Molecular basis for N-terminal alpha-synuclein acetylation by human NatB

NatB is one of three major N-terminal acetyltransferase (NAT) complexes (NatA-NatC), which co-translationally acetylate the N-termini of eukaryotic proteins. Its substrates account for about 21% of the human proteome, including well known proteins such as actin, tropomyosin, CDK2, and α-synuclein (αSyn). Human NatB (hNatB) mediated N-terminal acetylation of αSyn has been demonstrated to play key roles in the pathogenesis of Parkinson's disease and as a potential therapeutic target for hepatocellular carcinoma. Here we report the cryo-EM structure of hNatB bound to a CoA-αSyn conjugate, together with structure-guided analysis of mutational effects on catalysis. This analysis reveals functionally important differences with human NatA and Candida albicans NatB, resolves key hNatB protein determinants for αSyn N-terminal acetylation, and identifies important residues for substrate-specific recognition and acetylation by NatB enzymes. These studies have implications for developing small molecule NatB probes and for understanding the mode of substrate selection by NAT enzymes.

NatB is conserved from yeast to human in both complex composition and in its substrate specificity profile (Starheim et al., 2012). In Saccharomyces cerevisiae, the deletion of NatB subunits produces more severe phenotypes compared to the knockout of NatA or NatC subunits. Deletion of either NAA20 or NAA25 leads to similar phenotypes including slower growth rate, diminished mating, defects in actin cable formation, and aberrant mitochondrial and vacuolar inheritance (Polevoda et al., 2003). These observations suggest that the proper function of actin and tropomyosin requires NTA by the intact NatB complex (Polevoda et al., 2003). In humans, disruption of NatB (hNatB) by knockout leads to defects in proper actin cytoskeleton structure, cell cycle progression, and cell proliferation (Starheim et al., 2008;Ametzazurra et al., 2008;Ametzazurra et al., 2009;Neri et al., 2017). In addition, hNatB is upregulated in human hepatocellular carcinoma (Ametzazurra et al., 2008), where it has been suggested as a potential therapeutic target as silencing of this complex can block cell proliferation and tumor formation (Neri et al., 2017). hNatB-mediated NTA of a-synuclein (aSyn) has been shown to increase aSyn stability and lipid binding, and to reduce aggregation capacity (Dikiy and Eliezer, 2014;Watson and Lee, 2019;Mason et al., 2016;Maltsev et al., 2012;Trexler and Rhoades, 2012;Fernández and Lucas, 2018;Fauvet et al., 2012;Kang et al., 2012;Iyer et al., 2016). Since aSyn is a key protein in Parkinson's disease (PD; Halliday et al., 2011;Spillantini et al., 1998), hNatB might play an indirect role in PD pathogenesis in vivo as supported by a recent study (Vinueza-Gavilanes et al., 2020). It was also recently demonstrated that NTA of aSyn increases its propensity for lipid membrane binding without altering its structural properties of the bound state (Runfola et al., 2020).
Compared to the comprehensive structural and biochemical characterization of NatA (Liszczak et al., 2013;Magin et al., 2017;Knorr et al., 2019;Deng et al., 2019;Gottlieb and Marmorstein, 2018;Weyer et al., 2017), the study of NatB has been limited, particularly in humans. Recently, the crystal structure of Candida albicans (Ca) NatB bound to a bisubstrate CoApeptide conjugate was determined, providing important insights into substrate specificity and NTA by caNatB (Hong et al., 2017). However, hNAA20 and hNAA25 share only~40% and~20% sequence identity with the Candida albicans homolog (Hong et al., 2017), respectively. Moreover, nearly all biological studies of NatB have been conducted in Saccharomyces cerevisiae (Lee et al., 2014;Caesar et al., 2006;Singer and Shaw, 2003), Arabidopsis (Ferrández-Ayela et al., 2013;Huber et al., 2020), mouse (Ohyama et al., 2012), and human (Starheim et al., 2008;Ametzazurra et al., 2008;Ametzazurra et al., 2009;Neri et al., 2017) as model organisms. As a result, the mode of human NatB-mediated catalysis and aSyn-specific NatB recognition remains unresolved. In this study, we report the 3.5 Å resolution cryo-electron microscopy (cryo-EM) structure of the~130 KDa hNatB bound to a bisubstrate CoA-aSyn conjugate together with a structureguided analysis of mutational effects on catalytic activity. This analysis reveals functionally important structural differences between hNaB and related NAT enzymes, as well as insights into the molecular mechanisms that define aSyn and related substrates that are recognized for hNatB-mediated N-terminal acetylation.

