OUP user menu

Effect of on-field inoculation of Phaseolus vulgaris with rhizobia on soil bacterial communities

Darine Trabelsi , Alessio Mengoni , Haroun Ben Ammar , Ridha Mhamdi
DOI: http://dx.doi.org/10.1111/j.1574-6941.2011.01102.x 211-222 First published online: 1 July 2011

Abstract

The aim of this study was to assess the impact of inoculation of Phaseolus vulgaris with two indigenous rhizobia strains on plant growth promotion, nitrogen turnover processes, richness and structure of the Rhizobiaceae and total bacterial communities in the bulk soil. Both strains used induced a significant increase in nodulation and grain yield. Analysis of bulk soil fertility showed positive, negative and strain-dependent effects of inoculation on nitrate, phosphorus and ammonium, respectively. Terminal-restriction fragment length polymorphism profiling demonstrated that inoculation significantly increased the phylotype richness of the bacterial communities. No significant difference in richness between the strains used and no additive effect of co-inoculation were observed. However, differences between both inoculants and a clear additive effect of co-inoculation on heterogeneity were found. This work gives original insights into the effect of rhizobial inoculation outside the restricted rhizospheric area. Effects on bacterial structure and diversity are clearly sensed in the neighbourhood of 25 cm and in a limited time course. Both Alpha- and Gammaproteobacteria, together with Firmicutes and Actinobacteria, were enhanced by inoculation, No evidence of terminal-restriction fragment inhibition was found. However, it remains to be answered how the impact on taxonomic groups can be related to effects on functional capabilities of soil microbial communities.

Keywords
  • bacterial structure
  • soil fertility
  • richness
  • rhizobia
  • terminal-restriction fragment length polymorphism

Introduction

Analysing the consequences of biodiversity shifts on ecosystem functioning and stability has been a major concern in ecology (Wertz et al., 2007). However, the impact of a decline in soil microbial diversity on ecosystem sustainability remains largely unknown. This has been assessed for decomposition, which is ensured by a large proportion of the soil microbial community (Wertz et al., 2007), but not for more specialized and less diverse microbial groups.

The impact of biodiversity decrease on ecosystem functioning and, more particularly, on ecosystem stability following disturbance, have been evaluated in many studies (e.g. McNaughton, 1977; Tilman, 1996; van der Heijden et al., 1998). Ecosystem stability following disturbance comprises both resistance (i.e. the ability of the system to withstand disturbance) and resilience (the ability of the system to recover after disturbance) (Holling, 1973; Grimm & Wissel, 1997). Biodiversity, usually considered as species richness, may have an important role in the stability of ecosystem functioning (e.g. Li et al., 2010). Indeed, the presence of a large number of species that can respond differently to disturbance may provide insurance that minimizes the likelihood of large changes in ecosystem functioning after disturbance (Yachi & Loreau, 1999; Ives et al., 2000; McCann, 2000).

Inoculation of legume crops with nitrogen-fixing soil bacteria, collectively known as rhizobia, has been widely used to improve legume productivity in fields. However, problems of establishment of the introduced inoculants were frequently encountered (Triplett & Sadowsky, 1992; Toro, 1996). In such cases, the better-adapted indigenous populations tend to outcompete the inoculants for nodule formation and saprophytic survival. Thus, for inoculants to be effective, the strains used must not only be good nitrogen fixers but also be highly competitive. However, their potential ecological risks on microbial diversity should not be neglected. Therefore, in recent years there has been a growing interest in genes and transcripts coding for metabolic enzymes. Besides questions addressing redundancy and diversity, more and more attention is given to the abundance of specific DNA and mRNA in the different habitats. In recent studies, the quantification of bacterial genes encoding the nitrogenase reductase (nifH), ammonia monooxygenase (amoA) and nitrite reductase (nirK and nirS), as well as archaeal amoA genes within the nitrogen cycle was performed in rhizosphere samples (Hai et al., 2009). Results clearly demonstrated that effectiveness of rhizobial inocula is related to abundance of nifH genes in the late flowering phase of alfalfa (Babic et al., 2008). Moreover, other genes involved in nitrogen turnover had been affected by inoculation, for example higher numbers of amoA copies were observed during flowering when the more effective strain had been inoculated (Babic et al., 2008).

