A novel RNA-binding peptide regulates the establishment of the Medicago truncatula–Sinorhizobium meliloti nitrogen-fixing symbiosis


  • Philippe Laporte,

    1. Institut des Sciences du Végétal (ISV), Centre National de la Recherche Scientifique, Université PARIS XI, Avenue de la Terrasse, 91198 Gif sur Yvette, France
    Search for more papers by this author
    • Present address: Laboratoire des Interactions Plantes Micro-organismes (LIPM), UMR 441-2594 (INRA-CNRS), Chemin de Borde Rouge-Auzeville, 31326 Castanet Tolosan, France.

  • Béatrice Satiat-Jeunemaître,

    1. Institut des Sciences du Végétal (ISV), Centre National de la Recherche Scientifique, Université PARIS XI, Avenue de la Terrasse, 91198 Gif sur Yvette, France
    Search for more papers by this author
  • Isabel Velasco,

    1. Biomedal S.L., Avda. Américo Vespucio, 41092 Sevilla, Spain
    Search for more papers by this author
  • Tibor Csorba,

    1. Agricultural Biotechnology Center, 2100 Godollo, and Hungary Bay Zoltan Foundation for Applied Research, Institute for Plant Genomics, Human Biotechnology and Bioenergy, H-6726 Szeged, Hungary
    Search for more papers by this author
  • Willem Van de Velde,

    1. Institut des Sciences du Végétal (ISV), Centre National de la Recherche Scientifique, Université PARIS XI, Avenue de la Terrasse, 91198 Gif sur Yvette, France
    Search for more papers by this author
  • Anna Campalans,

    1. Institut des Sciences du Végétal (ISV), Centre National de la Recherche Scientifique, Université PARIS XI, Avenue de la Terrasse, 91198 Gif sur Yvette, France
    Search for more papers by this author
    • Present address: Commissariat à l’Energie Atomique. UMR217 CNRS-CEA, BP6, F-92265 Fontenay aux Roses, France.

  • Joszef Burgyan,

    1. Agricultural Biotechnology Center, 2100 Godollo, and Hungary Bay Zoltan Foundation for Applied Research, Institute for Plant Genomics, Human Biotechnology and Bioenergy, H-6726 Szeged, Hungary
    Search for more papers by this author
    • Present address: Istituto di Virologia Vegetale, Consiglio Nazionale dell Ricerche, Strada delle Cacce, 73-c/o Area di Ricerca di Torino, 10135 Torino, Italy.

  • Miguel Arevalo-Rodriguez,

    1. Biomedal S.L., Avda. Américo Vespucio, 41092 Sevilla, Spain
    Search for more papers by this author
  • Martin Crespi

    Corresponding author
    1. Institut des Sciences du Végétal (ISV), Centre National de la Recherche Scientifique, Université PARIS XI, Avenue de la Terrasse, 91198 Gif sur Yvette, France
    Search for more papers by this author


Plants use a variety of small peptides for cell to cell communication during growth and development. Leguminous plants are characterized by their ability to develop nitrogen-fixing nodules via an interaction with symbiotic bacteria. During nodule organogenesis, several so-called nodulin genes are induced, including large families that encode small peptides. Using a three-hybrid approach in yeast cells, we identified two new small nodulins, MtSNARP1 and MtSNARP2 (for small nodulin acidic RNA-binding protein), which interact with the RNA of MtENOD40, an early induced nodulin gene showing conserved RNA secondary structures. The SNARPs are acidic peptides showing single-stranded RNA-binding activity in vitro and are encoded by a small gene family in Medicago truncatula. These peptides exhibit two new conserved motifs and a putative signal peptide that redirects a GFP fusion to the endoplasmic reticulum both in protoplasts and during symbiosis, suggesting they are secreted. MtSNARP2 is expressed in the differentiating region of the nodule together with several early nodulin genes. MtSNARP2 RNA interference (RNAi) transgenic roots showed aberrant early senescent nodules where differentiated bacteroids degenerate rapidly. Hence, a functional symbiotic interaction may be regulated by secreted RNA-binding peptides.


Plants have evolved a wide variety of strategies to sense the environment and the presence of microorganisms. In particular, small peptides have been implicated in signaling events leading to specific responses during pathogenic or symbiotic interactions. Leguminous plants are able to establish symbiosis with bacteria (rhizobia) to form a new organ, the nitrogen-fixing root nodule, through the coordinated expression of plant and bacterial genes (Stacey et al., 2006). Among the plant genes specifically expressed in nodule tissues, a large family of putatively secreted cysteine-rich peptides only present in indeterminate nodule-forming leguminous plants have been found (Mergaert et al., 2003; Graham et al., 2004). Furthermore, peptide signaling is an emerging area in plants, as recent detailed bioinformatic approaches coupled with expression profiling have revealed under-predicted small peptides (Silverstein et al., 2007). Several secreted peptides are involved in defense responses, such as the well-known antimicrobial peptides (Garcia-Olmedo et al., 1998; Boman, 2003; Bulet et al., 2004; Theis and Stahl, 2004), but also in development [e.g. clavata 3-related (Fiers et al., 2007; Fukuda et al., 2007) or RALF peptides (Pearce et al., 2001)]. These peptides may function as ligands for specific cell-surface receptors and/or affect the function of different intracellular targets, as in the inactivation of ribosomal function by certain defensin peptides (Endo et al., 1987; Vivanco et al., 1999; Chen et al., 2002).

In Medicago spp., the early nodulin gene MtENOD40 is rapidly induced by rhizobia during the symbiotic interaction (Crespi et al., 1994; Fang and Hirsch, 1998; Compaan et al., 2003) as well as in other non-symbiotic organs (Crespi and Galvez, 2000). The ENOD40 genes code for RNAs containing only short open reading frames (sORFs), so-called npcRNAs (for non protein-coding RNAs). Both sORF-encoded oligopeptides and the RNA itself seem to be involved in gene function (Sousa et al., 2001; Rohrig et al., 2002, 2004). Transgenic Medicago truncatula plants overexpressing MtENOD40 exhibited accelerated nodulation, while plants with reduced expression of this gene formed only a few, modified nodule-like structures (Charon et al., 1999; Wan et al., 2007).

