CERBERUS, a novel U-box protein containing WD-40 repeats, is required for formation of the infection thread and nodule development in the legume–Rhizobium symbiosis


*(fax +81 29 838 7417; e-mail umehara@nias.affrc.go.jp).


Endosymbiotic infection of legume plants by Rhizobium bacteria is initiated through infection threads (ITs) which are initiated within and penetrate from root hairs and deliver the endosymbionts into nodule cells. Despite recent progress in understanding the mutual recognition and early symbiotic signaling cascades in host legumes, the molecular mechanisms underlying bacterial infection processes and successive nodule organogenesis are still poorly understood. We isolated a novel symbiotic mutant of Lotus japonicus, cerberus, which shows defects in IT formation and nodule organogenesis. Map-based cloning of the causal gene allowed us to identify the CERBERUS gene, which encodes a novel protein containing a U-box domain and WD-40 repeats. CERBERUS expression was detected in the roots and nodules, and was enhanced after inoculation of Mesorhizobium loti. Strong expression was detected in developing nodule primordia and the infected zone of mature nodules. In cerberus mutants, Rhizobium colonized curled root hair tips, but hardly penetrated into root hair cells. The occasional ITs that were formed inside the root hair cells were mostly arrested within the epidermal cell layer. Nodule organogenesis was aborted prematurely, resulting in the formation of a large number of small bumps which contained no endosymbiotic bacteria. These phenotypic and genetic analyses, together with comparisons with other legume mutants with defects in IT formation, indicate that CERBERUS plays a critical role in the very early steps of IT formation as well as in growth and differentiation of nodules.


Legumes can utilize atmospheric dinitrogen in symbiosis with soil bacteria collectively called rhizobia. The symbiosis between legumes and rhizobia is established by mutual and complex interactions. The ‘nod inducers’, mainly flavonoids secreted from host plant roots, induce rhizobia to synthesize lipochitin–oligosaccharide signal molecules, Nod-factors (NFs), which in turn trigger host plant responses mediating bacterial infection and induction of nodule formation (Stougaard, 2000; Oldroyd and Downie, 2008).

Recent developments in molecular genetics using the model legumes Lotus japonicus and Medicago truncatula have enabled the identification of genes involved in the legume–Rhizobium symbiosis. NFR1 and NFR5 of L. japonicus (Radutoiu et al., 2003, 2007) and LYK3 and NFP of M. truncatula (Limpens et al., 2003; Arrighi et al., 2006; Smit et al., 2007) have been postulated as NF receptors, while SYMRK/DMI2 (Endre et al., 2002; Stracke et al., 2002), CASTOR and POLLUX/DMI1 (Ane et al., 2004; Imaizumi-Anraku et al., 2005), and CCaMK/DMI3 (Levy et al., 2004; Tirichine et al., 2006) have been identified as the components of the ‘common symbiosis pathway’ (CSP) which is required for both rhizobial and mycorrhizal symbiosis (Kistner and Parniske, 2002). NSP1 and NSP2 (Kalo et al., 2005; Smit et al., 2005; Heckmann et al., 2006; Murakami et al., 2006), NIN (Schauser et al., 1999; Marsh et al., 2007) and genes encoding putative ERF transcription factors (Middleton et al., 2007; Asamizu et al., 2008) are positioned in the ‘nodulation specific pathway’ just after the CSP. These genes are involved in the very early steps of symbiotic interactions such as NF recognition and the immediate downstream signaling networks which precede the invasion of rhizobia into root hair cells through infection threads (ITs) and induction of cortical cell division. In contrast, the host legume genes involved in IT formation and/or progression, and their coordination with nodule organogenesis, are still largely unknown.

A number of symbiotic mutants have been isolated which are able to initiate nodulation but show defects in IT formation and/or its development. These mutants are mostly characterized by the arrest of nodule organogenesis at the stage of small bumps. These bumps and/or nodule-like structures lack endosymbiotic bacteria and these mutants are thus grouped into the Hist mutant category (Kawaguchi et al., 2002), which is characterized by defects in nodule histogenesis and rhizobial infection processes. The Lotus crinkle and alb1 mutants are typical examples of Hist mutants. crinkle forms ITs inside root hairs but IT growth is mostly arrested in the epidermal cells. Cortical cell division is induced to form nodule primordia but nodule development is arrested at the stage of small bumps (Tansengco et al., 2003). The latter, alb1, forms aberrant nodule structures in which tissue differentiation, such as the development of vascular bundles, is incomplete (Imaizumi-Anraku et al., 1997, 2000). The alb1 ITs are very thick and are mostly aborted in root hair cells. When ITs penetrate into nodule tissue on rare occasions, Rhizobia are not released from ITs into the nodule cells. (Imaizumi-Anraku et al., 1997; Yano et al., 2006). The CRINKLE and ALB1 genes have not yet been reported. Another class of Lotus mutant, cyclops, also shows defects in IT formation together with abortion of nodulation at the bump stage (Yano et al., 2008). In contrast to crinkle and alb1, cyclops is also impaired in the symbiosis with mycorrhizal fungi and is thus suggested to be part of the CSP. The causal gene, CYCLOPS (an ortholog of Medicago IPD3), was recently identified and shown to encode a protein that interacts with calcium and calmodulin-dependent protein kinase (CCaMK); (Yano et al., 2008; Messinese et al., 2007). NIN has also been shown to be required for the initiation of ITs. nin mutants display excessive root hair deformation with an absence of ITs inside the root hair cells and are unable to induce cortical cell division, thus being different from the Hist mutants described above (Schauser et al., 1999).

