Protein-protein interactions as a proxy to monitor conformational changes and activation states of the tomato resistance protein I-2.

Plant resistance proteins (R) are involved in pathogen recognition and subsequent initiation of defence responses. Their activity is regulated by inter- and intramolecular interactions. In a yeast two-hybrid screen two clones (I2I-1 and I2I-2) specifically interacting with I-2, a Fusarium oxysporum f. sp. lycopersici resistance protein of the CC-NB-LRR family, were identified. Sequence analysis revealed that I2I-1 belongs to the Formin gene family (SlFormin) whereas I2I-2 has homology to translin-associated protein X (SlTrax). SlFormin required only the N-terminal CC I-2 domain for binding, whereas SlTrax required both I-2 CC and part of the NB-ARC domain. Tomato plants stably silenced for these interactors were not compromised in I-2-mediated disease resistance. When extended or mutated forms of I-2 were used as baits, distinct and often opposite, interaction patterns with the two interactors were observed. These interaction patterns correlated with the proposed activation state of I-2 implying that active and inactive R proteins adopt distinct conformations. It is concluded that the yeast two hybrid system can be used as a proxy to monitor these different conformational states.

Introduction

The interaction between the soil-born, xylem-colonizing fungus Fusarium oxysporum f. sp. lycopersici (Fol) and its host tomato (Solanum lycopersiscum) is a model system to study the molecular basis of disease resistance in plants. Tomato plants defend themselves against fungal colonization by the secretion of antimicrobial components, pathogenesis-related proteins and by blocking the xylem vessels with tyloses, pectic gels, and gums (Beckman, 2000; Rep ). In a susceptible plant, the blocking of the xylem vessels by the spreading fungus and the responding plant’s reduction of water flow thereby leads to wilting and eventually death. Some plants, however, are resistant to particular isolates of Fol. Upon infection they respond faster and hence more effectively, restricting fungal colonization. This gene-for-gene type of resistance depends on the presence of a dominant resistance (R) gene in the plant that recognizes a matching avirulence factor (Avr) in the fungus (Flor, 1942). Many Avrs are in fact effectors and therefore gene-for-gene resistance is also called effector-triggered immunity (Jones and Dangl, 2006).

In the tomato–Fol interaction, three well-studied R/Avr pairs have been identified (Simons ; Rep ; Houterman ; Houterman ; Takken and Rep, 2010). The resistance gene that mediates defences against race 2 isolates of Fol has been cloned and is called I-2 (Immunity to race 2) (Simons ; Houterman ). I-2 is expressed in root and stem parenchyma cells that are in direct contact with the xylem tissue (Mes ). Avr2 recognition by I-2 and subsequent defence responses can be artificially induced in leaves and stems, either by virus-based overexpression of Avr2 in tomato carrying I-2, or in N. benthamiana leaves after transient co-expression with I-2 through agroinfiltration, and is visible as a hypersensitive response (Houterman ). I-2 belongs to the CC-NB-LRR class of R proteins that contains an amino-terminal coiled-coil (CC) domain, a central nucleotide-binding (NB) domain, and a C-terminal leucine-rich repeat (LRR) domain. The most conserved domain of this class of R proteins is the NB domain that is part of a larger region called the NB-ARC (Nucleotide Binding domain shared by Apaf-1, R proteins, and Ced-4; van der Biezen and Jones, 1998). With a purified recombinant form of I-2, which lacked the LRR domain, a role of the NB-ARC domain in ATP/ADP binding and ATP hydrolysis has been shown (Tameling , 2006).

Biochemical analyses of two constitutively active I-2 mutants (S233F and D238E) showed that they were affected in ATP hydrolysis, but not in ATP/ADP binding, suggesting that these mutants are locked in the ATP-bound state. When these mutations were combined with a mutation in the P-loop (K207R) that blocks nucleotide binding, the autoactivation phenotype was abolished. These observations show that nucleotide binding is required for activation of defence signalling and that the ATP-bound state most likely represents the activate state (Tameling , 2006). Binding of ADP was found to result in a stabile I-2-nucleotide complex, which implies that the different nucleotide-binding states exhibit different conformations. Based on these observations, a molecular switch model was proposed (Takken ). In the ‘off’ state, the R protein is tightly bound to ADP. It is assumed that upon Avr perception the conformation of the nucleotide-binding pocket changes, resulting in release of ADP. Subsequent binding of ATP results in a second conformational change (‘on’ state) that allows the protein to activate the down-stream defence-signalling cascade. Hydrolysis of the bound ATP by its intrinsic ATPase activity reverts I-2 to the ‘off’ state. In this biochemical model, the conformation of the protein is regulated by its nucleotide-binding state.

To get insight into the conformation of the I-2 NB-ARC domain, the crystal structure of the NB-ARC domain from the ADP-bound state of Apaf-1 was used as a template to obtain a 3D model of this domain (Riedl ; Takken ; van Ooijen ). The predicted structure allowed mapping of the amino acid residues in I-2 that are most likely involved in nucleotide binding and hydrolysis. Mutations of many of those residues, which are highly conserved in other R proteins, resulted in either a constitutively active- or a loss-of-function phenotype (Tameling ; van Ooijen ). Mutations in the corresponding residues in other R proteins conferred similar phenotypes (Dinesh-Kumar ; Tao ; Bendahmane ; van Ooijen ). These genetic data further support the above-mentioned molecular switch model in which a change in the nucleotide-binding state results in a conformational change representing the different activation states (on/off). An assumption of the switch model is that nucleotide exchange triggers a conformational change allowing R proteins to bind to, or dissociate from, interacting proteins. However, direct evidence for a conformational change or an altered interaction with other proteins is currently lacking.

The aim of this study is to investigate whether the activation state of I-2 affects its ability to interact with other proteins. Two proteins interacting with I-2 were identified. The functional involvement of these proteins in I-2-mediated defence was analysed using stable silenced tomato lines. Yeast two-hybrid constructs were used to map the minimal domains of I-2 that are required for the interaction with these proteins. I-2 mutants that differ in their proposed activation state show differences in their ability to interact with the two interactors. These distinct interaction patterns indicates that the different nucleotide-dependent conformational changes can be monitored in the yeast system and provide direct support for the switch model.

Materials and methods

Yeast two-hybrid constructs and protein expression

Bait constructs:

Construction of the tomato–Fusarium cDNA library and the I-2 baits used for screening I-2 FL and I-2N CC-NB-ARC have been described before (de la Fuente van Bentem ). The bait used for the library screen was constructed by subcloning the NcoI/SacI fragment of I-2 FL in the yeast two-hybrid (Y2H) bait vector pAS2-1 (Clontech Laboratories, Palo Alto, CA, USA), resulting in bait CC-NB-ARC-LRR1-12 (amino acids 1–872). Bait I-2N+:CC-NB-ARC-LRR1-5 (amino acids 1–643) was constructed by cloning the NcoI/PstI fragment into pAS2-1. Bait I-2NΔMHD:CC-NB-ARCΔMHD was obtained via exonuclease activity after digestion of the Tth111I restriction site in bait I-2 CC-NB-ARC. After religation and sequencing, the codons encoding amino acids 485–495 appeared to be deleted. Bait I-2CC (amino acids 1–168) was constructed by cloning the NcoI/EcoRV fragment of I-2 into pAS2-1.

