All peptides (see all sequences in Supplementary Table 1) were purchased at 95% purity (Genscript, Leiden, Netherlands). NAD was purchased from Roche (Basel, Switzerland). All other chemicals were purchased from Sigma-Aldrich (St. Quentin, France).

The full length (FL) sequence of SpNaa10 (177 residues) from pETDuet-SpNaa10(FL) was introduced between the NdeI and KpnI restriction sites of pETDuet-SpNaa15(FL)-SpNaa10(1–156; Liszczak et al., 2013) to replace the 1–156 truncated open reading frame with the FL and produce pETDuet-SpNaa15(FL)-SpNaa10(FL). Purification of SpNatA was as described in Gottlieb and Marmorstein (2019) with the exception of the last step of size exclusion chromatography, which was skipped as proven unnecessary for the sake of enzyme assays. Purity was assessed >99%.

The DNA fragment encoding amino acid residues 89–254 of GNAT2 from O. sativa (UniProt ID Q5KQI6) was inserted between the NcoI/BamHI restriction sites of the pETM41 vector featuring a dual N-terminal hexahistidine-Mannose binding protein (MBP) tag followed by a Tobacco Etch Virus (TEV) cleavage site. This plasmid (pETM41-OsGNAT2) was propagated in E. coli BL21-pLysS Rosetta2 cells (Novagen). After 0.25 mM IPTG induction at OD₆₀₀ = 0.6, cells were gown overnight for 20 h at 20°C in 2YT medium. Cells were lysed by sonication at 4°C in 20 mM Tris–HCl (pH 8.0), 0.5 M NaCl, and 5 mM 2-mercaptoethanol supplemented with 10 mM imidazole (Buffer A). The lysate was loaded directly onto an immobilized nickel ion affinity chromatography (HisTrap Crude FF, 5 ml, GE Healthcare) and protein eluted by a linear gradient of 10–250 mM imidazole in buffer A (2 ml/min, 100 ml). The pool of purified protein was next dialyzed 24 h against the storage buffer containing 20 mM Tris-HCl (pH 8.0); 0.2 M NaCl, 1 mM 1,4-Dithiothreitol (DTT), and 55% glycerol (Buffer B). The 6xHis-MBP-OsGNAT2 protein samples were stored at −20°C.

For preparation of OsGNAT2 devoid of MBP, the sample obtained after the aforementioned nickel affinity purification was further dialyzed overnight against a buffer containing 20 mM Tris–HCl (pH 8.0), 0.2 M NaCl, and 5 mM 2-mercaptoethanol supplemented with 10 mM imidazole (Buffer C) in the presence of 1 mg per 15 mg purified protein of a TEV protease version displaying a C-terminal polyhistidine-tag (To et al., 2015). The sample was next applied onto a nickel affinity column (HisTrap Crude FF, 5 ml, GE Healthcare), and the column was washed with 25 ml buffer C. The flow-through—corresponding to OsGNAT2 devoid of both the 6xHis-MBP tag and TEV-6xHis—was diluted four times in buffer D containing 20 mM Tris–HCl (pH 8.0), 1 mM DTT. The OsGNAT2 sample was loaded on an anion exchange column (HiTrap Q FF 5 ml), and the protein was eluted using a linear gradient of 0–1 M NaCl in buffer A (2 ml/min, 75 ml). Pulled fractions of the OsGNAT2 protein were then dialyzed against buffer B and stored at −20°C.

Plasmids used for the experiment were pETM41-OsGNAT2 (This work) for OsGNAT2 induction and pETM41-AtNAA15 for control induction as NAA15 does not have any acetylase activity (Linster et al., 2015). Cultures were grown in 30 ml LB media at 37°C until OD₆₀₀ reaches a value of 0.2. Next, induction was promoted by adding of 0.25 mM IPTG. Cultures were grown for another two generations or so in the exponential phase (OD₆₀₀ = 0.9–1). Cells were harvested, frozen in liquid nitrogen, and stored at −80°C. The sample was solubilized in 300 μl of a buffer containing 50 mM HEPES/NaOH pH 7.2; 1.5 mM MgCl₂; 1 mM EGTA; 10% glycerol; 1% Triton X-100; 0.15 M NaCl; 2 mM phenylmethylsulfonyl fluoride (PMSF); and one tablet of protease inhibitor cocktail (EDTA+; Merck) and was lysed by ultra-sonication. All further steps are reported in details in Bienvenut et al. (2020) and were performed without any deviation. The SILProNAQ pipeline for acetylation yield measurements is detailed in Bienvenut et al. (2017a,b).