In M. truncatula, BIT1/ERN1 and RPG were cloned and shown to be the host components essential for IT initiation, following activation of the early symbiotic signaling cascade(s). BIT1/ERN1 was proposed to be an ERF transcription factor and to be required for activation of early nodulin (ENOD) genes in the root epidermis in response to NFs (Middleton et al., 2007). RPG encodes a long coiled-coil protein in the nucleus, and is proposed to be involved in the process whereby rhizobia manage to dominate the process of induced tip growth for root hair infection (Arrighi et al., 2008). However, the exact functions of these genes in IT formation are still unclear. The M. truncatula mutants lin and api have been well characterized and they are also allocated to the ‘nodulation-specific pathway’ following the CSP, but their causal genes have not yet been reported (Kuppusamy et al., 2004; Teillet et al., 2008). An M. truncatula mutant, hcl, which has a defect in LYK3, an ortholog of LotusNFR1, displays Nod phenotypes, but the weak allele of the hcl mutants has been shown to exhibit abortion of IT growth on the curled root hair tips (Smit et al., 2007).

Here, we describe the characterization of a novel Hist mutant of L. japonicus, designated ‘cerberus’, which shows defects in very early steps of IT formation and forms small bumps after inoculation of Mesorhizobium loti. We cloned the causal gene, CERBERUS, which encodes a protein containing a U-box and WD-40 repeats. Based on the mutant phenotypes and the deduced structure of the CERBERUS protein, we discuss possible functions of the CERBERUS gene in the process of rhizobial infection and nodule development.


Isolation and growth of the cerberus mutants

The cerberus-5 (Ljsym101) mutant was first isolated from progeny of the plants regenerated from calli. cerberus-1 of the same complementation group was identified from a progeny of descendants after T-DNA transformation (Schauser et al., 1998), cerberus-2, -3 and -4 were from progenies of transformants carrying Ac launching constructs (Thykjaer et al., 1995) and cerberus-6 was found in an ethyl methanesulfonate (EMS) mutagenized population (Karas et al., 2005). Different from wild-type (WT) plants, the cerberus mutants formed small white bumps and displayed severe nitrogen deficiency symptoms, such as stunted stems, small chlorotic leaves and anthocyanin accumulation (Figure 1a–d). The growth of cerberus was fully restored when it was supplied with a sufficient amount of ammonium nitrate (Figure 1e), indicating that the growth retardation of cerberus was due to the lack of nodule nitrogen fixation. Mycorrhizal fungi could establish symbiosis normally with the cerberus-5 roots (data not shown). Segregation analysis of the F2 progeny after crossing cerberus-5 with WT B-129 Gifu plants resulted in a 102:33 segregation (WT/mutants, 3:1 χ2 = 0.022), indicating that the cerberus locus is monogenic and recessive.

Figure 1.

 Phenotypes of Lotus japonicus cerberus mutants.
(a) cerberus-5 (left) and wild-type (WT) Gifu (right) 28 days post-inoculation (dpi) with Mesorhizobium loti Tono.
(b) Nodules on WT Gifu roots 14 dpi.
(c, d) Bumps on cerberus-1 and -5 roots (14 dpi), respectively.
(e) Growth of cerberus-5 in comparison with WT plants in nitrogen-free conditions with inoculation of M. loti (left) and in the medium supplemented with 1 mm ammonium nitrate without M. loti inoculation (right).
Scale bars = 10 mm (a), and 1 mm (b–d).

Nodule development is impaired in the cerberus mutants

Internal structures of the bumps formed on cerberus roots were examined by light microscopy and compared with the WT nodules. Mature nodules formed on WT roots showed differentiation of the central infected zone surrounded by boundary layers and cortex in which vascular bundles were developed (Figure 2a). The infected cells were filled with bacteroids and contained a few vacuoles (Figure 2d). In contrast, cerberus-5 bumps did not show such tissue differentiation (Figure 2b,c). Occasionally, bifurcated vascular bundles were initiated towards the nodule bumps, but they were not fully developed (Figure 2c). The cells in the central zone of the cerberus-5 bumps were highly vacuolated and contained no bacteroids (Figure 2e). Almost all the cells in the cerberus-5 bumps contained large numbers of small amyloplasts, while in WT nodules large amyloplasts were only found in the uninfected cells (Figure 2d).

Figure 2.

 Micrographs of nodule and bump sections.
(a) Wild-type (WT) Gifu nodule 16 days post-inoculation (dpi).
(b, c) cerberus-5 bumps 16 (b) and 21 (c) dpi. Arrows in (c) indicate the initiation of vascular strands towards the nodule bump.
(d) Wild-type nodule infected zone 16 dpi.
(e) Cells in the central zone of cerberus-5 bump 16 dpi.
V and arrowheads in (d) and (e) indicate vacuoles and amyloplasts, respectively.
Scale bars = 100 μm (a–c) and 10 μm (d, e).

Initiation and growth of ITs are arrested in the cerberus roots

Infection events were visualized in the roots of WT and cerberus-1 and -5 seedlings by inoculation with DsRed- or LacZ-labelled M. loti (Figure 3). In WT roots, ITs were formed within curled root hair tips, grew through the base of the root hair cells (Figure 3a) and further branched into the root cortex and then into developing nodule tissues (Figure 3g). In contrast, rhizobia colonizing the infection pocket of curled root hair tips of cerberus mostly failed to enter into the root hairs (Figure 3b,d). Occasionally a few rhizobia were observed inside root hairs without forming a tight IT structures (Figure 3e). Though ITs were sometimes seen inside the short root hairs, their growth was aborted in root hair cells and they did not penetrate into root cortical cells (Figure 3c,f). As a consequence, the bumps formed on cerberus contained no endosymbiotic rhizobia (Figures 3h and 2e). The other cerberus alleles, cerberus-2 and -3, also showed bump formation (data not shown) and the infection phenotypes were essentially the same as cerberus-1 and -5 (Figure S1 in Supporting information).

