Apoplastic plant subtilases support arbuscular mycorrhiza development in Lotus japonicus


  • Naoya Takeda,

    1. Faculty of Biology, Genetics, University of Munich, Großhaderner Straße 2, 82152 Martinsried, Germany
    Search for more papers by this author
    • Present addresses: NT; National Institute of Agrobiological Science, 2-1-2 Kan-non-dai, Tsukuba, Ibaraki 305-8602, Japan. EA; Graduate School of Life and Environmental Sciences, University of Tsukuba, 1-1-1 Ten-noudai, Tsukuba, Ibaraki 305-8572, Japan.

  • Shusei Sato,

    1. Kazusa DNA Research Institute, 2-6-7 Kazusa-kamatari, Kisarazu, Chiba 292-0818, Japan
    Search for more papers by this author
  • Erika Asamizu,

    1. Kazusa DNA Research Institute, 2-6-7 Kazusa-kamatari, Kisarazu, Chiba 292-0818, Japan
    Search for more papers by this author
    • Present addresses: NT; National Institute of Agrobiological Science, 2-1-2 Kan-non-dai, Tsukuba, Ibaraki 305-8602, Japan. EA; Graduate School of Life and Environmental Sciences, University of Tsukuba, 1-1-1 Ten-noudai, Tsukuba, Ibaraki 305-8572, Japan.

  • Satoshi Tabata,

    1. Kazusa DNA Research Institute, 2-6-7 Kazusa-kamatari, Kisarazu, Chiba 292-0818, Japan
    Search for more papers by this author
  • Martin Parniske

    Corresponding author
    1. Faculty of Biology, Genetics, University of Munich, Großhaderner Straße 2, 82152 Martinsried, Germany
    Search for more papers by this author

*(fax +49 2180 74702; e-mail parniske@lmu.de).


In the arbuscular mycorrhiza (AM) symbiosis, plant roots accommodate Glomeromycota fungi within an intracellular compartment, the arbuscule. At this symbiotic interface, fungal hyphae are surrounded by a plant membrane, which creates an apoplastic compartment, the periarbuscular space (PAS) between fungal and plant cell. Despite the importance of the PAS for symbiotic signal and metabolite exchange, only few of its components have been identified. Here we show that two apoplastic plant proteases of the subtilase family are required for AM development. SbtM1 is the founder member of a family of arbuscular mycorrhiza-induced subtilase genes that occur in at least two clusters in the genome of the legume Lotus japonicus. A detailed expression analysis by RT-PCR revealed that SbtM1, SbtM3, SbtM4 and the more distantly related SbtS are all rapidly induced during development of arbuscular mycorrhiza, but only SbtS and SbtM4 are also up-regulated during root nodule symbiosis. Promoter–reporter fusions indicated specific activation in cells that are adjacent to intra-radical fungal hyphae or in cells that harbour them. Venus fluorescent protein was observed in the apoplast and the PAS when expressed from a fusion construct with the SbtM1 signal peptide or the full-length subtilase. Suppression of SbtM1 or SbtM3 by RNAi caused a decrease in intra-radical hyphae and arbuscule colonization, but had no effect on nodule formation. Our data indicate a role for these subtilases during the fungal infection process in particular arbuscule development.


Arbuscular mycorrhiza (AM) is a symbiotic interaction between land plants and AM fungi belonging to the Glomeromycota (Smith and Read, 2008). AM fungal hyphae absorb water and minerals including phosphate from the soil, which are then supplied to the plant root (Harrison and van Buuren, 1995; Harrison et al., 2002). In return, the AM fungus obtains carbohydrates, probably hexoses, from the host plant (Bago et al., 2003;Harrison, 2005; Parniske, 2008). During AM development, fungal hyphae on the root surface induce formation of a pre-penetration apparatus by the underlying plant cell, and subsequently penetrate this cell (Genre et al., 2005, 2008). After intracellular passage of the outer cell layers of Lotus japonicus, fungal hyphae exit plant cells and continue to grow through the intercellular space along the host root axis (Demchenko et al., 2004). Symbiotic structures (arbuscules and vesicles) are formed in the root cortex. Arbuscules are a major site of nutrient exchange between AM fungi and the host plant (Gianinazzi-Pearson, 1996; Harrison, 2005). Arbuscule development starts with a fungal hypha penetrating into a cortical cell. The fungal hypha branches repeatedly to form a tree-shaped structure (Javot et al., 2007). Each fungal arbuscule branch is surrounded by a plant-derived peri-arbuscular membrane, which prevents fungal contact with the plant cytosol and hence controls nutrient exchange between the symbionts. The apoplastic space between fungal plasma membrane and the peri-arbuscular membrane is called the peri-arbuscular space (PAS), a compartment that is still only poorly characterized (Bonfante and Perotto, 1995). Metabolite transport from fungus to the plant and vice versa involves crossing of the two membranes at the symbiotic interface. Consistent with this idea, expression of symbiosis-induced plant phosphate transporter genes was observed in arbuscule-containing cells (Bucher, 2007), and the Medicago truncatula phosphate transporter PT4 has been localized to the peri-arbuscular membrane (Harrison, 2005).

