The tripeptide glutathione (GSH) is implicated in various crucial physiological processes including redox buffering and protection against heavy metal toxicity. GSH is abundant in plants, with reported intracellular concentrations typically in the 1–10 mM range. Various aminotransferases can inadvertently transaminate the amino group of the γ-glutamyl moiety of GSH to produce deaminated glutathione (dGSH), a metabolite damage product. It was recently reported that an amidase known as Nit1 participates in dGSH breakdown in mammals and yeast. Plants have a hitherto uncharacterized homolog of the Nit1 amidase. We show that recombinant Arabidopsis Nit1 (At4g08790) has high and specific amidase activity towards dGSH. Ablating the Arabidopsis Nit1 gene causes a massive accumulation of dGSH and other marked changes to the metabolome. All plant Nit1 sequences examined had predicted plastidial targeting peptides with a potential second start codon whose use would eliminate the targeting peptide. In vitro transcription/translation assays show that both potential translation start codons in Arabidopsis Nit1 were used and confocal microscopy of Nit1–GFP fusions in plant cells confirmed both cytoplasmic and plastidial localization. Furthermore, we show that Arabidopsis enzymes present in leaf extracts convert GSH to dGSH at a rate of 2.8 pmol min−1 mg−1 in the presence of glyoxalate as an amino acceptor. Our data demonstrate that plants have a dGSH repair system that is directed to at least two cellular compartments via the use of alternative translation start sites.
Enzymes are usually not totally specific towards one substrate but can also act on other similar molecules [1,2]. This process—an enzyme acting on a metabolite other than its ‘true’ substrate—is known as enzymatic error, and it can result in the formation of a damaged metabolite . Damaged metabolites are often toxic [4,5] or useless . It is becoming apparent that all cells are equipped with dedicated metabolite repair systems to deal with the inevitable formation of damaged metabolites via enzymatic errors [7,8].
A metabolite is expected to be particularly prone to enzymatic error if: (i) the metabolite is abundant; (ii) the metabolite is similar to other metabolites; and (iii) the enzyme(s) that can inadvertently act on the metabolite are highly abundant. Glutathione (GSH), which is involved in multiple metabolic processes including redox buffering [9–11] and detoxification of heavy metals [12,13] and toxic compounds [14,15], fits all of these criteria. GSH is abundant in various plant tissues at reported concentrations of 2–15 mM [16–19]. Because GSH is a tripeptide (γ-glutamylcysteinylglycine), it has amino acid and peptide characteristics and is a potential alternate substrate for numerous enzymes that act on amino acids and proteins. Its γ-glutamyl moiety makes GSH a structural analog of glutamate, which is a common amino donor for aminotransferases . In fact, GSH has been shown to be an alternate substrate for a few mammalian aminotransferases in vitro , resulting in the deamination of the amino group of the γ-glutamyl moiety of GSH to form deaminated glutathione (dGSH) (Figure 1). Ophthalmic acid (OA), a close analog of GSH in which the central cysteine is replaced with 2-aminobutyrate , also undergoes transamination to form deaminated ophthalmic acid (dOA) (Figure 1). Both dGSH and dOA can exist in a linear form or as cyclized ketoacids that occur as two distinct anomers (Figure 1).
Glutathione and ophthalmic acid are damaged by transaminases and repaired by Nit1.
It was recently reported that the Nit1 enzyme of yeast and mammals is an amidase that efficiently hydrolyzes the linear forms of dGSH and dOA . It is not known how plants deal with dGSH, but they have a homolog of Nit1 with a predicted plastid targeting peptide. Because GSH is involved in several areas of metabolism and has the potential to be crucially damaged in plants, we investigated whether the Arabidopsis Nit1 homolog functions in dGSH repair. We first characterized the enzyme biochemically and determined its subcellular location. We then obtained evidence that the erroneous deamination of GSH by plant transaminases occurs under physiological conditions. Finally, we investigated the effect of ablating the Arabidopsis gene encoding Nit1 on the levels of dGSH and other metabolites.
Materials and methods
Protein sequences were from GenBank, MaizeGDB.org, and UniProt.org, and were verified using expressed sequence tags from the GenBank dbEST database. Phylogenetic analysis was performed with MEGA7  using the neighbor-joining method. Alignments were made with Multalin . Subcellular localization was predicted using the algorithms given in Supplementary Table S1.
