Endoreduplication-mediated initiation of symbiotic organ development in Lotus japonicus

In response to appropriate inductive conditions, plants have the capacity to form new organs from differentiated cells. Root nodulation is a form of such de novo organogenesis that occurs predominantly in leguminous plants (Crespi and Frugier, 2008). During early nodule development, a rhizobia-derived nodulation factor induces the dedifferentiation of root cortical cells. The activated cortical cells then proliferate to form the primordium of the symbiotic nitrogen-fixing root nodule (Yang et al., 1994; Geurts and Bisseling, 2002; Oldroyd et al., 2011). Recent identification and functional analyses of the putative cytokinin receptors Lotus japonicus LOTUS HISTIDINE KINASE 1 (LHK1) and Medicago truncatula CYTOKININ RESPONSE 1 have led to a greater understanding of how the activation of cytokinin signaling is crucial to the initiation of nodule organogenesis (Gonzalez-Rizzo et al., 2006; Murray et al., 2007; Tirichine et al., 2007). In particular, it has been shown that in the L. japonicus spontaneous nodule formation 2 (snf2) mutant, which possesses a gain-of-function form of LHK1, confers the constitutive activation of cytokinin signaling, resulting in the formation of spontaneous nodule-like structures in the absence of rhizobia (Tirichine et al., 2007).

Normal nodulation is achieved by interactive processes involving infection by rhizobia and nodule organogenesis (Madsen et al., 2010); the complexity of these interactions has made it difficult to study the regulation of these mechanisms separately. The use of spontaneous nodules, however, appears to be an efficient approach for focusing on nodule organogenesis because the effect of the rhizobial infection process can be excluded from spontaneous nodule development (Tirichine et al., 2006). Spontaneous nodule formation is suppressed by the mutation of any of the three putative transcription factors NODULE INCEPTION, NODULATION SIGNALING PATHWAY 1 (NSP1) and NSP2; these factors are involved in nodule organogenesis (Tirichine et al., 2007). But our understanding of the nodule initiation process is still limited because of the limited number of genes that have been identified.

To investigate the genetic mechanism involved in the onset of nodule organogenesis, we performed a screen for genetic suppressors of the snf2 spontaneous nodulation phenotype. A recessive semi-dwarf mutant named vagrant infection thread 1 (vag1) was identified, and the number of spontaneous nodules reduced in the vag1 snf2 double mutant compared with the snf2 single mutant (Fig. 1A). The number of spontaneous nodules of snf2 increased in the presence of the hypernodulating hypernodulation aberrant root formation 1 (har1) mutation as previously shown (Tirichine et al., 2007), but the enhanced spontaneous nodulation was strikingly suppressed by the vag1 mutation (Fig. 1A). In the vag1 single mutant, the initiation of nodule organogenesis was severely attenuated after inoculation of its rhizobial symbiont Mesorhizobium loti (Fig. 1B,C), without substantially altering overall root architecture (supplementary material Fig. S1A,B). The vag1 mutant appeared to respond normally to cytokinin, as shoot and root growth of the mutant, as well as of the wild type (WT), were inhibited in the presence of high cytokinin concentrations (supplementary material Fig. S2), suggesting that the vag1 mutant acts as a suppressor of snf2 without affecting cytokinin signaling in shoot and root growth.

During nodule development, rhizobia invade host cortical cells via a specialized structure called the infection thread (IT), which originates in a root hair cell (Murray, 2011). In WT, ITs penetrated the root hair and ramified within inner cortical cells (Fig. 1D; supplementary material Fig. S3). By contrast, ITs that formed in the vag1 mutant appeared to penetrate root hair cells normally but failed to ramify and reach cortical cells (Fig. 1E,F; supplementary material Fig. S3), suggesting that the mutant infection threads are blocked at the epidermal-cortical interface. Normal nodule development is achieved by interactive processes involving nodule organogenesis and IT formation (Madsen et al., 2010). To examine the two processes separately, we focused on the cyclops mutant, in which, despite a premature arrest of IT elongation in root hairs, the initiation of nodule organogenesis is nonetheless induced (Yano et al., 2008). In the vag1 cyclops-6 double mutant, although ITs showed the cyclops-type premature arrest phenotype, the initiation of nodule organogenesis was significantly suppressed (supplementary material Fig. S4), suggesting that the impairment of nodule initiation in the vag1 mutant is due to a defect in the nodule organogenesis process rather than its aberrant IT progression.

