Tissue-specific regulation of lipid polyester synthesis genes controlling oxygen permeation into Lotus japonicus nodules

Significance

Symbiotic nitrogen fixation reduces the dependency on costly and environmentally hazardous synthetic nitrogen fertilizers. The rhizobia nitrogenase that catalyzes the reduction of atmospheric nitrogen into ammonia is oxygen sensitive. Legumes protect the nitrogenase by creating a microaerophilic environment inside nodules. A long-standing model posits that oxygen diffusion into nodules is limited by a barrier in the nodule periphery. By exploring the natural diversity of Lotus japonicus accessions using comparative transcriptomics, we identified genes involved in the deposition of lipid polyesters on cell walls that are specifically expressed in the nodule periphery. Our findings demonstrate that deposition of lipid polyesters in specific cells restricts oxygen diffusion into nodules and is essential for efficient nitrogen fixation.

Abstract

Legumes establish endosymbiotic associations with nitrogen-fixing rhizobia, which they host inside root nodules. Here, specific physiological and morphological adaptations, such as the production of oxygen-binding leghemoglobin proteins and the formation of an oxygen diffusion barrier in the nodule periphery, are essential to protect the oxygen-labile bacterial nitrogenase enzyme. The molecular basis of the latter process remains elusive as the identification of required genes is limited by the epistatic effect of nodule organogenesis over nodule infection and rhizobia accommodation. We overcame this by exploring the phenotypic diversity of Lotus japonicus accessions that uncouple nodule organogenesis from nodule infection when inoculated with a subcompatible Rhizobium strain. Using comparative transcriptomics, we identified genes with functions associated with oxygen homeostasis and deposition of lipid polyesters on cell walls to be specifically up-regulated in infected compared to noninfected nodules. As hydrophobic modification of cell walls is pivotal for creating diffusion barriers like the root endodermis, we focused on two Fatty acyl-CoA Reductase genes that were specifically activated in the root and/or in the nodule endodermis. Mutant lines in a Fatty acyl-CoA Reductase gene expressed exclusively in the nodule endodermis had decreased deposition of polyesters on this cell layer and increased nodule permeability compared to wild-type plants. Oxygen concentrations were significantly increased in the inner cortex of mutant nodules, which correlated with reduced nitrogenase activity, and impaired shoot growth. These results provide the first genetic evidence for the formation of the nodule oxygen diffusion barrier, a key adaptation enabling nitrogen fixation in legume nodules.

The root nodule symbiosis between nitrogen-fixing rhizobia and legumes serves as the foundation to reduce inorganic nitrogen dependency in agriculture (1). This interaction is initiated by a bidirectional signal exchange between the symbionts, which on the plant side simultaneously elicits the development of specialized root organs called nodules and the formation of tubular infection structures that guide rhizobia into the developing nodule primordium (2). This culminates in the accommodation of differentiated bacteria, called bacteroids, inside living nodule cells (3).

Nodule cells undergo profound cell biological and metabolic changes to accommodate rhizobia. They go through endoreduplication (4), their vacuoles contract (5), and their tubulin cytoskeleton realigns (6), which coincides with a dramatic cell expansion (7). A single infected cell can host several thousand differentiated bacteroids (8). These are surrounded by a plant-derived membrane, which combined form the symbiosome, a specialized nitrogen-fixing organelle-like compartment (9). Across the symbiosome membrane, exchange of fixed nitrogen, reduced carbon, amino acids, and inorganic cations such as iron, copper, molybdenum, nickel, and cobalt is essential to fuel the activity of the nitrogen-fixing nitrogenase (10). As the nitrogenase is an oxygen-labile enzyme (11), legumes tightly control oxygen concentration in nodules by limiting its diffusion (12), relocating and clustering of mitochondria that consume oxygen via respiration (13), and by expressing symbiotic leghemoglobins that bind and deliver oxygen to satisfy the mitochondrial and bacteroid respiratory demand (14). This is crucial to support the energetic requirements of nitrogen fixation (15).

Despite the key importance of these adaptations for effective nitrogen fixation, many of the genes controlling for instance the diffusion of oxygen into nodules are still unknown. Forward and reverse genetics screens in the model organisms Lotus japonicus and Medicago truncatula and in crops such as soybean (Glycine max) and pea (Pisum sativum) have been instrumental to uncover the genetic architecture of this symbiosis (16). However, conventional genetic screens are of limited use when investigating multigene families that can exhibit genetic redundancy or dissecting complex intertwined traits. Exploration of the natural diversity using transcriptome analyses has become a powerful alternative to identify novel functions associated with complex traits (17).

In this study, we explored the natural diversity of L. japonicus to identify genes controlling adaptations to host endosymbiotic rhizobia. We hypothesized that these genes were specifically induced in infected compared to noninfected nodules. We exploited the diversity in nodule infection phenotypes elicited in Lotus by Rhizobium leguminosarum Norway, a strain that uses an infection thread-independent mechanism to colonize nodule cells (18). By applying comparative transcriptomics and coexpression analyses, we discovered that genes with functions associated with deposition of lipid polyesters in secondary cell walls were specifically up-regulated in infected nodules. To determine if these genes mediated the impregnation with hydrophobic substances of the nodule endodermis, which has been proposed to restrict the diffusion of oxygen into nodules (19), we inspected the activity of their promoters. Confocal microscopy of transgenic nodules expressing reporter constructs revealed that the promoters of these genes were specifically active in the nodule endodermis. Mutants in a Fatty acyl-CoA Reductase encoding gene had reduced deposition of lipid polyesters on the nodule endodermis, higher oxygen levels inside nodules, and impaired nitrogenase activity. Our work provides genetic evidence for the formation of an oxygen diffusion barrier in Lotus nodules, a key adaptation for endosymbiotic nitrogen fixation.

Results

L. japonicus Accessions Display Contrasting Nodule Infection Phenotypes.

To identify L. japonicus accessions in which nodule organogenesis and nodule infection were uncoupled, 40 accessions were phenotyped after inoculation with R. leguminosarum Norway. Most accessions developed well-defined nodules. However, some arrested development at the primordium stage, while others formed tumors with irregular or multilobular shape (SI Appendix, Table S1). R. leguminosarum Norway did not induce epidermal infection threads in any of the L. japonicus accessions. In contrast to other well-characterized interactions (e.g., L. japonicus and Mesorhizobium loti), the formation of nodules and their infection do not occur simultaneously upon R. leguminosarum Norway inoculation (18). Although R. leguminosarum Norway induced nodule organogenesis in all accessions, nodule infection ranged from highly infected to almost complete absence of bacteria (SI Appendix, Table S1). In addition, all nodules induced by R. leguminosarum Norway did not fix nitrogen, as described in other species (20).