Phusion® High-Fidelity DNA polymerase (New England Biolabs, Ipswich, MA, U.S.A.) was used for PCRs. All other chemicals and reagents were obtained from Sigma–Aldrich (St. Louis, MO, U.S.A.), unless otherwise stated. dGSH, dOA, and αKGM were synthesized and purified as described previously [21,25]. Briefly, 5 ml solutions containing 10 units of catalase from bovine liver, 10 mg of l-amino acid oxidase from Crotalus adamanteus, and 200 mM of either GSH, ophthalmic acid, or glutamine, pH 7.5 were incubated for 16 h at 37°C. Afterwards, the dGSH preparation was supplemented with 120 mM DTT to reduce disulfides. Solutions were then deproteinized with Amicon Ultra-4 10 000 NMWL centrifugal filters (Millipore), HCl was added to 0.1 mM, and the solutions were applied to a 5 ml AG50 cation-exchange column equilibrated with 0.1 M HCl. The eluate was neutralized with NaOH, lyophilized, and the powder was stored at −20°C until use. dGSH and dOA were quantified enzymatically with purified recombinant Arabidopsis Nit1 protein.
cDNAs and expression constructs
All primers used are listed in Supplementary Table S3. All constructs were sequence-verified. For transient expression of proteins fused C-terminally to GFP, the full-length open reading frame of the Nit1 gene was PCR amplified (primers 1 and 2) from cDNA prepared from Columbia-0 plants using the SuperScript™ III First-Strand Synthesis System (Invitrogen, Carlsbad, CA, U.S.A.) according to the manufacturer's recommendations. Amplicons were digested with BamHI and NheI and ligated into pUC18-mGFP-MCS  to give Nit1-FL-pUC18-mGFP-MCS. The M2L construct was created using the QuikChange II Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA, U.S.A.) according to the manufacturer's recommendations with primers 3 and 4 and Nit1-FL-pUC18-mGFP-MCS as a template to give Nit1-M2L-pUC18-mGFP-MCS. The Tr construct was created by PCR-amplifying the Tr open reading frame with primers 5 and 2, digesting with BamHI and NheI, and ligating into pUC18-mGFP-MCS to give Nit1-Tr-pUC18-mGFP-MCS.
For in vitro transcription/translation assays, the FL (primers 6 and 7), M2L (primers 6 and 7), and Tr (primers 8 and 7) coding sequences of the Arabidopsis Nit1 protein were PCR-amplified, digested with EcoRI and XbaI, and cloned into pGEM4Z.
For heterologous expression in Escherichia coli, the Tr coding sequence was PCR-amplified with either primers 9 and 10 or 11 and 12, digested with NdeI and XhoI or NcoI and XhoI, and ligated into pET28b to facilitate the expression of the predicted mature Nit1 protein with either an N-terminal or C-terminal His-tag, respectively.
Protein production and isolation
To produce recombinant proteins, E. coli strain BL21 (DE3) RIPL harboring the respective expression plasmid was grown in 200 ml of LB medium plus 50 mg l−1 kanamycin at 37°C until OD at 600 nm reached 0.9. Cultures were then supplemented with 0.5 mM isopropyl-β-D-thiogalactoside and incubated for a further 16 h at 22°C. Cells were collected by centrifugation and stored at −80°C. Pellets were resuspended in 8 ml of ice-cold lysis buffer (50 mM potassium phosphate, pH 8.0, 300 mM NaCl, and 10 mM imidazole) and sonicated (Fisher Scientific Ultrasonic Dismembrator, model 150E) using seven 12-s pulses at 70% power, cooling on ice for at least 1 min between pulses. The resulting lysates were centrifuged (16 000g, 10 min), and the supernatant was applied to 0.5 ml Ni-NTA superflow resin (Qiagen) columns and proteins were purified using the manufacturer's protocol. Purified proteins were passed through PD-10 columns (GE Healthcare) equilibrated with 100 mM KCl, 25 mM Tris–HCl, pH 8.0, and 10% (v/v) glycerol. Aliquots (25 µl) were snap-frozen in liquid nitrogen and stored at −80°C. Protein was determined by dye-binding . Maize ω-amidase (Nit2) was purified as previously described .
To create Arabidopsis cell-free protein extracts, 5 g of leaves from 3-week old soil-grown plants were ground for 20 min on ice in 6 ml of buffer containing 50 mM Tris–Cl, pH 8.2, 10% glycerol (v/v), 3 mM DTT, 1.25 mM EDTA, 2.5 mM MgCl2, 6 mM CHAPS, 1 mM PMSF, and 25 µM PLP. The resulting lysates were centrifuged (16 000g, 10 min), and the supernatant was applied to PD-10 columns (GE Healthcare) equilibrated with 100 mM KCl, 50 mM HEPES, pH 8.0, 0.5 mg l−1 BSA, and 2 mM DTT.