Results hNatB is potently inhibited by a CoA-aSyn conjugate
While attempts to express recombinant hNatB in E. coli were unsuccessful, we found that overexpression of hNatB complex with full-length hNAA25 (residues 1-972) and C-terminally truncated hNAA20 (residue 1-163 out of 178 total residues) in baculovirus-infected Sf9 insect cells produced soluble protein that could be purified to homogeneity ( Figure 1A). To evaluate the activity of the recombinant hNatB, we tested it against different peptide substrates. aSyn with an N-terminal sequence of 'MDVF' has been widely considered as an in vivo hNatB substrate (Van Damme et al., 2012;Anderson et al., 2006;Ohrfelt et al., 2011;Theillet et al., 2016). We, therefore, incorporated this sequence into a peptide substrate named 'MDVF' for an in vitro acetyltransferase assay ('MDVF' peptide sequence: NH 2 -MDVFMKGRWGRPVGRRRRP-COOH). In agreement with in vivo studies (Van Damme et al., 2012;Anderson et al., 2006;Ohrfelt et al., 2011;Theillet et al., 2016), we observed that the purified recombinant hNatB was active against this 'MDVF' peptide, while no activity could be observed in the absence of either the enzyme or peptide ( Figure 1B). hNatB also showed no observable activity if either the first residue 'M' or the first two residues 'MD' in this aSyn peptide substrate was removed ('DVFM' peptide sequence: NH 2 -DVFMKGLRWGRPVG RRRRP-COOH; 'VFMK' peptide sequence: NH 2 -VFMKGLSRWGRPVGRRRRP-COOH; Figure 1B), suggesting that peptide substrate recognition by NatB is highly dependent on the first two N-terminal residues. To further confirm the substrate specificity of hNatB, we tested it against several previously identified peptide substrates for other NATs ('SASE' peptide sequence (NatA-type): NH 2 -SA SEAGVRWGRPVGRRRRP-COOH; 'MLRF' peptide sequence (NatC-type): NH 2 -MLRFVTKRWGRPVG RRRRP-COOH; 'SGRG'/H4 peptide sequence (NatD-type): NH 2 -SGRGKGGKGLGKGGAKRHR- COOH; 'MLGP' peptide sequence (NatE-type): NH 2 -MLGPEGGRWGRPVGRRRRP-COOH; Figure 1B). Consistent with previous results (Van Damme et al., 2012), hNatB is only active toward its unique canonical substrate type, displaying no overlapping activity toward other NAT substrates.
In order to understand the mechanism of hNatB substrate recognition, we synthesized a bi-substrate inhibitor in which the first 10 residues of aSyn are covalently linked to CoA (Liszczak et al., 2013) for enzymatic and structural studies. Half-maximum inhibitory concentration (IC 50 ) determinations revealed that this CoA-aSyn conjugate had an IC 50 of about 1.63 ± 0.13 mM ( Figure 1C), significantly lower than the K m values we had determined for hNatB toward a 'MDVF' peptide (45.08 ± 3.15 mM) and acetyl-CoA (47.28 ± 5.70 mM) ( Table 1).
