MIR2111-5 locus and shoot-accumulated mature miR2111 systemically enhance nodulation depending on HAR1 in Lotus japonicus

Legumes utilize a shoot-mediated signaling system to maintain a mutualistic relationship with nitrogen-fixing bacteria in root nodules. In Lotus japonicus, shoot-to-root transfer of microRNA miR2111 that targets TOO MUCH LOVE, a nodulation suppressor in roots, has been proposed to explain the mechanism underlying nodulation control from shoots. However, the role of shoot-accumulating miR2111s for the systemic regulation of nodulation was not clearly shown. Here, we find L. japonicus has seven miR2111 loci, including those mapped through RNA-seq. MIR2111-5 expression in leaves is the highest among miR2111 loci and repressed after rhizobial infection depending on a shoot-acting HYPERNODULATION ABERRANT ROOT FORMATION1 (HAR1) receptor. MIR2111-5 knockout mutants show significantly decreased nodule numbers and miR2111 levels. Furthermore, grafting experiments using transformants demonstrate scions with altered miR2111 levels influence nodule numbers in rootstocks in a dose-dependent manner. Therefore, miR2111 accumulation in leaves through MIR2111-5 expression is required for HAR1-dependent systemic optimization of nodule number.

L and plants ensure their continuous growth under fluctuating conditions by sharing environmental information between the leaves and roots. For example, information about light conditions for the leaves 1 and nitrogen (N) deficiency in the roots 2 is transferred using a long-distance communication system 3 . These shoot-root communications are accomplished delivering signaling molecules between shoot and roots, such as proteins 1 , peptides 2,4-6 , phytohormones [7][8][9] , and RNA species [10][11][12] , which are produced locally, in response to various environmental cues. In Arabidopsis, HY5, a bZIP transcription factor, translocates from shoots to roots in response to light and regulates root growth and nitrate uptake 1 . C-TERMINALLY ENCODED PEPTIDE (CEP) family peptides produced in nitratedeficient roots serve as N-demand signals and systemically regulate nitrate uptake through shoot-acting receptors, CEPRs, and shoot-to-root mobile CEPD peptides 2,4,6 .
Leguminous plants have evolved shoot-mediated long-distance signaling systems to maintain mutualism with rhizobia, soil N 2fixing symbiotic bacteria, that inhabit specialized symbiotic organs, root nodules [13][14][15][16][17] . Through this symbiotic interaction, rhizobia fix atmospheric N 2 into ammonium, a plant-available form of N, in exchange for photosynthate from the host plant. Therefore, root nodule symbiosis is beneficial to host plants under N-limiting conditions. However, excessive nodule formation strongly inhibits host growth because N 2 -fixation is a highly energy-consuming process. To optimize the number of nodules, legumes utilize a long-distance negative-feedback mechanism known as autoregulation of nodulation (AON) 17 .
AON is achieved by signal transduction between roots and shoots through a shoot-acting leucine-rich repeat receptor-like kinase (LRR-RLK). The LRR-RLK was isolated from hypernodulation mutants of various legume species 13,18 , and designated HYPERNODULATION ABERRANT ROOT1 (HAR1) in Lotus japonicus [14][15][16] , SUPER NUMERIC NODULES (SUNN) in Medicago truncatula 19 , NODULE AUTOREGULATION RECEPTOR KINASE in soybean 20 , and SYM29 in pea 16 . Reciprocal grafting of these mutants demonstrated that these LRR-RLKs act in the shoot systemically to inhibit nodule formation in roots. These LRR-RLKs have an indispensable role in AON to perceive rootderived mobile CLV3/ESR-related (CLE) peptides, such as CLE ROOT SIGNAL 1, 2, and 3 (CLE-RS1, 2, and 3) in L. japonicus 5,21,22 , CLE12 and CLE13 in M. trunctula 23 , and RHIZOBIA-INDUCED CLE1 (RIC1) and RIC2 in soybean 24,25 . All of these peptides are synthesized in roots in response to rhizobial infection or high soil nitrate concentrations. In L. japonicus, triarabinosylation of CLE-RS2 peptide is required for the direct binding of HAR1 and suppression of nodulation in a HAR1dependent manner 5 . Moreover, the inhibitory effects of CLE peptides on nodulation in L. japonicus depend partially on the gene known as PLENTY, which encodes hydroxyproline-O-arabinosyl transferase (HPAT) 26 . In M. truncatula, mutants of ROOT DETERMINED NODULATION1 27,28 , an orthologous gene to PLENTY, show a similar phenotype to plenty. Thus, arabinosylation of CLE peptides by HPATs is likely to be required in order for them to be recognized by LRR-RLKs. After CLE peptides are perceived by the shoot-acting LRR-RLK, nodulation in roots becomes inhibited by TOO MUCH LOVE (TML), the gene for which was isolated from hypernodulation mutants of L. japonicus 29,30 . Since TML encodes an F-box/kelch-repeat protein, it is hypothesized that TML may be involved in ubiquitin proteasome-mediated degradation of unknown target proteins that are associated with early nodule development 30 . M. truncatula has two TML orthologues, TML1 and TML2, that inhibit nodule formation downstream of SUNN 31 . Hence, these factors that are involved in systemic signaling are well-conserved in legume species.
