Bacterial strains and plasmids are listed in Table 1. Escherichia coli strains were propagated in Luria–Bertani (LB) medium. Sinorhizobium meliloti strains were constructed and grown in LB medium supplemented with 2.5 mM CaCl₂ and 2.5 mM MgSO₄ (LBMC). Antibiotics, when required, were added at the following concentrations: streptomycin 100–300 μg ml⁻¹, tetracycline 5–10 μg ml⁻¹, gentamycin 40 μg ml⁻¹, carbenicillin 50 μg ml⁻¹, kanamycin 50 μg ml⁻¹.

To test the bacterial response to NO, S. meliloti was grown in Vincent minimal medium (VMM: 7.35 mM KH₂PO₄, 5.74 mM K₂HPO₄, 1 mM MgSO4, 18.7 mM NH₄Cl, 10 mM Na₂ succinate, 456 μM CaCl₂, 35 μM FeCl₃, 4 μM biotin, 48.5 μM H₃BO₃, 10 μM MnSO₄, 1 μM ZnSO₄, 0.5 μM CuSO₄, 0.27 μM CoCl₂, 0.5 μM NaMoO₄, pH = 7) at 28°C. The NO donor Spermine NONOate (SpNN) was purchased from Cayman Chemicals Coger (CAY-82150-M100, Paris, France). A 100 mM stock solution was prepared, just before use, in Na-phosphate buffer (0.1 M) pH 6.9 and then diluted in the cell culture medium at the appropriate concentration.

Medicago truncatula cv Jemalong A17 was used, and was cultivated as follows. Seeds of M. truncatula cv Jemalong A17 were scarified with H₂SO₄, surface-sterilized in a bleach solution, rinsed with sterile distilled water, germinated on agar plates in the dark and allowed to grow on nitrogen-free Farhäeus medium in test tubes or plates (covered with pouch paper) for 2 to 7 d before inoculation in a culture room (22–25°C) with a 16 h light/8 h dark photocycle (all details and medium used are described in the Medicago Handbook: http://www.noble.org/MedicagoHandbook/). For NO detection with 4,5-diaminofluorescein (DAF-2, Sigma-Aldrich) (Figs 1, 4), transgenic roots (Fig. 6), and quantitative reverse-transcription polymerase chain reaction (qRT-PCR) experiments (Fig. 7), S. meliloti strains were grown in liquid LBMC medium, cells were pelleted at 2500 g, washed twice with sterile distilled water, resuspended in sterile distilled water to a final optical density at 600 nm (OD₆₀₀) of 0.01, and plants were inoculated with 200 μl of bacterial suspension per root. In all other experiments, S. meliloti strains were grown on LBMC plates, cells were resuspended in sterile distilled water to a final OD₆₀₀ of 0.001, and plants were inoculated with 100 μl of bacterial suspension per root. When 2-(4-carboxyphenyl)-4,4,5,5-tetramethyl imidazoline-1-oxyl-3-oxide (cPTIO; Sigma) was used, 200 μl of a 1 mM solution prepared in sterile water was added along the whole length of the roots, 2 h before, and 2 h after inoculation, and then, every 24 h during 4 d.

For competition assays, plants were inoculated with a 1 : 1 mixture of wild-type (wt) and mutant strains (OD₆₀₀ = 0.001). In order to verify that the number of cells used for inoculation with each strain were identical, CFUs were determined for each strain. To determine the relative proportions of nodules occupied by mutant or wt bacteria, the wt strain contained a plasmid (pXLGD4) constitutively expressing the β-galactosidase enzyme. After 3 wk, nodules were wounded and incubated (1 h) in Z′ buffer (0.1 M Na₂HPO₄/NaH₂PO₄ pH 7.4, 10 mM KCl, 1 mM MgSO₄) containing 1.5% of glutaraldehyde under vacuum. Nodules were then washed three times in Z′ buffer and incubated in the same buffer containing potassium ferricyanide (5 mM) and X-gal (0.8% in dimethylformamide). Incubation was performed under vacuum at room temperature (90 min) and then at 37°C (90 min). Blue-stained nodules corresponding to nodules infected by GMI11545 (wt) and pink nodules corresponding to nodules infected by the mutant strain were counted. For each competition test, 50 –100 nodules from 10–20 plants were analysed, and the mean value of the percentage of nodule occupancy was calculated for each strain. When cPTIO was used in the test, nodules were taken at 4 wk instead of 3 wk after inoculation because of the delayed nodulation.