hNatB reveals potentially biologically significant structural differences with hNatA and caNatB We performed single particle cryo-EM of hNatB in the presence of the CoA-aSyn conjugate. A 3.46 Å -resolution cryo-EM three-dimensional (3D) map was determined from 982,420 particles, selected from 5281 raw electron micrographs (  . The central core region of the EM map contains excellent sidechain density with a local resolution of~2.5 Å , particularly around the catalytic subunit, hNAA20. Consistent with previous NAT structural studies, the atomic model of hNatB features a catalytic subunit, hNAA20, that adopts a canonical Gcn5-related N-acetyltransferase (GNAT) fold Neuwald and Landsman, 1997). Additionally, the model reveals that the auxiliary subunit hNAA25 is composed of a total of 39 a-helices among where the predicted first and second a-helices were built as poly-alanine due to a lack of resolvable sidechain density (Figure 2figure supplement 3). The 39 a-helices can be roughly divided into three groups: an N-terminal region: a1-a8; a core region: a9-a29; and a C-terminal region: a30-a39 ( Figure 2A and Video 1). The N-terminal region (residue 1-164) displays relatively weak EM density compared to other regions, suggesting that it is relatively flexible (Figure 2-figure supplement 2). The eight helices of the N-terminal region form four helical bundle tetratricopeptide repeat (TPR) motifs, which often participate in protein-protein interactions. While there are no visible contacts between the N-terminal TPR motifs and hNAA20, it is possible that this region participates in ribosome association, similar to the N-terminal region of the NAA15 auxiliary subunits of Schizosaccharomyces pombe (Magin et al., 2017) and Saccharomyces cerevisiae (Knorr et al., 2019) NatA. The 21 helices of the core region also form a number of TPR motifs, which come together to form a ring that completely wraps around and extensively contacts hNAA20 within its hollow center ( Figure 2A). Indeed, the interaction between hNAA20 and the TPR motifs of this core region buries a total interface area of about 2300 Å 2 . In the core region, it is noteworthy that there is a long a-helix (a28, ranging 30 residues) that traverses almost from one side of the complex to the other. The a28 helix closes the core ring structure, locking hNAA20 in position, and bridging the N-and C-terminal and regions. This is similar to the role played by a29-a30 of the hNAA15 auxiliary subunit of hNatA ( Figure 2A and B). The C-terminal region features helices that bundle together to protrude out of the plane of the core ring structure at an angle of~45˚( Figure 2B).  The sequence identity of the catalytic and auxiliary subunits of hNatA and hNatB are 20% and 15%, respectively. To understand how this translates to key structural differences, we superimposed the crystal structure of hNatA (PDB: 6C9M) with our model. Between the catalytic subunits, there is a high degree of superposition (1.151 Å root-mean-square deviation [RMSD] over 105 common C a atoms), except for an additional helix, a5, on the C-terminus of hNatA-NAA10, which is absent in hNatB-NAA20 ( Figure 2B). Between the auxiliary subunits, the core and N-terminal regions of both hNatA-NAA15 and hNatB-NAA25 display similar topology, although a higher degree of deviation than the catalytic subunits. The core regions both wrap around their respective catalytic subunits (8.369 Å RMSD over 262 common C a atoms), while the N-terminal regions jut off to the side (7.360 Å RMSD over 55 common C a atoms). By contrast, the C-terminal regions of hNatA-NAA15 and hNatB-NAA25 diverge significantly from one another. For hNA25, the C-terminal region of hNAA25 is oriented toward its N-terminal region, while the C-terminal region of hNAA15 is positioned~50 Å away from the relative position of the superimposed C-terminal domain of hNAA25 ( Figure 2B). The positioning of hNAA25 may serve to promote hNAA25 intra-termini communication, which is similar to the interaction of hNatA and its regulatory protein HYPK (Gottlieb and Marmorstein, 2018). HYPK, which does not interact with hNatB, interacts with both the N-and C-terminal domains of hNatA-NAA15, potentially serving as bridge to enable closer communication between these two domains ( Figure 2-figure supplement 4). Recent reports have described the role of the small molecule IP 6 (inositol hexakisphosphate) in hNatA activity, where it is found to act as 'glue' between the C-terminal and core domains in hNAA15 and hNAA10 via a series of hydrogen bonds and electrostatic interactions (Gottlieb and Marmorstein, 2018;Cheng et al., 2019). While no corresponding small molecules have been identified to play a similar role in hNatB, our model shows that this interaction is replaced by an extended loop that connects the a31 helix with the a32 helix of hNatB-NAA25. This loop, which is not present in hNatA-NAA15, appears to mediate hydrophobic interactions between hNatB-NAA25 and -NAA20, likely to serve a similar role as IP 6 (Figure 2-figure supplement 4B).