AM is genetically connected to the root nodule symbiosis (RNS), during which the host plant forms a new specialized organ, the root nodule, for the accommodation of rhizobia, which fix atmospheric nitrogen to ammonium (Oldroyd and Downie, 2004). RNS is thought to have evolved by recruiting so-called common symbiosis (SYM) genes from evolutionarily older AM (Kistner and Parniske, 2002; Markmann et al., 2008). The common signalling network controls the symbiotic development of AM and RNS, and comprises a symbiosis receptor kinase (SYMRK), two ion channels (CASTOR/POLLUX), a calcium–calmodulin-dependent protein kinase (CCaMK) and CYCLOPS, a nuclear protein that interacts with CCaMK (Endre et al., 2002; Stracke et al., 2002; Anéet al., 2004; Levy et al., 2004; Imaizumi-Anraku et al., 2005; Tirichine et al., 2006; Charpentier et al., 2008; Yano et al., 2008). In addition to the common symbiosis genes, the divergent nature of AM and RNS predicts that additional and AM-specific genes must govern AM development and maintenance (Harrison, 2005; Javot et al., 2007; Parniske, 2008). However, so far only two phosphate transporters have been genetically demonstrated to be required specifically for AM (Maeda et al., 2006; Javot et al., 2007). To obtain further insights into the molecular processes underlying AM development, we performed a transcriptome comparison of AM roots with non-inoculated roots, and isolated several AM-induced genes from L. japonicus (Kistner et al., 2005). Among the AM-induced genes, two serine protease genes, SbtM1 (previously called SbtM by Kistner et al., 2005) and SbtS, with sequence similarity to subtilisin-like serine protease (subtilase) genes (Siezen and Leunissen, 1997; Rawlings and Barrett, 1999), exhibited strong transcriptional up-regulation upon inoculation with AM fungi. Independent large-scale transcriptome analyses in M. truncatula (Liu et al., 2003) and the monocot rice (Oryza sativa) (Guimil et al., 2005) have consistently identified activation of subtilase genes during AM. This phylogenetically widespread occurrence among angiosperms suggests a conserved and ancient role in symbiosis. Eukaryotic proteases have been implicated in regulation of a diverse array of processes, including programmed cell death, non-self recognition and the response to developmental cues as well as pathogens (van der Hoorn, 2008). For example, the two Arabidopsis subtilases ALE1 and SDD1 are key regulators of embryo cuticle formation (Tanaka et al., 2001) and stomatal density (von Groll et al., 2002). SbtM1 has similarity to the pathogenesis-induced P69 subtilases from tomato (Tornero et al., 1996; Jorda et al., 1999). Like typical plant subtilases, they comprise a predicted secretion signal peptide, a pro-peptide domain and a peptidase domain. The mature protease is formed after removal of the signal peptide and the pro-peptide domain (Baker et al., 1993). The potential regulatory role of subtilases, together with their highly specific expression during symbiosis, encouraged us to functionally analyse the subtilase genes SbtM1 and SbtM3. Here, we investigated the expression patterns of AM-induced subtilase genes and analysed the phenotypic consequences of inhibiting the expression of SbtM1 and SbtM3, two genes that appear to be exclusively expressed during AM.


Genomic organization of AM-induced subtilases

In order to analyse the genomic organization of the SbtM gene family, genomic TAC and BAC libraries containing closely related members were sequenced in the context of the Ljaponicus genome project (Sato et al., 2008). Consistent with previous reports describing the occurrence of plant subtilase genes in physically linked clusters (Meichtry et al., 1999; Yamagata et al., 2000; Beers et al., 2004), we identified two SbtM gene clusters (Figure 1a). The SbtM1/SbtM2/SbtM3 gene cluster was located within a 14 kbp interval on genomic clone BM1477b (AP009544), located on the long arm of chromosome 2. SbtM4 was initially identified through tentative consensus sequence TC10429, which exhibited high sequence similarity to SbtM1. A cluster comprising SbtM4 and SbtM5 was subsequently annotated within 10 kb on genomic clone BM1764 (AP009542) located on the short arm of chromosome 4. SbtM2 and SbtM5 are predicted pseudogenes, each containing a premature stop codon. SbtM3 comprises two exons and encodes a complete subtilase protein (750 amino acid residues, 66% amino acid identity with SbtM1). SbtM4 encodes a protein of 755 amino acid residues that shares 57% identity with SbtM1. The predicted polypeptides SbtM1, SbtM3, SbtM4 and SbtS comprise a predicted secretion signal peptide, a pro-peptide region and a peptidase domain (Figure 1b) (Takeda et al., 2007). The peptidase domain contains a conserved catalytic triad (Dodson and Wlodawer, 1998) and a protease-associated (PA) domain with higher amino acid sequence divergence than other regions of the peptidase domain (Figure 1b) (Takeda et al., 2007). The PA domain was identified by sequence comparisons and is proposed to mediate protein–protein interaction and confer substrate specificity (Mahon and Bateman, 2000).

Figure 1.

SbtM gene clusters in Ljaponicus and predicted protein domains.
(a) The SbtM gene family resides within two gene clusters. SbtM3 contains an intron. SbtM2 and SbtM5 are probably pseudogenes because both carry a premature stop codon (arrowheads) at 763 and 1585 bp, respectively, from the A of the predicted start codon. No other genes were predicted in these clusters.
(b) Subtilase domain organization, comprising a predicted secretion signal peptide, a pro-peptide domain and a peptidase domain interrupted by a PA domain. The percentages of amino acid identity between SbtM1 and the other AM-induced subtilases are shown for the peptidase and PA domains.

The products of AM-induced subtilase genes isolated from rice and M. truncatula (Liu et al., 2003; Guimil et al., 2005) are closely related to SbtM1. Bioinformatic analysis of the rice genome sequence identified a cluster of two adjacent genes on PAC sequence AP003286 encoding subtilases BAD82002 (AM-induced; Guimil et al., 2005) and BAB89803. The M. truncatula subtilisin-like protease represented by EST AW584611 was found to be up-regulated during AM development (Liu et al., 2003). To determine the genomic environment of the corresponding gene, we performed PCR and obtained the complete ORF and 5′ UTR of this gene. Sequence comparison with the Lotus SbtM gene family revealed that SbtM1 is more closely related to this M. truncatula gene than to SbtM3. The similar expression patterns and close sequence relationship suggest that AW584611 represents the Medicago orthologue of SbtM1. Limited sequence analysis revealed a second subtilase gene approximately 4 kb upstream of the predicted ATG of the AW584611 gene, indicating the presence of a cluster of at least two subtilase genes in M. truncatula.

