  and used for plate counts, direct cell counts, Ecoplate inoculation and PCR-DGGE. Parallel, but larger (2–3 g soil dry wt.), samples were taken and analysed for their content of phospholipid fatty acids. All samples of soil and roots, as well as roots alone, were weighed, and the actual water content of the soil in each tube was measured. Plant weight was monitored at days 10, 21, 35 and 50 and showed an increase, but no difference between treatments (data not shown).
The number of P. fluorescens DR54-BN14 was counted on solid Luria–Bertani medium with kanamycin (50 μg ml−1) and delvocid (50 μg ml−1 delvocid, consisting of 50% natamycin and 50% lactose), to inhibit growth of bacteria other than DR54-BN14 and fungi , respectively. Appropriate dilutions were drop-plated (5×10 μl) and incubated at 25°C for 1 day. GFP positive colonies found by microscopic inspection were counted. Samples from the untreated and imazalil-treated plants were checked for contamination with P. fluorescens DR54-BN14.
Total number of culturable bacteria was counted after spread plating on Winogradsky agar (per l: 5.0 g K2HPO4, 2.5 g MgSO47H2O, 2.5 g NaCl, 0.05 g MnSO41H2O, 0.0025 g FeSO47H2O, 18 g agar) with 25 μg ml−1 natamycin and incubating for 4 weeks at 20°C.
Spores from the B. cereus group and total spore formers were counted after spread plating on T3 medium  with 50 μg ml−1 delvocid. 50 μl of appropriate dilutions, which had been heated to 65°C for 30 min, were plated and incubated for 3 days at 20°C. Colonies of the B. cereus group were recognised on the basis of their morphology. By this procedure >73% of the culturable spores of the B. cereus group are counted (B.M. Hansen, personal communication). For all plate counts, three replicate plates were counted per sample.
2.4Acridine orange direct cell count (AODC)
Bacteria in a 10−2 dilution of the original soil suspension were fixed by adding formalin to a final concentration of 4% and stored at 4°C. AODC  was performed on samples from the root base and tip at days 4, 14, 50 and 91, and bulk soil samples from days 0, 50 and 112.
2.5Community level physiological profile (CLPP)
The CLPP of the microbial communities were analysed by the BIOLOG® assay (BIOLOG, Hayward, CA, USA) by incubation of soil slurries in Ecoplates®. Ecoplates were inoculated with 150 μl well−1 diluted (10−2) soil suspension, resulting in inoculation of 0.6–2.2×104 CFU well−1. One Ecoplate with 31 substrates in triplicate was inoculated per sample. OD590 was measured after 3 days incubation at 20°C, with a microtitre plate reader (EL 340, Struers, Denmark). The difference in inoculation density is accounted for by calculation of the average well colour development according to Garland and Mills . CLPP was not measured at days 10, 21 and 77, and on bulk soil samples only at day 0.
The genetic diversity of bacteria was measured by running denaturing gradient gel electrophoresis (DGGE) of PCR products with a GC clamp. This procedure separates DNA differing in nucleotide sequence . DNA was extracted and purified from the soil samples with the fast soil purification kit (BIO 101, Vista, CA, USA) in accordance with the manufacturer's instructions . Two PCR-DGGE analyses were performed: one with the Bacteria-specific primers F341GC and R534  and one with the actinomycete-specific primers F243 and R513GC . DGGE was not performed on day 77 and the bulk soil samples were only analysed at day 0. Of the root tip samples only the non-treated were analysed and only with the Bacteria-specific PCR-DGGE.
