The Medicago truncatula lncRNA ENOD40 is a mediator of microRNA169-controlled NF-YA activity in nodule initiation

Significance

The involvement of the long noncoding RNA ENOD40 in legume root nodule formation is well documented. However, the molecular mechanism through which ENOD40 facilitates this process has not yet been elucidated. Here, we generate an enod40-1/ enod40-2 double mutant in the model legume Medicago truncatula and show that its impaired nodule initiation can be rescued by a region that harbors a highly conserved stretch of 24 nucleotides, designated as box2. Our data support a model in which box2 functions as a target mimic for miR169 and thereby relieves posttranscriptional repression of NF-YA1, which is an essential transcription factor in the nodulation process.

Abstract

The hallmark of the legume lncRNA EARLY NODULIN40 (ENOD40), involved in rhizobium-induced nodulation, is the presence of a highly conserved stretch of 24 nucleotides, designated box2, preceded by a small open-reading-frame (sORF) coding for a peptide of 12 to 13 amino acids. Although there is a well-established link between ENOD40 and nodulation, it is not fully clear by which mechanism ENOD40 functions in this process. Here, we show that a region harboring box2 can complement nodule formation in an ENOD40 knock-out mutant (enod40-1-2/1). The sequence of box2 bears the characteristics of a miR169 target mimic. We show that the artificial target mimic MIM169defg can indeed complement the reduced capability of nodule formation in enod40-1-2/1, and that box2 exhibits target mimic activity in a transient luciferase assay. In addition, the introduction of a miR169-resistant form of MtNF-YA1 also elevates the capacity to form nodules in enod40-1-2/1. We conclude that ENOD40 effectuates nodule initiation by posttranscriptional upregulation of the miR169 target NF-YA1, which encodes an essential transcription factor in this step of the nodulation process.

Long noncoding RNAs (lncRNAs) are RNA molecules of at least 200 nucleotides in size that do not possess substantial protein-coding capacity, although studies differ in their criteria for the maximum lengths of potential open reading frames (ORFs) (1–3). lncRNAs are widely spread in eukaryotes, with over 13 000 loci identified in the genome of Arabidopsis thaliana (Arabidopsis), and over 250 000 in that of humans. Despite their lack of significant protein coding capabilities, lncRNAs have been found to be key regulators in a wide range of biological processes (2, 4, 5). More recently, combined efforts of data mining and proteomics have uncovered large pools of peptides encoded by genes originally annotated as lncRNAs (6).

As the first lncRNA identified in plants, ENOD40 was discovered in the legume soybean among genes that are upregulated during nodule organogenesis, induced by soil-borne bacteria collectively known as rhizobia (7, 8). ENOD40 has drawn much attention, both for its dynamic early expression pattern during this process, as well as for its nature as an early example of a lncRNA harboring small open-reading frame (sORF) peptide. The presence of this sORF and its role as RNA shuttle for RNA-Binding Protein 1 (9, 10) has dubbed it as bifunctional RNA (11).

During the interaction between roots of the model legume Medicago truncatula (Medicago) and rhizobium, ENOD40 is induced in pericycle cells facing the proto-xylem pole at the onset of organogenesis and subsequently in the developing nodule primordium (12, 13). In mature nodules, ENOD40 genes are highly expressed in the pericycle of the nodule vasculature, the proximal meristem, and infection zone of the central zone of the nodule (14). Accordingly, RNAi of ENOD40 in both Medicago (15) and Lotus japonicus (Lotus) (16) dramatically reduced nodule number. In Medicago, overexpression of ENOD40 caused cell divisions in more proximal differentiated parts of the roots (17) and nodules induced closer to the root tip (18).

The overall nucleotide sequence homology of ENOD40 genes among legumes is low due to numerous insertions and trinucleotide replications. In contrast, a strong conservation in RNA structure has been uncovered, suggesting that gene activity is RNA-mediated (19). The most conserved region among ENOD40 RNAs is a stretch of 21 to 24 nucleotides designated as box2, which is flanked by two conserved stem-loop domains. In addition, legume ENOD40 RNAs contain a small region, designated as box1, which harbors an ORF for a small peptide of 12 or 13 amino acids upstream of box2 (13). Bombardment studies involving fragments of MtENOD40-1 in Medicago roots indicate that box1 and box2 regions separately are sufficient to induce cell divisions (20). Strikingly, ENOD40 genes are also present in nonlegume plants such as Arabidopsis (19), rice (21), tomato (22), suggesting that they play roles in processes unrelated to nodule formation.

Another group of lncRNAs are miRNA target mimics, also known as miRNA target decoys, that base-pair to miRNAs and thereby sequester them. Upon binding of the miRNA, target mimics fail to be cleaved, either because certain nucleotides between the miRNA and lncRNA are mismatching, or because the target mimic forms a bulge on the binding site (23). As a result of this, the miRNA remains associated to the target mimic, preventing the former from binding to and causing degradation of its target mRNAs. This lack of degradation in turn leads to an increased translation of these miRNA targets (24). After the discovery of the first target mimics, IPS1 and its close paralog AT4 in Arabidopsis (24), bioinformatics studies have identified many potential target mimics in plants, some of which have been experimentally validated (25, 26).

Several miRNAs have been shown to be involved in root nodulation, among them members of the miR169 family that target the transcription factor NF-YA1, that together with its paralog NF-YA2 has been shown to play crucial roles in nodulation (27–30). Strikingly, the expression of miR169, NF-YA1, and ENOD40 are all upregulated during nodulation in Medicago (14, 27, 31). Here, we show that a partial reverse-complementarity between box2 and miR169 leads to miR169 mimicry activity of box2 which is critical for the regulation of NF-YA1 mRNA levels during nodule initiation.