The I-2N mutant CC-NB-ARC (S233F, D283E) were reported before as I-2 FL (S233F, D283E) clones in pGreen1K (Tameling ). NcoI/XhoI fragments of those pGreen vectors were used to create I-2 FL baits with corresponding mutations in pAS2-1 digested with NcoI/SalI. Afterwards, I-2 FL baits (S233F, D283E) were digested with PstI to release a fragment thereby resulting in deletion of LRR5-29 and creation of I-2N+S233F, I-2N+D283E in pAS2-1.

To create the I-2N:CC-NB-ARC bait with the K207R mutation, an NcoI/XhoI fragment from pGEX-KG construct with I-2N (Tameling ) was cloned into pAS2-1 that was digested with NcoI and SalI. Replacing the NcoI/BamHI fragment of this clone with the pAS2-1:I-2N+ made the bait I-2N+K207R.

The Rx CC-NB-ARC and Rx full-length coding region in pBAD were obtained from Wageningen University (Hans Keller). An NcoI/EcoRI fragment of Rx CC-NB-ARC was first ligated into a pGEX-4T-1 (GE Healthcare) vector digested with the same enzymes. Vector pGEX-4T-1 was modified to contain extra restriction sites for NcoI, XbaI, and BglII. This plasmid encodes Rx amino acids 1–426 which is fused N-terminally to GST. Bait Rx:CC-NB-ARC was created following the same strategy where an NcoI/EcoRI fragment from pGEX-4T-1:Rx:CC-NB-ARC was cloned into pAS2-1 vector digested with NcoI and EcoRI. To create a full-length Rx bait, first a NcoI/SmaI fragment of the full-length Rx from pBAD was exchanged with the NcoI/PmeI fragment of pGEX-4T-1. This plasmid encodes full-length Rx protein fused N-terminally to GST and C-terminally to HIS tag. Subsequently a NcoI/SalI fragment was ligated to a pAS2-1 vector digested with the same enzymes.

To create the Mi1.2:CC-NB-ARC bait, a NcoI/BsmI fragment was cut from plasmid pKG6210 (Keygene, Wageningen, The Netherlands; Vos ) containing Mi-1.2 and transferred into pAS2-1 digested with NcoI/SmaI. The resulting plasmid was digested with NcoI, filled with Klenow polymerase, and religated in order to adjust the frame (amino acids 161–896). To create a full-length Mi-1.2 bait, several intermediate plasmids were constructed. First, two PCR amplifications of the template pKG6210 were performed using the primer pair FP181 (5′-CGGGATCCTGTCATTTCGATCACCGGTATGC-3′) and FP182 (5′-GGGGTACCCGGCTATTTCTTTACCGACATC-3′) and the primer pair FP183 (5′-GGGGTACCCCGTTGTGACAAATCGGCCGTG-3′) and FP184 (5′-CTGGTCGACCTACTTAAATAAGGGGATATTCTTCTG-3′). After digestion of the PCR products and a BamHI/KpnI/SalI three-point ligation, the resulting NB-ARC-LRR fragment of Mi-1.2 was cloned into a pUC19 vector, previously digested with BamHI/SalI. Then, a XhoI/MscI fragment of pKG6210 was cloned into this plasmid digested with the same enzymes. The resulting vector was used to create bait NB-ARC-LRR in pAS2-1 by ligating its BamHI/SalI fragment into pAS2-1 digested with the same enzymes. To adjust the reading frame, this plasmid was digested with NcoI, filled with Klenow polymerase, and religated. A AgeI/SalI fragment of resulting vector was exchanged with a AgeI/SalI fragment from the Mi1.2:CC-NB-ARC bait, described above, to create a nearly full-length Mi1.2 protein that lacks its first 161 amino acids. To create the final bait encompassing the full-length Mi1.2 protein, a fragment was amplified by FP194 (5′-CATGCCATGGAAAAACGAAAAGATATTGAAG-3′) and FP180 (5′-GGGGTACCGAGTTGAAACAGAGGTAAGAC-3′) and ligated into pKG6210 after NcoI digestions. Since vectors carrying full-length Mi1.2 are unstable in E.coli, the resulting vector was transformed directly to yeast.

For the I-2 C1:CC-NB-ARC bait, a cosmid containing this homologue (cosmid A2; (Simons ) was used as the starting material. Using primers F I-2C1 (5′-CAGATTTGAGCCATGGAGATTGG-3′) and R I-2C1 (5′-GGGCCGACATTGTTCCAACATATG-3′) and Pfu DNA polymerase (Stratagene), a DNA fragment was amplified that encoded amino acids 1–526 from I-2C1. The fragment was cloned into the pAS2-1 vector using the same restriction sites as used for the corresponding I-2 fragment (de la Fuente van Bentem ).

Prey construct:

The insert of SlFormin homologue SGN-U583099 was amplified using the primer pair FM13 (5′-TTCCCAGTCACGACGTTGT5-3′) and RHom (5′-CCGAATTCGTATACGAGCTGCCCGTGC-3′). The amplification products were digested with EcoRI and XhoI and ligated into the pACT2-1 vector digested with the same restriction enzymes.

Western blot analysis:

For all bait and prey constructs, Western blot analysis confirmed that the expected chimeric proteins were produced in the yeast host PJ69-4a. For these blots, yeast cells were collected and proteins were extracted as described in the Clontech Yeast Protocols Handbook (http://www.clontech.com). Protein samples were loaded on SDS-PAGE gels for immunoblotting. Blots were probed with either αGal4DB for bait proteins or with αGal4AD for prey proteins (Clontech) antibodies followed by goat antimouse antibodies conjugated to horseradish peroxidase (Pierce).

Yeast two-hybrid assays and library screens

The Y2H assays and library screens were performed as described (de la Fuente van Bentem ). The nearly full-length I-2 bait (amino acids 1–872) was used to screen the tomato–Fusarium interaction cDNA library and 7 × 106 yeast transformants were tested for growth on MM-HWL plates. After 2 days of growth at 30 °C, the plates were replica-plated on MM-AWL and MM-HWL selective plates and on MM-WL plates. The original plate was subjected to an X-Gal staining procedure (Duttweiler, 1996) for detection of the LacZ marker. A second series of plates (MM-AWL, MM-HWL, and MM-WL) was made from the first MM -WL replicate. After 5 days of growth at 30 °C, the growth on the second series of selective plates was determined.