Map-based cloning identified a gene that is responsible for the vag1 mutation (supplementary material Fig. S5). The vag1 mutation carries a G-to-A nucleotide substitution in the splice donor site of intron 9 in the gene, causing intron mis-splicing (Fig. 2A-C). The mutant nodulation phenotype of vag1 was rescued when a genomic fragment containing the wild-type gene was introduced into the mutant by hairy root transformation (Fig. 2D-F). Moreover, we isolated another mutant carrying a mutation in the VAG1 gene (hereafter we denote the original and the new vag1 mutant vag1-1 and vag1-2, respectively). In vag1-2, there is one base deletion in exon 1 producing the emergence of a considerably premature stop codon (Fig. 2A,B). Nodule initiation was completely compromised in the vag1-2 mutant (Fig. 1B). The expression of VAG1 seemed to be unaffected by M. loti infection (supplementary material Fig. S6A). To determine the spatial expression pattern of VAG1, we first identified a functional VAG1 promoter that could rescue the vag1 mutation when VAG1-coding sequence was expressed under the control of the promoter (Fig. 2F). Reporter gene analysis using the promoter (pVAG1::GFP-NLS) showed that VAG1 was expressed throughout the root, including proliferating cortical cells, to form nodules (supplementary material Fig. S6B-E).

VAG1 encodes a protein that is putatively orthologous to Arabidopsis ROOT HAIRLESS 1 (RHL1)/HYPOCOTYL 7 (HYP7) (Schneider et al., 1998; Sugimoto-Shirasu et al., 2005), a component of the plant DNA topoisomerase VI. In Arabidopsis, loss-of-function mutants of factors constituting topoisomerase VI have general defects in ploidy-dependent cell growth, and the number of highly endoreduplicated cells is reduced in the mutants (Hartung et al., 2002; Sugimoto-Shirasu et al., 2002, 2005; Yin et al., 2002), indicating that topoisomerase VI has a function as a positive regulator of endoreduplication. Although the rhl1 mutation prevents the formation of root hairs, they were formed normally in the vag1 mutants (supplementary material Fig. S1C,D). RHL1 could rescue the vag1 mutation when it was expressed under the control of the VAG1 promoter (Fig. 2F), suggesting that the VAG1 protein has a function similar to RHL1.

As represented by trichomes of the leaf epidermis in Arabidopsis and by endosperm of the embryo in maize, it is believed that endoreduplication contributes to increasing plant cell sizes, in particular to producing terminally differentiated cells (Lee et al., 2009). In legume-Rhizobium symbiosis, rhizobial-infected cells of nodules are terminally differentiated as a consequence of endoreduplication (Kondorosi and Kondorosi, 2004). To assess the involvement of VAG1 in controlling endoreduplication, we next investigated the nodule phenotype of vag1. The vag1-1 mutant could form nodules, although their numbers were small (Fig. 1B). The size of the nodules formed in the mutant was largely indistinguishable from those of the WT (Fig. 3A,B). An observation of vag1-1 mutant nodule sections showed that the sizes of cells located in the inner region of nodules were smaller and the numbers of potential rhizobia-colonized infected cells were reduced compared with the WT (Fig. 3C-F). In addition, the inner region of vag1-1 nodules comprised large numbers of small rhizobia-infected (as yet uncolonized) cells that resembled those located at surrounding regions of rhizobia-colonized infected cells in WT (supplementary material Fig. S7). In wild-type nodules, flow cytometry revealed that few diploid cells were detected, and cells underwent up to four rounds of endoreduplication to reach 32C (Fig. 3G,I). By contrast, vag1-1 nodules comprised increased proportions of diploid cells, whereas the proportions of endoreduplicated cells (>4C) were significantly reduced (Fig. 3H,I). The ploidy level of uninoculated vag1 roots was indistinguishable from that of WT (supplementary material Fig. S1E-G).

