Arabidopsis thaliana WT Col-0 and hot5-2 (GABI_315D11) plants were used in this study and grown at 40–100 μmol m^–2 s^–1 at a 16 h light/8 h dark schedule and 22/18°C.

Leaves from 4- to 6-week-old soil grown WT Col-0 and hot5-2 plants were used in the quantitative proteomics experiment. One biological replicate consists of nine plants and five biological replicates per genotype were used. Frozen leave plant material (200 mg) was ground in extraction buffer [25 mM HEPES-NaOH pH 7.7, 1 mM EDTA, 2.5% (w/v) SDS, protease inhibitor cocktail (Pierce A32955, Thermo Fisher Scientific, United States)] and subjected to trichloroacetic acid (TCA)/acetone precipitation. For that, 20% TCA (final concentration) were added to the protein extract and incubated for 45 min at −20°C. Proteins were sedimented by centrifugation at 16,000 × g for 10 min at 4°C and following three washes with ice-cold 70% acetone, proteins were resuspended in extraction buffer [25 mM HEPES-NaOH pH 7.7, 1 mM EDTA, 2.5% (w/v) SDS, protease inhibitor cocktail (Pierce A32955, Thermo Fisher Scientific, United States)]. Protein concentration was determined using a BCA assay (Pierce 23225, Thermo Fisher Scientific, United States). Fifty micrograms of total protein per sample was run for 15 min on an 4–20% SDS-PAGE system to separate proteins from lower molecular weight contaminants, and the entire protein region of the gel excised and subjected to in-gel trypsin digestion after reduction with 50 mM dithiothreitol (DTT, Sigma-Aldrich, United States) and subsequent alkylation with 100 mM iodoacetamide (IAM, Sigma-Aldrich, United States). Peptides eluted from the gel were lyophilized and resuspended in 20 μL of 0.1% (v/v) formic acid (FA). A 3 μL injection was loaded by a Thermo Easy nLC 1000 UPLC onto a 2 cm trapping column and desalted with 8 μL mobile phase A (0.1% FA in water). Peptides were then eluted at 300 nL min^–1 onto a 75 μm i.d. × 50 cm RSLC column (Thermo Fisher Scientific, United States) using a linear gradient of 5–35% mobile phase B (0.1% FA in acetonitrile) over 150 min. Ions were introduced by positive electrospray ionization (ESI) using a stainless-steel capillary at 2.1 kV into a Thermo Orbitrap Fusion tribrid mass spectrometer. Mass spectra were acquired over m/z 300–1750 at 120,000 resolution (m/z 200) with an automatic gain target (AGC) of 1e6, a cycle time of 1 s, and data-dependent acquisition at top speed selecting the most abundant precursor ions for tandem mass spectrometry by higher-energy C-trap dissociation (HCD) fragmentation using an isolation width of 1.6 Da, maximum fill time 110 ms and AGC target 1e5. Peptides were fragmented with a normalized collision energy 27, and fragment ion spectra acquired at turbo speed in the linear ion trap. Dynamic exclusion was applied with an exclusion of 15 s after observing the same ion twice within 15 s.

Mass spectra were searched against the Uniprot A. thaliana databases (UP000006548_3702 and UP000006548_3702_additional, downloaded 07/2019) using MaxQuant version 1.6.7.0 with a 1% false discovery rate (FDR) at the peptide and protein level, peptides with a minimum length of seven amino acids with iodoacetamide-dependent cysteine carbamidomethylation, N-terminal acetylation, and methionine oxidation as fixed modifications. Enzyme specificity was set as C-terminal to arginine and lysine using trypsin as protease and a maximum of two missed cleavages were allowed in the database search. The maximum mass tolerance for precursor and fragment ions was 4.5 and 20 ppm, respectively, with second peptides and match between runs enabled. Label-free quantification (LFQ) was performed with the MaxLFQ algorithm (Cox et al., 2014) using a minimum ratio count of 2.