Figure 3.

 Infection phenotype of cerberus-1 and -5.
(a) Infection threads (ITs) in wild-type (WT) root hairs (7 dpi).
(b) Microcolonies on curled root hairs of cerberus-1 (14 dpi).
(c) Infection thread formed inside short root hairs of cerberus-1 (14 dpi).
(d, e) Microcolonies on curled root hairs of cerberus-5 (7 dpi).
(f) Infection thread formed inside short root hairs of cerberus-5 (9 dpi).
(g) Infection threads in WT plants branching (shown by an arrow) into the central zone of developing nodules (14 dpi).
(h) A bump formed on cerberus-5 at 14 dpi. Microcolonies of Rhizobia are indicated by arrows.
Infection threads were visualized by DsRed-labeled (a–f) or LacZ-labelled Mesorhizobium loti MAFF303099 (g, h). Scale bars = 20 μm (a–f), 50 μm (g) and 100 μm (h).

Phenotypic comparisons with other infection mutants

To examine the symbiotic phenotype of cerberus in more detail, the frequency of infection events and the number of nodules and nodule primordia (bumps) of cerberus-5 were determined after inoculation with LacZ-labelled M. loti. For comparison, WT and the two infection mutants, nin-2 and cyclops-3, were included in the analyses. Total numbers of infection events, i.e. the sums of microcolonies in curled root hair tips (Figure 3b,d), ITs formed in short root hairs (Figure 3c,f) and ITs observed inside well-elongated root hairs (Figure 3a), were significantly reduced in all three mutants compared with WT plants (Figure 4a). Among the three mutants, the number of total infection events decreased in the order cerberus-5 > nin-2 > cyclops-3. Formation of ITs was almost completely inhibited in cyclops-3 and nin-2, while ITs were observed in cerberus-5 at a low but significant frequency.

Figure 4.

 The number of infection events and nodulation in wild-type (WT) plants and the infection mutants, cerberus-5, cyclops-3 and nin-2.
(a) Frequencies of the infection events per plant. ‘Short ITs’ means infection threads (ITs) formed inside the emerging short root hairs, and ‘ITs’ means those that developed into elongated root hairs and/or into cortex.
(b) Numbers of nodules and bumps per plant.
Infection events were visualized by LacZ-labelled Mesorhizobium loti at 1 weeks post-inoculation (wpi) (1) and 2 wpi (2). The data are presented as averages of 14 or 15 individual plants with the standard error (SE) as indicated by error bars.

The cyclops-3 mutants initiated cortical cell divisions and formed bumps containing no endosymbiotic bacteria, but the number of bumps was significantly lower than the number of nodules (and bumps) formed in the WT roots. The nin-2 mutants did not form a nodule primordium as previously reported (Schauser et al., 1999). In contrast to these mutants, cerberus-5 formed a large number of small bumps, about twice that of nodules and bumps formed in the WT plants at 2 weeks post-inoculation (wpi) (Figure 4b).

Expression profiles of the genes involved in the infection process or nodule organogenesis

Analyses of the morphological phenotypes of cerberus indicated that the causal gene is required for either IT formation or progression of nodule development. To analyze further the symbiotic defects in cerberus, we examined the expression of nodulin genes which are presumed to be involved in infection events and/or nodule organogenesis (Figure 5). ENOD40s have been shown to be required for the induction of cortical cell division and formation of successive nodule primordia, but are not involved in the early infection process (Charon et al., 1999; Kumagai et al., 2006). In cerberus-1 and -5, ENOD40-1 was induced at 3–6 days post-inoculation (dpi), although the expression levels were lower than in WT roots. At 12 dpi, the level of expression of ENOD40-1 was significantly lower than in WT nodules, possibly due to the developmental arrest of nodule vascular bundles in cerberus, as shown in Figure 2(b,c). Expression of NIN, which is induced in response to NFs and accompanying infection and nodule initiation, was also induced in cerberus-1 and -5 roots inoculated with M. loti. However, the level of expression was significantly lower than in WT plants. LjN6 is an ortholog of MtN6 which was proposed to be involved in the preparation of M. truncatula cells for progression of the IT (Mathis et al., 1999). Apparently, no significant induction of the LjN6 gene was detected in cerberus-1 and -5. These results are consistent with the severe inhibition of the rhizobial infection process in cerberus.

Figure 5.

 Expression analyses of several early nodulin genes in wild type (WT) and cerberus roots after inoculation with Mesorhizobium loti.
The data were obtained by real-time RT-PCR and are expressed as means of measurements of the RNAs from triplicate plant materials. The error bars represent standard errors.

Involvement of CERBERUS in nodule organ development

The legume mutants identified thus far which have defects in the IT formation and/or growth are accompanied by a lack of nodulation or by incomplete nodule organogenesis. This is also the case for cerberus, because nodule organogenesis is aborted at the stage of small bumps (Figure 2b,c). To assess the involvement of CERBERUS in nodule organogenesis, we transformed a cauliflower mosaic virus (CaMV) 35S promoter-driven gain-of-function CCaMKT265D allele into the cerberus-5 mutants. It was recently demonstrated that CCaMK plays a pivotal role in symbiotic signal transduction leading to nodule organogenesis, and its gain-of-function mutation was shown to confer spontaneous nodulation in the absence of rhizobia (Gleason et al., 2006; Tirichine et al., 2006). The results are shown in Figure 6 and Table 1. The WT roots formed spontaneous nodules in the absence of rhizobia (Figure 6a); these were rather smaller in size than the nodules formed by infection with M. loti (Figure 6b). cerberus-5 also formed spontaneous nodules without Rhizobium inoculation (Figure 6c), even though the frequency of nodule and bump formation was quite low compared with the WT plants (Table 1). Interestingly, the number of bumps formed on the pCaMV35S::CCaMKT265D transgenic roots of cerberus-5 increased significantly after inoculation with M. loti (Figure 6e and Table 1). Nodules and/or bumps formed on transgenic roots contained no endosymbiotic bacteria (supplementary Figure S2).