We also compared the structures from human and the previously described C. albicans NatB (CaNatB, PDB: 5K18). Although the two superimposed structures revealed a high degree of structural conservation (NAA20: 0.698 Å RMSD over 125 common C a atoms; NAA25 Core region: 3.267 Å RMSD over 266 common C a atoms), the N-terminal region of hNatB-NAA25 appears to overlay more closely to hNatA-NAA15 than to CaNatB-NAA25 ( Figure 2B and C). Compared to CaNAA25, the N-terminal regions of hNatB-NAA25 and hNatA-NAA15 are positioned more closely to the peptide substrate binding sites of the respective catalytic subunits. Based on the role that the N-terminal yeast NatA-Naa15p regions play in ribosome docking (Magin et al., 2017;Knorr et al., 2019), we propose that the relative shift in the position of the N-terminal regions of the human NAT auxiliary subunits, hNAA15 and hNAA25, may reflect a difference in the mechanism for ribosome association and cotranslational NTA in C. albicans compared with humans. In addition, the overlay of C-terminal regions of hNAA25 and CaNAA25 displays an RMSD of 15.960 Å over 133 common C a atoms. We observe that the main difference that Video 1. Overall view of the NatB complex. hNaa20 (light orange) and hNaa25 (cyan) are shown in cartoon. The CoA-aSyn conjugate inhibitor is shown in sticks and colored as magenta.
https://elifesciences.org/articles/57491#video1 contributes to this deviation in this region is the length of helices.

hNatB makes specific interactions with the first 4 N-terminal residues of aSyn
In the cryo-EM map, density for the CoA-aSyn conjugate bisubstrate inhibitor is well resolved, allowing us to confidently model the CoA portion and the first 5 N-terminal residues (of 10 residues present) of the aSyn portion ( Figure 4A and B, Video 2). Similar to other NATs, CoA enters the catalytic active site through a groove formed by a3 and a4 of the catalytic subunit, while the peptide substrate enters the active site on the opposite side of the catalytic subunit flanked by the a1-2 and b6-b7 loops (Deng and Marmorstein, 2020; Figure 4A). hNAA20 contains a conserved acetyl-CoA binding motif among NATs: R 84 R 85 XG 87 XA 89 (Figure 4-figure supplement 1). Here we observe that the positively charged hNAA20-Arg85 interacts with the negatively charged 3'-phosphorylated ADP portion of CoA to form a salt bridge while Arg84 makes Van der Waals interactions ( Figure 4C and D). A hydrogen-bonding network is formed mainly between the 5'-diphosphate group and backbone atoms of a few residues including Val79, Gly87, Ala89, and Ala90 ( Figure 4C and D), and mediated by the sidechains of Arg85 and Gln125. The CoA molecule anchors to the binding pocket through a series of van der Waals contacts formed by residues Ser67, Val79, Leu77, Leu88, Val118, Met 122, and Tyr123 ( Figure 4D).
