Dynamic of the genetic structure of bacterial and fungal communities at different developmental stages of Medicago truncatula Gaertn. cv. Jemalong line J5

Authors


Author for correspondence: Christophe Mougel Tel: +33 3 80 69 30 53 Fax: +33 3 80 69 32 24 E-mail: mougel@dijon.inra.fr

Summary

  • • The genetic structure of bacterial and fungal communities was characterized in the rhizosphere of Medicago truncatula Gaertn. cv. Jemalong line J5 at five developmental stages (three vegetative and two reproductive stages), and in three compartments (bulk soil, rhizosphere soil and root tissues).
  • • The genetic structure of microbial communities was determined by cultivation-independent methods using directly extracted DNA that was characterized by automated ribosomal intergenic spacer analysis (ARISA).
  • • Principal component analyses (PCA) indicate that, for all developmental stages, the genetic structure of microbial communities differed significantly by compartment, with a major shift in the community in root tissues corresponding to the most intimate compartment with the plant.
  • • Differences were also recorded during plant development, the most significant being observed during the transition between vegetative and reproductive stages. Throughout this period, plants were shown to establish the highest level of symbiotic association (mycorrhization, nodulation) with arbuscular mycorrhizal fungi and Rhizobia. During the reproductive stages, the dynamics of the genetic structure differed between bacterial and fungal communities. At the last reproductive stage, the genetic structure of bacterial communities became close to that recorded during the first vegetative stages, suggesting a resilience phenomenon, whereas the genetic structure of fungal communities remained different from the vegetative stages and also from the early reproductive stages, suggesting a persistence of the rhizosphere effect.

Introduction

Medicago truncatula is an omni-Mediterranean species closely related to the world's major forage legume, alfalfa. Medicago truncatula establishes root symbiotic associations with arbuscular mycorrhizal (AM) fungi and Rhizobia. With progress in structural, comparative and functional genomics, M. truncatula is nowadays considered a leguminous plant model (Cook, 1999; Young et al., 2003). In this context, many studies have been dedicated to interactions between M. truncatula and its symbionts, and between this host plant and specific beneficial free-living organisms (Sanchez et al., 2004), but so far little attention has been given to the interaction between these symbionts and other rhizosphere bacteria and fungi (Marschner et al., 2001; Marschner & Timonen, 2005). However, these organisms are likely to interact with the symbionts and affect the symbiotic interactions, as was shown for the interactions of specific populations of fluorescent pseudomonads preferentially associated with Laccaria bicolor that were able to promote the mycorrhization process of Douglas fir (Frey-Klett et al., 2004).

In the 1990s, most studies on the microbial ecology of the rhizosphere consisted of characterizing the diversity and genetic structure of culturable populations belonging to specific microbial groups, including bacterial groups such as fluorescent pseudomonads (Latour et al., 2003), or fungal groups such as Fusarium spp. (Edel et al., 1997). Over the past decade, new molecular methods based on the analysis of nucleic acids directly extracted from soil have been developed, giving access to a previously uncharacterized part of soil microbial diversity (for review see Kent & Triplett, 2002). Such approaches have allowed several authors to demonstrate, in a broad manner, differences in the genetic structure and diversity of microbial communities between bulk and rhizosphere soils (Smalla et al., 2001; Wieland et al., 2001; Marcial Gomes et al., 2003). However, none of the studies so far has taken into account the impact of symbiotic associations and related spatio-temporal modifications of the rhizosphere on the genetic structure of microbial communities. Furthermore, studies of the genetic structure of microbial communities have focused on either bacteria (Duineveld et al., 1998; Marschner & Timonen, 2005) or fungi (Marcial Gomes et al., 2003), and to our knowledge none has characterized both microbial communities in the rhizosphere within the same experimental design.