Assays of recombinant proteins (100 µl, 22°C) contained 50 mM HEPES–KOH, pH 8.5, 50 mM KCl, 0.5 mg ml-1 BSA, 2 mM DTT, 0.2 mM NADH, 5 mM ammonium acetate, 5 units of glutamate dehydrogenase, 5–200 µM dGSH or dOA, or 10 mM αKGM, and were started by adding 0.12 µg (dGSH), 0.06 µg (dOA), or 1.2 µg (αKGM) of Nit1 or 1.0 µg of ω-amidase (Nit2). Absorbance at 340 nm was recorded every 15 s for up to 10 min, and the rate of α-ketoglutarate formation was calculated. To determine enzyme kinetics, at least three independent data sets consisting of assays at six different substrate concentrations were performed. Kinetic constants were determined by fitting the experimental data to the Michaelis–Menten equation using Prism software (GraphPad Software, Inc., La Jolla, CA, U.S.A.).
Assays of cell-free protein extracts (100 µl, 30°C) contained 50 mM HEPES–KOH, pH 8.5, 50 mM KCl, 0.5 mg ml−1bovine serum albumin, 2 mM DTT, 10 mM GSH, 3 mM glyoxylate, and 50 µM PLP, and were started by adding 1.4–1.6 mg of total protein. After an 18 h incubation, samples were deproteinized using an Amicon Ultra-0.5 10K ultrafiltration column (21 000g, 10 min), and the amount of dGSH present in the flow-through was determined enzymatically with recombinant Nit1 as described above.
Coupled in vitro transcription–translation assays
In vitro transcription–translation was performed using a TNT® Coupled Wheat Germ Extract System (Promega, Madison, WI, U.S.A.), according to the manufacturer's directions with [3H]Leu (107.7 Ci/mmol, PerkinElmer) as label.
Expression and visualization of GFP fusion proteins in BY-2 cells
Expression and confocal laser-scanning microscopy (CLSM) analysis of Nit1–GFP fusion proteins in tobacco (Nicotiana tabacum) BY-2 suspension-cultured cells was performed as previously described , except that cells were co-bombarded with BCAT3-Cherry , which served as a marker for plastids in BY-2 cells. Confocal microscopy of co-transformed BY-2 cells was performed using a Leica DM RBE microscope (Leica Microsystems, Inc., Richmond Hill, Canada) as previously described . Figures were composed using Adobe Photoshop CS (Adobe Systems). All fluorescence images of cells shown in the figures are representative of >50 independent (transient) co-transformed cells from at least three independent transformation experiments.
Growth and validation of the Arabidopsis nit1 knockout line
An Arabidopsis line with a T-DNA insertion in the sixth exon of the Nit1 gene (At4g08790) was identified in the Salk collection (SALK_108849C). Seeds from this line and the Columbia-0 wild-type parent were germinated on ½ MS medium. Genomic DNA was isolated from young leaves as previously described . Standard PCRs were performed with wild-type allele primers (13 and 14) or T-DNA insertion allele primers (13 and 15), and amplification products were analyzed by agarose-gel electrophoresis. All SALK_108849C plants analyzed were homozygous for the T-DNA insertion allele. The amplicon obtained with the T-DNA insertion allele primers was cloned into pGEM-T easy (Promega) and sequenced to verify the T-DNA insertion site. Total RNA was isolated from homozygous plants using the RNeasy® Plant Mini Kit (Qiagen) with on-column DNase treatment. First-strand cDNA was created with the SuperScript™ III First-Strand Synthesis System (Invitrogen). PCRs were performed with primers (16 and 17) to amplify the full-length Nit1 coding sequence or a fragment of the actin-7 transcript (18 and 19). Amplicons were analyzed by agarose-gel electrophoresis.
Seeds from wild-type and nit1 knockout Arabidopsis plants grown under identical conditions were used for growth experiments and to prepare samples for enzymatic or metabolic analysis. For growth experiments and metabolic analysis, seeds were surface-sterilized, plated on ½ MS salts with 1% sucrose, vernalized for 3–5 days at 4°C, and placed under fluorescent lights (photosynthetic photon flux ∼150 µmol photons m−2 s−1) on a 12 : 12 h light/dark cycle at 21–23°C. Whole seedlings were harvested after 14–28 days of growth. For enzymatic or metabolic analysis of soil-grown plants, seeds were germinated in soil and grown for 21–28 days prior to harvesting aerial parts of the plants. For metabolic analysis, samples were immediately frozen in liquid nitrogen, powdered, lyophilized, and stored at −80°C until analysis. For enzymatic analysis, harvested material was immediately used to create protein lysates.
3D modeling of Arabidopsis Nit1
Homology modeling was conducted using the Rosetta Comparative Modeling protocol . Fragment 3-mer and 9-mer libraries were generated from the Arabidopsis Nit1 sequence using the Robetta server . Saccharomyces cerevisiae Nit1 (scNit1) structures containing α-ketoglutarate and oxaloacetate (Protein Data Bank ID: 4hg5, 4hg3) were used as initial models based on their 37% sequence identity with Arabidopsis Nit1. scNit1 and Arabidopsis Nit1 protein sequences were aligned using Clustal Omega  prior to threading the Arabidopsis Nit1 sequence onto the scNit1 structures. In total, 500 models were generated and, following visual inspection, the model with the lowest total score was selected for ligand docking. dGSH was manually placed into the active site of the Arbidopsis Nit1 structure using PyMol. RosettaLigand was then utilized to minimize the protein : ligand complex. The top 100 models with the lowest overall score were then pooled and the best model was selected as the model with the lowest interface score.