RNA-binding proteins have been implicated in mRNA splicing, transport, stability or translation. Numerous RNA-binding proteins contain conserved domains that can bind single-stranded RNA (Hall, 2002; Messias and Sattler, 2004). However, others are unique and contain specific motifs such as the RNA chaperones (Cristofari and Darlix, 2002). As secondary structures in the MtENOD40 RNA are required for biological activity (Sousa et al., 2001) and five RNA stem-loop domains are conserved in legume ENOD40 genes (Girard et al., 2003) we used the yeast three-hybrid system (SenGupta et al., 1996) to explore MtENOD40 RNA interactions. We identified a novel nuclear RNA-binding protein, MtRBP1, which is relocalized from nuclear speckles to the cytoplasm by MtENOD40 (Campalans et al., 2004).

During the three-hybrid screening for MtENOD40 RNA-interacting partners, we isolated several times two genes encoding very small proteins. We show here that these peptides have in vitro RNA-binding activity and are members of a novel family of nodulins, which we termed MtSNARP1 and MtSNARP2 (for small nodulin acidic RNA-binding protein). This new class of RNA-binding proteins exhibits a putative secretion signal peptide (SP). Using MtSNARP2 RNA interference (RNAi), histological and electron microscopy studies, we show that MtSNARP2 regulates the persistence of bacteroids in the symbiosome.


Identification of two novel MtENOD40 RNA-interacting proteins, MtSNARP1 and MtSNARP2

To screen for proteins interacting with the MtENOD40 RNA, we used the yeast three-hybrid system (Campalans et al., 2004) and isolated 12 independent clones corresponding to genes encoding two small proteins showing 62% of identity that we have named MtSNARP1 and MtSNARP2. An in silico BLAST search of the Institute for Genomic Research (TIGR) expressed sequence tag (EST) database allowed us to identify a gene family of eight members, all encoding small proteins of 67–83 amino acids (Figure 1). This new peptide family defines two novel conserved domains of unknown function (Figure 1a). Certain tentative consensus sequences of EST clusters (TCs) additionally contained an N-terminal signal peptide sequence as revealed by the SignalP 3.0 server (http://www.cbs.dtu.dk/services/SignalP/) (Figure 1a; Bendtsen et al., 2004). To verify the presence of the signal peptide coding sequence in the MtSNARP2 mRNA, we performed a 5′-rapid amplification of cDNA ends (RACE)-PCR experiment from mature nodule mRNAs. All sequences obtained (nine independent clones) exhibited the same 5′ end containing the putative signal peptide sequence (Figures 1b and S1).

Figure 1.

 The MtSNARP gene family.
(a) Alignment of the eight MtSNARP peptides. The predicted signal peptide cleavage site is indicated (grey line) and the peptide signal is highlighted. Conserved acidic domains and two basic amino acids (k, r) of this family can be deduced from the consensus.
(b) Alignment of the 5′ region spanning the signal peptide of MtSNARP2 corresponding to nine randomly chosen 5′-RACE PCR clones.

The activity of the predicted signal peptide in targeting MtSNARP2 to the secretory pathway was analyzed using C-terminal translational GFP fusions to this protein in transient expression assays in Arabidopsis protoplasts. Full-length MtSNARP2::GFP C-terminal fusions with or without the predicted signal peptide were compared (Figure 2). Complete MtSNARP2::GFP fusions localized in multiple reticulated particles surrounding the nuclear envelope, while the nucleus was devoid of GFP signal (Figure 2e–h, arrows). Moreover, the majority of the MtSNARP2::GFP fusion co-localized with a HDEL::mCherry fusion (Figure 2l–n), an endoplasmic reticulum (ER) marker. In contrast, cells transformed with a similar construct without this signal peptide accumulate GFP in the nucleus, as well as diffuse localization into the cytoplasm (Figure 2a–d), and did not co-localize with a HDEL::mCherry fusion (Figure 2i–k).

Figure 2.

 Subcellular localization of MtSNARP2::GFP fusion peptides in Arabidopsis protoplasts.
(a–d) Three-dimensional (3D) projection of Z-stack (a), and confocal sections (b–d) of a protoplast expressing the MtSNARP2::GFP fusion without a signal peptide. Sections through the nucleus (b and c), and at the nuclear envelope level (d) are shown. Diffuse GFP fluorescence is observed in all subcellular compartments.
(e–h) Similar 3D projection (e) and confocal sections (f–h) of a protoplast transformed with the full-length MtSNARP2::GFP fusion carrying the predicted signal peptide (f and g through the nucleus, and h above nuclear envelope). Note the lack of GFP fluorescence in the nuclei.
(i–k) Confocal section, through the nucleus, of a protoplast co-expressing the MtSNARP2::GFP fusion without signal peptide (i) and a HDEL::mCherry fusion [j, an endoplasmic reticulum (ER) marker]. (k) Merged image of (i) and (j).
(l–n) Similar section of a protoplast co-transformed with the full-length MtSNARP2::GFP fusion (l) and ER marker (m). (n) Merged image of (l) and (m) that allows us to observe that both fusion proteins are mainly co-localized to the ER.
N, nucleus; V, vacuole; ER, endoplasmic reticulum. Bars: (a–h) 10 μm; (i and j) 2 μm.

To verify the activity of this signal peptide in nodules, we introduced MtSNARP2::GFP fusions in Medicago truncatula composite plants (Boisson-Dernier et al., 2001). After kanamycin selection, composite plants were inoculated with a red-fluorescent Sinorhizobium meliloti Sm1021 strain containing pDG77 plasmid (DsRed; Gage, 2002) or a wild-type strain (Figure 3). In the infection zone of the nodule at 6 days post-infection (dpi), the MtSNARP2::GFP fusion signal appeared in membrane particles accumulating around infection threads that penetrate host cells (Figure 3a–d, arrowhead in a detailed magnification, e). In 14-dpi nodules, the MtSNARP2::GFP fusion exhibited a localization around the nuclear envelope excluding the nucleus (Figure 3f–i and arrow in detailed magnification j), similar to the localization obtained in Arabidopsis protoplasts (Figure 2). A control without fluorescent bacteria or MtSNARP2 fusion allows us to monitor the level of autofluorescence in our experiments (Figure 3k–o). Hence, MtSNARP2 is addressed to the secretory pathway via a functional signal peptide during the infection of host cells.

Figure 3.

