Defining the mutation sites in chickpea nodulation mutants PM233 and PM405

Like most legumes, chickpeas form specialized organs called root nodules. These nodules allow for a symbiotic relationship with rhizobium bacteria. The rhizobia provide fixed atmospheric nitrogen to the plant in a usable form. It is of both basic and practical interest to understand the host plant genetics of legume root nodulation. Chickpea lines PM233 and PM405, which harbor the mutationally identified nodulation genes rn1 and rn4, respectively, both display nodulation-deficient phenotypes. Previous investigators identified the rn1 mutation with the chickpea homolog of Medicago truncatula nodulation gene NSP2, but were unable to define the mutant rn1 allele. We used Illumina and Nanopore sequencing reads to attempt to identify and characterize candidate mutation sites responsible for the PM233 and PM405 phenotypes. We aligned Illumina reads to the available desi chickpea reference genome, and did a de novo contig assembly of Nanopore reads. In mutant PM233, the Nanopore contigs allowed us to identify the breakpoints of a ~ 35 kb deleted region containing the CaNSP2 gene, the Medicago truncatula homolog of which is involved in nodulation. In mutant PM405, we performed variant calling in read alignments and identified 10 candidate mutations. Genotyping of a segregating progeny population narrowed that pool down to a single candidate gene which displayed homology to M. truncatula nodulation gene NIN. We have characterized the nodulation mutation sites in chickpea mutants PM233 and PM405. In mutant PM233, the rn1 mutation was shown to be due to deletion of the entire CaNSP2 nodulation gene, while in mutant PM405 the rn4 mutation was due to a single base deletion resulting in a frameshift mutation between the predicted RWP-RK and PB1 domains of the NIN nodulation gene. Critical to characterization of the rn1 allele was the generation of Nanopore contigs for mutant PM233 and its wild type parent ICC 640, without which the deletional boundaries could not be defined. Our results suggest that efforts of prior investigators were hampered by genomic misassemblies in the CaNSP2 region of both the desi and kabuli reference genomes.