Results

The Initiation of Nodulation Is Reduced in ENOD40 Double Mutants.

RNAi studies (15) in Medicago hairy roots have shown that a reduction in ENOD40 expression correlates with a decrease in nodule numbers. In order to confirm this phenotype in stable ENOD40 knock-out plants, we generated the homozygous ENOD40 double mutant line enod40-1-2/1 in the Medicago R108 background. This line was obtained by crossing the MtENOD40-1 Tnt1 insertion line NF1186 (32), where Tnt1 is inserted 7 bases upstream of box2, with the MtENOD40-2 Tnt1 insertion line NF7811 (32), where Tnt1 is inserted 64 bases upstream of box2 (Fig. 1A). ENOD40 gene expression diminished in nodules of enod40-1-2/1, as measured by qPCR. (Fig. 1B).

Fig. 1.

Characterization of the Medicago enod40 1-2/1 double mutant. (A) Schematic presentation of MtENOD40-1 and MtENOD40-2 with the location of the TnT1 insertion sites in NF1186 and NF7811, respectively. The conserved box 1 and box 2 regions are indicated with “1” and “2,” respectively. (B) Neither MtENOD40-1 nor MtENOD40-2 exhibited significant expression levels in mature root nodules of the double mutant enod40-1-2/1. Relative expression data were determined by qPCR and normalized to 1 for expression in R108 using MtACTIN2 as a reference. Error bars show the SEM of three technical replicates. (C) Nodule numbers are significantly reduced in double mutants enod40-1-2/1, compared to R108. Each dot represents the count of nodules on a plant, 21 d after inoculation with S. meliloti 2011. Black bars represent the mean nodule number for each genotype. ****P < 0.00001.

To test nodulation efficiency, we grew seedlings of enod40-1-2/1 (n = 22) and R108 (n = 21) wildtype for 21 d in the presence of S. meliloti 2011. On roots of enod40-1-2/1, significantly less nodules were formed than on wildtype roots (Fig. 1C) (P < 0.00001). In line with the aforementioned findings (15), this demonstrates that ENOD40 genes facilitate nodule formation. However, as the mutant still produces nodules, the role of ENOD40 in this process is not essential.

To identify whether the decreased nodule formation results from a lack of initiation or defects during development, we performed a spot-inoculation on roots of both R108 wildtype and enod40-1-2/1 with S.meliloti 2011. Normally, at 8 d after application, the formation of small nodules can clearly be distinguished. This was the case for 11 out of 18 inoculated R108 roots, whereas on the 18 inoculated enod40-1-2/1 roots, only 1 nodule was macroscopically visible.

To determine whether the spots that failed to form nodules had developed an arrested primordium, we isolated 1-cm root segments containing these spots and imaged them using confocal microscopy. Because anticlinal pericycle divisions in the earliest stage of nodulation cannot be reliably distinguished from lateral root primordia (12), we focused on the identification of cortical cell divisions, indicative of primordia of stage II and later (12). These could be detected in only one of the nonnodulating enod40-1-2/1 roots, where anticlinal cortex divisions resembled a stage II primordium, although some abnormal anticlinal divisions in the second cortex layer were also present (SI Appendix, Fig. S1). The absence of observable nodule primordia on the other 16 inoculated enod40-1-2/1 roots indicates that the reduction in visible nodules is likely the result of a lower capacity to initiate nodule formation.

Next, microsections of 21d-old nodules formed on R108 and enod40-1-2/1 roots were made to enable more precise studies on the effect of the absence of ENOD40 expression in nodule development. Like in R108 (Fig. 2A), the central part of the enod40-1-2/1 nodule (Fig. 2A) consists of an apical meristem and proximal infection and fixation zones. Each zone in R108 and the enod40-1-2/1 nodule is built up of a comparable number of cell layers. At the nodule periphery, vascular bundles are positioned within the cortex that is bordered by an endodermis. Hence, the enod40-1-2/1 nodules that do develop, do not show any severe developmental defects.

Fig. 2.

enod40-1-2/1 nodules have wild type–like morphology and MtENOD40 and MtNF-YA1 have similar spatial expression patterns during nodule primordium formation. (A) Longitudinal plastic sections (7 µm thick) of R108 and enod40-1-2/1 nodules 21 d after inoculation with S. meliloti 2011, counterstained with toluidine blue. M—meristem, IZ—infection zone, FZ—fixation zone. Arrows point to nodule vascular bundles (Scale bar, 100 µm.) (B and C) Representative images of RNA in situ localization of MtENOD40 (A) and MtNF-YA1 (B) transcripts in nodule primordia at developmental stage V. Both genes are expressed in pericycle and cortical cell layers 4 and 5. c3, cortical cell layer 3; c4/5 cortical cell layers 4 and 5; en, endodermis; pc, pericycle. (Scale bar, 100 µm.)

ENOD40 box2 Is Required to Complement Nodule Numbers in enod40-1-2/1.

Previously it has been reported that introduction of ENOD40 box1 and box2 regions separately into Medicago roots led to cell divisions in the cortex (20), suggesting that they harbor activities that can provoke similar responses. We questioned whether these regions also have a comparable function in the initiation of nodules. To this end, we generated three constructs to express mutated variations of MtENOD40-1 under its own promoter: pMtENOD40-1::MtENOD40-1Δatg, in which the start codon of the box1 ORF is mutated to AAG; pMtENOD40-1::MtENOD40-1Δbox2, in which a 62-nt region comprising box 2 is deleted; and pMtENOD40-1::MtENOD40-1ΔatgΔbox2, which has both the aforementioned mutations. In two consecutive experiments, we transformed these constructs into roots of enod40-1-2/1, which were then inoculated with S. meliloti 2011 to compare their ability to initiate nodules to roots transformed with a negative control construct (pMtENOD40-1::GUS).

