In flowering plants, RNA editing involves deamination of specific cytidines to uridines in both mitochondrial and chloroplast transcripts. Pentatricopeptide repeat (PPR) proteins and multiple organellar RNA editing factor (MORF) proteins have been shown to be involved in RNA editing but none have been shown to possess cytidine deaminase activity.
The DYW domain of some PPR proteins contains a highly conserved signature resembling the zinc-binding active site motif of known nucleotide deaminases. We modified these highly conserved amino acids in the DYW motif of DYW1, an editing factor required for editing of the ndhD-1 site in Arabidopsis chloroplasts.
We demonstrate that several amino acids of this signature motif are required for RNA editing in vivo and for zinc binding in vitro.
We conclude that the DYW domain of DYW1 has features in common with cytidine deaminases, reinforcing the hypothesis that this domain forms part of the active enzyme that carries out RNA editing in plants.
RNA editing is a post-transcriptional process that modifies the genetic information contained in the RNA (reviewed by Takenaka et al., 2013a). In plants, this mostly consists of the deamination of specific cytidines (C) to uridines (U) in mitochondria and plastids. In chloroplasts, 43 editing sites have been found in Arabidopsis thaliana transcripts (Ruwe et al., 2013), whereas 619 editing sites have been found in mitochondrial transcripts (Bentolila et al., 2013).
The enzyme involved in RNA editing is still unknown in plants, but some proteins have been shown to be specificity factors for RNA editing. These specificity factors are all from the PLS class of the pentatricopeptide repeat (PPR) family, containing c.200 members in Arabidopsis (Lurin et al., 2004). The PLS class is so named because of the characteristic triplets of P-type, L-type and S-type variants of the typical PPR motif that make up the bulk of these proteins (Lurin et al., 2004). These PPR arrays specifically recognize the target RNAs by binding upstream of the editing site with the C-terminal PPR motif aligned with the nucleotide at −4 with respect to the C to be edited (Barkan et al., 2012; Takenaka et al., 2013b; Yagi et al., 2013). The PLS class editing factors possess a C-terminal extension generally divided into two c.100 amino acid domains, the E domain and the DYW domain (Lurin et al., 2004), the latter named after its characteristic terminal tripeptide, Asp-Tyr-Trp (Aubourg et al., 2000). Many PLS class proteins have truncated C-termini, lacking the DYW domain and sometimes part of the E domain as well (Chateigner-Boutin et al., 2013). They are nevertheless functional in vivo, probably via recruitment of the missing domains from other proteins (Boussardon et al., 2012). The best-studied example is the pairing of CHLORORESPIRATORY REDUCTION4 (CRR4) (Kotera et al., 2005) and DYW1 (Boussardon et al., 2012), both of which are required for the editing of the ndhD-1 editing site in Arabidopsis chloroplasts. A synthetic fusion between the two proteins is fully capable of complementing the double mutant crr4/dyw1(Boussardon et al., 2012).
The DYW domain of DYW1 and the other c. 90 DYW proteins in Arabidopsis contain a motif similar to the cytidine deaminase (CDA) signature HxE(x)nPCxxC (only the proline is not conserved in the DYW domain sequences) and its predicted secondary structure resembles that of known CDAs (Salone et al., 2007; Iyer et al., 2011). Despite the apparent structural and sequence similarity to CDAs and close phylogenetic correlation between the DYW domain and C-to-U RNA editing (Salone et al., 2007; Rüdinger et al., 2011, 2012), the enzymatic involvement of the DYW domain in RNA editing has remained unclear because attempts to detect any CDA activity in vitro using recombinant DYW proteins have failed (Nakamura & Sugita, 2008; Okuda et al., 2009).