Two types of molecules generated in shoots downstream of LRR-RLK that are involved in AON have been reported to date. One is a phytohormone, cytokinin (CK). In L. japonicus, CK accumulates in shoots in a HAR1-dependent manner after inoculation with rhizobia 7 . When exogenously applied to shoots, CK causes a decrease in nodule numbers in a TML-dependent manner. The other factor is microRNA (miRNA) miR2111, which targets TML mRNA and enhances nodulation 12,32 . In L. japonicus and M. truncatula, the accumulation of mature miR2111s is drastically reduced in both shoot and roots in response to rhizobial infection and nitrate treatments, depending on HAR1/SUNN. Thus, HAR1/SUNN negatively regulates the accumulation of miR2111s, which target TML mRNA to enhance nodulation under non-inoculated or nitrate-deficient conditions 12,32 .
A substantial portion of miRNA gene families in plants consists of multiple loci [33][34][35] . Each individual locus in a certain miRNA gene family often exhibits different expression levels and patterns and sometimes is involved in distinct developmental processes even though the target genes are shared in the gene family [36][37][38][39] . Three loci for miR2111 (MIR2111-1, -2, and -3) have been reported in L. japonicus. Ectopic overexpression of MIR2111-3 in hairy roots increases the amount of mature miR2111 and nodule numbers. Because the activity of MIR2111-3 promoter is detected predominantly in the phloem of leaves, translocation of shoot-derived mature miR2111s into roots has been proposed to explain anticorrelation between the levels of mature miR2111s and of TML mRNA in roots. However, it remains unclear whether or not MIR2111-3 is a responsible locus for AON since MIR2111-3 expression levels and its dependency on HAR1 have not been evaluated. Moreover, systemic regulation by shoot-accumulating mature miR2111s of nodulation has not been clearly shown.
In this study, we identified four L. japonicus miR2111 loci, the expression of which was detected in leaves and controlled by HAR1 LRR-RLK. We attempted reverse genetic analyses such as CRISPR and short tandem target mimic (STTM) technologies to elucidate the function of highly regulated miR2111 loci on shootmediated control of nodulation. Furthermore, we examined whether or not shoot-derived miR2111 is sufficient for the systemic regulation of nodulation in roots by grafting experiments using miR2111 overexpressed line and downregulated lines.

Results
Identification of MIR2111-2, -4, -5, and -7, the expression of which is repressed by rhizobial infection in a HAR1-dependent manner. Three genomic loci for miR2111 precursors (MIR2111-1, -2, and -3) have been reported in L. japonicus (Supplementary  Table 1). These miR2111 gene loci contain a total of five miR2111 hairpin sequences, which generate three mature miR2111 isoforms, miR2111a, miR2111b, and miR2111c 12 . We investigated additional potential miR2111 hairpin sequences using BLAST search with an improved reference genome assembly for L. japonicus 40 (Gifu v1.2), and hairpin structure prediction 41,42 . In addition to the five known miR2111 hairpin sequences, we identified five sequences that are potentially processed to become miR2111s and designated a total of 10 miR2111 hairpin sequences as miR2111-A to miR2111-J (Supplementary Table 1 and Supplementary Fig. 1a). The five new sequences were predicted to form hairpin structures containing mature miR2111 sequences in their stems ( Supplementary Fig. 1b). However, the predicted structures of three hairpin sequences (miR2111-F, -H, and -J) were not typical among reported miRNA structures since they have the miR2111 duplex at the proximal end of the hairpin structure [43][44][45] (Supplementary Fig. 1b).
Next, we carried out RNA-seq analyses in mature leaves of L. japonicus (2 weeks after germination) to examine the expression of miR2111 precursors and performed RNA-Seq-based gene prediction using Stringtie 46 to identify functional miR2111 loci containing predicted hairpin structures (Fig. 1a-d), because miRNAs are generated by processing of primary transcripts of miRNA genes (pri-miRNAs), which are polycistronic in some cases. We detected four pri-miR2111 sequences and named their genes MIR2111-2, MIR2111-4, MIR2111-5, and MIR2111-7 ( Fig. 1a-d). MIR2111-4 and MIR2111-5 are monocistronic miR2111 loci containing miR2111-F and miR2111-G, respectively. MIR2111-2 and MIR2111-7 are polycistronic and possesses dual-hairpin-structure sequences. MIR2111-2 contains the sequences for miR2111-B and miR2111-C, whereas MIR2111-7 possesses the sequences for miR2111-I and miR2111-J (Supplementary Fig. 1a, b and Supplementary Table 1). The nucleotide sequence of MIR2111-7 showed 96% identity with that of MIR2111-2, suggesting that these two loci are paralogous. Furthermore, we searched new potential miR2111 genes that were not detected from our RNA-seq data by BLASTn search using all miR2111 gene sequences as the query. We found a MIR2111-4 paralogous sequence and named it MIR2111-6, which showed 93% identity with MIR2111-4 and contained the sequence for miR2111-H ( Supplementary Fig. 1a).
