A tandem Kunitz protease inhibitor (KPI106)–serine carboxypeptidase (SCP1) controls mycorrhiza establishment and arbuscule development in Medicago truncatula


For correspondence (e-mail Natalia.requena@kit.edu).


Plant proteases and protease inhibitors are involved in plant developmental processes including those involving interactions with microbes. Here we show that a tandem between a Kunitz protease inhibitor (KPI106) and a serine carboxypeptidase (SCP1) controls arbuscular mycorrhiza development in the root cortex of Medicago truncatula. Both proteins are only induced during mycorrhiza formation and belong to large families whose members are also mycorrhiza-specific. Furthermore, the interaction between KPI106 and SCP1 analysed using the yeast two-hybrid system is specific, indicating that each family member might have a defined counterpart. In silico docking analysis predicted a putative P1 residue in KPI106 (Lys173) that fits into the catalytic pocket of SCP1, suggesting that KPI106 might inhibit the enzyme activity by mimicking the protease substrate. In vitro mutagenesis of the Lys173 showed that this residue is important in determining the strength and specificity of the interaction. The RNA interference (RNAi) inactivation of the serine carboxypeptidase SCP1 produces aberrant mycorrhizal development with an increased number of septated hyphae and degenerate arbuscules, a phenotype also observed when overexpressing KPI106. Protease and inhibitor are both secreted as observed when expressed in Nicotiana benthamiana epidermal cells. Taken together we envisage a model in which the protease SCP1 is secreted in the apoplast where it produces a peptide signal critical for proper fungal development within the root. KPI106 also at the apoplast would modulate the spatial and/or temporal activity of SCP1 by competing with the protease substrate.


Plant proteases and protease inhibitors are key components of cellular development, necessary not only to control protein turnover but also to make nutrients available, to protect the cell against attack and to produce developmental cues at specific times or locations (Ryan, 1990; van der Hoorn, 2008; Martinez et al., 2012). A subset of proteases and inhibitors are in addition devoted to controlling the interactions of plants with microorganisms, and the number of examples illustrating this is steadily increasing. An interesting example of this tandem relationship has been recently reported in Zea mays during the interaction with the pathogenic fungus Ustilago maydis. There, the plant protease inhibitor cystatin CC9 is transcriptionally induced upon penetration by U. maydis strains secreting the effector protein pep1 (van der Linde et al., 2012a). Induction of CC9 leads to the inhibition of several apoplastic papain-like cysteine proteases. This inhibition is essential for pathogenicity, as silencing of CC9 led to a strongly reduced fungal colonization. Therefore, the protease inhibitor CC9 can be defined as a plant compatibility factor essential to suppress host immunity and to cause disease (van der Linde et al., 2012a,b). Another prominent example of a protease–protease inhibitor pair playing a key role during plant–microbe interactions is the tandem Rcr3 and Avr2. The secreted protein effector Avr2 from the fungus Cladosporium fulvum, but also interestingly the effector GrVAP1 from the nematode Globodera rostochiensis, have been shown to inhibit Rcr3, a secreted cysteine protease from tomato (Kruger et al., 2002; Rooney et al., 2005; Lozano-Torres et al., 2012). Although Avr2 and GrVAP1 are not described as canonical protease inhibitors, they mediate Rcr3 inhibition by binding to the active site of the protease, inactivating its enzymatic activity, and thus they can be considered as protease inhibitors (Rooney et al., 2005; Lozano-Torres et al., 2012). Interestingly, and in contrast to other effectors from Phytophthora infestans also targeting Rcr3 (Song et al., 2009), inhibition of Rcr3 by Avr2 and GrVAP1 allows microbial colonization only in tomato plants lacking the R-protein Cf-2. This indicates that in those latter cases the extracellular plant immune receptor Cf-2 guards Rcr3 and mediates disease resistance upon modifications of the protease (Kruger et al., 2002; Lozano-Torres et al., 2012) while the modifications/inhibitions produced by the effector proteins of P. infestans are not monitored by Cf-2 (Song et al., 2009).

Symbiotic plant–fungal associations share several common features with those of plant–fungal pathogens, including the use of effector proteins to counteract plant defence responses (Kloppholz et al., 2011; Plett et al., 2011). However, less is known about the role of protease–protease inhibitor interactions during plant–fungal symbioses. The arbuscular mycorrhiza (AM) symbiosis is a beneficial biotrophic interaction between plant roots and fungi of the Glomeromycota phylum (Schüßler et al., 2001). The AM fungi enter the root using hyphopodia formed on the rhizodermis and colonize the root cortex inter- and intracellularly. There, the fungal hyphae invaginate the cytoplasm to form highly branched structures, the arbuscules, that serve as the main site of nutrient exchange between the symbionts (Harrison, 2012). The recognition of a suitable symbiotic partner and initiation of the symbiosis is carried out by a reciprocal signal exchange between the plant and the fungus (Bonfante and Requena, 2011). Thereupon, drastic alterations in plant gene expression follow in order to trigger reorganization processes within the cortical cells required for accommodation of the entering fungus and to prepare those cells for the nutrient exchange (Liu et al., 2003, 2007; Brechenmacher et al., 2004; Kuster et al., 2004; Guimil et al., 2005; Hohnjec et al., 2005; Gomez et al., 2009). Transcriptomic studies showed that several plant protease genes are specifically induced in M. truncatula, Lotus japonicus, rice and petunia when engaged in AM symbiosis (Liu et al., 2003; Guimil et al., 2005; Kistner et al., 2005; Breuillin et al., 2010). For example, the expression of SCP1, encoding a serine carboxypeptidase, is localized to arbusculated and adjacent cortical cells (Liu et al., 2003). Therefore, SCP1 serves as mycorrhizal marker gene and SCP1pro:GFP has been used to visualize cellular reorganization processes in cortical cells during arbuscule development by co-expression with different organelle markers (Pumplin and Harrison, 2009). In L. japonicus, Takeda et al. (2009) showed that silencing of LjSbtM1, a secreted mycorrhiza-specific subtilase, led to a significant decrease of arbuscule abundance. Furthermore the authors showed that LjSbtM1 localizes to the periarbuscular space, indicating that proteolytic cleavage at the plant–fungal interface might be required for arbuscule development and/or function.