First, to test whether box2 functions in nodule initiation, we compared the ability of pMtENOD40-1::MtENOD40-1Δatg and pMtENOD40-1::MtENOD40-1ΔatgΔbox2 to complement the low nodulation phenotype of enod40-1-2/1 by counting the number of transformed roots on which nodules were formed. Twenty-one days after rhizobium inoculation, 47/76 (62%) of roots transformed with pMtENOD40-1::MtENOD40-1Δatg roots developed nodules, whereas this was the case for only 21/49 (43%) of roots transformed with pMtENOD40-1::MtENOD40-1ΔatgΔbox2. Compared to the negative control, where 22/57 (39%) of transgenic roots developed nodules, the increase in nodulation capacity is only statistically significant for roots transformed with pMtENOD40-1::MtENOD40-1Δatg (P = 0.0078). Furthermore, the increase in nodulation capacity is also statistically significant between roots transformed with pMtENOD40-1::MtENOD40-1Δatg and pMtENOD40-1::MtENOD40-1ΔatgΔbox2 (P = 0.0374). From this, we conclude that the 62-nt region of ENOD40 containing box2 is positively involved in the initiation of nodule formation.

Next, we tested whether pMtENOD40-1::MtENOD40-1Δbox2 can also complement the low initiation of nodulation capacity of enod40-1-2/1. Twenty-one days after rhizobium inoculation, 7/17 (41%) of roots transformed with pMtENOD40-1::MtENOD40-1Δbox2 had developed nodules, which is comparable to the 17/36 (47%) negative control roots (P = 0.6798). Hence, the ENOD40-1 box 1 peptide on its own most likely does not play a role in the initiation of nodules.

Summarizing, these complementation experiments show that the 62-nt region containing box2, but not box1, of ENOD40 is involved in the initiation of nodule formation.

The Sequence of ENOD40 box2 Indicates a Molecular Function as miR169 Target Mimic.

Although the foregoing experiments highlight the importance of box2, they do not elucidate the molecular mechanism behind its function. Close inspection of the 21 to 24 nucleotide ENOD40 box2 sequences in different plant species reveals two especially conserved segments of ten and four nucleotides (Fig. 3A, marked in red), interspersed with a less conserved segment of six nucleotides (Fig. 3A, marked in blue). Sequence analysis shows that the two highly conserved segments are reverse-complementary to the sequence of micro-RNA 169 (example for MtENOD40-1 and MtmiR169g shown in Fig. 3B), suggesting a potential interaction between box 2 of ENOD40 and miR169.

Fig. 3.

The sequence of ENOD40 box2 hints at a molecular function as miR169 target mimic. (A) Alignment of the RNA sequences of ENOD40 box 2 in various angiosperm species. Nucleotides matching in all displayed sequences are marked in gray. The red bars indicate two highly conserved segments of 10 and 4 nucleotides; the blue bar indicates the less conserved segment in between them. (B) Reverse-complementary alignment of MtENOD40-1 box 2 and MtmiR169g RNA sequences. The highly conserved 10- and 4-nt regions in the sequence of ENOD40 are indicated in red, the less conserved interspacing segment is indicated in blue, corresponding to panel A. Reverse-complementarity to miR169 is found for the red segments but not for the interspacing blue segment, hinting at a molecular function as target mimic.

Strikingly, the less conserved segment that interspaces the two reverse-complementary segments does not exhibit reverse-complementarity to the micro-RNA (Fig. 3B). Furthermore, the absence of reverse-complementarity in this segment is conserved in all ENOD40 box 2 sequences, indicating that it could be of biological relevance. Such central mismatches between transcripts and miRNAs are typical for so-called miRNA target mimics, such as the miR399 target mimic IPS1 in Arabidopsis (24, 33). Interestingly, the artificial miR169 target mimics MIM169 and MIM169defg (34) have been shown to enhance the transcript levels of miR169-repressed NF-YA transcription factors (35). In Medicago, NF-YA1 is strongly induced during nodulation, and is repressed by miR169. NF-YA1 plays a crucial role in nodulation (27), as the nf-ya1-1 null mutant produces less nodules (29), similar to ENOD40 double mutants. We hypothesized that box2 of ENOD40 functions as a miR169 target mimic and thereby positively mediates nodulation through NF-YA1 activity.

The Induction of the miR169 Target NF-YA1 in Response to Nod Factor Is Diminished in Roots of enod40-1-2/1.

If ENOD40 functions as a miR169 target mimic, we expect that the level of NF-YA1 mRNA is reduced in double mutant enod40-1-2/1 compared to wildtype, since enod40-1-2/1 will have a larger pool of free miR169 that can bind to and cause degradation of NF-YA1 transcripts. To test this, we exposed enod40-1-2/1 and R108 wildtype roots to a 2- or 4-h treatment with S.meliloti 2011 secreted nodulation initiation signal molecules (nod factors) and quantified the expression of NF-YA1 in the susceptible zone by RT-qPCR (Fig. 4). NF-YA1 is about 10-fold induced in wildtype roots after 4 h of nod factor treatment, while induction is nearly abolished in enod40-1-2/1 treated roots. In contrast, the upregulation of the nodulation master regulator NIN is similar in wildtype and mutant, showing that enod40-1-2/1 roots are still responsive to nod factors (Fig. 4). These observations indicate that ENOD40 increases the level of NF-YA1 mRNA in Medicago.

Fig. 4.