In general, studies on rhizospheric microflora have stressed variations in the diversity of bacterial populations associated with different plant species (Latour et al., 1996; Grayston et al., 1998). They have further shown a gradient of reduced diversity from bulk soil, rhizosphere, rhizoplane to root tissues (Lemanceau et al., 1995; Clays-Josserand et al., 1995) and differences in bacterial populations according to the root zones (Liljeroth et al., 1991; Yang & Crowley, 2000; Baudoin et al., 2002). One may expect that these differences are a mirror of variations in rhizodeposits. In addition to plant species and root zones, variations in rhizodeposition have also been shown to evolve during plant development (Hamlen et al., 1972) and to be affected by symbiotic associations that correspond to a sink of carbon in the root system (nodules, mycorrhizae) (Söderberg et al., 2002). Utilization of a significant part of photosynthates by symbionts must contribute to a diminution and modification of the composition of rhizodeposits. Variations in rhizodeposition resulting from symbiotic associations and from plant development are likely to influence the genetic structure of microbial communities and their activity (Grayston et al., 1998).

The aim of this study was to study the genetic structure of bacterial and fungal communities in the rhizosphere of M. truncatula Gaertn. Jemalong line J5 during its development (three vegetative and two reproductive stages) and in three compartments (bulk soil, rhizosphere soil and root tissues), taking into account both time and space effects. The spatio-temporal dynamics of bacterial and fungal communities were characterized using molecular methods and analysed in relation to various plant traits: plant growth, root architecture, level of mycorrhization and nodulation.

Materials and Methods

Plant culture and sampling strategy

Medicago truncatula Gaertn. cv. Jemalong line J5 seeds were scarified and surface sterilized by gently shaking them in 98% sulphuric acid for 2 min, 95% ethanol for 5 min, and 3.5% NaClO solution for 10 min, and rinsed successively six times for 5 min in sterile demineralized water. Seeds were germinated on 0.7% (w : v) water agar plates at 25°C for 48 h. One germinated seed was sown in each container containing a silty clay soil (Châteaurenard, France). Major physicochemical characteristics of the soil were: 23.8% sand, 49.6% silt; and 26.6% clays, pH 8.5, carbon 11.22 g kg−1, nitrogen 1.02 g kg−1.

Plants were cultivated in a growth chamber at 16 : 8 h light : dark photoperiod, 23 : 18°C light : dark thermoperiod, PAR = 500 µE m−2 s−1 at pot height and 70% relative humidity. Water was added each day to maintain humidity at 70% soil water-holding capacity.

The plants were analysed at five developmental stages: four leaves (vegetative stage, VS1); shoot ramification of order 1 (VS2); shoot ramification of order 2 (VS3); flowering (first flowers, reproductive stage, RS1); pod production (RS2). The mass of soil per container differed by the duration of plant cultivation. The soil density (1.062 g cm−3) and container height (30 cm) were kept even by increasing the diameter size of the container (6 cm for VS1; 12.5 cm for VS2, VS3 and RS1; 30 cm for RS2). For each developmental stage 18 containers with plants and nine without plants (bulk soil) were randomly placed in the growth chamber and moved each day.

Quantification of plant growth and characterization of root morphogenesis

Plant growth was quantified for each developmental stage on four plant replicates both as root and shoot dry weight (DW) after drying the fresh material for 48 h at 80°C.

Characterization of root architecture was performed on four plants for developmental stages VS1, VS2, VS3 and RS1. The pod maturation stage (RS2) was not analysed because of technical limitations.

Whole root systems were separated from soil adhering to roots by gently washing the roots with sterile water. Then, whole or fractioned root systems were put in a transparent container (30 × 24 cm, Régent Instruments, Canada) with water, and digitized using a dedicated Epson 1680 scanner equipped with a special lighting system for root measurement, connected to a Pentium 4 computer. Digitized root images were analysed by winrhizo pro software (Régent Instruments, Canada) and morphometric parameters were evaluated, including total root length, total root surface area, total root volume, mean root diameter, number of root tips and root branching degree (represented by root tip number divided by total root length) (Gamalero et al., 2002, 2004).

Arbuscular mycorrhizal infections

The efficiency of the mycorrhizal infection was determined on five root systems after randomly sampling 30 pieces of root (1 cm long) per experimental condition. Two staining procedures were used: a nonvital stain (trypan blue) allowing visualization of total AM fungi; and a vital stain (alkaline phosphatase) allowing visualization of active AM fungi, as described previously (Tisserant et al., 1993). For both staining methods, the intensity of root cortex colonization by AM fungi was determined as described by Trouvelot et al. (1986) using mycocalc software (http://www.dijon.inra.fr/mychintec/Mycocalc-prg/download.html) and expressed as: frequency (F%) and intensity (M%) of colonization of the root cortex; colonization intensity of the mycorrhized root cortex (m%); arbuscular abundance of the root cortex (A%); and arbuscular abundance of the mycorrhized root cortex (a%).

