M. truncatula seeds were scarified with concentrated sulfuric acid for 5 to 8 min, rinsed with deionized water for five times, surface sterilized with 5% sodium hypochlorite solution and rinsed with autoclaved deionized water for several times. After vernalization at 4°C for 3 days, seedlings were transplanted to pots and grown in greenhouse with 16 hr/8 hr light/dark cycle and 150 E/m2/sec light intensity. Perlite and sand (volume ratio of 3:1) were used as the planting medium. The perlite and sand were separately rinsed with deionized water, and then mixed in proportion for use. The germinated seeds were transferred to a 72-cell tray (530 mm × 270 mm, with a single well volume of 40 mL), watered with 1 L of 0.125 mM KNO₃ (low-nitrogen nutrient) solution. Rhizobia were inoculated after one week of growth in the greenhouse.

The rhizobia strains were removed from a -80°C refrigerator. Two mL of the rhizobia were inoculated into 50 mL of TY liquid medium containing tetracycline and cultured at 28°C for 2 days. Then, the bacterial cultures were centrifuged at 4,000 rpm for fifteen min in a 50 mL centrifuge tube, re-suspended and diluted to OD600≈0.05 with the low nitrogen nutrient solution. Five mL of the above bacterial suspension were inoculated to each plant.

Three to five plants inoculated with S. medicae ABS7 strain carrying the hemA::LacZ reporter were harvested at 6 dpi, 8 dpi, 10 dpi, 13 dpi and 16 dpi (days post inoculation). Roots were washed with deionized water. The root nodules were fixed as previously described (Xi et al., 2013). For GFP studies, nodules from plants inoculated with the S. meliloti Sm2011 strain harboring hemA::mGFP reporter gene were washed with deionized water, embedded in 6% agarose, and sectioned with Vibratome™ 1000 Plus (Leica Biosystems, Germany). Nodule sections (50-70 μm) were mounted on glass slides and imaged using MVX10+DP74 microscope (Olympus).

Plants inoculated with S. medicae ABS7 strain were used for 50-70 μm nodule section preparation. 5 μM SYTO9 and 30 μM PI were added to 50 mM Tris (pH7.0) buffer to make a staining solution. Nodule sections were stained for 20 min at room temperature. Sections were placed on slides covered with coverslips and observed with a confocal microscope [Nikon Confocal laser scanning microscope (A1R+ Ti2-E)].

Five-micrometer semi-thin sections were placed on glass slides and stained with 1% potassium iodide phosphate buffer in dark condition for 10 min. Then, the image was observed with microscope [Axio Imager.Z2 (ZEISS)].

The nodules were placed in the fixing solution, vacuumed for 10 min, and fixed overnight at 4°C. The fixed nodules were dehydrated in 70%, 85% and 95% ethanol series for 30 min each, and in 100% ethanol for 2 h. Finally, 0.1% eosin Y3 was added to 100% ethanol to dye for half an hour. After the nodules turned red, samples were replaced in a mixture of 100% ethanol and base liquid Technovit 7100 resin (v/v=1:1) for 2 h. The nodule samples were transferred to the infiltration solution overnight (infiltration solution: 1 g hardener I to 100 mL base liquid Technovit 7100 resin. Allow approximately 5 min for complete dissolution). The nodule samples were embedded after the permeation treatment. The temperature was set to 40°C and the embedding mould was dried beforehand. The infiltrate and hardener II (v/v=15:1) were mixed gently, and 200 μL of the mixture was added to each well of the embedding mould. The nodule materials were carefully picked up with forceps and placed on the tip of the embedding mould, with materials placed vertically to facilitate sectioning. The samples were arranged neatly and then polymerized overnight in a 40°C incubator. The Leica RM2265 rotary microtome was used to cut sections with a thickness of 3-5 μm. The semithin sections were placed on glass slides and stained with 1% toluidine blue for 5 min. Finally, the stained sections were observed and imaged with microscope [Axio Imager.Z2 (ZEISS)].

