Structure-based phylogenetic analysis reveals multiple events of convergent evolution of cysteine-rich antimicrobial peptides in legume-rhizobium symbiosis

Nitrogen is essential for plant growth, yet its availability often limits agricultural productivity. Some legumes have evolved a unique ability to form symbiotic relationships with nitrogen-fixing soil bacteria called rhizobia, enabling them to thrive in nitrogen-deficient soils. In five legume clades, an exploitive strategy has evolved in which rhizobia undergo Terminal Bacteroid Differentiation (TBD), where the bacteria become larger, polyploid, and have a permeabilized membrane. Terminally differentiated bacteria are associated with higher N2-fixation and, thus, a higher return on investment to the plant. In several members of the IRLC (Inverted Repeat-Lacking Clade) and the Dalbergioid clades of legumes, this differentiation process is triggered by a set of apparently unrelated plant antimicrobial peptides with membrane-damaging activity, known as Nodule-specific Cysteine-Rich (NCR) peptides. However, whether NCR peptides are also implicated in symbiotic TBD in other legume clades and whether they are evolutionarily related remains unknown. Here, to address the molecular identity of NCR peptides and their evolution in different legume clades, we performed inter- and intra-clade comparisons of NCR peptides in representative species of four TBD-inducing legume clades. First, we collected genomic and proteomic data of species for which NCR peptides are known (1523 NCR peptides). We then used sequence similarity-based clustering to regroup the NCR peptides, resulting in over 400 different NCR clusters, each clade-specific. We obtained Hidden Markov Models for each cluster and used them to predict NCR peptides in 21 legume genomes (6 clades), including newly generated deep-sequenced root and nodule RNA-seq data of Indigofera argentea (Indigoferoid clade) and newly assembled high-quality transcriptomes of Lupinus luteus and Lupinus mariae-josephae (Genistoid clade), using tailored gene prediction pipeline and transcriptome matching. This resulted in 3710 NCR peptides in species that induce TBD. To date, the rapid diversification of NCR peptides that reduces the sequence similarities has masked the origin of NCR peptide evolution. We obtained high-confidence structural models for one sequence of each cluster. We performed structure-based clustering and phylogenetics, which resulted in 23 superclusters (14 inter-clade and nine clade-specific) that we represent in a structural distance-based tree. Our study revealed that the evolution of NCR peptides is a mix of divergent and convergent processes within each clade. We further chose nine independently evolved NCR peptides to test in vitro whether they are functional analogs in symbiosis.