OPEN ACCESS ARTICLE

This Article Free via Open Access: OA OA Abstract Full Text (PDF) OA All Versions of this Article: 140/4/1233    most recent pp.105.073510v1 Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Tameling, W. I.L. Articles by Takken, F. L.W. Search for Related Content PubMed PubMed Citation Articles by Tameling, W. I.L. Articles by Takken, F. L.W. Agricola Articles by Tameling, W. I.L. Articles by Takken, F. L.W.

CELL BIOLOGY AND SIGNAL TRANSDUCTION

Mutations in the NB-ARC Domain of I-2 That Impair ATP Hydrolysis Cause Autoactivation^1,[OA]

Plant Pathology (W.I.L.T., J.H.V., B.J.C.C., F.L.W.T.), Biomolecular Mass Spectrometry (J.A.B.), and Plant Physiology (M.A.H.), Swammerdam Institute for Life Sciences, University of Amsterdam, 1090 GB Amsterdam, The Netherlands; and Max Planck Institute for Informatics, 66123 Saarbruecken, Germany (M.A., T.L.)

	ABSTRACT

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

 
Resistance (R) proteins in plants confer specificity to the innate immune system. Most R proteins have a centrally located NB-ARC (nucleotide-binding adaptor shared by APAF-1, R proteins, and CED-4) domain. For two tomato (Lycopersicon esculentum) R proteins, I-2 and Mi-1, we have previously shown that this domain acts as an ATPase module that can hydrolyze ATP in vitro. To investigate the role of nucleotide binding and hydrolysis for the function of I-2 in planta, specific mutations were introduced in conserved motifs of the NB-ARC domain. Two mutations resulted in autoactivating proteins that induce a pathogen-independent hypersensitive response upon expression in planta. These mutant forms of I-2 were found to be impaired in ATP hydrolysis, but not in ATP binding, suggesting that the ATP- rather than the ADP-bound state of I-2 is the active form that triggers defense signaling. In addition, upon ADP binding, the protein displayed an increased affinity for ADP suggestive of a change of conformation. Based on these data, we propose that the NB-ARC domain of I-2, and likely of related R proteins, functions as a molecular switch whose state (on/off) depends on the nucleotide bound (ATP/ADP).

Most R genes encode proteins with a predicted trimodular structure (Inohara and Nunez, 2003; Martin et al., 2003). At their carboxy terminus, these trimodular proteins carry a Leu-rich repeat (LRR) domain believed to be the initial recognition domain for the following reasons: (1) This domain is highly variable and under divergent selection (Meyers et al., 2002); and (2) functional analysis has indicated that specific recognition of pathogens resides in this domain (Ellis et al., 2000). The trimodular R proteins have either an amino-terminal domain that contains coiled-coil (CC) motifs or a domain that shares homology to the Toll/interleukin-1 receptor (Pan et al., 2000b). Since the amino-terminal domain appears to determine the requirement for certain signaling components, such as EDS1 or NDR1, it is generally believed that this part is important for downstream signaling (Feys and Parker, 2000). At the center of the R protein, we find the NB-ARC (nucleotide-binding adaptor shared by APAF-1, R proteins, and CED-4) domain (van der Biezen and Jones, 1998; Aravind et al., 2004; Leipe et al., 2004) giving these proteins their name: NB-LRR proteins.

The NB-ARC domain itself consists of three subdomains: a nucleotide-binding (NB) domain and a C-terminal extension that forms a four-helix bundle (ARC1) and a winged-helix fold (ARC2; Albrecht and Takken, 2006). Mutations in the NB-ARC domain often abolish R-protein function, signifying the functional relevance of this domain (Dinesh-Kumar et al., 2000; Tao et al., 2000; Tornero et al., 2002a). We have shown that the NB-ARC domains for tomato (Lycopersicon esculentum) R proteins I-2 and Mi-1 are functional ATPase modules that specifically bind and hydrolyze ATP in vitro (Tameling et al., 2002).

Besides R proteins, the NB-ARC domain is found in animal ATPases APAF-1 and CED-4, which are both involved in apoptosis, various bacterial transcription factors, and many uncharacterized proteins (Leipe et al., 2004). Together, these proteins represent a clade of the signal transduction ATPases with numerous domains (STAND) proteins (Leipe et al., 2004). The NB domain of these STAND proteins has the so-called P-loop NTPase fold. Two well-conserved motifs in the NB site of P-loop NTPases can be distinguished: the P-loop (also called the Walker A or kinase 1a motif) and the Walker B (kinase 2) motif (Walker et al., 1982; Saraste et al., 1990; Traut, 1994). The Walker A motif has the general consensus sequence GX₄GK[T/S], wherein the invariant Lys residue is involved in coordination of the - and -phosphates of the bound NTP. Mutation of this Lys of I-2 into an Arg (K207R) greatly reduced the ability of the protein to bind ATP (Tameling et al., 2002). The Walker B motif is less conserved; hhhhD and, in R proteins, hhhhDD (where h stands for hydrophobic residue). The first Asp in this motif is invariant and involved in coordination of the Mg²⁺ ion in the catalytic site. The second acidic residue, which is frequently present in this motif, is believed to act as the catalytic base for ATP hydrolysis (Muneyuki et al., 2000). Besides the Walker A and B motifs, the NB-ARC domain of R proteins contains three other conserved motifs with unknown function: resistance nucleotide-binding site (RNBS)-A, RNBS-D, and MHD (for Met, His, Asp; van der Biezen and Jones, 1998; Meyers et al., 1999, 2003; Pan et al., 2000a).

To further our understanding of the role of nucleotide binding and hydrolysis for R-protein function, we introduced mutations in the conserved motifs of the NB subdomain of the NB-ARC of I-2. The tomato R protein I-2 confers resistance to the fungal pathogen Fusarium oxysporum (Ori et al., 1997; Simons et al., 1998). Two mutations were identified that lead to HR in the absence of an elicitor (the elicitor for I-2 has not been identified yet). Biochemical analysis of the mutant proteins revealed that they were disturbed in their ATPase activity. Three-dimensional (3D) modeling of the ATPase domain showed that both mutations map near the predicted active site of the NB subdomain. To determine whether these mutants are affected in ATP binding or in ATP hydrolysis, their reaction kinetics were studied. These data indicate that, for R-protein activation, ATP binding, but not hydrolysis, is required. Although ATP hydrolysis appears not to be needed for induction of signaling, our data suggest that it plays an important role in keeping R-protein activity in check in the absence of plant pathogens. Furthermore, our experiments show that the NB-ARC domain adopts different conformations depending on the nucleotide bound (ATP or ADP), providing novel insight into the role this domain plays as a molecular switch for R-protein activation.