Apart from MIR2111-1 on chromosome 1, each monocistronic and polycistronic miR2111 locus was tandemly paired and arranged near the pericentromeric region of chromosome 3 ( Supplementary Fig. 1a, b). It is noteworthy that three of the four pri-miR2111s that were expressed in leaves, MIR2111-4, MIR2111-5, and MIR2111-7, were different to known miR2111 loci. In addition, the MIR2111-2 transcript detected in this study was much longer (789 bp) than that reported previously (312 bp) 12 .
MIR2111-2 and MIR2111-5 promote nodulation and mature miR2111 accumulation. We evaluated nodulation phenotypes of L. japonicus roots ectopically overexpressing the miR2111 genes we identified to confirm whether or not the genes are functional. We chose MIR2111-2, MIR2111-4, and MIR2111-5 from among the four miR2111 genes that demonstrated expression in leaves. MIR2111-7 was not incorporated into the functional analyses since it was over 90% identical to MIR2111-2. For the overexpression assay, DNA fragments fully covering pri-miRNA regions (~700-900 bp) were expressed under an L. japonicus Ubiquitin promoter 47 ( Supplementary Fig. 1a-d). We generated hairy roots transformed with either proUBQ:MIR2111-2, proUBQ: MIR2111-4, or proUBQ:MIR2111-5. Nodule numbers were counted in transformed roots displaying fluorescence of a GFP transformation marker at 21 days after inoculation (dai) with DsRed-labeled M. loti. Furthermore, we conducted qRT-PCR analyses to examine the accumulation of mature miR2111s and TML mRNA in transformed roots at 5 dai. Nodule numbers were significantly increased in roots transformed with proUBQ:MIR2111-2 and proUBQ:MIR2111-5 compared with empty vector controls ( Fig. 2a-e); the mean nodule numbers were about fivefold and eightfold higher in proUBQ: MIR2111-2 and proUBQ:MIR2111-5 roots, respectively (Fig. 2e). In addition, these transformed roots produced mature nodules that were apparently smaller than those in control roots, similarly to tml roots 29 ( Fig. 2a-d). proUBQ:MIR2111-2 and proUBQ: MIR2111-5 also increased the accumulation of mature miR2111s about 20-fold and 180-fold over controls, respectively (Fig. 2f). Conversely, the abundance of TML mRNA was decreased in these transformed roots (Fig. 2f). These results indicated that MIR2111-2 and MIR2111-5 enhanced nodulation and inhibited TML mRNA accumulation similarly to MIR2111-3 12 . Therefore, we concluded that MIR2111-2 and MIR2111-5 were functional in promoting nodule formation by influencing the accumulation of TML mRNA.
However, proUBQ:MIR2111-4 influenced neither the number of nodules nor TML mRNA accumulation, even when the levels of mature miR2111 in the transformed roots we detected by qRT-PCR were comparable with those in roots transformed with proUBQ:MIR2111-2 (Fig. 2f). This result suggested that MIR2111-4 is not functional with respect to regulating nodule numbers.
MIR2111-5 is expressed predominantly in the phloem of mature leaves in a HAR1-dependent manner. Our RNA-seq analyses showed that, among L. japonicus miR2111 loci, MIR2111-5 most abundantly accumulated in leaves of noninoculated plants. In addition, compared with the other miR2111 loci tested, MIR2111-5 overexpression in roots was the most effective at increasing the number of nodules. Therefore, we focused on MIR2111-5 in further analyses.
In order to characterize the spatial expression pattern of MIR2111-5, a 3.0-kb DNA fragment upstream of MIR2111-5 was inserted upstream of a GUS reporter gene. Histochemical GUS staining assay was conducted using plants stably transformed with this reporter construct (proMIR2111-5:GUS). GUS staining was detected along with vascular bundles in true leaves and cotyledons of 2-week-old plants (Fig. 3a). At this point after germination, L. japonicus MG-20 seedlings have two sets of fully expanded leaves. GUS expression was detected strongly in the more mature leaves, while much weaker expression was seen in younger leaves (Fig. 3a). GUS expression tended to be clear in vascular bundles around the leaf margins and cotyledons, but was not detected in midveins (Fig. 3a). In contrast, GUS signals were negligible in stems, roots, and root nodules (Supplementary Fig. 3a-d). This GUS expression along with vasculatures was attenuated within 5 dai with M. loti (Fig. 3b). Cross-sections of leaves revealed that the GUS reporter was expressed mainly in the phloem (Fig. 3c, d), which is similar to the expression pattern previously reported for MIR2111-3 12 .