Figure 6.

 Spontaneous nodulation on the roots transformed with CaMV35S promoter-driven gain-of-function CCaMKT265D.
(a) Spontaneous nodulation of the transgenic wild-type (WT) roots in the absence of rhizobia.
(b) Nodules formed on the transgenic WT roots with rhizobial inoculation.
(c) Spontaneous nodulation of cerberus-5 roots transformed with CCaMKT265D without rhizobial inoculation.
(d, e) Nodules and bumps formed on cerberus-5 roots transformed with CCaMKT265D with rhizobial inoculation. The nodules never contain endosymbiotic rhizobia, and are thus designated as ‘Empty nodule’ in Table 1.
(f, g) Control experiments for the roots of WT (f) and cerberus-5 (g) transformed with pCaMV35S::CCaMKT265T and grown in the absence of Rhizobium bacteria.
The transgenic plants were grown for 6 weeks with or without Mesorhizobium loti MAFF303099. Both bright field (left) and GFP fluorescence (right) images are shown through panels (a–e). In panels (f) and (g), only GFP images are shown. Nodules and bumps are indicated by arrowheads and arrows, respectively. Scale bar = 1 mm.

Table 1.   Formation of spontaneous nodules and bumps on the roots transformed with pCaMV35S::CCaMKT265D with or without M. loti inoculation
PlantConstructMesorhizobium loti+Mesorhizobium loti
BumpsSpontaneous nodulesBumpsEmpty nodulesInfection
  1. The data represent the numbers of plants that formed bumps or nodules/total number of the independent transformants tested. Numerals in parentheses indicate the numbers of bumps or nodules as averages (±SE) of all transformants. The data were collected from plants 6 weeks post-inoculation. As the control experiments, the roots were transformed with pCaMV35S::CCaMKT265T (wild type CCaMK).

  2. n.d., not determined.

cerberus-5CCaMKT265D2/13 (0.23 ± 0.17)5/13 (0.62 ± 0.24)10/10 (4.0 ± 1.13)3/10 (0.33 ± 0.25)
CCaMKT265T0/80/86/6 (5.0 ± 0.58)0/6
WT-GifuCCaMKT265D9/10 (4.9 ± 1.41)10/10 (4.9 ± 0.81)n.d.n.d.n.d.

Map-based cloning of the CERBERUS gene

We initiated chromosomal mapping of the CERBERUS locus using a mapping population generated by crossing cerberus-5 with L. japonicus MG-20 Miyakojima. Simple sequence repeat (SSR) and derived cleaved amplified polymorphic sequence (dCAPS) (Hayashi et al., 2001) markers were assessed for co-segregation in 2205 F2 homozygous recessive mutant plants. As a consequence, we delimited the CERBERUS gene locus in a region of about 90 kb between two dCAPS markers on linkage group 5 (Figure S3a). A TAC clone, LjT135A04, covering this region was sequenced and we predicted nine open reading frames (ORFs). Sequence comparisons between these ORFs of cerberus-5 and the WT genomes allowed us to identify a mutation in ORF-5 (supplementary Figure S3b). Subsequent sequence analyses of ORF-5 in five mutants belonging to the same complementation group demonstrated that all of them contained mutation in the same ORF-5 (Figure 7a). In cerberus-1, a 59-bp nucleotide stretch over the fourth intron and the fifth exon was deleted, resulting in missplicing and a frame shift followed by an abrupt stop codon. In cerberus-2 to -5, insertions of putative retrotransposons were found at various positions in ORF-5. cerberus-6 carried a single nucleotide substitution at the splicing site of intron-16/exon-17 which resulted in a frame shift immediately followed by an abrupt stop codon.

Figure 7.

 Predicted protein structure of CERBERUS.
(a) Exon and intron structure of the CERBERUS gene. The boxes show exons. A lozenge indicates the position of deletion of 59 bp in cerberus-1 (Ljsym7). Solid arrowheads indicate the position of the retrotransposon insertion in cerberus-2 (Ljsym41), cerberus-3 (Ljsym55), cerberus-4 (Ljsym57) and cerberus-5 (Ljsym101). An open arrowhead indicates a single nucleotide substitution changing the AG/GT splice site of intron16/exon17 to AG/AT in cerberus-6 (LjS28-2B). Red lines (a) and (b) correspond to regions of the U-box domain and WD40-repeats, respectively.
(b) Sequence alignment of the U-box domain of CERBERUS with those of the plant U-box type E3 ubiquitin ligases, AtPUB17 (NM_102674), BoARC1 (EU344909) and OsSPL11 (AY652589). Identical amino acids are shown as black boxed. Well-conserved residues and those of high similarity are shown by dark gray and light gray shading, respectively. The residues shown to be crucial for the interactions with E2 enzymes (Stone et al., 2003; Yang et al., 2006) are indicated by asterisks.

A genomic DNA fragment of about 16 kb long which contains ORF-5 together with its promoter and terminator regions was transformed into cerberus-5 roots by hairy root transformation, resulting in restoration to fully effective nodulation comparable to WT plants (Figure S4). Taken together, we concluded that ORF-5 is the CERBERUS gene.

