Species-specific microsymbiont discrimination mediated by a Medicago receptor kinase

Abstract

Host range specificity is a prominent feature of the legume-rhizobial symbiosis. Sinorhizobium meliloti and Sinorhizobium medicae are two closely related species that engage in root nodule symbiosis with legume plants of the Medicago genus, but certain Medicago species exhibit selectivity in their interactions with the two rhizobial species. We have identified a Medicago receptor–like kinase, which can discriminate between the two bacterial species, acting as a genetic barrier against infection by most S. medicae strains. Activation of this receptor-mediated nodulation restriction requires a bacterial gene that encodes a glycine-rich octapeptide repeat protein with distinct variants capable of distinguishing S. medicae from S. meliloti. This study sheds light on the coevolution of host plants and rhizobia, shaping symbiotic selectivity in their respective ecological niches.

A plant receptor–like kinase enacts partner specificity between Medicago and two closely related rhizobial species.

INTRODUCTION

Legumes establish a root nodule symbiosis with nitrogen-fixing soil bacteria to fulfill their nitrogen needs (1, 2). These symbiotic interactions are characterized by a high degree of specificity, occurring both within and between host and bacterial species (3, 4). As a result, nodulation capacity and nitrogen-fixing efficiency vary enormously among different plant-bacteria pairs (5–10). At the species level, nodulation specificity is primarily determined by the evolutionary divergence of host receptors, enabling the recognition of distinct bacterial nodulation (Nod) factors (11–16). Although activation of Nod-factor signaling is typically essential and sufficient for bacterial infection and nodule organogenesis (12, 13), incompatible bacterial signals, such as secreted effectors, microbe-associated molecular patterns, and strain-specific cell surface molecules, can trigger host immune responses, thereby blocking bacterial infection (17–20). Understanding the genetic basis of symbiotic specificity would enable the development of strategies to overcome the host range barrier, which is also a critical step toward extending root nodule symbiosis to nonleguminous plants (21).

Sinorhizobium meliloti and Sinorhizobium medicae are two closely related species with a shared host range (22). While both species are effective symbiotic partners of Medicago sativa (alfalfa) and the model legume Medicago truncatula, certain Medicago species exhibit selectivity in their interactions with these Sinorhizobium species (22, 23). For instance, S. medicae is mostly associated with host plants such as Medicago polymorpha and Medicago murex that thrive in acid soils, whereas S. meliloti predominantly forms symbiotic relationships with Medicago littoralis and Medicago tornata, which naturally flourish in alkaline or neutral pH soils (23). These observations suggest that S. meliloti and S. medicae have finely tuned their adaptations to different Medicago species, aligning with the ecological niches these plants commonly occupy in their native environments. However, the genetic basis of these specialized interactions has not been explored. From a diverse collection of Medicago accessions, we identified the accession DZA220-H that inhibits nodulation by most tested S. medicae strains. DZA220-H was categorized as M. truncatula, but an analysis of its plastid genome suggests that it likely belongs to M. littoralis, which can interbreed with M. truncatula (24). Genetic analysis of both the host and bacteria allowed us to identify the genetic mechanism underlying this recognition specificity.

RESULTS

Identification of the Medicago accession DZA220-H incompatible with S. medicae ABS7

We screened a collection of 36 Medicago accessions that represent a wide range of genetic diversity and geographical origins by inoculation with the S. medicae strain ABS7 (ABS7) (table S1). This experiment revealed that ABS7 was unable to nodulate DZA220-H but able to nodulate all other accessions such as DZA045 (Fig. 1, A and B). While ABS7 failed to induce the development of mature nodules on DZA220-H, it did trigger root hair curling and occasionally the formation of nodule primordia (Fig. 1, B and D), two hallmark traits suggestive of successful Nod factor perception and signaling (1). To support this observation, we further characterized the expression of ENOD11, an early nodulin expressed during preinfection and infection stages of nodulation in root and nodule tissues (25). We found that ENOD11 was expressed during both compatible (S. meliloti Rm41) and incompatible (ABS7) interactions, but not in uninoculated roots or roots inoculated with a nodC mutant of Rm41 defective in Nod factor production (fig. S1). ABS7 frequently colonized the curled root hairs of DZA220-H, but in contrast to the compatible interaction with DZA045, fully developed infection threads were not observed (Fig. 1, C and D). As such, the nodule primordia on DZA220-H roots were not infected (Fig. 1, E and F), and because of the absence of infection, cortical cell division was arrested at an early stage, resulting in barely visible nodule bumps (Fig. 1B).