Cloning of SlFormin and SlTrax

To obtain a full-length cDNA sequence of SlFormin, primer-walking was performed with gene- and plasmid-specific primers using the tomato–Fusarium interaction cDNA library as the template (de la Fuente van Bentem ). After three steps, the assumed full-length coding region was identified and subsequently amplified with the primer pair I2I-1-F1 (5′-AGGGGCTTCAATCCATCTG-3′) and I2I-1-stop-R (5′-CAGTCGACCTACGGGCTTGAGCTCTCGT-3′). PCR fragments were cloned into pGEM-T-Easy (Promega) and three independent clones were sequenced. The consensus sequence was compared to the genomic sequence present in the SOL Genomics Network (SGN) database (SL1.00sc05390_84.1.1; SGN Tomato Combined; http://solgenomics.net/).

For SlTrax, the full-length cDNA was directly amplified from the cDNA library (de la Fuente van Bentem ) with the primer pair I2I-2-ATG-F (5′-ATGGCTTCAAAACCCCAGCGC-3′) and I2I-2-R (5′-TTCAATGTCTGGCATGCCCAA-3′). The forward primer was designed on the first ATG codon of SGN-U575744 sequence (http://solgenomics.net/). The C-terminal part of the protein encoded by this unigene was identical to the insert in I2I-2. PCR fragments were cloned into pGEM-T Easy (Promega). After sequencing three independent clones, the consensus was compared to SGN-U575744. All PCRs were performed with Pfu (Fermentas).

Design of RNAi construct

The RNA-interfering (RNAi) hairpin constructs were produced by fusing part of the target gene to a fragment of the GUS gene (Wroblewski ; Tomilov ). Briefly, around 300–400 bp fragments covering the 3′ or 5′ end of the gene-coding region was amplified with primers in which a SfiI restriction enzyme cleavage site was introduced. The specific primer sets for SlFormin were: Formin 3′ F1 Sfi-I (5′-ATGGCCATGTAGGCCGTCCTGAGTCTTTGCAAG-3′) and Formin 3′ R1 Sfi-I (5′-ATGGCCAGAGAGGCCGACAGTGAGAGGCTGTGG); Formin 5′ F1 Sfi-I (5′-ATGGCCATGTAGGCCCGATTAGGGGCTTCAATC-3′) and Formin 5′ R1 Sfi-I (5′-ATGGCCAGAGAGGCCGAATCCATACTTGCTCGG-3′) (15-bp adapters containing the SfiI cleavage site are in bold). For SlTrax the used primers combinations were: Trax 3′ F1 Sfi-I (5′-ATGGCCATGTAGGCCCTGCAGTTTTGTGCGTGA-3′) and Trax 3′ R1 Sfi-I (5′-ATGGCCAGAGAGGCCTGCCCAACAGTGGATAAC-3′); Trax 5′ F1 Sfi-I (5′-ATGGCCATGTAGGCCCAGCGCATTCGTCACTTG-3′) and Trax 5′ R1 Sfi-I (5′-ATGGCCAGAGAGGCCGAACCCCAGGAGAATATGC-3′). The obtained fragments were fused to the GUS gene and introduced into the binary vector pGollum to create an inverted repeat structure as described before (Wroblewski ; Ament ; Krasikov ).

Plant transformation and selection of transgenic plants

S. lycopersicum cv. Motelle was transformed with the silencing constructs described above using Agrobacterium-mediated transformation using a protocol optimized for plant transformation (Ament ; Krasikov ).

Transformants were selected as described before (Ament ; Krasikov ). In short, the presence and number of the T-DNA insertions in primary transformants was assessed by analysing the presence of the neomycin phosphotransferase gene (NPTII) using PCR followed by Southern blotting with the NPTII gene as probe (data not shown). Efficiency of gene silencing in T0 parents and T1 progeny was assayed by screening for reduced GUS expression. Reduction in GUS expression was visualized by a histochemical GUS staining (Jefferson ) of transgenic plants agroinfiltrated with the pTFS40:GUS vector (Jones ; Wroblewski ) (data not shown).

Per silencing construct, two lines (called a and b) that showed the highest level of GUS silencing were selected for further analysis. In these lines, SlFormin or SlTrax silencing efficiency was measured using quantitative PCR (q-PCR).

RNA isolation, cDNA synthesis, and quantitative PCR

Total RNA was isolated from roots of five 4-week-old seedlings using Trizol LS reagent (Invitrogen) followed by a chloroform extraction and isopropanol precipitation. Additional RNA purification was performed using RNeasy mini-columns (Qiagen) and contaminating DNA was removed with DNAse (Fermentas). cDNA was synthesized from 8 μl total RNA using M-MuLV Reverse Transcriptase (Fermentas) as described by the manufacturer in a 20 μl reaction. The concentration of total RNA was estimated on agarose gels and by using a NanoDrop ND-1000 spectrophotometer. PCR was performed in a ABI 7500 Real-Time PCR system (Applied Biosystems) using Platinum SYBR Green qPCR SuperMix-UDG kit (Invitrogen). 20 μl PCR reactions contained 0.25 μM of each primer, 0.1 μl ROX reference dye, and 20 ng cDNA. The cycling program was 50 °C for 2 min, 95 °C for 5 min, and 45 cycles at 95 °C for 15 sec and 60°C for 1 min. The primer pairs used were F6qFormin (5′-ACATGCGGAACAGGACATTA-3′) and R6qFormin (5′-AAAGAGACGCAAGCCTTCAT-3′) for SlFormin amplification and F1qTrax (5′-GAGGTTAGCAATCGGTCGAA-3′) and R1qTrax (5′-TCCATCTCTGGGGCAATAAG-3′) for SlTrax amplification. Amplification with selected primer pairs was confirmed for linearity with standard cDNA dilution series and analysis of primer melting curves. Expression level was normalized to the expression of α-tubulin (TC170178) detected with the primer pair qTubL (5′-CAGTGAAACTGGAGCTGGAA-3′) and qTubR (5′-TATAGTGGCCACGAGCAAAG-3′) and compared to relative expression in an empty vector (EV)-transformed tomato line. Two independent RNA isolations followed by two cDNA syntheses represented two biological replicas. Each was tested twice with q-PCR.

Fusarium bioassays

Ten-day-old tomato seedlings of the silenced lines and an EV control line were inoculated with either an avirulent isolate of Fusarium oxysporum f. sp. lycopersici (race 2 Fol007) or with a virulent isolate (race 3 Fol029) (Rep ) using the root dip method (Wellman, 1939). Mock-inoculated seedlings served as controls. Inoculations and scoring were performed as described (Rep ). Briefly, 21 days after inoculation, plant weight above the cotyledons was measured and the disease index was determined. The disease index is correlated to the extent of browning of vessels, which is illustrated on a scale from 0 (no symptoms, none of the vessels brown) to 4 (wilt disease; all vessels brown or plant dead). One-way ANOVA and pairwise comparison with Student’s t-test for the weight measurements and the non-parametrical Mann–Whitney test for the disease index was performed using GraphPad Prism 5.0 software. The experiment was performed twice, using 40–60 plants per line.