CERBERUS encodes a novel U-box protein containing WD-40 repeats

A cDNA for the CERBERUS mRNA cloned by 5′- and 3′-rapid amplification of [5′- and 3′-] complementary DNA ends (5′- (3′-)RACE) from the WT Gifu RNA was about 5 kb long and sequence alignment with the genomic sequence indicated that the CERBERUS gene consists of 17 exons (Figure 7a). Genomic Southern blot analysis indicated that CERBERUS is a single gene in the L. japonicus genome and is conserved among the legume species tested, namely Glycine max, M. truncatula and Pisum sativum (Figure S5).

CERBERUS encodes a conceptual peptide of 1477 amino acids which contains a U-box domain in the middle part and at least three WD-40 repeats in the C-terminal region (Figures 7a and S6). U-box domains are known as components of E3 ubiquitin ligases which interact with E2 ubiquitin-conjugating enzymes (Hatakeyama et al., 2001; Azevedo et al., 2001). Figure 7(b) shows an alignment of the U-box amino acid sequence in CERBERUS with those of plant proteins known to function as E3 ubiquitin ligases, AtPUB17 (Yang et al., 2006), BoARC1 (Stone et al., 2003) and OsSLP11 (Zeng et al., 2004). The CERBERUS U-box showed high similarity to other U-box sequences and the amino acid residues which are crucial for the interaction with E2 enzymes were perfectly conserved in the CERBERUS U-box. WD-40 repeats are proposed to coordinate interactions with many kinds of proteins and/or small ligands (Smith et al., 1999). The mutations in cerberus-2 to -5 lead to the lack of WD-40 repeats, whereas the mutation in cerberus-1 results in the lack of both U-box and WD-40 domains (Figure 7a).

We found CERBERUS homologs with the same domain structures in several angiosperms, Vitis vinifera, Populus trichocarpa, Oryza sativa and Sorgum bicolor, and a spikemoss, Selaginella moellendorffii. These homologous proteins appeared phylogenetically to belong to three clades (Figure S7). Two contain angiosperms and the other contains a moss. The biochemical function of these proteins, however, have not yet been determined.

Expression analyses of the CERBERUS gene

The transcripts of CERBERUS were detected by RNA gel blot analysis in uninfected roots and nodule primordia, as well as in mature nodules and in calli and young shoots at very low levels (Figure 8a). To examine the CERBERUS transcript levels in the roots during early stages of M. loti infection, we performed real-time reverse transcription-polymerase chain reaction (RT-PCR) analyses (Figure 8b). The CERBERUS transcripts appeared to increase in the early stages of rhizobial infection up to fivefold at 3–7 dpi and further increased 12 dpi in the WT roots, while expression in the cerberus roots was low and not significantly enhanced upon inoculation with M. loti.

Figure 8.

 Expression of the CERBERUS gene.
(a) Northern blot analysis of RNAs from various organs and during nodule development of Lotus japonicus.
(b) Real-time RT-PCR analysis of CERBERUS transcripts during early stages of nodulation of wild type (WT), cerberus-5 and cerberus-1. Total RNAs were sampled from entire root systems after inoculation with Mesorhizobium loti Tono. The data are the means ± SE of three individual plant samples.

To analyse the spatial expression of the CERBERUS gene, we employed histochemical observations with a CERBERUS promoter::β-glucuronidase (GUS) fusion in roots and nodules by hairy root transformation. In uninfected roots, GUS activity was detected in the whole root tissue, particularly in the root epidermis including root hairs (Figure 9a). Strong GUS activities were detected in nodule primordia (Figure 9b). In mature nodules, GUS activity was detected in the central infected zone (both infected and uninfected cells) and in vascular bundles (Figure 9c).

Figure 9.

 Histochemical detection of the CERBERUS promoter::GUS fusion by hairy root transformation.
(a) Uninfected root of wild type (WT) Gifu.
(b) Developing nodule primordium in WT root infected with Mesorhizobium loti (2 weeks post-inoculation, wpi).
(c) Mature nodule formed on WT root, 3 wpi.
(d) Root of nin-2 (2 wpi).
(e) Bump on cyclops-3 root (2 wpi).
(f) Bump on alb1-3 root (2 wpi).
Scale bars = 200 μm.

We further analyzed CERBERUS promoter activity in the infection mutants, nin-2, cyclops-3 and alb1-3. The nin-2 mutants showed neither rhizobial invasion nor nodulation, but expression of CERBERUS was clearly detected in the roots (Figure 9d). In cyclops-3 and alb1-3, CERBERUS was expressed in whole roots and more strongly in the bumps (Figure 9e,f).


Morphological phenotyping of the cerberus mutants indicated that CERBERUS is primarily involved in the rhizobial infection process. In comparisons with the other Lotus Hist mutants, cerberus showed some distinctive characteristics. In cerberus, the rhizobial infection process was aborted at a very early stage of IT formation, mostly at the stage of microcolonies in the curled root hair tips (Figure 3b,d,e). Even when ITs were occasionally formed inside short root hair cells, their extension was arrested within the epidermal cell layer (Figures 3c,f and 4a). In the crinkle mutants previously described, the initiation and extension of ITs in root hairs appears comparable to the WT, although most ITs end at the base of the root hair cells (Tansengco et al., 2003). Another Hist mutant, alb1, shows more advanced infection phenotypes; ITs formed on alb1-1 penetrate into the root cortex and nodule tissues, but rhizobia are not released from ITs into the nodule cells (Imaizumi-Anraku et al., 1997, 2000). In alb1-3, possibly a stronger allele of alb1, ITs are formed but are mostly arrested within the root hairs (Yano et al., 2006). In both alleles of alb1, however, nodule organogenesis is clearly more advanced than in cerberus. It is noteworthy that these previously described Hist mutants form, at low frequency, nodules with successful endosymbiosis with rhizobia (‘type-II’ nodules), and those nodules exhibit substantial nitrogen-fixing activity (Imaizumi-Anraku et al., 1997; Tansengco et al., 2003). In contrast, we did not find such type-II nodule formation in cerberus. These observations strongly suggest that CERBERUS acts at an earlier stage of IT formation than the causal genes for crinkle and alb1. NIN and NSP2 are both positioned in the early nodulation-specific pathway that follows CSP because those mutants show severe defects in IT formation and cortical cell division (Schauser et al., 1999; Murakami et al., 2006). Although the nin-2 mutants formed ITs inside short growing root hairs on very rare occasions, they never formed nodule primordia (Figure 4a). Thus, CERBERUS is possibly downstream of these genes. Induction of the NIN gene in cerberus roots upon inoculation with M. loti also supports this, though the level of NIN expression was less than in WT roots, probably reflecting a low infection frequency in cerberus.