Targeted analysis of dGSH and dOA
Powdered and lyophilized Arabidopsis leaf samples were stored at −80°C until processed. Sample preparation was carried out as described previously  at Metabolon, Inc. Briefly, samples were extracted with methanol under vigorous shaking for 2 min followed by centrifugation. The resulting extract was divided into five fractions, dried, and then reconstituted in solvents compatible to each of four methods and a backup. Each reconstitution solvent contained a series of standards at fixed concentrations to ensure injection and chromatographic consistency. One aliquot was analyzed using acidic positive ion conditions and chromatographically optimized for more hydrophilic compounds. In this method, the extract was gradient eluted from a C18 column (Waters UPLC BEH C18-2.1 × 100 mm, 1.7 µm) using water and methanol, containing 0.05% perfluoropentanoic acid (PFPA) and 0.1% formic acid (FA). Another aliquot was also analyzed using acidic positive ion conditions; however, it was chromatographically optimized for more hydrophobic compounds. In this method, the extract was gradient eluted from the same aforementioned C18 column using methanol, acetonitrile, water, 0.05% PFPA, and 0.01% FA, and was operated at an overall higher organic content. Another aliquot was analyzed using basic negative ion optimized conditions using a separate dedicated C18 column. The basic extracts were gradient eluted from the column using methanol and water, however, with 6.5 mM ammonium bicarbonate at pH 8. The fourth aliquot was analyzed via negative ionization following elution from a HILIC column (Waters UPLC BEH Amide 2.1 × 150 mm, 1.7 µm) using a gradient consisting of water and acetonitrile with 10 mM ammonium formate, pH 10.8. All methods utilized a Waters ACQUITY UPLC and a Thermo Scientific Q-Exactive high-resolution/accurate mass spectrometer interfaced with a heated electrospray ionization (HESI-II) source and Orbitrap mass analyzer operated at 35 000 mass resolution. The MS analysis alternated between MS and data-dependent MSn scans using dynamic exclusion. The scan range varied slighted between methods but covered 70–1000 m/z.
Metabolites were identified by automated comparison of the ion features in the experimental samples to a reference library of chemical standard entries that included retention time, molecular mass (m/z), preferred adducts, and in-source fragments as well as associated MS spectra and curated by visual inspection for quality control using software developed at Metabolon [33,34]. Identification of known chemical entities is based on comparison to metabolomic library entries of purified standards. dGSH and dOA detectable characteristics were registered into LIMS. Peaks were quantified using area under the curve (raw data).
Metabolomic and lipidomic analysis
Samples (10 mg of lyophilized plant tissue) were extracted using a biphasic extraction technique as previously described . Briefly, 225 µl of LC–MS grade methanol was added to lyophilized plant tissue in 2 ml Eppendorf tube, vortexed for 10 s, followed by the addition of 750 µl methyl tert-butyl ether (MTBE). Each sample was then vortexed again for 10 s, shaken on orbital shaker at maximum speed for 6 min, followed by the addition of 188 µl LC–MS grade water. Finally, each sample was vortexed again for 10 s and centrifuged for 2 min at 14 000g. The resultant two-phase extract was aliquoted into four clean 1.5 ml Eppendorf tubes. The result was two 350 µl aliquots of MTBE phase (top), and two 110 µl aliquots of methanol/water phase (bottom). Extraction was carried out at 4°C. All extract tubes were dried under vacuum and frozen at −80°C until LC–MS/MS analysis.
Lipidomics analysis followed methods of Cajka et al. . Briefly, extract from one aliquot of MTBE phase was resuspended in 110 µl methanol/toluene (9 : 1, v/v), vortexed for 10 s, centrifuged for 2 min at 14 000 rpm, transferred to HPLC vial, and stored at 4°C until LC–MS/MS analysis. Gradient, internal standards, mobile phases, and data collection methods were described recently . Sample injection volume was 3 µl on a Waters Acquity UPLC CSH C18 column (100 mm × 2.1 mm, 1.7 µm particle size). A Thermo Vanquish Focused UHPLC system coupled to a Thermo Q Exactive HF mass spectrometer was used for all untargeted analysis.
Untargeted polar metabolomic analysis followed methods of Niehaus et al. (2018). Briefly, one aliquot of methanol/water phase extract was resuspended in 100 µl of acetonitrile/water (4 : 1 v/v) with internal standards including 5 µg/ml Val-Tyr-Val. Analysis was carried out using a Waters Acquity UPLC BEH Amide (150 mm × 2.1 mm id, 1.7 µm particle size) column with identical mobile phase, gradient, injection volume, and data collection as recently described .