PVX screen

Creation of the binary PVX:Avr2 and PVX:Avr2R/H constructs and their transformation to Agrobacterium tumefaciens GV3101 was described before (Houterman ). Toothpick inoculation of 3-week-old tomato plants was performed as described (Takken ; Houterman ). Inoculated and systemic leaves were scored at 8, 10, and 12 days after inoculation for development of necrotic symptoms. The hypersensitive response (HR) was quantified on an arbitrary scale from 0–4: 0, no HR; 1, only inoculated cotyledons show necrosis, plant does not display visible HR on systemic leaves; 2, HR started to develop on systemic leaves; 3, HR is more developed on systemic leaves compare to 2, upper leaves started to curl down; 4, extensive systemic HR and strong curling of the leaves (Supplementary Fig. S1, available at JXB online). The experiment was repeated twice, which allowed analysis of ±35 plants per line. One-way Anova and pairwise comparison with the non-parametrical Mann–Whitney test was performed using GraphPad Prism 5.0 software.

Results

Opposite Y2H interaction patterns of I-2-interacting proteins suggests different conformational states of I-2

Using a nearly full-length I-2 protein (amino acids 1–872) as bait, an Y2H screen of a Fusarium–tomato interaction cDNA library was carried out. This screen resulted in the identification of two interacting clones (I2I-1 and I2I-2, for I-2 interacting clones 1 and 2, respectively). The interactions were confirmed using full-length I-2 protein (Fig. 1A; left panel). To test whether these proteins interact with I-2 specifically, this study also analysed their ability to interact with two other NB-LRR proteins: Mi-1.2 and Rx. Both Rx, conferring resistance to Potato virus X (PVX) (Bendahmane ; Bendahmane ), and Mi-1.2, conferring resistance to the nematode Meloidogyne incognita, aphids, psyllid, and white flies (Milligan ; Rossi ; Vos ; Nombela ; Casteel ) are Solanaceous proteins also belonging to the CC-NB-LRR resistance protein class. As shown in Fig. 1A, only I-2 interacts with the proteins encoded by the two cDNAs as reflected by their growth on selective medium. Similar results were obtained for the CC-NB-ARC baits of those three R proteins (Fig. 1A). Notably, the CC-NB-ARC bait of an I-2 paralogue (I-2C1) showed no interaction with I2I-1 or I2I-2. I-2C1 shows 75% sequence identity with I-2 and is located on the same gene cluster; however, the gene does not confer resistance to F. oxysporum (Simons ). All full-length and truncated bait proteins were expressed in yeast (Fig. 1A; right panel) and since both interactors bind to I-2 but not to Rx, Mi-1.2 or I-2C1, these data suggested that these interactions with I-2 are specific.

Fig. 1.

Identification of I-2 domains required for yeast two-hybrid (Y2H) interaction with I2I-1 and I2I-2. (A) Y2H analyses show interactions of I2I-1 or I2I-2 with full-length I-2 as bait, but not with Rx and Mi-1.2. I2I-1 and I2I-2 interact with the CC-NB-ARC bait of I-2 (amino acids 1–520), but not with bait constructs containing the CC-NB-ARC domains of I-2C1 (amino acids 1–526), Rx (amino acids 1–426), or Mi-1.2 (amino acids 161–896). The baits and preys did not show autoactivity when co-expressed with the pACT2 and pAS2-1 empty vectors. The test for activation of the HIS3 marker is shown after 10 days of growth (left panel). Expression of the bait proteins was confirmed by Western blot (WB) analyses using α-Gal4BD antibody detection on total yeast protein extracts (right panel). (B) Y2H analyses of I2I-1 and I2I-2 with various I-2 baits: I-2FL (full-length: amino acids 1–1266), I-2N+ (amino acids 1–643), I-2N (amino acids 1–520), I-2NΔMHD (amino acids 1–520, with deletion at amino acids 484–496), and I-2CC (amino acids 1–170). The different I-2 (sub)domains are indicated and coloured: orange, coiled-coil (CC); red, nucleotide-binding (NB); purple, Apaf-1, R proteins, and Ced-4 (ARC) 1; blue, ARC2; and green, leucine-rich repeat (LRR). The test for activation of the ADE2 marker is shown after 5 days of growth; the other two markers (HIS3, LacZ) gave identical results. +, activation of all three selectable markers; –, no activation of these markers. The smallest part of I-2 enabling interaction with I2I-1 is the CC domain and for I2I-2 the CC-NB-ARC domain.

Next, this study aimed to identify the minimal fragment of I-2 required for interaction with I2I-1 and I2I-2, by testing five Y2H baits encoding: (i) full-length I-2 (I-2FL); (ii) the CC-NB-ARC carrying the first five LRRs (I-2N+); (iii) the CC-NB-ARC domain (I-2N); (iv) an CC-NB-ARC variant carrying an internal deletion in its ARC2 subdomain that removed the conserved MHD motif (I-2NΔMHD); and (v) the CC domain alone (I-2CC) (Fig. 1B). As expected, both I2I clones interacted with the full-length I-2 protein, as was indicated by clear growth of the yeast cells on selective medium after 5 days. Remarkably, a truncation that removed the 18 C-terminal LRRs (I-2N+) resulted in loss of interaction with I2I-1, but not with I2I-2. When the remaining five LRRs were removed (I-2N), I2I-1 regained its ability to interact, whereas I2I-2 lost this ability. For I2I-2, the interaction was recovered when an internal deletion in the ARC2 subdomain was made (I-2NΔMHD), but this deletion compromised the interaction with I2I-1. Truncation of the entire NB-ARC domain resulted once more in an opposite interaction pattern, as the CC domain alone (I-2CC) bound only to I2I-1. These results showed that I2I-1 binds preferentially to I-2 baits that do not bind to I2I-2 and vice versa. The observation that each bait interacts with at least one I2I-clone proved that all baits are expressed and do accumulate to high enough levels for an interaction, as was confirmed by Western blot analyses (data not shown). The opposite interaction patterns observed for both I2I clones suggest that the various I-2 truncations may have different conformations that differ in exposure of their interaction interfaces. For I2I-1, this interaction interface is present in the CC domain whereas I2I-2 apparently interacts with a surface formed by both the CC and NB-ARC domains (the NB-ARC domain alone did not interact with I2I-1 or I2I-2; data not shown). Furthermore, it indicates that the presence or absence of domains and subdomains influences these interaction interfaces, most likely by intramolecular interactions.