The symbiosis between rhizobia and legumes is supposed as the driving force for a positive nitrogen balance in the soil. However, this effect does not only depend on the elemental capacity to fix nitrogen, but also on the internal nitrogen turnover as well as on the rhizosphere (Babic et al., 2008). Therefore, investigation of other processes involved, such as the dynamics of nonrhizobial and rhizobial communities, is strongly needed. What happens to soil microorganisms when a rhizobial strain is introduced at high levels in the rhizosphere of legumes? Unsurprisingly, the answer is, it depends. In relation to the soil–plant–environment background, certain groups may be enhanced, whereas others may be inhibited, or the introduced bacteria may not affect population structure (Nacamulli et al., 1997; Schwieger & Tebbe, 2000; Bacilio-Jimenez et al., 2001; Dobbelaere et al., 2003).

The aim of this work was to evaluate the long-distance effect of the changes in the rhizosphere after on-field inoculation of Phaseolus vulgaris with two local selected rhizobial strains on soil fertility and Rhizobiaceae and total bacterial communities. Rhizobium gallicum strain 8a3 was previously isolated from the North of the country and was shown to be highly effective and competitive (Mrabet et al., 2005; Mnasri et al., 2007c). Ensifer meliloti (formerly Sinorhizobium meliloti) strain 4H41 is a highly salt-tolerant strain isolated from common bean nodules from South Tunisia (Mnasri et al., 2007b). This strain did not nodulate Medicago species but was effective in nitrogen fixation on its original host. Thus, it was proposed as a novel biovar (bv. mediterranense) inside E. meliloti (Mnasri et al., 2007b) and was shown to be more competitive and more effective in nitrogen fixation with common beans under conditions of water deficiency than the commonly used inoculant strain CIAT899 (Mnasri et al., 2007a).

Materials and methods

Experimental design

The two indigenous rhizobial strains used for field inoculation were previously isolated from common bean nodules from Tunisian soils as stated above. Strains were replicated from cultures stored at -80 °C in yeast-extract mannitol (YEM) liquid broth (Vincent, 1970) in the presence of 25% glycerol.

The on-field trial was carried out in an experimental plot of the Technical Centre of Potatoes located in the region of Manouba, North Tunisia (36°48'28?N, 10°6'4?E). The soil was nonsaline and slightly alkaline and had been used for potatoes, cereals and faba bean cultivation but had not been cropped for 1 year. The soil is a fluvisol with 55% clay, 20% silt, 25% sand, 2.5% humus, 1.2 g kg-1 total nitrogen and pH 8.5. The experimental design consisted of five treatments, noninoculated, nitrogen-fertilized and inoculated with E. meliloti 4H41, R. gallicum 8a3 or both strains (1/2, 1/2). The experiment was conducted in a completely randomized block design. Each treatment was subdivided into three randomized subplots cultivated with P. vulgaris cv. Flamingo. The subplots were 5 and 6 m in size. The sowing density was 20 seeds m-2 (10 and 50 cm). A margin of 3 m separated the different plots. The experiment was carried out under drip irrigation between 25 March and 10 June 2009. The daily average temperature was 21.1 °C and ranged from an average maximum of 26.3 °C and an average minimum of 16 °C.

The inoculants were grown to late exponential phase in YEM (Vincent, 1970) and diluted with well water 1 : 40 and then used to inoculate two-leaf stage seedlings by manually applying about 5 mL on each plant along the seedling line (approximately 108 bacteria per plant) using a watering can. Inoculation was conducted on 15 April. Weeds were removed manually. The nitrogen-fertilized control received the equivalent of 100 kg KNO3 ha-1 fractioned twice, at the two-leaf seedlings stage and the beginning of pod formation. All treatments received the equivalent of 4 kg ha-1 of chelated iron by foliage spraying on 20 April.

Nodule and biomass determination

At flowering stage (18 May), 30 plants, 10 from each subplot, were randomly collected from each treatment and used for nodule counting. Shoots were oven-dried at 70 °C for 48 h and weighed. At final harvest (10 June), 30 plants were collected from each subplot and used for grain yield determination. Statistical analysis was performed using the statistica software by the anova/manova program. The HSD Tukey test (P<0.05) was used for comparison of the means.