Fig. 1. Nodulation phenotypes of the Medicago roots inoculated with GFP-tagged S. medicae ABS7.

(A) Mature nitrogen-fixing nodules on DZA045. (B) Small white nodule bumps on DZA220-H. (C) A representative infection thread observed on DZA045. (D) ABS7-induced root hair curling without infection thread formation on DZA220-H. (E) A representative nodule primordium on DZA045 containing infected bacteria. (F) A representative noninfected nodule primordium on DZA220-H, despite the presence of bacterial colonies on the epidermal surface of the nodule primordium. Scale bars, 10 μm (C and D) and 100 μm (E and F).

Identification of the NS2 locus in DZA220-H that blocks infection by S. medicae ABS7

For genetic analysis, we made a cross between DZA220-H and DZA045. Initial phenotyping of 257 F2 individuals inoculated with ABS7 revealed that the segregation of the nonnodulation (191) and nodulation (66) phenotypes statistically follows the 3:1 ratio (χ² = 0.064, df = 1, P = 0.80). This suggests that DZA220-H carries a single dominant locus conferring resistance to infection by ABS7, which we named NS2 (Nodulation Specificity 2). By genotyping this F2 population using the genome-wide single-nucleotide polymorphism (SNP) markers, we mapped the NS2 locus to a region on chromosome 8. In the reference genome of Jemalong A17, which forms nitrogen-fixing nodules by ABS7, this region spans approximately 375 kilobases (kb) and includes genes from Medtr8g014690 to Medtr8g015480 (M. truncatula genome version 4.0) (26, 27).

The NS2 locus maps to a complex gene cluster encoding receptor-like kinases

In the reference genome, the aforementioned chromosomal region harbors a cluster of 16 genes encoding leucine-rich repeat receptor-like kinases (LRR-RLKs) with a malectin-like domain (MLD) at the N terminus (Fig. 2A). These MLD-LRR-RLK proteins share similarities to NS1a (Medtr8g028110) and NS1b (Medtr8g028115) in the M. truncatula accession F83005, known to provide broad-spectrum resistance against S. meliloti strains (9). Through fine mapping with more than 3500 segregating individuals, we narrowed down the NS2 locus to an approximately 40-kb region inferred from the reference genome sequence. The region contains Medtr8g015120 and Medtr8g015150, both encoding MLD-LRR-RLK proteins (Fig. 2A).

Fig. 2. Map-based cloning and functional validation of the NS2 candidate genes.

(A) Diagrammatic representation of the NS2 locus. NS2 maps to a gene cluster encoding MLD-LRR-RLKs (gray arrows) that differ in copy number between different M. truncatula genotypes. Solid black arrows represent non–MLD-LRR-RLK genes. Allelic genes are connected by vertical lines. Arrows also indicate the transcriptional orientation of the genes. The candidate gene region is indicated by the gray box. (B and C) Complementation tests of the DZA220-H alleles of Medtr8g015120 (B) and Medtr8g015150 (C), showing that the transgenic hairy roots (blue) of DZA045 retained the ability to form nodules by ABS7. Transgenic roots were distinguished from those of wild-type (white) by GUS staining. (D and E) Functional validation of RLK-γ by CRISPR-Cas9–mediated gene knockout in the DZA220-H background. (D) A representative example showing that expression of sgRNA1.a (AGCGTTCAGTATGACAGTGA) knocked out RLK-γ, resulting in nodule formation on the mutant roots. (E) Top: Gene structure of RLK-γ and the CRISPR-Cas9 target site of sgRNA1.a. The exons and introns are indicated by boxes and lines, respectively. Arrow indicates the sgRNA targeting site on the third exon. Bottom: The nodulated blue roots shown in (D) were subjected to DNA sequencing, showing a single base pair (T) deletion at the target site. At least 40 transgenic hairy roots were obtained for each experiment, and all showed the same results.