	RESULTS

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

 

Point Mutations in the NB-ARC Domain of I-2 Result in Autoactivating Proteins

Little is known about the mode of action of the NB-ARC domain in plant R proteins. The NB-ARC domain of the tomato R protein I-2 contains five conserved motifs (i.e. the P-loop, the RNBS-A, the Walker B, the RNBS-D, and the MHD; van der Biezen and Jones, 1998; Meyers et al., 1999, 2003; Pan et al., 2000a; Fig. 1A ). The first three motifs are present in the NB subdomain, whereas the latter two are located in the helical extension. Previously, we have shown that the NB-ARC domain comprises a functional ATPase module (Tameling et al., 2002). Here, we study the role of NB and hydrolysis activity of this domain for I-2 protein function; therefore, we introduced mutations at specific residues in the three conserved motifs (Table I ) in the NB subdomain. These mutations were designed such that they could potentially negatively affect nucleotide binding and/or nucleotide hydrolysis. Two contiguous mutations were made in the P-loop; that is, the invariant Lys was substituted by Arg (K207R) and a highly conserved Thr was substituted by a Ser (T208S). In the RNBS-A motif, S233F was introduced because of the mutation S207F in the NB domain of the enhancer-binding protein NtrC from Escherichia coli. This S207F mutation in NtrC affected ATP hydrolysis, but did not affect nucleotide binding (Rombel et al., 1999). However, subsequent bioinformatics studies using sequence alignments revealed that this mutation in NtrC does not really correspond to S233F in I-2 as thought initially (data not shown). Furthermore, the Walker B motif was disrupted by mutating the two conserved Asp residues (i.e. D282C and D283E).

View larger version (55K): [in this window] [in a new window]   Figure 1. Selection of autoactivating mutations in I-2. A, Schematic representation of the tripartite modular structure of I-2, which consists of an N-terminal CC, a central NB-ARC, and a C-terminal LRR. In the indicated motifs, specific amino acids were subjected to mutational analysis to identify autoactivating proteins. Numbers indicate amino acid positions. B to I, Transient expression of wild type and a subset of mutant forms of I-2 in N. benthamiana leaves via agroinfiltration. In B, the response to wild-type I-2 is depicted. Autoactivating mutants I-2D495V, I-2S233F, and I-2D283E (C–E) induce a cell death response that is abolished when combined with the K207R mutation in the P-loop (F–I). Leaves were photographed 9 d after infiltration.

Previously, we have shown that the invariant Lys (K207) in the P-loop of I-2 is essential for nucleotide binding (Tameling et al., 2002). The corresponding residue had also been shown to be required for resistance mediated by the R protein N of tobacco (Nicotiana tabacum) and the resistance to Pseudomonas syringae 2 (RPS2) and resistance to Pseudomonas syringae pv maculicola 1 (RPM1<)?show [accolade7]>) R proteins of Arabidopsis (Arabidopsis thaliana; Dinesh-Kumar et al., 2000; Tao et al., 2000; Tornero et al., 2002a) and for autoactivation of Rx-MHV and L-MHV mutants (Bendahmane et al., 2002; Howles et al., 2005). To test whether an intact P-loop is required for HR triggered by the autoactivating mutations, double mutations were introduced in I-2 in which K207R was combined with either one of the three autoactivating substitutions. When transiently expressed in N. benthamiana, all three double mutants (K207R/D495V, K207R/S233F, and K207R/D283E), similar to the single K207R mutant, failed to induce an HR (Fig. 1, F–I). This result indicates that the ability of all three autoactivating mutants to trigger HR depends on a functional P-loop able to bind nucleotides.

Two Autoactivating Mutants Display Reduced ATP Hydrolysis

Both autoactivating mutations, D283E and S233F, in the NB subdomain were subsequently subjected to a biochemical study to explore the role of nucleotide binding and ATP hydrolysis for I-2 function in activating an HR. Since full-length I-2 is unstable in E. coli and yields are extremely low, truncated forms of I-2 (I-2N, I-2N^S233F, and I-2N^D283E) were produced in E. coli that encompass the amino-terminal part of the protein (i.e. the CC and NB-ARC domains, but not the LRR domain; Tameling et al., 2002). Equal purity and similar yields were obtained for wild-type and mutant proteins (data not shown). We then quantified the intrinsic ATP-binding and ATPase activity of wild-type (average of two independently purified batches) and mutant proteins. For measuring ATPase activity, purified proteins were incubated at 25°C with varying concentrations of [³²P]ATP. To measure the conversion of [³²P]ATP into [³²P]ADP in time, samples were taken at different time points and [³²P]ATP and [³²P]ADP were resolved by thin-layer chromatography (TLC). The conversion rate at each of the [³²P]ATP concentrations was used to calculate the maximal initial hydrolysis rate (V_max) using standard Lineweaver-Burk plots. The V_max of wild-type I-2N was determined to be 6.2 ± 1.5 pmol ATP hydrolyzed min^–1 µg^–1 protein (Fig. 2A ). The V_max values of I-2N^S233F and I-2N^D283E (Fig. 2A) were found to be 4.5- and 5-fold lower (1.4 ± 0.4 and 1.2 ± 0.2 pmol min^–1 µg^–1), respectively, indicating that the mutants are impaired in their ATPase activity.

View larger version (7K): [in this window] [in a new window]   Figure 2. Hydrolysis rate (Vmax) and affinity (Km) of I-2N and its mutants for ATP. Purified I-2N and mutant forms of this protein (S233F and D283E) were incubated with different concentrations of [32P]ATP. Conversion of ATP to ADP was measured in an ATPase assay and quantified. The conversion rates at each ATP concentration were plotted in a Lineweaver-Burk graph and used to determine the Vmax (pmol min–1 µg–1; A) and Km (µM; B). All reactions were done at least in duplicate. The Vmax and Km value of I-2N presented is the mean based on the outcome with two independently purified protein preparations; error bars indicate the SD.