Next, we examined the effects of rhizobial infection on the accumulation of mature miR2111s, MIR2111-5, and TML mRNA in leaves and roots by qRT-PCR. Mature miR2111s were clearly detected in both leaves and roots of mock controls, and accumulation decreased within 3 dai in a HAR1-dependent manner (Fig. 3e, f). TML mRNA levels were negatively correlated with the accumulation of mature miR2111s in roots (Fig. 3i). Similar to mature miR2111s, the abundance of MIR2111-5 in leaves also altered in response to inoculation with M. loti depending on HAR1 (Fig. 3g). However, the accumulation of MIR2111-5 in roots was sustained at very low levels or below detectable levels even in har1-7 roots where mature miR2111s accumulated in abundance (Fig. 3h). Patterns of endogenous MIR2111-5 accumulation were similar to those of GUS expression driven by its own promoter (Fig. 3a, b and Supplementary  Fig. 3a-d).
The accumulation of mature miR2111s in shoots is sufficient to enhance nodulation in roots. According to the current hypothesis, mature miR2111s are synthesized in leaves and translocated to roots via phloem vessels to activate nodulation through post-transcriptional inhibition of TML mRNA accumulation 12 . Consistent with this hypothesis, promoter activity of MIR2111-5 was predominantly detected in the phloem of leaves and cotyledons (Fig. 3a-d). Indeed, MIR2111-5 was abundant in leaves of uninoculated plants (Fig. 3g). Nevertheless, it was not clear whether shoot-derived miR2111 truly be able to systemically enhance nodulation in roots.
To clarify the functional relevance of shoot-derived miR2111s in nodulation, we generated stable proUBQ:MIR2111-5 transgenic lines (MIR2111-5ox) ( Supplementary Fig. 4). MIR2111-5ox displayed significantly increased nodule and infection thread numbers, indicating MIR2111-5 overexpressed lines successfully generated ( Supplementary Fig. 4). Next, we performed reciprocal grafting between MIR2111-5ox and WT plant. Scions overexpressing MIR2111-5 significantly increased nodule numbers in WT rootstocks (28 dai) (Fig. 4a-e). This result indicated that MIR2111-5 overexpression in shoots was sufficient to promote nodulation. Next, we examined the effects of MIR2111-5 overexpressed scions on the accumulation of mature miR2111s, MIR2111-5, and TML mRNA in rootstock using qRT-PCR. WT rootstocks grafted with MIR2111-5ox scions showed significant increases in mature miR2111s and a reduction of TML mRNA levels (5 dai) (Fig. 4f), indicating that MIR2111-5 overexpression in shoots can systematically regulate the accumulation of mature miR2111s and TML mRNA in roots. Grafting between MIR2111-5ox scions and tml-4 rootstocks or har1-7 scions and MIR2111-5ox rootstocks did not exhibit additive effects on nodule numbers compared with self-grafted mutant plants, suggesting that MIR2111-5 functioned in the same genetic pathway as HAR1 and TML in the control of nodule number ( Supplementary  Fig. 5). ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-19037-9 In contrast to mature miR2111s, MIR2111-5 levels in WT rootstocks grafted with MIR2111-5ox scions were very low or below detectable levels (Fig. 4f). To confirm whether or not shoot-derived mature miR2111s promote nodulation in roots, we designed a short tandem target mimic (STTM) construct 48 to prevent the accumulation of mature miR2111s (proUBQ: STTM2111) and generated stable proUBQ:STTM2111 transgenic lines (STTM2111). In leaves of STTM2111, the abundance of   miR2111s was reduced by less than half of that in WT plants, whereas the levels of MIR2111-5 were not influenced, indicating that STTM2111 successfully inhibits mature miR2111s ( Supplementary Fig. 6). Next, we carried out reciprocal grafting experiments between STTM2111 and WT plants. STTM2111 scions significantly reduced nodule numbers in the WT rootstocks to those equivalent to STTM2111 rootstocks (Fig. 4g). Thus, the amount of mature miR2111s generated in shoots is crucial to increasing nodule numbers in the roots.
The deletion of MIR2111-5 significantly decreased the nodule and infection thread number compared with the WT (Fig. 5b). Furthermore, the abundance of mature miR2111s in these leaves and roots was reduced to less than half of that seen in the WT (Fig. 5c). TML mRNA levels were not significant but tended to be higher in roots of mir2111-5 mutants than those of the WT (Fig. 5c). These results suggested that MIR2111-5 locus is necessary to increase the amount of mature miR2111s in leaves and roots and the number of nodules in roots.