Nodulation

Intensity of nodulation was evaluated on the same five plants used for mycorrhization analysis by measuring the total number and DW of nodules. The nodule DW was determined after drying for 48 h at 80°C.

Extraction and purification of total DNA from soil samples

DNA was extracted from three different compartments: bulk soil; rhizosphere soil compartment including bacteria and fungi present in the soil adhering to roots; and root tissue compartment including bacteria and fungi attached to the root surface (rhizoplane) and present in the roots. For each compartment, three replicates per sampling date resulting from the pooling of three different containers were used.

The bulk soil was obtained after pooling and homogenization of the soil from three different containers. One gram of dry bulk soil was sampled in each of the three subreplicates. The rhizosphere soil was obtained after manually separating the root system with adhering soil from the container. Three root systems were pooled for the three subreplicates. The rhizosphere soil was obtained by washing the root system under agitation (200 rpm for 10 min) in sterile distilled water. The volume of water was adjusted as a function of size of root system. The root system was discarded and the soil was collected after centrifugation at 9000g for 10 min. One gram of dry rhizosphere soil was sampled in each of the three subreplicates. The root system was ground with a mortar and pestle in liquid nitrogen. One gram of root system was sampled in each of the three subreplicates. Each sample was frozen in liquid nitrogen and conserved at −80°C for further use in DNA extraction.

The DNA extraction procedure was based on chemical and mechanical extraction as described previously Ranjard et al. (2003). DNA was separated from residual impurities and particularly humic substances by centrifugation through two types of minicolumns. Aliquots (100 µl) of crude DNA extract were loaded onto PVPP (polyvinyl polypyrrolydone) minicolumns (BIORAD, Marne la Coquette, France) and centrifuged at 1000g for 2 min at 10°C. The eluate was collected and DNA extract was purified a second time using the GENECLEAN Turbo Nucleic Acid Purification kit (Q Biogene, Illkirch, France) following the manufacturer's instructions.

Quantification of DNA extracts

Purified DNA samples were resolved by electrophoresis in a 0.8% agarose gel, stained with ethidium bromide and photographed under a camera (Biocapt, Vilber Lourmat, Marne la Vallée, France). Dilutions of calf thymus DNA (Bio-Rad, Marnes la Coquette, France) were included in each gel and a standard curve of DNA concentration (250, 125, 62.5 and 31.25 ng) was used to estimate the final DNA concentration in the purified extracts. The ethidium bromide intensities were integrated with ImageQuaNT software (Molecular Dynamics, Evry, France).

Automated RISA fingerprinting

The bacterial ribosomal IGS were amplified with the primers S-D-Bact-1522-b-S-20 (3′ end of 16S genes) and L-d-Bact-132-a-A-18 (5′ end of 23S genes) (Normand et al., 1996) for bacterial automated ribosomal intergenic spacer analysis (B-ARISA). Length distribution of IGS varied from 150 to 1500 pb (Ranjard et al., 2000). The fungal ITS1-5.8S-ITS2 region was amplified with the primers ITS1-F (3′ end of 18S genes) (Gardes & Bruns, 1993) and 3126T (5′ end of 28S genes) (Sequerra et al., 1997) for fungal automated ribosomal intergenic spacer analysis (F-ARISA). Length distribution of ITS1-5.8S-ITS2 regions varied from 390 to 1065 pb (Ranjard et al., 2001). PCR conditions, PCR template preparation for DNA sequencer loading and electrophoresis conditions were those described by Ranjard et al. (2003).

Statistical analysis

Plant growth and root architecture data were submitted to anova and to Fisher's least significant test (P = 0.05) using the statview statistics package. Mycorrhizal colonization data were statistically analysed by anova after arcsine transformation of percentage using statview software (SAS Institute Inc., Cary, NC, USA).

Principal component analysis (PCA) describing the dynamic of the genetic structure of microbial communities was performed as reported by Ranjard et al. (2003).