Identified protein groups generated by the MaxQuant program were uploaded to the Perseus program version 1.6.10.43 (Tyanova et al., 2016). Site only, reverse, and contaminant peptides were removed from the dataset and missing values were imputed using a normal distribution. Invalid values were then excluded, and empty columns were removed. The volcano plot function was used to identify proteins that were significantly changed using a T-test with a permutation-based FDR of 0.05 and an S0 of 0.1. For hierarchical clustering, LFQ intensities were first z-scored and significantly different proteins (two sample test, permutation-based FDR of 5%, 250 rounds of randomization) clustered using Euclidean as a distance measure for column and row clustering.

Arabidopsis thaliana AKR4C8 (AT2G37760.2), AKR4C9 (AT2G37770.2), AKR4C10 (AT2G37790.1), and AKR4C11 (AT3G53880.1) were cloned into pET23b-HIS₆-SUMO vector (Liu et al., 2019) by Gibson Assembly using the primers listed in Supplementary Table 1. Expression and purification were performed as previously, with slight modifications (Dreyer et al., 2021). Ulp1 cleavage of the N-terminal HIS6-SUMO tag after immobilized metal affinity chromatography (IMAC) was performed during overnight dialysis against phosphate buffered saline (PBS) supplemented with 2 mM Tris(2-carboxyethyl) phosphine (TCEP, Sigma-Aldrich, United States). Following dialysis, samples were reapplied to Ni-NTA material (Thermo Fisher Scientific, United States) equilibrated in Tris–HCl pH 7.9 and the flow through, containing the non-tagged AKR4C proteins, was collected and concentrated by centrifugation if necessary. Protein molecular mass and purity were analyzed by SDS-PAGE after desalting with PD-10 columns (GE Healthcare, United States) equilibrated with 30 mM Tris–HCl pH 7.9. The concentration was determined spectrophotometrically using a molar extinction coefficient at 280 nm (ε₂₈₀) of 50,420 M^–1 cm^–1 for AKR4C8, 61,420 M^–1 cm^–1 for AKR4C9, 59,930 M^–1 cm^–1 for AtAKR4C10, and 58,440 M^–1 cm^–1 for AtAKR4C11. The resulting homogeneous protein solutions were stored at −20°C.

S-nitrosoglutathione and CoA-SNO were prepared freshly as described previously (Stomberski et al., 2019b; Tagliani et al., 2021). Enzyme assays were performed in 100 mM Tris–HCl (pH 7.9) containing 0.2 mM NAD(P)H, 0.4 mM GSNO or SNO-CoA, and variable amounts of AKR4C proteins (50–200 nM) or A. thaliana leaf protein extracts (50–200 μg). Leaf protein extracts were obtained by resuspending 200 mg of plant material in extraction buffer [ratio 1:4 (w/v), 50 mM Tris–HCl (pH 7.9), 0.2% (v/v) Triton X-100, Pierce protease inhibitor cocktail (Pierce A32955, Thermo Fisher Scientific, United States), 0.5 mM DTT]. After desalting by gel filtration using spin columns (Zeba spin desalting columns, 7K MWCO, Thermo Fisher Scientific, United States), the protein concentration was determined by Bradford assay (BioRad protein assay, BioRad, United States). Reactions were performed in triplicate and initial rates were calculated from the absorbance decrease (340 nm) using a molar extinction coefficient of 7.06 mM^–1 cm^–1, which comprises both NAD(P)H and GSNO/SNO-CoA absorbance (Sanghani et al., 2006). The linear rate of the reaction was corrected to a reference rate without nitroso compounds.

