Two distinct biochemical activities have previously been reported for NB-ARC family proteins in plants. The I-2 and Mi-1 R-proteins of tomato are ATPases whose signaling capability is proposed to be determined by a structural state corresponding to the bound nucleotide (8
). The Pollen Signaling Protein (PSiP) of maize has a domain structure similar to R-proteins and has been proposed as an AC (17
). These findings suggest that the modes of biochemical activity among this class of proteins might be more diverse than previously suspected and we chose to investigate this intriguing possibility through an in depth analysis of NBD biochemistry.
A purified recombinant protein was required to investigate the biochemistry of plant R-proteins but a natively soluble active plant recombinant NBD has never been reported. We therefore tested a range of R-protein NBDs for expression of soluble recombinant protein in E. coli
. R-protein NBDs tested included those from ORFs At3g14460, At3g14470, and At3g50950 of A. thaliana
, AAX89832 of Glycine max
, CAD45035 of Hordeum vulgare
, AAX61322 of Phaseolus vulgaris
, and Os09g_13820 of Oryza sativa
. Among the proteins tested was Os02g_25900, an 809 amino acid orphan R-protein of rice carrying an ~200 amino acid N-terminal extension. This N-terminal extension contains a CC domain, a domain commonly associated with the N terminus of R-proteins (A
). Amino acids 197–334 of Os02g_25900, corresponding to the NBD (hereafter called R1-NB), was expressed at relatively low levels (~10 μg of purified protein/liter bacterial culture) and co-purified with a single contaminating protein after affinity purification. MALDI-TOF analysis of tryptic digests identified this protein as the Hsp70-family member DnaK, a chaperone involved in protein folding (B
). DnaK could be removed through a combination of high salt treatment and anion exchange chromatography to give the homogenous R1-NB protein (C
FIGURE 1. Expression of a R-protein NBD.
A, protein domains of Os02g_25900. CC, coiled coil domain; NBD, nucleotide binding domain; ARC, Apaf-1, R-protein, CED-4 domain, LRR, leucine-rich repeat. B, analysis of protein co-purifying with R1-NB (SDS/PAGE analysis (more ...)
We examined the range of nucleotides utilized by R1-NB. R1-NB catalyzed the conversion of all adenine nucleotides to adenosine (Ado) with a substrate preference order of AMP>ADP>ATP (A
; the ratios of nucleoside product to unused substrate as measured by peak areas are 2.9, 1.6, and 0.6 for AMP, ADP, and ATP, respectively). Radiolabeled substrates and analysis by TLC were used to validate this finding (B
). The identity of the product as Ado was confirmed by ion fragmentation in an electrospray ionization mass spectrometer. Product generated from R1-NB was indistinguishable from an Ado standard (C
). No cAMP was produced indicating that Os02g_25900 is not an AC as proposed for PSiP (14
). The Ado-generating activity was not due to another protein as no contaminants were visible by SDS-PAGE and electrospray ionization mass spectrometry in positive ion mode (D
). To prove that the observed nucleotidase activity was not an experimental artifact we used an unrelated plant protein (Arabidopsis
)) expressed from the same vector and purified and assayed in exactly the same manner as R1-NB as a control. This preparation gave no significant nucleoside generating activity (E
). As it may be formally possible that the activity observed is due to a minor contaminating protein that does not ionize in electrospray mass spectrometry we mutated a conserved amino acid (K211A) in the Walker-A/P-loop motif homologous to residues presumed to bind the β and γ phosphate moieties in P-loop ATPases (2
) and expressed as a recombinant protein (A
). We noted a significantly reduced specific activity in the K211A mutant compared with wild type protein (B
). To be absolutely certain that the catalytic activity observed for R1-NB is not due to a contaminant we chose to mutate additional catalytic residues. Modeling the structure of R1-NB using Apaf-1 as template identified Asp-294 and Asp-295 as likely residues to co-ordinate a catalytic metal ion (, C
). R1-NB D294A/D295A was expressed as a recombinant protein (A
) and assay revealed, again, a significantly lowered specific activity compared with wild type (B
FIGURE 2. R1-NB generates Ado from adenine nucleotides.
A, 5.2 μm R1-NB was assayed in 100 μl reactions in the presence of 100 μm ATP (upper panel), ADP (middle panel), or AMP (lower panel). Reactions were analyzed by HPLC. B, 9.75 μ (more ...)
FIGURE 3. R1-NB P-loop mutants show defective nucleotidase activity.