cerberus formed only small bumps with no endosymbiotic bacteria, and development of mature nodules was infrequently observed (Figures 1 and 2). It was recently shown that Lotus snf1 mutants (gain-of-function mutation in CCaMK) can develop empty nodules spontaneously in the absence of rhizobia (Tirichine et al., 2006). In contrast to this phenotype, the hit1 mutant forms excessive ITs penetrating into the root cortex without inducing timely cortical cell division (Murray et al., 2007). These findings indicate that the rhizobial infection process and nodule organogenesis are governed by genetically distinct programs, even though their concerted progression is essential for the establishment of effective symbiosis. Transformation of cyclops roots with the gain-of-function CCaMKT265D results in spontaneous formation of genuine nodule structures, despite the fact that cyclops mutants form only small bumps upon inoculation of M. loti. This indicates that CYCLOPS is not directly involved in the nodule organogenesis program (Yano et al., 2008). cerberus roots transformed with the same CCaMKT265D construct also formed structurally genuine nodules without rhizobial inoculation, although at very low frequency compared with WT plants (Table 1 and Figure 6). These nodules formed on the cerberus roots with CCaMKT265D appeared to be structurally well developed except for the lack of endosymbiotic rhizobia (Figure S2). The low frequency of spontaneous nodulation mediated by CCaMKT265D in comparison with WT roots was also the case in cyclops (Yano et al., 2008). Thus, like CYCLOPS, CERBERUS may not be indispensable to nodule organogenesis itself, and the arrest of nodule organ development in cerberus roots might be an indirect consequence of the arrest of the rhizobial infection process. In this regard, however, it should be noted that bump formation in CCaMKT265D-transformed cerberus roots was significantly reinforced by M. loti inoculation (Figure 6d,e and Table 1), and most of them failed to retain complete nodule development, even at 6 wpi. These observations indicate that formation of nodule primordia in the cerberus roots is strongly dependent on perception of the rhizobial symbiotic signals, even when the downstream pathway(s) of CSP was forcibly driven by the gain-of-function form of CCaMKT265D. Based on these observations, we hypothesize that CERBERUS plays a role(s) in the developmental program of nodule organogenesis in conjunction with bacterial infection and/or perception of NFs, besides its role in IT initiation and/or growth. Relatively strong expression of the CERBERUS promoter::GUS fusion gene in developing nodule primordia and the infected zone of mature nodules (Figure 9b,c) also supports this hypothesis.

It is noteworthy that a Medicago mutant, lin, displays a symbiotic phenotype very similar with that of cerberus. The lin mutants show low frequency of IT formation and extension, and nodule development fails to mature (Kuppusamy et al., 2004). The LIN locus has been mapped on the lower arm of linkage group 1 of M. truncatula, which corresponds, from genome synteny between Lotus and Medicago (Choi et al., 2004), to the region of Lotus linkage group 5 where CERBERUS resides. Therefore, it is very likely that LIN is the Medicago ortholog of the Lotus CERBERUS gene.

cerberus formed a large number of nodule bumps which amounted to nearly double the number of nodules formed on WT roots (Figure 4b). This is in contrast to crinkle which forms bumps in numbers equivalent to nodule numbers in WT plants (Tansengco et al., 2003). cyclops also forms bumps, but the frequency is lower than in cerberus and even lower than in WT plants (Figure 4b). It has been well documented that nodule number is controlled systemically by type of long-distance signaling termed ‘autoregulation’, mediated by a receptor kinase, HAR1 (see Oka-Kira and Kawaguchi, 2006; for review). Since a cerberus har1 double mutant (cerberus-6) and har1 mutants develop a comparable nodule number (Murray et al., 2006), bump formation in cerberus is essentially under the control of HAR1-mediated autoregulation. However, the low frequency of infection events in cerberus may cause incomplete operation of the autoregulation, thus resulting in recurrent bump formation. Alternatively, the absence of nitrogen-fixing nodules in cerberus may induce excessive bump formation, as suggested in the case of ineffective nodules formed on Fix mutants (Suganuma et al., 2003).