Besides these proteases, transcriptomic analyses during AM symbiosis have also repeatedly reported the up-regulation of several Kunitz protease inhibitors (Wulf et al., 2003; Grunwald et al., 2004; Hohnjec et al., 2005; Liu et al., 2007), suggesting they might also play an important role. Recently, in a screen for M. truncatula genes induced in response to diffusible signals produced by AM fungi, we also identified two genes, TC106351 and TC100804, annotated as putative Kunitz protease inhibitors (KPIs) (Kuhn et al., 2010). Kunitz protease inhibitors are small proteins with a molecular mass of 18–22 kDa that predominantly contain two conserved disulphide bonds and one reactive site. Kunitz protease inhibitors have been shown to be mainly active against serine proteases, but also against cysteine and aspartic proteases, whereas no inhibition of metalloproteases has yet been reported (Araujo et al., 2005; Zhou et al., 2008). Since the discovery of the soybean trypsin inhibitor (STI) by Kunitz in 1947, many KPIs have been characterized. Different plant storage tissues, especially leguminous seeds, contain a high content of KPIs where they carry out various functions. As storage proteins, KPIs are rapidly degraded after germination, thereby providing nutrients for survival (Candido Ede et al., 2011). Kunitz protease inhibitors also serve as protection against herbivores by interfering with their proteolytic digestive processes (Srinivasan et al., 2005; Oliva et al., 2011). But interestingly, several KPIs are induced under various stress conditions, including pathogen attack and during nodulation, pointing to a function during biotic and abiotic stresses (Valueva et al., 1998; Haruta et al., 2001; Lievens et al., 2004; Ledoigt et al., 2006; Jimenez et al., 2008; Li et al., 2008; Huang et al., 2010; Odeny et al., 2010).

The co-induction of proteases and protease inhibitors during mycorrhiza formation prompted us to functionally characterize the mycorrhiza-specific induced genes TC106351 and TC100804 encoding secreted KPIs and their potential target proteases in M. truncatula during symbiosis. Here, we show that the protease inhibitor KPI106 and the protease SCP1 interact and are key to the control of fungal growth and morphogenesis within the cortex. And thus, that protease–protease inhibitor interactions are not just an important feature shaping the fate of plant–pathogen associations, but are also required during the interactions of plant with mutualistic AM fungi.


Kunitz protease inhibitors are induced during mycorrhiza formation

TC106351 and TC100804 encode two putative KPIs, of 205 and 201 amino acid (aa) residues, respectively, herein named as KPI106 and KPI104. Both genes have been several times reported as being induced during mycorrhiza formation (Wulf et al., 2003; Grunwald et al., 2004; Hohnjec et al., 2005; Liu et al., 2007), including very early stages of the symbiosis (Kuhn et al., 2010). We analysed in detail the expression kinetics of KPI106 and KPI104 during colonization of M. truncatula roots by Rhizophagus irregularis. Quantitative real-time PCR results showed that both transcripts accumulate parallel to the development of the fungus within the root, with KPI106 having higher expression levels throughout all time points analysed (Figure 1a). Furthermore, KPI106 is expressed at low levels in non-mycorrhizal roots while KPI104 is not. A detailed sequence analysis of both genes in the recently released M. truncatula genome (Medicago hapmap3.5, http://www.medicagohapmap.org/) revealed that KPI106 and KPI104 belong to a cluster within a large family of KPIs (Figure 1b). In silico analysis of expression using the M. truncatula Gene Atlas (MtGEA; http://mtgea.noble.org/v3/) showed that only the genes within this cluster are mycorrhiza-specific. All deduced protein sequences consist of an N-terminal secretion signal peptide and a soybean trypsin inhibitor (STI) domain of protease inhibitors (Pfam PF00197). Alignment of the mycorrhiza-specific KPIs showed that they contain six conserved cysteines, indicative of three putative disulphide bonds in contrast to most described KPIs that contain two disulphide bonds, as for example the Kunitz STI (Figure 1c, Rawlings et al., 2004; Oliva et al., 2010).

Figure 1.

Expression and sequence analysis of Medicago truncatula mycorrhiza-induced Kunitz protease inhibitors (KPIs).

(a) Time course expression analysis of KPI106 and KPI104 during mycorrhiza symbiosis in M. truncatula root organ cultures ‘ARqua1’ colonized with Rhizophagus irregularis. Bars represent standard deviations of three independent biological replicates. Values are normalized to M. truncatula tef.

(b) Phylogenetic tree of M. truncatula KPIs according to the hapmap3.5 database. Expression of each KPI gene was analysed in the Gene Expression Atlas and revealed a cluster of mycorrhiza-induced genes (black dots). Numbers are bootstrap supported values.

