Medicago truncatula DMI1 Required for Bacterial and Fungal Symbioses in Legumes

See allHide authors and affiliations

Science  27 Feb 2004:
Vol. 303, Issue 5662, pp. 1364-1367
DOI: 10.1126/science.1092986


Legumes form symbiotic associations with both mycorrhizal fungi and nitrogen-fixing soil bacteria called rhizobia. Several of the plant genes required for transduction of rhizobial signals, the Nod factors, are also necessary for mycorrhizal symbiosis. Here, we describe the cloning and characterization of one such gene from the legume Medicago truncatula. The DMI1 (does not make infections) gene encodes a novel protein with low global similarity to a ligand-gated cation channel domain of archaea. The protein is highly conserved in angiosperms and ancestral to land plants. We suggest that DMI1 represents an ancient plant-specific innovation, potentially enabling mycorrhizal associations.

Mycorrhizal and rhizobial associations represent the two most important symbiotic relations between higher plants and microorganisms, providing access to otherwise limiting supplies of phosphate and nitrogen, respectively. Although better characterized at a molecular-genetic level, nitrogen fixation is the more recent of the two associations, primarily restricted to members of the Legume family and their close relatives (1). Mycorrhizal associations are considered to be ancestral to the lineage of land plants (2).

The initiation of symbiosis between leguminous plants and nitrogen-fixing bacteria is tightly controlled, with specificity determined by a molecular dialogue between the two partners. Legume roots exude metabolites, principally flavonoids, that induce the synthesis of lipo-chito-oligosaccharidic signals (called “Nod factors”) in the bacterial symbiont (3). Nod factors, in turn, elicit plant responses that culminate in infection by rhizobia and the development of the root nodule. Extremely low concentrations (10–9 to 10–12 M) of purified Nod factors trigger many of the same responses in the roots of legume hosts that are induced by symbiotic rhizobia, including specific ion fluxes, gene expression, and cell division (4). Genetic and molecular analyses in the model legumes Medicago truncatula and Lotus japonicus have identified multiple genes that are required for nodule development and also for early Nod factor responses (512). Phenotypic analyses have placed these genes into a hierarchy. NFP and NFR1/NFR5 reside at the top of the cascade, as mutants of these genes lack all known responses to Nod factor. DMI1 and DMI2 genes are necessary for the induction of calcium spiking in root hairs and appear to function upstream of DMI3, NSP1, and NSP2 (13). DMI1, DMI2, and DMI3 mutants are also blocked for mycorrhizal symbiosis (Nod Myc mutants), which demonstrates that the signaling pathways of nodulation and mycorrhization share common components (5, 14). DMI2 encodes a receptor-like kinase homolog with leucine-rich repeats (8). The cloning of DMI1 reported here, and an accompanying manuscript by Lévy et al. (15) that reports the cloning of DMI3, further characterize the known conserved elements for rhizobial and mycorrhizal associations.

DMI1 is located at one extremity of M. truncatula chromosome 2 (16). Genetic marker 1N1R, which defined the telomere end of linkage group 2 and was invariantly linked to the DMI1 locus in an F2 population of 499 Nod individuals, was used to seed a chromosome walk toward DMI1. In parallel to physical mapping, we used knowledge of chromosomal synteny between M. truncatula and M. sativa (17) to systematically position new genetic markers between 1N1R and the telomere. Genetic mapping of 1N1R in M. sativa revealed a single genetic marker, U212D, that was on the telomeric side of 1N1R in both species (Fig. 1A). Using U212D as a probe, we identified additional bacterial artificial chromosome (BAC) clones of M. truncatula and completed a 1.2-Mb contig extending from genetic marker 88P13S to the adjacent telomere of chromosome 2 (Fig. 1).

Fig. 1.

Defining the DMI1 locus. (A) Initial genetic mapping of the DMI1 region indicated that the gene is located between the 88P13S marker and the telomere. We constructed a 1.2-Mb contig covering this region by means of BAC fingerprinting. From a new population consisting of 1500 individuals, we localized DMI1 between the 57N18S marker and the telomere. We sequenced and annotated a minimum tiling path of five BAC clones covering this region. (B) The Arabidopsis and M. truncatula homologs, but not their rice counterpart (not shown on figure), reside in a region of conserved genome microsynteny, presumably indicative of the ancestral chromosomal context.