Differential Y2H interaction patterns coincide with active and inactive I-2 conformations

The results suggested that different conformational states of I-2 can be studied by monitoring their Y2H patterns with the two interacting clones. This hypothesis was tested by assaying the interactions between the two identified I2I clones and I-2 variants that are predicted to have different nucleotide-dependent conformations: notably wild-type I-2 and three mutants – S233F, D283E, and K207R. The nucleotide-binding states of these proteins have been determined in vitro and have been linked to their in planta phenotypes; e.g. autoactivation for I-2S233F and I-2D283E and loss-of-function for I-2K207R (Tameling ). Individual mutations were introduced into the two C-terminally truncated wild-type proteins exhibited differences in their interaction pattern with the two I2I clones, notably I-2N and I-2N+ (Fig. 1B). Interestingly, I-2NK207R showed an interaction pattern that differed from the wild type: the mutant could no longer bind I2I-1, but gained the ability to interact with I2I-2 (Fig. 2A). This confirms that I2I-2 is capable of binding I-2N without LRRs. Both I-2NS233F and I-2ND283E showed an interaction pattern identical to the wild type (Fig. 2A). Introduction of the K207R mutation into I-2N+ did not change the interaction pattern (Fig. 2B). When the effect of the individual S233F and D283E mutations in the extended bait was analysed, an interaction pattern was observed that differed from the wild type: both interacted with I2I-1 (Fig. 2B). Autoactivating mutations in the I-2N+ setting did not affect the interaction with I2I-2. Introduction of the K207R, S233F, or D283E mutation in full-length I-2 protein resulted in identical interaction patterns with I2I-1 and I2I-2 as those observed for I-2N+ (data not shown).

Fig. 2.

Yeast two-hybrid (Y2H) interaction of mutated I-2 N-terminal variants with I2I-1 and I2I-2: interaction patterns of mutated I-2N (A) and I-2N+ (B). The test for activation of the ADE2 marker is shown after 1 week of growth for I2I-1 and after 28 hours for I2I-2. The other two auxotrophic markers (HIS3, LacZ) are not shown, but gave identical results. +, activation of all three markers; –, no activation of these markers. None of the bait or prey constructs showed autoactivity. Three mutations in the NB-ARC domain were tested: a nucleotide-binding mutant K207R representing the empty state and two ATPase impaired mutants, S233F and D283E, representing the active state of I-2. In the lower part of the two panels, the interaction with wild-type (WT) baits are depicted.

To summarize, the Y2H interaction patterns of I2I-1 and I2I-2 with the various I-2 truncations and mutants are distinct and coincide with the proposed nucleotide-dependent conformational states of I-2.

Characterization of putative I-2 interactors

The distinct interaction patterns in yeast are suggestive for a functional involvement of I2I-1 and I2I-2 in I-2-mediated resistance. To learn more about the proteins encoded by these two cDNAs, this study aimed to clone the full-length cDNAs corresponding to the complete proteins. The insert of clone I2I-1 spans 1084 bp (GenBank accession number AY150043) and contains an open reading frame (ORF) of 648 nucleotides as well as a putative 3′ untranscribed region (UTR) and a poly(dA) stretch (19 residues). A BLAST search with this nucleotide sequence in the tomato genome database revealed a predicted gene (Solyc07g052730) of which the 3′ end matches 100% with the sequence of I2I-1. The I2I-1 cDNA sequence differs from this gene by six nucleotides (originating from the cloning adapter used) at its 5′ end and a poly(dA) stretch at its 3′ end. To validate the predicted gene model, this study aimed to clone a full-length cDNA. To this end a gene walking approach was employed using vector-specific and gene-specific primers to amplify and clone new sequences from the same cDNA interaction library that was used for the Y2H screens. For the amplification of a complete cDNA, one primer (I2I-1-F1) was designed corresponding to the most up-stream 5′ sequence obtained from the gene walking experiments, and one primer (I2I-2stop-R) was designed corresponding to the 3′ end of the ORF as determined in the I2I-1 clone. This gene-specific primer pair was used to amplify the full-length cDNA. Three independent full-length clones of approximately 3 kb were sequenced. These three sequences were used together with the I2I-1 insert to assemble a consensus I2I-sequence of 3425 nucleotides excluding the poly(A) stretch (GenBank accession number LN836738). The assembled consensus sequence contained a 5′ UTR of 172 nucleotides, an ORF of 2832 nucleotides encoding a protein of 944 amino acids and a 3′ UTR of 418 nucleotides, and corresponds to gene Solyc07g052730 on tomato chromosome 7 (nucleotides 58,506,937–58,513,630). Comparison of the full-length cDNA and the gene revealed a sequence identity of 100% to exon sequences and confirmed the positions of the first five introns in the gene model predicted by the SGN. A sixth intron was predicted to start one nucleotide up-stream of the stop codon in the cDNA. However, comparison of the genomic and cDNA sequences places the start of this last intron in the 3′ UTR, notably 12 nucleotides down-stream of the stop codon. The corrected gene organization is shown in Fig. 3A.

Fig. 3.

Gene model and predicted domain structure of the I-2 interacting proteins SlFormin and SlTrax. (A) SlFormin, gene Solyc07g052730 positions 58,506,937–58,513,630. SP, signal peptide; TM, trans-membrane domain; FH1 and FH2, Formin homology domains 1 and 2. (B) SlTrax, contig 1850142 (8881–13,284). NES, nuclear export signal; Translin, Translin domain. For both genes, a black line depicts the intron–exon structure, yellow boxes represent coding sequences (exons) interrupted by introns (//). The cDNA nucleotide positions of the exons are indicated above the line and the lengths of introns are given below the line. Numbers at the start and the stop refer to the position of the gene in the Sol Genomics Network database (http://solgenomics.net/). The proteins and their predicted domains are represented by rectangles and the positions of the domains are indicated above. The pink boxes indicate the regions required for binding to I-2.

The amino acid sequence encoded by the Solyc07g052730 ORF is predicted to contain a signal peptide of 35 amino acids (SignalP 3.0). However, the ORF contains an in-frame AUG triplet at codon-position 13 (as well as at position 18). A translational start at the second AUG codon would result in a protein with a signal peptide of 22 amino acids, in eukaryotes a more commonly observed signal peptide length. So, most likely the second AUG is the genuine start codon. Furthermore, the amino acid sequence contains in its C-terminal half a so-called Formin Homology 2 (FH2) domain, a hallmark domain of formins, as well as a transmembrane domain and a proline-rich domain that may correspond to the Formin Homology 1 (FH1) domain (Fig. 3A). From these observations can be concluded that Solyc07g052730 encodes a tomato Formin, and hence this gene, from which interacting clone I2I-1 was derived, is referred to as SlFormin. BLAST searches with the complete SlFormin cDNA sequence on the tomato genome in the SGN database revealed the presence of at least 13 Formin homologues. Since the I-2 interacting Formin apparently is a member of a larger protein family, this study tested whether I-2 also interacts with other Formin homologues. A close Formin homologue, SGN-U583099 (70% homology in the FH2 region) corresponding to Solyc10g006540, was selected and region corresponding to the I2I-1 encoded fragment was analysed for its ability to interact with I-2 in a Y2H assay. This homologue did not interact with I-2N (data not shown) suggesting a specific interaction of SlFormin with I-2.