Data processing was carried out using open source software MS-DIAL  version 2.90 for peak deconvolution, peak picking, alignment, accurate mass/retention time matching, and MS/MS matching (further settings found in Supplementary Table S4). MS/MS spectra matches were manually inspected to confirm annotation. Adducts, isotopes, and duplicate peaks were removed using MSFLO . Evidence required for annotation of a compound was accurate mass match in addition to (a) experimental MS/MS spectrum match to library spectrum and/or (b) a match to in-house accurate mass/retention time library created with authentic standards under identical chromatography conditions. Mass Bank of North America (MoNA), NIST17, and LipidBlast  MS/MS libraries were queried against for this analysis. Peak height was reported, and all samples were normalized to mean total ion count for all known compounds of each analytical platform. Raw data can be found at www.metabolomicsworkbench.org (project doi:10.21228/M8N965).
Results and discussion
Sequence features of plant Nit1 homologs
BlastP searches of Arabidopsis or maize protein databases using validated  Nit1 sequences as the query identified a single Nit1 homolog in Arabidopsis (At4g08790) and maize (GRMZM2G117642) as the best hit in all cases. The second best hit in Arabidopsis (At5g12040) and second or third best hits in maize (GRMZM2G169365 and GRMZM2G156486) have been previously shown to be ω-amidase (i.e. Nit2) proteins, which hydrolyze α-ketoglutaramate (α-KGM) as their preferred substrate (Ellens et al. ). A phylogenetic analysis was performed with the plant Nit1 and Nit2 homologs, as well as validated homologs from Saccharomyces cerevisiae, Mus musculus, Pseudomonas aeruginosa, and Escherichia coli. The plant Nit1 homologs clustered in the same clade as Nit1 proteins from other kingdoms, whereas Nit2 proteins clustered in another clade (Supplementary Figure S1), further supporting the inference that Arabidopsis and maize have single Nit1 proteins.
Plant Nit1 proteins have poorly conserved N-terminal regions (Supplementary Figure S2) that are predicted to be plastidial targeting (transit) sequences (Supplementary Table S1). All plant Nit1 proteins analyzed have a downstream methionine residue that would remove the putative targeting sequence if used to initiate translation (Supplementary Figure S2). The N-terminal regions of plant Nit1 proteins thus have features suggesting possible localization to both plastids and cytoplasm.
Enzymatic activities of recombinant Arabidopsis Nit1 protein
Recombinant Arabidopsis Nit1 proteins were assayed for amidase activity towards dGSH and dOA. The coding sequence of the predicted mature protein was cloned into expression vectors, thus facilitating expression of mature peptides with either N- or C-terminal His-tags. Proteins were purified to near homogeneity (Supplementary Figure S3). In preliminary assays, N- and C-terminally His-tagged proteins had essentially the same activity; we chose C-terminally His-tagged Nit1 for subsequent experiments because it was judged to be slightly purer (Supplementary Figure S3).
A coupled assay with glutamate dehydrogenase that monitors the release of α-ketoglutarate was used to assay Nit1. Because dGSH and dOA exist largely in their cyclic forms  and Nit1 is expected to act only on the linear form (Figure 1), we first determined assay conditions in which opening of the cyclic forms is not rate-limiting . Nit1 was shown to have high amidase activity towards dGSH and was even more efficient at hydrolyzing dOA (Table 1). Because Nit1 is a close homolog of Nit2, which was shown to hydrolyze α-KGM , we assayed Nit1 for α-KGM amidase activity. Nit1 had weak activity towards αKGM that was estimated to be ∼250-fold less than the activity towards dGSH. To test whether plant Nit2 enzymes can hydrolyze dGSH, we assayed recombinant maize Nit2. As expected, Nit2 had very low activity against dGSH and no measurable activity against dOA (Supplementary Table S2).
|Substrate||Vmax (µmol min−1 mg−1)||kcat (s−1)||KM (µM)||kcat/KM (mM−1 s−1)|
|dGSH||7.9 ± 0.4||4.2 ± 0.2||35 ± 5||121|
|dOA||5.6 ± 0.3||3.0 ± 0.2||6.7 ± 0.7||445|
|α-KGM||10.021 ± 0.002||–||–||–|
|Substrate||Vmax (µmol min−1 mg−1)||kcat (s−1)||KM (µM)||kcat/KM (mM−1 s−1)|
|dGSH||7.9 ± 0.4||4.2 ± 0.2||35 ± 5||121|
|dOA||5.6 ± 0.3||3.0 ± 0.2||6.7 ± 0.7||445|
|α-KGM||10.021 ± 0.002||–||–||–|
Value represents activity at 10 mM substrate.