A, purification of recombinant R1-NB K211A and R1-NB D294A/D295A (SDS/PAGE analysis and Coomassie Blue staining). 1.5 μg portions of protein were applied, and molecular mass standards (more ...)
R1-NB is therefore not an AC nor is it an ATPase similar to other characterized NB-ARC domain proteins. Unexpectedly, R1-NB was found to have a nucleoside generating catalytic activity. R1-NB was also able to utilize guanine nucleotides to generate guanosine with a similar substrate preference order as observed for adenine nucleotides (, A–C; the ratios of nucleoside product to unused substrate as measured by peak areas are 3.7, 0.9, and 0.5 for GMP, GDP, and GTP, respectively). R1-NB was able to utilize UTP and CTP as substrate (, D and E). A small amount of deoxynucleotide diphosphate was generated from the appropriate triphosphate with no evidence for the production of deoxynucleoside (, F and G). This activity was, however, very weak compared with the utilization of the corresponding nucleotides.
R1-NB generates nucleoside from nucleotides. 5.2 μm R1-NB was assayed in 100 μl reactions in the presence of 100 μm (A) GTP, (B) GDP, (C) GMP, (D) CTP, (E) UTP, (F) dATP, and (G) dGTP. Reactions were analyzed by HPLC.
To further characterize the R1-NB protein we examined the response of R1-NB to temperature and pH. A pH optimum of 7.5 and temperature optimum of 50 °C was found using ADP as substrate (, A
). Further assays demonstrated that divalent metal was required for R1-NB activity as would be expected for a P-loop family protein with a nucleotidase activity and an EC50
of 4.1 mm
is consistent with related molecules (C
). Kinetic analysis of R1-NB using ADP as substrate gave a Km(app)
of 53 ± 7 μm
of 16.6 ± 0.4 nmol Ado mg−1
(corresponding to a kcat
of 0.28 min−1
). Although the affinity of R1-NB for the nucleotide is lower than that of the affinity of the I-2 CC-NB-ARC domains for ATP (Kd
= 2.2 μm
)), an affinity in the μm
range demonstrates that the nucleoside generating activity is a likely biologically relevant activity and not an artifact of assay in vitro
under extreme conditions. A sigmoid for the substrate response gives a Hill Slope of 1.7 consistent with R1-NB acting as a multisubunit protein. Analytical gel filtration revealed a major peak at just over 45 kDa that contained only the R1-NB protein (17 kDa) as assessed by SDS-PAGE (D
). It is tempting to assign this peak as a R1-NB dimer structurally analogous to the asymmetric dimer formed by the CED-4 nucleotide binding α/β fold in octameric apoptosome assembly (18
). To provide additional evidence for dimer formation we performed chemical cross-linking of R1-NB using the E. coli
catabolite-activated protein (CAP) as positive control and ribonuclease A1 and chymotrypsinogen A1 as negative controls (E
). Cross-linking of R1-NB and CAP gave clear bands corresponding to a protein dimer that could be eliminated by reduction of the cross-link (lower arrow
for R1-NB). No such bands were evident for ribonuclease A1 or chymotrypsinogen A. The protein bands for the negative controls are reduced through smearing up the gel by nonspecific modification by the cross-linking agent but no clear multimer bands are observed. A faint higher molecular weight band in the CAP lanes (at around 72 kDa) is evident but was a minor contaminant from protein purification and not a band corresponding to cross-linking. A higher molecular weight band was consistently observed on cross-linking of R1-NB (upper arrow
for R1-NB). This band could be speculatively assigned as a trimer, consistent with analytical gel filtration and asymmetric multimerization, but an unambiguous assignment of the multimer status will require a full structural analysis. R1-NB was assayed in the presence of the reducing agents dithiothreitol or monothioglycerol (supplemental Fig. S1
) to examine whether a redox sensitive linkage might contribute to enzyme activity, for example through an intermolecular bond. Increasing concentrations of dithiothreitol had no influence on specific activity. Elevated concentrations of monothioglycerol gave a very small but statistically relevant reduction in specific activity. The lack of any effect of dithiothreitol and the small effect of monothioglycerol does not support a role for an intra- or intermolecular redox sensitive bond essential for activity.
FIGURE 5. Influence of assay variables on R1-NB activity assayed by TLC.
A, 3.76 μm R1-NB was assayed in the presence of 1 mm ADP at varying pH (n = 8). B, 3.76 μm R1-NB was assayed in the presence of 1 mm ADP at varying temperature (n = 8). C, (more ...)