Expression analysis of AM-induced subtilase genes

Quantitative RT-PCR analysis revealed that SbtM1, SbtM3, SbtM4 and SbtS were up-regulated in roots inoculated with the AM fungus Glomus intraradices, with differences in timing and relative expression (Figure 2). SbtM1 reached a high, plateau-like relative expression level from 4 days after inoculation onwards (Figure 2a). In contrast, the transcript level of SbtM3 continued to increase over the time course (Figure 2b). SbtM4 was already induced at 2 days after inoculation, the earliest time point analysed, but the fold induction was lower than that of the other AM-induced subtilases (Figure 2c). SbtS induction was observed as early as 2 days after inoculation, and transcript levels reached a plateau at 4 days after inoculation (Figure 2d).

Figure 2.

 Subtilase gene expression patterns in AM.
Transcript levels of SbtM1 (a), SbtM3 (b), SbtM4 (c) and SbtS (d) in roots inoculated with Gintraradices. Total RNA was extracted from 4 to 8 roots each at 2, 4, 7 and 14 days after inoculation (dai). The expression level was determined by quantitative RT-PCR relative to non-inoculated roots harvested at each time point (value for non-inoculated roots = 1). Error bars indicate the standard deviation for three independent experiments.

Of the four subtilases tested, only SbtM4 and SbtS were induced upon inoculation with Mesorhizobium loti. Interestingly, only SbtM4 was induced in nodules (Figure 3a,b), which is similar to two other nodulation-responsive genes, NIN and ENOD40-1 (Figure 3c,d). In contrast, SbtS transcript accumulation was transient in roots and not detected in nodules (Figure 3b), consistent with previous observations by Kistner et al. (2005). SbtM4 and SbtS also responded to Nod factor treatment, and the expression patterns were similar to those observed after inoculation with Mloti until 7 days after treatment (data not shown). SbtM1 and SbtM3 were not induced during RNS, and their transcript levels in nodules were at or below the detection threshold of RT-PCR. Transcripts of both of the two pseudogenes, SbtM2 and SbtM5, could be detected in roots, but levels did not increase during AM or RNS as determined by quantitative RT-PCR (data not shown). These results indicate that SbtM1, SbtM3, SbtM4 and SbtS are subject to differential regulation during AM and RNS.

Figure 3.

 Gene expression patterns in RNS. Expression of SbM4 (a), SbtS (b), NIN (c) and ENOD40-1 (d) in roots 2, 4, 7 and 14 days after inoculation with Mloti and mature nodules was compared with that in non-inoculated roots (value for non-inoculated roots = 1) by RT-PCR. Induction of SbtM4, NIN and ENOD40-1 was observed during all stages of RNS, but SbtS was not induced in nodules. Total RNA was extracted from 4–8 roots per sample in each experiment. Error bars indicate the standard deviation for three independent experiments.

Histochemical localization of subtilase promoter activity

To analyse the spatial patterns of subtilase expression, we used promoter fusions with the β-d-glucuronidase (GUS) reporter gene. Transformed roots carrying SbtM1pro:GUS showed GUS staining only in epidermal (Figure 4a) and cortical cells at the site of fungal infection (Figure 4b–d). External hyphae in the vicinity of the root did not activate the SbtM1 promoter in the absence of physical contact. GUS staining was observed along internal hyphae (Figure 4b–d), and cells containing arbuscules showed stronger staining than neighbouring cells (Figure 4c,d). During AM, the spatial distribution patterns of the GUS stain in roots carrying SbtM1, SbtM3, SbtM4 or SbtS promoter–GUS fusions appeared identical (Figure S1). However, different staining times were required to obtain similar GUS staining intensities in transgenic roots containing the SbtM1 (1–3 h) SbtM3 (3–6 h), SbtM4 (5–10 h) or SbtS (3–6 h) promoter–GUS constructs.

Figure 4.

 Histochemical localization of subtilase gene expression during AM and RNS. GUS-stained transgenic roots carrying SbtM1 (a–d), SbtS (e) or SbtM4 (f–h) promoter–GUS fusions. The transgenic roots were stained 7 days after inoculation with Gintraradices (a–d), or 4 (e), 7 (f) and 14 days after inoculation (g–h) with Mloti.
(a) In SbtM1pro:GUS roots, staining was observed in the vicinity of hyphal penetration into the host root.
(b) GUS staining was also observed in epidermal cells and cortical cells around the penetration point of AM fungi and along the internal hyphae (b). The arrow indicates an external hypha stained with black ink (b).
(c, d) Strong GUS staining was observed in cells containing an arbuscule (arrowheads).
(d) Fungal cell walls imaged by wheat germ agglutinin Alexa Fluor 488 (WGA) fluorescence.
(e) SbtSpro:GUS roots inoculated with Mloti exhibited GUS staining in epidermal cells.
(f) Merged image of bright-field (SbtM4pro:GUS root) and red fluorescence of Mloti containing DsRED. The GUS staining was localized to dividing cortical cells of nodule primordia.
(g, h) In nodules, SbtM4pro:GUS was observed in infected cells (g) where Mloti containing DsRED accumulates (h).
Scale bars = 50 μm (a–f) and 100 μm (g, h).