The binary vector pKm43GW-rolD::GFP (^{Van de Velde et al., 2006}) was modified by replacing the green fluorescent protein (GFP) with a red fluorescent protein (RFP) reporter cassette under the control of the rolD promoter to facilitate the analysis by confocal microscopy with the cell permeable 4,5-Diaminofluorescein diacetate (DAF-2DA) fluorescent probe (excitation 495 nm, emission max c. 515 nm; Sigma-Aldrich) and the selection of transgenic roots. The RFP cassette was amplified by PCR from the pK7RWG2 vector (http://gateway.psb.ugent.be) with the primers DsRedX and DsRedS containing, respectively, a XhoI and SacI restriction site and cloned in the pKm43GW-rolD by using these two restriction sites to get the pKm43GW-rolD::RFP.

The promoter from MtENOD20 (Accession: AC136953.17; GenBank accession no. X99467) (2050 bp) was amplified with the specific primers a74F and a74R containing the corresponding attB recombination sites. The promoter region was recombined into pDONR-P4-P1 (Invitrogen). The Multisite Gateway Three-Fragment Vector Construct kit (Invitrogen) was used to fuse the promoter region with the GUS gene (from pDONR207-GUS) or the hmp gene (from pDONR207-hmp) and the T35S terminator (from pENTR-R2-T35S-L3) in the pKm43GW-rolD-RFP vector. The hmp gene was amplified from S. meliloti genomic DNA with the specific primers a76F and a76R.

The bacterial NO biosensor was constructed as follows. The promoter of the S. meliloti gene sma1289 was amplified by PCR with the following primers: OCB744 and OCB745 containing the PstI and XbaI restriction sites, respectively. The PCR amplification was performed using genomic DNA of strain GMI11495 as a template. The amplified fragment (410 nucleotides) was ligated in pGEM-T (Promega). The cloned S. meliloti DNA region was verified by DNA sequencing. The pGEM-T plasmid containing the sma1289 promoter was then digested with XbaI and PstI, and ligated into the XbaI and PstI digested pCZ962 plasmid containing the promoterless lacZ gene. Plasmids (pCZ962 with or without the sma1289 promoter) were introduced in S. meliloti strains (GMI11495, CBT515, CBT602) by tri-parental mating using pRK2013 as a helper, and subsequent selection for antibiotic resistance.

Sinorhizobium meliloti cells were grown to exponential phase at 28°C in VMM. At OD₆₀₀ = 0.2, Spermine NONOate (0 or 25 μM) was added to the cultures which were then incubated at 28°C. After 1.5 h incubation, OD₆₀₀ was measured and aliquots (1 ml) of the cultures were centrifuged. The pellets were immediately frozen in liquid nitrogen and stored at −20°C. β-Galactosidase assays were performed on the thawed samples as described previously (^{Miller, 1972}).

For β-galactosidase detection, plants were inoculated with strains GMI11495 (wt), CBT515 (nnrR), or CBT602 (hmp⁺⁺) carrying the plasmid pSma1289–lacZ or with strain GMI11495 carrying the empty plasmid. Six days after inoculation, plant roots were incubated (1 h) under vacuum, in Z′ buffer containing 1.5% glutaraldehyde. Roots were then washed three times in Z′ buffer and incubated (room temperature) in Z′ buffer containing potassium ferricyanide (5 mM) and X-gal (0.8% in dimethylformamide) for 90 min or for a longer period when the staining was too weak.