Discussion
Land plants adapt to environmental stimuli, such as soil nutrient deprivation at the whole-plant level by shoot-mediated signaling communication via vascular bundles 3 . In this study, we investigated the functions of miR2111 genes in the shoot-mediated nodulation control system known as AON, focusing, in particular, on MIR2111-5 locus. L. japonicus possesses at least seven miR2111 genes, including those we identified and four genes expressed in leaves ( Fig. 1 and Supplementary Fig. 1a, b). Grafting experiments demonstrated that the accumulation of mature miR2111s in shoots was sufficient to increase root nodule numbers and reduce the accumulation of their targets, TML mRNA, in roots (Fig. 4). Moreover, we found that MIR2111-5 locus, which was expressed mainly in leaves ( Fig. 3 and Supplementary Fig. 3), was required for the regulation of nodule number and accumulation of mature miR2111s in roots (Fig. 5). Thus, mature miR2111s accumulation in leaves through MIR2111-5 expression is necessary to systemically optimize nodule number in a HAR1dependent manner.
Four miR2111 genes, MIR2111-2, MIR2111-4, MIR2111-5, and MIR2111-7, were expressed in leaves and repressed after rhizobial inoculation in a HAR1-dependent manner, whereas expression of the others was not detected by our RNA-seq analyses ( Fig. 1 and Supplementary Fig. 2). All of these miR2111 genes have complementary sequences to TML mRNA. Therefore, it is assumed that miR2111 genes that were expressed in leaves have functional redundancies and potentially work in the AON pathway (Fig. 1). Indeed, overexpression of MIR2111-5 and MIR2111-2, of which the sequence is 96% identical to that of MIR2111-7, significantly increased nodule numbers. In contrast, overexpression of MIR2111-4 did not affect nodule numbers; nonetheless, mature miR2111s accumulated at high levels in transformed roots (Fig. 2). This may have been due to the atypical secondary structure of miR2111-F derived from MIR2111-4 ( Supplementary  Fig. 1b). Unlike the hairpin structures of MIR2111-2 12 and MIR2111-5, the mature miR2111 sequence in the predicted miR2111-F secondary structure is located at the 5′ end of its stem ( Supplementary Fig. 1b). Nucleotide lengths proximal to the miRNA/miRNA* duplex affect the efficiency of correct miRNA processing in Arabidopsis 43-45 . Although we detected an increase of mature miR2111s in roots overexpressing MIR2111-4, miRNA generated from miR2111-F might not have nucleotide sequence patterns that are sufficient to target TML mRNA (Fig. 2). This notion is supported by the lack of a significant change in TML mRNA accumulation in roots overexpressing MIR2111-4. Although overexpression of MIR2111-3 increased nodule numbers in the previous study 12 , expression of MIR2111-3 was undetectable levels in our RNA-seq of L. japonicus leaves ( Supplementary Fig. 2). MIR2111-3 may be expressed under conditions different from those where the other four loci expressed.
It has been proposed that shoot-derived mature miR2111s are translocated to roots via the phloem to enhance nodulation 12 . However, it was unknown whether or not the accumulation of mature miR2111s in shoots leads to an increase in root nodule numbers. We carried out grafting experiments and demonstrated that WT rootstocks grafted with MIR2111-5 overexpressed scions exhibited increased root nodule numbers and mature miR2111 levels and repressed TML mRNA levels ( Fig. 4a-f). Since the levels of MIR2111-5 were not increased in WT rootstocks, it is unlikely that the increment in mature miR2111s in rootstocks was due to the movement of MIR2111-5 from shoots. These results indicated that MIR2111-5 overexpression in the shoot was sufficient to systemically regulate root nodule numbers and TML mRNA levels. Furthermore, results from grafting between MIR2111-5ox and AON mutants, har1-7 and tml-4, suggested that MIR2111-5 functions in the same genetic pathway as these AON factors (Supplementary Fig. 5). In contrast to MIR2111-5ox, scions constitutively expressing proUBQ:STTM2111 inhibited the accumulation of mature miR2111s (Supplementary Fig. 6) and decreased nodule numbers in WT rootstocks (Fig. 4f). Thus, the shoot-accumulating mature miR2111s influences root nodule numbers in a dose-dependent manner.