In the nucleotide deaminases that have been studied, the signature residues bind a zincion which is essential for catalysis (MacGinnitie et al., 1995). Deamination proceeds via deprotonation of zinc-bound water by the signature glutamate. This is followed by a nucleophilic attack of the cytidine C4 by the zinc-coordinated hydroxide, leading to irreversible hydrolysis and release of the C4-bound NH2 group as ammonia (Yao et al., 2005). Alterations of these conserved amino acids lead to the abolition of CDA function. The replacement of a cysteine (C97) by an alanine in the CDA signature of human APOBEC3G leads to the abolition of the multimerization of this cytidine deaminase (Opi et al., 2006). Substitution of a histidine (H238) with an alanine in mouse adenosine deaminase disabled the catalytic function (Sideraki et al., 1996). When the glutamic acid (E104) is switched to an alanine in Escherichia coli CDA, it destabilizes the enzyme-substrate complex (Carlow et al., 1996). We therefore set out to test the implication of the CDA-like signature in DYW proteins by studying mutants affected in various parts of the DYW domain including the signature motif. We show that point mutations in the DYW domain of DYW1 lead to decreased (or complete loss of) RNA editing at the ndhD-1 site. We also studied different mutant versions of DYW1 by inductively coupled plasma mass spectrometry(ICP-MS) to examine metal ion content.
Materials and Methods
Three dimensional (3D) structure prediction
A model of the DYW-domain of DYW1 was generated using the Robetta server. The 2.2 Å structure of cytidine deaminase from Vibrio cholera was detected as a suitable template (PDB entry 4EG2) despite limited sequence identity, and a model was generated by the RosettaCM protocol (Song et al., 2013). Models were inspected, and molecular graphics figures generated with PyMOL (The PyMOL Molecular Graphics System, Version 22.214.171.124; Schrödinger LLC, Portland, OR, USA).
Plant material, growth conditions and complementation analysis
Arabidopsis ecotype Columbia (Col-0) was used in this study. Seeds were surface sterilized, vernalized at 4°C for 3 d and grown on ½ Murashige and Skoog media containing 3% sucrose in vitro. Plates were placed in growth chambers under 16 h : 8 h, light : dark, at 25°C and with 45% humidity. Two-week-old seedlings were transferred onto soil and grown under 16 h : 8 h, light:dark at 21°C and 65% humidity. Isolation of the EMS mutants identified by TILLING is described in Boussardon et al. (2012). For complementation analysis, the complete DYW1 locus with its native promoter and terminator was amplified by PCR using DYW1_compF and DYW1_compR primers on genomic Arabidopsis Col-0 DNA, cloned into the pDNR207 vector by Gateway BP reaction (Invitrogen) and subcloned into pGWB1 vector (Nakagawa et al., 2007) by LR reaction. The complete list of oligonucleotides used in this study is summarized in Supporting Information Table S1.
To introduce point mutations into DYW1, two complementary internal primers (MxF and MxR) were generated containing the desired mutation. Two PCR fragments were generated using DYW1_compF and MxR and DYW1_compR and MxF. A secondround of PCR using DYW1_compF and DYW1_compR primers and the products of the first round reactions as template was carried out to produce a full-length DYW1 PCR product containing the desired mutation. Mutant DYW1 sequences were cloned into pDNR207 vector by Gateway BP reaction and subcloned into pGWB1 vector by LR reaction. For M7, the construct required only a single round of PCR with the reverse primer M7R and DYW1_compF.
Analysis of RNA editing
RNA from leaves of 18-d-old plantlets was extracted with the ‘RNeasy plant minikit’ from Qiagen. RNA was treated twice with DNase I (2 U μl−1; Ambion) 30 min at 37°C and cDNA was synthesized using Superscript II (Invitrogen). Reverse transcription polymerase chain reaction (RT-PCR) products were obtained with NdhD_AT_For and NdhD_AT_rev primers surrounding the ndhD-1 editing site (117166) and used as template for sequencing using the NdhD_AT_For primer.
Poisoned primer extension of RT-PCR products was performed as described by (Chateigner-Boutin & Small, 2007). RT-PCR products were obtained with NdhD_AT_For and NdhD_AT_rev primers and served as templates for the extension reaction from 5′FAM-labelled ndhD_PPE_C primer (SIGMA Genosys, Sydney, NSW, Australia) that anneals next to the editing site. The extension was stopped by the incorporation of ddCTP at the location of the editing site for unedited molecules producing a short unedited product. The extension was stopped at the next G/C for the edited molecules producing a longer edited product.