To further explore candidate NS2 genes, we retrieved the contiguous NS2-locus sequences from existing genome assemblies of the Medicago HapMap accessions accessible via the Medicago Analysis Portal (https://medicago.legumeinfo.org). Annotation and comparative genomic analysis revealed a high level of polymorphisms for both copy number and gene sequences of the MLD-LRR-RLK genes across different haplotypes. In some accessions, such as HM010 and HM023, we identified two additional MLD-LRR-RLK genes, named RLK-α and RLK-β, located between Medtr8g015120 and Medtr8g015150 (Fig. 2A). At that time, owing to the absence of the DZA220-H genome sequence, we regarded all four genes, Medtr8g015120, RLK-α, RLK-β, and Medtr8g015150, as potential NS2 candidates. However, as described below, none of these four genes were validated as the bona fide NS2.

NS2 alleles are absent in most genotypes compatible with S. medicae ABS7

We performed PCR amplification of the four candidate genes from DZA220-H, utilizing sequence information from Jemalong A17, DZA045, HM010, and HM023. While we successfully amplified Medtr8g015120 and Medtr8g015150, RLK-α and RLK-β were not amplified, suggesting their absence in the DZA220-H genome. Genomic DNA sequences of Medtr8g015120 and Medtr8g015150, including their putative promoters and 3′ untranslated regions, were used to develop genomic constructs for complementation tests. However, hairy root transformation of DZA045 using DZA220-H alleles of Medtr8g015120 and Medtr8g015150 failed to validate either of the two genes as NS2, as the transformed roots retained the ability to form nodules with ABS7 (Fig. 2, B and C). We therefore inferred that while the NS2 gene is present in DZA220-H, its allelic copies are absent in DZA045 and the reference genomes.

Sequencing the DZA220-H genome revealed NS2 as a genotype-specific gene encoding a receptor-like kinase

We sequenced the DZA220-H genome using PacBio high-fidelity (HiFi) long-read sequencing technology. The de novo assembled DZA220-H genome totaled 440.3 Mb with an N50 contig size of 9.56 Mb. Assembly and annotation of the genomic region encompassing the NS2 locus revealed an additional MLD-LRR-RLK gene, designated RLK-γ, positioned between Medtr8g015120 and Medtr8g015150 (Fig. 2A). The genomic DNA of RLK-γ spans approximately 14 kb, comprising 13 exons, with the longest intron exceeding 6.7 kb (Fig. 2E). It encodes a protein of 878 amino acids, predicted to have an N-terminal signal peptide, an MLD, five LRRs, a single-pass transmembrane domain, and a serine/threonine kinase domain at the C terminus. While an RLK-γ allele is also present in the ABS7-compatible accession HM017 (also a putative M. littoralis accession), this gene isoform is presumably nonfunctional due to a 4-bp frameshift deletion in the second exon (fig. S2). We thus postulated that RLK-γ in DZA220-H likely represents the NS2 gene responsible for resistance to ABS7 infection.

Because of the inability to clone RLK-γ genomic DNA into a transformation vector for complementation testing, we validated this gene through CRISPR-Cas9–based knockout in the DZA220-H genetic background. We designed two gene-specific single guide RNAs, designated NS2-sgRNA.a and NS2-sgRNA.b, targeting two distinct sites of RLK-γ (Fig. 2E and fig. S3). We then used these CRISPR-Cas9 constructs to perform hairy root transformation on DZA220-H and evaluated the nodulation phenotype of the transformed roots. Our experiments conclusively demonstrated that RLK-γ knockout enabled the mutant roots to form nitrogen-fixing nodules with ABS7 (Fig. 2D and fig. S3). Thus, we established RLK-γ as the NS2 gene essential for conferring resistance to ABS7 infection.