 Subcellular localization of MtSNARP2::GFP during nodulation.
(a–j) Localization of MtSNARP2::GFP expression during nodule development [6 days post-infection (dpi) (a–d) and 14 dpi (f–i)]. (e, j) Detailed magnification of (c, h), respectively.
Longitudinal vibratome sections (70 μm) of nodules expressing MtSNARP2::GFP fusion compared with similar sections of control nodules (k–o), allow us to observe the GFP signal in membrane particles around infection threads (e, arrowhead) or around the nuclear envelope excluding the nucleus (j, arrow).

The MtSNARP2 protein was purified after recombinant expression in Escherichia coli and used for in vitro RNA pull-down assays. Preliminary attempts to produce MtSNARP2 as a fusion protein with polyhistidine (6 × His) or glutathione-S-transferase (GST) tags suggest that this protein was toxic for E. coli. To avoid this toxicity problem, we used the highly controlled LYTAG®/CASCADE® system (http://www.biomedal.com/) to produce recombinant MtSNARP2 protein in E. coli, expressed and purified as a LYTAG-MtSNARP2 fusion protein without the predicted signal peptide. The recombinant LYTAG-MtSNARP2 fusion protein was shown to interact with various RNA molecules in vitro (Figure 4). To better characterize the RNA-binding activity of MtSNARP2, we performed gel-shift assays with synthetic single-stranded (ss) and double-stranded (ds) RNAs and DNAs of different sizes (Figure 5). The MtSNARP2 protein does not bind short [21, 24, 26 nucleotide (nt)] ss- or dsRNAs and DNA; however it efficiently binds 49-nt long ssRNAs with high affinity (dissociation equilibrium constant Kd = 93.2 nm; Figure 5a). Furthermore, the RNA-binding activity was not sequence specific, since MtSNARP2 bound the entire MtENOD40 RNA, its different subdomains (F1, F2 and F3) and the 49 synthetic RNA oligo with a related affinity (Kd between 79.5 and 120.8 nmFigure 5b–f). These results demonstrate that MtSNARP2 has ssRNA-binding activity in vitro.

Figure 4.

 Interaction between MtSNARP2 and MtENOD40 RNA.
(a) Expression and purification of the fusion protein LYTAG-SNARP2 and analysis by Coomassie blue staining. The arrow indicates the size of the expected fusion protein in the elution from the LYTAG purification column. Extracts from induced and non-induced bacteria are shown as well as the flow-through of the purification column.
(b) RNA pull-down analyses were performed using biotinylated MtENOD40 or MtRBP1 RNAs and purified LYTAG-SNARP2 or LYTAG alone (negative control) proteins. The figure displays a western analysis using anti-LYTAG2 antibodies. Inputs represent the amount of LYTAG-SNARP2 and LYTAG proteins present in the RNA-binding assay before the pull-down assay. Elutions of the affinity beads containing biotinylated MtENOD40 or MtRBP1 RNAs after pull-down are shown on the left. Note that only the large LYTAG-SNARP2 fusion but not the LYTAG alone are pulled down by the RNA-affinity beads.

Figure 5.

 MtSNARP2 protein binds to single-stranded RNAs in vitro.
LYTAG-SNARP2 (SNARP), or LYTAG2 (−) as well as positive control proteins (Carnation italian ringspot virus p19 protein and Reovirus sigma 3 protein; Lakatos et al., 2006) were incubated with single-stranded 21 nucleotide (nt) or 49 nt and double-stranded 21 nt or 49 nt RNA molecules (a) and loaded on 6% native acrylamide gels. In contrast, SNARP2 protein binds ENOD40 RNA (b), ss49 RNA (c), ENOD40 RNA domains F1 (d), F2 (e) and F3 (f). SNARP2 does not show affinity to single- or double-stranded DNA (c, e) or to short 21, 24 or 26 nt single-stranded RNAs or 21 nt double-stranded RNA (a, d). Arrow, alternative conformation (b), *degradation products (d, e).

Expression of MtSNARP2 is detected early during symbiotic interactions

Transcripts levels of MtSNARP2 were evaluated using real-time RT-PCR analysis, performed during the symbiotic interaction of S. meliloti with M. truncatula roots, for MtSNARP2, MtNIN (nodule inception; Marsh et al., 2007), and Leghaemoglobin (MtLb1; Gallusci et al., 1991) (Figures 6a,b). MtSNARP2 transcripts are induced concomitantly with MtNIN and before MtLb1 transcript accumulation, showing that it is an early induced nodulin. The expression of the SNARP family analysed in the M. truncatula Gene Expression Atlas (MtGEA) Web Server (Benedito et al., 2008) also revealed nodule-specific expression. MtSNARP2 and two other members (MtSNARP1 and TC102256) are expressed earlier than the other members of the family (Figure S2).

Figure 6.

 Expression patterns of MtSNARP2 transcripts.
(a–f) Real-time RT-PCR analysis of MtSNARP expression compared with MtLb1 and the early nodulin MtNIN of roots (0, 1, 3, 8 and 21 days post-infection (dpi)) and from 40-day-old nodules (a and b, T0–T40); of roots 0, 1, 3, 8 and 21 dpi with the Sinorhizobium meliloti nodH strain. (c, d) Analysis from 21 dpi roots with different rhizobial symbiotic mutants, the exoY strain and bacA strain, compared with 21 dpi root tissues with wild-type S. meliloti. (e, f) Quantification of specific PCR products was normalized with MtH3-l. The value for uninfected control roots (T0) is set to one as reference. Results of three biological experiments are combined (around 15 plants were used for each time point), and error bars indicate standard deviation. *Logarithmic scale representation of the Y-axis.

The nodH rhizobial strain, unable to interact with the plant partner, did not induce the expression of any gene except MtNIN (Figure 6c,d). However, the expression observed for MtNIN at 21 dpi is not significant compared with that observed in nodules (100-fold versus 5-fold; Figure 6b,d). The expression of these genes was also analyzed in roots infected with different rhizobial symbiotic mutants at 21 dpi (Figure 6e,f). The exoY strain induces empty nodules and premature abortion of infection threads (Gray et al., 1991). These nodules showed weak induction of MtSNARP2 and MtNIN and no expression at all for the late nodulin MtLb1 (Figure 6e,f). The interaction with a bacA strain, blocked in bacterial differentiation (Glazebrook et al., 1993), results in more developed and infected nodules. MtSNARP2 and MtNIN showed a similar induction of expression in these nodules as in the wild-type situation, in contrast to MtLb1 transcript levels (Figure 6e,f). Therefore, MtSNARP2 and MtNIN show a similar pattern of early induction during nodulation.