The S233F Mutation Results in a Compromised k_cat

The ATPase cycle consists of the following steps: (1) ATP binding, (2) ATP hydrolysis, and (3) dissociation of ADP from the active site:

As mentioned previously, both wild-type and autoactivating I-2 mutants display similar affinities (K_m) for ATP (step 1 in the cycle). However, unlike the wild type, these mutants were impaired in ATPase activity (shown by the lower V_max). This reduction could be caused by (1) a decreased hydrolysis rate (k_cat; step 2) or (2) an increased affinity of the mutants for ADP after hydrolysis (lower k_off-ADP; step 3). In the presence of ATP, a decreased hydrolysis rate will result in an accumulation of the I-2N·ATP complex, whereas an increased affinity for ADP (reduced k_off-ADP) will result in an accumulation of the I-2N·ADP complex. To test which explanation is correct, we determined the relative amounts of ATP and ADP that are bound to I-2N. Therefore, we incubated on ice wild-type I-2N and I-2N^S233F with 2 µM [³²P]ATP for 15 min. The samples were then spotted on polyvinylidene difluoride (PVDF) membranes and, after extensive washing, the nucleotides bound to the I-2N proteins were extracted under denaturing conditions. The extracted nucleotides were separated on TLC plates (Fig. 3A ) and the amounts of [³²P]ATP and [³²P]ADP were quantified (Fig. 3B). For I-2N^S233F, the amount of bound ADP was reduced by approximately 70% compared to wild-type I-2N (n = 9, using two independent wild-type protein isolations). This reduction in the amount of bound ADP suggests that the reduced ATPase activity in the I-2N^S233F mutant is due to a lower hydrolysis rate (k_cat) rather than to an increased affinity for ADP (reduced k_off-ADP). Moreover, because the wild-type and mutant protein have similar ATP-binding affinities, this result implies that, under steady-state conditions, the mutant is likely to accumulate a relatively higher level of the I-2N·ATP state.

View larger version (57K): [in this window] [in a new window]   Figure 3. TLC of nucleotides bound to I-2N. [32P]ATP was incubated in duplicate either without protein (no prot.), with I-2N, or with the I-2NS233F mutant. The nucleotides retained by I-2N on the filters were extracted and separated by TLC. Radioactivity of [32P]ATP and [32P]ADP was visualized by phosphor imaging (A) and quantified (B). The prevalent nucleotide bound by I-2N or I-2NS233F is ADP. M, Marker of [32P]-labeled ATP and ADP.

View larger version (15K): [in this window] [in a new window]   Figure 4. Stability of the I-2N·ADP complex. In filter-binding assays, radioactivity bound to wild-type I-2N was quantified and expressed as relative cpm. A, In a competition experiment, [32P]ATP was premixed with an increasing amount of either cold ATP or ATP--S. This graph shows that both nucleotides compete equally well for binding. B, Determination of koff. I-2N was incubated with [32P]ATP (for 15 min) and, after addition of an excess of cold ATP, samples were taken at different time points. From this graph, the half life of the I-2·ADP complex was estimated to be approximately 30 min. C, The same competition experiment as in B, but using [32P]ADP and an excess of cold ADP. This graph shows that ATP hydrolysis is not required to obtain a stable I-2N·ADP complex. D, The formation of the stable I-2N·ADP complex depends on Mg2+. I-2N was incubated with [32P]ADP in the presence or absence of MgCl2, and radioactivity bound to the protein was quantified.

I-2N Bound to ADP Has a Different Conformation Than I-2N Bound to ATP

The instability of the I-2N·ATP complex (and the I-2N·ATP--S complex) versus the stability of the measured I-2N·ADP complex suggests a conformational change of the protein resulting in an increased binding affinity for ADP. To investigate the I-2N·ADP complex further, its stability was measured. To this end, a filter-binding experiment was performed in which I-2N was incubated on ice with [³²P]ATP. After 15 min, an excess of cold ATP was added and the mixture was further incubated for different periods of time. The relative amount of radioactivity retained on the filter ([³²P]ADP) was plotted against the incubation time (Fig. 4B). Using this plot, the time at which 50% of the radioactivity had dissociated from the I-2N·ADP complex was calculated to be approximately 30 min (t_1/2 = 30 min). This extraordinarily long t_1/2 corresponds to a dissociation rate constant (k_off) of 3.9 x 10^–4 s^–1 (k_off = ln 2/t_1/2) for I-2N·ADP. Such a low dissociation rate constant for the I-2N·ADP complex is not compatible with the previously measured affinity for ATP (K_d = 2 µM) for the empty I-2N protein (Tameling et al., 2002) because it would imply an association rate constant (k_on; k_on = k_off/K_d) of 1.91 x 10² M^–1 s^–1. Such a low k_on value is not in agreement with the observed binding kinetics of ATP or ADP to I-2N and the experimental findings can only be explained when the protein has two distinct conformations having different affinities and k_off values for ADP: an empty and ATP/ADP-bound state having a K_d of 2 µM with a relatively high k_off and a stable I-2N·ADP state with a much lower K_d and a very low k_off (3.9 x 10^–4 s^–1). Unfortunately, the K_d value of the latter complex cannot be measured directly because release of ADP causes the protein to return into its initial low-affinity state. Only indirect evidence for the existence of the stable complex, as determined by the low k_off value, can be obtained.

To answer the question of whether the formation of the stable I-2N·ADP complex requires ATP hydrolysis or whether direct ADP binding is also sufficient, filter-binding assays in which I-2N was incubated with either [³²P]ADP or [³²P]ATP were performed. In both cases, similar amounts of ADP were retained by I-2N, showing that ADP binding itself is sufficient for the formation of a stable I-2N·ADP complex and that ATP hydrolysis is not required (data not shown). To firmly establish that the complex formed by ADP binding is similar to the one that is formed by ATP hydrolysis, we repeated the filter-binding experiment as described above, but using [³²P]ADP and cold ADP as competitor instead of [³²P]ATP and cold ATP. From this experiment, the t_1/2 for ADP dissociation was calculated to be approximately 34 min (Fig. 4C). This value is similar to the 30 min found with [³²P]ATP (Fig. 4B), confirming that ATP hydrolysis activity is not required for the formation of a stable I-2N·ADP complex per se, as direct ADP binding to the empty site also results in this stable complex.