Results

Plant development characterization

Three vegetative (VS1, VS2, VS3) and two reproductive (RS1, RS2) stages were chosen, corresponding to important changes in the plant's physiology (Table 1). Plant growth was expressed as root and shoot DW and root : shoot DW ratio (Table 1). Root and shoot DW showed an exponential increase from vegetative to reproductive stages. Root DW increased from 0.009 g at VS1 to 9.193 g at RS2, while shoot DW increased from 0.026 g at VS1 to 284.25 g at RS2. Therefore the root : shoot ratio decreased significantly from RS1. At RS2, dead roots with central cylinder present without cortical cells were observed (data not shown).

Table 1.  Characterization of development and level of symbiotic association during Medicago truncatula Gaertn. cv. Jemalong line J5 development
Time (d)Developmental stage*
VS1 4 leaves 13VS2 First ramification 21VS3 Second ramification 34RS1 Flowering 70RS2 Pod maturation 117
  • *

    VS, Vegetative stage; RS, reproductive stage.

  • Number of plants analysed.

  • Means designated with the same letter are not significantly different (P= 0.05) according to Student–Newman–Keuls's least significant difference test.

  • §

    nd, Not determined.

  • Lower case letters, total mycorrhization; upper case, for active mycorrhization: indicate significant differences between developmental stages (P= 0.05) according to Student–Newman–Keuls's least significant difference test.

Plant aspect (not to scale)inline imageinline imageinline imageinline imageinline image
Biomass (n= 5)
Shoot (g)0.026 a0.13 a1.625 a42.87 b284.25 c
Root (g)0.009 a0.045 a0.405 a3.122 a9.193 b
Root : shoot partitioning0.318 a0.396 b0.249 a0.073 c0.032 d
Root topology (n= 4)
Total root length (cm)116.62 a1129.37 a7978.76 b43403.11 cnd§
Total root surface (cm2)19.39 a164.23 a1549.88 b8185.8 cnd
Total root volume (cm3)0.257 a2.183 a24.935 b106.817 cnd
Average root diameter (mm)0.529 a0.532 a0.635 b0.559and
Number of root tips65.4 a970.75 a7338.5 b33361.59 cnd
Root branching degree0.561 a0.89 a0.925 b0.765 and
Nodulation parameters (n= 5)
Number0 a10 a195 a968 b4393 c
Dry matter (g)0 a0.0005 a0.009 a0.27 b1.13 c
Dry matter nodules/0 a0.435 b0.583 bc0.642 c0.402 d
Dry matter aerial part (%)
Mycorrhization parameters (n= 5): total/active
F (%)18.67 a/30.67 A66.67 b/27.33 A60.67 b/38.00 A84.67 b/66.67 B74.67 b/70.67 B
M (%)1.99 a/2.59 A28.75 b/4.93 A34.25 b/14.14 B38.17 b/23.49 BC44.24 b/36.09 C
m (%)9.40 a/8.76 A40.95 b/19.09 A54.51 b/37.24 B45.19 b/33.80 B59.27 b/48.99 B
A (%)1.58 a/2.01 A28.81 b/4.25 AB29.35 b/10.97 B26.71 b/10.23 B21.99 b/25.30 C
a (%)90.33 a/65.10 A93.08 a/80.09 A84.42 ab/73.89 A68.69 b/40.40 A48.05 bc/69.22 A

Values of root morphometric parameters (total root length, total root surface, total root volume, mean root diameter, root tip number, root branching degree) recorded for VS1–VS3 and RS1 are shown in Table 1. Total root length, total root surface area, total root volume and number of tips increased significantly between VS3 and RS1. The kinetics of these growth parameters were well fitted by an exponential curve (data not shown). In contrast, root-branching degree and average root diameter were well fitted by polynomial curves (data not shown) and had their highest values recorded at VS3.

Nodulation

Bacterial symbiotic status was quantified for the five developmental stages by evaluation of the number and DW of nodules (Table 1). No nodules were visible at VS1. At the other developmental stages, cylindrical-shaped nodules were seen. Colour varied according to developmental stage: white at VS2 and pink at VS3, RS1 and RS2. Some nodules were green and soft at RS2 (data not shown).