Genotyping of >1500 individuals identified a single recombination event that delimited DMI1 to a 550 – kilobase pair (kbp) interval between genetic marker 57N18S and the telomere. The observation of suppressed recombination in the subtelomeric region of chromosome 2 prompted us to abandon further genetic analyses in favor of a candidate gene approach to identify the gene. We sequenced a minimum tiling path composed of five BAC clones (∼550 kb in length) using a BAC sublibrary shotgun strategy. FGENESH (18) predicted 85 genes in the sequenced region, roughly half of which were supported by unigene assignments in the M. truncatula EST data. Oligonucleotide primers were designed for candidate genes, and the polymerase chain reaction was used to amplify both genomic DNA and cDNA isolated from independent DMI1 alleles, including three ethylmethane sulfonate and two fast neutron bombardment mutants. A single unknown protein gene on BAC mth2-54A24 was altered in each of five independent alleles (Table 1), and the mutations were predicted to strongly affect the transcript or the conceptually translated protein structure. The ability of a wild-type DMI1 transgene to complement the Nod phenotype of DMI1 alleles C71 and Y6 (Fig. 2) confirmed the identity of this gene as DMI1.

Fig. 2.

Complementation of the DMI1 phenotype by using A. rhizogenes–based transformation (25). Both panels show mutant dmi1 roots complemented by the DMI1 wild-type transgene. The full-length wild-type DMI1 cDNA was amplified and cloned into the pCR2.1 vector (Invitrogen). Of the promoter sequence, 1.6 kb was also amplified from BAC mth2-54A24 and cloned upstream the cDNA. This construct was introduced into the pCAMBIA-1303 binary vector (clone pCAMBIA-DMI1). dmi1 and wild-type roots were transformed with either pCAMBIA-DMI1 or pCAMBIA-1303 vector alone and inoculated with Sinorhizobium meliloti, then nodulation was scored visually after 15 days. Constitutively expressed uidA was used as a reporter to identify transformed tissues based on GUS staining. Among more than 80 independent dmi1 transformants, only those transformed by the DMI1 transgene were nodulated by S. meliloti. Transformation of wild-type plants with either DMI1 or control constructs had no effect on nodulation. Scale bar, 1 mm.

Table 1.

Summary of M. truncatula dmi1 alleles. EMS, ethylmethane sulfonate; FNB, fast neutron bombardment.

Mutagenesis DMI1 mutant Nature of mutations
EMS B129 G1068A point mutation that results in a premature stop codon truncating the protein at amino acid position 356.
EMS C71 (domi) G1264A point mutation that occurs at the 5′ splice site of the third intron and that causes a missplicing event.
EMS Y6 C913T point mutation that creates a premature stop codon truncating the protein at amino acid position 305.
FNB GY15-3F-4 Large deletion of all the 5′ of the gene and the promoter.
FNB GY15-1B-5.1 Large deletion of all the 5′ of the gene and the promoter.

A full-length DMI1 cDNA cloned from M. truncatula root mRNA encodes a 2649-bp open reading frame with a deduced protein of 883 amino acids. Predicted features of the protein include four transmembrane domains, the first two of which encompass a putative leucine zipper (LZ) motif, and a proline-rich domain that could facilitate protein–protein interactions (19). Other than these general features, the DMI1 protein lacks even moderate similarity to functionally characterized proteins or protein subdomains in plants, although the deduced protein is conserved throughout most lineages of land plants that have been sampled by EST and genome sequencing projects.

DMI1 is constitutively expressed in roots, with lower levels of transcript detected in pods, flowers, leaves, and stems (Fig. 3A). Treatments known to activate expression of plant “nodulin genes,” including inoculation with the compatible symbiotic bacterium Sinorhizobium meliloti (Fig. 3B) or 10–9 M cognate Nod factors, had no effect on DMI1 transcript levels.

Fig. 3.