When infected with Mloti, only background staining around the vascular bundle and the root tip resulted from expression of SbtM1pro:GUS and SbtM3pro:GUS constructs, similar to uninfected roots (data not shown). The lack of response to Mloti was consistent with the results of quantitative RT-PCR experiments. Inoculation of transgenic roots carrying SbtSpro:GUS with Mloti induced GUS expression in epidermal cells (Figure 4e). In agreement with the quantitative RT-PCR data (Figure 3b), there was no GUS staining in nodules. Only root hair cells on the nodule surface were sometimes stained. There was a remarkable difference in the behaviour of SbtSpro:GUS roots between the two types of root symbiosis. During AM, GUS staining was strong in the cortex throughout symbiosis (Figure S1d), but expression was transient during RNS and was only observed in epidermal cells (Figure 4e). GUS staining of SbtM4pro:GUS roots was localized to epidermal cells and cortical cells in the vicinity of infection threads (data not shown) in nodule primordia (Figure 4f) and infected cells of the nodule (Figure 4g,h), indicating that activation of SbtM4 coincides with the sites of rhizobial infection.

Localization of SbtM1 in the apoplast

Domain analysis using TargetP software predicted a secretion signal peptide at the N-terminal end of each of the subtilases. In order to test its functionality, the predicted secretion signal peptide in SbtM1 was fused to the YFP variant Venus (Nagai et al., 2002), and the chimeric sequence was expressed under the control of the SbtM1 promoter. Transgenic control roots carrying 35Spro:Venus exhibited fluorescence in the cytosol of epidermal and cortical cells (data not shown). Fluorescence of Venus protein expressed from a SbtM1pro:SbtM1 signal peptide:Venus fusion construct was observed in those parts of the root that were infected with G. intraradices. Expression of the translational SbtM1 promoter–Venus protein fusion without the signal peptide (SbtM1pro:Venus) resulted in fluorescence in plants cells that were in contact with fungal hyphae (Figure S2a–c). At the tissue level, both expression patterns were consistent with the promoter–GUS fusion results (Figure 4c). However, while Venus protein expressed without the signal peptide was exclusively located within the plant cell (Figure S2a–c), presence of the signal peptide led to a drastic alteration of Venus protein localization. In roots expressing the SbtM1pro:SbtM1 signal peptide:Venus fusion construct, weak fluorescence was observed around the cell walls of plant cells infected with fungal hyphae (Figure 5a,b), indicating that Venus protein was secreted, and hence demonstrating the functionality of the SbtM1 signal peptide secretion signal. Fluorescence was detected around intracellular hyphae in outer cell layers (data not shown). Fungal cell walls were stained using wheat germ agglutinin Alexa Fluor 594 (Figure 5d,g), and the localization was compared to that of the SbtM1 signal peptide–Venus fusion protein (Figure 5c,f). Strong fluorescence was observed in cells containing arbuscules in close proximity to fungal cells (Figure 5a,e,h), indicating that the protein accumulated in the intercellular spaces and the PAS. During arbuscule development, a fungal hypha penetrates into a cortical cell and forms several thick branches. Fluorescence of the fusion proteins was observed along the surface of the thick branches (Figure 5c,e). In well-developed arbuscules, the fine branches of the arbuscule exhibited much stronger wheat germ agglutinin Alexa Fluor signal than the thick branches, which showed no or only slight fluorescence (Figure 5g). In marked contrast, the fluorescence resulting from the SbtM1 signal peptide–Venus protein fusion construct was localized around the thick branches found at the base of arbuscules, and only weak fluorescence was observed around the fine branches (Figure 5f,h), which may be a reflection of the apparently larger PAS volume observed around the thick branches (Bonfante and Perotto, 1995; Gianinazzi-Pearson, 1996). A construct expressing the SbtM1 full-length protein fused to Venus protein (SbtM1pro:ORF:Venus) resulted in the same spatial distribution of fluorescence as the fusion with the SbtM1 signal peptide alone (Figure S2d–f). However, only free Venus protein was detected in an anti-GFP Western blot analysis of roots transformed with either the full-length or the signal peptide fusion (Figure S2g). This result is consistent with cleavage of the signal peptide during secretion and subsequent further proteolytic processing of the subtilase–Venus protein fusion. A similar release of GFP has previously been described for the full-length SDD1 subtilase from Arabidopsis (von Groll et al., 2002). Loss of the fluorescent protein tag appears to be a reoccurring problem when studying subtilase localization (van der Hoorn, 2008).

Figure 5.

 Localization of fungal cell walls and of Venus protein targeted by the SbtM1 signal peptide.
Transformed hairy roots carrying SbtM1pro:SbtM1 signal peptide:Venus were inoculated with Gintraradices and observed 7 days after inoculation.
(a–c, f) Venus protein fluorescence images (a, c, f) and bright-field image (b), using epifluorescent (a) and confocal microscopy (c, f).
(d, e, g, h) Images of fungal cell walls stained with wheat germ agglutinin Alexa Fluor 594 (WGA) after sectioning (d, g) merged with Venus protein images (e, h).
Venus protein fluorescence was localized in intercellular spaces (a) and arbuscule-containing cells (a, c, f). Strong Venus protein fluorescence was observed around AM fungal cell walls detected by Alexa Fluor fluorescence in a developing arbuscule (c–e). In a developed arbuscule, Venus protein fluorescence was localized along thick fungal branches and weaker fluorescence surrounded the fine branches (f). Strong Alexa Fluor fluorescence was observed along the fine branches, but not along the thick branches (g, h). Scale bars = 10 μm.

Suppression of SbtM1 or SbtM3 by RNAi impairs AM development

To analyse the function of the AM-specific subtilase genes, we used RNAi to inhibit their expression. Two RNAi constructs targeting SbtM1 or SbtM3 were stably introduced into Ljaponicus (Figure S3), and homozygous lines were identified that showed successful suppression of the target gene during AM.