GNATs activity was assayed at 30°C in a coupled assay with PDH (see reactions 1 and 2 in the section “Results and Discussion”). The reaction mixture contained 50 mM Tris–HCl (pH 8.0), 100 mM NaCl, 1 mM MgCl₂, 1 mM EGTA, 1 mM DTT, 0.2 mM thiamine pyrophosphate, 2 mM pyruvate, 2.5 mM NAD⁺, 0.150 units/ml of porcine heart PDH, 10–500 μM Ac-CoA (>93% pure, Sigma, A2056), and 10–5,000 μM peptides (Genscript). The reaction mixture was pre-incubated for 5 min at 30°C before starting the reaction by adding 10–500 μM Ac-CoA. A final volume of 200 μl was used in 96-well black plates (Grenier Bio One and Dominique Dutscher, Brumath, France; optical path for 0.2 ml is 0.44 cm). A value of 6,300 M⁻¹.s⁻¹ was used as the molecular extinction coefficient of NADH at 340 nm. An Infinite M Nano+ plate reader (Tecan, Lyon, France) was set at 340 nm to monitor the absorbance over time at 30°C.

Acetylation kinetics was monitored continuously for 5–120 min, and the data were fitted as to obtain the initial velocities associated to each peptide concentration. Curve fits to obtain kinetic parameters were achieved by non-linear regression with GraphPad Prism 9.1 (GraphPad Software, La Jolla, CA, United States). Parameters with SEs were computed for all parameters using the complete dataset including replicates. Both k_cat and K_m kinetic values were obtained by fitting to the Michaelis–Menten equation. When V_max could not be reached, the k_cat/K_m value was obtained with simple linear regression fit with Graphpad Prism 9.1.

The overall incubation conditions were identical to those of the PDH-coupled assay (see before) but the coupling reagents (i.e., MgCl₂, Pyruvate, NAD⁺, thiamine pyrophosphate, and PDH) and DTT were omitted. To ensure DTT would not challenge the reaction of DTNB with CoA, SpNatA was diluted strongly (>5,000-fold) from the 1 mM-DTT containing stock solution in reaction buffer without DTT leading to submicromolar concentration that eventually only poorly contributed to the signal. DTNB from a stock solution at 10 mM in ethanol was added at a final concentration of 0.1 mM in the incubation buffer. This corresponded to the best assay conditions compared to final DTNB concentrations of 1 mM obtained from 10 mM stock solutions solubilized either in ethanol or freshly prepared in potassium phosphate buffer solution, pH 7.2 containing 0.1 mM EDTA (Supplementary Figure 1). Absorbance was monitored continuously over time at 412 nm. A value of 13,600 M⁻¹.s⁻¹ was used as the molecular extinction coefficient of the thiophenolate (TNP) anion (Silverstein, 1975).

Matrix Assisted Laser Desorption Ionization Time-Of-Flight (MALDI-Tof) spectrometry was performed as described in Dian et al. (2020) as follows. About 100 μl of a mixture containing 100 mM Tris–HCl (pH 8.0), 200 mM NaCl, 1 mM EGTA, 1 mM DTT, 100 μM Ac-CoA, 5 μM OsGNAT2, and 0.5 mM of synthetic peptide TQTFIPGKDA (Genscript, Piscataway, NJ, United States) were incubated at 30°C for 60 min. Around 10 μl samples withdrawn before and after incubation were diluted in 90 μl of water/acetonitrile solution. The samples were then diluted five times in the matrix solution made of 5 mg/ml of α-cyano-4-hydroxycinnamic acid solubilized in water/formic acid/acetonitrile (50/50/0.1%). About 1 μl of each dilution was spotted on a metal target and dried. MS and MS/MS spectra of each sample were acquired with an AB SCIEX 5800 MALDI-Tof-Tof instrument in positive ion mode. Survey scans were performed using delayed extraction (390 ns) in reflector mode for a total of 15,000 shots. MS/MS scans were operated with collision energy of 1 kV. Peptide and fragment mass tolerances were set at 10 ppm and 0.8 Da, respectively. Mass spectra were analyzed with PeakView® 2.2 software (AB Sciex, Macclesfield, United Kingdom). The default threshold in MS/MS peak labelling and finding was 5%.

