Medicago truncatula and Sinorhizobium meliloti form a symbiotic association resulting in the formation of nitrogen-fixing nodules. Nodule cells contain large numbers of bacteroids which are differentiated, nitrogen-fixing forms of the symbiotic bacteria. In the nodules, symbiotic plant cells home and maintain hundreds of viable bacteria. In order to better understand the molecular mechanism sustaining the phenomenon, we searched for new plant genes required for effective symbiosis.
We used a combination of forward and reverse genetics approaches to identify a gene required for nitrogen fixation, and we used cell and molecular biology to characterize the mutant phenotype and to gain an insight into gene function.
The symbiotic gene DNF2 encodes a putative phosphatidylinositol phospholipase C-like protein. Nodules formed by the mutant contain a zone of infected cells reduced to a few cell layers. In this zone, bacteria do not differentiate properly into bacteroids. Furthermore, mutant nodules senesce rapidly and exhibit defense-like reactions.
This atypical phenotype amongst Fix− mutants unravels dnf2 as a new actor of bacteroid persistence inside symbiotic plant cells.
Most species of the legume family are able to form symbiotic associations with nitrogen-fixing soil bacteria, commonly referred to as rhizobia. These associations result in the development of a specialized organ of the root system, called the nodule, in which the bacteria invade the symbiotic plant cells. The early steps of the nodulation process have been studied extensively in the model legumes Medicago truncatula and Lotus japonicus. The very early steps of the plant–microbe interaction are highly similar in these two models; the genes and molecules involved in the reciprocal recognition of the two symbiotic partners are largely conserved (Oldroyd et al., 2011). The interaction starts with the exudation of flavonoids by the plant roots. These molecules are secondary metabolites recognized by the bacteria. In response, the rhizobia synthesize nodulation factors, called Nod factors, which are perceived in a host-specific manner by the plant to initiate the symbiotic process. At the root surface, root hairs curl to surround attached bacteria. These entrapped bacteria induce invagination of the plasma membrane and the formation of a progressing structure, called the infection thread, in which bacteria develop towards the root cortex. Nod factors trigger cortical cell dedifferentiation and the initiation of a nodule primordium. Infection threads reaching the primordium release bacteria within the plant cells. At this stage, nodule formation diverges between Medicago and Lotus. In Medicago, distal primordium cells will not become infected with rhizobia and will develop into an apical nodule meristem (Timmers et al., 1999). Persistence of the meristem results in continuously growing, indeterminate nodules. By contrast, the primordia formed on Lotus plants will not form a meristem, and therefore these nodules are of the determinate type.
Persistence of the apical meristem in indeterminate nodules of Medicago results in nodules harboring different cell types organized in zones (Vasse et al., 1990). The apical meristematic zone, or zone I, contains constantly dividing plant cells that ensure nodule growth (Timmers et al., 1999). In the infection zone, zone II, bacteria are released from the infection threads and invade the plant cells. The rhizobia and their surrounding plant-derived membrane form the symbiosomes. Within the symbiosomes, the bacteria undergo a morphological terminal differentiation involving cell elongation and genome amplification, a process which is governed by numerous plant nodule-specific cysteine-rich peptides, the so-called NCR peptides, which are addressed to the bacterium-containing compartments (Van de Velde et al., 2010). In the interzone II–III, bacteria fully differentiate into nitrogen-fixing bacteroids. In the fixation zone III, the bacteroids reduce atmospheric nitrogen and transfer the resulting ammonium to the plant cells (Vasse et al., 1990). Zone IV, also called the senescence zone, is formed in older nodules and is characterized by the degeneration of both the host cells and their bacteroids. A recent study has indicated that leaf senescence markers, such as cysteine proteinases, are induced in this zone (Van de Velde et al., 2006). The proximal zone V is characterized by the presence of saprophytic rhizobia which proliferate in the remains of the cells after their senescence (Timmers et al., 2000).
In M. truncatula, several genes involved in the first steps of the interaction, Nod factor signaling and the immediate downstream events have been described (Oldroyd et al., 2011). Far less is known in this model organism about the genes involved in the subsequent steps of the symbiotic process, once bacteria have been released into the plant cell cytoplasm. This contrasts with the knowledge acquired on other legume organisms, especially on L. japonicus, in which genes such as LjIGN1, FEN1 and SEN1 have been cloned and described (Kumagai et al., 2007; Hakoyama et al., 2009, 2012). In Lotus, bacteroids do not undergo terminal differentiation; thus, the study of Medicago mutants altered in these symbiotic steps is also of particular interest to understand the role of terminal differentiation. Starker et al. (2006) identified eight M. truncatula mutants altered in their capacity to support nitrogen fixation, that are, Fix− mutants (see also Mitra & Long, 2004). These mutants clustered into seven complementation groups and the corresponding genes were designated as dnf1 to dnf7, where dnf stands for does not fix nitrogen. Microarray experiments were performed on dnf1, 2 and 7, and the expression of nodulation markers was studied in all of them (Mitra & Long, 2004; Starker et al., 2006). Of the seven mutated genes involved in nitrogen fixation, only DNF1 has been cloned so far (Wang et al., 2010). This gene encodes a signal peptidase subunit required to target NCR peptides to the symbiosomes (Van de Velde et al., 2010).