We performed a detailed phenotypic analysis of one line for each RNAi construct at the macro- and micro-scopic level. Suppression of SbtM1 or SbtM3 was associated with a decrease of fungal colonization in the host root (Figure 6). In particular, a decrease in the level of colonization by arbuscules was especially obvious (Figure 6b). This effect of silencing was specific for AM, as all SbtM RNAi lines exhibited normal root growth (Figure 6d), and we did not notice any other pleiotropic effects of the transgene. SbtM1 and SbtM3 are not expressed during RNS, and therefore the RNAi constructs targeting their expression are not expected to result in a defective phenotype during RNS. In agreement with this, root growth as well as nodule number and morphology upon inoculation with Mloti were indistinguishable from those of the wild-type (Figure S4b).

Figure 6.

 AM phenotype of SbtM1 and SbtM3 RNAi lines.
(a–c) Colonization of wild-type (WT), SbtM1 RNAi and SbtM3 RNAi lines by hyphae (a), arbuscules (b) and vesicles (c). Plants were co-cultured with Gintraradices in chive nurse pots for 2 weeks. AM colonization of the main root was determined using the line intersect method, and is given as the percentage of root length colonized (%). The RNAi lines showed a significant decrease in hyphal and arbuscular colonization compared to WT.
(d) Mean root lengths of SbtM1 and SbtM3 RNAi lines were almost identical after 2 weeks of co-culture with Gintraradices. In a–d, one representative result (4 plants per sample) of three independent experiments is shown. Asterisks indicate significant (< 0.05) differences from WT colonization.
(e) Transcript levels of SbtM1, SbtM3, SbtM4, SbtS and PT4 in SbtM1 and SbtM3 RNAi lines compared with WT (=100%) after 2 weeks of co-culture with G. intraradices. Error bars indicate standard deviation for three independent experiments, each comprising 4–8 plants per sample.

To examine the symbiotic phenotype at the molecular level, we analysed gene expression levels during AM and nodulation in the SbtM1 and SbtM3 RNAi lines. Consistent with the unaltered nodulation behaviour of these lines in response to M. loti, induction of the nodulins NIN and ENOD40 in the SbtM1 and SbtM3 RNAi lines was comparable to that in the wild-type. Moreover, the nodulation-induced subtilases SbtM4 and SbtS showed unaltered transcript accumulation in the SbtM1 and SbtM3 RNAi lines (Figure S4a). These results confirm the AM-specific phenotypic effect of the RNAi lines. Importantly, they also demonstrate that SbtS and SbtM4 are not subject to off-target silencing by the RNAi constructs.

During AM, we followed the expression of SbtM1 and SbtM3 as well the marker genes SbtM4, SbtS and LjPT4 (AP010874) (Figure 6e). The phosphate transporter gene MtPT4 is specifically expressed in arbuscule-containing cells and constitutes a useful marker gene for this AM-specific cell type (Harrison et al., 2002). We therefore isolated a putative Lotus orthologue of MtPT4 by searching for phosphate transporter genes within the L. japonicus genome (Sato et al., 2008). The most closely related phosphate transporter sequence was called LjPT4 and exhibited 78% identity at the nucleotide level. Like MtPT4, expression of this gene was induced during AM (Figure 6e).

The transcript levels of all of the AM-induced genes were significantly reduced in the SbtM1 or SbtM3 RNAi lines, which we interpret as a molecular manifestation of the reduced fungal hyphae and arbuscule colonization levels in the RNAi lines. However, SbtM1 and SbtM3 transcripts were detectable at a level of 10–30% in the RNAi plants infected with G. intraradices. This remaining level of gene expression could be responsible for the observed incomplete suppression of AM in the RNAi plants. These gene expression data, in combination with the results of the microscopic analysis, demonstrate that SbtM1 and SbtM3 genes play a specific role in AM development.

Vesicle colonization also decreased in some RNAi plants, but there was no significant difference compared with wild-type (Figure 6c). In SbtM1pro:GUS and SbtM3pro:GUS roots, formation of vesicles was not correlated with expression of SbtM1 or SbtM3. Therefore, any decrease of vesicle number in the RNAi lines is likely to be a consequence of the overall reduction of fungal colonization.

In order to investigate whether the decrease in AM colonization was caused by early senescence of symbiotic structures or inhibition of their formation, we analysed AM development in time-course experiments using the wild-type and the SbtM1 RNAi-1 line (Figure S5). We measured AM colonization in roots at 1, 2 and 4 weeks after inoculation with Gintraradices. A decreased level of AM structures was already observed at 1 week after inoculation with Gintraradices. This early reduction suggested that the decrease in AM structures was caused by an impairment of fungal infection or AM development. To increase the resolution of the phenotypic analysis, we determined the effect of SbtM1 and SbtM3 suppression on external hyphae or the attachment of hyphae to epidermal cells. Fungal infection points on the root surface were inspected and counted, but no significant difference in shape or quantity between wild-type and the suppression lines could be detected (Figure S6). In particular, we did not observe aberrant balloon-like deformations of fungal hyphae at the site of infection, which are the hallmark of some AM-defective common symbiosis mutants (Demchenko et al., 2004; Kistner et al., 2005), and were also observed in SYMRK RNAi lines (data not shown). These results indicate that suppression of SbtM1 and SbtM3 affects the development of AM after fungal recognition and successful infection attempts.