The number of nodules and their DW increased significantly from VS3 to RS2 following a linear regression (y = 0.003x) with a positive correlation (R2 = 0.99). The nodule : shoot DW ratio (percentage) reached the highest value at RS1 (0.63%).

Mycorrhization

For each developmental stage, the efficiency of mycorrhizal colonization was assessed by quantifying total and active arbuscular mycorrhizal infections (nonvital and vital staining procedures, respectively) in two subfractions of the same root system. For each staining procedure, measurements of five parameters: frequency (F%) and intensity of colonization of the root cortex (M%), colonization intensity of the mycorrhized root cortex (m%), arbuscular abundance of the mycorrhized root cortex (A%) and arbuscular abundance in mycorrhizal parts of root fragments (a%) were determined (Table 1). Parameters describing total arbuscular mycorrhizal infections were higher than those of active infections.

Total arbuscular mycorrhizal infections increased during the two vegetative stages and then remained even during the reproductive stages. Compared with total arbuscular mycorrhizal infections, increase of active infections was delayed and started only at VS2, but was maintained at all further developmental stages.

Arbuscular abundance of the mycorrhized root cortex slowly decreased from VS2 to RS2 while that of active mycorrhized root cortex remained even throughout plant development.

DNA fingerprints of microbial communities

DNA fingerprinting of microbial communities provided complex profiles for all compartments (bulk soil, rhizosphere soil and root tissues) (Figs 1, 2). B-ARISA bands ranged from 200 bp (50 bp-IGS) to 1051 bp (900 bp-IGS) for bulk soil, and from 200 bp (50 bp-IGS) to 1350 bp (1200 bp-IGS) for rhizosphere soil and root tissue compartments (Fig. 1). For F-ARISA, bands were distributed from 200 to 1120 bp in these three compartments (Fig. 2).

Figure 1.

Bacterial automated ribosomal intergenic spacer analysis (B-ARISA) on DNA extracted from different compartments (bulk soil; rhizosphere soil; root tissues) at the five stages of plant development: VS1 (four leaves); VS2 (first ramification); VS3 (second ramification); RS1 (flowering); RS2 (pod maturation).

Figure 2.

Fungal automated ribosomal intergenic spacer analysis (F-ARISA) on DNA extracted from different compartments (bulk soil; rhizosphere soil; root tissues) at the five stages of plant development: VS1 (four leaves); VS2 (first ramification); VS3 (second ramification); RS1 (flowering); RS2 (pod maturation).

Visual observations of B-ARISA gels indicated that, for a given experimental condition, very few variations were detected among replicates (common bands between profiles with different relative intensity), whereas some differences were recorded in F-ARISA DNA fingerprints (bands specific for a profile or common to several profiles with different relative intensity).

Effect of compartment on genetic structure of microbial communities

The spatial effect of M. truncatula was assessed by comparing the genetic structure of bacterial and fungal communities in bulk soil, rhizosphere soil and root tissue compartments for each developmental stage. This effect is illustrated in Figs 3 and 4 for bacterial and fungal communities, respectively, at VS1 and RS1. The rhizosphere effect was recorded for both communities, but was expressed with a delay for fungal compared with bacterial communities. In the case of the bacterial communities for both developmental stages, the bulk soil and rhizosphere soil compartments were clearly separated from the root tissue compartment on the first axis, which explained 49 (Fig. 3a) and 47% (Fig. 3b) of total variability for VS1 and RS1, respectively; and the bulk soil was clearly separated from the rhizosphere soil compartments on the second axis, which explained 20 and 16% for VS1 and RS1, respectively. Similar data were obtained for the three other stages (data not shown). For fungal communities, this type of separation of compartments was not recorded at VS1, where the first and second axis explained 37 and 22% of total variability (Fig. 4a), but was expressed only from VS2 (data not shown). This indicates that the rhizosphere effect towards this community was delayed, this effect being maintained for further stages including RS1 (Fig. 4b), where the first axis explained 35% of total variability and the second axis explained 22% of total variability.

Figure 3.