Several zones can be identified in mature indeterminate nodules (Vasse et al., 1990), including the apical meristem (zone I), the infection zone (zone II) where bacteria are released from infection threads into the plant cell; the interzone II–III, a region characterized by an accumulation of amyloplasts and by major transcriptional changes in both partners; and the nitrogen-fixing zone (zone III). The spatial expression pattern of MtSNARP2 transcripts was analyzed using in situ hybridization on mature nodules (Figure 7). A signal associated with zones I and II and the periphery of the nodule, especially around vascular bundles, was observed (Figure 7a,d and g). This expression pattern coincides with that observed for MtENOD40 (Figure 7b,e and h), notably around the vascular bundles. This signal was absent in nodule sections hybridized with the sense probe (Figure 7c,f and i). The spatial expression patterns of MtSNARP2 and MtENOD40 further suggest that it is an early nodulin gene.

Figure 7.

 Spatial expression of MtSNARP and MtENOD40 genes in nodules.
In situ hybridization was performed on microtome sections (10 μm) of 20-day-old nodules using a digoxigenin (DIG)-labeled T3-riboprobe to detect MtSNARP transcripts and a DIG-labeled T7-riboprobe to detect MtENOD40 transcripts. A sense MtSNARP2 RNA probe was used as negative control.
(a, d, g). A nodule longitudinal section hybridized with an MtSNARP2 antisense probe (a) and higher magnification showing the invasion zone (d) and vascular bundles (g).
(b, e, h). A consecutive section to (a) hybridized with a MtENOD40 antisense probe.
(c, f, i) A nodule section hybridized with an MtSNARP2 sense probe.
VB, vascular bundles.
Bars: (a–f) 100 μm; (g–i) 50 μm.

MtSNARP2-silenced transgenic roots form numerous aberrant nodules

MtSNARP genes have not yet been sequenced in the M. truncatula genome program. RNA interference could have a much broader effect on the expression of similar genes, and this approach has been used in the literature to detect symbiotic phenotypes (Limpens et al., 2004; Gargantini et al., 2006; Colditz et al., 2007). We believe that appropriate controls are required to avoid any unspecific effect due to the activation of gene silencing. We used RNAi constructs targeting the uidA (GUS) gene encoding β-glucuronidase as controls for our experiments in order to generate small interfering (si) RNAs unable to target endogenous genes.

To analyze potential functions of this new small RNA-binding protein in root and nodule organogenesis, we silenced MtSNARP2 gene expression using RNAi under the control of the 35S CaMV promoter and Agrobacterium rhizogenes transformation (Boisson-Dernier et al., 2001). After kanamycin selection, composite plants transferred on a nitrogen-free medium were inoculated with the S. meliloti Sm2011 wild-type strain. At 40 dpi, the growth of the root system of MtSNARP2 RNAi composite plants was compared with control plants (expressing a GUS RNAi construct). This analysis was conducted on at least 150 independent composite plants in greenhouse conditions and, for each construct, three independent biological experiments were performed. No significant growth phenotype was observed for these composite plants (Table 1). We concomitantly confirmed the effect of the RNAi construct on the expression of MtSNARP2 in symbiotic nodules at 16 dpi (as these genes were not detected in non-symbiotic tissues). Quantitative RT-PCR analysis confirmed the specific reduction of MtSNARP2 expression in these nodules, whereas transcripts levels of MtNIN, MtENOD40 and different members of the MtSNARP family remain unchanged (Figure S3).

Table 1.  Growth characteristics of MtSNARP2 RNA interference (RNAi) composite plants compared with control GUS RNAi
dpiNodule number on GUS RNAi roots (±CI)Nodule number on MtSNARP2 RNAi roots (±CI) GUS RNAi dry root mass/shoot mass (±CI) MtSNARP2 RNAi dry root mass/shoot mass (±CI)
  1. dpi, days post-infection; CI, confidence interval.

88.3 ± 1.67.3 ± 1.5  
8.4 ± 1.59.7 ± 1.7  
7.7 ± 1.56.1 ± 1.3  
1610.5 ± 1.68.5 ± 1.4 8.9 ± 1.6  
10.5 ± 1.711 ± 1.6  
11.8 ± 1.5   
4034.7 ± 6.325.3 ± 8.40.38 ± 0.170.37 ± 0.17
40.6 ± 6.541.3 ± 3.60.38 ± 0.260.42 ± 0.29
43.6 ± 4.135.6 ± 3.50.41 ± 0.210.41 ± 0.25

The ability of MtSNARP2 and GUS RNAi roots to form nodules was determined at 8 and 16 dpi in composite plants. Afterwards, these plants were transferred to the greenhouse to determine nodule numbers at 40 dpi. No significant difference was observed in nodule number between control and RNAi roots at any tested time point after inoculation with Rhizobium (Table 1). However, we identified two different nodule types on both control and RNAi roots: regular pink wild-type nodules and ‘aberrant’ ones exhibiting a hypertrophied outer cortex and containing a brown pigment (instead of a pink colour; Figure S4). At 40 dpi, the number of aberrant nodules on the MtSNARP2 RNAi roots was significantly higher than in GUS RNAi control roots (Figure 8a; α ≤ 0.001, more than 4000 nodules were evaluated for each transgenic root type, MtSNARP2 and GUS RNAi). Complementary analyses of the aberrant nodules were conducted using a GFP-fluorescent S. meliloti Sm1021 strain containing pDG71 plasmid (Figure 8b,c,e and f; Gage, 2002) or a Sm2011 strain containing a ProHemALACZ reporter gene (Ardourel et al., 1994; Figure 8d,g). At early time points after infection (5 dpi) no differences in formation of the infection thread were detected in the MtSNARP2 RNAi roots and host cell invasion occurred normally (Figure 8b,e, arrowheads show infection threads). Later in development, control roots showed strong lacZ staining in 10-dpi nodules (Figure 8d). This was in contrast to the weak and patchy LacZ expression observed in nodules of MtSNARP2 RNAi roots. Staining was confined to the apical part of the nodule (Figure 8g). This result suggests that bacteria infect nodule cells but do not remain metabolically active to express the lacZ gene. Microscopic analysis of longitudinal sections from 35-dpi GUS RNAi (Figure 9a–e) and MtSNARP2 RNAi nodules (Figure 9f–j), was done using a light microscope under white or UV light. All observed GUS RNAi nodules (including 15 initially scored as ‘aberrant’ at a macroscopic level) presented a normal zonation (Figure 9b) with archetypal cells filled with autofluorescent bacteroids (a central vacuole surrounded by bacteroids; Figure 9c, arrowhead). Ultraviolet light revealed that the abnormal MtSNARP2 RNAi nodules also have a very thick outer cortex and endodermis (Figure 9g in comparison with Figure 9b). In contrast to wild-type nodules (Figure 9a–e), the presence of a nitrogen-fixing zone (zone III) containing strongly autofluorescent cells due to the presence of bacteroids was not observed in these nodules (Figure 9b,g). The aberrant development of MtSNARP2 RNAi nodules was characterized by the lack of clearly delimited zones I, II and III, and the formation of a closed endodermal layer completely surrounding the nodule (Figure 9g, arrow). Certain infected enlarged lacunae between cells could be observed (Figure 9h arrowhead) and many cells collapsed. Furthermore, Lugol staining showed the absence of amyloplasts in MtSNARP2 RNAi nodules (Figure 9i,j); in contrast to control nodules where amyloplasts were observed in the interzone II–III (Figure 9d,e).