(c) Protein domain structure of M. truncatula mycorrhiza-induced KPIs including an N-terminal secretion signal peptide (SP) and a C-terminal soy bean trypsin inhibitor (STI) domain. All KPIs contain three putative conserved disulphide bonds (black bars), whereas most of the typical STIs only have two (grey).

Overexpression of mycorrhiza-specific KPIs leads to aberrant arbuscule formation

To investigate the function of KPI106 and KPI104 during the mycorrhiza symbiosis we first used an overexpression approach and expressed both protease inhibitors in M. truncatula hairy roots under control of the cauliflower mosaic virus (CaMV) 35S promoter. Overexpression was confirmed by PCR on cDNA of non-mycorrhizal roots and results showed that KPI106 and KPI104 transcripts were present in all analysed samples, in contrast to non-mycorrhizal control roots carrying an empty vector (Figure 2a and Figure S1 in Supporting Information). Interestingly, constitutive overexpression of both protease inhibitors led to an aberrant mycorrhizal phenotype with a remarkably large number of malformed arbuscules and an increased number of septated hyphae (Figure 2b–d). Quantification of fungal colonization confirmed the morphological observations, and showed that the overall abundance of mature arbuscules (a%) in the overexpression lines was significantly reduced when compared with the control line (Figure 2e). In contrast, the frequency of mycorrhization (F%) was similar in all lines, although hyphae in overexpression lines were more often septated. Given that although they are fewer in number fully matured arbuscules can still be found at 17 days post-inoculation (dpi) in KPI-overexpressing roots, a possible hypothesis is that the corresponding proteins could be involved not only in the process of arbuscule formation but also in their turnover. In addition, the occurrence of more septated hyphae in KPI-overexpressing roots points to a general program controlling fungal growth within the cortex. To get more insight into the function of the mycorrhiza-specific protease inhibitors, a loss-of-function approach to silence KPI106 and KPI104 by RNAi was employed. However, reduction of KPI106 and KPI104 transcripts did not result in an obvious effect on the symbiosis (Figure S2a–e). Given the redundancy of mycorrhiza-specific KPIs in M. truncatula (Figure 1b), the absence of a phenotype in KPI104 and KPI106 RNAi lines could be attributed to functional compensation. In support of this, in RKPI106 lines, no changes in the expression of KPI104 and KPI111 could be observed (Figure S2f). Alternatively, if the function of mycorrhiza-specific KPIs were – as suggested from the overexpression experiments – related to arbuscule turnover and control of fungal growth within the cortex, it is possible that KPI inactivation does not produce a noticeable phenotype.

Figure 2.

Overexpression of Kunitz protease inhibitors (KPIs) leads to a disturbed arbuscular mycorrhiza symbiosis.

(a) Confirmation of overexpression of 35Spro:KPI106 and 35Spro:KPI104 root lines by reverse transcriptase PCR. Amplicons of KPI106 (146 bp) and KPI104 (392 bp) transcripts are detected in non-mycorrhizal overexpression lines but not in the control roots.

(b–d) Visualization of fungal structures using WGA-fluorescein in control lines (b), 35Spro:KPI106_1 (c), and 35Spro:KPI104_1 (d), 17 days post-inoculation (dpi). Many malformed arbuscules are visible in the overexpression lines (red arrowhead) and intraradical hyphae show septa (white arrowhead).

(e) In 35Spro:KPI106_1 and 35Spro:KPI104_1 the abundance of mature arbuscules (a%) is significantly reduced compared with the control roots, whereas the frequency of mycorrhization (F%) is not altered. **P-value <0.01; *P-value <0.05; n, number of root fragments.

Target identification of KPI106 and KPI104

To get information about the protease targets of KPI106 and KPI104, a non-directed Gal4 reporter-based yeast two-hybrid assay was carried out. We used the KPI106 bait protein to screen a cDNA library of M. truncatula roots colonized by R. irregularis and identified 156 positive clones. The intensity of the interaction using the α-Gal reporter expression and the presence of an N-terminal secretion signal peptide were used as criteria to select putative interaction partners. Results revealed a strong interaction between KPI106 and a clone encoding 150 aa of the C-terminus of a secreted M. truncatula cysteine protease (CP; TC133093; Figure S3a,b). The cDNA sequence of CP without a signal peptide was tested for interaction with KPI106 as well as with KPI104, and in both cases a positive interaction was found (Figure 3). However, a more detailed sequence analysis of CP revealed the presence of the –NPIR– vacuolar sorting motif within the N-terminal part (Figure S3c; Ahmed et al., 2000; Richau et al., 2012), and thus the interaction between the KPI106, KPI104 and CP in the apoplast is unlikely. However, we cannot exclude that they interact, as both proteins pass through the secretory pathway.

Figure 3.

Individual interaction patterns of Kunitz protease inhibitor (KPI) family members with mycorrhiza-induced proteases.

Mycorrhiza-induced KPIs show individual interaction patterns with the mycorrhiza-induced proteases SCP1 and the Medicago truncatula homologue of LjSbtM1. Controls include interactions of KPI-bait proteins with the Gal4 activation domain (Gal4-AD) and the previously identified cysteine protease (CP). As a control inhibitor KPIc did not interact with any of the tested proteases, confirming specificity of the interactions. Interaction of the yeast proteins RecT and p53 served as a positive control.