In recent years, ambitious projects have been developed to produce genomic resources useful for the study of the M. truncatula–Sinorhizobium meliloti symbiosis. Amongst these are extensive transcriptomic data (El Yahyaoui et al., 2004; Van de Velde et al., 2006; Benedito et al., 2008; Maunoury et al., 2010; Moreau et al., 2011), large insertional mutant collections (Tadege et al., 2008; Iantcheva et al., 2009; Cheng et al., 2011), developed thanks to the establishment of mutagenesis protocols (d'Erfurth et al., 2003; Rakocevic et al., 2009), and the recently published M. truncatula genome sequence (Young et al., 2011).
We made use of these resources to identify a symbiotic plant gene, DNF2, and we demonstrate its involvement in the nitrogen fixation process. Phenotypes of knockout mutants were examined, and we showed that this gene is required for symbiosome persistence, to prevent plant defense reactions and to avoid early nodule senescence.
Materials and Methods
Bacterial strains and inoculation
The S. meliloti strains used in this study, including Sm1021 (Meade et al., 1982), Rm41 (Kondorosi et al., 1984) and Sm2011 (Rosenberg et al., 1981), and their derivatives (Leong et al., 1985; Ardourel et al., 1994; Penmetsa & Cook, 1997), were cultivated at 30°C in YTB (Orosz et al., 1973) or YEB (Krall et al., 2002) medium supplemented with 0.5 mg.ml−1 of streptomycin for Sm1021 and with 0.01 mg.ml−1 of tetracycline for the strains harboring the pXLGD4 plasmid. To prepare plant inoculum, fresh liquid cultures were washed twice with sterile distilled water and the optical density at 600 nm (OD600 nm) of the suspensions was adjusted to 0.1.
All M. truncatula plants used in this study were from the R108 ecotype (Hoffmann et al., 1997), except for the Jemalong dnf2 mutant (Mitra & Long, 2004; Starker et al., 2006). The Tnt1 insertion mutant lines were generated at the Noble Foundation (Tadege et al., 2008). The MERE1 insertion mutant line MS240 corresponds to a somaclonal variant obtained by regeneration of a T-DNA-tagged Medicago line (Scholte et al., 2002; Brocard et al., 2006; Rakocevic et al., 2009). For expression studies using the β-glucuronidase reporter gene, transgenic plants expressing a DNF2 promoter::gus fusion (pMB132 construct, see later) were generated by Agrobacterium tumefaciens strain EHA105 transformation of M. truncatula R108, as described in Cosson et al. (2006). Of the eight independent transgenic plants tested, two did not express the reported gene and six expressed it specifically with the same expression pattern in the nodules. One of these transgenic lines was used for further studies.
Seeds were surface sterilized and germinated as described in the Medicago truncatula Handbook (Garcia et al., 2006; www.noble.org/MedicagoHandbook). For seed production or crosses, plants were cultivated in the greenhouse under long-day conditions (16-h light : 8-h dark cycle) at 24°C, 50% humidity. For the segregation study, 25 plants of the mutant line NF0217 were grown in the greenhouse on a mixture of sand and perlite (1/3, v/v) and watered with nitrogen-free ‘i’ medium (www.isv.cnrs-gif.fr/embo01/manuels/pdf/module1.pdf) to allow nodulation. Seedlings were inoculated with 10 ml of an S. meliloti suspension prepared as described above.
For in vitro nodulation experiments, seedlings were transferred onto square plates filled with Buffered Nodulation Medium (Ehrhardt et al., 1992) supplemented with 0.1 μM AVG [l-α-(2-aminoethoxyvinyl)-Gly]. The plates were then incubated in growth chambers (16-h light : 8-h dark cycle, 24°C). After 1 or 2 days, each plant root was inoculated with 50 μL of bacterial suspension. Acetylene reduction assays were conducted on plants at 27 days post-inoculation (dpi) as described in Hardy et al. (1968).
PCR reverse screening for Tnt1 and MERE1 insertion mutants in the MTR_4g085800 gene
Plant genomic DNA was extracted from leaves as described by d'Erfurth et al. (2003). Medicago truncatula Fix− mutant lines available in our laboratory were screened for the presence of insertion in MTR_4g085800 by PCR. Genomic DNA was tested in PCRs using three primers per reaction. Two primers correspond to the external parts of the retrotransposon (Tnt1 or MERE1), called LTR for Long Terminal Repeat: primers LTR4 and LTR6 for Tnt1 or MERE1 and MERE4 for MERE1. The third primer was specific to the MTR_4g085800 gene and hybridizes either to the start or the stop codon region for the forward and reverse primer, respectively. All primers are listed in Supporting Information Table S1.