Among the seven miR2111 loci, MIR2111-5 showed the highest level of TPM in L. japonicus leaves (Fig. 1). When overexpressing in roots, MIR2111-5 was the most effective at increasing nodule numbers among miR2111 genes we examined (Fig. 2). These results implied that MIR2111-5 might significantly contribute to the regulation of nodule numbers in L. japonicus. Knockout lines of MIR2111-5 support this notion. Even though there are multiple Fig. 2 Overexpression of MIR2111-2 and MIR2111-5 in hairy roots promoted nodulation. a-d Nodulation at 21 days after inoculation (dai) on hairy roots transformed with an empty vector (a), proUBQ:MIR2111-2 (b), proUBQ:MIR2111-4 (c), and proUBQ:MIR2111-5 (d). Bright images (upper) and corresponding fluorescence images as a GFP transformation marker and DsRed constitutively expressing in M. loti (lower). Scale bars: 2 mm. e Nodule numbers in hairy roots transformed with an empty vector, proUBQ:MIR2111-2, proUBQ:MIR2111-4, and proUBQ1:MIR2111-5 (21 dai). Bars indicate mean ± standard deviation. Different letters indicate significant differences (P < 0.05) from Tukey's honestly significant difference test. f qRT-PCR analyses of mature miR2111 and TML in transformed hairy roots (5 dai). n = 4 individual biological replicates for each treatment. Bars indicate mean values. Two-sided Student's t test was used to determine statistical difference compared with empty vector control: *P < 0.05; ** P < 0.01; n.s. not significant. e, f Scatterplots show individual biological replicates as dots. loci for miR2111 genes in the L. japonicus genome ( Fig. 1 and Supplementary Fig. 1a), as for other miRNA gene families in plants 33 , a single mir2111-5 mutation decreased miR2111 levels in leaves and roots to less than half of WT (Fig. 5b). In addition, the MIR2111-5 deletions resulted in the significant inhibition of nodule formation. As for MIR2111-3 expression patterns 12 , pro-MIR2111-5:GUS expression was observed in the phloem of leaves (Fig. 3a, c, d) but not in stems, roots, or nodules ( Supplementary  Fig. 3). The accumulation of MIR2111-5 was hardly detected in roots using qRT-PCR (Fig. 3h). Considering this expression pattern, it is likely that the decrease in nodule numbers and mature miR2111 levels in roots of MIR2111-5 knockout lines was caused by the attenuation of mature miR2111 levels in leaves (Fig. 5c). The expression in the phloem, which is convenient to explain the shoot-to-root movement of mature miR2111s through the vasculature, was diminished after rhizobial inoculation (Fig. 3b). Thus, MIR2111-5 accumulation in leaves is regulated at the transcriptional level. The decrease in MIR2111-5 Fig. 4 Reciprocal grafting between wild-type and either MIR2111-5 overexpression or STTM2111 plants. a-e Nodulation at 28 days after inoculation (dai) in wild-type (a, b) and MIR2111-5ox (c, d) rootstocks grafted with wild-type (a, c) and MIR2111-5ox (b, d) scions.    expression in leaves in response to rhizobial inoculation correlated well with the accumulation patterns of mature miR2111s in leaves and roots (Fig. 3). The HAR1-dependent transcriptional regulation of miR2111 genes including MIR2111-5 in leaves would systemically influence nodule numbers in roots. Taken together with the results of grafting experiments, it is likely that mature miR2111s generated in shoots are translocated to roots and post-transcriptionally silence TML mRNA. Phloem sap contains a defined subset of miRNA species, and some of these specifically accumulate under various nutrientlimiting conditions [49][50][51] . In Arabidopsis and tobacco, shoot-toroot mobile miR399s, which accumulate in shoots in response to the deprivation of inorganic phosphate (Pi), posttranscriptionally regulate mRNA for PHO2, which encodes ubiquitin-conjugating E2 enzyme 24 10,49 . Since PHO2 suppresses Pi uptake in roots, shoot-derived miR399s maintain Pi homeostasis through the regulation of their target mRNA levels. Interestingly, Pi-limitation induces the accumulation of mature miR2111 in rapeseed phloem sap 50,51 . Arabidopsis and tobacco miR2111s, which also accumulate in response to Pilimitation 50,52 , are predicted to target mRNA for TML orthologues 52 . In legumes, the control of miR2111 production depends on nitrate availability 12,32 . Recently, it has been reported that in M. truncatula, nitrogen-deprivation in roots induces the production of mature miR2111s in shoots through CEP family peptide-mediated long-distance signaling, in order to enhance nodulation 32 . Although the biological functions of TML orthologues and roles of miR2111s in Arabidopsis and tobacco have not yet been identified, shoot-derived miRNA accumulation in roots may be a widespread mechanism of adaptation to nutritional changes in plants.

Methods
Plant materials and bacterial resources. We used L. japonicus accession MG-20 53 as the WT. har1-7 29 and tml-4 30 mutants were derived from MG-20. For the nodulation and infection phenotyping assay, plants were transplanted on autoclaved vermiculite supplemented with B&D containing 0.5 mM KNO 3 and inoculated with M. loti carrying dsRED. Plants were inoculated 3 days after being transplanted and harvested within indicated days after inoculation (dai) to analyze the nodulation and infection phenotypes.
