Structural determinants and cellular environment define processed actin as the sole substrate of the N-terminal acetyltransferase NAA80

NAA80 Acetylates Acidic N Termini Using a Ternary Complex Mechanism.

A unique NAT enzyme, NAA80, responsible for cellular Nt acetylation of processed β- and γ-actins was recently revealed (6). To investigate the in vitro activity of NAA80, an enzyme activity screening using a broad substrate library modified from Kuhn et al. (14) on purified enzyme was undertaken. Among all potential substrates tested, only the peptides with N-terminal sequences of MDEL₂₄ (119 ± 5.61 μM), DDDI₂₄ (50.5 ± 1.0 μM), and EEEI₂₄ (42.4 ± 0.95 μM) (Table S1) were acetylated. Substrate specificity studies in the linear range of enzyme concentration revealed that the best substrate for NAA80 was MDEL₂₄, with a fivefold higher product formation compared with the second best substrate, MDDD₂₄, while its cellular substrate, processed β-actin, DDDI₂₄ ranked third (Fig. 1A). Both MDEL₂₄ (p65) and MDDD₂₄ (unprocessed β-actin) represent cellular NatB substrates (12). Interestingly, another NatB substrate, MELL₂₄, which as MDEL₂₄, also contains a Met residue followed by an acidic residue, was not detectably acetylated. Thus, it seems that NAA80 strongly prefers peptide substrates with acidic amino acid residues both at positions 2 and 3, while it is more promiscuous regarding the nature of the amino acid at position 1.

To design inhibitors for NAA80, we investigated the enzymatic mechanism by bisubstrate kinetics and product inhibition experiments. In contrast to previous data suggesting a ping pong mechanism for NAA80/NAT6/Fus-2 (15), our data clearly supported a ternary complex mechanism (Fig. S1).

Selective and Potent Inhibitors of NAA80 Can Be Developed.

Since a ternary complex mechanism enables inhibition by bisubstrate analogs, we pursued NAA80 bisubstrate analog inhibitors (Fig. 1B) based on the results of the substrate screening and the sequences of the cellular substrates-processed β- and γ-actin (Table S1). We synthesized bisubstrate conjugates of CoA coupled to the tetrapeptides MDEL-NH₂, DDDI-NH₂, EEEI-NH₂, and MLGT-NH₂ via an acetamide linker (Fig. 1B). Inhibition studies revealed that CoA-Ac-DDDI-NH₂ was the most potent NAA80 inhibitor, with an IC₅₀ value of 0.38 μM: 3- and 3.3-fold more potent than CoA-Ac-EEEI-NH₂ and CoA-Ac-MDEL-NH₂, respectively (Fig. 1C). The negative control, CoA-Ac-MLGT-NH₂ based on a NatC substrate (16), did not show detectable inhibition of NAA80 at the highest concentration tested of 1 mM. To determine the selectivity of the most potent NAA80 inhibitor, CoA-Ac-DDDI-NH₂ was tested against a panel of human NATs (Fig. 1D). With a K_i value of 43 nM, the inhibitor was established to be both potent and selective. Interestingly, the monomeric NAA10, which has a known preference for acetylating acidic N termini in vitro (17), was the second most inhibited NAT enzyme. However, CoA-Ac-DDDI-NH₂ showed a 88-fold reduced potency for NAA10 compared with NAA80.

Overall Structure of the Actin NAT DmNAA80 Reveals Unique Features.

Crystallization trials with Homo sapiens NAA80 (HsNAA80) did not produce well-ordered crystals. HsNAA80 has several predicted unstructured regions, particularly in the immediate N terminus, and a very long nonconserved β6–β7 loop that is not found in the other NATs that have been characterized (Fig. S2A). A sequence alignment of various model organisms revealed that Drosophila melanogaster NAA80 (DmNAA80) lacks both the long β6–β7 loop and the unstructured N-terminal region found in HsNAA80 (Fig. S2B). We, therefore, pursued crystallization trials using DmNAA80.