To quantify nodule infection, nine accessions that harbored well-defined nodules and had qualitative differences in nodule infection were selected. Some accessions had more infected nodules than others (SI Appendix, Fig. S1 A and B). As the percentage of nodules infected does not necessarily reflect the quantity of bacteria inside nodule cells, nodules were sectioned and the percentage of area colonized by R. leguminosarum Norway-GFP was quantified. Two groups were selected: i) a group that included accessions with the highest percentage of infected nodules (MG-70, MG-79, and MG-136) and ii) a group comprising accessions that had few to no infected nodules (MG-9, MG-112, and MG-115). In addition, the area of L. japonicus Gifu nodules infected by M. loti MAFF303099-DsRed, a compatible symbiont (21), was also quantified as a control. Nodules infected with M. loti MAFF303099 showed colonization with a mean value of around 60%. The first group of accessions had nodules significantly more colonized by R. leguminosarum Norway-GFP (mean = 20%) than the second group (mean = 5%) (Fig. 1A and SI Appendix, Fig. S1C). Based on these results, accessions were designated as infected and noninfected, respectively. These results show that upon inoculation with R. leguminosarum Norway there are L. japonicus accessions that uncouple the organogenesis and infection programs.

Fig. 1.

Natural diversity and transcriptomic variation in nodules with contrasting infection phenotypes. (A) Representative images of nodule sections for infected (MG-79, MG-70, MG-136) and noninfected (MG-9, MG-112, MG-115) accessions. Plants were harvested 5 wk after inoculation with R. leguminosarum Norway-GFP. (Scale bar, 100 μm). (B) Principal component analysis (PCA) showing the variation between the transcriptomes of infected and noninfected groups by plotting the first two principal components. The infected accessions are depicted in shades of green whereas the noninfected ones are depicted in shades of purple. (C) Upset plot for the DEGs shared among nine pairwise comparisons. The number of DEGs in each pairwise comparison is listed at the bottom left side of the plot. The blue bar indicates the intersection for all nine pairwise comparisons. The plot was created with the function “upset” from the UpSetR package (59). (D) Heatmap for the 167 DEGs (p-adjust < 0.01) in different biological replicates from the six different accessions. Hierarchical clustering dendrograms belonging to the accessions and the DEG genes are displayed at the top and the left side, respectively.

The Transcriptomes of Infected and Noninfected Nodules Are Significantly Different.

To identify genes specifically up-regulated in infected nodules, we used prime-seq, a sensitive bulk RNA sequencing method based on molecular crowding (22, 23). We sequenced the nodule transcriptomes of the top infected (MG-70, MG-79, and MG-136) and noninfected (MG-9, MG-112, and MG-115) accessions. Specifically, nodules highly colonized by R. leguminosarum Norway-GFP were selected for the infected group, whereas nodules completely lacking infection were sampled from the noninfected group. Sequencing resulted on average 4.2 million 3′ cDNA reads/sample that were mapped to the L. japonicus Gifu v1.2 genome. The L. japonicus MG-20 v3.0 genome was used as a comparison yielding similar results (SI Appendix, Fig. S2). Reads mapped to exonic (∼65%), intronic (∼4%), and intergenic (∼10%) regions similarly across samples and accessions (SI Appendix, Fig. S3). Using exonic and intronic reads, we detected on average 571,467 unique molecular identifiers, 19,547 genes per sample, and 29,670 genes in total, which is over 98% of the genes annotated in L. japonicus Gifu v1.2.

To visualize expression patterns, we plotted the first and the second principal components of the nodules' transcriptomes, which explained more than 60% of the variation in the dataset (Fig. 1B). With the exception of one outlier, expression profiles clustered mainly by the infection status, suggesting that this is the main factor influencing differences in gene expression. To identify differentially expressed genes (DEGs), nine pairwise comparisons between infected and noninfected accessions were performed and visualized using volcano plots (SI Appendix, Fig. S4). Comparisons generated between 1,310 and 4,053 DEGs (p-adjusted < 0.01; SI Appendix, Dataset S1 A–I). By defining common DEGs as the intersection of DEGs with at least a twofold change, we identified 167 candidate genes (SI Appendix, Dataset S2 and Fig. 1C). These common DEGs showed contrasting expression patterns between the two infection phenotypes and contained 53 down-regulated and 114 up-regulated genes with reference to the infected group (Fig. 1D and SI Appendix, Dataset S2).

Differentially Expressed Genes Cluster in Three Coexpression Modules.

To identify associations between the DEGs as clusters of coexpressed genes, a Weighted Gene Co-Expression Network Analysis (WGCNA) was performed, as genes involved in the same processes are often coexpressed (24). Transcriptome data from the Lotus base (https://lotus.au.dk) was extracted from seven different conditions (leaf, mature flower, seed, root, roots 15 dpi with arbuscular mycorrhiza, and nodules at 10 and 21 dpi with M. loti R7A) for the 167 DEGs. Three modules were generated with module 1 containing 92 genes, and the second and third modules harboring 47 and 28 genes, respectively (SI Appendix, Fig. S5 and Dataset S3 A–C). Genes in module 1 clustered separately as they had a weak correlation with genes in the other two modules and the majority were down-regulated in infected nodules (SI Appendix, Dataset S3A). Modules 2 and 3 grouped genes that were mostly up-regulated in infected nodules (SI Appendix, Dataset S3 B and C). Module 2 included genes that were specifically expressed in nodules at 10 dpi, while module 3 harbored genes that were slightly expressed at 10 dpi but were strongly up-regulated at 21 dpi (SI Appendix, Fig. S5).