Figure 8.

 Nodulation phenotype of MtSNARP2 RNAi roots.
(a) Quantification of ‘aberrant’ nodules was done on GUS RNA interference (RNAi) or MtSNARP2 RNAi roots 40 days post-inoculation (dpi) with Sinorhizobium meliloti 2011 strain. Three biological experiments are presented and error bars indicate standard deviation. (α 0.001, more than 4000 nodules were evaluated for each MtSNARP2 RNAi and GUS RNAi root).
(b, c e, f) Vibratome sections (70 μm) of, respectively, transgenic GUS RNAi roots and MtSNARP2 RNAi roots, infected with a GFP-fluorescent S. meliloti Sm1021 strain carrying the pVO185 plasmid containing a ProTrp:GFP. (b, e) Representative infection threads development (arrowheads). (c, f) Representative young 5-dpi nodule.
(d), (g) Histochemical staining of nodules inoculated with an S. meliloti strain expressing a ProHemA:LACZ reporter construct to follow the infection during nodulation. Representative infection events observed in S. meliloti-inoculated GUS RNAi nodules (d) and MtSNARP2 RNAi nodules (g) at 10 dpi are shown.
Bars: 200 μm.

Figure 9.

 Histological analysis of MtSNARP2 RNAi nodules 35 days post-inoculation (dpi).
(a–e) Longitudinal microtome sections (10 μm) of GUS RNA interference (RNAi) aberrant nodules: nodule structure in bright field (a), under UV (b, detail in c, delimited zones I, II and III; note bacteroid fluorescence in region III highlighted by arrowhead in c), and Lugol staining to reveal starch accumulation (d, detail in e).
(f–j) Longitudinal sections of MtSNARP2 RNAi aberrant nodules: nodule structure in bright field (f), under UV (g, detail in h; arrow indicates enclosed highly fluorescent endodermis; arrowhead showed interstitial bacteria), and Lugol staining (i, detail in j).
Bars: 200 μm.

To further characterize this phenotype, we performed electron microscopy studies (Figure 10) using 6-, 11- and 14-dpi nodules from MtSNARP2 RNAi and GUS RNAi roots. At an early stage of development, in 6-dpi nodules, archetypal rhizobial invasion could be observed in both MtSNARP2 RNAi nodules and the control ones (Figure 10a–e), in correlation with our previous experiments. Nevertheless, at a later stage (11 dpi), we began to see in MtSNARP2 RNAi nodules an increased peribacteroid space and bacteroid degeneration (arrowhead and arrow, respectively, Figure 10f,g,i and j compared with Figure 10h,k), indicating early nodule senescence and the lack of persistence of functional symbiosomes. At 14 dpi, the central region of the MtSNARP2 RNAi nodules showed striking differences in cell morphology (Figure 10l,m,o and p compared with Figure 10n,q) and no differentiated bacteria remain in the MtSNARP2 RNAi nodules. Cells appeared more disorganized within the tissue, often with their cytoplasm full of disorganized membranes (Figure 10l,m,o and p, arrowheads), suggesting a turgor defect. Transverse sections revealed the absence of differentiated rhizobia inside these disorganized plant cells contrasting with the elongated differentiated S. meliloti bacteroids (Figure 10n,q).

Figure 10.

 Ultrastructure of infected cells of GUS RNA interference (RNAi) and MtSNARP2 RNAi nodules.
(a–e) Six-day-old nodules. (a, b) Sections of a MtSNARP2 RNAi nodule exhibiting released bacteroids in host cells. (c) Central zone region (similar to b) of a GUS RNAi nodule presenting differentiated bacteroids. (d, e) Detailed images of (b and c), respectively, showing similar bacteroid differentiation at 6 days post-infection (dpi). (f–k) Eleven-day-old nodules. (f, g) Sections in the central zone of a MtSNARP2 RNAi nodule showing increased empty peribacteroid spaces (arrowheads) and degenerated bacteroids (arrows); (i, j) High magnification of (f and g). (h) A similar zone of a GUS RNAi nodule, detailed in (k), showed a filled peribacteroid space with differentiated bacteroids.
(l–q) Fourteen-day-old nodules. (l, m) Sections in the central zone of a MtSNARP2 RNAi nodule exhibiting flattened cells and abundant degraded membranes (arrowheads). Bacteria out of the infection threads do not elongate. (o, p) High magnification. (n) Central zone of a GUS RNAi nodule full of differentiated and elongated bacteroids. (q) Detail of (n).
B, bacteroïd; ER, endoplasmic reticulum; IT, infection thread; N, nucleus; Sm, Sinorhizobium meliloti undifferentiated.
Bars: (f, g, l, n) 5 μm; (a–c, h–k, m, o–q) 2 μm; (d, e) 1 μm.

These results indicate that MtSNARP2 plays a crucial role in the establishment of functional symbiosomes during the M. truncatulaS. meliloti interaction.