For immunoblot analysis, proteins were extracted from 4- to 6-week-old soil grown WT Col-0 and hot5-2 plants as described previously (Kim et al., 2012; Liu et al., 2021). Proteins were separated by SDS-PAGE using 12% acrylamide gels and electrophoretically transferred onto nitrocellulose membranes (45 μm, GE Healthcare, United States), blocked with 10% (w/v) fat-free milk, and probed with primary antibodies (anti-AKR4C8/C9/C10/C11, PHYTOAB, United States; 1:5000; and anti-actin, Agrisera AS13 2640, Sweden; 1:3000) over night at 4°C. Following washing with TBST [TBS containing 0.1% (v/v) Tween-20], membranes were incubated with secondary HRP-conjugated antibody (anti-rabbit IgG, PHYTOAB, United States; 1:10 000) for 60 min at RT and signals were obtained by enhanced chemiluminescent detection (ECL).

For nitrosative stress treatment, WT Col-0 seedlings were grown in 0.5× MS media [0.5× MS media, 0.5% (w/v) sucrose, 0.5% (w/v) MES, pH 5.7]. At day 10, media was replaced by fresh media and 2.0 or 5.0 mM of NO donors (Diethylenetriamine NON-Oate, DETA/NO, Cayman Chemical, United States; GSNO) or NO scavenger (Carboxy-PTIO potassium salt, CPTIO, Cayman Chemical, United States) were added, and seedlings treated for 3 h at 40–100 μmol m^–2 s^–1 and 22°C. Proteins were extracted and separated on 12% NUSEP nUView gels (NUSEP, United States) and blotted to nitrocellulose membranes. After blocking in 10% (w/v) dry milk in TBST, the membranes were incubated overnight at 4°C with primary antibodies anti-AKR4C8 (PHYTOAB, United States; 1:5000) and anti-actin (Agrisera AS13 2640, Sweden; 1:3000), and washed three times with TBST before incubation with HRP-conjugated secondary antibody (anti-rabbit IgG, PHYTOAB, United States; 1:10 000) for 1 h at RT followed by ECL.

Mutation of the GSNOR/HOT5 gene (At5g43940) has pleiotropic effects, altering plant growth and development as well as plant reproduction (Lee et al., 2008; Kwon et al., 2012; Xu et al., 2013; Shi et al., 2015). GSNOR is suggested to be the major regulator of NO homeostasis by acting in the specific NADH-dependent degradation of GSNO. Although RNS accumulation and enhanced nitrosative stress (e.g., increased level of low-molecular and protein-SNOs) are observed in hot5-2 plants, it is not clear how higher GSNO leads to the observed mutant phenotypes. To investigate further how the absence of GSNOR might affect plant phenotypes through altering the expression of other proteins, we examined the total proteome of leaf material from hot5-2 and WT plants using quantitative proteomics. Total proteins were prepared from five biological replicates of 4- to 6-week-old soil grown WT and hot5-2 plants (Supplementary Figure 1A) and subjected to shotgun mass spectrometric analysis. A total of around 2500 proteins were identified across all samples (Supplementary Figure 1B and Supplementary Data Set 1), and statistical correlation analysis revealed high Pearson correlation coefficients between replicates (Supplementary Figure 1C). However, principal component analysis showed distinct clustering of WT and hot5-2 samples, indicating that there are significant differences in protein composition between the genotypes (Supplementary Figure 1D). Approximately 23% of the identified proteins (559 proteins) were differentially regulated in hot5-2 compared to WT and could be divided into two clusters (Figures 1A,B and Supplementary Data Set 1). Cluster 1 (293 proteins) represents proteins that are significantly upregulated, while cluster 2 (266 proteins) comprises proteins that are downregulated in hot5-2. GO-term analysis (Figure 1C) revealed that upregulated proteins in cluster 1 were associated with chloroplasts and chloroplast organization along with RNA binding and translation-related components. Down-regulated proteins in cluster 2 also included chloroplast terms, dominated by chloroplast membrane-related terms.