For β-glucuronidase detection, roots 5 d post-inoculation (dpi) and 13 dpi transformed with pKm43W-pENOD20-GUS were incubated for 1 h in acetone 90% (diluted in a phosphate buffer: Na₂HPO₄, NaH₂PO₄ 0.1M pH = 7.4) at −20°C. Roots were then washed twice in the phosphate buffer and incubated at room temperature, protected from light, in phosphate buffer containing potassium ferricyanide (0.5 mM) and X-gluc (0.5 ng ml⁻¹) for 3 h and 16 h.

The roots were observed by optical microscopy (Axioplan imagin2; Zeiss). Pictures were taken with an AxioCam (Zeiss) and acquired with the corresponding software.

The construct pKm43GW-pENOD20-hmp was introduced into Agrobacterium rhizogenes strain Arqua1 (^{Quandt et al., 1993}). Medicago truncatula plants were transformed with Agrobacterium rhizogenes, as described by ^{Boisson-Dernier et al. (2001)}. Control plants were transformed with A. rhizogenes containing the pKm43GW-pENOD20-GUS. Selection of hairy roots took place 21 d after transformation. Plants were screened for transgenic roots by RFP with a MZFLII stereomicroscope (Leica, Wetzlar, Germany). One transgenic root per plant was retained, and the composite plants were transferred onto agar plates containing modified Fahräeus medium without nitrogen. Plants were inoculated 3 d after transfer.

The NO produced by both the wt and transgenic hairy roots, and released into the detection medium, was measured using the fluorescent probe DAF-2. At various times, the fluorescence of DAF-2T, the reaction product formed from DAF-2 and NO, was measured using a microplate reader spectrofluorimeter (Cary Eclipse; Varian, Les Ulis, France), excitation wavelength 495 nm/emission wavelength 515 nm. Plants were pooled into groups of seven to nine plants, and the roots were put into 3 ml of detection medium: 10 mM Tris-KCl detection buffer (DB: 10 mM Tris-HCl pH 7.4, 10 mM KCl) in the presence of 10 μM DAF-2 fluorescent probe. The roots were protected from light with tinfoil while the leaves were left in contact with air and light, allowing the photosynthesis to take place. Production of NO was measured at 10, 30, 60 and 90 min after the addition of the DAF-2 probe into the detection medium. Blank assays contained detection medium without plants.

Experiments were performed on young entire roots and fresh nodule slices (100 μm) obtained with a vibratome 1000 Plus (Labonord, Templemars, France). Plant material was then incubated for 30 min in the dark in 100–500 μl of DB containing 10 μM DAF-2DA the cell permeable analog of DAF-2 probe (Sigma-Aldrich). Plant tissues were subsequently washed for 30 min with DB and then mounted between slide and coverslip in DB for observation. When cPTIO was used, plant tissues were preincubated for 45 min in DB containing 1 mM of cPTIO, subsequently labelled with DAF-2DA and washed three times during 20 min with DB. The formation of DAF-2T following the NO reaction with DAF-2DA was visualized using a Zeiss LSM 500 confocal laser microscope upon excitation at 488 nm with an argon 2 laser. Dye emission was recorded using a 505–530 nm band-pass filter coupled with a 515 nm long-pass filter. Autofluorescence was judged negligible at all steps before acquisition with the DAF-2DA probe. Data acquisition settings (laser power, pinhole size, filters, detectors settings and scan conditions) were identical in all experiments. Images were processed and analysed using the Zeiss LSM510 META software and Adobe Photoshop.

For the cPTIO treatment used to identify NO-regulated genes, plantlets were grown on plates and then, at 4 d post-inoculation, were treated with 200 μl of 1 mM of cPTIO for 8 h. A control assay was not treated with cPTIO. Furthermore, in order to obtain a higher concentration of RNA from nodule primordia, a 2 cm region of the root in which most of the nodule primordia appear was harvested. To localize this root infection zone (IZ), we marked the position of the primary root apex on the day of the inoculation. Four days later, the IZ was mainly located around this mark.