Frailey et al. BMC Plant Biology (2022) 22:66 Chickpeas and most other legumes form root nodules, specialized organs that are inhabited by rhizobium bacteria. Chickpeas are nodulated by Mesorhizobium species [4]. Nodule formation allows legumes and rhizobia to form a symbiotic relationship. The plant provides carbon and other nutrients to the bacteria, which colonize the nodules, and the rhizobia provide fixed atmospheric nitrogen in a form usable by the plant [5].
The process of nodule formation is called nodulation, and it depends on gene products from both the rhizobia and the plant. Legume roots secrete flavonoids, which induce nodulation (nod) gene expression in the rhizobia. Rhizobial nod genes encode the biosynthetic machinery to synthesize lipo-chitooligosaccharide (LCO) molecules, which are known as Nodulation (Nod) factors. These are detected by the plant and trigger nodule formation. After Nod factor recognition, there are many additional plant genes encoding proteins involved in downstream signaling, nodulation regulation, and metabolic function [6].
Close to a hundred genes involved in nodulation have been discovered in the model legumes, Medicago truncatula and Lotus japonicus [7] through a combination of forward and reverse genetics studies. Many genes involved in nodulation were originally identified as nodulin genes, i.e., genes with increased expression levels in the nodules compared to other plant tissues [8]. Reverse genetics experiments showed that many of these genes encoded proteins involved in nodulation [9][10][11][12][13][14][15]. Of greater relevance to the presence study, forward genetic studies have identified additional genes involved in nodulation [16]. Several M. truncatula mutant lines deficient in nodulation were generated by γ-ray irradiation and EMS exposure [17,18]. The genes responsible for these mutant phenotypes were further characterized in followup studies [19][20][21][22].
Five of the mutant chickpea lines generated by Davis et al. [23,24] were the initial subjects of the present study, which ultimately focused on mutants PM233 and PM405. Mutant PM233 fails to form nodules when inoculated with rhizobia and grown in nitrogen-deficient media, and is classified as having a non-nodulation phenotype [23,33]. In contrast, mutant PM405 can form nodules in the presence of rhizobia, but these nodules are small and white, and the nodulated plants exhibit nitrogen deficiency, on which basis PM405 was classified as having an ineffective nodulation phenotype [25].
Of the seven chickpea nodulation mutants enumerated above, only one has been subjected to any form of molecular analysis. On the basis of linkage analysis [34] and cross-referencing of nodulation gene positions in the available kabuli chickpea [2] and M. truncatula [35] reference genomes, Ali et al. [34] identified the chickpea homologue of M. truncatula Nodulation Signaling Pathway 2 (NSP2) as the likely gene underlying the mutant phenotype seen in PM233. These investigators extracted the wild type chickpea homolog CaNSP2, numbered in the available kabuli chickpea reference genome annotation as Ca_13583 [2]; however, they did not recover or define the mutant rn1 allele at the DNA sequence level.
We have initiated investigations aimed at illuminating the nodulation gene mutations in the six mutagenically derived chickpea nodulation mutants described above. Resources available for this study included the wild type desi line ICC 640 and its derivative mutant lines, the currently available desi [36] and kabuli [2] chickpea reference genomes, and next generation sequencing data sets that we have generated from the wild type and mutant chickpea lines of interest. The results we report here provide new insights into the molecular basis of defective nodulation in chickpea mutants PM233 and PM405.
Nanopore sequencing provided an additional 19.2 Gb (22x coverage) of PM233 sequence and 24.6 Gb (28x coverage) of ICC 640 sequence, as exemplified by the ICC 640 length distribution shown in (Additional file 1: Fig. S1). De novo assembly of the ICC 640 Nanopore reads generated 4860 contigs totaling 584,434,807 bp, with a median length of 35,009 bp and an N50 length of 481,908 bp. Similarly, de novo assembly of the PM233 Nanopore reads generated 10,566 contigs totaling 438,365,021 bp with a median length of 19,712 bp, and an N50 length 85,636 bp.
Variant discovery using Illumina sequence aligned to the ICC 4958 v3.0 desi reference genome [36] identified numerous unique variants in each mutant. Upon implementation of our filtering pipeline, a set of final candidate genes was identified in each of these five mutants (Table 1).
For our initial report based upon ongoing analysis of these data, we have chosen to focus on two mutants: PM233 because it had been the subject of a prior study resulting in identification of the CaNSP2 candidate gene on chickpea chromosome Ca5 [34]; and PM405 because its analysis immediately yielded a promising candidate gene, as detailed below. Findings with respect to the remaining nodulation mutants will be reported as those studies reach fruition.

PM233
Upon their alignment to the ICC 4958 v3.0 desi reference genome, the absence of PM233 Illumina reads aligning to a 35 kb region on chromosome Ca5 that contains the CaNSP2 candidate gene region ( Fig. 1) was immediately apparent, suggesting that this region, ranging from chromosomal coordinates ~ 1255 kb to ~ 1290 kb in the Ca5 reference sequence, was deleted in PM233. There were also several small gaps in the alignment and stretches of N's in the assembly flanking the putative deletion site, preventing the determination of the precise deletion boundaries using Illumina reads alone (Fig. 1).
Based upon their targeted locations in the ICC 4958 v3.0 desi reference sequence [36], the primer pairs UpF/ UpR and DownF/DownR, each consisting of one primer external to, and one primer internal to, the putative deletional gap, were expected to produce amplicons of a predicted size in ICC 640, but no products from PM233 due to deletional loss of internal primer binding sites in the mutant allele. The UpF-UpR primer pair binding to the upstream target worked as expected, producing a PCR product in ICC 640 but not in PM233. However, the DownF-DownR primer pair for the downstream end did not produce an expected product in ICC 640, and the UpF-DownR primer pair did not produce the expected product in PM233 (Additional file 2: Table S1).
Out of concern that the failure to produce expected amplicons might be attributable to either sequence polymorphism between ICC 640 and ICC 4958, or perhaps an error in the v3.0 reference assembly, we identified informative contigs and reads in our Nanopore-based ICC 640 and PM233 contig assemblies. One of the ICC 640 contigs, number c636 (accession number MZ047314), was ~ 758 kb in length and spanned the entire deletional gap in PM233. When compared to both the ICC 4958 v3.0 desi reference genome [36] and the kabuli reference genome, an internal ~ 90 kb region of contig c636 aligned to the CaNSP2 candidate gene region on Ca5 of both reference genomes (Fig. 2). However, outside of this region of alignment, the flanking contig segments aligned elsewhere ( Fig. 2) on pseudochromosome Ca5, indicative of structural disagreement between our contig c636 and the two reference genomes.
The ~ 90 kb region of alignment between the ICC 640 Nanopore contig c636 and the two chickpea reference genomes encompasses most of the PM233 deletional gap, including the upstream deletion boundary. However, the downstream deletion boundary is outside of this region and is located where the contig and reference assemblies diverge. The PM233 Nanopore reads that aligned to the upstream end of the deletion did not align to the downstream end in the reference assembly ( Fig. 1), but did align continuously with the downstream sequence of our contig c636 (not shown).