Genes were functionally annotated by querying the corresponding L. japonicus protein sequence against the TAIR database (https://www.arabidopsis.org). The highest scoring genes are listed in SI Appendix, Table S2. Among the 28 genes belonging to module 3, four encoded proteins with high similarity to Arabidopsis proteins with functions associated with lipid polyesters, such as suberin. Suberin is a cell wall polymer that is deposited in specialized plant tissues, such as the root endodermis (25). These genes included a Cytochrome 86A1 encoding gene (LjCYP86A1, LotjaGi6g1v0111000), which is a homolog of Arabidopsis HORST (AT5G58860), and a gene encoding a Fatty acyl-CoA Reductase (LjFAR3.2, LotjaGi3g1v0478900). In Arabidopsis, orthologs of both genes are involved in the synthesis of aliphatic long-chain monomers, which are suberin precursors (26, 27). In addition, we identified a gene encoding a transporter of the ABC-G family (LjABCG2, LotjaGi1g1v0696000) that is a homolog of Arabidopsis ABCG2 (AT2G37360), which has been associated together with other members of this family with the transport of suberin monomers from the cytoplasm to the cell wall (28); and a gene encoding a NAC transcription factor (LjNAC, LotjaGi1g1v0398100) homologous to Arabidopsis AtNAC38 (AT2G24430), which has a putative role in suberin biosynthesis (29). When including genes from modules 1 and 2, the list expanded to contain a second Fatty acyl-CoA Reductase gene (LjFAR3.1, LotjaGi3g1v0175200), a Fatty acid Desaturase gene (LjFAD, LotjaGi1g1v0726200), a second ABC-G transporter gene (LjABCG1, LotjaGi5g1v0359700), and a GDSL esterase/lipase gene (LjGDSL, LotjaGi1g1v0221300). Interlinked macromolecules of suberin and lignin act as cellular fences with roles in abiotic and biotic protection, solute diffusion limitation, and structural support (25, 30). Thus, we hypothesize that the products of these genes are involved in the deposition of suberin or other lipid polyester in nodules.

Lipid Polyester Biosynthesis Genes Are Induced in Nodules.

Suberin is a complex polymer composed of a mixture of long-chain fatty acids and polyphenols originating from the phenylpropanoid pathway (25). To investigate the expression of genes encoding enzymes involved in the synthesis of suberin precursors, L. japonicus orthologs of phenylpropanoid pathway and fatty acid synthesis genes were extracted from the Lotus base (SI Appendix, Dataset S4). As a comparison, we included lignin biosynthesis genes. Genes associated with suberin biosynthesis were in general up-regulated in infected nodules compared to roots and root hairs, while lignin and phenylpropanoid pathway genes were preferentially expressed in roots (SI Appendix, Figs. S6 and S7). To independently validate these results, we conducted qRT-PCR of L. japonicus Gifu roots inoculated with M. loti MAFF303099, a compatible L. japonicus symbiont (31). The analysis included four genes with predicted suberin synthesis or regulation functions (LjCYP86A1, LjFAR3.1, LjFAR3.2, and LjNAC) and two genes with putative cell wall associated functions (LjRBOHB and LjCOMT). All genes were specifically up-regulated in rhizobia inoculated compared to mock-treated roots (SI Appendix, Fig. S8). LjCYP86A1, LjFAR3.2, and LjCOMT reached peak expression at 14 dpi, whereas LjFAR3.1 and LjRBOHB were expressed higher at 7 dpi. This validated that genes with putative suberin or lipid polyester-related functions are induced during the root nodule symbiosis.

Fatty acyl-CoA Reductase Genes Have Distinct Expression Patterns.

FAR enzymes catalyze the formation of primary alcohols, a key step in the synthesis of suberin and other lipid polyesters (SI Appendix, Fig. S6A) (27). The distinct expression patterns of LjFAR3.1 and LjFAR3.2 (SI Appendix, Figs. S6B and S8 A and B) suggested their functional diversification. To investigate the relationships between these genes, a maximum-likelihood tree was generated with homologs of legumes and nonlegume species. The FARs encoded by LjFAR3.1 and LjFAR3.2 grouped in the subfamily 3 branch. This subfamily expanded in legumes compared to nonlegumes (Fig. 2A). L. japonicus contains six FAR3 paralogs all of which are located in Chromosome 3. LjFAR3.1 neighbors LjFAR3.5, while LjFAR3.2, LjFAR3.3, LjFAR3.4, and LjFAR3.6 are located next to each other. Four out of six paralogs (LjFAR3.1, LjFAR3.2, LjFAR3.4, and LjFAR3.6) were induced in nodules. Interestingly, FAR3.1 was also expressed in inoculated roots (Fig. 2B). To determine the spatiotemporal control of the promoters driving the expression of LjFAR3.1 and LjFAR3.2, a 3 kb upstream region from the start codon was cloned and fused to the β-glucuronidase (GUS) gene. Constructs were introduced by hairy root transformation into L. japonicus Gifu and roots were inoculated with M. loti MAFF303099. Both promoters were active in nodule primordia and mature nodules. However, only FAR3.1_pro:GUS was also active in the root (Fig. 2C). These results suggest that LjFAR3.1 mediates the deposition of lipid polyesters in roots and nodules, while LjFAR3.2, LjFAR3.4, and LjFAR3.6 might have evolved to fulfill nodule-specific functions.

Fig. 2.

Expression of Fatty acyl-CoA Reductase genes in root nodules. (A) Maximum likelihood phylogeny of LjFARs protein family built using the RAxML-HPC BlackBox tool (64) in CIPRES. Bootstrap values over 60 are indicated in the nodes. FARs identified in the differential expression analysis are highlighted in bold and blue. Arabidopsis thaliana (At), Arachis hypogaea (Ah), Cicer arietinum (Ca), Euglena gracilis (Eg), Lotus japonicus (Lj), Medicago truncatula (Mt), Oryza sativa (Os), Parasponia andersonii (Pa), and Zea mays (Zm). (B) Heatmap illustrating the expression levels of the six L. japonicus FAR3 paralogs. Expression data were extracted from Lotus Base. (C) Representative images of GUS stained roots and nodules. (Scale bar, 100 µm). (D) Modified secondary cell walls in L. japonicus nodule stained with Nile red. White arrows indicate the nodule endodermis En(n), the root endodermis En(r), and the nodule vasculature endodermis En(nv). Infected cells, ic. (Scale bar, 100 μm). (E, F) Representative images of nodule sections displaying nuclear localized fluorescent reporter (NLS-2xYFP, yellow) driven by 3 kb promoters, infected cells (M. loti MAFF 303099-GFP, magenta), and auto-fluorescent secondary cell wall modifications (gray). The outer border of the nodule is marked with dashed lines. (Scale bar, 50 µm).