We further evaluated the proteomics dataset to visualize the most highly regulated proteins using volcano plot analysis (Figure 2 and Supplementary Data Set 1). As reflected in the GO-term search for chloroplast-associated terms, key proteins involved in chlorophyll biosynthesis such as magnesium chelatases CHLI1 and CHLI2 and protochlorophyllide oxidoreductase C (PORC) were upregulated (Figure 2A), while both nuclear and chloroplast-encoded subunits of the photosystem I reaction center PSI (e.g., PSAA, B and C, PSAL, and PSAF) and of photosystem II (e.g., PSBA, B, C, D and S, and LHC proteins) were down regulated (Figure 2B and Supplementary Data Set 1). We also observed differential regulation of stress and redox related proteins in the hot5-2 mutant (Figure 2C). While some proteins such as catalases (CAT1 and CAT2), dehydroascorbate peroxidases (DHAR2), heat shock proteins (HSP70 and HSP90), and glutaredoxins (GRXC2) are upregulated, other stress responsive proteins belonging to the peroxidase family (peroxiredoxins IIF and Q), cyclophilins (CYP18-3 and CYP20-3), and myrosinases (TGG1 and TGG2) are less abundant in hot5-2. Taken together, this suggests that tightly controlled NO homeostasis by GSNOR is important in the regulation of chloroplast processes and the general stress response.

The recent recognition that human AKR1A1 is involved in GSNO metabolism led us to also examine our data for changes in proteins belonging to the AKR superfamily; the disruption of NO metabolism in hot5-2 plants might impact expression of these enzymes, which are proposed to be involved in NO homeostasis (Stomberski et al., 2019b). Using the PFAM identifier PF00248¹ (for AKR domain-containing protein), out of the 31 proteins listed for this superfamily in A. thaliana, we found 11 proteins in our data set (Figure 2D, Table 1, and Supplementary Data Set 1). Notably, two proteins in the class 4C AKR protein family (AKR4C), AKR4C8 (At2g37760), and AKR4C9 (At2g37770), were significantly upregulated in hot5-2 (>2- and >1-fold, respectively; Table 1). The only other AKR superfamily member that showed a statistically significant increase (though <0.5-fold) was L-galactose dehydrogenase (LGALDH), which is a well-characterized enzyme of the ascorbate biosynthesis pathway (Smirnoff and Wheeler, 2000).

To understand the relationship of AKR4C8 and 9, as well as other A. thaliana AKRs to human AKR1A1, we recovered the closest homologs through a BLAST search of the A. thaliana protein database. Notably, AKR4C8 and 9, along with A. thaliana homologs AKR4C10 and 11, have the highest amino acid sequence identity to the human protein (Supplementary Table 2). Of the other AKRs identified in the proteomics data set, only At1g59960 and At2g21250 have greater than 25% sequence identity to AKR1A1, but these proteins were either down regulated or unchanged, respectively, in hot5-2 compared to WT (Figure 2D).

We used the AKR proteins in Supplementary Table 2 to construct a phylogeny including other human AKR proteins, along with selected AKR proteins from Saccharomyces cerevisiae and Chlamydomonas reinhardtii (Figure 3A). The tree clearly shows that the A. thaliana AKR4C proteins group together and that the plant and algal proteins have evolved independently from the human and yeast proteins. The phylogeny does not support a closer relationship between human AKR1A1 and any specific A. thaliana AKR protein. However, the increase in AKR4C8 and 9 in the hot5-2 mutant and higher sequence identity/similarity of these proteins to AKR1A1 (Supplementary Table 2) led us to focus further work on the AKR4C proteins.

Notably, the four AKR4C proteins are the most well-described plant AKRs with respect to their structure and in vitro oxidoreductase activity on a variety of substrates, though not including GSNO or SNO-CoA (Simpson et al., 2009; Saito et al., 2013). Their high degree of sequence identity is illustrated in Figure 3B, and atomic structures are available for both A. thaliana AKR4C8 (Figure 3C) and AKR4C9 (Simpson et al., 2009). They share the native fold that is typical of the AKR superfamily, consisting of a (α/β)₈-barrel (TIM barrel) with two additional α-helices on the periphery of the enzyme (Jez et al., 1997; Simpson et al., 2009). Three flexible, surface loops (A, B, and C, Figures 3B,C) are involved in defining the specificity of these enzymes toward the substrate, and the flexibility of these loops appears to allow interaction with multiple substrates (Simpson et al., 2009). Another conserved feature are four critical residues, Asp47, Tyr52, Lys81, and His114 (based on residue numbering in A. thaliana AKR4C8; Figure 3B), that form the catalytic tetrad (Mindnich and Penning, 2009).