Four N-terminal residues of aSyn participate in hNAA20 interactions. Anchoring of the aSyn peptide is mediated by protein hydrogen bonds with the backbone atoms of Met1 and Asp2 of aSyn. Hydrogen bonds are formed between the backbone N-H group of aSyn-Met1 and the backbone carbonyl group of hNAA20-Phe111, as well as the backbone carbonyl group of aSyn-Met1 with the sidechain of hNAA20-Tyr138. The backbone N-H and carbonyl of aSyn-Asp2 also form hydrogen bonds to the sidechain of Tyr27 and between the backbone carbonyl group of Asp2 and sidechain of hNAA20-Tyr27 ( Figure 4D). Remarkably, hNAA20 contacts each of the first 4 N-terminal residue sidechains of aSyn via van der Waals interactions. The only sidechain that forms a hydrogen bond with hNAA20 is aSyn-Asp2, which hydrogen bonds with a hNAA20-His73 ring nitrogen and the hNAA20-Thr75 sidechain ( Figure 4D). The more extensive van der Waals interactions include the following: aSyn-Met1 interacts with hNAA20 residues Glu25, Phe27, Tyr56, and Ala76; aSyn-Asp2 interacts with hNAA20 residues Tyr27, Thr75, His73, Phe111, and Tyr138; aSyn-Val3 interacts with hNAA20 residues Tyr137 and Tyr138; and aSyn-Phe4 interacts with hNAA20 residues Glu25 and Als140. aSyn-Met5 does not appear to make specific interactions ( Figure 4D). Consistent with the importance of the residues that mediate aSyn binding, most of the residues are highly conserved from yeast to humans (Figure 4-figure supplement 1).

Mutational analysis identifies key residues for hNatB catalysis and cognate substrate binding
To determine the functional importance of hNAA20 residues that appear to make important peptide or CoA substrate contacts in our model, we used an in vitro acetyltransferase assay to kinetically characterize WT and mutant hNatB proteins. Each mutant was purified to homogeneity and displayed identical gel filtration chromatography elution profiles (data not shown), indicating that they were all properly folded. We prepared alanine mutants of several residues involved in the CoA binding including Arg84, Arg85, Gly87, and Tyr123. Among them, R84A, R85A, and G87A did not show significant defects in overall protein catalytic function (Table 1). However, a Y123A mutant nearly abolished protein activity, with a 95% loss of protein activity, affecting both k cat and K m ( Table 1). To further interrogate the properties of this residue, we prepared a Y123F mutant which features a similar aromatic bulky sidechain but not the polar p-hydroxyl group. We observed that Y123F displayed a similar~88% loss of k cat , but had a negligible effect on the peptide K m ( Table 1). These data suggested that the Tyr123 hydroxyl group is critical for catalysis but not required for substrate binding, while the aromatic ring of Tyr123 plays a role in peptide substrate binding. Given that the hydroxyl group of Tyr123 is about 3.5 Å from the sulfur atom of the CoA-aSyn conjugate and 6.3 Å away from the aSyn N-terminus, it is in a position to play a role as a general base or acid for catalysis, potentially through an intervening water molecule ( Figure 5A). This is analogous to the proposed general base role of Tyr73 as a general base for hNAA50 catalysis (Liszczak et al., 2011;see Discussion).
We also prepared alanine substitutions for residues that appeared to play important roles in aSyn binding: Glu25, Tyr27, His73, Tyr137, and Tyr138. We were surprised to find that mutations of hNatB residues that mediated backbone hydrogen bond interactions, Y27A and Y138A, had relatively modest effects on aSyn peptide NTA with Y27A showing~twofold higher K m and Y138A showing~twofold reduced K cat , together suggesting that sidechain contacts might dominate the binding energy (Table 1). Consistent with this, and our structural observations, we found that H73A produced a~90% reduction in activity ( Table 1). This correlates with the importance of the His73 hydrogen bond and van der Waals contacts with aSyn-Asp2. Of note, other cognate hNatB sidechain residues at position 2, Glu, Gln, and Asn would also be well positioned to form hydrogen Video 2. Overall view of a-synuclein N-terminal interactions by NAA20. Amino acid sidechains that mediate hydrogen bond and van der Waals interactions with a-synuclein are highlighted in a cartoon model of NAA20.