For the RNA-seq, qPCR assay, and promoter GUS assay, plants were transplanted into autoclaved vermiculite containing B&D without a nitrogen source. Plants were inoculated either with or without M. loti carrying dsRED at indicated time points and harvested 14 days after germination.
For the grafting assay, plants were sown on vertical 0.9% agar plates for 2 days in darkness and 2 days under light/dark conditions. The 4-day-old seedlings were cut at the basal hypocotyls and scions were grafted on rootstocks, as described previously 29 . Grafted plants were grown horizontally on the sterilized waterabsorbed filter paper for 1 week and then transplanted to autoclaved vermiculite containing B&D either with or without 0.5 mM KNO 3 for the nodulation assay and qRT-PCR assay, respectively. Seven days after being transplanted, grafted plants were inoculated with M. loti carrying dsRED and harvested at the indicated time after inoculation.
Plasmid construction. All primers, oligonucleotides, and the synthetic DNA fragment are detailed in Supplementary Table 2.
For the overexpression analyses of MIR2111-2, MIR2111-4, and MIR2111-5, plasmids were constructed as follows. DNA fragments fully covering pri-miR2111 regions, as detected by RNA-seq were amplified from L. japonicus ecotype MG-20 genomic DNA by specific primer sets using the PrimeSTAR MAX DNA polymerase (Takara) and cloned into EcoRI-digested pENTR-1A (Invitrogen) using an In-Fusion HD Cloning Kit (Clontech). For hairy root transformation, all inserts were transferred to pUb-GW-GFP 47 by LR clonase (Invitrogen). The insert of MIR2111-5 fragment was also transferred to pUb-GW-Hyg 47 by LR clonase for use in whole-plant transformation.
For the GUS reporter assay of MIR2111-5, a 3.0-kb DNA fragment upstream of MIR2111-5 was amplified from L. japonicus ecotype MG-20 genomic DNA using specific primer sets by means of the PrimeSTAR MAX DNA polymerase (Takara) and cloned into EcoRI-digested pENTR-1A (Invitrogen) using an In-Fusion HD Cloning Kit (Clontech). The insert was transferred to pMDC162 55 by LR clonase.
For CRISPR/Cas9 constructs of MIR2111-5, targeting sites were designed using the CRISPR-P 2.0 program 56 . Two gRNAs were used for each construct to delete a nearby miR2111a sequence in MIR2111-5 on the L. japonicus genome. In order to create CRISPR/Cas9 constructs containing dual gRNA, we first modified the gRNA cloning vector pMR203 (provided by Dr. Mily Ron and Dr. Masaki Endo) as follows. DNA fragments of U6-26 promoter/BbsI/Chimera RNA (sgRNA) were amplified by two types of specific primers with multiple cloning sites (MCS) using the PrimeSTAR GXL DNA polymerase (Takara) from pMR203. Each DNA fragment was inserted between the ApaI and PstI sites of pMR203 to create new cloning vectors with different MCS, designated as pMR203_AB and pMR203_BC, which enable removal of a gRNA expression cassette after cloning of gRNA using standard restriction enzymes. The binary vector pMR285 (derived from pDe-Cas9) 57 was also modified as follows. Two oligo DNAs, forward and reverse of pMR285_oligo (Supplementary Table 2), were annealed and inserted into the BbsI site of pMR203 to create MCS. The DNA fragment of this MCS was removed by PstI and inserted into the PstI site of pMR285 to create new binary vectors with MCS, designated as pMR285_AD. Then, oligonucleotide pairs for gRNA were annealed at 95°C for 5 min and inserted into the BbsI site of pMR203_AB and pMR203_BC. The gRNA expression cassettes in pMR203_AB and pMR203_BC were removed using BamHI/SalI and SalI/XbaI, respectively. These fragments were inserted simultaneously between the BamHI and XbaI sites of pMR285_AD.
Hairy root transformation of L. japonicus roots was performed using Agrobacterium rhizogenes-mediated transformation, as described previously 58 . A. rhizogenes AR1193 strains carrying the plasmid of interest were grown on YEB medium containing 1.5% agar supplemented with appropriate antibiotics at 28°C for 2 days. A. rhizogenes AR1193 strains were collected and suspended in 6 mL of sterilized water. The seedlings of L. japonicus were grown on a 0.9% agar medium containing B&D for a total of 3 days (darkness for 2 days and 16 h of light, 8 h of darkness for 1 day). The seedlings were placed in the A. rhizogenes suspension and cut in the middle of their hypocotyls. The seedlings without roots were placed on a co-cultivation medium for hairy root transformation (1/2 Gamborg's B5 salt mixture, 0.01% sucrose, 0.9% agar, pH 5.5) for 3 days (24°C darkness). Then, they were transferred onto vertical hairy root induction medium (Gamborg's B5 salt mixture, Gamborg's vitamin solution, 1% sucrose, 12.5 μg mL −1 meropenem, 0.9% agar, pH 5.5) and grown for 10 days at 24°C (16 h of light and 8 h of darkness). The plants with transformed roots displaying fluorescence of a GFP transformation marker were selected and used for further analyses.