The insert of clone I2I-2 spans 732 nucleotides (GenBank accession number AY150044) and is part of the unigene SGN-U575744, an assemblage of 15 tomato ESTs (SGN). Full-length cDNA was amplified from the cDNA library using primers corresponding to a (putative) start (I2I-2-ATG-F) and end (I2I-2 stop-R) of the ORF of SGN-U57544, respectively, and three full-length, independent clones were sequenced. From the unigene and the newly obtained sequences, a consensus was derived of 1274 nucleotides (excluding the poly(dA) stretch present in AY150044) with an ORF encoding a protein of 285 amino acids. A BLAST search with the consensus sequence in the SGN database indicated Solyc04g05310.2.1 on chromosome 4 as the corresponding gene. The predicted gene model includes four introns, of which one is located in the 3′ UTR five nucleotides down-stream of the stop codon. The gene encodes a protein of 196 amino acids showing 100% identity with the sequence of amino acids 90–285 of the protein encoded by the consensus. This strongly suggests that the part of the gene encoding amino acids 1–89 of the ‘consensus’ protein is missing in the tomato genome database. Indeed, an additional search in the database with the first 270 nucleotides of the consensus sequence as query gave no hit. Blasting Solanum pimpinellifolium genomic sequences in the SGN database with the consensus identified a gene with seven introns (Fig. 3B) on contig 1850142 showing nearly 100% identity with the consensus sequence. Comparing the S. pimpinellifolium gene with Solyc04g05310.2.1 showed that the two sequences are nearly the same from more or less the middle of intron 3 onwards. A BLAST search with the sequence just up-stream of Solyc04g05310.2.1 identified an identical sequence in S. pimpinellifolium contig 1850142 at 6 kb up-stream of intron 3, suggesting that a 6-kb DNA fragment is missing in the S. lycopersicum genomic database.

As mentioned above, the consensus ORF encodes a protein of 285 amino acids. However, 36 nucleotides down-stream of the start of the ORF there is a second AUG. It is possible that this codon is used as translation initiation codon. Both the longer and shorter translation products carry a translin homology domain. Moreover, a potential nuclear export signal was predicted (amino acids 190–200) (Fig. 3B). The protein has homology to a human protein called translin-associated factor X (TRAX; E value 4e−34) (Aoki ). Human TRAX is present in a complex with Translin, which is involved in microtubule-depended mRNA trafficking and translational repression (Aoki ; Cho ). The tomato I2I-2 will be referred to as SlTrax.

Generation of stably silenced SlFormin and SlTrax tomato lines

The distinct interaction patterns observed in these Y2H analyses of SlFormin and SlTrax with the I-2 bait proteins suggest active binding – or release – of these interactors depending on the nucleotide-dependent conformational state of I-2. The interaction patterns make both interacting proteins prime candidates for involvment in I-2-mediated resistance. To test if SlFormin and SlTrax are involved in I-2-mediated resistance towards Fol, transgenic tomato lines were created in which expression of SlFormin or SlTrax was knocked down using post-transcriptional gene silencing.

RNAi silencing constructs were designed to silence simultaneously the genes of interest as well as a GUS reporter gene. The latter can be transiently expressed in leaves of stable transformants by agroinfiltration to simplify screening for plants exhibiting strong silencing (Wroblewski ; Ament ; Krasikov ). For each gene two constructs were created that targetted either the 3′ or the 5′ end of the transcribed gene sequence. For each transformation at least 10 independent transformants per construct were generated. Of these transformants, around 80% was diploid and strong GUS silencing was found in 40–60% of the cases. Lines having a single-copy insertion and showing strong GUS silencing were self-pollinated, and progeny homozygous for the transgene was selected. Two transgenic lines were chosen for each (5′ and 3′) end RNAi construct and were named 5a, 5b, 3a, and 3b, respectively. Subsequently SlFormin or SlTrax transcript levels in roots were quantified using q-PCR. All selected silenced tomato lines showed a significant reduction in relative expression of the targeted genes as compared to the control line transformed with an empty vector (Fig. 4A, B). The strongest silencing was observed for the Formin 3a and Formin 5b lines; up to 90% reduction in SlFormin expression levels (Fig. 4A). Of the SlTrax RNAi lines, Trax 5a and 5b showed the strongest reduction (80%) in SlTrax expression levels (Fig. 4B). Interestingly, constructs targeting the SlTrax 5′ end exhibited stronger silencing than the 3′ silencing constructs, which did not exceed a 50% reduction level (Fig. 4B). Remarkably, both silenced plants containing the 5′ construct for SlFormin, exhibited clearly aberrant phenotypes. The plants were smaller and yellowish, and their leaf edges showed small necrotic lesions (Fig. 4C, D). The 3′ silencing-inducing constructs for SlFormin did not exhibit these phenotypes. Also none of the SlTrax-silenced lines exhibited a visible phenotype. For the subsequent experiments, the two lines for each gene that showed the highest silencing level were selected: Formin 3a and 5b and Trax 5a and 5b.

Fig. 4.

SlFormin and SlTrax expressions levels in Formin and Trax RNA interference (RNAi) tomato plants. (A, B) Transcript levels were determined by real-time quantitative PCR relative to α-tubulin in roots of 5-week-old tomato plants transformed with the empty vector EV (red) or silenced for SlFormin (A) or SlTrax (B). Depending on the RNAi construct (3′ or 5′ end), four lines per gene were analysed: 3a, 3b, 5a, 5b. Mean ± SD transcription levels for two biological and two technical replicas are shown. (C) Five-week-old representative plants, showing that Formin 5b plants are smaller and yellowish compared to empty vector-transformed tomato plants (EV). (D) Necrotic lesions at the edge of a leaf from a Formin 5b plant.

Fusarium bioassay on RNAi SlFormin and SlTrax lines

Generation of silenced SlFormin and SlTrax tomato plants materialized possibilities to test whether decreased expression influenced I-2 function. First, the response of the silenced plants to F. oxysporum (Fol) infection was tested. Ten-day-old seedlings of the silenced and EV transformed (EV) tomato lines were infected with a race 2 isolate of Fol (Fol007). This race lacks Avr1, but expresses Avr2 and hence is avirulent on this I-2 containing cultivar (Mes ). The severity of disease development is scored using a disease index that corresponds to the number of brown, infected xylem vessels in the stem (Rep ). In a resistant cultivar, such as cv. Motelle (EV), none or few plants showed brown xylem vessels resulting in a disease index of 0 or occasionally 1 (Fig. 5A). Resistance towards a race 2 isolate is so robust that water and Fol007 treatments are normally equivalent in their disease scores (data not shown; Houterman ). As shown in Fig. 5A, no significant differences in disease indexes were observed between controls and silenced lines inoculated with Fol007. Hence, silencing of neither SlFormin nor SlTrax compromised I-2-mediated defence responses towards a race 2 isolate of Fol (Fol007). Use of the same isolate on susceptible tomato plants (cv. C32) caused wilt disease, confirming that the inoculum was infectious (data not shown). This study RNAi SlFormin, SlTrax and EV tomato lines were also inoculated with a race 3 isolate of Fol (Fol029) that carries Avr3 and is virulent on Motelle (I, I-2, i-3) (Marlatt, 1996; Rep ). All plants were severely diseased (Supplementary Fig. S2), which shows that disease development and hence susceptibility is not affected. Taken together, no significant change in either disease resistance or susceptibility was observed upon Fol infection in SlFormin or SlTrax silenced plants.