The whole genome sequencing (WGS) analysis was performed by Shanghai Oebiotech Co. The compound leaf samples were collected from wild type (WT) A17 and FN007 mutant, respectively. The genomic DNA was extracted using the DNeasy plant mini kit (Qiagen Shanghai, China), following the supplier’s instructions. The library was constructed after the DNA quality was passing the bioanalyzer test. The DNA samples were randomly sheared into 350 bp-500 bp fragments, and the TruSeq DNA LT Sample Prep Kit was used for library construction. The DNA fragments were treated with the following steps to complete library construction: end repair, addition of ployA tails, addition of sequencing connectors, purification, and PCR amplification. After library construction and qualification, pair-end sequencing was performed on the sequencer (HiSeq X Ten sequencer).

The raw sequencing data were processed in two stages. Sequencing data quality was mainly examined through the sequencing error rate, data volume, comparison rate, coverage and other statistics to assess whether the sequenced library meet the standard for subsequent analysis. Then, the high-quality sequences were compared to the reference genome to detect the variation information in the samples. SNP (single nucleotide polymorphisms) and InDel (insertion or deletion) detection was performed using the GVCF mode of the Haplotypecaller module of the GATK (Version: 4.1.0.0) software, based on the results of the sample comparison with the reference genome. To reduce the error rate of SNP and InDel assays, a criterion of QD more than 2.0 was chosen for filtering, and only mutant loci that met this criterion were retained. QD is the ratio of the quality of variation divided by the depth of coverage, which is actually the quality of variation per unit depth. The SNP and InDel results were annotated using SnpEff (Version: 4.4) software. Finally, the genomic variation information detected was visualized using Circos plots.

To eliminate the effect of sequencing errors on the results, the original sequencing data were cleaned using the pre-processing software Fastp. Clean reads were compared to reference genome using BWA, and the comparison results were converted to SAMtools, and de-redundant using Picard (http://broadinstitute.github.io/picard/), and the results were analyzed using the Qualimap software (Li and Durbin, 2009; Li et al., 2009; Chen et al., 2018). M. truncatula reference genome can download at the website: https://medicago.toulouse.inra.fr/MtrunA17r5.0-ANR/(MtrunA17r5.0).

Thermal asymmetric interlaced PCR (TAIL PCR) assay was carried out essentially as previously described (Liu and Chen, 2007). The specific primer P1 adjacent to the proposed deletion border was combined with five short randomly primers (AD1, AD2, AD3, AD5, AD6) with low Tm values and amplified by PCR with thermally asymmetric temperature cycling, using the FN007 mutant genomic DNA as template.

The resulting PCR product was amplified again to obtain a flanking sequence using the nested primer P2 of P1 ( Table S1 ) and the five random primers mentioned above. The PCR product was sequenced and compared with the genome sequence of M. truncatula A17 to confirm the deletion junctions in FN007 mutant.

M. truncatula FN007 mutant was isolated from a Fast Neutron (FN) induced deletion mutant collection in the Jemalong A17 background. Under Nitrogen replete condition, the FN007 mutant is comparable with Wild Type (WT) plants in growth and development processes, including leaf, seed pod, flower and root development (data not shown). But under Nitrogen-limited and symbiotic condition, the FN007 mutant plant develop nitrogen starvation symptoms including stunted growth and yellowish leaves at 16 days post inoculation (dpi) with S. medicae ABS7 strain, indicating that these phenotypes were caused by defective symbiotic nitrogen fixation ( Figures 1A, B ). Nodules developed on the FN007 mutant roots were small and spherical in shape, most with white colors (due to degradation of leghemoglobin), whereas in the WT plant, the pink cylindrical nodules were developed ( Figures 1C–F ). Moreover, nodule distribution was concentrated in WT plants but scattered in FN007 mutant and the WT plant had fewer nodule numbers than the FN007 mutant.