Promoters of Fatty acyl-CoA Reductase Genes Are Active in the Nodule Periphery.

Nodules, in addition to the suberized vascular endodermis, possess a barrier in the nodule periphery (32). This autofluorescent layer is located between the outer and inner cortexes (Fig. 2D) and has been called nodule endodermis (32). Suberization of this cell layer in broad bean has been proposed to limit oxygen diffusion into the nodule inner tissue (19), which is essential for efficient nitrogen fixation (14). To investigate if the activity of promoters driving the expression of LjFAR3.1 and LjFAR3.2, coincided with root and nodule vascular endodermises and/or the nodule endodermis, their 3 kb promoters were fused to a two-times nuclear-localized YFP (NLS-2xYFP) and constructs were introduced by hairy root transformation into L. japonicus Gifu. The FAR3.1_pro:NLS-2xYFP reporter was active in the root and nodule vascular endodermises and in the nodule endodermis. In contrast, the FAR3.2_pro:NLS-2xYFP reporter activity was mostly restricted to the nodule endodermis (Fig. 2 E and F).

LjFAR3.2 Mediates the Deposition of Lipid Polyesters in the Nodule Endodermis.

The putative function of FARs together with the activity pattern of the LjFAR3.1 and LjFAR3.2 promoters suggested that these genes could mediate the deposition of suberin or other lipid polyester in the root and/or nodule endodermis. To determine if disruption of these genes altered deposition of lipid polyesters on endodermal tissues, two independent LOTUS RETROTRANSPOSON 1 (LORE1) mutant lines for each gene were characterized (SI Appendix, Table S3). For LjFAR3.1 the LORE1 insertions were located in the fifth intron and in the eighth exon for far3.1-1 and far3.1-2, respectively (SI Appendix, Fig. S9A). While for the LjFAR3.2 gene both insertions were located in the last exon (Fig. 3A). All lines carried additional insertions in exonic and intronic regions of other genes (SI Appendix, Table S3). Therefore, homozygous plants for the wild-type allele from the respective segregating populations carrying the background mutations were used for comparison (referred to here as WT*). Lipid polyester staining of nodules was performed using Fluorol yellow. WT* nodules showed significantly higher Fluorol Yellow signal than far3.2 mutant nodules (Fig. 3 B and C) but no significant difference was observed compared to far3.1 mutant nodules (SI Appendix, Fig. S9B). To evaluate if these genes mediate the suberization of the root endodermis, uninoculated seedlings were stained with Fluorol Yellow. In Arabidopsis, the suberization pattern of the root endodermis is patchy at the onset of suberization but it becomes continuous with endodermal differentiation (33). Roots of uninoculated L. japonicus Gifu seedlings had a similar suberization pattern as Arabidopsis. Root suberization in far3.1 mutant lines did not significantly differ from the respective WT* lines. Although both far3.2 mutants had a slight increase in the percentage of continuous suberization compared to the respective WT* plants, only far3.2-2 showed a significant increase (SI Appendix, Fig. S10). This suggests that FAR3.2 primarily mediates the deposition of lipid polyesters on the nodule endodermis.

Fig. 3.

Characterization of nodule permeability to solutes and oxygen. (A) Intron/exon structure for the LjFAR3.2 gene. Arrowheads indicate the LORE1 retrotransposon insertion sites. 5′ and 3′ untranslated regions are indicated in red. (B) Representative images of far3.2-2 and WT* nodules stained with Fluorol yellow. (Scale bar, 200 μm). (C) Box plot displaying the Fluorol yellow signal quantification for the far3.2-1 and far3.2-2 and the respective WT* nodules. (D) Graphical depiction of the permeability ratio (PR) where toluidine distance (td) and nodule distance (nd) were measured in μm. (E) Representative images of 100-μm thick nodule sections of WT* and far3.2-2 mutant nodules stained with toluidine blue. (Scale bar, 100 μm). (F) Box plot displaying the PR quantification for the far3.2-1 and far3.2-2 mutants and the respective WT* nodules. The P values were * ≤ 0.05, ** ≤ 0.01, *** ≤ 0.001, and *** ≤ 0.0001 based on the Welch's t test. (G) Schematic representation of oxygen measurement setup. (H) Measurement on the outermost nodule cells (0 µm). (I) Measurement in the nodule inner cortex (75  μm). The P values were * ≤ 0.05, and ** ≤ 0.01 based on the Wilcoxon test.

Disruption of the Nodule Endodermis Increases Nodule Permeability to Oxygen and Reduces Nitrogen Fixation.

As hydrophobic barriers restrict the diffusion of solutes and gases (34), we hypothesized that disruption of the nodule endodermis in far3.2 mutants would affect nodule permeability. Toluidine blue staining has been used to evaluate root permeability in A. thaliana (35). Mutant and WT* nodules of both far3.2 lines were immersed in toluidine blue to evaluate their permeability to solutes. As nodules vary in size, a permeability ratio (PR) was calculated. This ratio was defined as PR = td/nd, where nodule distance (nd) measures the distance from the nodule border to the closest infected cell and the toluidine blue distance (td) measures the distance that the dye penetrates in the nodule (Fig. 3D). The dye permeated further into the mutant nodules compared to the WT* (Fig. 3E), as indicated by significantly larger PRs for both LORE1 lines (Fig. 3F). To evaluate the permeability of the nodule endodermis to oxygen, we measured the oxygen concentration in the outermost nodule cells and in the nodule inner cortex using an oxygen microelectrode (Fig. 3G). Mutant nodules had significantly higher oxygen concentrations compared to wild-type nodules (Fig. 3 H and I). This suggests that oxygen diffusion is restricted by deposition of lipid polyesters in the nodule endodermis.