Molecular cloning of CERBERUS revealed that it encodes a protein containing a U-box and WD-40 repeats. In addition to the fact that the U-box is known as a domain found in diverse isoforms of E3 ubiquitin ligases, some proteins with WD-40 repeats are also known to be involved in the ubiquitin–proteasome system in mammals (Higa et al., 2006; Hu et al., 2008). Therefore, it is very likely that CERBERUS is an E3 ubiquitin ligase, or at least has a function related to the ubiquitin–proteasome system. Recent evidence has shown that targeted proteolysis mediated by the ubiquitin–proteasome system plays a pivotal role in multiple plant cellular processes, including phytohormone signaling and defense reactions against pathogens. For instance, Arabidopsis AtPUB17 and tobacco ACRE276 were demonstrated to be involved in regulation of cell death and defense against pathogen attack (Yang et al., 2006), while AtPUB22 and AtPUB23 were recently shown to be required for responses to drought stress (Cho et al., 2008). Possible involvement of E3 ubiquitin ligases in the legume–Rhizobium symbiosis have also been reported. Seven in Absentia (SINA) E3 ligases of M. truncatula were recently suggested to be involved in common mechanisms underlying both IT and symbiosome formation, such as endomembrane vesicle trafficking (Herder et al., 2008). In L. japonicus, LjNSRING, a nodule specific putative E3 ligase, was suggested to be involved in IT initiation and successive nodulation (Shimomura et al., 2006). These examples are, however, both RING-finger type E3 ligases, and their knockdown caused not only symbiotic defects but also severe plant growth retardation. In contrast, cerberus exhibited no obvious pleiotropic abnormality in plant growth. Although our attempts to detect E3 ubiquitin ligase activity in vitro using various peptide fragments of CERBERUS including the U-box have so far been unsuccessful, it is one of the main objectives in our future experiments.

A number of proteins of high similarity with CERBERUS were found in angiosperms and a spike moss, which contain both U-box and WD-40 repeats (Figure S7), but none of their functions were characterized. One of the few proteins with U-box and WD-40 repeats which has been functionally characterized is yeast PRP19p (Ohi et al., 2005). It is a subunit of the spliceosome and plays a role in splicing of pre-mRNAs. A human homolog of PRP19p, SNEV, has been demonstrated to be involved in pre-mRNA splicing, repair of DNA double-strand break from stress damage, and transport of ubiquitinated substrate to the proteasome (Voglauer et al., 2006). Overall sequence similarity of CERBERUS with these proteins was low except for the domain structures. However, it is noted that CERBERUS is predicted to have a high possibility of being localized in the nucleus by the programs WoLFPSORT and LOCtree, despite the fact that it contains no obvious nuclear localization signal. Thus, the involvement in such a function as mRNA splicing might be a possible alternative biochemical function for CERBERUS. Our attempts to examine the intracellular localization of the CERBERUS protein by means of GFP-fusion have so far been unsuccessful, but localization analysis of CERBERUS in nodule cells is one of our tasks for the future.

Although there is no direct experimental evidence to show the exact function of CERBERUS, we speculate that it is most likely an E3 ubiquitin ligase. To explore the functions of CERBERUS in the legume–Rhizobium symbiosis, analysis of the intracellular localization of CERBERUS protein as well as screening proteins that interact with CERBERUS U-box and/or WD-40 repeats is currently under way.

Experimental procedures

Plant materials and growth conditions

The mutants were all derived from Lotus japonicus B-129 Gifu. cerberus-5 (Ljsym101) was isolated from a mutant library of the plants regenerated from calli. cerberus-1(Ljsym7) was from T-DNA-transformed lines (Schauser et al., 1998). cerberus-2, -3 and -4 (Ljsym41, 55 and 57) cerberus alleles and nin-2 were from progenies of the transformants carrying the Ac launching construct (Thykjaer et al., 1995). cerberus-6 (LjS28-2B) was from har1 plants mutagenized with EMS as described elsewhere (Karas et al., 2005). cyclops-3 and alb1-3 were described previously (Szczyglowski et al., 1998; Yano et al., 2006).

Surface-sterilized seeds were germinated on 0.9% (w/v) agar plates containing 1/10-strength B5 medium salts for a week. For nodulation tests, the seedlings were transferred to vermiculite pots watered by 1/2 strength B&D (Broughton and Dilworth, 1971) medium, inoculated with M. loti Tono or MAFF303099 (Kawaguchi et al., 2002), and grown in an artificially lit growth cabinet at 24°C for 16 h (light) and 22°C for 8 h (dark). For harvesting the seeds, the seedlings were transferred to pots of horticultural soil (Kureha Chemical Co. Ltd, http://www.kureha.co.jp/en/index.html) covered by a vermiculite layer and grown in a greenhouse at about 28°C (day) supplemented with mercury lamps for 16 h and 24°C (night).

Observation of the infection events

Infection events in the mutants and WT Gifu plants were observed by inoculation with M. loti MAFF303099 constitutively expressing the β-galactosidase (LacZ) gene (Kumagai et al., 2006) or DsRed (T. Kato and M.H., unpublished). Plant growth and detection of β-galactosidase activity were performed as described (Yano et al., 2006). Infection events were observed with a fluorescence stereo-microscope MZFLIII (Leica Microsystems, http://www.leica-microsystems.com/) and a confocal laser scanning microscope (Radiance-II, BioRad, http://www.bio-rad.com/).

Light microscopy

The samples were fixed in 4% paraformaldehyde and 2.5% glutaraldehyde in 0.1 m sodium phosphate buffer (pH 7.2) and embedded in paraffin according to procedures described previously (Kouchi and Hata, 1993). Some samples were post-fixed in 2% osmium tetroxide in 0.1 m sodium phosphate and embedded in an epoxy resin (Quetol-812, Nisshin-EM, http://www.nisshin-em.co.jp/home/index.html) as described in Kumagai et al. (2007). Sections of 1.5 μm (for epoxy resin) or 8 μm (for paraffin) were made and stained with 0.1% (w/v) toluidine blue in 0.5% (w/v) sodium tetraborate (pH 9.0) and observed by a light microscope (LEITZ DMR; Leica).