Total RNA was extracted from plant organs using the RNeasy Plant Mini Kit (Qiagen, http://www.qiagen.com/). For cDNA synthesis, RNA samples were treated with RNase-free DNAse I to remove traces of genomic DNA. Reverse transcription (RT) was then performed from 1 μg of total RNA using a poly-T primer and the First Strand cDNA Synthesis Kit from Fermentas (http://www.fermentas.de). Quantitative RT-PCRs were performed using the LightCycler® FastStart DNA Master SYBR Green I Kit on a LightCycler® 480 machine according to the manufacturer's instructions (Roche, http://www.roche.com). Primers are listed in Table S1. The PCR conditions were as follows: 95°C for 10 min (one cycle); 95°C for 5 s, 60°C for 5 s and 72°C for 15 s (40 cycles). PCR amplification specificity was verified using a dissociation curve (55–95°C). Two biological replicates were analyzed. The RBP1 gene was used as a constitutive control for the quantitative RT-PCR (Lelandais-Briere et al., 2009).
For the in situ expression study, a Gateway vector harboring a 4-kb DNA fragment corresponding to the upstream sequence of the DNF2 start codon was used. This vector, derived from pMDC163, was kindly supplied by Dr Dong Wang (Stanford University, Stanford, CA, USA). The site resulting from L/R recombination between the β-glucuronidase AUG and the 5′ UTR of the DNF2 gene was removed by digestion with the enzymes SphI and AscI. The fusion was restored by ligation of an adaptor generated with primers Fusprom1 and Fusprom2 (Table S1) into the linearized vector, resulting in the pMB132 vector.
For the study of alternative splicing variants, cDNAs were cloned into pGEM-T Easy (Promega) after amplification using primers 4-1/4-3, 2-b/4-3, AUG/4-3, 2-d/ex5-6 (Table S1), and then sequenced.
Thin sections (7 μm) of Technovit-embedded nodules were prepared as described by Van de Velde et al. (2006) and stained with 0.02% toluidine blue. Semi-thin sections (50–70 μm) of 6% agarose-embedded nodules were prepared, and β-galactosidase and β-glucuronidase activities were revealed, as described in Boivin et al. (1990) and Welchen & Gonzalez (2005), respectively. X-Gal-stained nodules were first fixed in 2.5% glutaraldehyde in 0.1 M PIPES (piperazine-N,N'-bis(2-ethanesulfonic acid)), pH 7.2. Sections in Fig. 3 were observed and photographed with a Leica DMI 6000B inverted microscope, and sections in Fig. 6 were observed and photographed with an Olympus BX41 epifluorescence microscope using transmitted light for the bright field picture and fluorescent light from the BrightLine GFP filter set EX47 (Hitschfel Instruments, Inc., St. Louis, Missouri, USA).
Free-living rhizobia and bacteroids extracted from nodules were stained with 4′,6′-diamidino-2-phenylindole (DAPI, 5 μg.ml−1) and observed by fluorescence microscopy, as described by Mergaert et al. (2006).
The LIVE/DEAD® BacLight™ Bacterial Viability Kit was used as described by Haag et al. (2011). Briefly, immediately after slicing, semi-thin sections (80 μm) of 6% agarose-embedded nodules were stained for 20 min in a mixture of Syto9 (3.34 mM) and propidium iodide (20 mM). In addition to nuclei, propidium iodide stains only bacteria with compromised membrane integrity. Sections were then briefly rinsed in distilled water and observed on a Leica TCS SP2 inverted confocal microscope under appropriate illumination, as described by Haag et al. (2011).
Phenolic compounds were revealed by potassium permanganate staining, as described previously (Vasse et al., 1993). Freshly sliced agarose-embedded nodule sections were subjected to post-fixation in 0.04% KMnO4, 10 mM PIPES, pH 7.2, for 1 h at room temperature. Sections were rinsed with 10 mM PIPES, pH 7.2, and subsequently stained in 0.01% methylene blue for 10 min. Stained sections were cleared briefly in bleach (2.4% active chloride), rinsed with water and photographed using bright-field microscopy.