To test whether the observed phenotypes were indeed caused by the transgenes, we followed segregation of the SbtM3 RNAi construct in the T2 progeny of primary transformants (T1), and analysed the AM phenotype of individual transgenic and non-transgenic siblings. Transgenic plants with or without the transgene were identified among a population of T2 plants using GFP fluorescence as a co-transformed marker on the T-DNA. The SbtM3 RNAi-1 lines carried a single transgene insertion in the genome as evidenced by the segregation pattern of the GFP fluorescent marker (45 plants with GFP fluorescence versus 10 plants without; χ2 = 1.36, P = 0.05). In four independent experiments T2 siblings with the transgene showed reduced arbuscule colonization compared to their non-transgenic siblings, suggesting that the RNAi construct caused the phenotype (Table 1).

Table 1.   Co-segregation of SbtM3 RNAi construct and the AM phenotype
 HyphaeArbusculeVesicleTotal number of plants
  1. AM colonization in the transgenic [SbtM3 RNAi-1 (+)] and non-transgenic [SbtM3 RNAi-1 (−)] siblings are compared with wild-type colonization (100%). Values are means ± standard deviation (= 4 independent experiments). The asterisk indicates a significant (< 0.05) difference from wild-type and SbtM3 RNAi-1 (−) colonization.

SbtM3 RNAi-1 (+)69.1 ± 21.538.1 ± 22.2*69.2 ± 69.624
SbtM3 RNAi-1 (−)91.7 ± 8.586.4 ± 18.2111.6 ± 83.619


SbtM1 and SbtM3 are required for AM development

Here we investigated the relevance of two AM-specific subtilases, SbtM1 and SbtM3, for AM development by negatively interfering with their expression through RNAi. A detailed phenotypic analysis revealed that plants in which SbtM1 or SbtM3 were successfully targeted for suppression by RNAi showed decreased levels of hyphal colonization of the roots and arbuscule frequency (Figure 6). SbtM1 or SbtM3 suppression did not affect the stages prior to root infection, as the number of fungal penetration attempts into outer cell layers was not altered. Our localization data show that SbtM1 is targeted for secretion, and is localized in the apoplastic spaces, including the peri-fungal spaces and the PAS. The presence of secretion signal peptides at their N-terminal ends predicts similar apoplastic localization for all subtilases of this study. The intercellular space and the PAS are parts of the interface between the symbiotic partners, at which signal and nutrient exchange are likely to occur. Therefore, the enzyme composition and physicochemical conditions within the interfacial compartment are likely to influence the development and efficiency of AM. The combined phenotypic and subcellular localization analysis indicates that SbtM1 and SbtM3 play an indispensable role during the development of intracellular infection structures, most prominently the arbuscules. The predicted proteolytic activity of SbtM1 and SbtM3, together with their specific expression and localization pattern, suggest that cleavage of a specific substrate present in the peri-fungal space is essential for proper arbuscule development.

Predicted substrates of AM-induced subtilases

Accumulating evidence indicates that plant proteases play important roles in determining the outcome of plant–microbe encounters. For example, the tomato cysteine protease RCR3 is an important target for protease inhibitors secreted by the phytopathogenic fungus Cladosporium fulvum (Rooney et al., 2005; Shabab et al., 2008). At the same time, RCR3 is subject to surveillance by the plant disease resistance protein Cf2, which probably detects structural changes associated with RCR3 binding to AVR2 (Krüger et al., 2002). Similarly, protease inhibitors of the phytopathogenic fungus Phytophora infestans target the SbtM-related P69B subtilase in tomato (Tian et al., 2005). At present, the substrates of these proteases are unknown. Proteases can convert precursor proteins into bioactive forms such as peptide hormones (Leibowitz and Wickner, 1976; Chasan and Anderson, 1989; Nakayama, 1997; Steiner, 1998) or cleave specific targets to inactivate them (Paton et al., 2006). Such selective proteases comprise key regulators of cellular development and responses to the environment (Seidah and Chretien, 1999). It is therefore possible that some of the plant extracellular proteases are involved in the generation or clearance of peptides with signalling quality. The developmental defects observed in the subtilase mutants ale1 or sdd1 of A. thaliana may indicate a direct or indirect signalling function of their products (Tanaka et al., 2001; von Groll et al., 2002). By analogy, the AM-induced subtilases may participate in symbiotic communication by processing signalling peptides or their precursors.

Alternative hypothetical substrates with more structural role have been proposed in a study of AIR3, the closest homologue of SbtS in A. thaliana. AIR3 was proposed to weaken cell–cell connections by degradation of structural proteins in the apoplastic space, thus facilitating emergence of a lateral root (Neuteboom et al., 1999). Cleavage of structural proteins within and between plant cell walls by SbtM1 and SbtM3 may be required for elongation of fungal hyphae in the intercellular space and for penetration into the host cell during arbuscule formation. Due to their localization in the interface with the fungal partner, potential substrates include proteins from both fungus and plant. Isolation of the target substrate will be a key step towards understanding the molecular functions of SbtM1 and SbtM3. However, the lack of successful strategies for identification of substrates is currently the limiting factor in understanding the function of plant proteases in general (van der Hoorn, 2008).