Protein expression and ICP-MS analysis
DYW1 coding sequences were amplified from Arabidopsis thaliana gDNA as follows. The 673 bp DYW1_WT form was amplified using primers DYW1_full_attb_int_TEV-F and DYW1_attb_STOP-R. To obtain DYW1ΔCDA, the first part of DYW1 was amplified using primers DYW1_full_attb_int_TEV-F and DYW1_del_CDA-R and the secondpart was amplified using primers DYW1_del_CDA-F and DYW1_attb_STOP-R. The two PCR products were then mixed and amplified using DYW1_full_attb_int_TEV-F and DYW1_attb_STOP-R to obtain a product of 577 bp. Similarly, to obtainDYW1_HE, products obtained with primers DYW1_attb_int-F plus DYW1_CDA_HE_mut-R and DYW1_CDA_HE_mut-F plus DYW1_attb_STOP-R were mixed and amplified using DYW1_full_attb_int_TEV-F and DYW1_attb_STOP-R to obtain a product of 673 bp. NP_416648 was amplified from Escherichia coli DNA using primers NP_416648_attb_TEV-F and NP_416648_attb-R. These fragments were cloned into pDONR207 using BP clonase II (Invitrogen) according to the manufacturer's instructions. The inserts were then sequenced and cloned into the pDEST15 expression vector using LR clonase II (Invitrogen). The recombinant proteins, now fused to a GST tag, were expressed in E. coli Rosetta 2 cells. Bacterial cultures (1 l) containing the expression vector were grown at 37°C until OD600 was 0.5. Cultures were then transferred to 16°C and incubated for 30 min. Protein expression was induced with IPTG (Sigma) to a final concentration of 0.4 mM. The cultures were incubated at 16°C for 16 h and harvested. The cell pellets were dissolved in 40 ml cold lysis buffer (×1PBS, pH 7.4) with addition of a protease inhibitor tablet (Complete Ultra; Roche). Cells were lysed by five rounds of homogenization (Emulsiflex-C5; Avestin, Ottawa, Ontario, Canada) and the supernatant was recovered after centrifugation for 15 min, 13 000 g at 4°C. Expressed proteins were purified using Glutathione Sepharose 4B (GE Healthcare, Little Chalfont, Buckinghamshire, UK) according to the manufacturer's instructions. Eluted proteins were concentrated with a centrifugal concentrator (Vivaspin 20; GE Healthcare) to 3 ml and dialysed in ×0.5 PBS, pH 7.4. Protein concentrations were determined with Coomassie Plus (Thermo Scientific, Waltham, MA, USA). BSA was used as the standard.
Dialysed products and the dissolution buffer were analysed by TSW Analytical Pty Ltd (University of Western Australia, Crawley, WA, Australia) using ICP-MS analysis.
Alignment and structural prediction of the DYW domain
Thousand DYW domain sequences from different plants were recovered from Genbank using the Arabidopsis DYW1 sequence for a blastp search (Altschul et al., 1997). A sequence logo representation of the conservation of this domain across this data set is shown in Fig. 1. The CDA signature is invariant and the final DYW triplet is extremely well conserved, particularly the terminal tryptophan. The tertiary structure of DYW1 was modelled based on the structure of cytidine deaminase from Vibrio cholera (Fig. 2a). The model revealed an arrangement of the conserved His, Glu and Cys residues that could plausibly form a metal-binding site corresponding to the active site of CDA (Fig. 2b). The model aligns the terminal DYW sequence with an EYL motif that forms part of the dimer interface in the CDA structure, suggesting a possible role for the DYW motif in homotypic or heterotypic protein–protein interactions.