Successive developmental stages of the symbiotic interaction are strictly coordinated by genes from both symbiotic partners. Recent genomic approaches identified large numbers of genes induced during nodule development. In this study, we identified a new nodulin family comprising eight MtSNARP genes that exhibit two conserved unknown domains, show RNA-binding activity in vitro and carry putative secretory signals.

The RNA-binding activity of MtSNARP2 does not seem to be sequence-specific, even though a specific interaction in vivo cannot be excluded, and does not depend on a RNA recognition motif (RRM). Peptides can have nucleic acid folding chaperone activity (Cristofari and Darlix, 2002), an activity also characterized by RNA binding with weak sequence specificity. These RNA chaperones allow rapid annealing of two complementary nucleic acid strands or catalyze strand exchange between a nucleic acid duplex and another nucleic acid strand to promote the most stable interaction. RNA chaperones do not share a common structure. The SNARP class of ssRNA-binding protein has acidic properties, although, in general, basic domains are associated with RNA binding (Messias and Sattler, 2004). MtSNARPs present a region containing several basic amino acids, but the most conserved domains are highly acidic and may be involved in other interactions. The progression of infection threads could release the SNARP peptides upon hydrolysis and they may become signals re-entering the plant cell or the bacterial symbiont. Proline-rich propeptides anchored to the cell walls are processed into peptide signals for systemic wound signaling in solanaceous plants (Narvaez-Vasquez et al., 2005). In addition, certain glycine-rich proteins (GRPs) could be involved in cell-wall-anchored defense responses, notably owing to their immobilization via Tyr residues that have the ability to link the aromatic residues of lignin (reviewed in Cassab, 1998). Such secreted peptides could play an extracellular role by interacting with bacteria inside cell walls during penetration of infection threads in root tissues. Another family of over 300 nodule-specific genes carrying a signal peptide (NCRs, for nodule cysteine-rich proteins) was identified in Medicago truncatula by Mergaert et al. (2003).

The silencing of the RNA-binding protein MtSNARP2 led to premature nodule senescence, even though the 35S CaMV promoter may have reduced expression at late nodulation steps (Auriac and Timmers, 2007). However, the early expression of MtSNARP2 during nodulation and the lack of functional symbiosomes in MtSNARP2 RNAi nodules indicate that MtSNARP2 has a role in symbiosis and validates the use of this construct.

Many Medicago mutants showed aberrant invasion phenotypes with concomitant derepression of defense responses and accelerated nodule senescence. The Medicago lin (lumpy infection) mutant (Kuppusamy et al., 2004) seems to be affected in a host defense mechanism, preventing infection threads from reaching the nodule primordium. Abortive infection threads are also observed after inoculation with a Sinorhizobium meliloti exoY mutant (Cheng and Walker, 1998). The Medicago nip (numerous infections and polyphenolics) mutant is also disturbed in defense response with proliferation of ineffective infection threads and accumulation of polyphenolic compounds (Veereshlingam et al., 2004). However, we found no evidence for accumulation of such compounds or of phenylalanine ammonia-lyase (PAL) transcripts (an enzyme that participates in phenylpropanoid biosynthesis pathway) in MtSNARP2 RNAi nodules (data not shown). Even though SNARP2 is induced in the infection zone, normal infection threads could be observed during symbiosis. Other invasion phenotypes were linked to the depletion of the LRR receptor kinase DMI2 (Limpens et al., 2005) and the knock-down of the transcription factor HAP2-1 (Combier et al., 2006) using RNAi-mediated knock-down. In MtSNARP2 RNAi nodules, electron microscopy showed early senescence (Van de Velde et al., 2006), characterized by rapid cell death.

A link between MtSNARP2 and the action of ENOD40 RNA in the cytoplasm of the nodule cells could not be established. The ENOD40 RNA relocalizes the nuclear MtRBP1 protein into a cytoplasmic particle of unknown function (Campalans et al., 2004) and this RNP may be linked to the secretory system similar to the well known signal recognition particle and 7S RNA (Sauer-Eriksson and Hainzl, 2003). We cannot exclude that SNARPs may re-enter the plant cell (due to their small size). Although we produced antibodies against MtSNARP members, they failed to recognize any specific signal in nodule tissues preventing us from determining the subcellular localization of MtSNARPs in nodules. It is well known that small peptides are difficult to immunolocalize and/or to isolate from plant tissues using western analysis or HPLC approaches (Ryan and Pearce, 2001).

These small RNA-binding proteins may target the bacteria, as the presence of a signal peptide does not exclusively represent apoplast targeting. For example, ENOD8 carries a signal peptide and localizes into the symbiosome (Coque et al., 2008). Recently, a chloroplast-located protein in higher plants was shown to be transported via an alternative route through the secretory pathway (Villarejo et al., 2005). In fact, this peculiar α-carbonic anhydrase (CAH1) is targeted to the ER via a signal peptide but ends in the chloroplast. During the primitive endosymbiotic interactions leading to the formation of eukaryotic cells, the ancestral chloroplast probably needed to absorb secreted proteins before developing a specific sorting pathway (Keegstra and Cline, 1999; Jarvis and Soll, 2001; Leister, 2003). Perhaps a related mechanism has been developed for targeting the bacteroid during the endosymbiotic interaction between rhizobia and legumes. Hence, certain peptides, such as MtSNARPs, carrying secretory signal peptides could end up in the bacteroids.

Alternatively MtSNARPs may interact with extracellular RNA from plant or microorganism-like symbionts, commensals or pathogens in the apoplast. Small peptides involved in the inactivation of ribosomal RNA in pathogen responses are also secreted RNA-interacting peptides (Endo et al., 1987; Vivanco et al., 1999). In addition, in Caenorhabditis elegans, part of the molecular dialogue between cells is RNA-mediated through a transmembrane protein, SID-1, that enables the transport of dsRNA inside the cell (Feinberg and Hunter, 2003; Saleh et al., 2006). As RNA-binding proteins play roles in a variety of post-transcriptional processes in bacteria (Pichon and Felden, 2007), we can speculate that the plant host may regulate these processes in the microsymbiont through the MtSNARP genes.