NS2 expresses only in rhizobium-inoculated roots and nodules

A single-pass transmembrane domain is predicted between the LRR and kinase domains, suggesting a topology with MLD and LRRs in the ectodomain and the kinase domain in the intracellular compartment. Supporting this prediction, transient expression of an eYFP-tagged NS2 protein in Nicotiana benthamiana leaves indicated its localization at the plasma membrane (Fig. 3A). Reverse transcription polymerase chain reaction (RT-PCR) analysis revealed the expression of NS2 in inoculated roots but not in leaves, flowers, and uninoculated roots. This induced expression occurred during both compatible (Rm41) and incompatible (ABS7) interactions (Fig. 3B). We further characterized the spatial expression of NS2 by assessing its promoter activity in the transformed roots of DZA220-H using the pNS2::GUS reporter. Consistent with the RT-PCR analysis, GUS activity was undetectable in the noninoculated transformed roots but strongly induced in the roots inoculated with either incompatible (ABS7) or compatible (Rm41) strains (Fig. 3, C and D). Notably, GUS activity was restricted to the susceptible zones or zones forming nodule primordia. In the mature nodules induced by the compatible strain Rm41, NS2 was mainly expressed in the meristematic and infection zones. These expression patterns suggest that the NS2 expression depends on host perception of Nod factors. Supporting this speculation, NS2 expression was not detected in the roots inoculated by a nodC mutant of Rm41 deficient in Nod-factor biosynthesis (Fig. 3B).

Fig. 3. Expression and subcellular localization of NS2.

(A) Localization of NS2 in the plasma membrane. NS2 tagged with eGFP at its C terminus was transiently expressed in N. benthamiana epidermal cells and analyzed by confocal laser scanning microscopy. As a control, FM4-64 was used to stain the plasma membrane. The reason for the presence of the green fluorescent puncta at the periphery of the membrane is unknown, possibly due to protein internalization. Scale bars, 20 μm. (B) RT-PCR analysis of NS2 expression in the roots/nodules of DZA220H inoculated by ABS7, Rm41, and a nodC mutant of Rm41 (Rm41ΔnodC), as well as in the leaf and flower organs. The expression of a Medicago Actin gene was used as a control. (C and D) Expression of pNS2::GUS in transgenic hairy roots of DZA220H inoculated by ABS7 (C) and Rm41 (D). GUS staining images of Agrobacterium rhizogenes–mediated transformed hairy roots were taken 14 days after inoculation. GUS activity was observed only in susceptible zones or nodules. Scale bars, 100 μm.

Identification of a bacterial gene required for activating NS2-mediated nodulation restriction

We conducted transposon-mediated random mutagenesis of ABS7 and isolated a single mutant strain that acquired the ability to nodulate DZA220-H. Through genome sequencing of this mutant strain, we identified four distinct transposon insertion sites, each resulting in the disruption of a specific gene (table S2). By generating the targeted deletion mutants for the individual genes, we confirmed one of the genes, named rns2, as being required for activation of NS2-mediated nodulation restriction (Fig. 4, A and B). Supporting this conclusion, expressing the rns2 allele of ABS7 under control of its native promoter in S. meliloti 1021 (Sm1021) resulted in the loss of its inherent capability to nodulate DZA220-H (Fig. 4, C and D).

Fig. 4. Functional and structural characterization of the bacterial rns2 gene.

(A) ABS7 induced a Nod− phenotype on DZA220-H. (B) An rns2 deletion mutant of ABS7 (ABS7Δrns2) gained the ability to nodulate DZA220-H. (C) S. meliloti 1021 induced nodulation on DZA220-H. (D) Sm1021 expressing the rns2 version of ABS7 failed to nodulate DZA220-H. Inset figures are zoomed-in views of nodules for better visualization. (E) Alignment of the Rns2 isoforms of ABS7 and Sm1021. The glycine-rich octapeptide repeats are underlined. (F) Sequence logos showing differential frequency distributions of the first amino acid residue between S. medicae and S. meliloti Rns2 isoforms. (G) Structural model of Rns2 from S. meliloti (left) and S. medicae (right). The predicted transmembrane region (red), a small N-terminal β sheet (orange), and the C-terminal β-repeat domain (blue) are indicated. The 18 repeats form a tightly packed β-repeat domain composed of nine complete turns. The β-repeat domains generally share the same overall arrangement with a root mean square deviation (RMSD) of 2.1 Å, but there is a notable difference in the size of each repeat generated form the variation in octapeptide and/or hexapeptide.