DNA extraction

Bulk soils were sampled at sowing (T0), flowering (T1) and grain harvesting (T2) from each subplot. Root-free samples (15 cm of topsoil) were taken with a soil corer at the middle of the rows (25 cm from each) in five replicates and checked to ascertain the absence of roots. The samples were then combined and homogenized and stored at -80 °C later for DNA extraction after shock-freezing in liquid nitrogen.

DNAs from each of the 45 samples (five treatments × three subplots × three time points) were extracted from 0.5 g soil using the fast prep extraction soil-DNA kit of Q-Biogene. The quality and quantity of the DNA extracts were checked on 1% agarose gel using the standard quantification marker of 1 kbp (Qiagen). DNA extracts were stored at -20 °C until use. As a preliminary analysis, extraction from 5 g of soil was performed and terminal-restriction fragment length polymorphism (T-RFLP) profiles were compared with those from 0.5 g. Results were identical using the analysis setup described below. For convenience, all samples were then extracted from 0.5 g of soil.

Chemical analysis of soil samples

The same soil samples collected at sowing (T0) and grain harvesting (T2) were also used for chemical analysis. Nitrate (NO3-), ammonium (NH4+) and available phosphorus (P2O5) were determined by colorimetry according to the method of Dabin (1965).

T-RFLP analysis of the Rhizobiaceae and total bacterial soil communities

The bacterial 16S rRNA genes were amplified using the universal bacteria specific primers 27f and 1492r as already reported using 20 ng of soil DNA as template (Trabelsi et al., 2009). The Rhizobiaceae 16S rRNA genes were amplified using the universal forward primer 27f and the Rhizobiaceae-specific reverse primer (RHIZ primer) (Tom-Petersen et al., 2003). The forward universal primer 27f was labelled with hexachloro-carboxyfluorescein and 6-fluoro-carboxyfluorescein for the bacterial and the Rhizobiaceae communities, respectively. T-RFLP was performed with RsaI and MspI restriction enzymes following previously described procedures (Mengoni et al., 2007).

Labelled restriction products were analysed by capillary electrophoresis on an ABI310 Genetic Analyzer (Applied Biosystems) using ROX 1000 (Applied Biosystems) as size standard. Analysis of T-RFLP profiles was performed following previously described procedures (Mengoni et al., 2007). Putative taxonomic assignment of terminal-restriction fragments (TRFs) was done using the MiCA3 server application (Shyu et al., 2007). TRFs <36 bp were not considered for analysis to avoid confusion with primer dimers. Derivative T-RFLP profiles of the different enzymes were combined and transformed to a binary vector (1, 0), in which presence or absence of each TRF was scored. To simplify the interpretation of TRFs, the triplicate reproducible (3/3 or 2/3) result was scored for each treatment. The vectors of binary profiles were then compared to compute the similarity matrix using the Jaccard index as implemented in the software ntsyspc 2.02 (Rohlf, 1990). The matrix of Jaccard similarity values was then used to construct an UPGMA dendrogram using the module present in the ntsyspc 2.02 (Rohlf, 1990).

Mean number of pairwise differences (?pi) were computed for treated and control samples with arlequin 3.5.1.2 software (http://cmpg.unibe.ch/software/arlequin3/) and used as a measure of intersample heterogeneity of bacterial community. As reported previously (Mengoni et al., 2009), the distribution of the variance of T-RFLP profiles within and among samples was tested by amova (Excoffier et al., 1992).

The past 2.02 software package (Hammer et al., 2001) was used to compute nonmetric multidimensional scaling (N-MDS) based on Jaccard distance. The effect on microbial community structure was tested using one- or two-way anosim (Clarke, 1993). N-MDS considers two parameters, the number of TRFs present and their relative size. The closer the scattered data points were to each other, the more similar the bacterial communities were with respect to TRF composition.