The peribacteroid membrane is made up of membrane from the infection-thread, ER-derived membrane and de novo synthesis (Roth and Stacey, 1989). If MtSNARP2 peptides are not secreted, they may remain associated with the ER and contribute to the functionality of the peribacteroid membrane to maintain the viability of the symbiosomes.

Overall, this work has identified a new peculiar nodulin, a ssRNA-binding peptide with a functional signal peptide, that plays a crucial role in the stability of functional symbiosomes during the M. truncatulaS. meliloti interaction. This may offer interesting perspectives for defining novel roles of RNA-binding peptides in cell-to-cell communication events.

Experimental procedures

Plant material

Medicago truncatula cv. Jemalong seeds were sterilized as described (Charon et al., 1999) and seedlings were grown on nitrogen-deficient i medium (Blondon, 1964). Roots were spot-inoculated (OD600nm = 0.05, 0.4% agar) with Sinorhizobium meliloti Sm1021 strain or mutant strains (Sm1021 bacA, Sm1021 exoY, Sm2011 nodH). Three biological experiments were performed (n > 15 plants per experiment).

Bioinformatic analyses

BlastN and BlastX searches were performed for ESTs on TIGR databases (http://compbio.dfci.harvard.edu/tgi/). Putative signal peptides (Bendtsen et al., 2004) were identified using SignalP3.0. (http://www.cbs.dtu.dk/services/SignalP/). Peptide and gene alignments were performed using ‘Multiple sequence alignment with hierarchical clustering’ (Corpet, 1988) or ESPript.cgi version3.06CGI3.05.

Expression of MtSNARP2::GFP fusion proteins

Arabidopsis protoplasts were transformed with the constructs pPK100-35S:SNARP2::GFP (without a signal peptide, starting at the [ATG]28) or p2GWF7-35S:(SP)SNARP2::GFP (full length MtSNARP2 with a signal peptide; Figure 1) using the polyethylene glycol (PEG)-mediated method (Mathur and Koncz, 1998). The HDEL::mCherry fusion was co-transformed with the previous fusion.

Medicago truncatula roots were transformed using A. rhizogenes with the construct pK7FWG2-35S:(SP)SNARP2::GFP, and inoculated with a DsRed-fluorescent Sm1021 strain carrying the pDG77 plasmid (Gage, 2002). Longitudinal sections (70 μm) were taken from material embedded in 3% agarose using a Leica VT1200S vibratome. Localization of MtSNARP2::GFP fusions was visualized using a Leica TCS SP confocal laser-scanning microscope (Leica Microsystems, http://www.leica.com).

Recombinant expression and purification of SNARP2

SNARP2 was expressed in bacteria as an N-terminal tagged protein and purified using the LYTAG Purification System (Biomedal). The SNARP2 coding sequence was PCR-amplified from the yeast plasmid (Campalans et al., 2004) using primers MAO15(5′-CGCGGATCCATGGTTACCACTTTCGTGGG-3′; BamHI site underlined) and MAO16(5′-CGCGGATCCTCAATGAAGAGAATCCATAT-3′; BamHI site underlined), and fused in frame with the choline-binding domain of the LYTAG coding sequence in the BamHI site of vector pALEX2b. The resulting construct, pALEX2b-SNARP2-3, encodes a LYTAG-SNARP2 fusion protein with a theoretical molecular mass of 23.63 kDa.

The specialized Escherichia coli host strain REG-1 was transformed with plasmid pALEX2b-SNARP2-3. A 250-ml culture at 37°C and 200 g was induced (OD600 0.9) by the addition of 2 mm salicylate for 4 h at 30°C. Harvested cells were resuspended in 40 ml of lysis buffer (20 mm potassium phosphate pH 7, 0.1% Triton X-100, 1.5 m NaCl), and disrupted by sonication (10 pulses of 20 sec, 60%). The resulting lysate was centrifuged at 9000 g for 15 min and the supernatant was incubated with 0.5 ml of LYTRAP resin at 4°C for 120 min as described by the manufacturer. Expression and purification of the LYTAG protein domain, used as a control, was carried out similarly using a culture of strain REG-1 transformed with the vector pALEX2b.

SNARP2–RNA interaction in vitro

The PCR-DNA template for in vitro synthesis of MtENOD40 or MtRBP1 RNAs was obtained from plasmid pSKMtenod40 or pSKMtRBP1, respectively (Campalans et al., 2004), using Pfu (Bioneer, http://www.bioneer.com/), and primersT3 (5′-ATTAACCCTCACTAAAGGGA-3′) and Mar6 (5′-CAGAAACTGAAACAAGAAC-3′) or T7 (TAATACGACTCACTATAGG). Biotinylated RNAs were synthesized with the MEGAscript®T3 Kit (Ambion, http://www.ambion.com/), using 1 μg of PCR product in the presence of 0.33 mm Biotin-16-UTP (Roche, http://www.roche.com/), in a 20 μl reaction, and purified with the MEGAclearkit (Ambion).

To study the SNARP2–RNA interaction, 1.7 μg of purified LYTAG-SNARP2 or LYTAG control was incubated with 2 μg of biotinylated RNA and 25 μl of μMACS-streptavidin magnetic beads (Miltenyi, http://www.miltenyibiotec.com/) as described by the manufacturer. The RNA-bound protein was recovered in the magnetized support, resolved by 15% SDS-PAGE, transferred to a polyvinylidene fluoride (PVDF) membrane (Hybond-P; Amersham Biosciences, http://www1.gelifesciences.com), and analyzed by western blot using anti-LYTAG rabbit polyclonal antibody and alkaline phosphatase detection reagent (SIGMA FAST®BIP/NBT; Sigma, http://www.sigmaaldrich.com/).

Gel mobility shift assay

Gel mobility shift assays have been done using P32-labeled synthetic ssRNA, dsRNA, ssDNA, dsDNA (21, 24, 26, 49 and 60 nt length, sequence available upon request) or T7 transcripts (Fermentas, http://www.fermentas.com/) of full length ENOD40 or ENOD40 domains (fragment 1, F1: 43–128 nt, fragment 2, F2: 110–220 nt and fragment 3, F3: 201–352 nt of the MtENOD40 sequence) in a concentration of 0.5–2 nm. The purified LYTAG-MtSNARP2 protein and the RNAs or DNAs (0.5 nm) were incubated in band shift buffer (10 mm HEPES pH 7.5, 50 mm KCl, 1 mm EDTA, 1 mm DTT, 0.5% Triton X-100, 10% glycerol) at 25°C for 30 min. A fraction of the binding reactions was separated on 0.5 × TBE 6% acrylamide native gels at 4°C. Gels were dried and exposed on phosphor-image screens (GE Healthcare Europe, http://www.gehealthcare.com/eueu/).