Induction of MtNF-YA1 in response to nod factors is diminished in enod40-1-2/1. Relative transcript levels of MtNF-YA1 (Left) and MtNIN (Right) in the root susceptible zone of wild-type R108 and enod40 1-2/1 seedlings, at 2- and 4-h following nod factor application, along with a mock treatment. Compared to R108, enod40 1-2/1 shows a significant reduction in MtNF-YA1 transcript levels following nod factor inoculation, while the transcript level of MtNIN is comparable. Relative expression data were determined by qPCR and normalized to 1 for expression in R108 mock using MtACTIN2 as a reference. Shown graphs are the means ± SEM of three biological replicates. The value of each biological replicate is the mean of three technical replicates. (n.s. = not significant; *P < 0.05; **P < 0.01.)

A prerequisite for cell-autonomous regulation of NF-YA1 transcript level by ENOD40 is that the expression patterns of the two genes, as well as that of miR169, are overlapping. To test this, we performed in situ hybridization on nodule primordia of stage V. NF-YA1 and ENOD40 are coexpressed during nodule formation in the pericycle and primordium (Fig. 2 B and C), which agrees with promoter-GUS expression studies of these genes (12, 30). A thorough spatiotemporal analysis on the expression of miR169 is difficult due to the presence of 12 annotated precursors in miRbase and slight variations in the mature sequences thereof (36). However, root expression and upregulation of miR169 in response to nodulation have been reported in an RNA-Seq study (31). Taken together, the near abolishment of NF-YA1 induction following nod factor treatment in enod40-1-2/1 and the colocalization of ENOD40 and NF-YA1 expression further support the hypothesis that ENOD40 is involved in the positive regulation of NF-YA transcript levels.

Complementation of enod40-1-2/1 with MIM169defg.

So far, we have shown that a 62-nt region containing box2 is required for the function of ENOD40 on nodule initiation, that the sequence of box2 exhibits the characteristics of a miR169 target mimic, and that ENOD40 is positively involved in the regulation of the miR169-repressed nodulation transcription factor NF-YA1.

If ENOD40 affects nodule initiation through miR169 target mimicry as hypothesized, we expect that the introduction of an artificial miR169 target mimic MIM169defg (34) under the promoter of MtENOD40-1 (pMtENOD40-1::MIM169defg) in roots of enod40-1-2/1 will affect the capacity to initiate nodule formation in a way similar to pMtENOD40-1::MtENOD40-1Δatg. To test this, we compared the ability of enod40-1-2/1 roots transformed with pMtENOD40-1::MIM169defg to form nodules to roots transformed with the negative control pMtENOD40-1::GUS.

Twenty-one days after inoculation with S. meliloti 2011, nodules had developed on 39/58 (67%) of roots transformed with pMtENOD40-1:MIM169defg, compared to 21/47 (45%) of roots transformed with the negative control. This increase is statistically significant (p = 0.0202), indicating that, like ENOD40 box2, a miR169 target mimic can complement the low nodule initiation in enod40-1-2/1.

Box2 of MtENOD40-1 Shows Target Mimic Activity in a Transient Luciferase Assay in Nicotiana Leaves.

To further consolidate the function of ENOD40 box2 as miR169 target mimic, we conducted a transient luciferase assay in leaves in Nicotiana benthamiana, similar to previous experiments for other target mimics (23, 24, 33). We generated constructs for the 35S-mediated expression of variations of the MtENOD40-1 sequence: MtENOD40-1ΔATGΔbox2, MtENOD40-1ΔATG, MtENOD40-1ΔATGbox2-3’mut (with the 3’ region of box2 mutated GGCA > ccgA), and MtENOD40-1ΔATGbox2-5’mut, (with the 5’ region of box2 mutated CGGCAAGUCA > gGcCAAcUgA). The 35S::Fluc-NF-YA1 3’UTR (wt) construct was also created to express firefly luciferase fused to the miR169-sensitive 3’UTR of MtNF-YA1.

We coinfiltrated leaves of N. benthamiana with Agrobacterium tumefaciens carrying each ENOD40 variant, 35S::Fluc-NF-YA1 3’UTR (wt), 35S::miR169g for miR169g overexpression, and 35S::Rluc for the miR169-independent expression of renilla luciferase as internal standard. We also included a positive control for target mimicry, substituting ENOD40 for MIM169defg. Protein samples were extracted 3 d postinfiltration, assayed for Fluc and Rluc activity (Fig. 5).

Fig. 5.

MtENOD40-1 box2 acts as a miR169 target mimic in a transient luciferase assay. The ratio of activity of firefly luciferase (Fluc)-fused to the MtNF-YA1 3’UTR-to renilla luciferase (Rluc)-used as internal control-was quantified following transient expression in N. benthamiana leaves. Constructs for the expression of Fluc-MtNF-YA1 3’UTR and Rluc were co-agroinfiltrated with constructs for the expression of MtmiR169g and either variants of the MtENOD40-1 sequence or the artificial target mimic MIM169defg. Significance groups were determined using a one-way ANOVA with Tukey’s HSD post hoc test (P < 0.05). (A) When Fluc is fused to the wildtype, miR169-susceptible MtNF-YA1 3’UTR, coexpression with MtENOD40-1ΔATG and MIM169defg leads to significantly higher Fluc/Rluc ratios compared to MtENOD40-1ΔATGDbox2. MtENOD40-1ΔATGbox2-3’mut also shows significant upregulation of Fluc-NF-YA1 3’UTR activity compared to MtENOD40-1 ΔATGDbox2, while MtENOD40-1ΔATGbox2-5’mut does not. n = 18 per coinfiltration. (B) When Fluc is fused to a miR169-resistant version of the MtNF-YA1 3’UTR, no significant difference in Fluc/Rluc ratio is measured between MtENOD40-1ΔATG and MtENOD40-1 ΔATGDbox2. n = 12 per coinfiltration.