Table 1 Outcomes of variant and candidate gene identification pipeline in chickpea mutants
Legend: The listed variant counts are those remaining after implementation of each respective filtration step. The Total column includes all variants found before filtering. The CDS (coding sequence) column includes remaining variants after filtering only for variants found within coding sequence. The Unique column are all variants found only within a single chickpea line. The 2 Alleles column are variants with 2 alelles. The Read Depth column are variants with a read depth of at least 2. The Min Allele Frequency are variants where the alternate allele (the one not matching to the reference sequence) had a frequency of at least 0.9. The Manual IGV (Integrative Genomics Viewer) Check column included variants confirmed after manual confirmation using IGV (i.e. variants which were present in other chickpea lines but not called as variants were excluded)

Sample
Filter Criterion When our contigs spanning the deletion site c636 and c2 (accession number MZ047315) from ICC 640 (wild type) and PM233 (mutant), respectively, were aligned, they were entirely collinear, except for the contig c2 deletional gap, the position of which is depicted in yellow in contig c636 (Additional file 3: Fig. S2).

PM405
We did not find any sizable regions of the desi ICC 4958 v3.0 reference assembly that were so poorly covered by PM405 Illumina reads as to suggest any large deletion in PM405. After filtering for read depth, alternate allele frequency, and location within CDS as specified by the ICC 4958 v3.0 genome reference genome [36], we identified 10 sequence variants present only in PM405 and not in ICC 640 or the other nodulation mutant read sets ( Table 2). These variants were distributed over four of the eight reference genome pseudochromosomes (Table 2).
Of these ten coding sequence variants, eight were substitutions, of which six were synonymous and two were non-synonymous (missense). The remaining two variants were deletions, one within-frame and one causing a frameshift ( Table 2). The frameshift mutation consisted of a single base deletion in PM405.
Each of the ten PM405 variants was in a different gene. When the functional annotations of these ten genes were examined, only one, the Ca_06500 gene at position 38,602,732 on chromosomes Ca2 of the ICC 4958 v3.0 genome assembly had an obvious functional relationship to nodulation. This gene exhibits orthology to the Medicago NODULE INCEPTION (NIN) gene, which is known to be essential for nodule formation. The coding sequence mutation in this gene is a single base deletion (CAA -> CA_), and is located at the 2189th nucleotide position of the CDS of a gene annotated as Ca_06500, and produces a frameshift starting  [36], and expected amplicons (brown bars) A1 through A4, as defined in (Additional file 2: Table S1). The location of the CaNSP2 gene is depicted as a red arrow in the reference track. Its 5′-3′ orientation is left to right. The deletional gap in PM233 is shaded in yellow. The PM233 Nanopore reads that aligned to the left of the deletional gap are not continuous with those that aligned to the right of the gap at that position (Additional file 4: Fig. S3). This deletion was present in all PM405 reads aligned to the site and none of the reads from ICC 640 or the other mutant lines. The full Ca_06500 gene from start codon to stop codon is 3216 bp and spans from positions 38,600,214 to 38,603,430 on pseudochromosome Ca2 of the reference assembly [36], according to which the CDS covers a total of 2888 bp (Additional file 4: Fig. S3).
The gene annotated as Ca_06500 comprises four exons and three introns (Additional file 4: Fig. S3). When we ran a BLASTn alignment of the Ca_06500 CDS against the M. truncatula NCBI database, the top result was the nodule inception (NIN) gene. The reciprocal BLASTn of the M. truncatula NIN gene against the ICC 4958 v3.0 chickpea genome aligned to Ca_06500 as the top result. The chickpea Ca_06500 and M. truncatula NIN gene coding regions were aligned in MegAlignPro and shared 72.6% nucleotide sequence identity and 68.5% amino acid identity. The frameshift in PM405 occurs in the fourth exon between predicted RWP-RK and PB1 domains and removes the PB1 domain in the predicted protein encoded by the PM405 allele. On the basis of its potential functional impact, we chose to study this deletion further as a candidate for the rn4 mutation in PM405.