https://elifesciences.org/articles/57491#video2 bonds with His73. Together, hNAA20-His73 appears to play a critical role in cognate substrate recognition by hNatB. hNatB-Asn116 is a highly conserved NatB residue (Figure 4-figure supplement  1) that caps the hNAA20 a4 helix, which also harbors the putative catalytic residue, Tyr123, and is also in position to make a water-mediated hydrogen bond with to CoA pantetheine nitrogen ( Figure 5A). This observation suggests that Asn116 could play an important functional role. To test this, we prepared and evaluated an N116A mutant and, consistent with our hypothesis, Ans116, we found that this mutant leads to~90% loss in activity (Table 1).
Taken together, our structural and mutational analysis of hNatB highlight the functional importance of hNatB-NAA20 residues His73, Asn116, and Tyr123 in hNatB-mediated N-terminal acetylation. While Tyr123 appears to play a critical catalytic role, potentially as a general base and/or acid for catalysis; His73 appears to play an important role in the recognition of substrate residue two and Asn116 likely plays a structural role ( Figure 5A). Each of these residues could employ a bridging water molecule to mediate their functional roles, although these putative water molecules are not visible at the current resolution of our structure.

Discussion
Since the identification of hNatB more than a decade ago, many studies have shown that it N-terminally acetylates important proteins such as actin, tropomyosin, CDK2, and a-Syn, and its function has connections to diseases such as hepatocellular carcinoma and Parkinson's disease (Starheim et al., 2008;Polevoda et al., 2003;Ametzazurra et al., 2008;Neri et al., 2017;Halliday et al., 2011;Spillantini et al., 1998). Despite its clear biological importance, hNatB-mediated NTA by hNatB remained poorly understood. Here we developed a CoA-aSyn conjugate hNatB inhibitor, determined the cryo-EM structure of CoA-aSyn inhibitor-bound hNatB, and carried out associated structure-guided mutagenesis and activity assays. This has led to the identification of functionally important differences with human NatA and C. albicans NatB. These studies have also provided evidence for important hNatB-specific elements responsible for aSyn recognition and N-terminal acetylation, providing direct implications for NatB recognition of other canonical substrate proteins.
Consistent with previous studies, we have demonstrated that hNatB acetylates a cognate 'MD' N-terminus and is unable to N-terminally acetylate non-cognate N-termini that are substrates for other NATs such as NatA, NatC, and NatE. We demonstrated for the first time that hNatB can acetylate an a-Syn peptide in vitro, directly linking hNatB to NTA of a-Syn. We have also demonstrated that aSyn peptides lacking the Met1 or both the Met1 and Asp2, do not serve as hNatB substrates, confirming the strict substrate specificity of hNatB. This is consistent with our structural model showing significant interactions between hNatB-NAA20 and both the first and second N-terminal residues of an aSyn peptide, with important but less extensive interactions with the third and fourth residues. This hierarchy of interactions likely explains how NatB enzymes can accommodate cognate substrates that diverge at positions three and four.