For the electron microscopy study, wild-type (WT) and MS240 nodules were dissected from the root system and cut longitudinally to improve the infiltration process. Nodules were placed in a fixation solution of 1.5% paraformaldehyde and 3% glutaraldehyde in 100 mM phosphate buffer (pH 6.8). Nodules were vacuum infiltrated with the fixation solution twice for 10 min, followed by four times for 5 min, and further fixed for 1 h under gentle shaking at room temperature. Samples were subsequently rinsed four times for 10 min with 100 mM phosphate buffer. The nodules were stained for 1 h with 1% tannic acid in 100 mM phosphate buffer and rinsed four times for 10 min with 100 mM phosphate buffer. They were then subjected to post-fixation with 0.1% OsO4 in 100 mM phosphate buffer for 45 min at room temperature. After rinsing 10 times for 5 min with distilled water, nodules were dehydrated in a series of increasing ethanol concentrations (20%, 30%, 50%, 70%, 85%, 90% and 100%) at 4°C. Propylene oxide was used as a transition solvent before Epon 812 (Electron Microscopy Sciences, Hatfield, Pennsylvania, USA) infiltration. Samples were polymerized in Epon 812 at 60°C for 24 h. Sections were obtained using a Reichert Depew, New York, USA ultramicrotome EM UCS (electron microscopy ultracut S). To localize zones II and III precisely, thick longitudinal nodule sections (500 nm) stained with toluidine blue were observed with a light microscope. The distance between the zone of interest and the edge of the block was evaluated using ImageJ software. After turning the block 90° and trimming the excess of resin to reach the zone of interest, transverse thin sections (70 nm) were obtained and observed with an FEI (Hillsboro, Oregon, USA) CM10 electron microscope. Images were recorded with an XR60 CCD camera (AMT, Woburn, Massachusetts, USA) at 80 kV.
Flow cytometry measurements of ploidy levels in nodule cell nuclei were performed as described previously (Cebolla et al., 1999). For each experiment, 11 nodules from five plants at 28 dpi were pooled, and each experiment was repeated twice. The mean values of the repetitions were calculated. The fractions of nuclei at ploidy levels from 2 to 64°C were expressed as a percentage of the total number of nuclei measured. The endoreduplication index (EI), expressing specifically the levels of 16, 32 and 64°C endoreduplicated nuclei, was calculated as described previously (Maunoury et al., 2010). Bacteroids were extracted from nodules (Mergaert et al., 2006), and stained with DAPI (5 μg ml−1). Measurements were performed with a Beckman-Coulter (Danvers, Massachusetts, USA) ELITE ESP flow cytometer.
A candidate symbiotic gene identified using Medicago genomics resources
The Fix– mutant NF0217 was identified during a screen of the M. truncatula Tnt1 mutant collection at the Noble Foundation (Tadege et al., 2008; Pislariu et al., 2012). Flanking sequence tags (FSTs) corresponding to the Tnt1 insertion sites were available for this line (http://bioinfo4.noble.org/mutant/index.php). The expression profiles of the Tnt1-tagged genes were analyzed using the M. truncatula gene expression atlas (MtGEA; Benedito et al., 2008; He et al., 2009; http://mtgea.noble.org/v2/index.php). Of the 35 genes tagged by Tnt1 in NF0217, only MTR_4g085800 (Fig. 1a) is exclusively expressed in nodules. Relative expression levels of probeset Mtr.630.1.S1_at corresponding to MTR_4g085800 in different organs and during nodulation are presented in Fig. 1(b). The genomic sequence of MTR_4g085800 is 2084 bp long from the predicted start to the stop codon. According to the sequences available in public databases, this gene is composed of eight exons (Fig. 1a), which together form the TC119183 cDNA sequence. In NF0217, Tnt1 is inserted in the sixth exon, 1487 bp downstream of the AUG start codon (FST NF0217Ase4; Fig. 1a). The predicted 334-amino-acid protein encoded by TC119183 displays a signal peptide and a phosphatidylinositol-specific phospholipase C X domain (PI-PLCXD; Fig. S1). It belongs to a protein family that is widespread in plants, with multiple genes in each species, including M. truncatula. Despite the presence of the PI-PLCXD, members of this family differ from canonical PI-PLC by lacking the other PI-PLC functional domains. The family includes proteins of Arabidopsis thaliana, Phaseolus vulgaris and Glycine max, as well as a Lotus japonicus nodulin of unknown function (LJNUF; AB167409.1; Kinoshita et al., 2004; Fig. S1).