Bacteria-specific PCR-DGGE: Each PCR tube contained a total volume of 25 μl, with 17.3 μl of DNase-free water, 2.5 μl 10×PCR buffer (Boehringer Mannheim), 2.5 μl bovine serum albumin (10 mg ml−1, Sigma), 1 μl of deoxynucleoside triphosphate (10 mM), 0.25 μl (0.05 mM) each of primers F341GC and R534 (DNA Technology A/S, Århus, Denmark), 0.20 μl Taq DNA polymerase (Boehringer Mannheim) and 1 μl sample as template. The PCR conditions were as follows: 5 min at 94°C; 30 cycles of 1 min at 94°C, 1 min at annealing temperature (four cycles 71°C, four cycles 70°C, five cycles 69°C, five cycles 68°C, four cycles 67°C, four cycles 66°C, four cycles 65°C), 2 min at 72°C; 6 min at 72°C; and final cooling at 4°C. The size and purity of PCR products were checked on 1.5% agarose gels. DGGE was performed by using 8% acrylamide gels (ratio of acrylamide to bisacrylamide, 37.5:1) with a 40–60% denaturant gradient, where 100% was defined as 7 M urea plus 40% formamide . 15 μl PCR product plus 4.5 μl loading buffer III  was loaded per sample. To enable comparison between gels, a marker consisting of mixed PCR products from three strains, Streptomyces lavendulae DSM 40069T, Micrococcus luteus DSM 20030T and Cytophaga arvensicolor DSM 3695 was run in three lanes on each DGGE gel. To localise the position of DR54-BN14, PCR product of this strain was run in a separate lane. The gels were electrophoresed at 60°C/150 V for 3.5 h by using 1×TAE buffer (Bio-Rad), and the D Gene System (Bio-Rad). The gels were silver stained and dried as described by Heuer et al. . DGGE gels were digitised as described in Christoffersen et al. . The number of bands and their position on the DGGE gel were analysed and dendrograms were created based on Dice coefficient of similarity and the UPGMA method by Dendron® (Solltech Inc., Oakdake, IA, USA).
Actinomycete-specific PCR-DGGE: Each PCR tube contained a total volume of 100 μl, with 72 μl of DNase-free water, 9.8 μl 10×PCR buffer (Boehringer Mannheim), 9.8 μl bovine serum albumin (10 mg ml−1, Sigma), 4 μl of deoxynucleoside triphosphate (10 mM), 0.5 μl (0.1 mM) each of primers F243 and R513GC (DNA Technology A/S, Århus, Denmark), 0.4 μl Taq DNA polymerase (Boehringer Mannheim) and 3 μl sample as template. The forward primer F243 was specific for the actinomycetes (but see below), while the reverse primer R513GC targeted all actinomycetes in addition to other groups. The PCR conditions were as follows: 8 min at 95°C; 35 cycles of 30 s at 95°C, 30 s at 65°C, 30 s at 72°C; 6 min at 72°C; and final cooling at 4°C. The size of PCR products was checked on 1.5% agarose gels, and the PCR product cleaned with QIAquick PCR purification kit (Qiagen GmbH, Hilden, Germany) and eluted in 30 μl of the elution buffer in the kit. Loading and DGGE were performed as for Bacteria-specific PCR-DGGE with a 50–65% denaturant gradient. A marker consisting of PCR products of the four strains Corynebacterium glutamicum DSM 20300, Gordonia terrae DSM 43249, Streptomyces albus DSM 40313 and Rhodococcus ruber DSM 43338 was used. The gels were electrophoresed at 60°C/150 V for 4 h. Gel staining, drying and band analyses were done as already described.
When the PCR protocol was run on DSM strains with known sequences with zero, one, two or three mismatches to primer F243, strains having up to two mismatches were amplified. Therefore, the specificity of primer F243 was checked in February 2001 on the Ribosomal Database Project (RDP) database . If two mismatches were accepted (not in the 3′-end), only genera in the group of high GC Gram-positive matched, though some members of this group (half of the genera in the ‘Arthrobacter subdivision’, including Arthrobacter/Micrococcus, the Radiotolerans, Atopobium and Acidimicrobium groups, and a few genera throughout the remaining actinomycete groups) had at least three mismatches. Because of the low specificity of the primer F243, the PCR-amplified genera were cloned and sequenced as follows: A PCR product from the control treatment after 91 days was cloned using the TOPO TA Cloning Kit (Invitrogen, Carlsbad, CA, USA). Clones containing the entire insert were identified by PCR, using primers F243 and R513GC as described above. The inserts of 49 clones were PCR amplified using primers M13R and M13F contained in the TOPO TA Cloning Kit, purified by QIAquick PCR purification kit, and sequenced at the sequencing facility at MWG Biotech, Ebersberg, Germany. Sequence quality was checked using Chromas 1.61. To find the closest relative, alignment of the actinomycete sequences was performed with NCBI BLAST against the GenBank database. Check of chimeric molecules was done against the RDP database.