To further characterize the nodule development processes of the FN007 mutant, time-course experiments were carried out. Under low nitrogen condition (0.125 mM KNO₃), the growth and development of root nodules in the FN007 mutant and WT plant inoculated with S. medicae ABS7 strain was continuously monitored and recorded. Morphological and color differences between FN007 mutant and WT nodules became apparent over time ( Figure 2 ). At 6 dpi, WT ( Figure 2A ) and FN007 mutant ( Figure 2F ) nodules appeared to be similar, both white and spherical shape. At 9 dpi, WT nodules ( Figure 2B ) became pale pink color and conical in shape, while FN007 mutant nodules ( Figure 2G ) were still white color and cone-shaped. At 12dpi, WT nodules ( Figure 2C ) were pink and cylindrical shape, but FN007 mutant nodules ( Figure 2H ) gradually increased in size with increasing growth time, and remained white color. At 15 dpi, WT nodules ( Figure 2D ) continued to grow, almost entirely pink, and were already mature and capable of efficient nitrogen fixation. FN007 mutant nodules ( Figure 2I ) grew slowly, with one or two white and rod-like, and the remainder were basically white spheres or cones shape. At 19 dpi, WT nodules ( Figure 2E ) were generally the same as those at 15 dpi, but the number of mature nodules increased, while FN007 mutant nodules ( Figure 2J ) still appeared as white rod shape. Time-course experiments showed that early nodule development (6 dpi) appears to be normal in the FN007 mutant. However, differences began apparent at 6-9 dpi with leghemoglobin accumulation in WT nodules. Even though the FN007 nodules did not show the pink functional leghemoglobin, some nodules were able to elongate and grow into rod-like shape during later development stage, indicating that the nodule meristem still had normal developmental functions.

Under low nitrogen condition (0.125 mM KNO₃), the FN007 mutant and WT seedlings grown for one week in greenhouse were inoculated with S. medicae ABS7 strain containing the hemA::LacZ reporter gene. Nodules were harvested at different development stages after inoculation and stained with LacZ staining buffer. At 12 dpi, in FN007 mutant nodules ( Figure 3B ), there were none or few rhizobia in nodule cells. For WT plant at the same time ( Figure 3A ), rhizobia filled the entire nodule cells. At 15 dpi, the number of rhizobia-containing nodule cells in FN007 mutant ( Figure 3D ) was decreased, and nodule cell debris was observed, at which point a distinct senescence zone (rather than a nitrogen-fixing zone) had formed, which was significantly different from the WT ( Figure 3C ). We speculated that between 6-7 dpi, the FN007 mutant failed to form a nitrogen-fixing zone and went directly into senescence, forming a region that accounted for an increasing proportion of nodules over time. To confirm this result, another rhizobia strain with GFP reporter gene, S. meliloti Sm2011-GFP, was inoculated. At 11 dpi, the fluorescence signal in wild type nodules was very strong, indicating that rhizobia released and filled with nodule cells ( Figure 4A ). However, in FN007 mutant nodules, though the fluorescence intensity of zones I and II was no different from the WT, zone III fluorescence was distributed in a punctiform pattern and no mature symbiont was formed ( Figure 4B ), indicating that rhizobia release into nodule cells during infection was normal, which was consistent with the results of nodule sections inoculated with S. medicae ABS7 strain. At 14 dpi, the WT nodules showed characteristic partitioning and a dominant green fluorescent signal in mature symbiotic cells completely filled with bacteroids ( Figure 4C ), whereas in the FN007 mutant nodules an increasingly severe degradation of the bacterial cycle in the nitrogen-fixing zone was observed ( Figure 4D ). This observation suggests the FN007 mutant nodules may undergo early senescence.