Previously, we have shown that I-2, like other P-loop ATPases, requires Mg²⁺ as a cofactor for the hydrolysis of ATP (Tameling et al., 2002). To test whether Mg²⁺ is also required for the formation of the stable I-2N·ADP complex, a filter-binding assay with [³²P]ADP was performed in the absence of Mg²⁺. When Mg²⁺ was omitted from the reaction mixture, no radioactivity was retained on the filters (Fig. 4D), which indicates that Mg²⁺ is not only required for ATP hydrolysis (Tameling et al., 2002), but also for the formation of the stable I-2N·ADP complex. Based on these data, it is not possible to distinguish whether Mg²⁺ is required for initial binding of ADP or for the conformational change leading to the stable complex or both.

Autoactivating Mutations Cluster Together in a 3D Model of the NB Domain

To reveal the relative position of Ser-233 and Asp-283 in the 3D structure of I-2, closely related NB-ARC and NACHT domains of plant and human proteins were assembled into a structure-based multiple sequence alignment (Fig. 5A ). The inclusion of both plant and human proteins allowed us to modify our previous alignment of human NACHT and NB-ARC domain proteins (Albrecht et al., 2003a). Based on structure prediction results from the BioInfo.PL metaserver and the associated 3D-Jury evaluation, including secondary structure predictions, the x-ray crystal structure of the ATPase domain of the cell division control protein Cdc6 (Protein Data Bank [PDB] identifier 1fnnA; Liu et al., 2000) was chosen as the best modeling template for the NB domain of I-2 (Fig. 5A). This template structure was also used to model the ATPase domains of the NACHT domain proteins Nod2 and Pypaf1 (Albrecht et al., 2003a), which shows that NB-ARC and NACHT domains share very similar structures. Based on the alignment with Cdc6, a 3D model of the NB domain structure of I-2 bound to ADP could be constructed (Fig. 5B). In this model, Lys-207, Ser-233, and Asp-283 (marked yellow) cluster together in close proximity to the phosphates of ADP, suggesting that these residues could indeed be directly involved in ATP binding and/or hydrolysis. Our autoactivation mutations might indirectly affect hydrolysis by disrupting the binding pocket (S233F) or directly by preventing correct bonding to the catalytic water required for the nucleophilic attack, which is essential for hydrolysis of the - phosphate bond (D283E). It might also directly interfere with nucleotide binding (K207R).

View larger version (36K): [in this window] [in a new window]   Figure 5. Structure-based multiple sequence alignment of the NB subdomain and structural model of this subdomain in I-2. A, Structure-based multiple sequence alignment of the NB subdomain of NB-ARC and NACHT domain proteins including the PDB structures 1fnnA, 1e32A, and 1bmfF. The DSSP secondary structure and the corresponding predictions by the PSIPRED server are depicted in the top and bottom part of each alignment row (-helices are represented by curled lines, -strands by horizontal arrows). Alignment columns with identical residues are highlighted in dark-gray boxes, those in which more than 60% of the residues are physicochemically conserved are shown in light-gray boxes. Mutations in I-2 and disease-associated mutations in Nod2 and Pypaf1 are denoted. Residues important in ATP binding and/or hydrolysis as well as the putative nucleotide sensor-1 residues (analogous to AAA+ proteins) are also annotated. B, Structural 3D model of the NB subdomain of I-2 (based on the template 1fnnA). -Helices are colored in red and -strands in blue; the locations of mutations are annotated in yellow and are shown as sticks together with the bound ADP and magnesium ion (Mg). -Helices and -strands are numbered according to the order in which they occur in the multiple sequence alignment.

	DISCUSSION

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

Previously, we have shown that the NB-ARC domain of I-2 functions as an ATPase (Tameling et al., 2002) allowing the protein to go through an ATP hydrolysis cycle. In this cycle, the protein binds and hydrolyzes ATP and subsequently releases ADP, allowing it to return to its empty state. Here, we provided data that ATP binding and not necessarily ATP hydrolysis by the NB-ARC domain may be the crucial step for R-protein activation. Mutagenesis of conserved motifs that together form the ATP-binding fold of the NB-ARC domain resulted in the identification of two mutations that lead to autoactivation of the defense response as manifested by a macroscopic HR, similar to I-2^D495V reported earlier (De la Fuente van Bentem et al., 2005). Both mutants, S233F and D283E, were affected in ATP hydrolysis but not in ATP binding (I-2^D495V has not been subjected to biochemical analysis because the mutation lies outside the NB subdomain). When we combined mutations S233F and D283E with a mutation in the P-loop that strongly reduces ATP binding (K207R; Tameling et al., 2002), HR was not induced in N. benthamiana. From this, we conclude that activation of I-2 requires nucleotide binding to the NB subdomain prior to activation of downstream signaling. Moreover, because the two autoactivating mutants are disturbed in ATP hydrolysis (and hence likely accumulate in the ATP-bound state), the active conformation of I-2 able to initiate an HR might be the ATP-bound state. Such a model, in which the NTP-bound state of a protein is the active state, is analogous to that for other NB-signaling proteins such as the distantly related G-proteins, in which the GTP-bound state is the active state (Sprang, 1997).

We favor a model for I-2 function in which the resting or off state of the NB-ARC domain is represented by the ADP-bound state. The ATPase activity of I-2 would reset the protein from the ATP or on state to its off state, trapping ADP in the binding pocket. This idea agrees with the high stability of the I-2N·ADP complex compared to the instability of the I-2N·ATP complex (Fig. 3). The high stability is probably caused by a conformational change upon nucleotide binding because the initial NB affinities are equal for ATP and ADP as measured in competition assays (see "Results;" Tameling et al., 2002). This finding is reminiscent of the apparently highly stable ADP-bound state of APAF-1 (Riedl et al., 2005). Because direct ADP binding to the empty protein could also induce this conformational change, resulting in a stable I-2·ADP complex, hydrolysis is not required per se. However, in the cell, this change will probably rely on hydrolysis of ATP because ATP is by far the most prevalent nucleotide in vivo. Our data suggest that ATP binding is sufficient to switch the I-2 protein into the active conformation that triggers defense signaling.