N-acetyltransferases transfer the acetyl moiety from Ac-CoA to the N-terminal ammonium of peptide substrates. They release coenzyme A (CoA), which can be used by decarboxylating enzymes such as those of the Krebs cycle in the presence of NAD⁺ to produce NADH. NADH is often used in spectrometric assay as a reporter because of its ability—unlike NAD⁺—to strongly absorb light at 340 nm and fluoresce at 465 nm when excited at 340 nm. PDH or α-ketoglutarate dehydrogenase were already used successfully as coupling enzymes using pyruvate or α-ketoglutarate, respectively, to produce NADH. Examples involve histone, palmitoyl, or myristoyl acyl transferases (Boisson and Meinnel, 2003; Berndsen and Denu, 2005; Traverso et al., 2013; Hamel et al., 2014). When using Ac-CoA as the acyl donor of the primary reaction, PDH regenerates Ac-CoA. This offers the advantage over another dehydrogenase of keeping the second substrate (Ac-CoA) constant during the reaction time, improving initial velocity duration and avoiding any aberration of the associated kinetics caused, for example, by product inhibition. We chose therefore PDH as a valuable coupling enzyme to start with. GNAT activity (equation 1) was assayed by continuously monitoring the formation of NADH in a coupled assay using PDH activity (equation 2) as follows (where R is any chemical compound):

In order to set the conditions of the assay, we chose as model NAT SpNatA, an enzyme extensively studied in the past using the discontinuous radioactive test (Gottlieb and Marmorstein, 2019). In the radioactive assay, the model peptide is composed of a fusion of the N-terminal heptapeptide from the yeast threonyl-tRNA synthetase (Ths1p), which was previously shown to be an excellent substrate in vivo of yeast NatA (Polevoda and Sherman, 2003; Liszczak et al., 2013). This peptide was fused to a C-terminal 12-mer tag derived from AdrenoCorticoTropic Hormone (ACTH) featuring reinforced strong basic character—including three Lys to Arg substitutions—aimed at tightly retaining the peptide on anionic beads including phosphocellulose filter papers such as P81 or LSA-50 (Arnesen et al., 2009; Appelmans et al., 2021). This allows to separate the labeled acetyl group attached on the peptide product from the bulk of radiolabeled Ac-CoA, which is negatively charged and does not bind the filter. This19-amino acid peptide SASEAGVRWGRPVGRRRRP defines the canonical NatA substrate (reference peptide) that was used throughout this study.

In order to set a relevant, robust coupled enzymatic assay, the initial mandatory step was to determine all conditions allowing PDH not to be rate limiting; under such conditions the NADH readout over time only relies on the first reaction. We used a PDH concentration slightly above the value previously validated as facilitating kinetic data recovery (Traverso et al., 2013). Next, we challenged the SpNatA activity by progressive increase of its concentration in the presence of the reference peptide substrate (SASEAGVRWGRPVGRRRRP) and Ac-CoA at equimolar concentrations (0.5 mM). At SpNatA concentrations beyond 0.1 μM, the reaction rates were no longer proportional and resulted in a loss of apparent velocity rate indicating that PDH was limiting (Supplementary Figure 1A). The SpNatA rate measured in the linear part of the curve was in perfect agreement with the data previously published under the same conditions but obtained with the radioactive assay (2 vs. 1.8 s⁻¹; Liszczak et al., 2013). With this information, we concluded that relevant rate measurements could be obtained provided the rate was lower than 0.2 μM.s⁻¹. This rate limit was applied to all further experiments. Any rate reaction exceeding this value had to be performed again with a lower concentration of enzyme to slow down the reaction and make it compatible with non-limiting PDH rate.