Oxygen denatures the nitrogenase enzyme (36). To assess nodule functionality, nitrogenase activity and plant growth were quantified. The nitrogen fixation potential was quantified using an acetylene reduction assay (ARA). Mutant lines showed a decline in the nitrogenase activity compared with the respective WT* (Fig. 4A). This correlated with a significant reduction in the shoot length of both mutant lines when grown under nitrogen-limiting conditions (Fig. 4 B and C). Mutant lines also exhibited shorter roots (SI Appendix, Fig. S11A). To investigate if the reduced shoot growth was caused by a reduction in the nitrogen fixation capacity of nodules or due to pleiotropic effects of the far3.2 mutation, plants were grown under nitrogen-limiting and nitrogen-sufficient conditions in the absence of rhizobia. Irrespective of the nitrogen concentration, the shoots of far3.2 mutant plants did not significantly differ from the respective WT* (SI Appendix, Fig. S12). Further characterization of the lines included quantification of infection threads, nodule number, and nodule colonization. Only far3.2-2 had significantly fewer nodules, whereas far3.2-1 had significantly fewer infection threads (SI Appendix, Fig. S11 B and C). Both lines had similar nodule colonization levels (SI Appendix, Fig. S11D). Overall, these results suggest that FAR3.2 is required for optimal nodule function, but not for organogenesis or infection. Combined, our results show that deposition of lipid polyesters on the nodule endodermis reduces oxygen diffusion into nodules and is essential for efficient nitrogen fixation.

Fig. 4.

Plant growth and nitrogenase activity phenotypes. (A) Box plot displaying nitrogenase activity. Each point was calculated from a five time-point linear regression. (B) Representative image of the shoot length phenotype of mutant and the respective WT* plants. (Scale bar, 1 cm). (C) Box plot displaying the shoot phenotype for the far3.2-1 and far3.2-2 mutants and the respective WT* plants. The P values represent * ≤ 0.05, ** ≤ 0.01, *** ≤ 0.001, and *** ≤ 0.0001 based on the Welch's t test. (D) Model of diffusion barrier formation in nodules. Genes encoding proteins with putative lipid polyester-related functions are expressed in infected nodules in a cell-specific manner. Lipid polyesters are deposited on the walls of cells located in the periphery of the nodule creating an oxygen diffusion barrier. Limiting oxygen diffusion protects the oxygen-sensitive nitrogenase enzyme. Oxygen is bound by nodule leghemoglobins to ensure its supply for respiration. En: endodermis, r: root, nv: nodule vasculature, n: nodule.

Discussion

One of the long-standing goals in the root nodule symbiosis field is to transfer nitrogen fixation to nonlegume crops. This is a monumental task, as this symbiosis is a complex trait regulated by hundreds of genes (37, 38). For this endeavor to be successful, it is essential to not only understand the developmental programs that control nodule organogenesis, but also the cell biology that underlies the hosting of bacteria, and the metabolic adaptations that fuel nitrogen fixation. However, the high interconnectivity between different pathways makes it difficult to genetically dissect the contribution of candidate genes to each process. Exploration of natural diversity harbors great potential to investigate context-dependent interactions, such as epistasis or developmental dependency (39). Moreover, the avenue of modern sequencing technologies makes the study of multiple nonmodel species or accessions possible. Here, we sequenced the nodule transcriptomes of six L. japonicus accessions to bypass the epistatic effect of nodule organogenesis over nodule infection and reveal genes the expression of which was specifically associated with infected nodules.

The phenotypic diversity of L. japonicus accessions in response to R. leguminosarum Norway was used to identify combinations that disconnected nodule formation from infection. We determined the transcriptomic response in nodules of accessions with contrasting infection phenotypes using the prime-seq method (22). The combination of the distinct phenotypes with sensitive transcriptomics provided a unique opportunity to identify genes directly connected to nodule infection, independent of genes associated to nodule organogenesis or nitrogen fixation. Among the 167 genes that were differentially regulated in infected compared to noninfected nodules, we not only identified genes from the common symbiotic signaling pathway, such as the Symbiosis receptor kinase (SymRK) and Nodulation Signaling Pathway 2 (NSP2), but also identified genes encoding transporters, proteins mediating the deposition of lipid polyesters in cell walls, and proteins involved in oxygen homeostasis (SI Appendix, Dataset S2). These constitute a valuable resource to explore the genetic base of the adaptations required to host bacteria inside nodules.

Oxygen is both a potent denaturing agent of the nitrogenase enzyme complex and essential for bacteroid respiration (36). To reconcile this apparent contradiction, the host carefully controls oxygen content in the nodules by limiting oxygen diffusion into the inner cortex and expressing high quantities of leghemoglobin proteins (40). Witty and collaborators proposed more than three decades ago the presence of a barrier in the nodule periphery that restricts oxygen diffusion into the nodule (12, 41). Oxygen rapidly drops in the nodule inner cortex reaching concentrations in the nanomolar range (41–43). In common bean, only few air spaces through which oxygen can freely move are visible between cells located in the transition from the outer to the inner cortex in the nodule periphery (41). In soybean, the outer and the inner cortex are separated by a cell layer with modified secondary cell walls. In broad bean the cell walls of this layer are suberized and their chemistry resemble the one of the root endodermis and has thus been denominated nodule endodermis (19). However, the ontogenesis of this tissue has not been studied and genetic evidence supporting the connection between this putative endodermal barrier and oxygen permeability was missing.

Coexpression analyses revealed that genes with functions associated with regulation, biosynthesis, and export of lipid polyesters such as suberin, coexpressed and were induced in nodules (SI Appendix, Figs. S6–S8). Promoters driving the expression of two of these genes (LjFAR3.1, LjFAR3.2) were active in nodules, in specific in the nodule endodermis (Fig. 2). FAR3.1 and FAR3.2 had distinct expression patterns. LjFAR3.1 had the highest absolute expression in roots among all FARs, but it was also expressed in nodules, while LjFAR3.2 was specifically expressed in nodules (Fig. 2B). The activity pattern of the FAR3.1 and FAR3.2 promoters was also distinct, as LjFAR3.1pro was active in root and nodules, while LjFAR3.2pro was primarily active in nodules (Fig. 2). LjFAR3.1 and LjFAR3.2 are closely related to other four genes, two of which are also expressed in nodules. It is tempting to speculate that the nodule-specific paralogs have evolved through successive gene duplications and have gained nodule-specific cis-regulatory elements. The closest Arabidopsis ortholog AtFAR3 is associated with cuticular wax synthesis and expressed in leaves, stems, flowers, siliques, and roots (44). In Arabidopsis FAR1, FAR4, and FAR5 have been associated to suberin biosynthesis in roots. They are expressed at suberin deposition sites, and their promoters are active in root endodermal cells (27). FAR1, FAR4, FAR5 cluster in a neighboring branch with no legume protein (Fig. 2A). Thus, the closest legume homologs are the FAR3 proteins. Poplar FAR3 orthologs are induced in suberized bark tissue (45). This suggests that FAR3 genes could have a role in suberin deposition in other plant species. We posit that at least FAR3.2 and possibly other homologs have been co-opted from the root endodermis program and mediate the deposition of a lipophilic polyester in the nodule endodermis. As shown for broad bean, it is likely that in Lotus this polyester is suberin (19). However, additional chemical characterization is required as variation has been described among nodule cortical tissues of different legume species (32).