If MtENOD40 box2 acts as a miR169 target mimic, the ratio of Fluc/Rluc activity will be higher in leaves infiltrated with t MtENOD40-1ΔATG compared to MtENOD40-1 ΔATGDbox2, as only the former can alleviate miR169-mediated repression of the Fluc-NF-YA1 3’UTR transcript. Indeed, both MtENOD40-1ΔATG and positive control MIM169defg show significantly higher Fluc/Rluc ratios compared to MtENOD40-1 ΔATGDbox2. Interestingly, MtENOD40-1ΔATGbox2-3’mut still shows significant upregulation of Fluc-NF-YA1 3’UTR activity, at least to a partial degree, while MtENOD40-1ΔATGbox2-5’mut does not (Fig. 5A). This indicates that the 5’ 10-nt conserved region of box2 is critical to its function, while the 3’4-nt conserved region is less so.

As an additional control, we repeated the experiment for MtENOD40-1ΔATGΔbox2 and MtENOD40-1ΔATG, now fusing Fluc to a mutated miR169-resistant version of the NF-YA1 3’UTR (27) (35S::Fluc-NF-YA1 3’UTR (res)). If box2 functions as target mimic, its presence will not affect the Fluc/Rluc ratio, as Fluc is not being repressed by miR169. Indeed, no significant difference in Fluc/Rluc ratio is observed between MtENOD40-1ΔATGΔbox2 and MtENOD40-1ΔATG (Fig. 5B). Together, these results show a function for box2 as target mimic, with a critical role for its 5’ 10-nt conserved region.

Complementation of enod40-1-2/1 with pMtENOD40-1::MtNF-YA1.

Finally, to test whether ENOD40 effectuates nodule initiation through upregulation of NF-YA1, we generated a construct to express NF-YA1 under the promoter of ENOD40-1 (pMtENOD40-1::MtNF-YA1res). To prevent degradation of NF-YA1 transcripts because of increased miR169 activity in enod40-1-2/1, the miR169 binding sites in the 3’-UTR of NF-YA1 are mutated as described previously (27). In the same fashion as the previous experiments, we generated transgenic roots in enod40-1-2/1 with this construct, as well as pMtENOD40-1::GUS. Twenty-one days after inoculation with S. meliloti 2011, 50/67 (75%) of roots transformed with pMtENOD40-1::MtNF-YA1res had formed nodules, which is significantly more than the 28/52 (54%) of roots transformed with pMtENOD40-1::GUS (P = 0.0179).

To exclude the possibility that the observed increase in nodulation ability did not result from a rescue of NF-YA1 activity but instead from an overall increase thereof, the same experiment was performed on transgenic roots generated on wildtype R108 plants. Twenty-one days after inoculation with S. meliloti 2011, 20/34 (59%) of roots transformed with pMtENOD40-1::MtNF-YA1res had developed nodules, which is lower but not statistically different from the 28/38 (74%) of roots transformed with pMtENOD40-1::GUS (P = 0.1817). Taken together, we can conclude that the expression of a miR169-resistant NF-YA1 under the promoter of ENOD40-1 only enhances nodulation ability in the absence of ENOD40 activity, indicating that the inhibited nodule initiation of ENOD40 double mutants is the result of decreased NF-YA1 expression.

In conclusion, our studies indicate a mechanism where ENOD40 expression, through its control of NF-YA1 mRNA level, mediates successful nodule initiation.

Discussion

Our data indicate that box2 of ENOD40 functions as a miR169 target mimic and thereby increases the transcript level of NF-YA1, an essential nodulation transcription factor. In this respect, the activity of ENOD40 is comparable to IPS1 in Arabidopsis, which acts as a target mimic for miR399 that targets PHO2 mRNA (24). Several studies based on sequence analysis have predicted the presence of many more lncRNAs functioning as miRNA target mimics in plants (25, 26, 37), but experimental studies to underpin such a role are rare (38, 39). ENOD40 distinguishes itself from other miRNA target mimics in the sense that it has also been shown to encode for a small peptide. In this respect, ENOD40 can be grouped among recently identified micropeptide encoding LncRNAs (6). As the sequence of the encoded peptide is conserved within but not between plant families, it is likely that its function has diversified during plant evolution.

Besides its target mimicry and peptide-encoding functionalities, ENOD40 is also reported to act as a shuttle between cytoplasm and nucleus for the splicing involved protein MtRBP1, independently of box2 (9, 10, 40). This latter function is derived from gain-of-function experiments in Arabidopsis, and hence the potential relevance for the nodulation process remains to be determined. Regardless, it is likely that next to its function as target mimic, ENOD40 possesses more biological modes of action, possibly through its peptide or strongly conserved RNA structure (19).

Consistent with RNAi experiments (15), nodule number is reduced in the ENOD40 double mutant enod40-1-2/1. We showed that this reduction in nodules results from a decreased capacity to initiate nodule formation, as indicated by the absence of cortical cell divisions typically observed for nodule primordia in the vast majority of spot-inoculated enod40-1-2/1 roots. Interestingly, in wildtype roots, even the earliest stages of nodule initiation coincide with an increased activity of ENOD40 in both the pericycle and the cortex (12). This suggests that ENOD40 is involved in the induction of cell divisions at the onset of nodule development. Furthermore, the observation that dividing pericycle cells during the formation of lateral roots show decreased ENOD40 expression (12, 22) indicates that the involvement of ENOD40 in the induction of cell divisions is specific for the nodulation program. Previously, it has been reported that overexpression of MtENOD40 leads to induction of cortical cell divisions in Medicago roots in the absence of Rhizobium (17). However, this increase in cell division activity does not lead to either an increase or ectopic localization of nodules once roots become exposed to S.meliloti 2011 (18). This suggests that the ENOD40-mediated induction of cell divisions is not sufficient to develop nodules.