Fig. 5.

Silencing of SlFormin or SlTrax does not affect I-2-mediated resistance to Fusarium oxysporum. (A) Ten-day-old seedlings of empty vector (EV) and SlFormin or SlTrax silenced tomato lines were infected with race 2 Fusarium oxysporum f. sp. lycopersici (Fol007). The disease index was determined for the indicated number of plants at 21 days post inoculation. Distribution of plants over the different disease indexes is depicted. Anova analysis reveals no significant differences in the distribution of plants over the different disease indexes between silenced (Formin 3a, Formin 5b, Trax 5a, Trax 5b) and non-silenced (EV) tomato lines upon infection with Fol007. The presented data represent two independent experiments. (B) Four-week-old tomato plants were tooth-pick inoculated with Agrobacterium carrying PVX::Avr2. Development of hypersensitive response (HR) as a result of Avr2 recognition by endogenous I-2 was quantified after 10 days post inoculation. The graph illustrates the percentage of plants with an HR intensity ranging from 1 to 4 for each line. Anova analysis indicated that the Formin 5b line showed a significantly higher HR intensity upon PVX::Avr2 inoculation. The presented data represent two independent experiments. For an illustration of the HR intensity scale, see Supplementary Fig. S1.

PVX::Avr2 screen on SlFormin and SlTrax RNAi lines

As the above-described bioassays did neither confirm nor exclude the possibility that SlFormin or SlTrax is involved in I-2 function, this study next examined the ability of the silenced lines to respond to Avr2. To express Avr2 in tomato plants, the Potato virus X (PVX)-based expression system was used (Takken ). A recombinant viral replicon carrying the Avr2 transgene was cloned into a binary vector that can be introduced in a plant by toothpick Agrobacterium-mediated inoculation. Systemic spread of the PVX::Avr2 virus, which expresses Avr2, triggers defence responses in an I-2-carrying plant. Induction of defence is visible as a hypersensitive response that spreads from the inoculated cotyledons to the non-inoculated leaves (Houterman ). The prediction was, that if SlFormin or SlTrax is involved in recognition or down-stream signalling of I-2, it could affect the timing or the extent of the HR in the silenced tomato lines inoculated with PVX::Avr2. To quantify differences in HR development, this study devised a scale from 1 (HR development only on primary inoculated leaf) to 4 (extensive systemic HR and strong curling of the systemic leaves) (Supplementary Fig. S1).

This study used as a negative control PVX::Avr2R/H, which expresses a variant of Avr2 that is not recognized by I-2 and hence does not trigger HR on an I-2 plant (Houterman ). PVX::Avr2R/H was used to exclude the possibility that the virus itself induced cell death in this screen (data not shown). Ten days post inoculation (dpi) of PVX::Avr2 on EV (EV) control plants, around half of the plants showed a class 1 HR, while the remainder showed a response of 2 or more (Fig. 5B). This intermediate stage of HR development on the controls suggested that 10 dpi is an appropriate time point to score for enhanced or reduced HR on silenced lines. One line showed significant differences in the HR distribution; the Formin 5b line. Most of the Formin 5b plants showed an HR score >1, the majority residing in group 4. The same plants, when non-inoculated, were yellowish and small and had necrotic spots on their leaf edges (Fig. 4C, D). Since appearance of necrotic lesions (HR) was used to discriminate the groups, a phenotypic influence of the Formin 5b phenotype affecting the scoring cannot be excluded. Furthermore, the retarded growth of Formin 5b plants may facilitate systemic spreading of the virus resulting in faster HR development. Nevertheless, these results indicate a trend that both SlFormin-silenced lines, regardless of their visible phenotypes, exhibited an enhanced HR development (Fig. 5B).

Discussion

I-2 Y2H interaction patterns support the molecular switch model

Upon screening an Y2H tomato–Fusarium cDNA library two proteins, SlFormin and SlTrax, were found to interact with I-2, a tomato R protein of the CC-NB-LRR class. Each interactor showed distinct binding preferences for specific truncated and/or mutated forms of I-2 (Figs. 1B and 2). The interaction patterns of SlFormin and SlTrax with the various I-2 baits were different and often contrasting, in that most I-2 baits interacted either with one or the other interacting protein. Apparently, I-2 adopts at least two conformations, a SlFormin- or a SlTrax-binding conformation. Some I-2 baits, such as the full-length I-2 and the I-2N+ ATPase mutants, bind both interactors. These baits could have an intermediate conformation or, more likely, both I-2 conformations could be present in an equilibrium in yeast. The exclusive ability to interact either with SlFormin or with SlTrax correlated with the active (I-2S233F and I-2D283E) or inactive/resting (I-2K207R) I-2 state (Tameling ). The observed interaction patterns of the mutants (Fig. 2) suggest that the SlFormin-binding conformation represents the active I-2 state, while the SlTrax-binding conformation corresponds to the resting or inactive state. Since extension of I-2N with five LRRs shifts the equilibrium towards the SlTrax-binding, it is concluded that the N-terminal LRR stabilizes the resting state.

The hypothesis that the LRR domain acts as a negative regulatory module for R proteins is in agreement with previous models (reviewed by Lukasik and Takken, 2009). These models are based on studies in which for instance deletion of the LRR domains from RPS5 (Ade ) and RPP1A (Weaver ) or swapping of LRR subdomains with those from homologous proteins, as shown for Rx (Rairdan and Moffett, 2006) and Mi-1.2 (Hwang ; Hwang and Williamson, 2003) resulting in constitutively active R proteins. These autoactive mutants induce defence signalling in the absence of the corresponding pathogen. The region required for negative regulation of Rx (Rairdan and Moffett, 2006) was pinpointed at the interface between the N-terminal part of the LRR and the ARC2 subdomain. Similarly, in the current Y2H assays, deletion of the complete LRR domain led to an interaction with SlFormin, which most likely represents the activated state. Either deletion of the MHD-motif in the ARC2 subdomain or adding the first five LRRs that are predicted to interact with the ARC2 subdomain, abolished its ability to bind SlFormin, and allowed the protein to interact with SlTrax indicative for the inactive state.