In order to assess the association between the Ca_06500 mutation and the ineffective nodulation phenotype in PM405, we phenotyped and genotyped a segregating F2 population from the cross ICC 640 x PM405. Of 24 seeds planted, 19 F2 seeds from this cross germinated to produce F2 plants, which were phenotyped in Leonard Jars using the procedures described. Fourteen of these plants formed normal looking root nodules and five lacked nodules.
In parallel, genotyping of the candidate Ca_06500 mutation site in these 19 F2 plants was accomplished using allele specific primers designed to produce amplicons of differing sizes from the wild type and mutant alleles. A common forward primer (06500F) was used in combination with two different, allele-specific reverse primers (Additional file 5: Table S2, Additional file 6: Fig.  S4), 06500Rwt and 06500Rm. The 3′ ends of the allelespecific reverse primers span the single bp deletion, and differ in sequence at that site. The mutation-specific primer 06500Rm also included a 21-nucleotide 5′ tag (Additional file 6: Fig. S4), thus creating an allele-specific mobility shift. This PCR assay (Additional file 5: Table S2) allowed us to genotypically characterize the 19 F2 plants.
PCR using the 06500Rwt reverse primer produced bands in all fourteen nodulating chickpeas and none of the non-nodulating. PCR using the 06500Rm reverse primer produced bands in all five of the non-nodulating plants and nine of the nodulating plants (Additional file 6: Fig. S4). Thus, the normally nodulating F2 plants were of two types: five possessing only the wild type allele, and by inference homozygous; and nine possessing both wild type and mutant alleles as diagnostic of heterozygosity. The five non-nodulating plants possessed only the mutant allele, for which they were presumably homozygous. Within this small data set, the Ca-06500 candidate gene was linked without evident recombination to the nodulation phenotype.
Of the nine other candidate mutations identified in PM405 (Table 2), only one of them was located on the same pseudochromosome as the Ca_06500 gene. Genotyping of this marker established the presence of recombination between this gene and the phenotypically defined locus, excluding this gene from further consideration, as detailed in legend of (Additional file 6: Fig.  S4). Based on these analyses, a mutation in the Ca_06500 (NIN) gene is indicated as the most likely cause of the phenotype in chickpea mutant PM405.

Discussion
Our analyses have provided new insights into the identities and phenotypic influences of two chickpea root nodulation genes, PM233 and PM405. The variant calling pipeline identified several (ranging from six to fifteen) mutation sites that fulfilled the filtration criteria in each mutant ( Table 1). The respective candidate gene lists are provided here for mutants PM233 (see Additional file 8: Table S4) and PM405 (Table 2), while those for the other three mutants will be presented as components of future studies. Based on the previous report of Ali et al. (2014), our a priori focus in PM233 was the nodulation gene CaNSP2, which in both the kabuli [2] and desi [3] reference assemblies is located on pseudochromosome Ca5. However, it was also of interest to apply our variant identification pipeline to PM233 to determine whether mutations in other genes of potential interest might be present, with specific attention to those that might have obvious nodulation functions (if any) located on Ca5. Eight such variant-containing genes were identified on PM233 Ca5 (see Additional file 8: Table S4); three of which were functionally uncharacterized and the other five had no obvious functional connection to nodulation.