DYW1 binds zinc
It has been proposed that the CDA signature HxE(x)nCxxC is involved in metal ion binding (Hayes et al., 2013). In order to test this hypothesis, different versions of DYW1 with various mutations were expressed as GST fusions in Escherichia coli. The cytidine deaminase from E. coli NP_416648 was cloned and expressed as a positive control for metal ion binding.
Proteins were purified on a glutathione column, eluted with glutathione buffer and dialysed into ×0.5 PBS, pH 7.4. The samples were analysed by ICP-MS to measure elemental composition–specifically zinc, cobalt and nickel. We first analysed the PBS buffer we used for dialysis to check for contaminating metal ions. The measured concentrations were considered background levels: 16 ppb (parts per billion) of zinc, 0.091 ppb of cobalt and 2.09 ppb of nickel (Fig. 3). The wild type DYW1 sample contained 289 ppb of zinc at a molar ratio of c. 0.7 zinc atoms per protein molecule (Fig. 3). The NP_416648 sample contained 447 ppb zinc at a molar ratio of c.0.8 zinc atoms per protein molecule (Fig. 3). Given the close similarity of the two molar ratios and the high zinc content compared to the buffer alone, we conclude that both NP_416648 and DYW1 contain a bound zinc ion. We found no evidence for significant binding of cobalt or nickel, suggesting that the zinc binding is relatively specific.
Mutant versions of DYW lacking the cytidine deaminase signature due to deletion of residues H163 to C194 (DYW1ΔCDA) or modification to AxA(x)nCxxC (DYW1-HE) contained much lower quantities of zinc (Fig. 3). We conclude that the CDA signature is necessary for proper zinc binding by the DYW domain.
The HxE(x)nCxxC CDA signature and the terminal tryptophan are necessary for RNA editing
We also tested the effects of mutations within DYW1 on RNA editing. Fifteen mutants resulting from the original TILLING screen that gave dyw1-1 (Boussardon et al., 2012) were analysed and seven targeted mutants constructed by introducing point mutations or by PCR-induced deletions (Table 1).
Table 1. Location of DYW1 mutations investigated in this study
TP, transit peptide; E, E domain, DYW, DYW domain.
For each mutant, the name of the EMS line (dyw1-x) or targeted PCR mutant (Mx) is given, together with the amino acid change and which domain of the protein is affected.
Of the random EMS-induced mutants, dyw1-2 carries a mutation in the putative organelle targeting sequence, dyw1-3 to dyw1-7 contain mutations N-terminal of the E and DYW domains, dyw1-8 to dyw1-10 have mutations in the short remnant E domain and dyw1-11 to dyw1-16 carry mutations in the DYW domain. Three backcrosses to WT Columbia (Col-0) plants were performed to remove unlinked background mutations. Plants that were homozygous for the mutations were used for analysis; all mutants exhibit a superficially WT phenotype, as do dyw1-1 and crr4-3 mutants (Kotera et al., 2005; Boussardon et al., 2012). At least three plants of each backcrossed line were analysed by RT-PCR and direct sequencing of the ndhD-1 editing site and one, previously analysed by sequencing, was analysed by poisoned primer extension.
Five of these mutants, dyw1-2, dyw1-4,dyw1-11,dyw1-12 anddyw1-16,showed a quantitative defect in ndhD-1 editing compared to WT Col-0 plants (Figs 4a, Supporting Information Fig. S1). In order to confirm these editing defects, we complemented these lines with full-length DYW1 and showed that this restored normal levels of editing (Fig. 4a; Supporting Information Table S2). Of the seven targeted mutants, M1-3 are mutated in the CDA signature sequence, M4-7 are mutated in the DYW triplet. These constructs were expressed under control of the native DYW1 promoter and used to complement the dyw1-1 mutant.
In mutant M1, the glutamate of the CDA signature has been altered to alanine (HSA). Six independent transformants were analysed and none showed any detectable editing of the ndhD-1 site (Fig. 4b). Similarly, in the M2 line, both histidine and glutamate (HSE) were altered to alanine (ASA). These mutations are identical to those that led to loss of zinc-binding (DYW1-HE in Fig. 3). RNA editing at the ndhD-1 site was not restored in any of six transformants tested (Fig. 4b). In the M3 line, the two cysteines of the CDA signature (CGDC) were modified to alanine (AGDA). Once more, none of six independent transformants showed detectable RNA editing of ndhD-1 (Fig. 4b). These results clearly show that the CDA signature is essential for DYW1 to edit ndhD-1.