Aproximately 100–200 mg of plant material (root) were ground in liquid nitrogen and total RNA was isolated using Trizol Reagent (Invitrogen). The integrity of total RNA was checked on agarose gels and its quantity as well as purity was determined spectrophotometrically. About 500–1500 ng of RNA was used as a template for reverse transcription reaction in 20 μl reaction volume using the Omniscript RT Kit (Qiagen). Quantitative real-time RT-PCR was carried out using the qPCR Mastermix Plus for SYBR Green I reagent (Eurogentec, http://www.eurogentec.com). Reactions were run on the Chromo4 Real-Time PCR Detection System (Bio-Rad) and quantification was performed with the Opticon Monitor analysis software v. 3.1 (Bio-Rad). Each reaction was set up in three technical replicates. For each reaction, 5 μl of 100-fold-diluted cDNA and 0.3 μM primers were used. The initial denaturing time was 10 min, followed by 40 PCR cycles at 95°C for 10 s, and 60°C for 1 min. The specificity of the amplification was confirmed by a single peak in a dissociation curve at the end of the PCR reaction. Data were quantified using Opticon Monitor 2 (MJ Research, Waltham, Massachusetts, USA) and analysed with RqPCRBase, an R package working on R computing environment for analysis of quantitative real-time PCR data (T. Tran & F. Hilliou, unpublished). The mRNA levels were normalized against three constitutively expressed endogenous genes, Mtc27 (Mtr.27459.1.S1_s_at) (^{Van de Velde et al., 2006}) and two genes (Mtr.10324.1.S1_at; Mtr.31250.1.S1_at) selected as reference genes because of their stable expression profile in different microarray results (^{Benedito et al., 2008}). The PCR reactions for each biological replicate were performed in triplicate. For each experiment, the stability of the reference genes across samples was tested using GENORM software (^{Vandesompele et al., 2002}). The absence of genomic DNA contaminations in the RNA samples was tested by PCR analysis of all samples using oligonucleotides bordering an intron in the glutathione synthetase gene of M. truncatula. The gene-specific primers used are listed in the Supporting Information, Table S1.

To detect NO during the early steps of the M. truncatula–S. meliloti interaction, we used the cell-permeable NO-specific fluorescent probe, DAF2-DA (^{Kojima et al., 1998}). Between 2 dpi and 6 dpi, the formation of the fluorescent triazol derivative, DAF-2T, was visualized in infection threads (ITs). For all these experiments, roots were inoculated with a S. meliloti DsRed strain harbouring the pDG77 vector, which exhibits a constitutive red fluorescence. A yellow colour from the merging of the two fluorescent markers occurred (Fig. 1a–d) in the infection pockets, where rhizobia are enclosed within an apoplastic space created by root hair tip curling (the so-called ‘shepherd's crooks’) upstream of the IT. Interestingly, the red fluorescence from the bacteria and the green fluorescence from the DAF probe did not overlap along the ITs (Fig. 1c–e). On a higher magnification (Fig. 1e), the DsRed-labelled bacteria can be visualized in the centre of the IT surrounded by the green fluorescence. In Fig. 1(c) and (d), we can observe the position of the nucleus close to the IT tip indicating that the IT is growing as mentioned by ^{Fournier et al. (2008)}. To validate the specificity of the green fluorescence observed, the roots were treated with the NO scavenger (cPTIO). The cPTIO treatment extinguished the fluorescence and further established that it was representative of NO production (Fig. 1f).