In parallel to the non-targeted approach, we considered the possibility that the mycorrhiza-specific described proteases SCP1 (serine carboxypeptidase1; Liu et al., 2003) and SbtM1 (subtilase M1; Takeda et al., 2009) could be targets of KPI106, KPI104 or KPI105 and KPI111. In a direct yeast two-hybrid assay, SCP1 and the M. truncatula closest homologue of SbtM1 were analysed. As a positive control for the yeast two-hybrid assay we included the CP identified in our experiments. To rule out unspecific interactions between the mycorrhiza-specific KPIs and the target proteases, we used a Kunitz inhibitor control (KPIc, Medtr6g009650.1). According to the MtGEA, KPIc is a Kunitz inhibitor not expressed during mycorrhiza symbiosis but in embryogenesis. Furthermore, we tested all KPI bait proteins for interaction with the Gal4 activation domain (Gal4-AD) to exclude self-activation of the reporter genes (Figure 3). Interestingly, and despite the high sequence similarity among the KPIs, we found that individual interaction affinities could be observed for each one. Thus, while KPI106 showed a stronger affinity for SCP1 than for CP, KPI111 only interacted with SCP1. In contrast, KPI104 only interacted with CP while KPI105 did not interact with any of the analysed proteases. None of the KPIs interacted with SbtM1. Moreover, KPIc did not interact with any of the proteases, supporting the specificity of the identified interactions.

KPI106 Lys173 is involved in the interaction with the protease SCP1

Proteinaceous protease inhibitors carry out their function by using a scissile bond that serves as a substrate for the protease (Ozawa and Laskowski, 1966; Laskowski and Qasim, 2000). The residues next to the scissile bond towards the N-terminus are labelled P1, P2 and so on, and P1', P2' in the C-terminal direction. In most cases, the formation of a protease–protease inhibitor complex is a highly complementary interaction, in which the residues adjacent to P1 form a reactive side loop that enters the active centre of the protease in a key–lock manner (Rawlings et al., 2004). To get more detailed information about the interaction between KPI106 and SCP1 we carried out in silico docking analysis to map potential residues eligible for the KPI106 P1 residue. Plant KPIs belong to the I3A family of protease inhibitors according to the MEROPS database (Rawlings et al., 2012). In contrast to most described Kunitz inhibitors, all mycorrhiza-specific KPIs identified possess three disulphide bonds (Figure 4a) and thus belong to the clan I3B (Rawlings et al., 2004). The protease inhibitor API-A from Sagittaria sagittifolia serves as reference structure for protein fold recognition of KPIs of the I3B clan in the Phyre2 database (http://www.sbg.bio.ic.ac.uk/phyre2/). Using this API-A as a template we modelled KPI106 (Figure 4b). Results show a highly similar three-dimensional (3D) structure of the two proteins with a comparable arrangement of three disulphide bonds and of a prominent Lys residue that has been experimentally assigned as the P1 residue in API-A (Bao et al., 2009). Docking analyses using the Patchdock server (http://bioinfo3d.cs.tau.ac.il/PatchDock/) to model the interaction with SCP1 showed that Lys173 from KPI106 perfectly matches the active pocket of SCP1 containing the catalytic triad Ser226, Asp414 and His466 (Figure 4c–e). This result suggests that Lys173 could act as P1 in KPI106. However, KPI104 does not interact with SCP1 but still contains a Lys residue in position 170 in the analogous loop, albeit followed by two different amino acids. We thus decided to test experimentally the importance of the Lys residue for the interaction using site-directed mutagenesis. We mutated the Lys to Gly in both KPIs and analysed the interaction with SCP1 in the yeast two-hybrid system. Results showed that the interaction between SCP1 and KPI106Lys173Gly was stronger than with KPI106 (Figure 5a). Accordingly, the Lys170Gly mutation in KPI104 enabled the interaction with SCP1. This indicates that the Lys173 of KPI106 indeed plays a role in determining the specificity and strength of the interaction with SCP1. We performed 3D structure alignments of the loops containing the Lys173/170 residues in KPI106, KPI104 and of their mutated forms. Results showed that the Lys to Gly mutation induces a torsion in the respective loops that changes the steric conformation of the residues adjacent to the Lys173/170 (Figure 5b). This suggests that in KPI106 not only Lys173 as putative P1, but also its very next adjacent residues (Phe174–Glu175) are important in mediating specificity towards SCP1.

Figure 4.

In silico docking of KPI106 and SCP1 reveals Lys173 as a putative P1 residue of KPI106.

(a) Alignment of mycorrhiza-induced Kunitz protease inhibitors (KPIs) with other Kunitz inhibitors including API-A from Sagittaria sagittifolia, soybean trypsin inhibitor (STI) and the control inhibitor KPIc. Mycorrhiza-induced KPIs and API-A show conservation of four cysteines, whereas STI and KPIc only contain two.

(b), (c) API-A (b) that has been used as reference structure to build the model of KPI106 (c) for in silico docking. API-A is the only crystallized Kunitz inhibitor present in the database containing three disulphide bonds as does KPI106. The P1 residue of API-A is specified by Lys169. KPI106 carries Lys at position 173.

(d), (e) In silico docking of KPI106 with SCP1 in steps. KPI106 Lys173 enters the pocket of the active centre of SCP1 containing the catalytic triad Ser226, Asp414 and His466 and is therefore hypothesized as a putative P1 residue of KPI106. KPI106 Phe174 and Glu175 are located next to the putative P1 Lys173.

Figure 5.

Substitution of KPI106 Lys173 to Gly leads to a stronger interaction with SCP1.