We next calculated the Michaelis–Menten kinetic parameters of the reference peptide using an array of concentrations ranging from 0.1 to 1 mM. We could calculate a k_cat of 9 ± 1 s⁻¹ and a K_m of 0.47 ± 0.12 mM (Table 1; Figure 1A). These data, albeit close to the previously reported values of 3 s⁻¹ and 0.34 mM, respectively, with the same peptide, feature a significantly higher k_cat value and a similar K_m. Higher k_cat values obtained with a coupled assay as compared to a radioactive assay were already observed with other acetyltransferases (see Table 1 in Kim et al., 2000). One most likely explanation lies in the inherent inhibitory effect of the product CoA on SpNatA (Gottlieb and Marmorstein, 2019). As already mentioned, CoA does not accumulate in the assay because it is immediately recycled in Ac-CoA by PDH, ensuring constant concentration of Ac-CoA. This therefore prevents the reaction from any possible inhibitory effect of CoA on the putative studied GNAT. This effect could also be associated to progressive decrease of the Ac-CoA concentration in the radioactive assay, which is not saturating and, decreases as the reaction progresses (see below). As a result, any decrease of Ac-CoA leads to a decrease of the velocity. By investigating further this phenomenon, we noticed that the concentration of Ac-CoA used in the assay was tightly associated with the quality of the raw data and signal to noise ratio. The lower the Ac-CoA concentration, the better the kinetics (Supplementary Figure 1B). The background noise—observed at high Ac-CoA concentrations—was assigned to the presence of free CoA allowing PDH to produce NADH without SpNatA. Free CoA can originate either from the Ac-CoA solution (only >93% pure at most in the stock solution) or from the hydrolysis of Ac-CoA in the reaction mixture. Indeed the Ac-CoA in stock solutions is preserved from hydrolysis by maintaining a slightly acidic pH; therefore, following mix in the neutral to slightly alkaline pH (8.0) reactor, hydrolysis of Ac-CoA may occur. As shown in Supplementary Figure 1B, a near-saturating amount of Ac-CoA of 500 μM induced a low signal-to-noise ratio as well as the establishment of the stationary phase duration giving the curves a sigmoid shape. This made uneasy the assessment of the initial velocity rate. In an attempt to optimize the assay, we used sub-saturating amounts of Ac-CoA (100 μM). Such conditions resulted in significant increase of the quality of the data and improved fit to the Michaelis–Menten equation. The newly calculated kinetic parameters for the reference peptide using 100 μM Ac-CoA in the coupled PDH assay, 9.4 ± 1.7 s⁻¹ and 1.0 ± 0.3 mM (Table 1; Figure 1A) respectively, showed a K_m value three times higher than that previously reported, and led to catalytic efficiency reduced by 2-fold compared to the value obtained at 500 μM Ac-CoA (Table 1). We finally observed that the catalytic efficiency determined with the coupled PDH assay at 100 μM Ac-CoA was almost identical to that assessed with the radioactive assay at 500 μM (9,279 vs. 8,800 M⁻¹.s⁻¹; Table 1). This indicated that SpNatA catalysis was working as efficiently in either assay, whatever the relative Ac-CoA concentration.

As the k_cat values obtained with 100 and 500 μM of Ac-CoA were about the same, we wondered if the K_m value of 59 μM for Ac-CoA previously determined for SpNatA was similar if measured with the PDH assay, which escaped product inhibition compared to the radioactive assay. As the peptide previously used to determine the K_m for Ac-CoA was the poly-R tailed reference peptide, we used it for comparison purposes. Although the peptide could not be incubated at saturating amounts because of its inhibition effect on the PDH (see below), in the attempt to obtain the most accurate K_m value for Ac-CoA, we performed several preliminary experiments for several concentration of Ac-CoA. An array of peptide concentrations was used to determine the initial velocity (v_i) and fit the Michaelis–Menten equation, providing a v_i/[E] value for each Ac-CoA concentration that reflects the value that would have been obtained with a saturating amount of peptide (Figure 1).

Of note, in kinetic analysis of multiple-substrate enzymes, the single-substrate study of Michaelis–Menten experiments is available only at saturating amounts of the second substrate; otherwise, the obtained kinetic parameters are only apparent values conditioned by both the concentration used and the K_m value of the second substrate. Our data show that the K_m for Ac-CoA is 20 ± 3 μM, three times lower than the value obtained with the radioactive assay. Therefore, we recommend the use of a concentration of Ac-CoA of 100–150 μM when using the PDH coupled assay with any unknown GNAT. At 100 μM, this ensures reaching >83% of the maximal theoretical velocity (V_max) of SpNatA, while optimizing the quality of the kinetics and minimizing the effect of the background noise observed at the highest Ac-CoA concentrations (see previous section). Note that whatever the assay considered up to now, only apparent V_max values are obtained as the Ac-CoA is not saturating; for instance, saturation is 89% with the radioactive assay at 500 μM Ac-CoA.