2.7Phospholipid fatty acid (PLFA) analysis
The procedure for extraction of phospholipid fatty acids was modified after Frostegård et al. . Root pieces with adhering soil were agitated for 10 min in Teflon centrifuge tubes (Oak Ridge, Nalge Nunc Int.) with 6 ml citrate buffer (0.15 M, pH 4.0). After removal of the root pieces 2–6 g dry soil was scored for extraction purposes. Chloroform (7.6 ml) and methanol (15 ml) were supplied and the tubes were capped, agitated for 10 min and left overnight. After centrifugation (2500×g, 10 min) the supernatant was transferred to glass tubes. The pellet was subjected to an additional washing/centrifugation step and the two supernatants were pooled and dried under streaming N2.
Purification and derivatisation of the polar lipid fraction, the GC equipment used and its settings, were described in . The fatty acids were identified as described by Frostegård et al.  following the fatty acid nomenclature of Tunlid and White . PLFA analyses of rhizosphere and decomposing root zone were not performed on days 4, 10, 77 and 112. Principal component analysis (PCA) was performed on nine PLFAs, using PLFAs known to be indicative of specific microbial groups, instead of all the 26 detected PLFAs. This increased the resolution in the loading plots slightly, apparently because blur associated to the non-specific PLFAs was avoided (see ). The selected PLFAs were the following: Gram-positive bacteria (PLFAs i15:0, a15:0, i16:0, i17:0, a17:0) , Gram-negative bacteria (18:1ω7, cy17:0, cy19:0) , fungi (PLFA 18:2ω6,9) .
CFU and AODC data were log-10 transformed and analysed by a two-way analysis of variance (ANOVA) for effects of sampling time, treatments, and interaction between the two factors by the procedure ANOVA or GLM (general linear model) using SAS V8 for Windows (SAS Institute, Cary, NC, USA). To compare significant effects, least significant difference (LSD) values, at a 95% confidence level were used. CLPP data were analysed by principal component analysis (PCA) using SAS V8 for Windows. Principal components (PC) 1, 2 and 3 were analysed by a two-way ANOVA for the effects of sampling time and treatment. Furthermore, the turnover of selected carbon sources from the Ecoplates was analysed by a two-way ANOVA. The PLFA data were analysed by PCA after log-10 transformation using the Unscrambler software (CAMO ASA, Norway). From a PCA on PLFAs of root base samples PC1 and 2 were tested for significant effects of treatment at individual sampling days by a one-way ANOVA followed by a Tukey test using SigmaStat ver. 2.0 (SPSS Inc.).
3.1Fate of P. fluorescens DR54-BN14
In the root base rhizosphere of DR54-BN14-treated plants, DR54-BN14 CFU decreased by a factor of 100 during the first 14 days and eventually stabilised at approximately 105 CFU g−1 soil (Fig. 1). At the root tip the declining pattern was similar, though the CFU amounted to less than 0.5% compared to the root base. Post harvest the number declined further.
3.2Total bacterial CFU and AODC
Imazalil and DR54-BN14 had no effect on either the culturable heterotrophs or the total cell numbers (AODC) (Fig. 2). In general, AODC and CFU were higher at the root base than at the root tip and lowest in the bulk soil. Total cell numbers were stable over time, with no significant differences between sampling days. Total CFU were also generally stable, except for the root base, which showed a slow increase with time.