Results

Effect of inoculation on nodule number and grain yield

On-field inoculation of P. vulgaris with strains 8a3 and 4H41 showed that nodule number, shoot dry weight and grain yield of common bean were increased significantly comparing to the control plants (Table 1). The improvement in grain yield was more than twofold compared with the nitrogen-fertilized control, with a prevalence of strain 8a3. The noninoculated plants were poorly or not nodulated.

View this table:
1

Nodule number and grain yield of Phaseolus vulgaris cv. coco inoculated with Rhizobium gallicum strain 8a3 and/or Ensifer meliloti strain 4H41 under field conditions

TreatmentNodule number (per plant)Shoot dry yield (g per plant)Grain yield (g per plant)
Inoculated with R. gallicum 8a328a6.2a22a
Inoculated with E. meliloti strain 4H4117ab6.1a14b
Co-inoculated with 8a3 and 4H4112bc6.3a14b
Nitrogen-fertilized control0c6.0a10c
Noninoculated control1c4.9b6d
  • Each value corresponds to the mean of 30 replicates (10 from each subplot). Values of the same column with identical letters are not significantly different (P<0.05).

Effect of inoculation on soil fertility

Ammonium (NH4+), nitrate (NO3-) and phosphorus (P2O5) were quantified in the soil samples from the different treatments on sowing (T0) and at the end of plant cycle (T2). The results are summarized in Fig. 1. An initial variability in ammonium concentrations was observed between the different treatments due to small-scale heterogeneity in the subplots. No significant variation was observed between stages T0 and T2 for the noninoculated, the nitrogen-fertilized and the co-inoculated treatments. However, contrasting variations were found for the treatments inoculated with strains 8a3 and 4H41. Inoculation with strain 4H41 induced an increase in soil ammonium concentration, whereas inoculation with strain 8a3 induced a decrease.

1

Ammonium (NH4+), nitrate (NO3-) and phosphorus (P2O5) concentrations in the bulk soil of Phaseolus vulgaris cv. coco inoculated with Rhizobium gallicum strain 8a3 and/or Ensifer meliloti strain 4H41 under field conditions during sowing (T0) and grain harvesting (T2). Each value is the mean of three measures performed on the three replicated subplots. Values of the same panel with identical letters are not significantly different (P<0.05).

An initial variability in nitrate concentrations was also observed between the different treatments. At the end of the plant cycle (T2), the amounts of nitrate were below the detectable limit for the noninoculated and the nitrogen-fertilized treatments. However, for the inoculated treatments a residual nitrate is still detectable at stage T2.

The concentration of phosphorus increased in stage T2 for the noninoculated and the nitrogen-fertilized controls; however, it was unchanged for the three inoculated treatments.

Effect of inoculation on Rhizobiaceae and total bacterial diversity

T-RFLP analysis of bacterial and Rhizobiaceae 16S rRNA genes was carried out on total DNA extracted from soil samples. The yield of DNA extraction was estimated to be 1.7–2.8 µgDNA g-1 of soil. The recovered DNA migrated in agarose gel as fragments around 20 kb, with no significant smearing.

In total, 37 and 66 different TRFs were recorded for Rhizobiaceae and total bacteria, respectively, on the whole dataset of 45 samples. An average of 9–23 and 13–49 TRFs were detected per sample for Rhizobiaceae and total bacteria, respectively. Some TRFs were always found in the different subplots and during the whole life cycle.

At the beginning of the experiment (T0), the bacterial communities showed a similar number of ribotypes. During plant development, the number of the different TRFs did not change significantly (P>0.05) in the noninoculated and nitrogen-fertilized controls for both communities (Rhizobiaceae and total bacteria). In contrast, the ribotype richness of Rhizobiaceae and bacterial communities clearly and significantly (P<0.05) increased in the inoculated and co-inoculated samples (Fig. 2). Moreover, the total bacteria seemed to be more influenced by inoculation than the Rhizobiaceae community, with a two- to threefold increase in the number of TRFs. Interestingly, the co-inoculated samples showed a quite similar increase in the number of TRFs to samples inoculated with single strains. The richness of both Rhizobiaceae and bacterial communities was positively correlated (R2=0.9).