In an attempt to validate the viability of the PDH coupled assay, we challenged the results with another, already established continuous assay, relying on the reactivity of the Ellman’s reagent (5,5′-dithio-bis 2-nitrobenzoic acid, DTNB) with CoA (Kohlhaw, 1988). Reduction of the disulfide bond of DTNB to TNB allows simple continuous monitoring with a spectrophotometer at 412 nm. This assay was previously used with a number of acetyltransferases, including lysine acetyltransferases (Bode et al., 1993), p300 HAT (Thompson et al., 2004), human NAA50 (Foyn et al., 2017), A. thaliana and human NAA60 (see the section “Materials and Methods” in Støve et al. (2016) and Linster et al. (2020), and A. thaliana and Candida albicans NatB (Hong et al., 2017; Huber et al., 2020). Because of the large reactivity of DTNB, such an assay requires that the reaction mixture only contains traces of any thiol-reactive agent such as dithiothreitol or 2-mercaptoethanol and any non-thiol reducing agent cleaving disulfide bonds such as Tris(2-carboxyethyl)phosphine (TCEP). Despite these restraints, this assay is highly valuable as it is also almost independent of peptide sequence although cysteine-containing peptides need to be avoided. We could check with the PDH coupled assay that the specific activity of the enzyme assayed in the absence of reducing agents was unchanged. In contrast, when using the DTNB assay, time-course absorbance monitoring indicated that initial velocity never properly established and that the reaction rate was slowing down at least 100-fold more rapidly over time than in the case of the PDH coupled assay (Supplementary Figure 1C). This suggested rapid reaction of DTNB with an important cysteine side-chain of SpNatA, which modification induced inhibition. Pseudo-initial velocity rates could only be determined in the very first seconds of the reaction. The SpNatA catalytic efficiency as measured with the DTNB-based assay was one third of the coupled PDH and radioactive assay data (Supplementary Figure 1D; Table 1).

The high reactivity of DTNB with at least one exposed cysteine residue on the surface of SpNatA might indeed promote dissociation of the two subunits. An inhibitory effect of DTNB was previously described for other members of the GNAT family (Rudnick et al., 1992). We concluded that the DTNB assay was most risky to be used a priori, as it might promote biased conclusions depending on the GNAT used. DTNB was therefore useful in a case-by-case study, but only provided it has been validated with another assay before, as it was achieved with AtNAA60 (Linster et al., 2020). Therefore, DTNB coupled assay is not recommended in primary screening and discovery of any new GNAT activity. Nevertheless, a discontinuous version of the DTNB assay can be used provided the GNAT is purified and active without any reducing agent. This was successfully achieved for screening and assessing inhibitors of human NAA10 and NAA50 (Foyn et al., 2013) and assays of the Saccharomyces cerevisiae NatC complex (Grunwald et al., 2020).

The above data further demonstrated the need of a continuous assay for NAT studies but also that the coupled PDH assay needed further investigation for validation. We therefore increased the peptide concentration above the previous upper value of 1 mM in an attempt to characterize more precisely the kinetic parameters. Surprisingly, higher concentrations resulted in a decrease of the initial velocity rate (Figure 2A). As the data poorly fitted with the equation for substrate inhibition model (Cornish-Bowden, 1995), this suggested another explanation to the phenomenon (Figure 2A). We, therefore, investigated the hypothesis that the reference peptide could directly interfere with PDH activity due to its polybasic C-tail tag. PDH alone was mixed in presence of increasing peptide concentrations (up to 10 mM) and its activity measured in the presence of PDH. The data showed that beyond 1 mM, the reference peptide indeed caused critical loss of PDH activity (Figure 2B).