Comparison of the genetic structure of bacterial communities by principal component analysis of bacterial automated ribosomal intergenic spacer analysis (B-ARISA) profiles from the different compartments (bulk soil, rhizosphere soil and root tissues) of Medicago truncatula for the plant developmental stages VS1 (a) and RS1 (b). Ellipses represented 90% confidence limits.

Figure 4.

Comparison of the genetic structure of fungal communities by principal component analysis of fungal automated ribosomal intergenic spacer analysis (F-ARISA) profiles from different compartments (bulk soil; rhizosphere soil; root tissues) of Medicago truncatula for plant developmental stages VS1 (a) and RS1 (b). Ellipses represent 90% confidence limits.

Together, these PCA analyses stress a larger modification of the genetic structure of both bacterial and fungal communities between the bulk soil and rhizosphere soil compartments vs the root tissue compartment than between the bulk soil vs rhizosphere soil compartment. These analyses further show, for all developmental stages, a more important effect of M. truncatula in the vicinity of roots (root tissues) than in the soil fraction influenced by the root system (rhizosphere soil).

Effect of M. truncatula development on the genetic structure of microbial communities

The genetic structure of bacterial communities in bulk soil was similar for all five sampling dates (Fig. 5a). The five bulk soils were not separated on the first or second axis in the PCA analysis, which explained 29 and 17%, respectively, of total variability. In contrast, the genetic structure of bacterial communities in the rhizosphere soil compartment (Fig. 5b) and in the root tissue compartment (Fig. 5c) evolved with time during M. truncatula development. RS1 was separated from VS1, VS2, VS3 and RS2 on the first axis, which explained 38% of the total variability for rhizosphere soil (Fig. 5b) and 24% of the total variability for the root tissue compartments (Fig. 5c). VS1, VS2, VS3 and RS2 were separated on the second axis, which explained 15% of the total variability for the rhizosphere soil (Fig. 5b) and 19% of the total variability for the root tissue compartments (Fig. 5c). Finally, the genetic structure of bacterial communities for the RS2 stage was similar to that for VS1 and VS2 in the rhizosphere soil (Fig. 5b) and root tissues (Fig. 5c).

Figure 5.

Comparison of the genetic structure of bacterial communities by principal component analysis of bacterial automated ribosomal intergenic spacer analysis (B-ARISA) profiles obtained during the five stages of plant development (VS1, ▪; VS2, ★; VS3, ⧫; RS1, ▴; RS2, •) from different compartments: (a) bulk soil; (b) rhizosphere soil; (c) root tissues. Ellipses represent 90% confidence limits.

Fungal communities in bulk soils sampled on the five dates were separated on the first and second axes, which explained 19 and 15%, respectively, of total variability (Fig. 6a), indicating that their genetic structure differed with time in bulk soil. However these differences appeared lower than those recorded at these five dates in the rhizosphere compartments, when considering the smaller relative amounts of variation (19 and 15% on the first and second axes, respectively) compared with those recorded in the rhizosphere compartments (Fig. 6b,c). The genetic structure of fungal communities in the rhizosphere soil changed significantly at the five dates during the development of M. truncatula (Fig. 6b). The three vegetative stages (VS1, VS2 and VS3) were separated from RS2 on the first axis, which explained 31% of the total variability, whereas VS1, VS2, VS3 and RS2 were separated from RS1 on the second axis, which explained 21% of the total variability. The genetic structure of fungal communities for VS1, VS2 and VS3 were similar (Fig. 6b). Together these data indicate that the genetic structures of fungal communities during the three vegetative stages were similar, whereas those of the two reproductive stages differed significantly compared with each other, and compared with the vegetative stages.

Figure 6.

Comparison of the genetic structure of fungal communities by principal component analysis of fungal automated ribosomal intergenic spacer analysis (F-ARISA) profiles obtained during the five stages of plant development (VS1, ▪; VS2, ★; VS3, ◆; RS1, ▴; RS2, •) from the different compartments: (a) bulk soil; (b) rhizosphere soil; (c) root tissues. Ellipses represent 90% confidence limits.

This observation was even clearer for the root tissue compartment (Fig. 6c) where the genetic structures of the communities in the three vegetative stages were not separated at all. These vegetative stages were separated from RS1 on the first axis, which explained 22% of the total variability, and RS2 was discriminated on the second axis from the four other stages with 18% of the total variability.