(a) KPI106Lys173Gly shows a stronger interaction with SCP1 in a yeast two-hybrid test. Furthermore, KPI104Lys170Gly shows interaction with SCP1, whereas KPI104 does not.

(b), (c) Three-dimensional structure alignments of the loop containing Lys173/170 of KPI106 (light green) with KPI104 (dark green, b) and KPI106Lys173Gly (lilac) with KPI104Lys170Gly (red, c). Mutation of Lys173/170 changes the steric conformation of the adjacent residues.

Mycorrhiza-induced SCPs in M. truncatula

The individual interaction patterns of the different KPI family members despite their high homology led us to screen for potential SCP1 homologues in the M. truncatula genome. Blasting the SCP1 cDNA sequence against the hapmap3.5 database we identified 20 SCP genes, of which 12 were shown to be mycorrhiza-specific according to the MtGEA (black dots, Figure 6a). They can be grouped into six pairs, each composed of two gene copies located on chromosome 1 (herein SCP1–6) and chromosome 3 (herein SCP1a–6a), respectively. The SCP genes are thereby localized in clusters that display identical copies on the respective chromosomes. Each SCP open reading frame is composed of seven exons, and Pfam analysis of the encoded protein sequence revealed the presence of an N-terminal secretion signal peptide and a protease domain of the S10 family with the catalytic residues S, D, H (Figure 6b). The SCP proteins within one cluster share a very high sequence homology, between 85 and 92%. It is known that some SCPs do not exhibit proteolytic activity but function as acyltransferases (Lehfeldt et al., 2000; Stehle et al., 2009). However, true serine carboxypeptidases contain the diagnostic pentapeptide –GESYA– around the catalytic serine residue, in contrast to –GDSYS– in acyltransferases (Stehle et al., 2009). As shown in Figure 6c, all mycorrhiza-induced SCPs contain the –GESYA– motif, indicating they are true serine carboxypeptidases.

Figure 6.

Medicago truncatula genome analysis of SCP1 homologues.

(a) Identification of 20 SCP genes within the M. truncatula genome. According to the M. truncatula Gene Atlas, 12 of them are exclusively induced during arbuscular mycorrhiza symbiosis (black dots, SCP1–6; SCP1a–6a). Numbers are bootstrap supported values.

(b) Genomic localization of mycorrhiza-induced SCPs. SCP1–6 are localized on a cluster on chromosome 1, whereas a copy of this SCP cluster is present on chromosome 3, containing SCP1a–SCP6a. Each SCP gene is composed of seven exons and proteins contain an N-terminal secretion signal peptide and a C-terminal S10 peptidase domain including the catalytic residues S, D, H.

(c) All mycorrhiza-specific SCPs contain the highly conserved –GESYA– motif (grey box), with S as catalytic residue (*), that is characteristic for SCPs with proteolytic activity.

Kunitz protease inhibitors and SCP1 are likely secreted into the apoplast

To investigate the localization of KPI106, KPI104 and SCP1, we expressed them in N. benthamiana epidermal cells as C-terminal GFP fusions under control of the 35S promoter. A DsRed cassette expressed under control of the ubiquitin promoter was included as a positive control. DsRed localizes in the cytoplasm and in the plant nucleus. For both KPI-GFP proteins and for SCP1-GFP the same GFP fluorescence localization pattern was observed. The proteins localized in dots distributed throughout the cell and accumulating at the cell periphery as well as in the endoplasmic reticulum (ER) around nuclei, indicating that fusion proteins follow the conventional secretory pathway (Figure 7a). In contrast, in plants infiltrated with the construct 35Spro: eGFP, the fluorescence was restricted to the cytoplasm and to the plant nucleus fully coincident with the DsRed localization in all constructs. Integrity of the C-terminal fusion proteins was demonstrated by Western blot using an anti-GFP antibody on tobacco protein cell extracts. KPI106-GFP (50 kDa), KPI104-GFP (50 kDa) and SCP1-GFP (84 kDa) proteins migrated according to their expected molecular size (Figure 7b). Free GFP resulting from cleavage during protein extraction was also observed migrating at 27 kDa.

Figure 7.

KPI106, KPI104 and SCP1 are secreted into the apoplast.

(a) Expression of 35Spro:KPI106-GFP, 35Spro:KPI104-GFP, 35Spro:SCP1-GFP in N. benthamiana is localized in dots around the cell periphery and the nuclei, indicating the secretory pathway of the fusion proteins. Free DsRed marks the nucleus. Free GFP serves as a localization control.

(b) Confirmation of the integrity of the C-terminal GFP fusion proteins by Western blot with an anti-GFP antibody. The GFP fusion proteins migrate at the correct size (red arrowheads). In all lanes free GFP is migrating at 27 kDa due to GFP cleavage during protein extraction.

Inactivation of the SCP proteins phenocopies the overexpression of the Kunitz inhibitors KPI106 and KPI104

Taken together, our results above suggest a possible role for tandem interactions between mycorrhiza-specific KPIs and SCP proteins. To challenge this hypothesis we decided to gain insights into the cellular function of the protease SCP1 using RNAi. Given the extensive nucleotide identity between different SCP homologues it was likely that the silencing construct did not just target SCP1 (Figure S4). Indeed, silencing of SCP1 transcript and all homologues was confirmed by quantitative PCR comparing expression in mycorrhizal RNAi hairy roots (17 dpi) versus mycorrhizal control roots (Figure 8a). Morphological analysis of roots colonized by R. irregularis at 17 dpi revealed that SCP-silenced roots contained a large proportion of malformed arbuscules (Figure 8b) when compared with control roots (Figure 8c). While there were no changes in the extension of the roots colonized by R. irregularis, the overall number of mature arbuscules in the SCP-silenced roots was significantly lower (Figure 8d). This phenotype resembles the overexpression of KPI106 and KPI104, and suggests that tandems of SCPs and KPIs might work together during arbuscule development and control of fungal growth within the root cortex. In particular, our interaction analysis suggests that SCP1 is the specific counterpart for KP106. We hypothesize that a precise balance between this protease and its inhibitor is indispensable for proper fungal development in planta.