MTR_4g085800 is required for nitrogen fixation
In order to link the MTR_4g085800 gene mutation with the Fix− phenotype, we analysed 25 plants from the progeny of an NF0217 heterozygous R1 line. Of these 25 plants, six had a Fix− phenotype, indicating that the mutation responsible for the phenotype is recessive. Genotyping of the 25 plants showed co-segregation between NF0217Ase4 homozygote mutants and the Fix− phenotype, indicating that the mutation in MTR_4g085800 is probably responsible for the Fix− phenotype. In order to further validate the involvement of this putative PI-PLCXD protein in the symbiotic process, we screened our Fix− mutant collection for additional lines harboring insertions in this gene. Two additional mutant lines altered in the same gene were identified by PCR: NF2496 and MS240. These two lines have a Tnt1 insertion in the third exon (544 bp downstream of the AUG start codon) and a MERE1 insertion in the sixth exon (1481 bp downstream of the start codon) of MTR_4g085800, respectively (Fig. 1a). Backcrosses to WT R108 for the three lines confirmed the recessive character of the mutations. Allelic tests were performed between the three R108 mutant lines NF0217, NF2496 and MS240. Crosses were made as follows: NF0217 was mated with NF2496 and MS240; in addition, MS240 was mated with NF2496. The Jemalong dnf2 mutant was mated with MS240. For all crosses, the resulting F1 plants (nine for each cross) displayed a Fix− phenotype, proving that the four lines form only one complementation group. These four independent lines altered in MTR_4g085800 were generated using three distinct mutagenesis protocols in two different genetic backgrounds. Characterization of multiple, independent mutant alleles clearly indicates that mutations in MTR_4g085800 are responsible for the Fix− phenotype in these lines. Subsequently, Jemalong dnf2, NF0217, NF2496 and MS240 were referred to as dnf2-1, dnf2-2, dnf2-3 and dnf2-4, respectively. The knockout nature of the mutations was evaluated using quantitative reverse transcription-polymerase chain reaction (qRT-PCR). DNF2 mRNA was not detectable in nodules of dnf2-2 and dnf2-4, indicating that these alleles are knockout mutants. The signal obtained with dnf2-3 nodules was close to the negative control (Fig. S2).
The phenotypes of the three R108 mutants, dnf2-2, dnf2-3 and dnf2-4, were indistinguishable. At the whole-plant level, each line exhibited a typical Fix− phenotype, that is, under symbiotic conditions, the mutants were smaller than WT plants, displayed chlorosis and anthocyanin accumulation, which are commonly associated with nitrogen deficiency (Fig. S3a), and formed more nodules than the WT (Figs 2a, S4). Mutant nodules were smaller and, in contrast with the pink-pigmented WT nodules (Fig. 2a,b), they were either white or brownish (Figs 2a,c,d, S4). Furthermore, as described previously for the Jemalong dnf2-1 mutant (Mitra & Long, 2004; Starker et al., 2006), the R108 dnf2-4 mutant does not support either the expression of the bacterial nifH gene responsible for nitrogenase production (Fig. S3b,c) or the acetylene reduction activity (Fig. S3d). When grown in nitrogen sufficient conditions, the mutants were identical to the WT (data not shown). Taken together; the genetic analyses showed that MTR_4g085800 corresponds to the DNF2 symbiotic gene required for effective nitrogen fixation.
DNF2 transcripts are nodule specific, alternatively spliced and accumulate from the infection zone to the distal part of the fixation zone
The MtGEA dataset was generated using M. truncatula ecotype Jemalong, whereas the Tnt1 mutant collection was generated in M. truncatula ecotype R108. In order to determine whether the DNF2 expression profile is similar in R108, relative transcript levels were measured in leaves, stems, petioles, roots and nodules of WT plants by qRT-PCR. Transcripts were detected only in nodules (Fig. 1c), in agreement with data from Jemalong, reported in MtGEA (Fig. 1b). Relative gene expression was detectable in the youngest nodule sample at 6 dpi (Fig. 1c), when pink pigmentation was first visible. Transcript levels were higher at 10 dpi and remained high in mature nodules (Fig. 1c). Remarkably, when sequencing cDNA products, we observed alternative splicing of the DNF2 gene transcript. Our observations indicate that at least five splice variants exist for the primary DNF2 transcript. Alternative splicing results in the nonsplicing of the first intron, and in variations in the third exon length (Figs 1a, S5).
In order to gain an insight into the putative DNF2 function, the expression pattern of DNF2 was investigated further in situ. A genetic construct carrying the β-glucuronidase reporter gene under the control of the DNF2 promoter was introduced into M. truncatula R108 through Agrobacterium tumefaciens transformation to produce transgenic plants. Histological analyses revealed that the cloned fragment activates β-glucuronidase expression from the infection zone to the distal part of the fixation zone (Fig. 1d). Taken together, the data show that DNF2 is expressed specifically in nodules in the invasion zone and interzone II–III, and that alternative splicing may generate different peptides which may have divergent functional features.