2

Ribotype richness, evaluated as number of different T-RFLP bands (TRFs), among T-RFLP profiles of Rhizobiaceae and bacterial 16S rRNA genes in the bulk soil of Phaseolus vulgaris cv. coco inoculated with Rhizobium gallicum strain 8a3 and/or Ensifer meliloti strain 4H41 under field conditions during sowing (T0), flowering (T1) and grain harvesting (T2). Each column is the average of the three replicated blocks. Values with identical letters are not significantly different (P<0.05). N, nitrogen-fertilized control; C, noninoculated control.

Concerning the inter-subplot variability, the inoculated and co-inoculated samples showed a high heterogeneity (that is, a higher number of pairwise differences) between subplots which increased from T0 to T1 and T2 (Fig. 3). Nitrogen-fertilized and noninoculated controls were more homogeneous, showing lower levels of inter-subplot variability.

3

Intersample heterogeneity of total bacterial community measured as mean number of pairwise differences (?pi) for the different treatments during sowing (T0), flowering (T1) and harvesting (T2). Each column is the average of the three replicated blocks. Values with identical letters are not significantly different (P<0.05). N, nitrogen-fertilized control; C, noninoculated control.

To determine the difference between the investigated treatments and replicated blocks inside treatments, N-MDS of all TRF frequencies was applied (Fig. 4). T-RFLP patterns obtained from the three different replicated blocks (subplots) of the same treatment and same plant development stage were similar and tightly clustered. The composition of the bacterial and Rhizobiaceae communities was significantly affected by inoculation treatments. N-MDS recognized some evidence for subgroupings of bacterial communities according to rhizobial inoculation, which was apparent within each plant stage development (Fig. 4). At each stage, N-MDS showed that bacterial communities of inoculated plots were homogeneous and similar. The second coordinate of N-MDS separated the bacterial community of the inoculated treatments at flowering stage (T1) from the final harvesting stage (T2). Similarly, the structure of the Rhizobiaceae community was primarily affected by rhizobial treatment, and there was also a secondary effect of plant stage development within each treatment. At each plant stage, N-MDS shows that Rhizobiaceae communities of the inoculated plots were heterogeneous, different and separated. The first coordinate of N-MDS separated the Rhizobiaceae community at the same stage for the different treatments (Fig. 4).

4

Multivariate statistical analysis of N-MDS showing distance (resemblance) in structure for the Rhizobiaceae (a) and total bacterial (b) communities in the bulk soil of Phaseolus vulgaris cv. coco inoculated with Rhizobium gallicum strain 8a3 and/or Ensifer meliloti strain 4H41 under field conditions during sowing (T0), flowering (T1) and grain harvesting (T2).

Putative taxonomic assignment of TRFs

Individual TRFs were assigned by the MiCA3 (Shyu et al., 2007) to distinct putative phylogenetic groups (Fig. 5). Four Rhizobiaceae TRFs originating from RsaI digestion could not be assigned to a known phylogenetic group, but all the remaining TRFs were assigned to putative phylogenetic groups. Several TRFs shared more than one phylogenetic group. The TRFs identified revealed differences in the relative bacterial population structures.

5a

Putative TRF assignation (MiCA3 application, Shyu et al., 2007) and dynamic of the Rhizobiaceae (a) and total bacterial (b) communities in the bulk soil of Phaseolus vulgaris cv. coco inoculated with Rhizobium gallicum strain 8a3 and/or Ensifer meliloti strain 4H41 during plant life cycle. To simplify the interpretation of TRFs, the triplicate reproducible (3/3 or 2/3) result was combined for each subplot. For total bacteria, only TRFs deriving from MspI are shown. The TRF bands corresponding to the inoculant strains are highlighted. N, nitrogen-fertilized control; C, noninoculated control.

Most TRFs were affiliated to Proteobacteria (Alpha-, Beta-, Gamma- and Delta-), followed by TRFs related to Bacillales, Actinobacteria and Acidobacteria. Some TRFs affiliated to Firmicutes (mainly Bacillales), Alphaproteobacteria (mainly Azospirillum) and Gammaproteobacteria (mainly Pseudomonas) were constitutively found in all the treatments and during the whole plant life cycle.

The expected TRFs associated to the inoculated strains, R. gallicum 8a3 and E. meliloti 4H41, were clearly identified in both soil communities at stages T1 and T2 (Fig. 5).