We have observed that the ENOD40 box2 region, the artificial target mimic MIM169defg, and a miR169-resistant NF-YA1 under the promoter of MtENOD40-1 all complement the nodule number, and thereby the induction of cell divisions at the onset of nodulation, in enod40-1-2/1. This supports a model in which the effect of ENOD40 on the induction of cell divisions is modulated through miR169 target mimicry and NF-YA. This is in line with the previously reported effect of NF-YAs on cell divisions in Lotus. Here, it was demonstrated that ectopic overexpression of LjNF-YA1 and LjNF-YB1 induces extra cell divisions in and around lateral root primordia, as well as ectopic cell divisions in the cortex (41).

In addition to a decrease in nodule number, overexpression of miR169, RNAi on NF-YA1, and loss-of function of NF-YA1 cause disorganization of nodule zonation, severe defects in nodule infection, and premature arrest of nodule growth (27, 29). In contrast however, mature nodules of enod40-1-2/1 do not show any of these phenotypes (Fig. 2B). The expression of NF-YA1 during later stages of nodulation has been shown to be regulated by a polypeptide translated from an alternatively spliced upstream ORF (uORF), which represses the expression of the full-length NF-YA1 protein and thereby ensures proper nodule zonation (28). Strikingly, gain-of-function phenotypes for both miR169 (27) and the NF-YA1 uORF peptide (28) resemble the two classes of nodule phenotypes described for nf-ya1 mutants (12). As the expression of NF-YA1 is regulated through multiple mechanisms, the increase of miR169 activity in enod40-1-2/1 might simply be insufficient to trigger overt nodule phenotypes. To further investigate the intricacies of the regulation of NF-YA1 activity by ENOD40 target mimicry, a comprehensive understanding of the spatiotemporal expression pattern of miR169 is essential. Although this is currently challenging due to the presence of many miR169 isoforms and variations in their mature sequences, ongoing and rapid technological advances in single-cell RNA sequencing and spatial transcriptomic hold promise for addressing these problems in the future.

Previously, RNAi studies in Medicago suggested that down-regulation of ENOD40 genes in Medicago nodules also affected bacteroid development (15). However, such phenotype was not observed in either RNAi studies in Lotus (16), or in our present study. Therefore, we postulate that the effect on bacteroid development, observed in the Medicago RNAi study, might be due to off-target effects of the introduced RNAi construct. On the other hand, all these studies reveal a reduction in nodule initiation, further corroborating ENOD40 involvement in nodule initiation.

Combined, our data indicate that a main function of ENOD40 is to promote posttranscriptional increase of NF-YA1 transcript levels, a key transcription factor in nodulation. Interestingly, in soybean, this regulation operates in the opposite direction, as NF-YA-c, which is the ortholog of Medicago NF-YA1, induces transcription of ENOD40 (42), suggesting a positive feedback-loop mechanism for NF-YA-c regulation. In Lotus, contrarily, LjENOD40-1 is strongly induced by overexpression of NIN, but not of NF-YA1 (41), pointing to an induction in parallel of NF-YA1 and its positive regulator of mRNA levels. These findings indicate that the miR169 related role of ENOD40 is embedded in additional regulatory mechanisms to ensure fine-tuned levels of NF-YA1 for modulating nodule development.

Materials and Methods

Hairy Root Transformation and Nodulation Assay.

Medicago seeds were sterilized by incubating for 7 min in concentrated H₂SO₄, washing 6 times with demineralized water, followed by incubating for 7 min in commercial bleach, and washing 6 times with sterilized demineralized water. To increase germination efficiency, the seeds were left in ~3 ml of water with ~200 µM of GA3 for 90 min after the final washing step. After the GA3 treatment, the water was removed and the seeds were put on circular 9 cm Fahräeus plates, which were sealed with parafilm and wrapped in aluminum foil. The plates were subsequently put at 4 °C for 24 h, and upside-down at 22 °C for 24 h to cold-shock and germinate the seeds.

Agrobacterium rhizogenes MSU440 was transformed with the desired construct by electroporation and grown on solid LB medium with the appropriate antibiotic. Hairy root transformation was conducted as described (43).

After transformation, Medicago plants were transferred to agra-perlite saturated with nitrate-free Fahräeus medium and grown for 3 wk under long-day conditions.

S. meliloti 2011 transformed with GFP (43) was grown at 28 °C in YEM medium until the culture was saturated. The bacteria were pelleted at 4,000×g for 10 min and resuspended in tap water to an OD600 of 0.1 to 0.2. This suspension was poured onto the plants in the perlite, using 150 ml for each ~50 × 50 cm tray of ~80 plants. The plants were grown for 21 more days to allow for nodules to form.

Three weeks after inoculation with S. meliloti 2011, each plant was carefully dug out and the root system was washed under a flowing tap to remove residual perlite. Using a fluorescence binocular, transgenic roots were identified by red fluorescence, separated from the other roots, and cut off at the hypocotyl. The transgenic roots were subsequently spread out on agar plates and checked for nodules using a brightfield binocular. Transgenic roots of total length <5 cm that failed to develop nodules were discarded. Statistical comparison was performed with a chi-square test.

Identification and Characterization of Insertion Lines in MtENOD40-1 and MtENOD40-2.