The kinetic constants of Nit1 with dGSH and dOA as a substrate are typical of metabolic enzymes towards their preferred substrates . Considering that the actual substrates of Nit1 are the linear forms of dGSH and dOA (Figure 1), which are much less abundant than their cyclized ketoacid forms, the Km of Nit1 towards the ‘actual’ substrates is much lower than we report. Thus, our kinetic data suggest that Nit1 should be very effective at hydrolyzing dGSH and dOA in vivo.
3D modeling of Arabidopsis Nit1 supports in vitro substrate specificities
Homology modeling reveals that Arabidopsis thaliana Nit1 (AtNit1) has a similar active site to yeast Nit1 (PDB: 4hg5) and the Caenorhabditis elegans enzyme (PDB: 1ems), as expected from the high sequence identity. Unlike the mouse Nit2 (PDB: 2w1v), AtNit1 is active towards dGSH and dOA. From a structure overlay of AtNit1 and MmNit2, it is evident that AtNit1 has a much larger binding pocket to accommodate the large αKGM derivatives, dGSH and dOA (Figure 2). This is a result of the β2α2 loop of MmNit1 placed further from the active site when compared with AtNit1. The preferential activity towards larger substrates may be explained by two key amino acid variations found in AtNit1 relative to MmNit2—Tyr49 in MmNit2 is Gly54 in the Arabidopsis enzyme and Tyr216 in MmNit2 is Arg224 in the plant homolog. These two positions are a critical determinant for whether the enzyme acts as Nit1 or Nit2 . Generally, Nit1s contains a small, uncharged amino acid at the position equivalent to Gly54 in AtNit1 (on the β2α2 loop), while Nit2s contain either tyrosine or phenylalanine . A small, uncharged amino acid at this site may allow the β2α2 loop to adopt a conformation that would otherwise not be possible with a large, aromatic amino acid and therefore create a larger binding pocket. At the site equivalent to Arg224 in AtNit1, Nit1s contain a basic amino acid (lysine or arginine), while Nit2s a tyrosine . Arg224 in the AtNit1 homology model is oriented toward the carbonyl group of the cysteinylglycine moiety of dGSH, forming a salt bridge interaction, while Tyr49 in MmNit2 is oriented toward the active site (Figure 2). A basic amino acid at this position stabilizes large α-KGM derivatives, while a tyrosine would occupy the binding pocket and therefore create a smaller binding site.
3D model of Arabidopsis Nit1 compared with mouse Nit2.
Subcellular localization of Arabidopsis Nit1
To begin to assess the subcellular localization of Arabidopsis Nit1, we first determined which of the potential start sites of the protein are functional in vitro using a coupled transcription–wheat germ translation system. The native full-length (FL) cDNA gave two translation products, while a modified cDNA with the second predicted start codon (at position 22) replaced with a codon for leucine (M2L) gave only the larger product, and cDNA truncated at the second start codon (Tr) gave only the smaller one (Figure 3). These results establish that Nit1 has two distinct alternative start sites, at least in vitro.
Arabidopsis Nit1 is dual targeted to the cytoplasm and plastids.
Next, we employed CLSM to analyze Arabidopsis Nit1 C-terminal green fluorescent protein (GFP) fusions expressed transiently in Bright Yellow-2 (BY-2) tobacco suspension-cultured cells, which are a well-known model system for studying plant protein localization in vivo [42,43]. The Nit1–GFP fusion constructs included the abovementioned wild-type full-length, N-terminal-truncated, and full-length M2L versions of Nit1. As shown in Figure 3B, the wild-type full-length version (Nit1-FL-GFP) localized to both plastids and the cytoplasm, as evidenced by its partial colocalization with the co-expressed plastid marker fusion protein, BCAT3-Cherry , and the diffuse fluorescence attributable to Nit1-FL-GFP throughout the rest of the cell (Figure 3). On the other hand, the Tr version (Nit1-Tr-GFP) localized exclusively in the cytoplasm, while the M2L version (Nit1-M2L-GFP) primarily localized to plastids (Figure 3). Taken together, these results are largely consistent with the in vitro data and indicate that, depending on the translation start site used, the Nit1 protein can localize to the cytoplasm or plastids. These results are also consistent with a previous study that analyzed the localization of the first 100 N-terminal residues of Arabidopsis Nit1 (which includes both potential methionine start sites) fused to GFP; the fusion protein was observed in plastids, however, the presence of a diffuse fluorescence pattern suggests cytoplasmic localization as well .
The use of alternate translation start sites to target an enzyme to multiple compartments has precedents among metabolite repair systems in plants. Plant enzymes involved in the repair of damaged NAD(P)H and α-KGM make use of organellar targeting peptides and alternate translation start sites to facilitate localization to multiple compartments [5,27]. There was no indication that Nit1 localized to mitochondria (Figure 3B) even though GSH  and aminotransferases  are known to occur there. It is possible that another dGSH repair system exists in mitochondria or that dGSH is exported to the cytoplasm. dGSH transport could be mediated by pore-regulating proteins, which are known to facilitate mitochondrial and nuclear GSH transport in animal cells  and may do the same in plants .