PM233
Here, our focus was to test the hypothesis of Ali et al. (2014) that the CaNSP2 gene was the site of the rn1 mutation conditioning non-nodulation in chickpea mutant PM233, which one of us (Davis) had initially isolated as part of an induced mutagenesis project [23,24]. Furthermore, our aim was to characterize the rn1 mutant allele at the DNA sequence level, where previously only the sequence of the wild type CaNSP2 (= Rn1) allele had been [34]. Using a combination of Illumina and Nanopore sequencing data and constructed contigs from PM233 and wild type ICC 640, we were able to identify and characterize the mutation in PM233 as a ~ 35 kb deletion that completely deletes the CaNSP2 gene as well as some flanking sequence, but no additional annotated genes. This finding is consistent with the recessive character of the PM233 rn1 allele [24,33]. A possible reason why we were able to define the rn1 allele at the DNA sequence level, while Ali et al. [34] were not, is explored in a later section of the Discussion. Confirmation of the identity between CaNSP2 and Rn1 via molecular complementation remains to be accomplished, but was beyond the scope of the present study.

PM405
Of the ten candidate genes identified by our filtration pipeline in PM405, only one had a functional annotation that suggested a role in root nodulation. The respective mutation site on chromosome Ca2 consisted of a single bp deletion that resulted in a frame shift in a gene identified as Ca_06500 in the ICC 4958 v3.0 assembly. To test the hypothesis that this mutation coincided with the rn4 mutation in mutant PM405, we generated a segregating population by crossing PM405 with ICC640 for the purpose of assessing marker-trait linkage association.
Based upon the desi reference sequence [36] corresponding to the mutation site region, we designed allelespecific primers with a WT-Rev primer intended to only amplify the wild type allele, and a Mut-Rev primer intended to only amplify the mutant allele. Based upon this design, a heterozygous individual was expected to produce bands of differing sizes, depending on which of the reverse primers was used. Genotyping of the segregating progeny by allele-specific PCR provided evidence supporting the hypothesis that the observed Ca_06500 mutation is responsible for the nodulation mutant phenotype of PM405. We also genotyped the population for a mutation in a nearby gene, Ca_06416, annotated as Heat Shock Protein 90-4. There were three recombination events between Ca_06500 and Ca_06416 but only one event between Ca_06416 and the rn4 locus. The Rn4 allele is dominant so we could only detect recombination events between the candidate gene loci and the rn4 locus when homozygous recessive (rn4/rn4). The recombination event between Ca_06416 and the Rn4 locus suggests this mutation was not responsible for the non-nodulating phenotype.
Ca_06500 is predicted as an NLP1 gene through NCBI automatic annotation. However, a BLAST alignment of the CDS against the M. truncatula database showed a highest amount of sequence identity with NIN. A reciprocal BLAST of NIN showed Ca_06500 as the alignment with the highest sequence identity, suggesting Ca_06500 may be an ortholog of NIN, not of NLP1.
NIN expression is induced in M. truncatula following Nod factor recognition. The NIN protein contains an RWP-RK-type DNA-binding domain and a Phox and Bem1 (PB1) domain necessary for protein-protein interactions. The predicted protein encoded by Ca_06500 includes the RWP-RK and PB1 domains. The single bp deletion in PM405 occurs between the two domains, and results in loss of PB1 due to a frameshift. Plants with mutations in NIN in M. truncatula do not form nodules [38].
The mutant segregants in the PM405 segregating population had no nodules, instead of ineffective nodules as previously reported in PM405 [25]. The mutant phenotype is apparently conditional in PM405, and under prior experimental conditions permitted formation of very small, ineffective nodules while under our current experimental conditions the plants were non-nodulating. The nodulation phenotypes of two of the other mutants, PM665 and PM679, were previously shown to also be conditional: both were temperature dependent [24].

Assembly issues
The ICC 4958 v3.0 desi reference assembly had multiple stretches of N's and regions of low read coverage in the region containing the PM233 deletion, making it difficult to determine the precise breakpoints using the aligned Illumina reads alone. We designed primers in an attempt to amplify and sequence the two ends of the deletion region, but were unable to amplify the downstream end of the PM233 deletion in ICC 640 using primers designed from the ICC 4958 v3.0 desi assembly. To resolve the deletion breakpoints, we generated our own Nanopore sequencing data from ICC640 and PM233 DNA and performed a de novo assembly of the ICC640 reads. Our assembled contigs agreed with the upstream end of the deleted region of the ICC 4958 v3.0 genome assembly, but diverged from the downstream end. The PM233 Nanopore reads suggest our assembled contig is correct because the same reads align to both the upstream and downstream ends of the deleted region in our assembled wild type contig, while in the ICC 4958 v3.0 genome reference assembly they align only to the upstream end. Our assembled wild type contig also aligns with the M. truncatula version Mt4.0 assembly chromosome 5 from ~ 32.25 Mb to ~ 32.95 Mb.
Our contigs assembled from Nanopore reads disagree at the region of the PM233 deletion with both the ICC 4958 v3.0 genome and kabuli reference assemblies. The fact that our c636 contig includes three individual Nanopore reads spanning the entire deletion and also that our contig assembly shows collinearity with the M. truncatula version Mt4.0 assembly suggests that the ICC 4958 v3.0 genome and kabuli reference assemblies are misassembled in this region, perhaps due to the abundance of repetitive sequence in the region. This putative misassembly could explain why the PM233 deletion boundaries could not be resolved by previous investigators [34].