This reaction scheme differs from that for the structurally related NB-ARC protein APAF-1. Binding cytochrome c to the WD-40 repeat domain of APAF-1 results in hydrolysis of bound dATP and subsequent exchange of the produced dADP by dATP/ATP (Kim et al., 2005). Both dATP hydrolysis and nucleotide exchange are essential for formation of the activated state of APAF-1 that is able to form the apoptosome (Kim et al., 2005). Recently, the crystal structure of CED-4, the Caenorhabditis elegans homolog of APAF-1, has been solved (Yan et al., 2005). Like APAF-1, CED-4 requires ATP binding for function; however, it has been reported that CED-4 is unable to hydrolyze or exchange the nucleotide, suggesting that in this protein ATP binding could merely have a structural role (Yan et al., 2005). Pop and Salvesen (2005) suggested that these data could mean either that evidence for hydrolysis or nucleotide exchange has been missed for CED-4 or that indeed conformational cycling of this NB-ARC protein differs from that of APAF-1. In the latter case, this would imply that, although the NB-ARC domains of I-2, APAF-1, and CED-4 are structurally related, they seem to regulate the activity of the complete protein in different ways. However, the common theme in the regulation would be that their active state requires a NB pocket that is bound to ATP.

Autoactivating Mutations Inhibit ATPase Activity, Locking I-2 in the On State

To gain insight into how the autoactivating mutations may affect ATP hydrolysis, an alignment of the I-2 NB-ARC domain was assembled with related proteins of the NB-ARC and NACHT clade, both belonging to the STAND family. The NB-ARC domain consists of three subdomains: the two C-terminal subdomains called ARC1 and ARC2 (Albrecht and Takken, 2006) and the N-terminal ATPase domain that forms a 3D structure consisting of a parallel -sheet flanked by -helices (Albrecht et al., 2003a; Fig. 5B). The proposed model is very similar to the crystal structure of APAF-1 bound to ADP (Riedl et al., 2005). The ATPase fold of I-2 corresponds to the core structure of P-loop-containing NTPases (Vetter and Wittinghofer, 1999). Our 3D model of the ATPase domain of I-2 predicts that the mutations leading to constitutive activity, S233F and D283E, cluster together at the surface near the bound nucleotide (Fig. 5B).

S233F substitutes a Ser that is part of a conserved region referred to as the RNBS-A motif (Meyers et al., 1999; Pan et al., 2000b) and maps close to mutations associated with autoinflammatory disease in the NACHT domains of Nod2 (R334W/Q) and Pypaf1 (R260W). These latter mutations have been shown to cause constitutive activation as well (Chamaillard et al., 2003; Dowds et al., 2004; Tanabe et al., 2004). Similar mutations in the -subunit of the F₁-ATPase (E181 and R182, PDB structure 1bmf, chain F) were also shown to affect ATP hydrolysis (Senior et al., 2002). Together, these findings indicate that the S233F mutation disturbs an essential and conserved region required for ATP hydrolysis in R and possibly in other STAND proteins. The D283E mutation in I-2 targets the second Asp of the Walker B motif that, in R proteins, as well as in AAA+ proteins like Cdc6 (1fnnA; Liu et al., 2000), ATP-binding cassette (ABC) transporters, and helicases, has the consensus hhhhD-D/E (where h is mostly a hydrophobic residue). The second acidic residue is believed to act as the general catalytic base in ATP hydrolysis (Muneyuki et al., 2000). In several ATPases (e.g. Cdc6, helicases, and ABC transporters), this residue was indeed found to be important for hydrolysis (Herbig et al., 1999; Orelle et al., 2003; for review, see Geourjon et al., 2001). The strict conservation of this Asp residue in the NB subdomain of NB-LRR proteins suggests that this residue may also act as the catalytic base in ATP hydrolysis in NB-LRR proteins.

The biochemical basis for the autoactivation of the D495V mutation is not known. The mutation lies close to the C terminus of the ARC2 subdomain and is part of the conserved MHD motif. In the structurally related APAF-1 protein, the imidazol residue from the analogous His forms a hydrogen bond with the -phosphate of the bound ADP (Riedl et al., 2005). Riedl and coworkers (2005) propose that this residue not only coordinates the nucleotide, but also connects the ARC2 subdomain to the binding pocket, thereby deeply burying ADP. Because of the significant, but weak, interaction of ARC2 with ADP, they suggest that this domain is prone to conformational changes. It is tempting to speculate that the MHD motif in I-2 has a similar function. The autoactivation phenotype caused by the D495V substitution would then be due to destabilization of ADP binding. This would allow the I-2 protein to release ADP, bind ATP, and hence switch to its active state in the absence of an elicitor. Currently, we are investigating this possibility.

Novel Intramolecular Interactions

Moffett and coworkers recently proposed a model for R-protein activation based on intramolecular interactions in the R-protein Rx, also a CC-NB-LRR (Moffett et al., 2002). They were able to demonstrate that the NB-ARC domain of Rx interacts with the LRR domain and likely with the CC domain as well (CC interacts with the NB-LRR part). The interaction with the LRR is disrupted upon recognition of the virally encoded avirulence protein that serves as the elicitor of the Rx-mediated defense response. So far, our attempts to perform similar experiments were unsuccessful because we have not been able to detect full-length I-2 or truncated versions of I-2 produced in planta using polyclonal antibodies raised against part of the ARC2 subdomain. Also N- or C-terminal versions of epitope-tagged I-2 expressed in planta could not be detected upon immunoprecipitations or pull-down assays. Nevertheless, we believe that the role of the NB-ARC domain in I-2 is consistent with the observations made for Rx.

We here propose a refined model for R-protein function in which the NB-ARC plays a pivotal role in regulating signaling (Fig. 6 ). As proposed for Rx, the resting state of the R protein is represented by a situation where the NB-ARC domain interacts with both the CC and LRR domain. This interaction requires a functional P-loop as observed for Rx. Moreover, we could show that I-2 function also requires an intact P-loop (K207R mutant). The switch from the off (ADP) to the on (ATP) state is triggered by either direct or indirect recognition of the elicitor. For Rx, this results in dissociation of the LRR from the CC-NB-ARC part of the protein. Our data would add that elicitor perception by the LRR domain triggers a conformation change in the NB-ARC domain leading to destabilization of the ADP-bound state (step 1), allowing rapid dissociation of ADP (step 2). A positive regulatory function of the LRR domain can be deduced from the observation that deletion of the LRR does not result in constitutive activity for Rx (Moffett et al., 2002), the I-2 wild type, or the hydrolysis mutants (data not shown). Also, many truncated and splice variants of R proteins that lack an LRR are encoded by the plant genome and these variants do not spontaneously induce a defense response (Lawrence et al., 1995; Anderson et al., 1997; Parker et al., 1997; Ayliffe et al., 1999; Dinesh-Kumar et al., 2000). A positive regulatory function for the LRR also fits with the finding that the elicitor-independent HR, triggered by a constitutively active Mi-1 chimera, could be abolished by numerous independent mutations in this domain (Hwang and Williamson, 2003). Furthermore, a positive regulatory function for the LRR can be inferred from our observation that in vitro the I-2N form, lacking the LRR domain, has a low, instead of a high, dissociation rate for ADP.