Figure 8.

Silencing of serine carboxypeptidases (SCPs) leads to a phenocopy of the Kunitz protease inhibitor (KPI) overexpression phenotype.

(a) Confirmation of successful silencing of SCPs in two RNA interference (RNAi) lines by quantitative PCR, 17 days post-inoculation (dpi). Values were normalized to tef. Error bars represent means of three independent biological replicates.

(b), (c) Visualization of fungal structures in R-SCP_1 (b) and control lines (c), 17 dpi, using WGA-fluorescein. Many arbuscules in the RNAi line show an aberrant morphology (red arrowhead) and intraradical hyphae contain septa (white arrowhead).

(d) Quantification of mycorrhizal colonization. The frequency of colonization (F%) is comparable in R-SCP and control lines, whereas arbuscule abundance (a%) is significantly lower in the RNAi lines which phenocopies the disturbed mycorrhizal morphology as shown for KPI106 and KPI104 overexpression. **P-value <0.01; n, number of root fragments.


The establishment of the arbuscular mycorrhizal symbiosis within the root requires the coordination of plant cellular programs to permit fungal accommodation while suppressing defence reactions. In particular, the development of the key organ of the symbiosis, the arbuscule, represents the culmination of this coordination, imposing on the fungus a morphogenetic change from polarized growth to dichotomous hyperbranching. That this process is mainly controlled by the plant cellular program is evident by the phenotype of several plant mutants in which this fungal differentiation is aborted or impaired (Javot et al., 2007; Reddy et al., 2007; Floss et al., 2008; Takeda et al., 2009; Baier et al., 2010; Feddermann et al., 2010; Kuhn et al., 2010; Pumplin et al., 2010; Zhang et al., 2010). In this paper we show a further example of plant control over fungal differentiation where a tandem between a mycorrhiza-specific plant KPI and its target serine carboxypeptidase is key to the full progress of fungal colonization and particularly arbuscule development.

The KPIs KPI106 and KPI104 are induced in response to diffusible molecules from the AM fungus R. irregularis (Kuhn et al., 2010). Other KPIs have also been shown to be induced in response to microbial elicitors, such as AtKTI1 from Arabidopsis thaliana that is transcriptionally activated in response to culture filtrates from Erwinia carotovora (Li et al., 2008). Similarly, in symbiotic associations, the KPI SrPI1 was found to be one of the earliest genes induced during nodule development between Sesbania rostrata and Azorhizobium caulinodans (Lievens et al., 2004). The SrPI1 transcript accumulates at very early stages, possibly in response to Nod factors, since bacterial strains unable to produce them did not induce expression (Lievens et al., 2004). KPI106 and KPI104 are mycorrhiza-specific and are further induced at later stages of the symbiosis, indicating that they might play a prominent function in the association. This is supported by the finding that KPI106 clusters within a group of M. truncatula KPI-encoding genes with mycorrhiza-specific expression patterns. Interestingly, all members of this cluster contain six conserved cysteine residues predicted to form three disulphide bonds which positions these proteins within the I3B clan of inhibitors (Rawlings et al., 2004). The only other characterized proteins within this clan are the double-headed protease inhibitors API-A and API-B from S. sagittifolia that contain two reactive sites instead of one like most of the KPIs (Bao et al., 2009). It is therefore tempting to speculate that the three disulphide bonds structure present in all mycorrhiza-specific KPIs are is essential for their symbiotic function.

The relevance of mycorrhiza-specific KPIs for the symbiosis was demonstrated by altering the expression levels of either KPI106 or KPI104. Overexpression of either of them impaired the development of the fungus within the root, causing a significant reduction in the number and developmental status of arbuscules. Degenerate arbuscules and septate hyphae were more often seen in those roots. Arbuscular mycorrhiza fungi are coenocytic fungi and the presence of septa is indicative of hyphal decay. Pre-symbiotic hyphae start septating if they do not encounter an appropriate host for colonization, retrieving the cytoplasm back to the spore (Logi et al., 1998). While septa in arbuscules are associated with arbuscule turnover (Bonfante-Fasolo, 1990), septa in intercellular hyphae are less common.

Conversely, inactivation of KPI106 and KPI104 by RNAi did not lead to any visible change in the development of the symbiosis. This phenotype was somehow surprising, because the expression pattern of KPI106 and KPI104 indicated that it is the presence and not the absence of the corresponding proteins that is necessary for the symbiotic function. However, we interpreted these results assuming that the redundancy of mycorrhiza-specific KPIs could compensate for the lack of one of them. In contrast, altering the concentration/localization of the corresponding proteins by ectopic expression within the cell produced more severe effects, and we thus hypothesized that this could be due to excessive or inappropriate spatial inhibition of the corresponding target protease.