Expression analysis

Total RNAs were extracted from frozen roots using TRIzol® reagent (Invitrogen, http://www.invitrogen.com/). First-strand cDNA was synthesized from 1 μg of total RNA using the Superscript® II first-strand synthesis system (Invitrogen). Primer sequences used are listed in Table S1. Real-time RT-PCR reactions were performed using the LightCycler® FastStart DNA MasterPLUS SYBR Green I kit (Roche Diagnostics, http://www.rochediagnostics.fr) on a Roche LightCycler®1.5 instrument (cycling conditions: 95°C for 10 min, 50 cycles at 95°C for 5 sec, 58°C for 5 sec, and 72°C for 15 sec). The PCR amplification specificity was confirmed by analysis of dissociation curves (55–95°C), and primer combinations showing a minimum amplification efficiency of 90% were used (Table S1). A negative control without a cDNA template was always included for each primer combination. Three technical replicates (independent synthesis of cDNAs) and three independent biological experiments were performed in all cases. Ratios were calculated with MtH3-l and MtRBP1 genes as constitutive controls (Table S1, Figure S5). The ratio value of the experimental control condition was set up to one to determine relative expression levels. For semi-quantitative RT-PCR, specific primers (Table S1) for MtENOD40, MtRBP1, MtSNARP and Mtc27 were designed and Mtc27 was used as a constitutive control (Crespi et al., 1994). Amplification conditions were: one cycle of 2 min at 94°C, 25 cycles of 30 sec at 94°C, 30 sec at 60°C and 1 min at 72°C.

The 5′-RACE-PCR assays were performed as described in Hirsch et al. (2006).

Agrobacterium rhizogenes root transformation.  The MtSNARP2 and GUS sequences were amplified by PCR and cloned into the pFRN destination vector (derived from pFGC5941; NCBI accession number AY310901) using Gateway® Technology (Invitrogen). Primers used were: MtSNARP2-5′, TGGACATCCAGATCAATACCC; MtSNARP2-3′, TTCTAAATGGAACTGCAACCAC; GUS-5′, GGCCAGCGTATCGTGCTGCG; GUS-3′, GGTCGTGCACCATCAGCACG.

The resulting constructs were introduced into Agrobacterium rhizogenes ARqua1 and used for M. truncatula root transformation as described in Boisson-Dernier et al. (2001). For all A. rhizogenes-transformed roots experiments, at least three biological experiments were performed. Specificity and efficiency of silencing was checked using real-time RT-PCR on several individual clones.

Nodulation experiments with Agrobacterium rhizogenes-transformed roots.  Kanamycin-selected A. rhizogenes-infected composite plants (2 weeks) were transferred onto growth pouch papers (mega international, http://www.mega-international.com/) on nitrogen-free Fahraeus medium (Truchet et al., 1985) for 4 days. Transgenic roots were inoculated with 20 ml of a S. meliloti Sm2011 suspension (OD600nm = 0.05) per plate for 1 h. Nodulation was evaluated at 8 dpi (non-nitrogen-fixing nodules) and at 16 dpi (nitrogen-fixing nodules). For analysis at later times, composite plants obtained in vitro were transferred to the greenhouse (16 h light/8 h dark, 22°C, 60–70% hygrometry) on a perlite:sand/4:1 mixed substrate imbibed with i medium, and the nodule number was determined at 40 dpi. Three biological experiments were performed with a minimum of 150 independent transgenic roots per construct analyzed.

Histochemical assays

A Sm1021 strain carrying the pDG71 plasmid (Gage, 2002) or a Sm2011 derivative strain (GMI6526; Ardourel et al., 1994) carrying the pXLGD4 plasmid containing a ProHemA:LACZ transcriptional fusion was used to inoculate RNAi transgenic roots. Vibratome sections (70 μm) of roots infected with Sm1021 pDG71 strain were observed on a Leica DMI6000B microscope equipped with a Leica DFC300 FX digital camera. Roots infected with Sm2011 pXLGD4 strain were stained for β-galactosidase activity as described in Ardourel et al. (1994) and observed using a Reichert Polyvar microscope equipped with a Nikon digital DXM1200 camera (Nikon Corporation Instruments Company, http://www.nikon.com/).

Electron microscopy analysis.  Histological techniques, starch staining using paraffin-embedded material and specimen processing for electron microscopy were performed as described in Hawes and Satiat-Jeunemaître (2001). For electron microscopy, nodules were embedded in Spurr resin. Ultrathin (90 nm) sections were obtained using a Leica ultramicrotome EM UC6, counterstained with 2% uranyl acetate and lead citrate, and observed with an electron microscope JEOL JEM-1400 (JEOL Ltd, http://www.jeol.com/). Images were recorded with an SC1000 Orius CCD camera (Gatan Inc., http://www.gatan.com/) at 120 kV.

In situ hybridization.  Medicago truncatula mature nodules were processed for in situ hybridization as described in Valoczi et al. (2006).

Accession numbers.  Accession numbers from EMBL/GeneBank and TIGR databases are provided in Figures 1 and S1.


Peter Mergaert is acknowledged for providing Sm1021 bacA, Sm1021 exoY, Sm2011 nodH strains, GFP and DsRed-fluorescent Sinorhizobium meliloti Sm1021 strains and certain RNA samples. Angel Cebolla (Biomedal S.L., Seville, Spain) is acknowledged for his help and ideas for in vitro RNA interaction assays, Joaquin Giner for technical assistance in recombinant expression and purification of SNARP2, Alexis Maizel for the HDEL::mCherry fusion and the Cell Biology Unit of the Imagif platform of the ‘Centre de Recherche de Gif sur Yvette’, supported by the ‘Conseil Général de l’Essonne’ and the IFR87 for imaging facilities. We also thank Peter Mergaert for careful reading of the manuscript. PL was the recipient of a fellowship from the ‘Ministère de l’Education Nationale et de la Recherche’ (MENR). This work was supported by the FP6 EEC-RIBOREG project.