To validate that DmNAA80 had comparable in vitro activity with HsNAA80, activity assays using purified recombinant DmNAA80 against selected substrate peptides were conducted. The results showed a preference for the DDDI₂₄ peptide (Fig. S2C), which matches well with the substrate preference of HsNAA80. Moreover, a rescue assay using NAA80 KO cells transfected with a control plasmid of empty V5, HsNAA80-V5, or DmNAA80-V5 verified the ability of DmNAA80 to Nt acetylate β- and γ-actin in human cells (Fig. 2A) (6). These in vitro and cellular data strongly support NAA80’s functional conservation as an actin-specific NAT in Drosophila.

We obtained crystals from an N-terminal deletion construct of DmNAA80, 9–168 (from here on only referred to as DmNAA80) bound to either Ac-CoA or the bisubstrate inhibitor CoA-Ac-DDDI-NH₂ and determined their crystal structures to resolutions of 2.00 and 1.76 Å, respectively. The structures were determined using molecular replacement with the bacterial NAT RimI [Protein Data Bank (PDB) ID code 2CNT] (18) prepared with CHAINSAW to remove Rim1 side-chain residues as a search model. Refinement statistics can be found in Table S2.

The DmNAA80/Ac-CoA and DmNAA80/CoA-Ac-DDDI-NH₂ structures are virtually identical, with an rms deviation of 0.33 Å for all shared atoms (Fig. S3). In the remainder of the text, the DmNAA80/CoA-Ac-DDDI-NH₂ structure will be described unless otherwise indicated. DmNAA80 has a general Gcn5-related N-acetyltransferase fold (19, 20) with a central conserved Ac-CoA binding core region flanked by N- and C-terminal segments, similar to other NAT structures (21–26), with a total of four α-helices and seven β-strands (Fig. 2B). While the overall fold of DmNAA80 is similar to other NATs, the substrate binding loops differ noticeably (Fig. 2C) and help explain the substrate specificity of NAA80.

In the NAT family (excluding NAA40), the α1–α2 and/or β6–β7 loops play important roles in mediating N terminus-specific substrate recognition. In particular, these loop regions constrict the peptide binding site, such that polypeptides cannot sit across the protein to insert internal lysine substrates for acetylation, explaining why NATs are restricted to N-terminal peptide acetylation (27). Notably, both of these regions differ in NAA80 (Fig. 2C). The α1–α2 substrate binding region is more open in NAA80 than other NATs, as it shifts about 7 Å away from the β6–β7 loop compared with the other NATs. This is accompanied by a shift of the substrate peptide, which makes extensive contacts to conserved residues in the α1–α2 substrate binding region (Fig. 2C). The β6–β7 loop is also in an open conformation in the structure. Together, the α1–α2 and β6–β7 loops make the substrate binding groove wider compared with the other NATs (Fig. 2C) (21, 23, 24).

A surface representation of DmNAA80, Schizosaccharomyces pombe NAA10 (SpNAA10), and HsNAA20 (Fig. 3) highlights the wider substrate binding groove in DmNAA80. Interestingly, the wider substrate binding groove in DmNAA80 is reminiscent of the lysine acetyltransferase (KAT) Gcn5, which has a more open peptide substrate binding site (Fig. 3C), and could suggest that DmNAA80 could accommodate protein lysine substrates if not for the active site specificity for the DDD N-terminal sequence as will be described below. Also of note is the closed conformation of Candida albicans NAA20 (CaNAA20), which Nt acetylates unprocessed β- and γ-actin (Fig. 3D).

DmNAA80 Substrate Binding Site Accommodates the Acidic Actin N Termini.

There was clear density for all four residues in the bisubstrate inhibitor (Fig. 4A). The DmNAA80 substrate binding site shows an electrostatic surface with many positively charged residues that would be able to accommodate the acidic β-actin N terminus (Fig. 4B), thus indicating substrate specificity toward acidic residues in the binding site. Both the backbone amides and the side chains make extensive interactions with the enzyme and ordered water molecules.