Reduction of Fluorol yellow staining in the nodule endodermis in far3.2 but not in far3.1 mutants (Fig. 3 B and C and SI Appendix, Fig. S9) along with an increase in solute and gas permeability in far3.2 mutant lines (Fig. 3 F and I) show that primarily FAR3.2 mediates the formation of a diffusion barrier in the nodule periphery. In addition, the significant reduction in nitrogenase activity (Fig. 4A) and in shoot length in these lines (Fig. 4C), show that alteration of this barrier compromises nodule function. However, to accurately estimate how much does the barrier protect the nitrogenase from oxygen, more research is required. Redundancy of other paralogs such as FAR3.4 and FAR 3.6 and potential compensatory effects from leghemoglobin proteins still need to be investigated. Our results support a model in which tissue-specific induction of lipid polyester biosynthesis genes mediates the formation of the oxygen diffusion barrier in nodules, a key adaptation for endosymbiotic nitrogen fixation (Fig. 4D). It is tempting to speculate that this is not the only function of the nodule endodermis, as diffusion barriers in plants reduce nutrient leakage and prevent pathogen infection (46, 47). It remains to be investigated what other functions this barrier has and how the formation of this barrier is regulated by biotic and abiotic cues.

Materials and Methods

Bacterial Growth Conditions.

Strains used in this work are listed in SI Appendix, Table S4. Liquid cultures of R. leguminosarum Norway and M. loti MAFF303099 were grown in tryptone yeast extract broth (48) supplemented with 5 mM CaCl₂ for 48 h at 28 °C. Agrobacterium rhizogenes 1193 and Escherichia coli TOP10 were grown on Luria Bertani (LB) medium (49) for 1 d at 28 °C and 37 °C, respectively. Media were supplemented with antibiotics as follows: tetracycline (Tc, 2 to 10 µg mL⁻¹); fosfomycin (Fm, 15 µg mL⁻¹); streptomycin (Sm, 100 µg mL⁻¹); gentamicin (Gm, 25 µg mL⁻¹); kanamycin (K_m, 50 µg mL⁻¹); spectinomycin (Sp, 100 µg mL⁻¹); rifampicin (Rf, 50 µg mL⁻¹); and carbenicillin (Cb, 50 µg mL⁻¹).

Plant Material, Growth Conditions, and Inoculation.

L. japonicus seeds (SI Appendix, Table S5) were surface sterilized with 1.2% NaClO and 0.1% SDS, soaked with sterile water for 2 h, and germinated in 0.8% agar plates with 1/2 Gamborg B5 (50) medium. Seeds were kept at 24 °C for 3 d in darkness, followed by 3 d under a long day photoperiod (16h:8h, light:dark). For phenotypic screening, 30 seedlings of each L. japonicus accession were transplanted into sterile Tulip-shaped Weck jars (WECK). Each jar contained 10 seedlings grown on 300 mL of a sand:vermiculite mixture (1:2) supplemented with 40 mL of Fabaceae (FAB) medium (51). Two days after transferring, seedlings were inoculated with 1 mL of a R. leguminosarum Norway-GFP suspension (OD₆₀₀ = 0.005) previously washed with sterile water. Plants were grown under a long day photoperiod at 24 °C. Five weeks post inoculation (wpi) plants were harvested and phenotyped. For LOTUS RETROTRANSPOSON 1 (LORE1) line phenotyping, independent mutant lines (SI Appendix, Table S3) were inoculated with M. loti MAFF303099-DsRed (OD₆₀₀ = 0.005) instead. Mutant lines were screened at 3 wpi for number of nodules, shoot length, root length, infection threads, Fluorol yellow staining, permeability to toluidine blue, oxygen concentration, and acetylene reduction. For general growth assessment, lines were grown for 3 wk in Weck jars containing a sand:vermiculite mixture supplemented with 50 mL of FAB medium containing 0.1 mM (low) or 10 mM (high) KNO₃. For root suberization assessment, surface sterilized seeds were stratified for 2 d, and germinated in darkness at 21 °C for 2 d. Germinated seedlings were grown on agar plates with 1/4 BD medium and 0.5 mM KNO₃ for 5 d.

Nodule Infection Screening.

Roots systems were fixed with 4% paraformaldehyde and immersed in ClearSee solution for 24 h, as described previously (52). Forty L. japonicus accessions were screened qualitatively for the formation of nodules, bumps, tumors, and presence/absence of nodule infection 4 wk after R. leguminosarum Norway-GFP inoculation. Ten accessions, which displayed round-shaped nodules, were quantitatively screened for the percentage of infected nodules. Three accessions with either the highest or the lowest number of infected nodules were selected to quantify the area colonized. For each individual accession, 10 nodules were embedded in 6% low melting agarose and sectioned with a VT1000S vibratome (Leica Biosystems) longitudinally. The 50-µm thin sections were visualized with a DM6 B upright microscope (Leica Microsystems). Nodule colonization was calculated for each section as the percentage of infected tissue (nodule cells with GFP signal) relative to the total inner tissue (inwards from the nodule inner cortex). Areas were marked manually in Fiji (53) and quantified using the default function Measure. Each data point represents the average of the values of three sections from the middle of a nodule.

RNA Extraction, Library Preparation, and Sequencing.