Impact of ablating Arabidopsis Nit1 on dGSH levels
To investigate the effect Nit1 has on in vivo dGSH levels, we identified an Arabidopsis line (SALK_108849C) with an insertion in the sixth exon. Plants were analyzed to verify the insertion location and to measure Nit1 transcript level (Supplementary Figure S4). No Nit1 transcript was detected in homozygous mutant plants, indicating that the T-DNA insertion eliminates Nit1 expression.
nit1 mutant plants and the Columbia-0 parental line were grown for 14 or 28 days prior to analyzing whole-plant samples by LC–MS. A peak corresponding to one anomeric form of dGSH was detected in both wild-type and mutant samples. This peak was significantly >70-fold and >45-fold larger in the mutant compared with wild type in 14- and 28-day-old plants, respectively (Figure 4). A peak corresponding to the other anomeric form of dGSH was readily detected in mutant plants but not in wild-type plants at both 14 and 28 days (Figure 4). Likewise, oxidized (i.e. disulfide) forms of dGSH were readily detected in mutant but not in wild-type plants (Figure 4). dOA was not detected in any samples, but OA levels were significantly lower in mutant plants at both time points (Figure 4), suggesting that Nit1 plays some role in OA metabolism. There were no significant differences in GSH or oxidized GSH (GSSG) levels between wild-type and mutant plants (Figure 4). We were unable to reliably quantify the absolute levels of dGSH by LC–MS because it occurs cyclized as two distinct anomers and several disulfides in unpredictable ratios; however, it is clear that dGSH massively accumulates in the Nit1 mutant (Figure 4). These results indicate that Nit1 is responsible for controlling dGSH levels in plants.
Relative dGSH, GSH, and OA levels in wild-type and nit1 mutant Arabidopsis plants.
Arabidopsis contains transaminases that catalyze dGSH formation in vitro
We predicted that dGSH is formed in plants by the inadvertent action of transaminases on GSH . To test this possibility, we prepared desalted protein extracts from leaves of wild-type or nit1 mutant plants and measured their ability to convert GSH to dGSH. Assays contained glyoxylate as the amino acceptor and were performed at physiological GSH concentration (10 mM). After incubation, dGSH formation was measured enzymatically with purified recombinant Arabidopsis Nit1 protein. dGSH formation was detected with nit1 mutant extracts but not with wild-type protein extracts (Table 2). When the amino acceptor was omitted from the assay, GSH deaminase activity in the nit1 mutant protein extract was undetectable (Table 2), indicating that aminotransferases in the protein extract are the catalysts. If our measured rate of GSH damage occurred in vivo, dGSH would accumulate in leaves to ∼60 µM in 1 day in the absence of a repair system (assuming a protein content of 15 mg g−1 fresh weight ). It is worth noting that the reason GSH damage could not be detected with wild-type protein extract is likely due to the presence of endogenous Nit1 enzyme—dGSH formation was countered by an efficient repair system. It was first necessary to ablate Nit1 in order to reliably estimate the scale of GSH damage.
|Plant source of protein extract||Activity (pmol min−1 mg−1)|
|Nit1 mutant||2.76 ± 0.10|
|Nit1 mutant (no glyoxylate)||1n.d.|
|Plant source of protein extract||Activity (pmol min−1 mg−1)|
|Nit1 mutant||2.76 ± 0.10|
|Nit1 mutant (no glyoxylate)||1n.d.|
Not detected, detection limit = 0.33.
Phenotypic impacts of ablating Arabidopsis Nit1
Untargeted metabolomics was performed to determine the metabolic effects of abolishing Nit1. Because metabolic changes can be dynamic and variable , we analyzed four independent sets of wild-type and nit1 mutant plants, which allowed us to identify metabolites whose levels are consistently changed across experiments. The four sample sets consisted of plants grown in soil or on defined medium for 3–4 weeks prior to harvesting. Using strict criteria (metabolites that showed significant changes in the same direction in at least three sample sets with >2-fold change or <0.5-fold change in at least one sample set), 16 metabolites were identified that were consistently affected in mutant plants (Figure 5). The levels of the GSH breakdown product cysteinylglycine and sulfur-related intermediate cystathionine were both increased in the mutant, indicating that GSH and/or sulfur metabolism is affected. There were differences in the levels of several amino acids and intermediates—particularly positively charged amino acids—between mutant and wild-type plants. Interestingly, levels of sinapine, an antinutrient , were dramatically lower in mutant plants. The levels of several phenylpropanoid derivatives were also significantly lower in mutant plants. In plants, glutathione-S-transferases (GSTs) are responsible for conjugating several phenylpropanoids . It is possible that dGSH is a competitive inhibitor of GSTs and that the reduced phenylpropanoid levels in mutant plants are flagging a GST effect.