With regard to backbone interactions, the N terminus of the peptide makes a hydrogen bond to a backbone carbonyl oxygen of S124. The backbone oxygen of D1 also makes a hydrogen bond to the backbone nitrogen of S88 in the β-bulge (Fig. 4C). W36 serves to orient the peptide substrate by acting as a hydrogen bond acceptor to the amide nitrogen of D2 (28) and through extensive van der Waals contacts. The amide nitrogen of D3 hydrogen bonds with the side chain of S88, and the backbone oxygen of I4 hydrogen bonds with the side chain of R38 (Fig. 4C). Unusually, the amide nitrogen of I4 makes an intramolecular hydrogen bond with the side chain of D2 (Fig. 4C). Most generalized NATs do not make a hydrogen bond with the amide nitrogen of the fourth residue in the substrate peptide; the hydrogen bonds stop at the third residue (23). Notably, NAA40 (the other specific NAT) makes a hydrogen bond with the fourth backbone amide as well (22). This could be another way that NAA80 mediates specificity for Asp or Glu in the second position.

The interactions with the side chains of the peptide portion of the CoA-Ac-DDDI-NH₂ bisubstrate inhibitor help explain the substrate specificity of NAA80. D1 of CoA-Ac-DDDI-NH₂ interacts with the substrate binding site via hydrogen bonding to the side chain of S124 and a water-mediated hydrogen bond to the backbone amide of I126 and the side chain of T125 (Fig. 4D). D2 of CoA-Ac-DDDI-NH₂ forms a salt bridge with R43 of DmNAA80 (Fig. 4E), and R43 is additionally anchored in place by hydrogen bonds between the guanidino group, the backbone oxygen of W36, and the side chain of N33 (Fig. 4E). Hydrogen bonds are also made between D3 of CoA-Ac-DDDI-NH₂ and the side chains of residues S46 and S88 (Fig. 4F). The side chain of S46 is hydrogen bonded to the side chain of K73, and the side chain of D3 makes a water-mediated contact with R43. These indirect contacts with positively charged residues may help mediate specificity for negatively charged residues in this position. Finally, the side chain of I4 makes van der Waals contacts with DmNAA80 A42 of the α1 helix (Fig. 4F). Combined, these observations in the structure suggest a substrate preference of DmNAA80 determined largely by residues 2 and 3 in the substrate peptide, where the D2 residue seems to be the anchoring residue of the inhibitor. Importantly, the residues that make contacts to the peptide substrate are highly conserved among NAA80 orthologs (Fig. S2B).

The majority of the peptide interactions between the CoA-Ac-DDDI-NH₂ bisubstrate inhibitor and the enzyme are mediated by the α1–α2 region, with few contributions from the β6–β7 region, in contrast to the other more promiscuous NATs (excluding NAA40), where this is a major site for substrate recognition. In these NATs, three conserved tyrosines, one in the α1–α2 loop and two in the β6–β7 loop, hydrogen bond to the backbone amides and orient the peptide in place (21, 23, 24, 26). These tyrosines are not conserved in NAA80 (Fig. S2A). Indeed, the architecture of the peptide binding site in NAA80 causes the α1–α2 loop to shift away from its position relative to the other NATs (Fig. 2C), such that it can make the observed contacts with the substrate peptide. This difference in the structure helps account for the substrate preference of NAA80.

A structure alignment of CaNAA20 and DmNAA80 with the CoA-Ac-DDDI-NH₂ bisubstrate inhibitor implicates why NAA20, which acetylates unprocessed actin, may not be able to also Nt acetylate processed actin. Specifically, a comparison of the CaNAA20 and DmNAA80 active sites reveals that CaNAA20 contains a hydrophobic pocket for M1 that would not accommodate an acidic residue at this position, which is accommodated by the more polar DmNAA80 site for residue 1 (Fig. 4G).

Kinetic and Mutational Analyses of DmNAA80 Are Consistent with Structural Observations.

We conducted a kinetic and mutational analysis of DmNAA80 against a variety of substrate peptides. We observed activity of DmNAA80 against peptides mimicking the N terminus of β- or γ-actin with or without the iMet as well as an MDEL₂₄ peptide substrate. The catalytic efficiency of DmNAA80 was similar for the two iMet-starting peptides and the two processed actin peptides without the iMet (Table S3). Interestingly, the activity was higher against the peptides with iMet and much higher against the MDEL₂₄ peptide. The major change in kinetic parameters between these peptides is the k_cat, while the K_m remains relatively unchanged (Table S3).