Expression profiles of DMI1 in M. truncatula and A. thaliana. (A) M. truncatula DMI1 and A. thaliana At5 g49960 cDNA analysis by semiquantitative RT-PCR in different tissues. (B) M. truncatula DMI1 expression is not affected by inoculation with wild-type S. meliloti as compared with the nodulation-associated control gene ENOD11 (25). Purified total RNA was normalized with the use of the Ribogreen RNA quantification kit (Molecular Probes) and analyzed using the PowerScript RT-PCR kit (Clontech Laboratories).

The signaling pathway leading to nodulation is highly conserved across the characterized legume species (812), and thus, diverse legume species are expected to contain orthologous signaling components. Southern blot experiments indicate that DMI1 is a gene with a low copy number in M. truncatula and that homologs are present in many other leguminous plants, including Medicago sativa, Melilotus alba, Vicia hirsuta, Lotus japonicus, Sesbania, Cassia, Trifolium, Desmodium, Vigna, Macroptilium, and Vigna radiata. BLASTN analyses identified similar expressed sequence tags (ESTs) in soybean and L. japonicus, which indicated that the sequence homologs in these species represent expressed genes. The sym8 mutations of pea (Pisum sativum), which are phenotypically similar to DMI1 mutants of Medicago truncatula, map to a syntenic location between the two genomes (16), consistent with the possible identity of sym8 as pea DMI1.

BLASTX and TBLASTN searches of the National Center for Biotechnology Information (NCBI) nonredundant and EST databases predicted similar proteins in 28 monocot and dicot species, as well as a more distant homolog in the nonvascular plant Physcomitrella patens. Related sequences were not identified in the fully sequenced genomes of animals, fungi, or green algae. Relatively close homologs were identified in only two eubacterial genera, namely the nitrogen-fixing symbiont Mesorhizobium loti and Streptomyces spp. Homologous genes were not evident in any other sequenced bacterial genome, including close relatives of M. loti such as Sinorhizobium meliloti and Agrobacterium tumefaciens. We used phylogenetic reconstruction based on parsimony analysis to infer the evolutionary history of this gene family. Analysis of several full-length deduced proteins from M. truncatula, Arabidopsis, rice, and bacterial species yielded three well-supported clades, corresponding to two paralogous groups of plant genes and a single group of bacterial genes (Fig. 4B).

Fig. 4.

Alignment and phylogenetic analysis of DMI1 protein homologs. (A) Partial alignment of DMI1 homologous proteins sequences by means of ClustalW with default parameters in the region sharing homologies with cation channels. This region contains the TrkA homologous region and also the third and fourth transmembrane domains of the DMI1 orthologs. A. thaliana At5 g49960 and rice BAB64102 DMI1 orthologs share about 80% similarity with the M. truncatula protein. A. thaliana At5 g02940 and rice AAN06856 proteins are 73% similar and share ∼37% similarity with the orthologous grouping of DMI1-related proteins in M. truncatula, Arabidopsis, and rice. Mesorhizobium loti NP 102608 and the Streptomyces proteins represent a third clade of proteins, with 60% similarity between bacterial homologs and ∼37% similarity to the DMI1 orthologs. Transmembrane domains (purple) were predicted using THMM, TMpred and TopPred 2 software programs (26). The presence of signal peptides (yellow) was assessed by using SignalP V2.0 software (27). The region homologous to cation channels was determined according to Anantharaman et al. (21) and annotated according to Jiang et al. (22). (B) Unrooted phylogenetic tree of DMI1 homologs. Full-length protein sequences were aligned using ClustalW with default parameters. The branch-and-bound parsimony method was used to infer phylogenetic relations and to identify two paralogous groups of plant genes and a single group of bacterial orthologs. Percentage bootstrap support is given to the side of each branch.

The complete sequence of DMI1 putative orthologous genes was available for the A. thaliana At5 g49960 and rice BAB64102 genes, corresponding to single-copy genes on chromosomes 5 and 1, respectively. The Arabidopsis and M. truncatula loci, but not their rice counterpart, reside in a region of conserved genome microsynteny, presumably indicative of the ancestral chromosomal context (Fig. 1B). The A. thaliana and rice predicted proteins share 80% identity with DMI1, with the highest similarity found in the C-terminal region. Although ESTs corresponding to the A. thaliana gene were absent from the public databases, semiquantitative reverse transcription polymerase chain reaction (RT-PCR) analysis indicates that the Arabidopsis gene is expressed in roots but not in leaves (Fig. 3A).