The rns2 of ABS7 encodes a protein of 220 amino acids, characterized by a predicted transmembrane domain located at its N terminus and 18 consecutive tandem repeats of a glycine-rich octapeptide [(T/S/G)GSGGGQD]. In contrast, Sm1021 encodes an Rns2 isoform of 250 amino acids (SMc04236), featuring 18 intact octapeptide repeats [(T/S/E)GSGGGQD], interspersed with 5 hexapeptide repeats [(S/E)GSGGG] that lack the QD sequence present within the octapeptide repeats (Fig. 4, E to G). Multiple sequence alignments of the Rns2 isoforms of S. meliloti and S. medicae from the GenBank revealed that, aside from other fixed amino acid differences, the presence or absence of interspersed hexapeptide repeats serves as a distinctive marker to differentiate between the two species, despite variations in the count of octapeptide and/or hexapeptide repeats across strains.

NS2 blocks infection by a wide spectrum of S. medicae strains

Given the dichotomy of Rns2 isoforms between S. medicae and S. meliloti, we hypothesized that NS2 might be able to discriminate between the two bacterial species. To test this hypothesis, we inoculated DZA220-H with 30 strains of S. meliloti and 24 strains of S. medicae (table S3). The experiment revealed that DZA220-H inhibited nodulation by 22 of 24 S. medicae strains, while allowing nodule formation by the majority of S. meliloti strains (28 of 30) (table S3). We further conducted genetic tests of the nodulation phenotype on five selected S. medicae strains (WSM419, KH53B, USDA1004, USDA1606, and USDA1631), all of which are incompatible with DZA220-H. By using CRISPR-Cas9–mediated knockout of the NS2 gene in DZA220-H, we demonstrated that the mutant roots were able to form nodules with these five strains (Fig. 5, A to C, and fig. S4). Moreover, targeted deletion of rns2 in WSM419 (WSM419Δrns2) enabled the mutant strain to nodulate DZA220-H (Fig. 5D). We thus conclude that NS2 is responsible for symbiotic incompatibility with most, if not all, of the S. medicae strains.

Fig. 5. Genetic characterization of the incompatible interaction between DZA220-H and S. medicae WSM419.

(A) DZA220-H exhibited a Nod− phenotype when inoculated by WSM419. (B) A representative example showing that expression of sgRNA1.b (GACCTTTCATACCTATGTAC) knocked out NS2, leading to nodule formation on the mutant roots inoculated by WSM419. Arrowheads indicate nodules on the roots. (C) The nodulated blue roots shown in (B) were subjected to DNA sequencing, showing a 14-bp deletion at the target site. (D) An rns2 deletion mutant of WSM419 (WSM419Δrns2) gained the ability to nodulate DZA220-H. The inset figure is a zoomed-in view of nodules for better visualization.

Evolution of the NS2 locus

The MLD-LRR-RLK gene cluster at the NS2 locus comprises approximately half of this gene family in the M. truncatula genome (fig. S5). This type of gene clusters tends to evolve rapidly among distantly related species, making it challenging to infer orthologous relationships for individual family members. Nevertheless, the genes surrounding the NS2 locus, such as Medtr8g014670, Medtr8g014900, Medtr8g014910 and Medtr8g015220, are highly syntenic across the legume family, providing insights into the evolutionary history of this gene cluster. Comparative genomic and phylogenetic analyses of these syntenic genes revealed an ancient duplication event that predates legume speciation (fig. S6). MLD-LRR-RLK genes are present in the syntenic regions of Medicago, soybean (Glycine max), common bean (Phaseolus vulgaris), cowpea (Vigna unguiculata), and lentil (Lens culinaris), with copy numbers ranging from 1 to 19, but not in chickpea (Cicer arietinum). These genes are typically found in only one of the duplicated regions, except in soybean, which underwent an additional genome duplication event after legume speciation.