To enable further studies of their expression levels and enzymatic activities, we cloned and purified all four A. thaliana AKR4C proteins (Supplementary Figure 2). We raised antibodies against each of the four proteins individually, tested their specificity and reactivity, and used them to confirm AKR4C levels are increased in hot5-2 as seen in the proteomics data. Not surprisingly, given the sequence similarity of the four proteins, antibodies against each AKR4C reacted with all four proteins (Supplementary Figure 3). We chose the anti-AKR4C8 antisera to probe for differences in AKR4C levels in hot5-2 compared to WT. Consistent with the proteomics data, we found an increase in AKR4C signal in the immunoblot analysis, where four different reactive bands were visualized with anti-AKR4C8 antisera, two of which were significantly increased (Figure 4A). We cannot definitively identify the individual bands, but quantification of all four bands indicates a minimum of a 50% increase compared to WT (Figure 4B), supporting the increases seen in the proteomics data.

Considering the global conservation of structural and catalytic-related elements of the AKR4C proteins with human AKR1A1 (Supplementary Figure 4), we determined the ability of these proteins to catalyze GSNO/SNO-CoA degradation. The NAD(P)H-dependent activity of purified enzymes was tested in presence of both nitroso-compounds. As shown in Figure 5A, all A. thaliana AKR4Cs efficiently degraded GSNO when NADPH was used as a source of reducing equivalents. Specific activities were roughly similar for AKR4C8 and AKR4C10 (2.27 and 2.76 μmol min^–1 mg^–1, respectively), while AKR4C9 and AKR4C11 were more efficient with specific activities ∼1.5- and ∼3-fold higher compared to AKR4C8/C10 (3.98 and 7.12 μmol min^–1 mg^–1, respectively). The strict specificity toward NADPH dependence was confirmed for all AKR4Cs with the sole exception of AKRC8, which exhibited NADH-dependent GSNO degradation activity that was, however, ∼20% of that measured in the presence of NADPH (0.51 μmol min^–1 mg^–1, Figure 5B). To test the substrate specificity of AKR4Cs, we monitored NADPH oxidation in the presence of SNO-CoA, replacing GSNO. As shown in Figure 5C, we observed the lowest specific activity for AKR4C8, whereas the other AKR4Cs were more efficient, with three- to fourfold higher activities than those measured in the presence of GSNO. Altogether, these results demonstrate that AKR4Cs are capable of degrading GSNO, but with opposite specificity toward the cofactor (NADPH vs. NADH) and with 20- to 60-fold lower specific activities than plant GSNORs (Guerra et al., 2016; Tagliani et al., 2021). Moreover, SNO-CoA appears as a suitable substrate for all A. thaliana AKR4Cs as observed for human AKR1A1 (Stomberski et al., 2019a).

Given the ability of A. thaliana AKR4Cs to catalyze GSNO degradation and the overexpression of AKR4C members in plants lacking GSNOR, we assessed NADPH-dependent GSNO degradation activity of leaf protein extracts from 4- to 6-week-old WT and hot5-2 plants (Figure 5D). We observed a ∼ 2-fold increase in GSNO degradation in the mutant lacking GSNOR, suggesting that the upregulation of specific AKR4C proteins in hot5-2 is responsible for the higher NADPH-dependent GSNO reduction. Thus, AKR4C activity may play a significant role in GSNO homeostasis partially compensating for the lack of GSNOR.