In indeterminate nodules, rhizobia released into nodule cells differentiate before nitrogen fixation begins, and the differentiated rhizobia are morphologically longer than free-state rhizobia and have endoreduplication of their chromosomes (Mergaert et al., 2006). From GFP and LacZ stained sections experiments, we found that the FN007 mutant was normal in early nodule development stage including infection thread development and rhizobia release. To further determine the development and differentiation of rhizobia released into nodule cells, 5 μm semithin longitudinal nodule sections were stained with Toluidine blue for microscopic observation. In healthy symbionts, rhizobia undergo intracellular replication and differentiation into rod-like structures evenly distributed around a central vacuole, but rhizobia in infected cells need to differentiate to exert nitrogen-fixing function properly. At 13 dpi, WT nodules ( Figure 5A ) had formed four distinct zones: zone I (zones of meristem, ZI), zone II (infection, ZII), zone II-III (a transition zone between ZII and ZIII, IZ), and zone III (nitrogen fixation, ZIII). Zone II-III was a transition zone from zone II to III, when rhizobia had already entered into infected cells, and with a compartment area of about one-fifth of the total rhizobia area. Infected cells were filled with intact symbiosomes in WT. In this period, the FN007 mutant nodules appeared to have four zones in terms of cell morphology: I, II, IZ, and IV (senescence zone), but did not appear to have the nitrogen-fixing zone (ZIII) as observed in WT nodules, entering directly into the senescence zone. Symbiosomes in most infected cells were disintegrated at the proximal end of the root in the FN007 mutant ( Figure 5D ). In the WT, the senescence zone was formed at 35 dpi in the absence of environmental stress, but in the FN007 mutant, the senescence zone was formed at 13 dpi, which clearly showed an early senescence phenotype. Therefore, it was assumed that symbiotic nitrogen fixation defect occurs in the II-III transition zone, and rhizobia cannot complete the transition after its release into the infected cells in the FN007 nutant. In WT nodules ( Figures 5B, C ), infected cells were morphologically normal, mostly subrounded, with a large number of symbiosomes containing rod-shaped bodies regularly arranged around a large central vacuole. In the FN007 mutant, infected cells ( Figures 5E, F ) were smaller and irregularly shaped compared to that in WT nodules, with less symbiosomes and most were not rod-shaped, resembling the free-living cells.

To further characterize the symbiotic phenotype of the FN007 mutant, semithin sections of nodules at 13 dpi were stained for starch granules with potassium iodide (KI). The WT nodules had almost no accumulation of starch granules at the distal root end ( Figures 6A, B ), but in the nitrogen fixation zone ( Figure 6C ), there was a small number of starch granules deposition around the cell membrane. Conversely, starch granules in the distal end of the FN007 nodules were evenly distributed and surrounded the cell membrane ( Figure 6D ), where clearly cell contours were observed and the starch granules were similar in size. At the proximal root end of the FN007 mutant nodule ( Figures 6E, F ), the number of deposited amyloplasts was increased and distributed in a disorganized manner. Overall, the number of starch granules in FN007 mutant nodules were greatly exceeded that of the wild type. These results indicated that the FN007 mutant cannot form a functional fixation zone, and thus many metabolic activities can not be carried out in infected cells, which accumulating more energy material (starch grains).

To confirm whether the FN007 mutant nodules undergo early senescence, we performed live/dead staining experiment, using fluorescence dyes SYTO9, which stains live bacteroids and nucleic acids of plant cells and propidium iodide (PI), which stains dead bacteroids. The FN007 mutant and WT plants were inoculated with S. medicae ABS7 strain. A large green fluorescence was observed within the WT nodule ( Figures 7A–C ), and clearly visible symbiosomes were formed within infected cells and regularly arranged around the central vacuole. At 13 dpi, the FN007 mutant nodules were mostly stained red fluorescence compared with the WT ( Figures 7D–F ), and high magnification images showed that infected cells contained both red and green scattered bacteroids, indicating that some bacteroids in the FN007 mutant were dead at this time. These observations suggest that nodule cells of the FN007 mutant exhibited signs of an early senescence phenotype (12 dpi).

To identify mutations in the FN007 mutant, whole genome resequencing was carried out. Raw reads were cleaned after quality control and QC results were obtained ( Table 1 ). Normally, for next generation sequencing values, Q20 should be over 95% (at worst, not below 90%), and Q30 is required to be more than 85% (at worst, not below 80%). The sequencing quality of the FN007 mutant and WT was high, with both Q20 values over 97% and both Q30 values over 93% ( Table 1 ). The distribution of base error rates for pair-end reads in length range of 1 to 150 bp is shown in Figure 8 . Regardless of read length, the average base error rate was less than 0.05%, indicating high sequencing quality.

Whole-genome sequencing uses non-strand-specific library building. Theoretically, G and C, and A and T bases should be equal in each sequencing cycle, and the whole sequencing process should be stable. The FN007 mutant base distribution sequencing data is shown in Figure 9 . The start position of the sequence is linked to the primer junction during sequencing, so A, C, G and T will fluctuate at the beginning but stabilize later.