Conclusions
In summary, this study characterized two mutations resulting in nodulation deficient chickpea, confirmed the previously proposed identity of the Rn1 gene by defining the nature and boundaries of the rn1 mutation in PM233, and identified a candidate gene potentially responsible for the rn4 mutation in PM405. Further molecular studies on the identified gene in PM405 could provide more information about the function of this gene. We have also generated Illumina sequencing data from three other nodulation deficient mutant chickpea lines: PM638, PM665, and PM796. Our Nanopore reads and contigs could be used to help identify and characterize mutations in these mutant lines, and could also provide a resource for a next generation desi reference genome assembly.

Plant culture and phenotyping
The original wild type chickpea line ICC 640 was part of a large set of chickpea germplasm obtained by the University of California, Davis from ICRISAT, India in 1981. The nodulation mutant lines PM233, PM405, PM638, PM665, PM679, and PM796 were generated by induced mutagenesis from ICC 640 in 1982 by one of us (T.M. Davis), and have been maintained by T.M. Davis throughout the intervening time period. This information is documented in Davis et al. 1985 [23]. Chickpea plants were grown at the University of New Hampshire Macfarlane Research Greenhouse. Plants grown for leaf tissue and seeds were grown in 8″ diameter plastic pots of ProMix ® potting soil on greenhouse benches in natural light. Plants grown for nodulation phenotyping were grown in Leonard jars [39] (see Additional file 9: Fig. S6) in a growth chamber at 25 °C with a photoperiod of 16/8 h and relative humidity of 70%. Leonard jar upper compartments were filled with Whittemore Super Coarse Horticultural Graded Vermiculite, which is N-deficient. The vermiculite was covered with a top layer of perlite to prevent contamination. Assembled Leonard Jars were wrapped in foil and autoclaved prior to use.
Rhizobium ciceri TAL 620 (USDA10050) cultures were grown on LMB medium (2 mM Kh 2 PO 4 + 2 mM Na 2 HPO 4 , 0.8 mM MgSO 4 7H 2 O, 4.5 mM CaCl 2 ) and used to inoculate the upper chamber of the Leonard Jar at the time of seed planting. N-deficient nutrient solution (2 mM MgSO 4 , 1 mM K 2 SO 4 , 1 mM K 2 HPO 4 , 2.5 mM CaSO 4 , 0.98 g/L Murashige and Skoog Basal Salts) was supplied to the plants via addition to the lower chamber of each Leonard jar. Plants were uprooted 28 days after planting and phenotyped based on presence or absence of nodules.

DNA isolation and sequencing
Young leaf tissue was collected from chickpea plants, and DNA was extracted using a modified (see Additional file 10: Supplemental Methods) CTAB extraction method [40]. Isolated genomic DNA was quantified using an Invitrogen Qubit Fluorometer and standardized to 10 ng/μL through dilution with sterile deionized water. For use in Nanopore sequencing, the genomic DNA isolation procedure was performed more gently than for Illumina sequencing, as indicated (see Additional file 3: Supplemental Methods).
Genomic DNA samples from nodulation mutants PM233, PM405, PM638, PM665, and PM796, and from wild type parent ICC 640 were subjected to Illumina sequencing by the Hubbard Center for Genome Studies (HCGS) at the University of New Hampshire (UNH). Illumina libraries for each genotype were prepared using the KAPA HyperPlus Kit, and sequencing was done on a HiSeq 2500 instrument to generate sets of 150 × 150 bp paired end reads. In addition, ICC 640 and PM233 were subjected to Nanopore sequencing on an Oxford Nanopore GridION instrument at the HCGS. Base calling was done by the HCGS using Guppy.