View larger version (9K): [in this window] [in a new window]   Figure 6. Proposed model of the NB-ARC domain functioning as a molecular switch in the regulation of R-protein-mediated signaling. In the resting (off) state, the CC and LRR domains are bound to the NB-ARC domain. The NB-ARC tightly binds ADP and this interaction is stabilized by the LRR domain. (1) Upon elicitor recognition, the LRR-NB-ARC interaction is disturbed, resulting in reduced affinity for ADP. (2) This reduced affinity results in rapid dissociation of ADP. (3) The free state binds ATP that triggers a conformational change, resulting in disruption of the CC-NB-ARC interaction. This active (on) state is competent to activate signaling, leading to induction of defense responses. (4) Hydrolysis of the bound ATP by the intrinsic ATPase activity returns the protein to its resting state. The steps affected by the mutations in the NB subdomain are indicated.

Hydrolysis of the bound ATP by the intrinsic ATPase activity (step 4) brings the protein back into its resting (off) state completing the ATPase cycle. We propose that this ATPase cycle also runs in vivo, albeit at a lower speed, in the absence of the pathogen. Evidence for that is the observation that there is a certain degree of constitutive activity of R-gene pathways in plants in the absence of elicitors (for review, see Nimchuk et al., 2003), resulting in an overall fitness penalty (Korves, 2004). A change in the reaction speed of any of the different steps in the ATPase cycle, either by pathogen perception or mutations in the NB-ARC domain, will influence the amount of I-2·ATP. Presumably, when the concentration of this complex exceeds a certain threshold, defense signaling will be initiated. This dynamic behavior of R proteins can explain why small disturbances in the amount of R protein in the cell can result in induction of defense signaling (Jin et al., 2002; Tornero et al., 2002b; Bieri et al., 2004). The fact that no HR is observed upon transient expression of wild-type I-2 in N. benthamiana indicates that, in this case, this threshold is not exceeded. It agrees with the finding that very low amounts of I-2 protein are produced in planta because the levels are below our detection levels. In summary, it is likely to be of crucial importance to the plant to tightly regulate the expression levels of R proteins.

Downstream Signaling

How the proposed active, ATP-bound state of I-2 triggers downstream signaling is not known. The related AAA+ ATPase proteins often rely on oligomerization for function (for review, see Hanson and Whiteheart, 2005) and oligomerization has also been shown for several STAND proteins, including MalT, APAF-1, C2TA, and Nod1 (for review, see Leipe et al., 2004). Recently, for the tobacco N R protein, a TIR-NB-LRR protein, it was shown that oligomerization is induced in the presence of the elicitor (Mestre and Baulcombe, 2006). So far oligomerization has not been shown for CC-NB-LRR proteins, and an alternative mechanism could be that the activated, NTP-bound state of the protein functions as a monomer, similar to G-proteins (Sprang, 1997).

Future experiments may reveal the molecular mechanism through which activated R proteins trigger downstream signal transduction cascades. In addition, more direct measurements of changes in protein conformation are required to establish how the nucleotide state controls this conformation and thereby the activity of the protein. Furthermore, we need to understand the switch function of the NB-ARC domain in relation to intramolecular interactions and other factors that may influence the rate of hydrolysis and nucleotide exchange similar to what has been found for G-proteins (Sprang, 1997).

	MATERIALS AND METHODS

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

 

Constructs

Mutations were introduced by overlap extension PCR (Higuchi et al., 1988) using I-2 as a template. For each mutation, two flanking primers and the following set of mismatch primers were used: T208S (5'-CCAGGGCAAGTCTACACTTGC and 5'-GCAAGTGTAGACTTGCCCTGG); S233F (5'-TGGTATTGCGTTTTTGAAGGATTTGA and 5'-TCAAATCCTTCAAAAACGCAATACCA); D282C (5'-TCCTTATTGTTTTGTGTGATGTGTGGA and 5'-TCCACACATCACACAAAACAATAAGGA); and D283E (5'-TCCTTATTGTTTTGGATGAAGTGTGGA and 5'-TCCACACTTCATCCAAAACAATAAGGA) All mutations were confirmed by sequence analysis. All fragments were digested with NcoI and Tth111I to replace the corresponding fragment in the yeast (Saccharomyces cerevisiae) two-hybrid bait vector pAS2-1 (CLONTECH Laboratories) containing I-2N (1–519; described in De la Fuente van Bentem et al., 2005; herein referred to as CC-NB-ARC). Restriction fragments containing the mutation were excised from these vectors with SalI and BamHI to replace the corresponding fragments in the pGreen 1K-derived binary vector encoding full-length I-2F (1–1,266; WP42). In this vector, the gene is under the control of the cauliflower mosaic virus 35S promoter and a nopaline synthase transcription terminator. WP42 and the binary vector harboring I-2F^D495V (WP45) are described in De la Fuente van Bentem et al. (2005). The SalI-BamHI fragments containing the sequences coding for the S233F and D283E mutations were used to replace the corresponding fragments in I-2 present in the Escherichia coli expression vector pGEX-KG (Tameling et al., 2002). In this article, construction of pGEX-KG containing the I-2N^K207R mutant is described. To construct the binary vector that contains I-2F^K207R/D495V, the SalI-BamHI fragment containing the K207R mutation was isolated from the pGEX-KG construct containing I-2N^K207R, and used to replace the corresponding fragment in WP45. The pGEX-KG construct was used as a template for overlap extension PCR, as described above, to gain DNA fragments containing a second mutation coding for either S233F or D283E. These fragments were digested with SalI and BamHI and used to replace the corresponding fragment in WP42. This resulted in pGreen 1K containing the double mutants I-2F^K207R/S233F and I-2F^K207R/D283E.