Fine mapping of the CERBERUS locus

An F2 mapping population was established by crossing the cerberus-5 mutant with L. japonicus MG-20 Miyakojima (Kawaguchi et al., 2001). A total of 2205 F2 plants were assessed for co-segregation analysis with DNA markers (http://www.kazusa.or.jp/lotus/) flanking the CERBERUS locus.

DNA manipulation and cDNA cloning

All conventional DNA manipulations were carried out according to standard procedures (Sambrook and Russell, 2001). A CERBERUS cDNA was cloned from cDNAs reverse-transcribed with oligo(dT)18 as a primer from nodule total RNA by PCR under the following conditions: 98°C (2 min), 35 cycles of 98°C (30 sec), 60°C (30 sec), 72°C (3 min) and 72°C (10 min) using the Phusion enzyme (Finnzymes, http://www.finnzymes.com/). 5′- and 3′-RACE was performed with the SMART RACE amplification kit (Clontech, http://www.clontech.com/). All the primers used throughout this study are listed in Table S1. Genomic Southern blot hybridization was performed according to standard procedures (Sambrook and Russell, 2001) for DNAs isolated from the seedlings or leaves with 32P-labelled probes prepared from CERBERUS cDNA fragments.

Complementation test and CCaMKT265D transformation

A binary vector pC1300GFP–AscI was made by insertion of an AscI linker into the SmaI site of pC1300GFP (Kumagai and Kouchi, 2003). A MluI–SalI fragment (16 kb) containing the entire CERBERUS gene was excised from a transformation-competent artificial chromosome (TAC) clone, LjT135A04, followed by ligation into the AscI–SalI site of pC1300GFP–AscI. The construct was introduced into Agrobacterium rhizogenes LBA1334 and transformed into cerberus-5 by hairy root transformation (Diaz et al., 2005). The plants with transformed hairy roots were inoculated with M. loti MAFF303099 and grown with ½ B&D medium supplemented with 0.5 mm ammonium nitrate for 1 month in an artificially lit growth cabinet. Transformed roots were selected by GFP fluorescence and observed by a fluorescence stereomicroscope.

For the transformation with gain-of-function CCaMKT265D, the binary vector containing CaMV35S promoter-driven CCaMKT265D was constructed as previously described (Yano et al., 2008). Wild-type Gifu and cerberus-5 were transformed by the hairy root transformation procedures, and grown with ½ B&D medium supplemented with 0.5 mm ammonium nitrate. M. loti MAFF303099 was inoculated at transplanting to vermiculite pots for the inoculated plants. Microscopic observation of inoculated and non-inoculated plants was performed 6 weeks after transplanting.

Expression analysis

Total RNAs were extracted from various tissues of WT Gifu and cerberus-1 and -5 by RNeasy Plant Mini Kit (Qiagen, http://www.qiagen.com/). Northern blot hybridization was done according to standard procedures (Sambrook and Russell, 2001). The probe was amplified by PCR and labelled with 32P-dCTP by random priming. Real-time RT-PCR was carried out after reverse transcription using the QuantiTect Reverse Transcription Kit (Qiagen). The resultant cDNA was used as templates for real-time PCR with Light Cycler (Roche Diagnostics, http://www.roche.com/) according to the manufacturer’s instructions, with primer sets specific for CERBERUS, ENOD40-1, NIN and LjN6, respectively (Table S1). Ubiquitin was used as an internal standard with PCR primers described by Flemetakis et al. (2000).

Promoter GUS analysis

The promoter region (3338 bp upstream of the translation start) and the terminator region (1001 bp downstream of the stop codon) of CERBERUS were amplified by PCR using the corresponding primer set (Table S1). The promoter fragment treated with AscI and BamHI was inserted into pC1300GFP-AscI, followed by insertion of the terminator fragment between the BamHI and SalI sites. Then, the reading frame cassette A of the Gateway vector conversion system (Invitrogen, http://www.invitrogen.com/) was introduced into the BamHI site between the promoter and terminator. A GUS Plus gene was amplified from pCAMBIA1305.1 by nested PCR, subcloned in pDONR/Zeo (Invitrogen) by Gateway BP reaction, and then transferred into the destination vector described above by LR reaction.

The construct was transfected into L. japonicus Gifu by hairy root transformation and inoculated with M. loti MAFF 303099. The roots were harvested 2–3 wpi and stained with 0.5 mg ml−1 5-bromo-4-chloro-3-indolyl β-d-glucuronic acid cyclohexylammonium salt, 5 mm potassium ferricyanide, 5 mm potassium ferrocyanide and 10 mm EDTA in 100 mm sodium phosphate (pH 7.0) and were incubated at 37°C in the dark. The roots were bleached in 100% ethanol before observation with a stereomicroscope. The nodules were embedded in 5% agar and sectioned (200 μm thick) by a microslicer, stained and observed by a light microscope (BZ-9000; Keyence Co., http://www.keyence.com/).

Computer analysis

Analyses of the primary protein structures and domain search were aided by SMART (http://smart.embl-heidelberg.de/), PredictProtein (http://www.predictprotein.org/main.php), WoLFPSORT (http://wolfpsort.org), and LOCtree (http://www.predictprotein.org/cgi/var/nair/loctree/).

Accession number

Genomic and mRNA sequences of CERBERUS are deposited in the DDBJ/EMBL/GeneBank databases under accession numbers, AB505797 and AB505798, respectively.


This work was supported by the Core Research for Evolutional Science and Technology (CREST) program from the Japan Science and Technology Agency to YU and MH, the Special Coordination Fund for Promoting Science and Technology from the Ministry of Education, Culture, Sports, Science and Technology, Japan to HK and YU, and the Danish National Research Foundation to JS and NS The authors thank to Robert W. Ridge of the International Christian University for critical reading of the manuscript.