DNA sequence analysis -Illumina
Illumina reads from each of the five nodulation mutants and wild type ICC 640 were trimmed with Cutadapt [41] and aligned independently to the Cicer arietinum desi variety ICC 4958 v3.0 reference genome (NCBI Project Number PRJNA78951) [36,42] using bwa-mem [43]. Picard Tools [44] was used to clean, sort, and remove duplicate reads from the alignment files.
For the purpose of candidate gene identification, GATK HaplotypeCaller was used to call variants separately in chickpea mutant lines PM233, PM405, PM638, PM665, and PM796, and wild type parent ICC 640. In the variant filtering pipeline (Fig. 3, Additional file 11), variants were retained if they were located within CDS using VCFTools [45], and were unique to the mutant using an in-house Python script (see Additional file 6). Using Microsoft Excel, only variant sites with two alleles, read depth greater than two, and an alternate (non-reference) allele frequency of greater than 0.9 were retained. In a complementary approach, alignments were scanned for occurrence (if any) of large deletions unique to one mutant based on read depth comparisons performed using SAMtools.
The sequential ordering of the filtration steps in the variant identification pipeline is depicted.
The resulting chickpea candidate gene sequences were then aligned to the M. truncatula sequence database in NCBI using BLASTn [46]. Chickpea and M. truncatula sequences extracted from the respective databases were also aligned with each other using Lasergene MegAlignPro [47].

DNA sequence analysis and contig assembly -Nanopore
Nanopore sequencing was performed only for mutant PM233 and wild type ICC 640. Adapters were trimmed from Nanopore reads using Porechop [48]. Summary statistics and histograms were generated using NanoPlot [49]. Nanopore reads from ICC 640 and, separately, PM233 were assembled into contigs using the Canu genome assembler [50], followed by polishing with Pilon using the Illumina ICC 640 or PM233 reads, respectively [51]. Nanopore reads from PM233 were then aligned to the assembled ICC 640 contigs, and also to the desi ICC 4958 v3.0 reference assembly [36], using minimap2 [52]. SAM files were sorted and converted to BAM files using SAMtools [53]. Nanopore contigs and specific reads of interest were aligned to the desi ICC 4958 v3.0 reference [36] and kabuli reference [2] assemblies using Mauve [54] and BLASTn [55], and to the M. truncatula version Mt4.0 genome [56] downloaded from the M. truncatula Genome Database.

PCR-based characterization of mutations sites
Primer pairs targeting mutation sites of interest in PM233 and PM405 were initially designed on the basis of ICC 4958 v3.0 reference genome sequence [36] with the aid of Primer3 [57], and later on the basis of our own ICC 640 Nanopore contig sequences as needed. Each reaction consisted of ~ 10 ng template DNA, 2.5 L NEB 10X Standard Taq Reaction Buffer, 0.5 μL 10 mM dNTPs, 0.5 μL 10 mM forward primer, 0.5 μL 10 mM reverse primer, and 0.125 μL NEB Taq DNA polymerase 10. PCR was performed on an Eppendorf thermocycler. The program was 95 °C for 30 s; 30 cycles of denaturation at 95 °C for 30 s, annealing at 60 °C for 45 s, and elongation at 68 °C for 1 min/kb of expected product size, and a final elongation step at 68 °C for 5 min. Amplicons were visualized on 0.8% agarose (LE Quick Dissolve Agarose) gels.

Candidate gene-trait association analysis in PM405
Wild type ICC 640 was crossed as female with PM405 to generate hybrid seeds. A single F1 hybrid plant was allowed to self-pollinate to generate a segregating F2 population, which was subjected to phenotypic screening in Leonard Jars under conditions permitting symbiosis, and to PCR-based genotyping following genomic DNA isolation. Amplicons of mutant and wild type alleles were differentiated by employment of allele-specific primers, or by restriction digests that revealed polymorphisms. Where applicable, restriction digest reactions were set up using 1 μL of NEB AvaI restriction enzyme, 10X NEB CutSmart Buffer, and 1 μg of PCR product. Reactions were incubated at 37 °C for 1 h prior to electrophoresis.