dnf2 nodules have a reduced zone with infected cells and senesce prematurely
In order to determine which steps of the symbiotic process are altered in the mutant, we characterized the structure of dnf2 nodules. The results presented in Fig. 3 do not reveal any differences between WT and dnf2 nodules in the meristem and the infection zones (Fig. 3a,b), but highlight striking differences in zones III and IV (Fig. 3e,f). In dnf2 nodules, the equivalent of WT zone III, which we named zone III*, is properly invaded by the rhizobia at 9 dpi (Fig. 3a,b), but consists of only a few cell layers after 11 dpi (Fig. 3e,f). Furthermore, most infected cells in zone III* have a granular appearance, indicating that the bacteroids in these cells are aberrant (Fig. 3c,d). The senescence zone appeared earlier in dnf2 nodules than in the WT. Zone IV can be observed as early as 11 dpi in the mutant, when it is still completely absent in WT nodules (Fig. 3e,f). The senescence zone first appeared at c. 20 dpi in the WT in our experiments. Consistent with these observations, we found that the expression of a nodule cysteine proteinase senescence marker, TC100437 (MTR_5g022560; Van de Velde et al., 2006), was strongly induced, as early as 10 dpi, in dnf2 relative to WT (Fig. 4a). The differentiation of plant symbiotic cells, characteristic of a functional nitrogen fixation zone, was investigated using flow cytometry. In agreement with the observed reduced size of zone III (Fig. 3e,f), we observed a drastic reduction in EI in the dnf2 mutant relative to the WT (Table S2), reflecting a smaller number of differentiated plant cells, and hence the reduced zone III* in dnf2 nodules. Nevertheless, the presence of a significant number of nodule cells with ploidy levels of 16C, 32C and 64C in dnf2 (Table S2) indicates that the symbiotic cells had grown to a mature size (Vinardell et al., 2003; Maunoury et al., 2010).
DNF2 is required to produce fully differentiated bacteroids
Light microscopy suggests that the bacteroids within symbiotic cells of dnf2 nodules are abnormally developed (Fig. 3c,d). In order to determine the effects of the dnf2 mutation at the ultrastructural level, dnf2-4 and WT nodules at 24 dpi were observed by transmission electron microscopy (TEM; Fig. 5). We could not detect differences between WT and dnf2-4 nodules in the meristematic region and in the infection zone. By contrast, and in agreement with the light microscopy observations, zone III was drastically modified in the dnf2-4 mutant. In WT nodules, bacteroids displayed the characteristic elongated shape of differentiated bacteria, reaching > 5 μm in length (Fig. 5a), and completely filling the symbiosomes. In these WT symbiosomes, the peribacteroid space appeared very thin or not visible (Fig. 5a). By contrast, in the mutant nodules, the peribacteroid membrane seemed to be pulled away from the bacteroid, leaving a large peribacteroid space that occupied most of the symbiosome (Fig. 5b). A lack of bacteroid differentiation was confirmed in the dnf2-4 mutant using flow cytometry and optical microscopy on isolated bacteroids (Fig. S6).
Nodules of the three R108 dnf2 mutants often exhibited dark brown pigmentation in symbiosis with the Rm41 strain (Fig. 2a,d). This was also observed when using Sm1021 (Fig. S4) and Sm2011 (data not shown). Brown pigmentation was visible immediately below the few layers of invaded cells (Fig. 6b) and was never observed in the apical zone. The brown pigments fluoresced green under UV illumination when using a green fluorescent protein (GFP) filter (Fig. 6d), indicating that they may be phenolic compounds. The presence of phenolic compounds in mutant, but not in WT, nodules was confirmed with a potassium permanganate/methylene blue staining procedure (Fig. 6e–h) that detects such compounds (Vasse et al., 1993). Remarkably, brownish nodules were never observed on WT plants, even when senescence had started at 24 dpi (Figs 2a, 4a, S4). Thus, the accumulation of phenolic compounds was not typical of nodule senescence. Because the accumulation of phenolic compounds is often associated with plant defense reactions (Nicholson & Hammerschmidt, 1992), we measured the transcript levels of pathogenesis-related protein 10 (PR10, TC94217/MTR_2g035150; Fig. 4b, Table S1), which is a molecular marker of plant defense (Samac et al., 2011). Elevated levels of PR10 transcript were found in dnf2-4 mutant nodules at 10 dpi. By contrast, PR10 transcript levels remained low in WT, even at 32 dpi (Fig. 4b).
DNF2 is required for bacteroid viability
In view of the evidence for plant defense-like responses and early senescence in dnf2 nodules, we tested the viability of bacteroids using a staining procedure which allows live and dead bacteria to be distinguished. Using sections of 19-dpi nodules, we found that bacteroids in the fixation zone of WT nodules were mostly alive (Fig. 7a,c), whereas only a few bacteroids in zone III* of dnf2-4 nodules had intact membranes, as indicated by the green fluorescence of Syto9 (Fig. 7b,d). Bacterial cells in the proximal, brown-pigmented region of dnf2-4 nodules stained red with propidium iodide, indicating that their membranes were compromised, and hence dead or dying (Fig. 7d). Thus, DNF2 helps to prevent bacterial cell death in infected nodule cells.
Here, we have described a new symbiotic gene of M. truncatula, DNF2 (MTR_4g085800), that is required for nitrogen fixation, but not for nodule development. Three R108 insertion mutant lines harboring different mutant alleles of this gene were isolated and shown to be allelic to the previously described dnf2 Jemalong deletion mutant (Mitra & Long, 2004; Starker et al., 2006).