To prove that, we screened for targets of KPI106 using the yeast two-hybrid approach and identified a constitutively expressed cysteine protease. However, in contrast to KPI106, which was predicted to localize in the apoplastic space, the cysteine protease contained a vacuolar sorting signal, making it unlikely that these two proteins would interact in vivo. This prompted us to look for other possible protease targets of KPI106 and KPI104. Among the mycorrhiza-induced proteases two types have received special attention because they are specific and induced very early similar to KPI106 and KPI104. One is a subtilase family, identified in L. japonicus, from which two isoforms SbtM1 and SbtM3 conform to these criteria (Kistner et al., 2005). Interestingly, inactivation of any one of these mycorrhiza-specific proteases also leads to an impaired mycorrhizal phenotype (Takeda et al., 2009). However, rather than malformed symbiotic structures, the subtilase RNAi plants showed a general drastic decrease in fungal colonization. The second type of mycorrhiza-specific protease that has been described is the serine carboxypeptidases (Liu et al., 2003; Guimil et al., 2005). SCP1 from M. truncatula is specifically expressed in colonized roots and it has homology to serine carboxypeptidase II (Liu et al., 2003). However, little was known about the function of SCP1 during symbiosis.

Interaction assays using the closest M. truncatula homologue of L. japonicus SbtM1 and the Medicago protein SCP1 versus all identified mycorrhiza-specific KPIs showed that SbtM1 is probably not a substrate for any of these KPIs. In contrast, the strong and specific interaction between SCP1 and KPI106 or KPI111 suggests that SCP1 might be their target protease. Given the sequence similarities between the mycorrhiza-induced KPIs, they showed different interaction affinities. This is supported by the fact that the other characterized KPIs from the I3B family, API-A and API-B, containing more than 91% sequence identity, have different inhibitory specificities (Bao et al., 2009). Three-dimensional modelling of KPI106 and SCP1 showed that KPI106 fits into the catalytic pocket of SCP1. This is in accordance to the Laskowski model of protease inhibitory function predicting that the inhibitor behaves as a substrate for the protease (Rawlings et al., 2004). Mutation analyses of the Lys residues of KPI106 that fits into the catalytic pocket showed that this residue, and probably also the adjacent residues conserved between KPI106 and KPI111 (–KFE–) and different in KPI104 (–KVQ–), are important in determining the strength and specificity of the interaction.

That KPI106 interacts with SCP1 is consistent with the fact that most Kunitz inhibitors act on serine proteases and with the localization pattern of SCP1 that is identical to that of KPI106. But more importantly, inactivation of SCPs phenocopies the overexpression of KPI106. This strongly suggests that both proteins might work together to control the same process. SCP1 is expressed in arbuscule containing cells and in adjacent cells, as well as in cells passed by fungal hyphae on their way to the inner cortex (Liu et al., 2003; Gomez et al., 2009; Pumplin and Harrison, 2009). We could envisage a model in which the protease SCP1, transcriptionally induced by the fungus in its way to the inner cortex, is required to produce a peptide signal that triggers further symbiotic development. This peptide signal can also move and act on adjacent cells to prepare the path that the fungus will take during colonization. Interestingly, evidence of the existence of such a signal that moves from one colonized cell into adjacent cells during AM colonization is also provided by microscopic observations of the formation of the pre-penetration apparatus (PPA) (Genre et al., 2008). In those, the path that the fungus will follow during colonization can be predicted by the formation of the PPA, a tunnel-like structure surrounded by ER and cytoskeletal elements. In this scenario, the Kunitz inhibitor KPI106 would ensure that the protease is not ubiquitously or permanently active, by competing with the substrate of the protease. It will therefore be necessary in the future to identify the target/s of the protease to determine precisely which peptide signal(s) triggers symbiotic development in the cortex.

Experimental Procedures

Plant growth and hairy root production and AM fungal material

Transformation of M. truncatula var. Jemalong mediated by Agrobacterium rhizogenes (strain ARquaI; Quandt et al., 1993) was carried out according to Boisson-Dernier et al. (2001), resulting in composite plants. For mycorrhization assays, the bi-compartmental plate system with R. irregularis DAOM 197198 (formerly Glomus irregulare) as described in Kuhn et al. (2010) was used. For localization studies, leaves of 5-week-old N. benthamiana plants were transiently transformed using Agrobacterium tumefaciens (strain GV3101) as described in Schütze et al. (2009).

Constructs used for overexpression, RNAi silencing and localization experiments

For overexpression analyses, KPI104 (TC100804, DFCI gene index) and KPI106 (Medtr8g059790.1, M. truncatula hapmap3.5) open reading frames (ORF) were amplified from cDNA, cloned into pENTR®/D-TOPO® (Invitrogen, http://www.invitrogen.com/) and subcloned into the binary Gateway vector pCGFP-RR (Karimi et al., 2002; modified according to Kuhn et al., 2010) by using the primers no. 1–4. For RNAi-mediated silencing of SCP1 (Medtr1g100630.1, M. truncatula hapmap3.5) a 259 bp fragment (+1 to +260) was amplified using the primers no. 11–12 and cloned into pK7GWIWG2D(II), 0 (Karimi et al., 2002) (See also Methos S1). For localization studies KPI104, KPI106 and SCP1 were expressed as C-terminal eGFP fusion proteins. Cloning was carried out by deletion of stop codons of the respective ORFs (primers no. 12–17) using the QuikChange® mutagenesis protocol according to the manufacturer's instructions (Stratagene, http://www.stratagene.com) and subcloned into the binary vector pCGFP-RR (Karimi et al., 2002; modified according to Kuhn et al., 2010). Mutation of Lys173/170 in KPI106 and KPI104 to Gly has been carried out using the QuikChange® mutagenesis protocol with primers no. 18–21.