To determine the effect of the Tnt1 insertions (32) in the two ENOD40 genes on MtENOD40 gene expression, we analyzed the location of the Tnt1 insertion in each gene. On isolated genomic DNA we conducted PCR with primers (Dataset S1) flanking the genes (Mt41F, Mt41R and Mt42F, Mt42R, respectively) and primers pointing outward of the Tnt1 insertion (tnt1F and tnt1R). Fragments were purified and subjected to DNA sequence analysis.

To test the effect of the insertion on the transcription of the mRNA we conducted qPCR on cDNA derived from RNA isolated from roots of the plants. Primers were used that are in the part upstream of the insertion site (qMt401box1F and R for MtENOD40-1 expression and qMt402box1F and R for MtENOD40-2 expression).

Next both lines were crossed, and the offspring was self-pollinated to obtain plants homozygous for insertions in both ENOD40-1 and ENOD40-2 This led to lines enod40-1-2/1 and enod40-1-2/2, of which the first is being used in our studies. Nodulation capacity of enod40-1-2/1 was determined 3 wk after inoculation by S. meliloti 2011 as described above.

Cloning.

Gateway cloning (Thermo Fisher Scientific, Waltham, MA, USA) was used to create all constructs for experiments in Medicago. Ligation mixtures were transformed into One Shot TOP10 chemically competent cells (ThermoFisher) according to manufactory protocol. Cells were plated on LB plates with the appropriate antibiotics and incubated at 37 °C overnight. Entry clones were checked by digestion with a suitable restriction enzyme, and dideoxy sequencing using M13F and/or M13R primers. Expression clones were checked by digestion with a suitable restriction enzyme.

A ~3.8 kb DNA fragment upstream of MtENOD40-1 transcribed region was amplified on genomic Medicago DNA with primers pMt40BPF and pMt40BPR provided with site enabling recombination of the purified PCR fragment through BP clonase into pDONR244.

The truncated sequences of MtENOD40-1ΔATG and MtENOD40-1ΔATGΔbox2 were synthesized at Twist Biotech (Dataset S2) with CACC at the 5’ end and ligated into pENTR/D-TOPO.

To create MtENOD40-1Δbox2, two DNA fragments flanking box2 were obtained by PCR on cDNA of nodules, involving Mt401F and Mt4015R for the 5’ end and Mt4013F and Mt401R for the 3’ end. The DNA fragments were diluted 1:500 and used as a template in a second PCR to introduce short overlaps (primers Mt401F and Mt4015XR and Mt4013XF and Mt401R, respectively). Subsequently, after purification these PCR products were diluted 1:500 and used in a third PCR with primers Mt401F and Mt401R to create a single amplicon. This final PCR fragment was cloned into pENTR/D-TOPO yielding MtENOD40-1Δbox2.

The original pGreen MIM169defg plasmid (34) was used as template to amplify a DNA fragment suitable for ligating into pENTR/D-TOPO with primers MIM169F and MIM169R.

To clone the miR169-resistant NF-YA1 gene, we first amplified the NF-YA1 transcript up to its 3’-UTR by PCR using primers NF-YA1F and NF-YA1R, which contained respectively CACC TOPO-cloning and Acc65I sites, and ligated the PCR fragment into pENTR/D-TOPO. The mutated NF-YA1 3’-UTR was synthesized at Twist Biotech with Acc65I and PauI restriction sites at the 5’ and 3’ ends, respectively, and fused to the NF-YA1 transcript in pENTR/D-TOPO by classical restriction/ligation cloning using Acc65I and PauI.

The aforementioned pENTR/D-TOPO clones were recombined in a multisite LR reaction with pDONR244– pMtENOD40-1 and pENTR p2rp3 MCS Stop-Term into the pKGW-RR -MGW backbone vector containing AtUBQ10::DsRED1 as a selection marker, to create the expression clone for hairy root transformation.

Constructs for the luciferase assay were generated with Golden Gate cloning using the MoClo Toolkit and Plant parts (Addgene Kits #1000000044 and #1000000047) (44).

The Golden Gate-domesticated sequences of MtNF-YA1 3’-UTR variants, wildtype and miR169-resistant (27), fused to the Nos terminator, the MtmiR169g precursor (27), MtENOD40-1ΔATGΔbox2, and MIM169defg (34) were synthesized by Twist Biotech with BpiI restriction sites. A box 2-customizable version of MtENOD40-1ΔATG was similarly prepared, where box 2 was replaced with a BsmBI-flanked selection cassette expressing mRFP under the lac promoter.

Both variants of MtNF-YA1 3’-UTR-NosT were cloned into pICH41276 and combined with pICH51277 (CaMV 35S promoter + TMV Omega leader) and pICSL80001 (Luciferase CDS) into pICH47732 (level 1 acceptor). The MtmiR169g precursor was cloned into pAGM9121 and combined with pICH41388 (CaMV 35S promoter) and pICH41414 (CaMV 35S terminator) into pICH47742. The same was done for MtENOD40-1ΔATGΔbox2, MIM169defg, and the box 2- customizable MtENOD40-1ΔATG using the pICH47752 level 1 acceptor. Subsequently, all level 1 cassettes were separately cloned into level 2 acceptor pICSL4723 using the appropriate dummies and end linkers. To generate the expression constructs for ENOD40 with wildtype and mutated box 2 variants, two annealed oligonucleotides comprising the desired box 2 sequence and compatible 5’-overhangs were scarlessly cloned into the customizable MtENOD40-1ΔATG level 2 vector in a Golden Gate reaction with BsmBI.

All constructs were validated by restriction digestion, and all level 0 constructs as well as level 2 constructs with cloned box2 sequences were validated by dideoxy sequencing.