Discussion

The dynamics of the genetic structure of bacterial and fungal communities were characterized during the development of M. truncatula in relation to developmental stage (from plantlet to production of seeds) and to the intensity of symbiotic associations with AM fungi and Rhizobia.

The genetic structure of microbial communities was assessed by A-RISA DNA fingerprints as described by Ranjard et al. (2003). Several genetic fingerprinting techniques are now available to analyse the genetic structure of microbial communities from complex environments. Ribosomal intergenic spacer analysis has recently been optimized and standardized in its resolution and sensitivity of band detection by conducting the electrophoresis in an automated system (ARISA) (Ranjard et al., 2001; Ranjard et al., 2003). This procedure has been shown to be sensitive and robust enough to detect changes consecutive to different perturbations. At a finer level, RISA was used to differentiate communities from different plots of a given site (Ranjard et al., 2001) and subcommunities associated with various micro-environments of a given soil (Ranjard et al., 2000).

The data indicate that the structure of both bacterial and fungal communities always differs according to the compartment from which the corresponding DNA was extracted. These differences were expressed as a gradient from root tissues to bulk soil, confirming the rhizosphere effect and indicating the good level of resolution of the analysis strategy applied.

The genetic structure of bacterial and fungal communities was shown to change significantly during plant development in both vegetative (plantlet, first ramification, second ramification) and reproductive (flowering, pod maturation) stages. To our knowledge this study is one of the first to characterize the dynamics of both bacterial and fungal communities during the full development of a host plant. Our data show that the dynamics of the genetic structure of bacterial communities differed from that of fungal communities during the developmental course of M. truncatula.

During vegetative development of the host plant, the genetic structure of bacterial communities varied at each stage, whereas that of fungal communities did not change significantly, indicating a lower reactivity of fungal communities than of bacterial communities during that period. In contrast, major modifications were recorded at the first reproductive stage (flowering) for both bacterial and fungal communities. Finally, differences in the evolution of structures between bacterial and fungal communities were again detected at the last reproductive stage (pod maturation). At that stage, the structure of bacterial communities was submitted to a major change and came back to a structure similar to that recorded in the initial vegetative stage. This evolution during the life cycle of M. truncatula suggests that, after major modifications during plant development, bacterial communities were almost unaffected by the host plant at the pod maturation stage. These observations suggest that the rhizosphere effect towards bacteria is resilient, using the definition of Gunderson (2000), who defined resilience as the ability of a system to return to a steady state after a perturbation. The genetic structure of fungal communities experienced a major change at pod maturation, leading to a genetic structure different from that of the flowering stage and also from those of the vegetative stages. This last evolution shows the persistence of the rhizosphere effect of M. truncatula towards fungal communities, in contrast to the resilience of bacterial communities towards this effect.

Rhizodeposits are known to account for variations of genetic structure and diversity of microbial communities in the rhizosphere (Marilley & Aragno, 1999; Smalla et al., 2001). The modifications of the genetic structure of bacterial and fungal communities that we recorded during the development of M. truncatula are then expected to mirror variations of the composition of rhizodeposits with time. These rhizodeposits include both water-soluble exudates and more complex organic compounds resulting from dead cells sloughed off roots (Bowen & Rovira, 1991). The proportion of photosynthates released in the rhizosphere and composition of the corresponding rhizodeposits have been shown to vary during the plant's life cycle according to changes in plant physiology during the course of development and the level of symbiotic associations (Gransee & Wittenmayer, 2000). In addition to these variations, bacteria and fungi are known to differ specially in their ability to use organic compounds: bacteria have been shown to have a higher metabolic reactivity than fungi in their ability to use these readily available organic compounds, whereas fungi are able to express enzymatic activities enabling them to use complex organic compounds for their metabolism (De Boer et al., 2005). The integration of these two types of information may give some insight into the variations of the rhizosphere effect towards bacteria and fungi recorded during plant development.