An observation of the attenuation of nodule initiation in the vag1 mutant and VAG1 implication in endoreduplication of infected cells within nodules led us to postulate that VAG1-mediated endoreduplication is required for the onset of nodule organogenesis. We previously produced transgenic L. japonicus plants, in which nuclear localized GFP is expressed under the control of a synthetic auxin-responsive element, DR5 (Suzaki et al., 2012). Use of these plants enabled us to monitor the nuclei of cortical cells relevant to the nodulation cell lineage. As a result, we observed that nuclei of the outermost cortical cells immediately before initiation of division were larger than those of the surrounding cortical cells in WT (Fig. 4A). The enlarged nuclei were continuously observed after the initiation of cortical cell proliferation (Fig. 4B,C). In addition, ITs appeared to elongate towards the cortical cells with the enlarged nuclei (Fig. 4A-C). By contrast, in the vag1-1 mutants there were no enlarged nuclei in cortical cells beneath root hairs with ITs and no cortical cell division was induced (Fig. 4D,E), despite the induction of auxin response (supplementary material Fig. S8). As DNA size is reflected in nuclear size (Bourdon et al., 2011; Iwata et al., 2011), we next estimated the DNA size of the enlarged nuclei by quantifying the size of DAPI-labeled nuclei. In WT, the size of the enlarged nuclei was more than 8-fold larger compared with those of root cap cells (2C control) (supplementary material Fig. S9), suggesting that their DNA size reaches up to at least 16C. In the vag1 mutants, the size of nuclei in cortical cells beneath root hairs with IT was largely comparable to that of cortical cells of uninoculated roots (supplementary material Fig. S9).

In plants, CCS52A isoforms are necessary and sufficient for the progress of endocycling (Cebolla et al., 1999; Larson-Rabin et al., 2009; Mathieu-Rivet et al., 2010; Baloban et al., 2013), which skips the mitotic phase and re-enters the S phase without cytokinesis, resulting in the duplication of the nuclear DNA content. In M. truncatula, MtCCS52A is shown to be expressed before endoreduplication and to regulate endoreduplication of infected cells within nodules (Vinardell et al., 2003). We isolated two L. japonicus CCS52A1 genes (LjCCS52A1-LIKE1 and LjCCS52A1-LIKE2) with high similarity to Arabidopsis CCS52A1. Promoter-GUS reporter analysis and in situ hybridization revealed that the two genes were expressed within nodules (supplementary material Fig. S10). The genes were also expressed in dividing cortical cells during early nodule development (Fig. 4F-I), indicating that the progress of endocycling could occur during the nodule-initiation process.

In this study, we have shown that topoisomerase VI complex containing VAG1 has an essential role in regulating the onset of nodule development. Based on the results, we propose a hypothesis in which VAG1-mediated endoreduplication of cortical cells may be required for the initiation of nodule organogenesis (supplementary material Fig. S11). The primary defect of vag1 could be endoreduplication preceding cortical cell division, which could lead to an attenuation of the phase of the cortical cell proliferation. Thus, we cannot examine the direct effect of the vag1 mutation on the phase of the cell cycle related to the cortical cell proliferation, or entirely rule out the possibility that a more general defect in the cell cycle could be responsible for the vag1 nodulation phenotype. During the nodule initiation process, the cortical cell sizes with enlarged nuclei are almost identical to adjacent cortical cells. This observation implies that the endoreduplication that occurs in this developmental context is unrelated to cell growth process, although its biological significance remains elusive. Given that change of ploidy level can be associated with regulation of gene expression involving epigenetic gene silencing, as previously shown (Comai et al., 2000; Wang et al., 2004; Yu et al., 2010), it is possible that the endoreduplication of cortical cells may be involved in silencing of genes that determine the state of differentiated cortical cells. This cellular-level epigenetic control may trigger the activation of sets of genes relevant to the nodule development. In order to identify a key molecular mechanism involved in the cell fate conversion, the elucidation of the detailed role of endoreduplication of cortical cells is required.

In addition, we have shown that ITs penetrate host cells in the direction of the endoreduplicated cortical cells. Loss of such endoreduplicated cells in the vag1 mutants causes a misguided elongation of ITs. As the phenotype of misguided infection is reminiscent of hyperinfected 1, a loss-of-function mutant of LHK1 (Murray et al., 2007), the onset of endoreduplication may occur downstream of the cytokinin response. Overall, during root nodule symbiosis, the endoreduplication of cortical cells may have another important role of acting as a determinant to guide rhizobia to host cells (supplementary material Fig. S11).