The crystal structure of SpNatA in complex with a bisubstrate inhibitor (Liszczak et al., 2013) shows that the binding and selectivity of NAA10 catalytic subunit of SpNatA mostly rely on the first four amino acids of the peptide target substrate (Figure 3A). The tail tag of the reference peptide, which is used for the detection of the product in the course of the radioactive assay or for HPLC-separation purposes, is not required for the PDH coupled assay. Therefore, we performed the assay with shorter peptides devoid of the poly-R tag. Unlike the reference peptide, these peptides poorly inhibit PDH (Figure 2B) and could be used at higher concentrations in the assay. The 7- and 8-mer peptides SASEAGV and SASEAGVR showed much higher K_m values (19 and 12 mM, respectively; Table 1; Figure 2C), while displaying identical turnover numbers (10 s⁻¹). Trimming the reference peptide eventually from 19 to 7 residues led to a 10-fold reduction of the catalytic efficiency. We next used an octapeptide featuring two prolines at positions 2 and 4. Such residues are known to promote inhibition of NatA reactivity (Van Damme et al., 2011). We could verify that this peptide indeed is not a substrate of NatA, as expected (Table 1). The data indicated that octapeptides fully reconstitute NatA specificity; this result could be verified only thanks to the PDH coupled assay as the radioactive assay requires a long poly-R tail. The data also suggested that either the tail tag or the length of the peptide, or both played a critical role in increasing the binding of the enzyme to its substrates. To investigate this hypothesis, we first analyzed the acetylation of a 4-mer peptide SASE. The results showed that this peptide, though still a substrate, was less prone to catalysis than longer peptides (Table 1; Figure 2C). Thus, the catalytic efficiency of SpNatA is positively correlated with peptide size. This showed that peptide length critically affects the ability of SpNatA to acetylate any polypeptide substrate (Table 1).

Because the poly-R tail has been widely used—not only with the radioactive assay—in the NAT literature—we next investigated its effect on the catalytic efficiency of SpNatA. We tested a peptide with the same length but originating from the natural sequence of yeast protein Ths1p. Compared to the reference peptide, this natural 19-amino acid N-terminal sequence (SASEAGVAEQVKKLSVNDS) is weakly charged, eliminating the probable interference with the PDH that we assigned to the strong basic character of the tail tag. This peptide showed indeed decreased inhibitory effect on PDH compared the reference peptide (Figure 2B). The data showed the catalytic efficiency was 1.5 fold higher, with the associated k_cat of the neutral peptide increased by a factor of 3.5 with respect to the reference, basic peptide. In contrast, the K_m value was two-fold higher than that of the reference peptide (Table 1). This suggested that the basic character of the C-tail improved peptide affinity to SpNatA. A structural analysis of the protein SpNatA revealed that a strongly negatively charge labile β₃β₄ loop, located at the extremity of the peptide binding site, nicely explains this increased affinity, as the Arg residues could form salt bridges with the conserved four aspartates and glutamates of the loop (Figures 3A,B). It is also interesting that yeast substrates of NatA show enriched composition starting with Ser and Ala to a smaller extent, followed by small chain amino acids, and feature one or two basic residues at positions ranging between 9 and 16 (Figure 3C). Taken together, these results show that the first four N-terminal residues, which determine enzyme selectivity, are not sufficient to promote both sufficient substrate binding and the highest efficiency to SpNatA. This impact was so far unknown as it was missed from early analysis only focused on the first five residues of the acetylated sequence (see Figure 1C in Colaert et al., 2009). This long-range effect could be unveiled and experimentally confirmed thanks to the new PDH coupled assay.