Nod Factor Inoculation of Medicago Roots.

For nod factor treatment, a 200 µl sample of a DMSO-dissolved nod factor extract of S. meliloti 2011 was added to 20 ml of liquid Fahräeus medium. For the mock treatment, 200 µl of DMSO was added to 20 ml of liquid Fahräeus medium.

Medicago seeds were sterilized, stratified, and germinated as described above. Seedlings were grown on 0.25 mM nitrate plates provided with 1 μM aminoethoxyvinylglycine (AVG; Sigma-Aldrich Company Ltd, Darmstadt, Germany) for five days prior to the experiment.

The seedlings were transferred to fresh filter-equipped square 12 cm plates containing Fahräeus medium with 0.25 mM nitrate salts/ 1 mM AVG in such a way that the root tips were at approximately the same height. The plates were positioned at a slight angle by placing them on top of their lids. Using a pipette, the bottom ~3 cm of the roots were sprayed with 1 ml of liquid Fahräeus medium with either the bacterial extract or the mock. This solution was allowed flow to the bottom of the diagonally positioned plate for 1 min, after which as much liquid as possible was sucked up and reapplied on the bottom ~2 cm of the roots, this time with the plate positioned horizontally. The plates were kept in this horizontal position for 5 min, after which they were sealed with parafilm and wrapped in aluminum foil in the same fashion as before. After the treatment, the bottom 1 cm of the roots was harvested for RNA isolation.

Luciferase Assay Using Transient Expression in Leaves of Nicotiana benthamiana.

Agrobacterium tumefaciens strain C58(pMP90) was transformed with the desired level 2 constructs and plated on selective LB medium with 25 mg/l rifampicin and 50 mg/l kanamycin. Culturing of agrobacteria and infiltration of Nicotiana benthamiana leaves was performed as described earlier (45). For infiltration, agrobacterium suspensions (OD600: 0.5) carrying constructs for the expression of firefly luciferase fused to the 3’-UTR of MtNF-YA1, MtmiR169g, variations of MtENOD40-1/MIM169defg, and the renilla luciferase internal standard (45) were mixed in a volumetric ratio of 1:1:5:0.2. Every tested combination was infiltrated into at least 3 leaves on separate plants. 3 d after infiltration, 3 to 4 disks of 15 mm in diameter were harvested from each leaf and assayed for luciferase activity using a Glomax™ 96 microplate luminometer (Promega) as described before (45).

Microsectioning.

Embedding of nodules in Technovit 7100 (Heraeus Kulzer), sectioning and staining were performed according to (12). Sections were analyzed with the DM5500B microscope equipped with a DFC425C camera (Leica)

In Situ Hybridization.

RNA in situ hybridization was conducted using Invitrogen ViewRNA ISH Tissue 1-Plex Assay kits (Thermo Fisher Scientific) according to the manufacturer’s user guide and optimized for Medicago root and nodules sections (46). RNA ISH probe sets were designed and synthesized at Thermo Fisher Scientific. Catalogue numbers are VF-20311 for NF-YA1 (MtrunA17_Chr1g0177091) and VF1-17674 for ENOD40 (MtrunA17_Chr5g0413247). Images were taken with an AU5500B microscope equipped with a DFC425c camera (Leica).

Spot Inoculation and Imaging of Cell Divisions in Medicago Roots.

Seeds were sterilized and germinated as described above. Seedlings were first grown for 5 d on Fahräeus medium containing 0.25 mM nitrate salts and 1 mM AVG. Spot-inoculation was conducted by applying 0.5 µl of S.meliloti 2011 culture of OD600 0.02 to the susceptible zone of the roots. Eight days after spot inoculation, a ~1 cm segment, with the inoculated spot in the center, was cut off from all roots that were not visibly developing a nodule. These segments were incubated in ClearSee solution (47) for 3 wk, refreshing the solution every 3 d. Then, the root segments were transferred to ClearSee provided with 100 µg/ml Calcoflour White. After 60 min, roots were transferred to a microscope slide with ClearSee and analyzed for cell divisions using a Leica TCS SP8 confocal microscope.

RNA Isolation and Expression Studies.

Plant material was sampled in a 2 ml tube containing a steel ball. The tubes were frozen in liquid N, after which the samples were homogenized using a Qiagen TissueLyser LT at maximum speed for ~20 s. After homogenization, samples were put back into liquid N and RNA was isolated using the E.Z.N.A. Plant RNA Kit (Omega Bio-Tek), following the manufacturer’s protocol. From these RNA samples, 1 µg was used for cDNA synthesis using the iScript cDNA synthesis kit (Bio-Rad), following the manufacturer’s protocol. Each 20 µl cDNA reaction mixture was diluted to 100 µl using Milli-Q water, this dilution was used for the qPCR analysis.

qPCR was performed using a CFX Connect thermal cycler (Bio-Rad) and IQ SYBR Green Supermix (Bio-Rad). MtAct2 (MtrunA17_Chr2g0278591) was used as reference gene. Bio-Rad CFX Maestro software was used for quality control of the data and determination of C_T values. The 2^−(ΔΔC_T) method (48) was used to determine the relative expression level of each biological replicate using the mean C_T of its technical replicates. Statistical analysis was performed on relative expression levels of biological replicates, using a Student’s t test.

ENOD40 DNA Sequences.

M. truncatula X80262 (14); L. japonicus AJ271787 (49); A. thaliana AK220907 (19); Oryza sativa AB024054 (21); Solanum lycopersicum AY388519 (22); Zea mays CD990776 (13).

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Data, Materials, and Software Availability

All study data are included in the article and/or supporting information.