In Arabidopsis, topoisomerase VI is generally involved in diverse cell growth processes that are accompanied by increased DNA levels. It is known that the pattern of ploidy-dependent cell growth is different among plant species; this may account for the observations that non-symbiotic shoot and root development is largely normal in the vag1 mutants. Currently, the detailed molecular mechanism that links topoisomerase VI to endoreduplication remains unknown. We show that the expression patterns of VAG1 and LjCCS52A1-LIKE genes overlap during nodule development. Further investigation of the potential interactions among the genes will be needed to elucidate how topoisomerase VI achieves progress of the endocycle.

The Miyakojima MG-20 ecotype of L. japonicus was used as WT in this study. The vag1-1 mutant was isolated from the M₂ generation of snf2 plants that had been mutagenized with ethylmethane sulfonate (EMS) (Miyazawa et al., 2010; Suzaki et al., 2013). The vag1-2 mutants were isolated by a screening of nodulation-deficient mutants using EMS-treated MG-20 plants, which were provided by Legume Base. DR5::GFP-NLS/vag1-1 plants were produced by crossing DR5::GFP-NLS transgenic plants (Suzaki et al., 2012) with vag1-1 plants. For the analyses of rhizobial-induced nodulation or spontaneous nodulation, plants were grown with or without M. loti MAFF 303099, respectively, as previously described (Suzaki et al., 2013).

The primers used for PCR are listed in supplementary material Table S1. For the complementation analysis, a 10.9 kb genomic DNA fragment including the VAG1 candidate gene was amplified by PCR from wild-type genomic DNA. This fragment included 3.0 kb of sequence directly upstream of the initiation codon, and was cloned into pCAMBIA1300-GFP-LjLTI6b (Suzaki et al., 2012). For the VAG1 expression analysis, a gateway-cassette (GW) fragment was cloned into pCAMBIA1300-GFP-LjLTI6b to create the new binary vector pCAMBIA1300-GW-GFP-LjLTI6b. Next, 3.0 kb of sequence directly upstream of the initiation codon of VAG1 was amplified by PCR and cloned into pCAMBIA1300-GW-GFP-LjLTI6b to create the vector pCAMBIA1300-pVAG1-GW-GFP-LjLTI6b. VAG1- or RHL1-coding sequence (cds) was amplified by PCR from a template cDNA prepared from wild-type L. japonicus or Arabidopsis Col-0 plants, respectively, and cloned into a pENTR/D-TOPO vector (Invitrogen). The VAG1 and RHL1 cds in pENTR/D-TOPO and GFP-NLS in pDONR221 (Suzaki et al., 2012) were inserted downstream of the VAG1 promoter by LR recombination reactions. To create the pLjCCS52A1-LIKE2::GUS construct, 2.8 kb of sequence directly upstream of the initiation codon of LjCCS52A1-LIKE2 was amplified by PCR and cloned into pCAMBIA1300-GW-GFP-LjLTI6b to create the vector pCAMBIA1300-pLjCCS52A1-LIKE2-GW-GFP-LjLTI6b. The β-glucuronidase (GUS) gene in pENTR-gus (Invitrogen) was inserted downstream of the LjCCS52A1-LIKE2 promoter by the LR recombination reaction. The recombinant plasmids were introduced into Agrobacterium rhizogenes strain AR1193 and were transformed into roots of L. japonicus plants by a hairy-root transformation method previously described (Okamoto et al., 2009).

Nodules were removed from wild-type or vag1-1 roots, cut into pieces with razor blades in extraction buffer (Partec) and incubated for 2 min at room temperature. The suspension was filtered through a 30 µm Celltrix filter (Partec) and stained with staining buffer (Partec). Flow cytometry was performed using a Ploidy Analyzer PA (Partec). Data are shown as mean±s.e.m. of three biological replicates.

Sequence data from this report can be found in the GenBank/EMBL data libraries under the following accession numbers: VAG1, AB871650; RHL1, At1g48380; LjCCS52A1-LIKE1, AB871651; LjCCS52A1-LIKE2, AB871652; CCS52A1, At4g22910.

We thank Sachiko Tanaka, Michiko Ichikawa, Yuko Ogawa and Kiyoshi Tatematsu for technical support; and Makoto Hayashi for providing M. loti MAFF303099 expressing DsRED. This work was supported by NIBB Core Research Facilities and NIBB Model Plant Research Facility.