In order to check the validity of the PDH coupled assay and apply its relevance to the discovery of novel GNATs, we chose a homolog of the A. thaliana GNAT within the NAA90 family, the latest NAA family described so far (Giglione and Meinnel, 2021). This homolog from rice (O. sativa) was pointed out recently and surmised to display similar activity as AtGNAT2 (Table 2 in Giglione and Meinnel, 2021). This GNAT remains yet uncharacterized. It shows the highest homology with AtGNAT2 and we referred to it as OsGNAT2. Like AtGNAT2, OsGNAT exhibits an N-terminal extension predicted to act as a plastid-targeting signal. We cloned OsGNAT2 and expressed the protein devoid of its N-terminal targeting pre-sequence. OsGNAT2 was purified to homogeneity in frame with MBP fused at the C-side. The purified protein was used in the PDH coupled assay along with 100 μM Ac-CoA, the SpNatA reference peptide, and a significant activity could be detected (Table 2). OsGNAT2 specific activity was very similar to that of AtGNAT2 obtained with a different assay and with a dissimilar peptide substrate (Table 2). In contrast, very low or no activity was observed with natural amino acids (Lys and Arg) or metabolites (serotonin, 5-methoxytryptamine), which are classic small substrates of several GNATs (Table 2). This indicated that OsGNAT2 works on polypeptide and behaves as a NAT. Upon release of the His-MBP tag following TEV cleavage, we observed that OsGNAT2 devoid of MBP displayed an activity very similar to that of the MBP fused protein on the Tsh1p unmodified peptide (35 vs. 24 nmol.min⁻¹.μmol⁻¹, respectively). This indicated that the MBP tag did not influence the activity of the fused GNAT. Next, we observed that the Tsh1p unmodified peptide was far less efficient than the reference peptide with both forms. This indicated that OsGNAT2 likely displays specificity determinants in the far-remote C-terminal side as SpNatA does. Finally, we observed no significant improvement of the catalytic efficiency when increasing Ac-CoA concentration up to 500 μM, indicative of a K_m value of Ac-CoA lower than 30 μM. We concluded that the 100 μM Ac-CoA concentration selected for SpNatA was also perfectly suited for OsGNAT2.

To further challenge the N-terminal activity of OsGNAT2, we performed a GAP assay. This assay consists in expression of the NAT-MBP fusion in E. coli, an organism where few, and well-known proteins undergo NTA and which basic level is known (Bienvenut et al., 2015; Dinh et al., 2015; Schmidt et al., 2016). Next, the SILProNAQ pipeline, dedicated to proteome range quantitative analysis of protein N-termini, was applied (Table 3). The complete list of 1,299 non-redundant protein N-terminal peptides revealed by the GAP assay is compiled in Supplementary Dataset 1. Among them, 753 were quantified (368 in the control and 385 when OsGNAT2 was expressed). The data reveal an increase not only of the number of NTAed substrates but also of the average NTA yield of the N-termini when cells were grown in the presence of OsGNAT2 (Table 3; Figure 4A). Forty-nine E. coli proteins showed a huge increase of their NTA yield resulting from the expression of OsGNAT2. Figure 4B shows that OsGNAT2, like AtGNATs, exhibits a relaxed substrate specificity similar to Archaeal NATs (Liszczak and Marmorstein, 2013; Bienvenut et al., 2020) and that the first N-terminal amino acid is the major determinant (M, S, or T are retrieved but also A, I, and V; Table 3). Like SpNatA the following residues appear to be uncharged and hydrophilic. We cannot formally exclude that the observed relaxed specificity results from heterologous production of OsGNAT2 involving incomplete or heterologous folding of the recombinant protein (Baneyx and Mujacic, 2004; González-Montalbán et al., 2007; Rosano and Ceccarelli, 2014). Nevertheless, this relaxed specificity of both OsGNAT2 and AtGNAT2—was similar to that deduced in planta from the N-acetylation defects detected in the gnat2 knockout A. thaliana line (Bienvenut et al., 2020). This further confirmed the accuracy of the GAP analysis for plastid GNAT substrate specificity analysis, as already observed with the cytosolic plant forms (see Giglione and Meinnel, 2021 and references therein). The GAP assay allows to determine 21 proteins strongly acetylated by both OsGNAT2 (Supplementary Dataset 1) and AtGNAT2 (Bienvenut et al., 2020). OsGNAT2 and AtGNAT2 display therefore tightly overlapping specificity determinants featuring little preference for the N-terminal residue (Figures 4B,C). It is therefore most likely that both fulfill similar physiological role.

We selected decapeptide substrate representatives from the E. coli most strongly acetylated proteins in the OsGNAT2 GAP assay (grayed in Table 3) and exhibiting various classes of N-termini, namely A, S, and T. Among them, TQTFIPGKDA derived from the N-terminus of ribosomal protein S12 methylthiotransferase accessory factor (YCAO, P75838). SNSYDSSSIK derived from DNA gyrase subunit B (GYRB, P0AES6). AVAANKRSVM derived from stringent starvation protein (SSPA; P0ACA3). The three peptides were tested with the PDH coupled assay. Although we observed a high K_m value preventing to reach the V_max, the data indicated NTA activities of OsGNAT2 for the three peptides were comparable to that of the reference peptide (Table 2). However, these small peptides showed much improved performances compared to those of the reference NatA peptide devoid of the poly-R tail (SASEAGV and SASEAGVAEQVKKLSVNDS; Table 2). To additionally validate our GAP test, we characterized N-terminal peptides derived from a protein, which NTA was not affected by the expression of OsGNAT2 in bacteria (RTNPTTS and RTNPTTSNPE). As expected, the assayed peptides, independently of their length, were not substrates of OsGNAT2 in our PDH assay (Table 2).