AM-induced subtilases are secreted at the symbiotic interface

The vast majority of Arabidopsis subtilases carry signal peptides, which, in the absence of retention signals elsewhere in the protein sequence, predict an extracellular localization (Rautengarten et al., 2005). Likewise, all subtilases in this study carry a predicted signal peptide at their N-terminus. Signal peptides typically target peptide synthesis to the rough ER, and are cleaved of during translation into the ER lumen. We tested the functionality of the SbtM1 signal peptide by fusing it to the N-terminus of Venus protein. Weak Venus protein fluorescence was observed in the apoplast around infected cells. This observed localization within the plant cell wall is fully consistent with the predicted targeting by the SbtM1 signal peptide to the secretory pathway. However, stronger signals were detected around fungal hyphae traversing plant cells and in the PAS (Figure 5). Our data suggest that the trafficking of proteins towards the PAS follows the canonical secretory pathway. The strong accumulation of Venus protein in the PAS may be explained by two factors. First, due to the small volume of the PAS, there is less dilution of proteins compared to the proteins secreted at the cell wall. Second, the strong increase in membrane surface during arbuscule development may cause a significant portion of trans-Golgi vesicles to fuse with the peri-arbuscular membrane, and hence redirect a significant portion of proteins that enter the secretory pathway to the PAS. This model does not exclude the existence of additional targeting signals that specifically target proteins to or specifically exclude proteins from the PAS. However, our sequence analysis did not reveal any additional subcellular targeting signals or transmembrane domains within the SbtM protein.

Genetic redundancy and cross-silencing between SbtM1 and SbtM3

Although SbtM1 and SbtM3 share only 66% overall amino acid sequence identity (Figure 1b), we cannot exclude the possibility that SbtM1 and SbtM3 or other subtilase genes have redundant functions. However, the enhanced sequence divergence between the PA domains of SbtM1 and SbtM3 (Figure 1b), which are predicted to mediate protein–protein interaction and selection of target substrates (Mahon and Bateman, 2000), may reflect distinct substrate preference.

In both RNAi lines, in addition to reduced induction of the target SbtM gene, we observed a lower expression of AM-regulated marker genes. We interpret this reduced expression as a molecular manifestation of the reduced arbuscular colonization level caused by the RNAi construct. However, for SbtM1 and SbtM3, there is at least a theoretical possibility that the two RNAi constructs targeted both of these sequence-related genes to some extent. RNAi can interfere with the expression of sequence-related genes through production of small RNA species (20–26 nt) that promote cleavage of RNAs that share close sequence identity (Brodersen and Voinnet, 2006). SbtM1 and SbtM3 share two regions with more than 21 nt identity (Figure S3). It is therefore possible that RNAi for either gene caused suppression of both SbtM1 and SbtM3 expression. While we could not determine the degree of cross-silencing between SbtM1 and SbtM3, we can rely on SbtM4 and SbtS as molecular markers for AM development in these experiments as they were certainly not directly targeted by the RNAi constructs, as evidenced by their unaltered transcript accumulation during nodulation in the SbtM1 and SbtM3 RNAi lines (Figure S4a).

Genetic redundancy between the SbtM1/SbtM3 pair and the nodulation-induced subtilases SbtM4 and SbtS is less likely because the timing of expression during AM is distinct (Figures 2 and 3), and the levels of SbtM1 and SbtM3 transcript levels are often below the detection limit during RNS. Also, SbtS and SbtM4 are likely to have distinct roles as they differ significantly with regard to the kinetics of gene expression; in contrast to SbtM4 (Figure 3a), SbtS transcripts do not accumulate in mature nodules and are only transiently expressed during the early stages of nodulation (Figure 3b). These distinct expression patterns suggest that the group of encoded subtilases, as a whole, influence multiple steps of symbiotic development.

Experimental procedures

Plant growth

Lotus japonicus Gifu B-129 and Miyakojima MG-20 seeds were sterilized with sodium hypochlorite for 10 min and germinated on 0.8% Bacto Agar (Difco, http://www.bd.com) plates. The plants were grown in a growth chamber (24°C, 16 h light/8 h dark) or in a glasshouse. The seedlings were transferred to symbiont inoculation substrate 2–3 days after germination. AM fungus Gintraradices was inoculated into the plants using a chive (Allium schoenoprasum) nurse pot system (Demchenko et al., 2004). Mesorhizobium loti MAFF303099 wild-type or carrying DsRED was inoculated using autoclaved expanded clay particles (Seramis, Mars GmbH, http://www.botanikpflanzen.de) and half-strength B&D medium (Broughton and Dilworth, 1971).

Expression analysis

Total RNA was extracted from roots using a NucleoSpin® RNA plant kit (Macherey-Nagel, http://www.macherey-nagel.com). Synthesis of first-strand cDNA and real-time PCR were performed using a SuperScript III Platinum® two-step quantitative PCR kit with SYBR® Green (Invitrogen, http://www.invitrogen.com/) and an iCycler (Bio-Rad, http://www.bio-rad.com/) according to manufacturer’s instructions. Real-time PCR primers were designed using Primer3 (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi) (primer pairs 1, 7, 10, 12, 13, 14, 15, 16 and 17 in Table S1). Ubiquitin and elongation factor (EF) 1α transcripts were used as a reference. The PCR efficiency of the primers was determined for each primer set. Fifty nanograms of total RNA were used for reverse transcription in a 20 μl reaction mixture, and 1 μl of the reverse transcription product was added to a 20 μl real-time PCR reaction mixture with 50 nm fluorescein (Invitrogen) as a reference dye. Three technical replicates were prepared per sample and experiment. Thermal cycler conditions were 94°C for 5 min, and 40 cycles of 94°C for 30 sec, 58°C for 30 sec and 72°C for 30 sec. Transcript expression levels were calculated using the PCR efficiency, and normalized to the level of ubiquitin transcript. At least three biologically independent experiments were performed and the data used to calculate means and standard deviations.

Total proteins extracted from transgenic roots containing 35Spro:Venus (10 μg), SbtM1pro:SbtM1 signal peptide:Venus (150 μg) or SbtM1pro:SbtM1 ORF:Venus (200 μg) were separated by 12% SDS–PAGE, and blotted onto Hybond-P membranes (Amersham, http://www5.amershambiosciences.com/). The blotted proteins were detected using anti-GFP antibody (Roche, http://www.roche.com), horseradish peroxidase-labelled secondary antibody (anti-mouse; Santa Cruz, http://www.scbt.com) and an ECL advanced Western blotting detection kit (Amersham).