Effect of Nit1 ablation on the Arabidopsis metabolome.
We also performed a lipidomics analysis on the four sample sets described above. Using the same strict criteria, 22 triglycerides were identified whose levels significantly decreased in nit1 mutant plants grown on defined medium (Figure 6). No other lipids were significantly affected. The levels of most affected triglycerides were at least 10-fold less in mutant plants compared with wild-type plants (Figure 6). This surprising result indicates that Nit1 plays a major role in determining triglyceride levels in Arabidopsis.
Effect of Nit1 ablation on Arabidopsis lipid levels.
When grown under our standard laboratory conditions, nit1 mutant plants showed no obvious phenotype. Because our data show that Nit1 controls dGSH levels, which may affect GSH metabolism, we grew plants under conditions reported to impair growth of other GSH mutants. These conditions include treatment with cadmium [12,13], paraquat , hydrogen peroxide , or formaldehyde [54–56], or limiting sulfur in the medium [57,58]. No significant difference in the growth of wild-type and nit1 mutant plants was observed under any of these treatments (Supplementary Figure S5).
Ablation of other plant metabolite damage repair systems also shows no obvious growth phenotype  or a phenotype that manifests only under specific growth conditions . Like in Arabidopsis, mouse nit1 mutants show no obvious growth phenotype despite having a dramatic increase in dGSH levels . Although we could not detect a negative growth phenotype, the fact that genes encoding Nit1 proteins are found almost ubiquitously in all domains of life suggests that this damage repair system is somehow beneficial. It seems likely that Nit1 confers some fitness advantage under yet-to-be defined growth conditions. It was previously shown that Arabidopsis Nit1 is one of a few dozen genes whose expression significantly increases in response to cadmium treatment , suggesting that it may play a role in heavy metal detoxification. Another study concluded that Arabidopsis Nit1 is involved in root growth and development .
Our results establish that At4g08790 encodes a dGSH repair enzyme that is directed to both plastids and the cytoplasm via the use of alternative translation start sites. We propose the name NIT1 (Nitrilase-like protein 1) for this gene, following the precedent in mammals and yeast . Because plant Nit1 proteins are highly similar in primary sequence (e.g. Arabidopsis and maize mature proteins share 72% identity and 94% similarity), they are likely orthologous and the Nit1 designation can be reliably propagated to all plant genes encoding this enzyme. Plant Nit1 proteins can now join the expanding list of validated, cross-kingdom metabolite repair systems .
Nit1 belongs to the same protein family as Nit2; both are amidases that perform analogous reactions on analogous substrates that differ mainly in size. Both amidases appear to have metabolite repair roles to clear-out the inevitable damage caused by transaminases acting on alternative substrates.
Arabidopsis nit1 mutant plants showed no obvious growth phenotype despite having highly elevated dGSH levels. However, we detected some specific metabolic disturbances, such as a decrease in the levels of several triglycerides. GSH peroxidases have been shown to inhibit lipid peroxidation caused by reactive oxygen species in Arabidopsis . If dGSH competitively inhibited GSH peroxidases, then elevated dGSH levels would lead to an increase in lipid peroxidation, which could explain the decrease in triglyceride levels observed in nit1 mutant plants. In principle, dGSH could affect all GSH-dependent enzymes and alter several areas of metabolism, including GSH homeostasis. dGSH may also serve other roles, such as signaling for elevated transaminase activity.
The mild phenotypes associated with metabolite damage control systems may partially explain why these systems have been historically overlooked—they escape genetic screens that rely on obvious phenotypes to be effective. It is becoming apparent that metabolite damage is an unavoidable consequence of life and that many metabolites are susceptible to enzymatic error or other damaging reactions. As such, many repair systems have been recently elucidated and there are certainly many others that function to mitigate metabolite damage.
confocal laser-scanning microscopy
deaminated ophthalmic acid
green fluorescent protein
methyl tert-butyl ether
Nitrilase-like protein 1
T.D.N. and A.D.H. conceived the project. T.D.N., A.D.H., J.S.F., J.A.P., D.C.A., R.T.M., S.D.B. and O.F. designed research. T.D.N., J.F., J.A.P., D.C.A., B.S.M. and M.P. performed research. All authors analyzed data. T.D.N. wrote the article.
This research was supported by U.S. National Science Foundation awards MCB-1611711 to A.D.H. and S.D.B. and MCB-1611846 to O.F., by a grant from the Natural Sciences and Engineering Research Council of Canada [no. 217291] to R.T.M., and by an endowment from the C.V. Griffin Sr. Foundation to A.D.H.
The Authors declare that there are no competing interests associated with the manuscript.