A principal component analysis of the carbon utilisation patterns of all samples showed a clear separation of rhizosphere and decomposing root samples, but no effects of antifungal treatments. Furthermore, root base and root tip samples were separated (Fig. 3). Two-way ANOVA of the root base data alone showed that both PC1 (accounting for 35% of variance) and PC2 (10% of variance) significantly depended on the time of sampling, but not on treatments (data not shown). PC3 (7% of variance) showed significant effect of the DR54-BN14 treatment at days 4, 7, 14 and 50. However, further analysis showed that this only reflected the presence and activity of DR54-BN14 in the Ecoplates, and not a changed physiological potential of the indigenous bacteria. Of the five substrates explaining most of the variation in PC3, four showed significant effect of the treatment in a two-way ANOVA. When DR54-BN14 could degrade the substrate (i-erythritol, glycyl-L-glutamic acid), an enhanced degradation was seen during the first two to five sampling occasions, and when DR54-BN14 could not degrade the substrate (D-glucosaminic acid, α-D-lactose), a depressed degradation was observed during the first two to four sampling occasions.
3.4Genetic diversity of Bacteria
The PCR-DGGE analysis generally showed complex band patterns. In particular bulk soil samples were complex with many indistinct bands, which could not be separated. Rhizosphere samples on the other hand showed an enrichment of some bands, with ca. 10 dominating bands. In general, this pattern changed over time approaching a complex band pattern from degrading roots (days 91 and 112), though different from bulk soil. As is also evident from the dendrograms, a natural variation between replicates of rhizosphere samples was found (Fig. 4). By visual observations of the gels, a strong band at the DR54-BN14 position was found in DR54-BN14-treated samples until day 14.
During the experimental period, the PCR-DGGE showed no systematic changes in band pattern caused by the treatments. This is illustrated in Fig. 4A, showing the similarity of bulk soil at day 0 and the three treatments at the root base at days 4 and 7, where the impact of the treatments was expected to be largest. As expected, the bulk soil samples showed a low similarity to the rhizosphere samples. The rhizosphere samples at days 4 and 7 differed, and this succession is further visible in Fig. 4B showing the succession of the microbial community in root base non-treated samples. Replicates clustered with similarities from ca. 97% (day 63) down to ca. 41% (day 112). The similarities between sampling times were fairly low (24–56%) with samples from the living rhizosphere and the decomposing root zone being mixed. A succession was, however, observed within the rhizosphere and the decomposing root zone, separately. The first and last rhizosphere samples (days 4 and 50) clustered with a similarity above 40% and the two intermediate samples (days 14 and 35) clustered with a similarity above 30%. Thus a major change occurred in the microbial community between days 4 and 14, and again between days 35 and 50. The decomposing root zone samples showed less similarity over time with a major change occurring during the initial decomposition between days 63 and 91. The similarity between replicates of the decomposing root zone decreased with time, which indicates individual succession in each decomposing root zone.
Root tip samples could be separated from root base samples, even though there were many common bands (data not shown). Root tip samples had a higher similarity over time than root base samples, but still rhizosphere and decomposing root samples were separated (days 14–50 vs. days 63 and 91) (Fig. 4C). After 91 days the similarity between the replicates was as low as 54%.
A PCA analysis on the phospholipid fatty acid profiles from the root base samples (Fig. 5A) showed an effect of the imazalil and DR54-BN14 treatments in the beginning of the growth period, as both PC1 and PC2 of the control at day 7 were significantly different from the two treatments. Also PC1 after 14 days showed significant differences between the antifungal treatments. Later, no systematic effect of either treatment was found, except at the final sampling at day 91, where PC2 of the imazalil treatment was significantly different from the control. Exclusion of the DR54-BN14 fatty acids from the PCA did not change this pattern. Succession over time is illustrated by the separation of early rhizosphere samples from older rhizosphere samples and decomposing root samples.
When analysing root tip samples and the corresponding root base samples, PC1 separated the samples according to time, and PC2 separated root base and root tip samples (Fig. 5B). This was also evident when plotting all treatments (data not shown).