Six biological replicates were collected for each accession (Infected: MG-70, MG-79, MG-136; Noninfected: MG-9, MG-112, MG-136). Each biological replicate comprised 20 nodules from at least 5 different plants. Nodules were harvested 5 wpi. Total RNA was extracted using the Spectrum Plant Total RNA kit (Sigma-Aldrich) according to the manufacturer’s instructions and treated with DNase I (Ambion). RNA integrity was examined with a 1% agarose gel in TAE buffer. A total of 36 libraries was prepared using the prime-seq method (22). Briefly, 4 ng of RNA were mixed with 5 µl of reverse transcriptase mix (Maxima H Minus reverse transcriptase 0.15 µL, Maxima RT 5× buffer 2 µL, 25 mM dNTP 0.4 µL, 100 µM TSO 0.1 µL, and UltraPure water 2.35 µL) and 1 µL barcoded oligo dT (10 µM). The reaction was incubated at 42 °C for 90 min. Cleaning of pooled samples was done with homemade SPRI beads (Sera Mag SpeedBeads in 22% PEG, 1 M NaCl, 0.01 M Tris⋅HCl, 0.001 M EDTA, 0.01% IGEPAL, and 0.05% Sodium Azide) and two times washing with 1 mL of 80% EtOH each. The final elution volume was 17 µl in UltraPure water. The cDNA was treated with exonuclease I for 20 min at 37 °C and 10 min of inactivation at 80 °C. A second purification step was carried out as described above. The cleaned cDNA was amplified (25 µL KAPA HiFi 2× RM, 3 µL 10 µM preamp primer, and 2 µL UltraPure water) by PCR. Thermocycler conditions were 98 °C 3 min, 10 cycles of 98 °C 15 s, 65 °C 30 s, and 4 min at 72 °C, followed by 10 min at 72 °C for final extension. Quality and quantity were assessed using the 2100 Bioanalyzer (Agilent) with High Sensitivity DNA Analysis Kit and with Quant-iTPicoGreen dsDNA, respectively. The library was prepared with the Nextera XT Library Prep Kit. Three replicates with 0.8 ng of cDNA were tagmented following the manufacturer’s protocol. A three prime specific primer was used to amplify the barcode and UMI sequences introduced in the reverse transcription step. Libraries were paired-end sequenced at the LAFUGA Gene Center, Munich, Germany with an Illumina HiSeq 1500. Deep sequencing was between 5 to 10 Mio raw reads per sample. Raw data were processed using the zUMIs pipeline v2.5.4 (54) and mapped to the L. japonicus Gifu v1.2 (55) and L. japonicus MG-20 v3.0 (56) reference genomes obtained from the Lotus Base (https://lotus.au.dk/).

Differential Expression Analysis.

The DESeq2 package v.1.36.0 (57) within R was used for differential expression (DE) analysis. A total of nine pairwise comparisons were performed between the transcriptomes of infected and noninfected accessions. For each individual analysis, a false discovery rate (FDR) < 0.05, α = 0.01, and a log₂fold change > 1 were set as a threshold to identify DEGs. Volcano plots were obtained with the package EnhancedVolcano (58) with a foldchange cutoff of 2 and a P value cutoff of 10⁻¹⁰. To find the shared DEGs across all nine pairwise comparisons, the UpSetR package (59) was used. An Upset plot was generated to display intersections between all combinations.

Coexpression Analysis.

Gene clusters that coexpressed together were identified by a weighted correlation network analysis (WGCNA) (60). A network dendrogram was created with 167 DEGs, which were shared across all pairwise combinations, and normalized expression data of those genes in seven different Lotus tissues and treatments (leaf, mature flower, seed, root, AM experiment at 15 dpi, and nodule at 10 and 21 dpi). Expression values for the different conditions were retrieved from the L. japonicus Gifu RNA-seq data in the Lotus Expression atlas tool within the Lotus Base (https://lotus.au.dk/expat/). Due to the low number of genes, a one-step gene network construction and module detection were used. First, a network topology analysis was done to select proper soft thresholding based on a Pearson correlation. Second, the WGCNA function blockwiseModules was used to detect modules of coexpressed genes. The minimum number of genes detected by a module was set to 30 with a standard merging threshold of 0.25. Results were plotted with the function plotDendroAndColors, also within the WGCNA package v.1.71. The heatmap for each module was created with the function pheatmap (61).

Phylogenetic Analyses.

Accession numbers of all protein sequences used in this study are listed in SI Appendix, Table S6. Phylogeny of the Fatty acyl-CoA Reductases (FARs) was created by retrieving the protein identifiers of FARs from Arabidopsis thaliana [At], Zea mays [Zm], Oryza sativa [Os], Solanum lycopersicum [Sl], Parasponia andersonii [Pa], Euglena gracilis [Eg] and the legumes L. japonicus Gifu [Lj], Medicago truncatula [Mt], Cicer arietinum [Ca], and Arachis hypogaea [Ah]. Protein sequences were retrieved from a previously published analysis (62) and the NCBI (https://www.ncbi.nlm.nih.gov/proteins/) by using blastP. The protein alignment was done in MAFFT using default settings (63) and nonconserved regions were removed by manually trimming gaps in the alignment. Maximum-likelihood phylogeny trees were constructed in CIPRES (http://www.phylo.org/) using default parameters of the RAxML-HPC BlackBox tool version 8.2.12 (64). Trees were displayed with Interactive Tree Of Life v5 (https://itol.embl.de/).

Cloning.

Primers and plasmids used in this work are listed in SI Appendix, Tables S7 and S8, respectively. All promoters were cloned using the Golden Gate toolkit (65). For each promoter a 3 kb region in the L. japonicus MG-20 genome was amplified by PCR with Phusion DNA polymerase (Thermo Fisher Scientific). Blunt end cloning with StuI or SmaI was employed to insert each fragment into a level 1 pUC57 backbone and subsequently moved via Esp3I into level 3 expression vectors encoding a NLS-2xYFP fluorescent reporter or a GUS reporter. All level 1 plasmids were sequenced using the Sanger method (Sequencing service, Biocentre LMU Munich).

Hairy Root Transformation.

Transient root transformation was conducted by the hairy root method (66). Roots of L. japonicus Gifu seedlings were cut and hypocotyls were submerged in an A. rhizogenes 1193 suspension transformed with the reporter constructs (SI Appendix, Table S8). Treated hypocotyls were placed on 1/2 Gamborg’s B5 agar (50) and incubated for 2 d in the dark at room temperature. Hypocotyls were then incubated under a long day photoperiod for 3 d. To remove A. rhizogenes, hypocotyls were transferred to Gamborg’s B5 agar plates supplemented with cefotaxime (300 µg mL⁻¹). Plants were then screened for the presence of an NLS-2xmCherry transformation marker under a M165FC stereo microscope (Leica Biosystem) equipped with a red filter. Three weeks post cutting, plants bearing transformed roots were transferred into sterile Weck jars with 300 mL of a sand:vermiculite mixture (1:2) supplemented with 40 mL of a low nitrogen FAB medium. After 2 d, plants were watered with 10 mL of FAB medium containing M. loti MAFF303099-DsRed (OD₆₀₀ = 0.005).