Agroinfiltration

All binary vector (pGreen 1K derived) constructs were transformed to Agrobacterium tumefaciens GV3101 and agroinfiltration was performed as described in Van der Hoorn et al. (2000). The cells were spun down and resuspended to an OD₆₀₀ of 2. Nicotiana benthamiana plants used for infiltration were 4 to 5 weeks old. Plants were grown in a greenhouse at 25°C under supplemental lighting.

Multiple Sequence Alignment and Structural Modeling

Protein sequences were retrieved from the UniProt (Apweiler et al., 2004) database and protein domain architectures from the Pfam (Bateman et al., 2004) and SCOP (Andreeva et al., 2004) databases. The synonymous names of the protein sequences and UniProt accession numbers are as follows: I-2, Q9XET3; Mi-1, O81137; Rx, Q9XGF5; Prf, Q96485; Rpi-blb1/RGA2, Q7XBQ9; RPM1, Q39214; RPS2, Q42484; N, Q40392; L6, Q40253; PYPAF1/CIAS1/CRYOPYRIN/NALP3, Q96P20; NOD1/CARD4, Q9Y23; NOD2/CARD15/IBD1, Q9HC29; PYPAF2/NALP2/NBS1/PAN1, Q9NX02; PYPAF3/NALP7/NOD12, Q8WX94; PYPAF4/NALP4/PAN2, Q96MN2; PYPAF5/NALP6/PAN3, P59044; PYPAF6/NALP11/NOD17, P59045; PYPAF7/MONARCH-1/NALP12/PAN6, P59046; PYPAF8/MATER/NALP5, P59047; DEFCAP/CARD7/NALP1/NAC, Q9C000; CLAN/CARD12/IPAF, Q9NPP4; and CIITA/MHC2TA, P33076.

Protein structures were obtained from the PDB database (Bourne et al., 2004). The secondary structure assignments of PDB structures were taken from the DSSP database (Kabsch and Sander, 1983). A single capital letter appended to the actual PDB identifier denotes the chosen structure chain. The state-of-the-art online server PSIPRED (McGuffin et al., 2000) predicted the secondary structure of NB-ARC and NACHT domain proteins. Because of recent improvements of the PSIPRED method, we also recomputed the secondary structure of NACHT domain proteins discussed in our previous studies (Albrecht et al., 2003a, 2003b). The structure-based multiple sequence alignment of NB-ARC and NACHT domain proteins (Fig. 5A) was assembled using precomputed Pfam domain alignments and T-COFFEE (Poirot et al., 2003). We also improved them manually by minor adjustments based on structure prediction results and pairwise superpositions of all PDB structures. The respective superpositions were computed by the program CE (Shindyalov and Bourne, 1998). For the alignment construction, we used the sequences of the NB-ARC domain proteins I-2, Mi-1, Rx, Rpi-blb1, Prf, RPS2, RPM1, L6, N, APAF-1, and CED-4 and the NACHT domain proteins PYPAF1-8, DEFCAP, CLAN, CIITA, and NOD1/2. In addition, the more distantly related structure of the -subunit of the F₁-ATPase (PDB identifier 1bmfF) was included because the role of its functionally relevant amino acids has been investigated in detail. Furthermore, we investigated the results of all state-of-the-art fold recognition methods available via the online metaserver BioInfo.PL (Bujnicki et al., 2001), which contacts a dozen other prediction servers (whose names are listed on the Web site http://Bioinfo.PL/Meta). The associated 3D-Jury system allows for the comparison and evaluation of the predicted 3D models in a consensus view (Ginalski and Rychlewski, 2003). To obtain a 3D model of the NB domain structure of I-2, we extracted the sequence structure alignment of I-2 to 1fnnA from the multiple sequence alignment of NB-ARC and NACHT domain proteins (Fig. 5A) and submitted it to the 3D modeling server WHAT IF (Rodriguez et al., 1998). An alternative modeling template with a very similar structure despite low sequence identity would have been the vasolin-containing protein VCP, also known as membrane fusion ATPase p97 (PDB identifier 1e32A; Zhang et al., 2000). Further possible templates would be the crystal structures of APAF-1 and CED-4, which were published after the completion of this study (Riedl et al., 2005; Yan et al., 2005). The sequence alignments depicted in the figures were prepared in the SEAVIEW editor (Galtier et al., 1996) and illustrated by the Web service ESPript (Gouet et al., 2003). The protein structure images were drawn in the Accelrys Discovery Studio ViewerLite.

Purification of Glutathione S-Transferase Fusion Proteins from E. coli

I-2N, I-2N^K207R, I-2N^S233F, and I-2N^S283E were expressed as glutathione S-transferase fusions from the pGEX-KG-derived constructs in E. coli. Expression, purification, and renaturation were performed as described previously (Tameling et al., 2002). Protein concentration was determined by the Bradford method using bovine serum albumin as a standard.

ATPase Assay

ATPase assays were performed as described previously (Tameling et al., 2002). To measure the V_max and K_m at 25°C, either 42 or 118 ng active I-2N, I-2N^K207R, I-2N^S233F, and I-2N^D283E were incubated with three different [³²P]ATP concentrations (0.6, 2, and 8 µM with specific activities of 7.4 x 10⁵, 2.2 x 10⁵, and 5.6 x 10⁴ cpm/pmol, respectively) in a reaction volume of 16 µL. Samples were taken 0, 10, 20, and 40 min after starting the reaction. The amount of [³²P]ADP and [³²P]ATP in each sample was quantified from a TLC plate by using phosphor imaging (Storm; Molecular Dynamics). Data points were plotted and the hydrolysis speed ( in pmol min^–1) for each reaction was determined by linear curve fitting. The speed values were used for each protein preparation to create Lineweaver-Burk plots from which the V_max (pmol ATP hydrolyzed min^–1 µg^–1 active protein) and K_m (µM) values were extracted.

[³²P]ADP Production Using Hexokinase

[³²P]ADP was produced by 5-min incubation at 25°C of [³²P]ATP with 2 units of hexokinase (Sigma-Aldrich) in a reaction volume of 120 µL that contained 20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 5 mM MgCl₂, and 100 mM Glc. To remove the hexokinase, the reaction was loaded on a microcon YM-10 filter unit (Millipore), centrifuged for 30 min at 13,000 rpm in a microcentrifuge, and subsequently washed with 100 µL water. The flow through was collected and the preparation was heated for 2 min at 100°C. No hexokinase activity could be detected in this final preparation. Full conversion to [³²P]ADP was confirmed by TLC and the concentration was quantified using liquid scintillation counting.