3.6Genetic diversity of actinomycetes
The band patterns from the DGGE analysis were highly conservative for all samples analysed, including bulk soil samples, showing eight dominating bands and some indistinct bands (data not shown). Consequently, no effect of treatment or sampling time could be found for actinomycetes.
All the 49 sequences tested from the control sample at day 91 aligned to the group of actinomycetes, with 96–100% homology to database sequences. The resolving power of the 285 bp long sequence was however limited, as 35% of the sequences showed good homology to more than one actinomycete genus. Furthermore, when aligning the sequences with each other, many of the sequences appeared to be quite similar (data not shown). However, only seven sequences were found in duplicate, and two in triplicate. Good homology (>96%) was found to the following genera (excluding sequences with best homology to more than one genus): Streptomyces (29%), Rhodococcus (5%), Mycobacterium (11%), Nocardioides (8%), Kineococcus (8%), Frankia (5%), and Tsukamurella (3%). 29% of the sequences showed good homology to unidentified soil isolates. 43% of the sequences had the potential to be chimerical when checked in the RDP database . The suggestions from the RDP database for best fits of potential chimeric sequences, and the sequences with best homology to more than one genus suggest 15 additional genera to be added to the list of genera in our soil. These results suggest that a high diversity of actinomycetes was present in the soil, in spite of the rather simple DGGE pattern found. Furthermore, they suggest that the genus Streptomyces, which has been found to dominate the culturable actinomycete population in many soils , holds a dominating position when using a technique independent of cultivation.
3.7CFU of spore formers and the B. cereus group
The number of total spore formers (ca. 106 CFU g soil−1 (dry wt.)) and the B. cereus group showed a small increase (2–3 times) from day 4 to days 77–91 followed by a minor decrease with the B. cereus group constituting ca. 10% of the total spore formers. No significant effects of treatment or time were found for either spore type.
4.1Microbial succession in the rhizosphere and around decomposing roots
With methods dependent as well as independent of cultivation, root base and root tip microbial communities were found to differ. In the other report on the same mesocosm experiment, culturable Pseudomonas and actinomycetes were also found to differ along the root . This is in accordance with Baudoin et al.  using the CLPP method. Using Bacteria-specific DGGE, root tips and mature roots could be differentiated although many common bands were found [4,34].
Harvesting the plant shoots at day 50 made it possible to compare the microbial succession in the rhizosphere with the subsequent succession around the decomposing root. A succession of the microbial community in the rhizosphere followed by a shift and another succession around the decomposing root was observed using both the Bacteria-specific PCR-DGGE, CLPP and PLFA techniques. For the DGGE, though, the succession in the rhizosphere was less clear and mixed with the succession in the decomposing root zone. Microbial communities of different composition were thus dominant in the rhizospheres and around the decomposing roots. Specifically, the culturable and total populations of Pseudomonas and actinomycetes were found to oscillate in the rhizosphere and decomposing root zone . Similarly, De Leij et al.  found a change from fast to slower growing colonies in the rhizosphere during root ageing, a tendency that continued after harvest. Using the PLFA technique, Steer and Harris  also found shifts in diversity in a grass rhizosphere over time. Using DGGE patterns of Bacteria-specific PCR products, no succession was observed in the rhizosphere of barley  and chrysanthemum , but was found in the rhizosphere of strawberry, oilseed rape and potato . Using general as well as group-specific primers for TGGE, seasonal shifts were demonstrated in the rhizosphere of maize grown in tropical soil . The reason for these contrasting results of community development in rhizosphere, using the DGGE/TGGE technique, is not obvious. Differences in rhizodeposition and succession patterns of different plant species and soil types are possibilities.