To confirm the presence of NO in the infection threads of M. truncatula root hairs, and to ascertain whether bacteria respond to the presence of NO in this environment, we constructed a reporter plasmid (psma1289-lacZ) carrying a transcriptional lacZ fusion to the S. meliloti sma1289 gene promoter. The sma1289 gene was chosen since it is known to be upregulated by NO via the NO-specific regulator NnrR (^{de Bruijn et al., 2006}; ^{Meilhoc et al., 2010}). As expected, in free-living conditions, expression of the reporter fusion, assessed by measuring β-galactosidase activity, was upregulated approx. 15-fold in the presence of the NO donor spermine NONOate (25 μM) in the wt strain. By contrast, no significant upregulation of the reporter fusion was observed in an nnrR mutant, or in a wt strain carrying the empty vector pCZ962 (Fig. 2). Thus, this strain was used as a NO biosensor strain. The bacteria carrying the reporter fusion were used to inoculate M. truncatula plantlets, and expression of the fusion was assessed, 6 d post-inoculation, by X-Gal staining and microscopic analysis of the roots. Blue-stained bacteria entrapped in ‘shepherd's crooks’ of root hairs (Fig. 3) were detected on roots inoculated with the wt strain, whereas no staining was observed with the nnrR mutant or with the wt strain carrying the empty vector pCZ962 (data not shown). Also, no staining was detected when we used a S. meliloti strain overexpressing the flavohaemoglobin Hmp from plasmid pBBR-MCS5 (^{Meilhoc et al., 2010}). Together, these results confirm that NO is present in the infection pockets and infection threads, and indicate that bacteria respond to NO in infection pockets.

The DAF-2T fluorescence was also observed 4 dpi in the bumps of the root infection zone in the region of infection thread progression. To more precisely localize this NO production, we carried out confocal microscopic analysis on fresh root slices (100 μm) from the infection zone (Fig. 4). All the pictures showed the fluorescence in dividing cortical cells of the nodule primordia (Fig. 4a,b). At the same time, we also detected NO production in lateral root primordia (data not shown), as described in lateral root development in tomatoes (^{Correa-Aragunde et al., 2004}). As expected, this green fluorescence was significantly diminished by treating root slices with cPTIO (Fig. 4c,d). These results show that NO is present in the early stages of the symbiotic interaction.

To determine whether the NO detected at early steps of symbiosis is involved in the infection process, we used an approach to deplete the NO levels. For this, we first tested the ability of the S. meliloti hmp⁺⁺ strain to compete with the wt strain for nodule occupancy. Medicago truncatula plants were coinoculated with a mixture (1 : 1) of the wt and hmp⁺⁺ strain. To differentiate nodules occupied by each strain, the wt harboured plasmid pXLGD4, which displays a constitutive and strong β-gal activity. Nodules occupied by the wt and hmp⁺⁺ strains could thus be numbered after addition of X-Gal by counting blue and white nodules, respectively. In control experiments, the wt strain harbouring pXLGD4 was mixed (1 : 1) with either the empty wt strain, or the wt strain carrying the empty vector pBBR-MCS5.

The competition experiment revealed a lower competitiveness of the hmp⁺⁺ strain compared with the wt (31.6% vs 68.4% nodules occupancy, respectively; P < 0.05 according to Student's t-test) whereas in control experiments, all strains were as competitive as the wt (Table 2). Even when the competition experiment was done with a 1 : 3 mixture of the wt vs hmp⁺⁺, the hmp⁺⁺ strain still remained significantly less competitive (36.5%, n = 1). It has been shown that overexpression of hmp in bacteria can lead to elevated concentrations of superoxide and peroxide (^{Gilberthorpe et al., 2007}; ^{McLean et al., 2010}). To rule out the possibility that such an effect is responsible for the lower competitiveness of the hmp⁺⁺strain, the coinoculation of wt and hmp⁺⁺ strains was repeated in the presence of the NO scavenger (cPTIO). In these conditions, the wt and hmp⁺⁺ strains recovered a similar competitiveness. This confirmed that the low competitivity of the hmp⁺⁺ strain was indeed the result of its greater ability to degrade NO. These results therefore indicate that NO production has a positive effect on the early steps of bacterial infection.