Despite their highly degenerate nature, synteny can also be observed between the genes flanking the NS2 locus in Medicago and those bordering a cluster of 12 MLD-LRR-RLK genes on chromosome 1 of Arabidopsis (fig. S7). These findings suggest a common ancestor of these gene clusters, which subsequently underwent independent gene loss or expansion within and between species. Notably, several Arabidopsis MLD-LRR-RLK genes in the cluster displayed differential expression patterns in response to pathogen or elicitor treatments, implying their roles in plant-microbe interactions (28). Such genes include IOS1 (At1g51800), which contributes to disease susceptibility caused by filamentous (hemi)biotrophs but confers resistance to bacterial pathogens (29), and SIF2 (At1g51850), which regulates stomatal immunity against bacterial pathogens (30). Meanwhile, some of the Medicago homologs, such as NS1a, NS1b, and NS2, are exclusively expressed in rhizobium-inoculated roots or nodules, aligning with their roles in regulating the partner specificity of root nodule symbiosis (9).

DISCUSSION

Plants continuously encounter pathogens and symbionts in their growing environments. Because of the divergent outcomes of these interactions, plants have evolved unique mechanisms to accommodate beneficial symbionts while defending against harmful pathogens. However, the presence of shared features between pathogenic and symbiotic microorganisms poses a challenge for the host to distinguish between them. Consequently, mutualistic microbes may be erroneously perceived as pathogens, thereby triggering immune responses to prevent infection. These mechanisms can also drive the diversification of both host and microsymbionts, facilitating specialized interactions that adapt to their respective environmental niches.

We previously identified the Medicago NS1 locus, which prevents infection by a wide range of S. meliloti strains (9). This nodulation blockade is regulated by two tandem genes that encode MLD-LRR-RLKs. The activation of NS1-mediated nodulation restriction is contingent upon the presence of a functional bacterial rns1 gene, which codes for a type I–secreted protein with homology to polysaccharide lyases (31). While rns1 is present in approximately half of the surveyed S. meliloti strains, it is notably absent in S. medicae. The occurrence of the NS1 haplotype is rare among M. truncatula accessions, suggesting a selection pressure against the spread of this resistance locus in natural populations.

In contrast to the NS1 locus, the NS2 locus inhibits infection by most strains of S. medicae. NS2-mediated resistance to infection depends on S. medicae-specific variants of the rns2 gene, which encodes a putative glycine-rich cell wall structural transmembrane protein. Given that both NS1 and NS2 genes encode homologous receptor kinases and exhibit similar expression patterns, one can speculate that their recognition mechanisms might be similar. However, the molecular and biochemical functions of Rns1 and Rns2 remain to be elucidated. It has been reported that rns2 of S. meliloti (SMc04236) is potentially a direct target of the ExoS/ChvI two-component signaling pathway (32). The ExoS/ChvI system plays a critical role in establishing a root nodule symbiosis with plant hosts and also regulates many free-living bacterial phenotypes such as exopolysaccharide production, motility, and cell envelope integrity.

We attempted but were unsuccessful in establishing physical interactions between NS1a/NS1b and Rns1, and between NS2 and Rns2. We thus hypothesize that Rns1 and Rns2 likely are not the direct ligands perceived by the receptor kinases but are instead the metabolic products involving Rns1 and Rns2. Further research is needed to elucidate the genetic pathways on the plant’s side and the molecular signals on the bacterial side to better understand how the receptor-like kinases perceive and transmit bacterial signals and how the signaling pathway interacts with nodulation and/or immunity signaling, ultimately leading to the specialization of host-bacterial symbiosis.

DZA220-H, which carries the NS2 gene, likely belongs to M. littoralis or is an introgressant derived from it, based on the presence of a 45-kb inversion (rpl20-ycf1) in its plastid genome (24). In this context, it is likely that the NS2 gene originated from M. littoralis, a species with a stronger association with S. meliloti rather than S. medicae (23). Our study highlights a genetic mechanism that legumes use to distinguish between different rhizobial species for symbiosis development.