Promoter GUS Assay.

Transgenic roots carrying pro:GUS constructs (SI Appendix, Table S8) were inoculated with M. loti MAFF303099-DsRed and harvested at 7 and 21 dpi. Roots were incubated in staining buffer containing 100 mg mL⁻¹ of X-Gluc in DMSO, 100 mM phosphate buffer (pH = 7.0), 0.5 mM EDTA (pH = 7.0), 0.5 mM K₃[Fe(CN)₆] and 0.5 mM K₄[Fe(CN)₆] at 37 °C for 8 h, as described previously (67). Roots were washed with 100 mM phosphate buffer (pH = 7.0) and fixed with 2.5% glutaraldehyde in 0.1 M sodium phosphate buffer (pH = 7.0) under vacuum for 45 min. Samples were inspected under a VHX-6000 digital microscope (Keyence).

Quantitative RT-PCR.

Differential expression of candidate genes was validated by qRT-PCR. L. japonicus Gifu root systems inoculated with M. loti MAFF303099 or mock-treated were collected at 3, 7, and 14 dpi, frozen immediately in liquid nitrogen, and total RNA was extracted. cDNA synthesis was performed according to the manufacturer's instructions using SuperScript III reverse transcriptase (Invitrogen). qRT-PCR was performed in a Quantstudio5 system (Thermo Fisher) in a final volume of 7 µL with 2× SYBR Green master mix (Thermo Fischer), 1:10 (vol/vol) dilution of the cDNA, and 0.3 µM of each primer. LjPPA2A was used as reference. The thermal cycler conditions were: 95 °C for 2 min, 40 cycles of 95 °C for 30 s, 56 °C for 30 s, and 72 °C for 20 s. qRT-PCR primers used in this study are listed in SI Appendix, Table S7.

Permeability Assay.

Intact root systems were incubated for 30 mins in a 0.1% toluidine blue solution. Nodules were then sectioned (100 µm) as previously described and observed under a CTR 6000 upright microscope (Leica Microsystems). Permeability to the dye was estimated as a permeability ratio (PR). The PR was defined as PR = td/nd where nodule distance (nd) indicates the length from the nodule border to the closer infected cell border and the toluidine blue distance (td) measures the distance that the dye penetrates. nd and td were measured in Fiji (53).

Root and Nodule Staining.

Nile red staining was adapted from a protocol to detect secondary cell wall modifications in Arabidopsis roots (68). Nodules were submerged overnight in a ClearSee solution supplemented with 0.05% Nile red for staining of hydrophobic polyesters, such as suberin. Subsequently, nodules were washed in ClearSee for 1 h with constant shaking. The solution was replaced every 20 min. Fluorol yellow staining was performed as previously described (69). Nodules were collected 21 dpi, fixed in a 4% formaldehyde solution, and cleared using ClearSee. Samples were embedded in 6% NuSieve GTG agarose, sectioned (100 μm thickness) using a VT1000S vibratome (Leica), and stained sequentially with 0.01% (vol/vol) Fluorol yellow 088 and 0.5% Aniline blue solutions. Signal intensity in the area of interest was quantified using ImageJ as (https://theolb.readthedocs.io/en/latest/imaging/measuring-cell-fluorescence-using-imagej.html). Roots were collected 7 d after gemination and stained as nodule sections. Start of suberization was determined at the position of the first suberized endodermal cell. Start of continuous suberization zone was determined at the position where endodermal cells at phloem pole are continuously suberized and the first unsuberized endodermal cell is longitudinally sandwiched by suberized endodermis.

Confocal Microscopy.

Sixty-five–micrometer thin agarose sections of transgenic or stained nodules were observed under a TCS SP5 confocal microscope (Leica Microsystems) equipped with a 20× HCX PL APO water immersion lens. Secondary cell wall components were excited by a diode laser and detected at 405 to 450 nm. The green fluorescent protein (GFP) from the tagged M. loti and the NLS-2xYFP reporter were excited with an argon laser at 488 and 514 nm, and emission was detected at 493 to 515 and 520 to 540 nm, respectively. Nile red was excited at 561 nm using a diode pumped solid-stage laser and detected at 600 to 650 nm. Fluorol yellow signal in nodules was imaged using an Axio Zoom.V16 stereo microscope (Zeiss) equipped with GFP excitation (450 to 490 nm) and emission (500 to 550 nm) filters. Root suberization was imaged using an LSM 980 (with Airyscan 2) confocal (Zeiss).

Oxygen Measurements.

A precalibrated profiling oxygen microsensor PM-Pst7 (PreSens) was used to measure oxygen concentrations at room temperature. Recording was done with the PreSens Measurement Studio2 (PreSens). Fresh nodules were embedded in 7% low melting agarose. The microelectrode was inserted perpendicularly to the top of the nodule using a manual micromanipulator (PreSens). Six measurements were recorded at the top of the nodule (0 μm) and in the nodule inner cortex (75 μm) and averaged for each biological replicate.

Acetylene Reduction Assay.

The nitrogenase activity was quantified by measuring the reduction of acetylene into ethylene using gas chromatography-flame ionization detection (GC-FID). Two independent LORE1 lines were inoculated with M. loti MAFF303099-DsRed, as described previously, and assayed at 21 dpi. Five biological replicates were analyzed. A single replicate comprised two plants with nodulated roots in a 25 mL glass tube with 500 μL of FAB medium and sealed with a rubber stopper. Subsequently, 1 mL of air was extracted and replaced with 1 mL of acetylene. Per each replicate the gas mixture was sampled at 0, 20, 40, 60, and 80 min while keeping the plants at 28 °C in a water bath. At each time point, 1 mL of the mixture was injected into a GC 2010 Pro (Shimadzu). The activity was calculated as nanomoles of ethylene per hour using a linear regression.

Statistical Analysis.

All statistical analyses were conducted in R (70). The DESeq2 v.1.36.0 (71) and the WGCNA v.1.71 (60) packages were used for DEG and for Coexpression analysis, respectively. Tukey’s HSD and ANOVA tests were performed with the package agricolae v.1.3-5 (72). Pairwise comparisons were performed using the function compare_means within the package ggpubr v.0.4.0.999 (73).

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information. RNA sequencing data have been deposited in GEO (GSE216502) (74).