To investigate the role of NO in the establishment of the symbiotic interaction, we first tested the effect of the NO scavenger cPTIO on the nodulation ability of M. truncatula plants inoculated by the S. meliloti wt strain. The plants grown in tubes were treated by addition of 1 mM cPTIO on the whole length of the root 2 h before and 2 h after inoculation, and then every 24 h for 4 d. Appearance of nodules was then monitored. The results show that the NO depletion caused by cPTIO treatment led to a significant delay (3–4 d) in nodule appearance (Fig. 5a). Similarly, plant inoculation with the hmp⁺⁺ strain led to a slight but significant delay in nodule formation (Fig. 5b).

To corroborate these results, transgenic hairy roots overexpressing the bacterial hmp gene, encoding a flavohaemoglobin able to scavenge NO, were obtained. To avoid any effect on root development, the hmp gene was expressed under the control of a nodule specific promoter pMtENOD20. As described in the literature (^{Vernoud et al., 1999}), we observed the expression of the pMtENOD20-GUS construction in dividing inner cortical cells corresponding to the site of nodule primordium from the fifth day of nodule development (see the Supporting Information, Fig. S1). However, compared with the results of ^{Vernoud et al. (1999)}, we did not observe glucuronidase (GUS) staining in associated root hairs displaying shepherd's crook curling. Detection of NO by fluorescent microscopy with DAF2-DA revealed a decrease of fluorescence on the Hmp-overexpressing transgenic nodule slices compared with the control transgenic nodules (Fig. 6a). Using a spectrofluorometric method, we estimated a 40% reduction of the NO production in hairy roots overexpressing the Hmp protein compared with the control hairy roots at 5 dpi (Fig. 6b) as well as at 13 dpi (data not shown). In addition, the overexpression of hmp under the control of this promoter significantly delayed the nodulation (Fig. 6c), with a 28% reduction in the number of nodules compared with the control transgenic roots. This result is in line with the results demonstrated with cPTIO treatment (Fig. 5a) or when plants were inoculated with the hmp⁺⁺S. meliloti strain (Fig. 5b). These results show that decreasing the NO level results in a delay in the nodulation process.

To further investigate the involvement of NO in this process, we measured the expression levels of genes known to be associated with nodule development and potentially regulated by NO. We first analysed the expression of MtCRE1, a gene encoding a cytokinin receptor involved in the early symbiotic interaction (^{Gonzalez-Rizzo et al., 2006}) as it was shown to be NO regulated in M. truncatula (^{Ferrarini et al., 2008}). The qRT-PCR experiments performed either on inoculated plants treated with cPTIO (during 8 h) or on transgenic hairy roots overexpressing the Hmp protein (Fig. 7a,b) showed that NO depletion downregulates the expression of MtCRE1. This result therefore confirms that NO upregulates the expression of MtCRE1, and further points to the presence of NO during the early steps of nodule development.

We then measured the expression levels of several well-studied genes associated with various infection and nodule developmental stages: MtRR4, MtNIN, MtN6, MtENOD40 and MtCCS52A (^{Pichon et al., 1992}; ^{Charon et al., 1997}; ^{Cebolla et al., 1999}; ^{Vinardell et al., 2003}; ^{Gonzalez-Rizzo et al., 2006}). Only one of them, MtCCS52A, displayed a lower expression level upon NO depletion by either cPTIO or Hmp overexpression (Fig. 7 and data not shown) suggesting that the gene expression could be induced by the NO production observed at this stage. During nodule organogenesis, the cell cycle-switching gene MtCCS52A triggers selected cells within the primordium to switch from mitotic cycles into endoreduplicating cycles (^{Cebolla et al., 1999}; ^{Vinardell et al., 2003}). These data therefore strengthen the role of NO in nodule development.