Another argument in support of a regulatory role for the ARC2 subdomain is that an internal deletion abolished the interaction of I-2 with SlFormin, which suggests that the deleted part is important for the protein to adopt the activated state. Among the 12 amino acids that were deleted is the conserved ‘MHD’ motif. Point mutations in this highly conserved motif confer either an autoactive or a loss-of-function phenotype for I-2, as well as for other R proteins, indicating its importance for R protein function (Takken ; van Ooijen ). For Mi-1.2 and I-2, the MHD motif has been proposed to fulfil the function of a sensor II motif, which regulates subdomain interactions and coordination of the bound nucleotide to control R protein activity (van Ooijen ). Delimiting the smallest interacting domain of I-2 brought another interesting observation. The active I-2 conformation seems to have an surface exposed on the CC domain that makes interaction possible with SlFormin, while in the inactive state this surface is buried as a consequence of which SlFormin binding is not possible. Apparently, activation of the R protein results in exposure of this surface, which would require relaxation of the intramolecular interaction between the CC and NB-ARC-LRR domain. This model fits that proposed for Rx where release of its CC domain from the remainder of the protein was observed upon recognition of its cognate Avr (Moffett ). Although the intramolecular interaction of the Rx CC with the NB-ARC-LRR domain was shown previously, it remained elusive to which specific domain the CC binds. Based on the current findings, these CC-mediated intramolecular interactions may involve the NB-ARC domain because, depending on the NB-ARC domain truncations or mutations, opposite interaction patterns for SlFormin and SlTrax were observed. The presence of both the NB and ARC1 subdomains apparently shields the SlFormin-interaction regions of the CC, and exposes the SlTrax binding surface.

Taken together, these observations support the refined switch model for R protein activation that states that activated and resting states of R proteins have distinct conformations controlled by their nucleotide-binding state (Lukasik and Takken, 2009). Biochemical analysis of tomato I-2, barley Mla27, and flax M proteins indicate that the activated state of an R protein is ATP bound, while the resting state is ADP bound (Tameling , 2006; Maekawa ; Williams ). For the current Y2H assays, with SlTrax and SlFormin, the same loss-of-function and autoactivating I-2 mutants were used as for the biochemical studies and I-2 activity can now be linked to at least two different conformations: a SlFormin/ATP-bound active state that requires an intact MHD motif, and a SlTrax/ADP-empty resting state that is stabilized by the LRR domain.

The use of Y2H as a proxy to study conformational changes in R proteins may be more universal. For RPM1, conferring resistance to Pseudomonas syringae, five proteins interacting with its N-terminus have been reported: TIP49a (Holt ), RIN2 and 3 (Kawasaki ), RIN13 (Al-Daoude ), and RIN4 (Mackey ). Like for I-2, their interactions in an Y2H assay required distinct regions of the RPM1 N-terminus. Consistent with the current findings, the reported interaction patterns changed depending on the NB-ARC domain truncation analysed. It would be interesting to check for these interactors whether their binding abilities correlate with the proposed activation state of the R protein.

The differences in Y2H interaction patterns for specific R protein variants are not unique for RPM1 and I-2. Many proteins interacting with either the N-terminal CC or TIR domain did no longer interact when the bait was extended to encompass the NB subdomain [NRIP1 (Caplan ), WRKY (Shen ) and RIN4 (Mackey )]. Alternatively, also for interacting proteins that require the NB subdomain, their ability to interact is often affected by the length of the bait [RIN2 and 3 (Kawasaki ); TIP49a (Holt ); RIN13 (Al-Daoude )]. The trend is that although the minimal interacting region is present in specific baits, often no interaction is found in extended baits. An explanation could be that large baits cannot enter the yeast nucleus. However, this study found that the full-length I-2 protein interacts with at least SlFormin and SlTrax showing that a full-length R protein can enter the nucleus. Together these observations imply that specific variations in interaction patterns correlate with different R protein folding states and hence can serve as markers to reflect their conformation.

The function of the interacting proteins in I-2-mediated resistance

The R protein interactors mentioned above are functionally involved in disease resistance (Holt ; Mackey ; Al-Daoude ; Kawasaki ; Shen ; Caplan ). The function of Trax in plants is unknown and, to our knowledge, this protein has not been linked to disease resistance signalling before. In mammals, TRAX regulates, together with Translin, translation by microtubule-dependent trafficking of mRNAs. SlTrax carries a nuclear export signal, similar to its animal counterparts, which suggests that it may shuttle between nucleus and cytoplasm (Fig. 3B) (Jaendling and McFarlane, 2010).

SlFormin contains a signal peptide and a transmembrane domain, which implies membrane localization (Fig. 3A). Such a localization would be similar to that observed for the orthologous class I Formins of Arabidopsis: Fh6, Fh4, Fh8, and Fh5 (Favery ; Deeks ; Ingouff ). SlFormin homologues are nucleating factors necessary for actin polymerization and stress fibre formation (Ingouff ). Operative assembly or disassembly of cytoskeleton filaments is required for proper plant growth and disruption of this process in RNAi Formin 5 plants could explain their reduced stature and the formation of necrotic lesions on their leaf edges (Fig. 4C, D). Depolarization and cytoskeleton reorganization in plants has been linked to decreased non-host resistance towards fungal pathogens such as Blumeria graminis and Colletotrichum in Arabidopsis (Yun ; Shimada ).

Functional involvement of SlFormin and SlTrax in I-2-mediated resistance was analysed using the silenced tomato lines created in this study (Figs. 4 and 5). No compromised I-2-mediated resistance or a change in susceptibility in Fol bioassays (Fig. 5A and Supplementary Fig. S2) was observed. This could imply that SlFormin and SlTrax are not involved in I-2-mediated resistance or that their involvement could not be detected using this experimental set up. A possible role of SlFormin and SlTrax in resistance could be masked by redundancy of these genes in the host. However, the results from this study do not favour this explanation as SlTrax is a single-copy gene, and an interaction between Formin and I-2 was only detected for this homologue and not for a closely related one (data not shown). Alternatively, the silencing levels in the knock-down lines (80–90%; Fig. 4) may be insufficient to confer a phenotype. This latter option can only be resolved by creating knock-out lines. Hence an alternative explanation was tested, in that I-2-mediated resistance is not compromised but enhanced, a phenotype that would be undetectable in the Fol bioassays. Avr2 was systemically expressed in the silenced tomato lines using PVX::Avr2 and onset and severity of HR was quantified. The PVX::Avr2 assay revealed a statistically significant enhancement of I-2-mediated responses to Avr2 for the Formin 5b line and a similar tendency for the Formin 3a line (Fig. 5B). However, these data have to be interpreted with care, as the possibility that the Formin 5b phenotype affected PVX spreading and HR quantification cannot be excluded.

Taken together, functional involvement in I-2-mediated resistance for the two interactors could be confirmed nor disproved. However, their interaction patterns correlate with the proposed activation states of I-2, making these interactors excellent markers to investigate I-2 activation-dependent conformational changes. As far as is known, this is the first report in which the proposed nucleotide-dependent conformation of an R protein is correlated to its ability to interact with specific proteins. The findings suggest that Y2H assays using N-terminal interactors can be used to monitor conformational changes in R proteins. For I-2, at least two conformational states appear to be present in vivo.

Supplementary material

Supplementary data are available at JXB online.

Quantification of the hypersensitive response (HR) in tomato: PVX::Avr2 HR intensity scale.

Bioassay with Fusarium oxysporum f. sp. lycopersici race 3.