Microscopic observation. An SZX16 stereomicroscope or a BX50 microscope (Olympus) was used to observe roots, nodules, and GUS-stained whole plants. Nodules with the neck at the basal region were counted in nodulation phenotype assays. For plastic sections, GUS-stained leaves, stems, or nodules were fixed with Formalin-Acetic-Alcohol buffer for 12 h at 4°C and embedded in Technovit 7100 resin (Haraeus Kulzer). Sections were cut with a microtome RM2255 (Leica) at a thickness of 5 μm and counterstained with 0.1% safranin for 10 min at 55°C.
Library preparation and sequencing. The total RNA was extracted from true leaves of L. japonicus grown under the conditions described using PureLink Plant RNA Reagent (Thermo Fisher Scientific) and purified using RNeasy plant mini kit (Qiagen). The quality of RNA was evaluated using an Agilent 2100 Bioanalyzer with Agilent RNA 6000 Nano kit. RNA-seq libraries were prepared using NEBNext Ultra II RNA Library Prep Kit for Illumina (New England Biolabs) according to the manufacturer's protocol. The prepared RNA-seq libraries were qualified by the Agilent 2100 Bioanalyzer using Agilent High Sensitivity DNA kit and quantified using a KAPA Library Quant Kit (Kapa Biosystems). Each library was diluted to 2 nM. Libraries were sequenced using NextSeq 550 (Illumina) and generated 81 bp single-end reads.
Bioinformatic analysis. All acquired RNA-seq reads were qualified by FastQC (ver. 0.11.7) and adapter trimmed by Trimomatic 59 (ver. 0.36, options: CROP:80 LEADING:30 TRAILING:30 SLIDINGWINDOW:4:15 HEADCROP:6 MIN-LEN:36). Trimmed reads were mapped to the L. japonicus genome (Gifu v1.2) using Hisat2 60 (ver. 2.1.0). The mapped reads with a MAPQ score of <30 were filtered out using SAMtools 61 (v1.9, options: view -bq 30) to handle only high mapping-quality scored reads for downstream analysis. The number of raw reads, trimmed reads, and mapped reads are shown in Supplementary Table 3. Filtered mapped reads were assembled and calculated TPM values using Stringtie 46  In order to predict new miR2111 loci, we searched miR2111a, b, and c sequence from new L. japonicus genome assembly (Gifu v1.2) by BLASTn 41 . We identified a total of ten positions for miR2111a, miR2111b, or miR2111c from the L. japonicus genome. In all, 100 bp-upstream and -downstream sequences from the identified 21-bp miR2111 regions were extracted, and the secondary structure was predicted using the minimum free energy (MFE) algorithm of RNAstructure program 42 (ver. 6.1). Next, we searched for pri-miR2111s expressing in L. japonicus using mapped our RNA-seq data. In order to detect pri-miR2111 expression from our RNA-seq data, RNA-Seq-based gene predictions were performed using Stringtie (ver. 1.3.4d) with the default setting. We then searched for pri-miR2111s from predicted potential L. japonicus genes based on our RNA-seq data by referring to the position of the miR2111 hairpin sequences, including those we predicted. Furthermore, we searched new potential miR2111 genes that were not detected in our RNA-seq data using BLASTn search of sequences for all miR2111 genes against the L. japonicus genome and identified one new potential miR2111 gene. Secondary structures of all new miR2111s were re-estimated using the full-length sequence of the predicted miR2111 genes using the MFE algorithm of RNAstructure (ver. 6.1). The genomic positions of miR2111s were visualized by R software (ver. 3.6.2) with chromoMap package 64 (ver. 0.2).
qRT-PCR analysis. The total RNA including miRNA was extracted from either roots or true leaves of plants grown under the conditions described using NucleoSpine miRNA (MACHEREY-NAGEL Inc). For qRT-PCR of mRNA and miR2111 precursors, extracted RNA was reverse-transcribed using PrimeScript RT reagent Kit (Perfect Real Time) (Takara). Mature miR2111s were reversetranscribed and adapter-ligated using the Mir-X miRNA First-Strand Synthesis Kit (Takara). Mature miR2111-specific primer and adapter-specific primer contained in Mir-X miRNA First-Strand Synthesis Kit were used for the amplification of mature miR2111s. qRT-PCR was carried out by LightCycler 96 (Roche Applied Science) using TB Green Advantage qPCR Premix (Takara). Ubiquitin and ATP synthase were used for normalization of expression levels. Primers of Ubiquitin 69 and ATP synthase 12 were synthesized as described previously. All primers are listed in Supplementary Table 2.