Crystal structure of barley agmatine coumaroyltransferase, an N-acyltransferase from the BAHD superfamily

The structure of barley agmatine coumaroyltransferase, a member of the BAHD acyltransferase superfamily, was elucidated. This is the first report of the structure of an N-acyltransferase from the BAHD superfamily.

Abstract

The enzymes of the BAHD superfamily, a large group of acyl-CoA-dependent acyltransferases in plants, are involved in the biosynthesis of diverse secondary metabolites. While the structures of several O-acyltransferases from the BAHD superfamily, such as hydroxycinnamoyl-CoA shikimate hydroxycinnamoyl transferase, have been elucidated, no structural information on N-acyltransferases is available. Hordeum vulgare agmatine coumaroyltransferase (HvACT) is an N-acyltransferase from the BAHD superfamily and is one of the most important enzymes in the secondary metabolism of barley. Here, an apo-form structure of HvACT is reported as the first structure of an N-acyltransferase from the BAHD superfamily. HvACT crystals diffracted to 1.8 Å resolution and belonged to the monoclinic space group P2₁, with unit-cell parameters a = 57.6, b = 59.5, c = 73.6 Å, α = 90, β = 91.3 , γ = 90°. Like other known BAHD superfamily structures, HvACT contains two domains that adopt a two-layer αβ-sandwich architecture and a solvent-exposed channel that penetrates the enzyme core.

1. Introduction  

Acyltransferases are present in a wide range of organisms from bacteria to mammals. In addition to crucial biological functions in plants, such as control of gene expression through histone acetylation and regulation of fatty-acid metabolism, acyltransferases play important roles in the biosynthesis of secondary metabolites such as phenolic compounds. The diverse biological functions of phenolic compounds are derived from their structural diversity, which is partly conferred by acylation steps during their biosynthesis. Plant acyltransferases acting on phenolic acyl donors are categorized into two families: BAHD and SCPL acyltransferases (Bontpart et al., 2015 ▸). The most prominent functional difference between the BAHD (the acronym is based on the first four characterized enzymes) and SCPL (serine carboxy­peptidase-like) families is the type of activated acyl groups that serve as their acyl donors; the SCPL family uses glycose esters, whereas the BAHD family uses acyl-CoA thioesters.

The BAHD superfamily is composed of acyltransferases that catalyse condensation between acyl-CoA (an acyl donor) and diverse alcohols and amines (acyl acceptors) to yield esters and amides. The BAHD superfamily was first organized into five phylogenetic clades (I–V) based on the sequences of 46 biochemically or genetically characterized acyltransferases (D’Auria, 2006 ▸) and was later expanded into eight clades: clades I, III and V were each subdivided into two (for example, clades Ia and Ib; Tuominen et al., 2011 ▸). BAHD acyltransferases are monomeric enzymes of ∼50 kDa. Two motifs are conserved across all of the clades: the HXXXD and DFGWG motifs (St-Pierre & De Luca, 2000 ▸; D’Auria, 2006 ▸). The histidine residue in the former motif is considered to be the catalytic centre that acts as a base in the deprotonation of the nucleophile (acyl acceptor) to facilitate attack on the carbonyl group of the acyl-CoA (Ma et al., 2005 ▸). Although the latter motif was also assumed to participate in catalysis, some crystal structure analyses suggested that it was instead involved in maintaining the structure of the enzyme (Ma et al., 2005 ▸; Walker et al., 2013 ▸). Overall, the structures of the acyl acceptors of the BAHD superfamily are much more diverse than those of the acyl donors, indicating that the wide range of structural diversity of BAHD acyltransferases reflects the diversity of the acyl acceptors. While numerous BAHD family members have been identified from a large number of plant species to date, the tertiary structures of only a few more than ten enzymes are currently available, about half of which are structures of hydroxycinnamoyl-CoA:shikimate hydroxy­cinnamoyl transferases (HCTs; EC 2.3.1.133; Ma et al., 2005 ▸; Unno et al., 2007 ▸; Lallemand et al., 2012 ▸; Walker et al., 2013 ▸; Levsh et al., 2016 ▸; Manjasetty et al., 2012 ▸; Chiang et al., 2018 ▸; Eudes et al., 2016 ▸). Therefore, limited information is available on the recognition mechanisms of the acyl acceptors.

Barley (Hordeum vulgare) accumulates high concentrations of hydroxycinnamic acid amides (HCAAs). These HCAAs are formed through condensation between CoA esters of cinnamic acid derivatives and amines. The predominant HCAA species in barley are p-coumaroylagmatine and feruloylagmatine, and an agmatine coumaroyltransferase (ACT; EC 2.3.1.64) is responsible for their biosynthesis (Burhenne et al., 2003 ▸; Supplementary Fig. S1). Initially, H. vulgare ACT (HvACT) was the only member of clade IV of the BAHD superfamily (D’Auria, 2006 ▸); however, the size of this clade has continually expanded owing to the increasing availability of genome sequences. Nonetheless, it remains the one of the smallest clades in the BAHD superfamily and is basically monocot-specific (Tuominen et al., 2011 ▸; Peng et al., 2016 ▸). HvACT is a noteworthy enzyme for two reasons: it represents clade IV and catalyses the transfer of acyl groups onto N atoms (an N-acyltransferase). All of the crystal structures obtained to date in the BAHD family are, to our knowledge, those of O-acyltransferases, and the structure of an N-acyltransferase has not yet been reported. Therefore, in the present study, we crystallized HvACT and analysed its crystal structure to study the structural basis of substrate recognition and the catalytic mechanism of the N-acyltransferases of the BAHD superfamily. These results provide fundamental insights into the structural factors that contribute to the diversity of the substrate specificity of the BAHD superfamily enzymes.