Here we have presented hNatB-aSyn interactions that can be used to rationalize the substrate specificity of hNatB: N-terminal sequences containing 'MD-', 'ME-', 'MN-', and 'MQ-'. aSyn-Met1 sits in a hydrophobic pocket that comfortably accommodates a methionine residue, whereas shorter sidechains or longer polar or charged sidechains would fit poorly ( Figure 5B-C). The nature of this binding pocket is similar to the described hNAA50 recognition of Met1 (Liszczak et al., 2011;Figure 5B-C). Although both of hNAA50 and hNAA20 can N-terminally acetylate peptides with Met at the first position, no overlapping activity has been observed. This can be rationalized based on the chemical properties of the second residue in the cognate peptide. We find that the aSyn-Asp2 sidechain forms hydrogen bonds with the hNAA20 sidechains His73 and Thr75. These polar residues in the peptide binding site of hNAA20 would likely serve as poor acceptors of the largely hydrophobic residues targeted by hNAA50. The hNatB substrate client profile featuring D-, E-, N-, or Q-residues in position two is consistent with the mechanisms of substrate recognition observed in the binding pocket for aSyn-Asp2. The aliphatic regions of each of these sidechains (D, E, N, Q) would all benefit from the extensive hNatB van der Waals interactions surrounding the aliphatic region for aSyn-Asp2 (Tyr27, His73, Thr75, Phe111, and Tyr138). This second residue would also form a hydrogen bond interaction with His73 and Thr75, which may accommodate the carboxyl sidechains of both the shorter D-and N-and longer E-and Q-sidechains. Notably, His73 and Thr75 are strictly conserved from yeast to man ( Figure 5 and Figure 2-figure supplement 3). In contrast, shorter polar or nonpolar sidechains would less efficiently fill the pocket for residue two, while larger polar or charged sidechains would likely result in steric clashes. In agreement with this, we have demonstrated that H73A mutation has a severe impact on hNatB catalysis ( Table 1).
The hNatB/CoA-aSyn structure has implications for the mode of hNatB catalysis. While previous studies have suggested that hNAA50-Tyr31 plays an important role in catalysis, mutation of the corresponding hNaa20 residue, Tyr27, had minimal effects on hNatB kinetic parameters. Strikingly, our mutational analysis has identified the functional importance of hNatB-NAA20 residues His73, Asn116, and Tyr123, although Tyr123 is the only residue that is in position to play a catalytic role ( Figure 5A). Specifically, Tyr123 is in position to play a catalytic role, potentially as a general base and/or acid, through a bridging water molecule (although a water molecule is not visible at the current resolution). Interestingly, hNAA50 contains a tyrosine residue at the same position (hNAA50-Tyr124), although the mechanistic significance of this tyrosine residue has not yet been described (Liszczak et al., 2011). It would be of interest to determine if hNAA50-Tyr124 also plays an important catalytic role, similar to the corresponding hNAA20-Tyr123 of hNatB.
The biological importance of hNatB and its connection to various disease processes highlights it as an important target for probe and inhibitor development. Indeed, a recent study highlights hNatB as a therapeutic target for aSyn toxicity (Vinueza-Gavilanes et al., 2020). Our development of a CoA-aSyn conjugate bisubstrate with an IC 50 of~1.6 mM represents a step in this direction, although the structural information provided here could further aid to the rational development of more druglike hNatB inhibitors with possible therapeutic applications.