The Jemalong dnf2 mutant displays no acetylene reduction activity. In addition, this mutant does not support the expression of the bacterial gene, nifH, which is required for nitrogen fixation, suggesting that the maintenance of bacteroids may be affected (Starker et al., 2006). Consistent with the observation that DNF2 is maximally expressed at the onset of nitrogen fixation, our work revealed no differences between dnf2 and WT until bacteria were released from infection threads. At this stage, and in contrast with WT, rhizobia in dnf2 nodules do not differentiate into N2-fixing, fully elongated bacteroids. The electron microscopy study showed that, in the mutant, the peribacteroid space was larger and the bacterial cells remained small, confirming that they were not differentiated (Fig. 5b). The enlarged peribacteroid space indicates that the peribacteroid membrane is continuously produced in the mutant despite the cessation of bacterial enlargement and differentiation, suggesting that the two processes are disconnected. We also observed that bacteroids lose their viability (Fig. 7b,d) after being released into dnf2 symbiotic cells, in contrast with WT (Fig. 7a,c), possibly because they are influenced by the toxicity of NCR peptides, as is the bacterial bacA mutant (Haag et al., 2011), or because they trigger active defense-like responses in the dnf2 mutant. Consistent with the loss of bacterial viability, bacteroids appear to be degraded more rapidly, and the nitrogen-fixing zone is replaced by a zone of empty cells containing cellular debris in dnf2 nodules (Fig. 3e–h).
The cysteine proteinase marker TC100436 is expressed prematurely in the dnf2 mutant (Fig. 4a), consistent with the early degeneration of infected cells observed in zone III* (Fig. 3). Interestingly, senescence markers, such as cysteine proteinases, are induced in the L. japonicus sen1 mutant (Suganuma et al., 2004), as well as in ineffective nodules induced by a nifH::Tn5 S. meliloti mutant on alfalfa (Barsch et al., 2006). This suggests that senescence reflects a general response of nonfunctional nodules (Maunoury et al., 2010).
The impaired differentiation and bacterial death are correlated with activation of plant defense-like reactions in dnf2 mutant nodules, as indicated by the expression of the PR10 defense gene (Fig. 4b) and the accumulation of phenolic compounds (Fig. 6). PR10 induction and necrotic nodules are triggered when inefficient symbiosis results from boron deprivation in Pisum sativum and Phaseolus vulgaris (Redondo-Nieto et al., 2007). However, this phenotype is not a typical feature of Fix− mutants. Indeed, in the R108 background, the majority of Fix− mutants described in Pislariu et al. (2012) do not accumulate these compounds. dnf2-1 has also been suggested to accumulate phenolic compounds near the infection zone (Pislariu & Dickstein, 2007); thus, defense-like reactions are not likely to be a specific trait of dnf2 in the R108 background. PR10 gene induction and the accumulation of phenolic compounds were not observed in R108 WT nodules (Figs 4b, 6), even at 32 dpi when late senescent nodules are present, as attested by the induction of the selected senescence marker (Fig. 4a). Furthermore, defense-like reactions were not described in the L. japonicus sen1 mutant (Suganuma et al., 2004) or in nodules induced by the nifH or nifA Fix− mutants (Barsch et al., 2006; Maunoury et al., 2010). This supports the idea that, during aborted symbiosis, early senescence is not necessarily associated with defense-like reactions. The DNF2 gene encodes a PI-PLCXD-containing protein that has homologs in many plant species, including nonlegumes and nonmycorrhized plants, with multiple homologs in each. Although some homologs of DNF2 obviously fulfill roles unrelated to symbiosis, the nodule-specific expression of DNF2 and the mutant phenotype indicate that the DNF2 PI-PLCXD-containing protein has been recruited to play a specific role in the rhizobium–legume symbiosis with M. truncatula. The expression profile of the DNF2 homolog, LjNUF in L. japonicus (Kinoshita et al., 2004), suggests that it may play a similar role in the determinate nodule symbiosis.
PI-PLCXD-containing proteins are distinct from bona fide eukaryotic phosphatidylinositol-specific phospholipase C proteins (PI-PLCs). PI-PLCs have a multidomain architecture consisting of various regulatory domains and of a PLC catalytic core, which is composed of X and Y regions, separated by a linker. By contrast, PI-PLCXD-containing proteins only contain an X domain (Heinz et al., 1998). A subgroup of PI-PLCXD-containing proteins, which includes DNF2, is found in eukaryotes and in bacteria. These proteins have a similar overall structure, consisting of an N-terminal signal peptide, indicating probable entry of the proteins in the secretory pathway, followed by the PI-PLCXD, and with a total size of 320–420 amino acids. However, the biochemical and biological functions of these proteins remain unclear and, to our knowledge, DNF2 is the first example for which a biological function can be implied based on the phenotype in knockout mutants.