The M5 construction (DYW to DAW) complements dyw1-1 perfectly showing that the tyrosine can be substituted by alanine without affecting editing of ndhD-1 (Fig. 4b). The M4 construct (DYW to AYW) only partially complemented ndhD-1 editing. The M6 (DYW to DYA) and M7 (ΔDYW triplet) constructions failed to complement at all (in six independent lines tested) (Fig. 4b, Supporting Information Table S1). Thus the C-terminal DYW triplet is important for RNA editing, although it can tolerate some alterations without loss of activity.
RNA editing in plant organelles has been shown to proceed by cytidine deamination (Yu & Schuster, 1995) and to be sensitive to zinc chelators in the case of chloroplasts (Hegeman et al., 2005), suggesting that the enzymatic reaction involves a classical cytidine deaminase, all of which contain a zinc ion at the active site (MacGinnitie et al., 1995). The active site motif consists of a HxE(x)nCxxC signature including the zinc-binding histidine and cysteine residues (MacGinnitie et al., 1995). The same signature is conserved in the DYW domain found in plant RNA editing factors (Salone et al., 2007; Iyer et al., 2011). Our hypothesis was that if the resemblance between the DYW domain signature and the CDA active site signature is not simply coincidental, then mutations in this signature sequence should eliminate both zinc binding and RNA editing.
As previously reported (Hayes et al., 2013), we confirmed that wild-type DYW1 binds zinc, although our results differ in that we found only a single bound zinc per protein molecule, as would be typically expected for a CDA. Mutation of the putative zinc-binding residues in the CDA-like signature greatly decreased zinc binding, suggesting that this signature sequence is the zinc-binding site. Furthermore, mutation of these same residues completely eliminated the ability of the DYW1 protein to act in RNA editing. This effect is relatively specific, as a number of other mutations that were tested had little or no effect on editing when tested in the same way. Taken together, these results show the DYW domain contains a zinc-binding site that is essential for RNA editing, in accordance with the hypothesis that this domain is (or is part of) an RNA-dependent cytidine deaminase. We cannot eliminate the possibility, though, that loss of activity is due to loss of structural integrity rather than a direct effect of the mutations on catalysis.
In addition to the CDA-like signature, DYW domains contain other highly conserved motifs of unknown function. The DYW terminal triplet is one of the most interesting of these motifs because highly conserved C-termini are unusual and generally associated with very specific functions. Complementation of dyw1-1 plants with DYW1 mutated in its DYW triplet showed that only replacement of the final tryptophan with alanine completely abolished editing. Replacement of the aspartate with alanine (M4) or glutamate (EMS line dyw1-16) reduced but did not abolish editing and the tyrosine could be replaced by an alanine (M5) without affecting editing at all. These results are consistent with the natural variation observed at these three positions, where in addition to the canonical DYW, other triplets are occasionally found, notably EYW or DFW. Several strictly C-terminal protein–protein interaction signals are known (e.g. the familiar K/HDEL ER retention signal and the SKL peroxisome targeting signal (Chung et al., 2002)) and the DYW terminus has very similar characteristics (charged residue followed by hydrophobic residues) that suggest it may be a docking signal bound by an as yet unidentified protein.
The authors thank Swaminathan Iyer (School of Chemistry and Biochemistry, University of Western Australia (UWA), Australia) for valuable discussions. Research at UWA was supported by Australian Research Council Centre of Excellence grants CE0561495 and CE140100008. Collaboration between UWA and URGV was supported by Australian Government International Science Linkages grants FR060030 and CG120098. The authors thank Sven Fjastad from TSW Analytical Pty Ltd for ICP-MS experiments.