We next observed that peptide length appeared to have a relatively poor effect on OsGNAT2 catalysis (Table 2). About 4- to 22-residues long peptides of the YCAO series indeed displayed similar activity (Table 2). The substrate specificity of OsGNAT2 appears therefore to be fully contained into the very first residues. This is unlike SpNatA, which needs long peptides to gain full efficiency. We therefore searched for other longer-range impacts. As the OsGNAT2 substrates displayed a similar pattern as the ones for SpNatA, with positively charged residues at positions 9–10, we challenged the influence of the basic character of the peptide on the catalytic parameters. To do so, we tested the 22 amino acid-long peptide TQTFIPGKDARWGRPVGRRRRP, featuring the basic C-tag fused to the TQTFIPGKDA YCAO substrate. The measured activity was 25 times higher than that of the 22-residues long peptide derived from the natural sequence (Table 2). This indicated that OsGNAT2 was very sensitive to the occurrence of basic residues remote from the N-terminus. This was finally confirmed by changing the three remote positive charges of YCAO 22-mer and the associated reduced catalytic rate associated with the new peptide variant (Table 2).

Overall, our data clearly indicate that, like SpNatA, efficient catalysis with OsGNAT2 requires both well suited N-terminal sequence and most likely also remote positively charged residues. Furthermore, this suggests that a negatively-charged patch functionally corresponding to the β₂β₄ loop of SpNatA (Figure 3) could be involved in the binding of substrates with a positively charged region. One such patch does occur at the N-terminus of OsGNAT2 between positions 83 and 95 (APIEEEEPLPEE). The role of remote positive charges as supplemental pattern contributing to substrate selection of GNAT2 in addition to the key role of very N-terminal residues is supported by the sequence logo, which was retrieved from the GAP study performed with either A. thaliana or O. sativa GNAT2. Indeed, polylysine tracks are clearly identified between positions 8 and 10 of both patterns (Figures 4B,C).

To finally check whether our data were fully in line with the logo displayed in Figure 4B, we selected a peptide sequence which could not be quantified in our analysis but which sequence did match the pattern shown in Figure 4B. We selected the MNMPMTERIR decapeptide. This peptide displays no internal K residue and therefore should bring unambiguous answer through the PDH-coupled assay. The sequence was derived from galactoside O-acetyltransferase (THGA, P07464). We could verify that the derived N-terminal decapeptide was indeed a substrate of OsGNAT2 with catalytic efficiency similar to the other decapeptides (Table 2).

We reported that plastid GNATs like OsGNAT2 display dual NAT and KAT activity. In the previous paragraph, several peptides devoid of K residues were substrates of OsGNAT2, in favor of NAT activity of the GNAT. Nevertheless, a number of selected peptides feature a K in their sequence. For instance, the YCAO decapeptide displays both a free N-terminus and an internal lysine at position 8. Therefore, we attempted to discriminate the N-acetylation type(s) OsGNAT2 ensured by mass spectrometry. After incubation of the peptide in the presence of OsGNAT2, MALDI-Tof mass spectrometry analysis revealed that the peptide was acetylated only in the presence of Ac-CoA and OsGNAT2 (Figures 5A,B). Next, fragmentation of the acetylated product of the reaction by MS/MS revealed the presence of four proteotypic ions of the N-terminal acetylated peptide only (Figures 5C,D). We could not observe any proteotypic ion of the K acetylation even after long incubation at higher concentrations of OsGNAT2. This does not mean that OsGNAT2 does not display KAT activity but rather that KAT substrates need yet to be discovered. We cannot exclude that fully folded proteins could be more appropriate substrates than peptides for this purpose. Such substrates would exhibit the reactive Lysine in a better-suited environment, which mimic the post-translational KAT activity of GNATs.