Canonical PI-PLCs cleave phosphatidylinositol (PI) or phosphorylated phosphatidylinositol (PIP) derivatives to produce inositol phosphate and diacylglycerol. These compounds are second messengers or precursors of second messengers, and can participate in plant defense reactions (Canonne et al., 2011). It is difficult to reconcile this with the observed phenotype. As DNF2 does not present all the characteristic domains of eukaryotic PI-PLC, one could consider the possibility that DNF2 interacts with PI(P) without triggering cleavage. A speculation that could fit with the two observations would be that, by binding to PI(P), DNF2 prevents cleavage. This would result in the repression of defense reactions normally triggered by PI(P) degradation products. In eukaryotes, phosphatidylinositides play important roles in membrane trafficking (Michell, 2011). It is also known that some intracellular bacterial pathogens can produce PI(P)-binding proteins or PI(P)-modifying enzymes to disturb membrane trafficking and to persist inside animal cells (Weber et al., 2009). DNF2 might act in a similar way through phosphatidylinositides to avoid the premature lysis of the symbiosome. It will be interesting to determine whether DNF2 supports bacteroid persistence in plant cells by acting on PI(P), especially given the induction of plant defense-like reactions in the dnf2 mutants.
The localization of the DNF2 protein will probably contribute to an understanding of its molecular mechanism. Our study has revealed the existence of five mRNA splice variants for DNF2. All variants conserved the region coding for the PI-PLCXD. The most abundant transcript, corresponding to TC119183, encodes a signal peptide-containing protein, but additional DNF2 mRNAs are predicted to generate proteins lacking the signal peptide (Fig. 1a). Putative splice variants of the third exon that include the first intron of a Phaseolus vulgaris DNF2 homolog are produced in nodules of common bean (Ramirez et al., 2005; Fig. S5). Such a splicing event is not detectable amongst Arabidopsis expressed sequence tags (ESTs), suggesting that this phenomenon could be legume specific. The different amino acid sequences of the predicted DNF2 proteins, resulting from the alternative splicing of the primary DNF2 transcript, indicate that they may be addressed to different compartments of the cell. Some signal peptide-containing proteins that are expressed in symbiotic cells are known to be transported to the symbiosomes (Liu et al., 2006; Coque et al., 2008; Hohnjec et al., 2009; Van de Velde et al., 2010; Meckfessel et al., 2012). Future studies will be necessary to determine the subcellular location of the different peptides and their capacity to restore the WT phenotype.
Taken together, our results indicate that DNF2 is necessary for complete differentiation of bacteroids (Figs 5b, S6) and to maintain symbiosomes (Figs 3, 7). Bacterial mutants altered in the nitrogenase structural gene nifH or in micro-oxic respiration fixL undergo terminal differentiation resulting in fully elongated bacteroids (Maunoury et al., 2010). This indicates that the lack of elongation of bacteroids does not result from an expression defect of nif or fix genes in dnf2, and thus that dnf2 acts upstream of the establishment of micro-aerobic respiration.
In addition to the bacterial lack of differentiation, mobilization of plant defense-like reactions was observed in the DNF2 mutants (Figs 4b, 6). It is possible that these phenomena are linked and that incomplete bacteroid differentiation or a defect in symbiosome maintenance in the dnf2 mutants results from, or triggers, the induction of defense-like reactions. Arguments exist supporting both hypotheses. For instance, the undifferentiated bacA bacterial mutant does not trigger plant defense-like reactions. By contrast, only a fraction of the dnf2 mutant nodules are brownish, suggesting that plant defense-like reactions are only secondary effects of the DNF2 mutation. Future studies are now required to establish the sequence of events resulting from DNF2 mutations.
In summary, in this work, we have identified the DNF2 gene and have further defined its role in the symbiotic process as an actor in symbiosome maintenance. The challenge for future studies will be to determine the molecular function of DNF2 in order to explain its role in the persistence of the symbiosome inside the symbiotic cells.
This work was supported by the Centre National de La Recherche Scientifique (CNRS) and the grant Agence Nationale de la Recherche (ANR) Blanc International SVSE 6.2010.1 (LEGUMICS) from the Agence National de la Recherche to P.R. M.B. was supported by a PhD fellowship from the French Ministry of Research. This work was also supported by the National Science Foundation Plant Genome Research Program, Grant No. DBI-0703285, and The Samuel Roberts Noble Foundation. We are grateful to Dr D. Wang and to Prof. Dr S. Long (Stanford University) for providing the Jemalong dnf2-1 seeds and the modified pMDC163. We thank Philipe Laporte (Laboratoire des Interactions Plantes-Microorganismes (LIPM), Toulouse) for help with Acetylene Reduction Assay (ARA).