To further probe the residues important for substrate binding and catalysis, we prepared mutant forms of the enzyme that targeted conserved residues, which make contacts with the peptide. We carried out complete kinetic curves for the mutant enzymes against DDDI₂₄, which represents the form of β-actin that NAA80 encounters in the cell (Table S3). Activity could not be detected for both the W36A and R43A mutants, corresponding to the residues that interact with D2. K73A and S88A, the residues that interact with D3, both had reduced catalytic efficiency compared with the WT enzyme, which was driven by a higher K_m (Table S3).

The structure of DmNAA80 suggests that the only residue in a position to act as a general base is E87, a highly conserved residue located in β4 (Fig. S4). Although it is not in a position to directly deprotonate the N terminus of the peptide, it is hydrogen bonded through a network of two ordered water molecules, which could mediate the deprotonation of the peptide substrate amino group (Fig. S4). We, therefore, made the mutant proteins E87A and E87Q and tested their activity. Surprisingly, these mutants were active and showed the same pH dependence on activity as the WT enzyme, suggesting that this residue does not play a role as a general base (Fig. S4E).

Previous studies of the NATs (23, 24) have described different residues important for catalysis for different members of the enzyme family. Some NATs use a general base to deprotonate the N terminus (21, 24). However, not all of the NATs use direct deprotonation of their substrates via a dedicated general base residue. The situation in NAA10 and NAA40 is ambiguous. Both of these enzymes have glutamate residues that are essential for catalysis yet are not directly positioned to deprotonate the N terminus based on the crystal structures (22, 23). Recently, the crystal structure of NatB was determined, and no residue was identified as a general base (26).

It is interesting that this is another NAT that does not use a general base for catalysis. This could be due to the fact that the pK_a of protein N termini is generally 2–3 units lower than lysine side chains (29). Intriguingly, in many NATs, there is a water molecule positioned directly by the N terminus, and this water molecule may be responsible for the direct deprotonation of the N terminus and does not need the assistance of a base due to the relatively low pK_a (Fig. S4).

The Narrow in Vivo Substrate Specificity of NAA80 Is Influenced by a Dominant Cotranslational NatB Activity.

A natural question arises of whether recognizing the second and third acidic residues of an N terminus is sufficient for limiting NAA80 to only acetylating β- and γ-actin in vivo. When searching for potential NAA80 substrates by only looking for N termini with acidic residues in the second and third positions (M-[D/E]-[D/E]), there are over 400 human proteins matching these criteria, which clearly cannot account for the specificity of this enzyme. Indeed, it was initially puzzling to us that the only substrates observed from cells were the processed acidic N termini of β- and γ-actin (6), yet NAA80 readily acetylates peptides beginning with methionine in vitro (Fig. 1A). In the cell, NatB (catalytic subunit NAA20) cotranslationally acetylates all proteins with an initial methionine followed by an acidic residue (12). Thus, to study the true in vivo capacity of NAA80, we made a yeast strain lacking NAA20/NAT3 (naa20Δ), and the strain was transformed with a plasmid carrying the human NAA80 gene. Fig. S5 shows tandem mass spectra of actin N-terminal peptides from the different yeast strains. As expected, no acetylated actin could be found in naa20Δ yeast, but surprisingly, the acetylation of the Met starting yeast actin was restored on expression of human NAA80. Furthermore, this pattern was observed for several other NatB-type N termini (MD, ME, MN, and MQ). The N-terminal peptides matching the NAA20 specificity, which were Nt acetylated in the WT strain and for which acetylation status could be determined in the NAA80 complemented strain, were assessed. Of the 11 unique NatB-type N-terminal peptides analyzed, 7 were acetylated in the naa20Δ strain expressing HsNAA80, while 4 were not (Fig. S5E). Although many of the substrates did not contain an acidic residue at both positions 2 and 3 of the N termini, all but one contained an acidic residue in at least one of these positions. In total, this supports the idea that NAA80 has the intrinsic capacity to acetylate a number of NatB-type substrates in vivo, but due to the cotranslational mode and high processivity of NatB, the only in vivo substrates of NAA80 are the processed actins.