The matrix plot (Fig. 5) showed that the rhizobial inoculation had a time effect on the Rhizobiaceae community. Many TRFs were recorded only at the last stage (T2). These TRFs corresponded mainly to Mesorhizobium, Rhizobium and Rhodospirillaceae members. Nevertheless, some TRFs from the Rhizobiaceae community were clearly inhibited by nitrogen fertilization, including Rhizobium, Mesorhizobium, Ensifer and Agrobacterium.

For the bacterial community, 25 TRFs resulting from MspI digestion (data for RsaI digestion are not shown) were induced by inoculation (Fig. 5). These TRFs could be associated to a wide variety of genera from different phylogenetic groups, including Alpha- and Gammaproteobacteria, Actinomycetes and Firmicutes.

Discussion

Rhizobial inoculation-based agricultural systems are complex and subject to a wide range of management practices. As soil microorganisms support numerous ecosystem processes underpinning production, sustainability and environmental impacts, there exists potential to manage these processes to achieve further gain. For this, knowledge about the response of soil microbial communities is needed (Gupta & Ryder, 2003). This means reliably measuring and assessing management-induced shifts in microbial communities (phylogenetic) and functional levels must be in place. Such technology is now a routine and an established component of soil microbiology ecology (e.g. Wakelin et al., 2007) and is moving rapidly into larger capacity formats (Gao et al., 2007; He et al., 2007; Yergeau et al., 2007). Previous studies have shown that the bacterial communities were structurally different in the bulk soil, rhizosphere and rhizoplane habitats and the species richness in the bulk soil was obviously higher than in the rhizosphere (Graff & Conrad, 2005).

Inoculation-independent factors which could affect microbial community structure include soil type, crop species, environmental factors and anthropogenic effects, such as those caused by farming practices. However, in our experiment, these factors were similar for all treatments. Consequently, the changes observed in our experiment would be mainly ascribed to inoculation, through a direct effect exercised by rhizobia on soil community and/or an indirect effect mediated by the root system.

The resulting effects of rhizobial inoculation on soil microbial communities were addressed at two levels: (1) by comparing community diversity using similarity analysis of T-RFLP profiles and (2) by comparing the community structures. The on-field scale would give also answers about the resistance and the resilience of the bacterial communities to different conditions such as nitrogen fertilization and inoculation. The number of TRFs was used as index to measure richness of bacterial phylotypes. The context of the study was not to determine total microbial richness but rather to comparatively explore the response of phylotype diversity among treatments, an application considered entirely valid (Danovaro et al., 2007). Previous studies have shown that trifolitoxin antibiotic-producing strains of rhizobia dramatically reduce the diversity of rhizosphere microbial communities of field-grown P. vulgaris (Robleto et al., 1998).

In our study, we found that in addition to changing the overall composition of soil microbial species (16S rRNA gene genotypes), rhizobial inoculation also increased richness of the bacterial and Rhizobiaceae communities. We revealed that the bacterial community at the final harvesting stage (T2) in the inoculated treatments is distinct from the other stages and from the noninoculated and nitrogen-fertilized controls. The T-RFLP patterns of the inoculated treatments revealed an increase in the number of TRFs at flowering and mainly harvesting stages, with respect to the noninoculated and nitrogen-fertilized controls (Fig. 2). However, no additive effect was seen after co-inoculation with both rhizobial strains. No significant difference was observed between both strains as regards the number of TRFs. However, differences in heterogeneity between both rhizobial strains used and a clear additive effect of co-inoculation were observed (Fig. 3).