Protein expression and purification
hNAA20 with a C-terminal truncation (1-163 out of 178 residues) and full-length hNAA25 were cloned into two separate insect cell expression vectors pFASTBac HTA. hNAA20 was untagged, while hNAA25 contained a Tobacco-etch virus (TEV)-cleavable N-terminal 6xHis-tag. Human NatB complex (hNAA20 1-163 /hNAA25 FL ) was obtained by co-expressing these two plasmids in Sf9 (S. frugiperda) cells (ThermoFisher, cat# 12659017), and purified as described previously (Gottlieb and Marmorstein, 2018). Sf9 cells were grown to a density of 1 Â 10 6 cells/mL and infected using the amplified hNAA20 1-163 /hNAA25 FL baculovirus to an MOI (multiplicity of infection) of 1-2. The cells were grown at 27˚C and harvested for 48 hr post-infection by centrifugation. Cell pellets were resuspended in lysis buffer (25 mM Tris, pH 8.0, 300 mM NaCl, 10 mM Imidazole, 10 mM b-ME, 0.1 mg/ mL PMSF, DNase, and complete, EDTA-free protease inhibitor tablet) and lysed by sonication. After centrifugation, the supernatant was isolated and passed over Ni-NTA resin (Thermo Scientific), which was subsequently washed with 10 column volumes of lysis buffer. Protein was eluted with a buffer with 25 mM Tris, pH 8.0, 300 mM imidazole, 200 mM NaCl, 10 mM b-ME, which was dialyzed into a buffer with 25 mM HEPES pH 7.5 50 mM NaCl 10 mM b-ME. Ion-exchange was carried out with an SP ion-exchange column (GE Healthcare) in dialysis buffer with a salt gradient (50-750 mM NaCl). Peak fractions were concentrated to~0.5 mL with a 50 kDa concentrator (Amicon Ultra, Millipore), and loaded onto an S200 gel-filtration column (GE Healthcare) in a buffer with 25 mM HEPES, pH 7.5, 200 mM NaCl, and 1 mM TCEP. Proteins were aliquoted, snap-frozen in liquid nitrogen, and stored at À80˚C for further use. Protein harboring mutations were generated with the QuickChange protocol (Stratagene) and obtained following the same expression and purification protocol as described for the wild-type protein. Primers synthesized for the generation mutant constructs are listed in Supplementary file 1.

Acetyltransferase activity assays
All acetyltransferase assays were carried out at room temperature in a reaction buffer containing 75 mM HEPES, pH 7.5, 120 mM NaCl, 1 mM DTT as described Deng et al., 2019

Cryo-EM data collection
For initial sample screening, 0.6 mg/mL fresh hNatB sample with three-molar excess bisubstrate was used. hNatB particles on these grids exhibited a severe preferred orientation, which generated an incorrect 3D initial model (data not shown). To solve this issue, 1 mL of 0.05% NP-40 was mixed with 20 mL of hNatB (4 mg/mL). 3 mL of this sample was applied to glow-discharged Quantinfoil R1.2/1.3 holey carbon support grids, blotted and plunged into liquid ethane, using an FEI Vitrobot Mark IV. An FEI TF20 was used for screening the grids and data collection was performed with a Titan Krios equipped with a K3 Summit direct detector (Gatan), at a magnification of 105,000Â, with defocus values from À0.1 to À2.0 mm. Each stack was exposed in super-resolution mode with a total dose of 45 e -/Å 2 , resulting in 35 frames per stack. Image stacks were automatically collected with Latitude software (Gatan, Inc).

Cryo-EM data processing
Original image stacks were summed and corrected for drift and beam-induced motion at the micrograph level using MotionCor2 (Zheng et al., 2017), and binned twofold, resulting in a pixel size of 0.83 Å /pixel. Defocus estimation and the resolution range of each micrograph were performed with Gctf (Zhang, 2016). About 3000 particles were manually picked to generate several rough 2D class averages. Representative 2D classes were used to automatically pick~1,927,673 particles from 5281 micrographs in Relion 3.0 (Kimanius et al., 2016;Zivanov et al., 2018). All particles were extracted and binned to accelerate the 2D and 3D classification. After bad particles were removed by 2D and 3D classification, 982, 420 particles were used for auto-refinement and per-particle CTF refinement. After refinement, a mask was created in Relion with an initial binarization threshold of 0.005, covering the protein complex and extending the binary map and soft-edge by 12 pixels. The map was sharpened with the created mask by estimating B-factor automatically in Relion. The final map was refined to an overall resolution of 3.46 Å , with local resolution estimated by Resmap (Kucukelbir et al., 2014). We attempted particle polishing on this data set but this surprisingly resulted in artifactual density in the resulting map. We believe that this was due to some small defects in the K3 camera during data collection, which corrupted the particle polishing process. We, therefore, did not perform particle polishing on this data set. Raw micrographs were deposited in EMPIAR with access ID of EMPIAR-10477.
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.