Comparison of the deduced partial protein Physcomitrella patens with either group of angiosperm paralogs indicates that all three groups are roughly equally diverged from one another (i.e., 26 to 28% identity and 53 to 56% similarity), consistent with an origin in the nonvascular plants. Taken together with the absence of homologs in the fungal, animal, and green algal lineages, these results suggest that the DMI1 protein represents a plant-specific innovation that potentially arose near the base of the land plant lineage. The fact that the M. loti and Streptomyces sequences also branch from the base of the DMI1 lineage, near the inferred plant duplication, is suggestive of horizontal transfer from an ancient plant genome or genomes to a limited number of bacterial species.

DMI1 is predicted to act in genetic proximity to two additional M. truncatula genes, DMI2 and DMI3, which together make up the set of known genes common to both mycorrhizal and rhizobial symbioses (5). All three DMI genes are implicated in transduction of the Nod factor signal; DMI1 and DMI2 are required for Nod factor–induced calcium spiking, which by analogy to animal systems is predicted to have a causal role in the Nod factor signal transduction cascade. Despite the fact that DMI1 homologs have been identified in the genomes of many plant species, DMI1 represents the first member of this gene family with an assigned phenotype and verified gene structure. As with the NORK receptor kinase (DMI2) (8), DMI1 is predicted to be a membrane-spanning protein and may participate in the formation of a receptor-complex for symbiotic signals. The fact that proline-rich and leucine zipper domains often participate in protein-protein interactions supports the idea of such a protein complex, but these motifs are present in the N-terminus of DMI1, a region of the protein that is poorly conserved among the various homologs. The DUF1012 protein family comprises the homologs of Arabidopsis, rice, Mesorhizobium, and Streptomyces proteins in the Pfam database (20), without functional assignment. The DMI1 DUF1012 domain overlaps substantially with a region of low, but global, similarity to the NAD-binding TrkA domain of certain bacterial potassium channels (21). In fact, the central portion of DMI1, from the third predicted transmembrane domain through the entire TrkA homology region, is predicted to share distant homology with multimeric, ligand-gated potassium channels of archaea (22) (Fig. 4A). DMI1 is conserved throughout both the pore and ring domains that are thought to constitute the functional cation channel (22) but is degenerate in the filter region, located between transmembrane domains 3 and 4, where structural modifications have been correlated with altered cation specificity (23). Whether DMI1 and its close phylogenetic relatives in plants and bacteria might also function as ligand-gated cation channels remains uncertain, but the possibility is particularly intriguing given the genetic requirement for DMI1 function in Nod factor–induced calcium oscillations within Medicago root hair cells.

Identifying DMI1-interacting proteins should expand our understanding of the role of the DMI1 protein in symbiotic interactions. Candidate proteins to interact with DMI1 include LYK3 and 4 and NFR1 and 5 (10, 11, 12, 24), NFP (6), DMI2 (8, 9), and DMI3 (5).

About 80% of terrestrial plants establish mycorrhizal symbioses. These beneficial plant-fungal associations act to expand the effective root-soil interface; in addition to their importance in natural and agricultural ecosystems, they may have facilitated the colonization of land by plants by increasing access to vital soil nutrients in the primitive rootlike structures of nonvascular plants (2). The biological and molecular function of this group of proteins is likely to be conserved because of the high degree of conservation for DMI1 orthologs among angiosperms (>80% similarity among rice, Arabidopsis and Medicago proteins) and because only a single gene with high similarity to DMI1 is present in the fully sequenced genomes of rice and Arabidopsis. Other genes identified for Nod factor signaling (i.e., LYK homologs, DMI2) are less well conserved among angiosperms, and some may be members of large, fast evolving gene families. It is widely anticipated that nodulation in legumes arose from an ancient, conserved pathway for mycorrhizal associations, and we suggest that DMI1 is a pivotal component of this pathway.

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article