Results showed that members of Alphaproteobacteria (mainly Azospirillum), Bacillales and Pseudomonadaceae were constitutively present in all treatments and during the whole trial. Members of these genera are typical of soil environments and exhibit an aerobic-heterotrophic metabolism and play a crucial role in soil fertility and sustainability (Slepecky & Hemphill, 1992; Bashan et al., 2004; Rabie & Almadini, 2005; Saikia et al., 2007). Bacillus species were also reported to be dominant in the rhizosphere of barley (Normander & Prosser, 2000), chrysanthemum (Duineveld et al., 2001) and cycad plants (Lobakova et al., 2003), whereas Alphaproteobacteria were most frequently found in the rhizosphere of grass (McCaig et al., 1999). Members of these taxonomic groups have been observed as dominant populations in an agricultural soil (Ovreas & Torsvik, 1998). Schwieger & Tebbe (2000) showed that field release of S. meliloti strain L33 affected the structural diversity in the rhizosphere by reducing the number of members of the Gammaproteobacteria and increasing the number of members of Alphaproteobacteria. However, in our work, we found that both Alpha- and Gammaproteobacteria, together with Firmicutes and Actinobacteria, were enhanced by inoculation. No evidence of TRF inhibition caused by inoculation was found. Nevertheless, some TRFs from the Rhizobiaceae community were inhibited by nitrogen fertilization, suggesting inhibition of growth of rhizobia in the rhizosphere by nitrogen fertilization, which may be related to previous reports on the negative effect of nitrogen fertilization on nodulation (Streeter, 1988).

However, it is interesting to note that, although T0 soil chemical parameters were slightly different, a homogenous composition of bacterial communities was present. Moreover, inoculated and co-inoculated samples were more heterogeneous among subplots in bacterial community composition, and a clear trend of increase in the heterogeneity was present up to the development stage (T2). These data may suggest that the perturbation of the community due to biotic agents (inoculation) is higher than that due to the fertilization and that the evolution of the community in response to inoculants is someway more stochastic, suggesting that the introduction of exogenous bacteria in a community is likely to produce more long-term unpredictable effects than nitrogen supply. The consequences of these changes are not clear and need to be investigated.

The results of this study agree with previous reports (Mrabet et al., 2005; Mnasri et al., 2007a, c) about the effectiveness and the competitiveness of these selected inoculant strains. Regarding the fitness of the plant, we obtained not only higher nodule number and grain yields in the inoculated plots, but also a more vigorous plant development compared with the other treatments. The soil–mineral–nitrogen was analysed and showed that at the early plant development stage T1, ammonium concentrations were high in all samples. This might be due either to high mineralization rates at this time point (Mrkonjic-Fuka et al., 2008) or to a low ammonium demand of the plant. The simultaneously low nitrate concentrations at this stage might indicate a certain activity of ammonia-oxidizing bacteria (AOB). As most of the known ammonia-oxidizing microorganisms have an autotrophic life-style, high organic carbon concentrations can reduce their activity (Verhagen & Laanbroek, 1991; Strauss & Lamberti, 2000). Therefore, at sowing, the observed nitrate concentrations might be explained by the low number of AOB. Interestingly, an inoculant strain effect was detected, as in samples inoculated with strain 4H41 there was an increase in soil ammonium concentration, while those inoculated with strain 8a3 showed a decrease. This finding suggests that genomic variation between strains may have considerable variation in the produced effect. However, no data are available to address the physiological and genetic basis of such differences. Similarly, the response of the phosphorus concentration to inoculation was not significant in the inoculated plots, whereas it increased in the noninoculated and nitrogen-fertilized controls. It could be explained by the high demand of the symbiosis to phosphorus (Gentili & Huss-Danell, 2003).

In conclusion, we demonstrated that the soil microbial community is complex and dynamic; varying in composition between different compartments and that inoculation causes changes in the number and composition of the taxonomic groups. These changes could well lead to changes in beneficial soil functions such as nitrogen fixation or nitrogen-cycling bacteria. This work gives original insights into the effect of rhizobial inoculation outside the restricted rhizospheric area. Effects on bacterial structure and diversity are clearly sensed in the neighbourhood of 25 cm and in a limited time course. To our knowledge, this is the first report, using a more informative culture-independent approach, to demonstrate the direct response of bulk soil microbial communities to inoculation, including Rhizobiaceae and total bacterial communities.

However, it remains to be answered how the impact on taxonomic groups can be related to effects on functional capabilities of the soil microbial communities. The consequences of these changes are not understood. Further studies are needed to determine the metabolic potential of these microorganisms and their importance to soil ecosystems.

Acknowledgements

The authors wish to thank the Technical Centre of Potatoes (Saida, Tunisia) for assistance in field trials.

References

View Abstract