The generation of nucleoside from nucleotide could occur by two distinct mechanisms. The first possibility is that R1-NB is a nucleotide monoester hydrolase that cleaves nucleotides at the nucleoside/α-PO42−
bond. The second possibility is that R1-NB is a nucleotide phosphatase that sequentially cleaves terminal phosphates. 31
P NMR analysis of R1-NB using ADP as substrate was used to select between these two alternatives. A single peak corresponding to PO42−
was observed (A
, black trace
) and no peak corresponding to P2
was observed indicating that R1-NB is a nucleotide phosphatase. A formal possibility is that P2
is generated and subsequently cleaved by R1-NB, however, R1-NB was unable to convert the non-hydrolyzable ATP analog adenosine 5′-(β,γ-imido)triphosphate demonstrating that the terminal phosphate bond is the target (B
). NMR in the presence of 40% (v/v) H2
revealed an asymmetric PO42−
peak with a shoulder, indicating a second signal in this peak that is not present when the assay is performed in H2
, red trace
). Deconvolution of the PO42−
peaks demonstrated the presence of a minor peak with a small chemical shift in the presence of H2
) but not with H2
). This finding is consistent with the production of P16
in the assay performed with H2
). Taken together all the data are consistent with a mechanism whereby divalent metal activated water mounts a nucleophilic attack on the terminal ester bond, releasing the phosphate. The substrate preference order demonstrates that all phosphates are sequentially cleaved with the eventual release of the nucleoside.
FIGURE 6. R1-NB is a nucleotide phosphatase.
A, 31P-NMR PO42− peak obtained from an assay in which R1-NB was provided with ADP as substrate in the presence of H2O16 (black line) or 60:40 H2O16: H2O18 (red line). B, 5.2 μm R1-NB was assayed in 100 (more ...)
We chose to establish that the nucleotide phosphatase activity of R1-NB was not an artifact of the truncated nature of the protein. Previous work with refolded preparations of Mi-1 and I-2 R-proteins utilized constructs that encompassed the entire NB-ARC domain (8
). Residues within the tandem ARC domains may confer an additional specificity on catalytic activity and nucleotide binding. Unfortunately we were unable to recover a soluble recombinant protein corresponding to the complete NB-ARC domain for Os02g_25900 and therefore turned our attention to other R-proteins. We specifically investigated the expression of the NB-ARC domains of PSiP and the well-characterized R-protein Rpm1 (hereafter called PSiP-NBARC and Rpm1-NBARC, respectively). Both Rpm1 and PSiP are derived from the CNL clade of NB-ARC family proteins and enable us to investigate whether the nucleotide phosphatase activity of R1-NB is present in other R-proteins. PSiP was an attractive experimental target for protein expression as this would enable us to validate whether an AC activity is associated with this ORF as hypothesized. PSiP-NBARC expression was comparable to R1-NB and it could be purified to homogeneity using similar methodology (A
). Rpm1 was an attractive experimental target for protein expression as this has an established role in the innate immune response. Rpm1-NBARC was expressed at extremely low levels and was often lost on purification but a single pure preparation of protein was obtained on which multiple experiments could be performed (B
). Analysis of PSiP-NBARC nucleotide utilization by HPLC showed the generation of the nucleoside with an identical substrate preference to R1-NB of AMP>ADP>ATP (C
) and GMP>GDP>GTP (D
). Specific activity using ADP as substrate was 6.0 ± 0.5 nmol Ado mg−1
protein; SD) equivalent to values obtained for R1-NB. PSiP-NBARC contains all residues and protein subdomains required for nucleotide binding/catalysis in P-loop ATPases based on bioinformatic and structural analysis (2
). PSiP is therefore not an AC. Using the limited Rpm1-NBARC material available we demonstrated that Rpm1-NBARC catalyzed the production of Ado from ATP and ADP demonstrating that the nucleotide phosphatase activity observed in R1-NB and PSiP-NBARC is also present in an established R-protein (E
). Specific activity using ADP as substrate was 38.5 ± 2.3 nmol Ado mg−1
protein; SD). Together these data demonstrate that the substrate specificity in the NB-ARC domain proteins is inherent to the NBD and that our data is unlikely to be due to an altered activity caused by protein truncation.
FIGURE 7. Purification and assay of recombinant NB-ARC domains (SDS/PAGE analysis and Coomassie Blue staining).
A, purification of recombinant PSiP-NBARC. A 1.5 μg portion of protein was applied and molecular mass standards (in kDa) are indicated. B, purification (more ...)