Filter-Binding Assays

Filter-binding assays were performed according to Tameling et al. (2002), except that the reaction mixtures contained 10 mM MgCl₂.

Nucleotide Extraction from Filters
I-2N (0.31 µg active protein in 60 µL) or buffer only were incubated with 2 µM [³²P]ATP (5.8 x 10⁴ cpm/pmol) on ice and subjected to a filter-binding assay. Glycerol was omitted from the washing buffer, which did not affect ATP-binding characteristics. Deoxynucleotides were extracted from the filters by soaking them for 1 h in 700 µL 60% methanol (method adapted from Palmer and Cox, 1994). After incubation, 550 µL of the extracts were transferred to clean Eppendorf tubes and the liquid was evaporated using dry, flowing air. Pellets were dissolved in 10 µL water, of which 6 µL were used for TLC similar to that described for the ATPase assay (Tameling et al., 2002). [³²P]ATP and [³²P]ADP separated on the TLC plate were visualized by autoradiography using x-ray films (Fuji Photo Film) and the amount of radioactivity was quantified by using phosphor imaging (Storm; Molecular Dynamics).

Competition Filter-Binding Experiment
I-2N (0.24 µg active protein) was incubated for 15 min on ice with 0.15 µM [³²P]ATP (5.8 x 10⁵ cpm/pmol) and varying amounts of either cold ATP or ATP--S (Fermentas and Roche) in 55-µL standard filter-binding reactions. Fifty microliters of the reactions were spotted on PVDF (Millipore) membranes and radioactivity was measured by liquid scintillation counting.

Nucleotide Dissociation Experiments
I-2N (0.18 µg active protein) was incubated on ice with 2 µM [³²P]ATP (6.8 x 10⁴ – 1.1 x 10⁵ cpm/pmol) in 35-µL reaction mixtures. After 15 min, 20-µL reaction buffer containing 2.5 mM cold ATP was added (to dilute free [³²P]ATP). These mixtures were incubated for different time periods prior to spotting on the PVDF membranes. To measure ADP dissociation, the same experiment was performed except that 2 µM [³²P]ADP (2.65 x 10⁴ cpm/pmol) and cold ADP were used.

Because of the apparent high stability of the I-2N·ADP complex, it was possible to measure the absolute amount of [³²P]ADP bound to I-2N. This was done in triplicate at the concentration required for half-maximal binding (2 µM). The amount at saturation was calculated to deduce the amount of active protein in the preparation.

MgCl₂ Dependency of ADP Binding
I-2N (0.29 µg active protein) was incubated on ice with 0.2 µM [³²P]ADP (7.6 x 10⁴ cpm/pmol) in 55 µL of either standard reaction mixtures or mixtures without MgCl₂ added and in the presence of 10 mM EDTA (to chelate MgCl₂ [0.1 mM] carried over from the [³²P]ADP preparation).

	ACKNOWLEDGMENTS

 
We thank Martijn Rep, Harrold van den Burg, and David Baulcombe for critical reading of the manuscript and the referees of this paper for their useful comments.

Received October 28, 2005; returned for revision January 25, 2006; accepted January 26, 2006.

	FOOTNOTES

 
¹ This work was supported by Bioseeds/Keygene (J.V.), the German Research Foundation (contract no. LE 491/14–1 to M.A. and T.L.), the German National Genome Research Network, and the BioSapiens Network of Excellence funded by the European Commission (grant no. LSHG–CT–2003–503265).

² These authors contributed equally to the paper.

³ Present address: The Sainsbury Laboratory, John Innes Centre, Norwich Research Park, Colney, Norwich NR4 7UH, UK.

⁴ Present address: Laboratory of Phytopathology, Wageningen University, Binnenhaven 5, 6709 PD Wageningen, The Netherlands.

The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (www.plantphysiol.org) is: Frank L.W. Takken (takken{at}science.uva.nl).

^[OA] Open Access articles can be viewed online without a subscription.

Article, publication date, and citation information can be found at www.plantphysiol.org/cgi/doi/10.1104/pp.105.073510.

^* Corresponding author; e-mail takken{at}science.uva.nl; fax 31–20–5257934.

	LITERATURE CITED

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

 
Albrecht M, Domingues FS, Schreiber S, Lengauer T (2003a) Structural localization of disease-associated sequence variations in the NACHT and LRR domains of PYPAF1 and NOD2. FEBS Lett 554: 520–528[CrossRef][ISI][Medline]

Albrecht M, Lengauer T, Schreiber S (2003b) Disease-associated variants in PYPAF1 and NOD2 result in similar alterations of conserved sequence. Bioinformatics 19: 2171–2175[Abstract/Free Full Text]

Albrecht M, Takken FLW (2006) Update on the domain architectures of NLRs and R proteins. Biochem Biophys Res Commun 339: 459–462[CrossRef][ISI][Medline]

Anderson PA, Lawrence GJ, Morrish BC, Ayliffe MA, Finnegan EJ, Ellis JG (1997) Inactivation of the flax rust resistance gene M associated with loss of a repeated unit within the leucine-rich repeat coding region. Plant Cell 9: 641–651[Abstract]

Andreeva A, Howorth D, Brenner SE, Hubbard TJ, Chothia C, Murzin AG (2004) SCOP database in 2004: refinements integrate structure and sequence family data. Nucleic Acids Res 32: D226–D229[Abstract/Free Full Text]

Apweiler R, Bairoch A, Wu CH, Barker WC, Boeckmann B, Ferro S, Gasteiger E, Huang H, Lopez R, Magrane M, et al (2004) UniProt: the universal protein knowledgebase. Nucleic Acids Res 32: D115–D119[Abstract/Free Full Text]

Aravind L, Iyer LM, Leipe DD, Koonin EV (2004) A novel family of P-loop NTPases with an unusual phyletic distribution and transmembrane segments inserted within the NTPase domain. Genome Biol 5: R30[CrossRef][Medline]

Ayliffe MA, Frost DV, Finnegan EJ, Lawrence GJ, Anderson PA, Ellis JG (1999) Analysis of alternative transcripts of the flax L6 rust resistance gene. Plant J 17: 287–292[CrossRef][ISI][Medline]