Yeast-two-hybrid experiments

For yeast-two-hybrid analyses, KPI106, KPI104, KPI105, KPI111 and KPIc ORFs, lacking the secretion signal peptides, were amplified from cDNA with primers (no. 22–31) containing restriction sites and cloned via the restriction sites into the bait vector pGBKT7 (Clontech, http://www.clontech.com/). For direct interaction tests, ORFs of CP (TC133093, DFCI gene index), SbtM1 (AW584611, DFCI gene index) and SCP1, lacking secretion signal peptides, were amplified from cDNA using primers (no. 32–37) containing restriction sites and cloned into the prey vector pGADT7 (Clontech) via the respective restriction sites. The bait and prey vectors were co-transformed in Saccharomyces cerevisiae strain AH109 and at least 10 colonies per co-transformation were tested for potential interaction on SD-LWHA medium (See also Methods S2).

All primer sequences used for cloning can be reviewed in Table S1.

Quantitative real-time PCR

Extraction of total RNA, cDNA synthesis and quantitative real-time PCR were carried out as described in Helber et al. (2011). Relative transcription levels of target genes were normalized to the M. truncatula trans-elongation factor 1-alpha (tef, TC106470, DFCI gene index). Respective primer sequences for all transcripts analysed by quantitative real-time PCR can be reviewed in Table S1.

Phenotypical analyses, staining procedures and microscopy

For each RNAi and overexpression construct, 70 composite plants were screened, of which 10 independent transformed roots were excised and propagated resulting in a hairy root line. The rate of silencing or overexpression of each line was tested by quantitative PCR and the best lines were selected for mycorrhization assays. For quantification, mycorrhizal hairy roots were stained with ink as described in Vierheilig et al. (1998), and quantification of mycorrhizal colonization was carried out according to Trouvelot et al. (1986) using the program Mycocalc (http://www2.dijon.inra.fr/mychintec/Mycocalc-prg/download.html). Between 60 and 100 replicates of each mycorrhized hairy root line were analysed and figures are representative of average results. P-values for mycorrhizal colonization parameters were calculated by Student's t-test as described in Kloppholz et al. (2011). For phenotypical analyses, mycorrhizal hairy roots were treated with 10% KOH for 45 min and neutralized with 2% HCl. After three washes with 1 × PBS, roots were put into staining solution (10 μg ml−1 WGA-Fluorescein, 0.02% Tween20, 1 × PBS) overnight at 4°C and analysed using a Leica TCS SP5 confocal microscope (http://www.leica-microsystems.com/) with excitation at 494 nm and emission collected from 505 to 525 nm. Microscopical analyses of transiently transformed N. benthamiana leaves were carried out as described in Kloppholz et al. (2011). All microscopic pictures represent maximal projections of Z-stack images.

Phylogenetic analyses

All phylogenetic trees are based on cDNA sequences obtained from the Medicago hapmap3.5 database (http://www.medicagohapmap.org/). In silico expression analyses were carried out using the Gene Expression Atlas (http://mtgea.noble.org/v2/; Benedito et al., 2010). Alignments were created with ClustalX (http://www.clustal.org/download/current/; Larkin et al., 2007) and the respective trees by the Mega5.0 software (http://www.megasoftware.net/; Tamura et al., 2011) using the neighbour-joining algorithm and a bootstrap value of 1000.

Protein analyses, 3D models and in silico docking

Amino acid sequences were analysed using the SignalP4.0 server (http://www.cbs.dtu.dk/services/SignalP/; Petersen et al., 2011) and the pfam database (http://pfam.sanger.ac.uk/; Punta et al., 2012). The protein model for SCP1 was generated by the protein structure prediction server Phyre2 (http://www.sbg.bio.ic.ac.uk/phyre2/html/page.cgi?id=index; Kelley and Sternberg, 2009) using PSI-Blast to find suitable template structures. For KPI106 the Phyre2 output for structure templates revealed 3E8L_C to be best model template. The respective 3E8L_C protein API-A from S. sagittifolia contains three disulphide bonds as does KPI106. The KPI106 model was adjusted to API-A using the Swiss-Model server (http://www.expasy.org/structural_ bioinformatics; Schwede et al., 2003). Ramachandran plots (http://mordred.bioc.cam.ac.uk/~rapper/rampage.php">http://mordred.bioc">http://mordred.bioc.cam.ac.uk/~rapper/rampage.php; Furnham et al., 2008) as well as Z-score and energy plots (http://prosa.services.came.sbg.ac.at/prosa.php; Wiederstein and Sippl, 2007) were applied to evaluate the quality of the protein models. In silico docking was carried out with the docking server Patchdock (http://bioinfo3d.cs.tau.ac.il/PatchDock/; Schneidman-Duhovny et al., 2005) using a geometry-based molecular docking algorithm. With the parameters 4.0 as clustering root-mean-square deviation and the selected complex type ‘Enzyme-Inhibitor’, a docking model of KPI106 and SCP1 was generated. The docking model with the highest score of 13 964, an area of 2152.3 and an Atomic Contact Energy (ACE) of 300.43 was used for further analysis and visualized with UCSF chimera (http://www.cgl.ucsf.edu/chimera/; Pettersen et al., 2004).


This work was supported by the German Research Foundation (DFG) RE1556/4-2 and SSR was funded by the Landesgraduiertenförderung of the state of Baden-Württemberg, Germany. We declare that authors do not have any conflict of interest.