The amount of carbon allocated belowground is usually shown to decrease with plant age (for review see Kuzyakov & Domanski, 2000) and, in agreement with Zhu et al. (1998), we have recorded a significant decrease in the root : shoot partitioning of carbon during the reproductive stages. This decrease indicates a greater use of photosynthates in the shoot than in the root, which may have led to a lower level of release of organic compounds in the rhizosphere during reproductive stages than during vegetative stages. The composition of rhizodeposits has also been shown to change with age and developmental stage (Martin, 1977; Kraffczyk et al., 1984; Hütsch et al., 2002). Changes in the carbohydrate composition of root exudates have been recorded during lucerne development, especially at the flowering stage (Hamlen et al., 1972). Soluble root exudates are released mostly during the vegetative development of the plant, whereas dead cells sloughed from the root, and organic compounds derived from dead pieces of roots, are more abundant as the plant gets older, especially at seed maturation (Eissenstat & Yanai, 2002). During the vegetative stages, the significantly stronger rhizosphere effect towards bacteria than towards fungi could be ascribed to the expected higher release of rhizodeposits mostly as soluble root exudates, which are more favourable to bacteria. In contrast, at pod maturation the expected reduction in the amount of rhizodeposits plus the increased proportion of complex organic compounds in these rhizodeposits would favour the fungi and account for the persistence of the rhizosphere effect at that stage.

The quantity of rhizodeposits was also previously shown to be affected by the intensity of symbiotic associations (Whipps, 1990). The release of photosynthates by mycorrhized and nodulated roots in the rhizosphere was shown to be 4–20 and 13–28% lower, respectively, than by roots without any symbiotic association (Vance & Heichel, 1991; Bago et al., 2000; Johnson et al., 2002). These reports indicate that root symbiosis leads to a sink for carbon allocated to roots including root exudates, and their amount is expected to be reduced (Bonkowski et al., 2001). This reduction would then explain the reduced rhizosphere effect towards bacterial communities that relies mostly on these soluble compounds. In contrast, during the reproductive stage when the level of symbiosis is high, the roots are also becoming older, leading to increased sloughing off of dead cells supporting the growth of fungal populations. This would explain the persistence of the rhizosphere effect recorded in this type of organism.

Finally, variations of the rhizosphere effect during the course of plant development may also be related to the evolution of root architecture. Previous studies have shown the relationships between root branching, root diameter and root exudation (Bonkowski et al., 2001). In our study root branching and root diameter appeared to increase during the vegetative stages and then to decrease during the flowering period, indicating again that a major change in the plant's physiology may have been associated with variation in the rhizodeposits.

Besides variations in the amount and composition of rhizodeposits in relation to the plant's physiology, differences in the dynamics of genetic structure between bacterial and fungal communities may also be related to interactions between these two types of community. For example, symbiotic microorganisms, bacteria and fungi may affect and be reciprocally affected by free-living organisms in the rhizosphere. Marschner et al. (2001) described changes in the bacterial 16S rDNA community composition in the rhizosphere of mycorrhizal maize inoculated with Glomus mosseae or Glomus intraradices. Some populations preferentially associated with mycorrhizae or nodules may include bacterial populations that favour mycorrhization or nodulation (Garbaye, 1994), but also populations that may be harmful to these symbiotic associations or more generally to fungal populations (De Boer et al., 2001).

In conclusion, using a cultivation-independent method we have shown that the dynamics of the genetic structures of bacterial and fungal communities differ during the time course of M. truncatula development. Specifically, we have shown the resilience of bacterial communities and the persistence of fungal communities towards the rhizosphere effect. Characterization of plant development and of the level of symbiotic associations has enabled us to propose hypotheses to explain these differences. Studies on the dynamics of carbon flux and composition in the rhizosphere now need to be performed during plant development to support these hypotheses. Further analyses of bacterial and fungal communities, with characterization of the groups explaining the variations in structure during plant development, are currently being performed in order to specify the evolution of the diversity of these communities.

Acknowledgements

The authors are grateful to G. Duc (URLEG-INRA, Dijon, France) for seeds of M. truncatula, S. Picault for M. truncatula drawings, K. Klein for correcting the English text and D. Pouhair for the control of plant growth conditions in the growth chamber. We thank G. Ganesh for support during this work. This work was supported by INRA – Santé des Plantes et Environnement (‘Rhizosphere Ecology of Annual Medics’ program) and the Burgundy regional project (02511CPO2S188).

Ancillary