MATERIALS AND METHODS

Plant material and nodulation assay

The Medicago accessions used in this study are listed in table S1. The F2 mapping populations were derived from a cross between DZA045.5 (DZA045) and DZA220-H. Seedlings were grown in a substrate mixture comprising a 1:1 ratio of vermiculite (PVP Industries Inc.) and turface (Turface Athletics) in a growth room under conditions of 16 hours light and 8 hours dark at 22°C. For the nodulation assay, the roots of 1-week-old seedlings were flood-inoculated with rhizobial bacteria. Each plant received approximately 1.0 ml of cell suspension with an optical density of OD₆₀₀ = 0.1. We documented nodulation phenotypes 3 weeks after inoculation.

Rhizobial strains and growth conditions

The S. medicae and S. meliloti strains used in this study are listed in table S3. The ABS7 strain (33) used for fluorescence microscopy contained the plasmid pHC60 carrying a constitutively expressed green fluorescent protein (GFP). The strains were cultured on TY agar medium at 28°C, supplemented with the appropriate antibiotics.

Genetic mapping

Fine mapping was based on SNPs identified around the NS2 locus between the two parental genotypes, DZA045 and DZA220-H. These SNPs were genotyped by conversion into cleaved amplified polymorphic sequences markers or by direct sequencing.

Complementation tests and CRISPR-Cas9–mediated mutagenesis

The genomic fragments of Medtr8g015120 and Medtr8g015150, which include both the promoter and 3′-untranslated regions, were amplified from DZA220-H and subsequently cloned into the binary vector pCAMBIA1305.1 using the In-Fusion Advantage PCR Cloning Kits (Clontech). Notably, the pCAMBIA1305.1 vector carries a GUS reporter gene, allowing for differentiation of transgenic and nontransgenic roots through GUS staining. The CRISPR-Cas9 gene knockout constructs were created using the pKSE401 vector (34). Two pairs of oligos were designed to target two distinct sites of RLK-γ (NS2). The oligo pairs were first annealed to produce double-stranded fragments with 4-nt 5′ overhangs at both ends, and then ligated into the Bsa I–digested pKSE401 vector. To facilitate phenotypic analysis, we amplified the GUS gene expression cassette from pCAMBIA1305.1 and cloned it into the pKSE401 vector.

Hairy root transformation and characterization of transgenic roots

Hairy root transformations were conducted following established protocols using the Agrobacterium rhizogenes strain ARqua1 (35). The resulting composite transgenic plants were transferred to plastic pots filled with a sterilized mixture of vermiculite and turface in a 1:1 ratio. These plants were allowed to grow for 1 week before inoculation. GUS staining was used to distinguish transformed roots from wild-type roots. For CRISPR-Cas9–based knockout experiments, transgenic roots underwent a series of steps, including DNA isolation, PCR amplification, and DNA sequencing to confirm the presence of target DNA mutations. In cases where the primary sequencing indicated the presence of numerous heterogeneous mutant alleles, the PCR products were ligated into the pGEM T-Easy Vector System (Promega), and 10 to 15 randomly selected colonies were subjected to sequencing.

Promoter activity analysis

The putative promoter region of NS2, spanning approximately 1.5 kb upstream of the start codon, was amplified using primers that included the attB sites. The fragment was first cloned into the entry vector pDONR/Zeo (Invitrogen) and subsequently subcloned into the destination vector pMDC163, which carries a GUS reporter gene using the Gateway cloning system (Invitrogen). The resulting plasmid was then introduced into the A. rhizogenes strain ARqua1 for hairy root transformation.

Subcellular localization

The full-length cDNA of NS2 was initially cloned into the entry vector pDONR/ZEO (Invitrogen) via the Gateway BP reaction (Invitrogen). Subsequently, this fragment was integrated into the pSITE_2NB vector, with an enhanced green fluorescent protein (eGFP) reporter fused at its C terminus under control of a double 35S promoter (36). The plasmids were introduced into the Agrobacterium tumefaciens strain LBA4404. For agroinfiltration in N. benthamiana, we followed the previously described procedure (9). Fluorescence imaging was performed using an Olympus Fluoview FV1000 confocal microscope. As a control, we used FM4-64 (Thermo Fisher Scientific) to stain the plasma membrane.