For species with a reference genome, their genomic information or complete transcript information can be downloaded from public databases, and then the sequenced reads can be compared to the reference genome. By quickly comparing each read to the reference genome, information about read position and match quality on the compared genome or other reference sequences can be obtained, and then the genes or transcripts can be annotated and quantified. After comparing the clean reads of the FN007 mutant and WT with the reference genome, the statistical results are shown in Table 2 . It was suggested that reads from the FN007 mutant is comparable with WT. For the FN007 mutant, more than 189 million pair-end reads (2×150 bp) were produced, which represented a 65.2× sequencing depth. For the WT plant, 151 million reads were produced, representing 52.3× sequencing depth ( Table 2 ). These reads were mapped to the latest reference genome of M. truncatula (MtrunA17r5.0). The total mean Mapping quality is nearly 50% in both mutant and WT. These result indicated that the WGS data has higher quality of the alignment.

SNP and InDel detection were performed using the GATK4 software Haplotypecaller module, based on the comparison of samples with the reference genome. QD is the ratio of variant quality divided by the depth of coverage, which is actually variant quality per unit depth, and most false positive variants have a QD of less than 2. Compared to transversion, transition is the most common mutation type, with T to C and A to G being the most common substitution types ( Figure 10A ). Of all the InDels obtained, deletions less than 4 bp accounted for 85.28% and insertions less than 4 bp accounted for 94.53% ( Figure 10B ). All the mutations detected were integrated into a Circos plot to visualize the genomic data ( Figure 10C ).

The FN007 mutant resequencing data revealed that there was a copy number variation (CNV) in large segment of chromosome 3. CNVs may be biologically important for species-specific genome composition, species evolution and phylogeny, as well as for the expression and regulation of genes in specific genome regions. We used IGV software to visualize detection CNVs in the FN007 mutant, and found a possible large deletion on chromosome 3 from location 4,518,720 to 4,871,880 ( Figure 11A ). To determine the specific interval of the mutant deletion, four pairs of primer ( Table S1 ) were first designed on both sides of each proposed boundary provided by the IGV data, and these primers were amplified by PCR in the FN007 mutant and WT ( Figure 11B ). The results showed that primers 3 and 4 could amplify the target bands in the FN007 mutant, but in contrast, primers 1 and 2 could not ( Figure 11C ). This suggests that the IGV data matches the actual deletion junction in the FN007 mutant. To identify the exactly deletion site, TAIL PCR was performed to identify the deletion regions in the FN007 mutant ( Figure 11D ). Results showed that the deletion border in the FN007 mutant was located at position 4,519,828-4,872,290 on chromosome 3. Based on the TAIL PCR results, we designed left and right border primers and verified the results, conformed that there was a deletion region in the FN007 mutant compared with WT ( Figure 11E ).

Whole genome sequencing analysis revealed that the FN007 mutant has a large deletion of approximately 348 kb on chromosome 3. To characterize the nature of the FN007 mutation, we generated F2 populations from cross-pollination between the FN007 mutant and its WT parent (Jemalong A17).

We analyzed 163 individual F2 plants and observed a ratio of 3.075:1 segregation of WT and mutant nodulation phenotypes in F2 populations, indicating a single recessive locus for the FN007 mutant phenotype. Next, we performed a linkage analysis, using pools of DNA samples isolated from individual F2 plants. The 123 plants with the WT phenotype were divided into 12 groups and 40 plants with the mutant phenotype were divided into 4 groups (the DNA from each group was pooled together). The deletion regions were identified by PCR using the corresponding primers ( Figure 12A ). The results showed that all the mutants plant could amplify the deletion border by MDB-F and MDB-R primer (data not shown), but could not get the corresponding band by the MDB-F and DB-R primer, while the WT plants can amplify the correct bands in the deletion region ( Figure 12B ). These results indicated that the fix^- phenotype of the FN007 mutant was tightly linked to the deleted region on chromosome 3.

Within the deletion region on chromosome 3, there are 53 annotated genes base on M. truncatula A17 r5.0 genome ( Figure 12C ; Table 3 ). This, together with the results of the genetic linkage analysis, suggests that these genes are candidates for the cause of the phenotype of the FN007 mutant. Though we have successfully identified the deletion region in the FN007 mutant, it will take more effort to pinpoint the causal gene. To identify which is the target gene, RNA-seq technology would be used to narrow down the gene list in the deletion region. Finally, genetic complementation assay should be carried out, using Agrobacterium rhizogenes-mediated root transformation.