Preparation of fusion and RNAi constructs

The promoter regions of SbtM1 (−688 to +7 bp, start codon +1 to +3 bp), SbtM3 (−1030 to +4 bp) and SbtM4 (−1034 to +4 bp) were amplified using specific primers (primer pairs 2, 8 and 11 in Table S1), and cloned into the pKGWFS7 GFP–GUS fusion vector (Karimi et al., 2002). The Venus protein fusion vector was constructed by modifying pCAMBIA1300 (http://www.cambia.org/). The Venus gene (35Spro:Venus:35Sterm) was integrated into pCAMBIA1300 between the HindIII and EcoRI sites, resulting in pCAMBIA.Venus. A DNA fragment containing the SbtM1 promoter (−688 bp) and the signal peptide region (36 amino acids from the N-terminus) or full-length ORF (750 amino acids) of SbtM1 was amplified using specific primers comprising SphI (forward) and NcoI or SalI (reverse) adaptor sequences (primer pairs 3, 4 and 5 in Table S1). The amplicons were cloned between the SphI and NcoI or SalI sites of pCAMBIA.Venus to replace the 35S promoter. The eGFP co-transformation marker gene (rolDpro:eGFP:35Sterm) was cloned into the AatI site of the pK7GWIWG2(I) post-transcriptional silencing vector (Karimi et al., 2002). The coding sequences of SbtM1 (360 bp) and SbtM3 (522 bp) were amplified using specific primers (primer pairs 6 and 9 in Table S1), and cloned into the modified pK7GWIWG2(I) plasmid containing the eGFP marker gene.

Transformation of L. japonicus

Lotus japonicus MG20 was transformed using Agrobacterium tumefaciens AGL1 containing SbtM1 or SbtM3 RNAi binary vector constructs. For stable transformation, callus formation and re-differentiation were induced from infected hypocotyls (Kato et al., 2005). T1 transgenic plant lines were used in all phenotype and expression analyses in this study. Transgenic hairy roots were induced using Agrobacterium rhizogenesAR1193. Arhizogenes colonies were suspended in sterilized water and spread on an autoclaved filter paper. Roots of 6-day-old seedlings (grown for 3 days in the dark and for 3 days in 16 h light/8 h dark, 24°C) were cut off below the hypocotyl on the filter paper. The infected shoots were incubated for 5 days in a growth chamber (24°C, 16 h light/8 h dark) on B5 medium (Sigma, http://www.sigmaaldrich.com/) solidified with 1.0% bactoagar, and subsequently transferred to the same medium containing 300 μg ml−1 claforan (Sanofi-Aventis, http://en.sanofi-aventis.com) for 10 days to 3 weeks until bacterial growth was suppressed.

AM phenotyping

For AM phenotypic analysis, main roots inoculated with G. intraradices were stained with ink, and internal hypha, arbuscule and vesicle colonization were quantified using the line intersect method (McGonigle et al., 1990). To analyse segregation of the transgene and the phenotype in the SbtM3 RNAi-1 line, plants were grown in chive nurse pots side-by-side with non-transformed Ljaponicus Gifu plants. After 1, 2 and 4 weeks, the presence of the transgene was scored by GFP fluorescence, and the transgenic and non-transgenic/non-transformed control plants stained for AM structures. The levels in the T1 siblings were calculated relative to the level in the non-transformed control plants. The Mann–Whitney U-test was used to compare colonization between wild-type and RNAi lines.

GUS staining

Transgenic roots containing promoter–GUS constructs were inoculated with Gintraradices or Mloti and stained with GUS staining buffer (0.5 mg ml−1 X-Gluc, 100 mm phosphate buffer pH 7.0, 100 mm EDTA, 0.5 mm K4[Fe(CN)6], 0.5 mm K3[Fe(CN)6] and 0.1% Triton X-100) at 37°C. Fungal hyphae were stained with ink (Demchenko et al., 2004). For double staining, ink staining was performed after GUS staining.

Fluorescence microscopy

Wheatgerm agglutinin Alexa Fluor 488 or 594 (Invitrogen) were used for fluorescent staining of fungal cell walls (Harrison et al., 2002). Transgenic roots containing the Venus protein fusion constructs were embedded in 5% agarose without fixing, and sectioned using a VT1000S vibrating-blade microtome (Leica, http://www.leica.com). The sectioned roots (80–100 μm) were stained with 1 μm wheatgerm agglutinin Alexa Fluor 594 in PBS for >20 min at room temperature. Fluorescence microscopy was performed using an inverted microscope (DMI6000B; Leica) or a confocal microscope (SP5; Leica) equipped with 40× dry objective (numerical aperture 0.6) or a 63× water immersion objective (numerical aperture 1.2). Acquired images were analysed using the Leica LAS-AF software package or Adobe Photoshop (Adobe, http://www.adobe.com).


We are grateful to Sonja Kosuta (present address: Agriculture and Agri-Food Canada, Ontario, Canada) for providing transgenic seeds containing the SbtS promoter–GUS fusion. This work was supported by the Priority Program ‘Molecular Basis of Mycorrhizal Symbioses (MolMyc)’ of the Deutsches Forschungsgemeinschaft. N.T. was supported by a Postdoctoral Fellowship for Research Abroad awarded by the Japan Society for the Promotion of Science.

The Genbank accession numbers for sequences of TAC or BAC clones carrying the Lotus japonicusSbtM1/SbtM2/SbtM3 cluster, SbtM4/SbtM5 cluster and Lotus japonicus PT4 are AP009544, AP009542 and AP010874, respectively.