The number of total spore CFU and B. cereus group spore CFU was not different between bulk soil, root tip and root base, as also found by Miller et al. . Also, no difference in actinomycete-specific DGGE patterns over time could be found, which was somewhat surprising, as the CFU of filamentous actinomycetes and actinomycete-specific DNA were found to oscillate . Almost identical actinomycete-specific primers for TGGE were able to differentiate between actinomycete population structure in an English and a Cuban soil  while limited seasonal diversity changes in rhizosphere of maize were reported . The clone library analysis of PCR-DGGE products performed in the present study questioned the sensitivity of the actinomycete-specific DGGE method and indicated high diversity.
Though all methods employed showed differences between root tip and base, only the community diversity estimates of CLPP, PLFA and Bacteria-specific PCR-DGGE revealed succession in rhizosphere and decomposing root zone. In an earlier report on the same study Pseudomonas and actinomycetes were also found to fluctuate. Together this indicates that specific changes occur within the communities during the succession, and that methods aimed at differentiating within the microbial community should be used.
4.2Fate of P. fluorescens DR54-BN14
The fate of DR54-BN14 is similar to an earlier report of a 100-fold reduction in CFU in the upper (near seed) rhizosphere during the first 14 days followed by a slower decline . The poor colonisation of root tips by DR54-BN14 is in agreement with [14,41]. Both studies measured an overall reduced activity of DR54-BN14 at the single cell level over time, indicating the effects of starvation and competition from the indigenous microorganisms. Indigenous Pseudomonas have been found to proliferate on dead barley roots , but with DR54-BN14 this was not the case, rather the number declined after harvest.
4.3Short-term effects of DR54-BN14 and imazalil
No effect of DR54-BN14 and imazalil was found on total CFU, AODC, CLPP, actinomycete-specific PCR-DGGE, CFU of spore formers and the B. cereus group. With Bacteria-specific PCR-DGGE, a strong band at the DR54-BN14 position was detected in DR54-BN14-treated root base samples until day 14, but no other changes in band patterns could be ascribed to the treatments. This is in agreement with previous studies applying Bacteria-specific PCR-DGGE, showing no effects of P. putida and Serratia grimesii in potato rhizosphere , and Alcaligenes faecalis in rice rhizosphere , except for early and transient antagonist-specific bands. Seed coating with imazalil was used as a positive control, to compare eventual effects of the antagonistic strain with the conventionally used fungicide. We found the effect of imazalil to be below detection limit and are not aware of other reports addressing the effect of this sterol biosynthesis-inhibiting fungicide on the microbial diversity in soil.
The PLFA analysis showed a clear but transient effect of both DR54-BN14 and imazalil on the structure of the microbial community, as the PCA clearly separated the controls from the treated samples at day 7. The effects of imazalil might be due to microbial enrichment following degradation of the compound or inhibition of part of the community. An exclusion of the DR54-BN14 fatty acids from the PCA did not change this pattern, indicating that it was not the presence of DR54-BN14 itself that caused the effect. The effect was probably caused by an initial high number of active DR54-BN14, displacing the indigenous bacteria and disappeared because of the low survival. DR54-BN14 was shown to initially displace indigenous pseudomonads , and this has also been observed by others .
We did not see any effect of the treatments by the CLPP method, which is in contrast to the effects on CLPP of introduced P. fluorescens CHA0  and P. chlororaphis strains .
None of the methods showed lasting effects of the introduced antagonistic bacterium P. fluorescens DR54-BN14, which is in accordance with previous reports [6,9,13,35]. The range of methods employed in this study of the microbial communities of rhizospheres and decomposing root zones shows a large resilience and robustness of the microbial communities towards introduced microorganisms. The effects of growing and decaying roots were in all instances larger than the effects of the introduced organism. The only method showing a transient effect was the PLFA analysis. Thus, this method deserves special emphasis in future laboratory and field studies of side effects.
L.T. and A.J. were both funded by the ‘Centre for Effects and Risks of Biotechnology in Agriculture’ under the Danish Environmental Research Programme (SMP2). We thank Hap Pritchard for helpful criticism on the manuscript, and Cillus A/S, Herlev, Denmark, for providing ‘Fungazil A’.