Random transposon mutagenesis of ABS7

We introduced the Mariner transposon delivery vector (pSAM_Rl) (37) into a chloramphenicol-resistant ABS7 derivative by bi-parental mating, following the same procedures as previously described (9). The mutant library was used to inoculate the roots of DZA220-H. The nodules induced by the mutants were surface sterilized with 70% (v/v) ethanol, washed in sterile distilled water, and then crushed in 200 μl of water. A dilution series of the suspensions were spread onto LB plates with kanamycin (200 μg/ml) and chloramphenicol (100 μg/ml). Genomic DNA samples were isolated from the colonies, digested with Xba I and Eco RI, and ligated into the pBluescript vector. After transformation of the ligated DNA into Escherichia coli MDS42 RecA blue cells (Scarab Genomics Inc., USA), colonies resistant to both ampicillin and kanamycin were isolated. Sequences flanking the transposon insertion sites were determined using the transposon-specific primers.

Construction of the targeted gene deletion mutants

A pair of PCR fragments of equal size (~1.0 kb) that flank the targeted genes, along with a spectinomycin resistance cassette to replace the deletion, were assembled in a pK18mobSacB vector derivative (38) equipped with Bsa I sites by Golden Gate Cloning (39). The constructs were then introduced into rhizobia by tri-parental mating. The bacteria with allelic exchanges by double homologous recombinations were selected on LB medium containing sucrose (5% w/v) and spectinomycin (100 μg/ml).

Construction of Rns2-expressing vectors

A DNA fragment of the rns2 allele of ABS7 was amplified from the genome of S. medicae strains ABS7. The PCR fragment was digested with the Bgl II and Hind III enzymes and cloned at the Bam HI and Hind III sites of a pCAMBIA3301 derivative in which the T-DNA region was replaced by a lacZα-encoding gene with the multiple cloning site from the pBlueScript vector. The clones were introduced into the S. meliloti and S. medicae strains with tri-parental mating using pRK600 (40) as the helper plasmid.

Phylogenetic analysis

Protein sequences were retrieved from Phytozome (https://phytozome-next.jgi.doe.gov/) and the Medicago Analysis Portal (https://medicago.legumeinfo.org). Syntenic gene sequences were concatenated into a super-gene alignment, which was then analyzed to generate the phylogenetic tree. Sequences were aligned by MUSCLE (41) and the unrooted phylogenetic trees were constructed by MEGA (42), using the maximum likelihood method with a bootstrap analysis of 1000 replicates.

Long-read genome sequencing and assembly

Genomic DNA was sheared to an average fragment size of 14 kb using a Megaruptor 3 instrument (Diagenode Inc., Denville, NJ, USA). The sheared DNA fragments were then converted into a library using the SMRTBell Express Template Prep kit 2.0. Sequencing was performed on a PacBio Sequel II platform (Pacific Biosciences, Menlo Park, CA, USA). HiFi reads were generated using the Circular Consensus Sequence (CCS) software with the following parameters: “min-passes = 3, min-rq = 0.99.” In total, 15.2 Gb of HiFi sequences were obtained, with an average length of 14.5 kb, providing a genome coverage of approximately 33-fold. The HiFi reads were assembled de novo into contigs using hifiasm v0.15.3 (43). Redundancies in the assembled contigs were removed using Purge Haplotigs (v1.1.2) with default parameters. Assembled contigs were also compared against the NCBI nonredundant nucleotide database to identify and remove possible contaminated sequences from organelle and microorganism genomes. Protein-coding genes were predicted from the NS2 locus using a combination of ab initio (Fgenesh) and homology-based (Fgenesh+) approaches (Softberry Inc.)

Structural analysis of Rns2

AlphaFold2 was used to generate the models of Rns2 from S. meliloti and S. medicae. S. meliloti Rns2 encodes a protein of 250 residues and the S. medicae Rns2 is composed of 220 residues. No template was used in the modeling process. Structural analysis and figures were made in PyMOL version 2.5.4 (Schrödinger LLC).

Supplementary Materials

This PDF file includes:

Figs. S1 to S7

Tables S1 to S3