2. Materials and methods  

2.1. Expression and purification of HvACT  

Information on the cloning of HvACT is shown in Table 1 ▸. The cDNA of HvACT (accession No. AB334132) was prepared from 72 h old barley (cv. Betzes) shoots using the polymerase chain reaction (PCR). After confirming the sequence, the coding sequence for HvACT was introduced into pET-28a to generate an N-terminally His-tagged protein via In-Fusion cloning (Clontech, California, USA) followed by transformation of Escherichia coli B834(DE3)pLysS cells.

Table 1. Information on macromolecule production.

Source organism	H. vulgare cv. Betzes
DNA source	cDNA
Forward prime for isolation	5′-TACGCACGTTCGCCGTCGACA-3′
Reverse primer for isolation	5′-AACGGGACCTAGTCGAGGCT-3′
Forward prime for subcloning†	5′-CAGCAAATGGGTCGCAAGATCACCGTGCACTCT-3′
Reverse primer for subcloning†	5′-CTCGAATTCGGATCCCTAGTCGAGGCTGTAGCAGCA-3′
Cloning vector	pBluescript
Expression vector	pET-28a
Expression host	E. coli B834(DE3)pLysS
Complete amino-acid sequence of the construct produced‡	MGSSHHHHHHSSGLVPR GSHMASMTGGQQMGRKITVHSSKAVKPEYGACGLAPGCTADVVPLTVLDKANFDTYISVIYAFHAPAPPNAVLEAGLGRALVDYREWAGRLGVDASGGRAILLNDAGARFVEATADVALDSVMPLKPTSEVLSLHPSGDDGPEELMLIQVTRFACGSLVVGFTTQHIVSDGRSTGNFFVAWSQATRGAAIDPVPVHDRASFFHPREPLHVEYEHRGVEFKPCEKAHDVVCGADGDEDEVVVNKVHFSREFISKLKAHASAGAPRPCSTLQCVVAHLWRSMTMARGLDGGETTSVAIAVDGRARMSPQVPDGYTGNVILWARPTTTAGELVTRPVKHAVELISREVARINDGYFKSFIDFANSGAVEKERLVATADAADMVLSPNIEVDSWLRIPFYDMDFGGGRPFFFMPSYLPVEGLLILLPSFLGDGSVDAYVPLFSRDMNTFKNCCYSLD

^†Common sequences with the vector are underlined.

^‡The extra amino acids originating from the cloning vector are underlined. Amino acids removed by thrombin are shown in italics.

Luria–Bertani (LB) broth (1.2 l) supplemented with antibiotics (34 µg ml⁻¹ chloramphenicol and 50 µg ml⁻¹ kanamycin) was inoculated with a colony of E. coli B834(DE3)pLysS cells harbouring HvACT and cultured at 310 K to an OD₆₀₀ of 0.7. Recombinant protein expression was induced by adding isopropyl β-d-1-thiogalactopyranoside to a final concentration of 1 mM, followed by further culture for 24 h at 293 K. The cells were harvested by centrifugation at 5000g for 8 min and resuspended in 40 ml 50 mM HEPES–NaOH pH 7.5. After cell disruption by ultrasonication (20 s, 15 cycles), the lysate was centrifuged twice at 15 000 rev min⁻¹ for 15 min to remove the cell debris. The expressed protein was bound to 800 µl TALON metal-affinity resin (Clontech) pre-equilibrated with 50 mM HEPES pH 7.5 containing 200 mM NaCl. After washing the resin with the same buffer, HvACT was eluted using 50 mM HEPES pH 7.5, 200 mM NaCl, 200 mM imidazole. To remove the His tag, 2 U of thrombin per milligram of protein was added and the mixture was incubated overnight at 277 K. After concentration by ultrafiltration using Amicon Ultra-4 10K centrifugal filter devices (Merck Millipore, Germany), the solution was subjected to gel filtration on Superdex 200 Increase 10/300 GL (GE Healthcare, USA) with a buffer consisting of 50 mM HEPES pH 7.2, 5 mM dithiothreitol (DTT), 150 mM NaCl at a flow rate of 0.5 ml min⁻¹. The enzyme was further purified via anion-exchange chromatography using a Mono Q 5/50 GL column (GE Healthcare). The protein was applied to a column equilibrated with 50 mM HEPES pH 7.2, 5 mM DTT and then eluted by increasing the NaCl concentration linearly (0–100 mM in 3 ml followed by 100–300 mM in 20 ml) at a flow rate of 1.0 ml min⁻¹. Fractions corresponding to the largest peak in absorbance at 280 nm were collected and concentrated to ∼10 mg ml⁻¹ via ultrafiltration using Amicon Ultra-0.5 10K centrifugal filter devices (Merck Millipore). All of the purification procedures were performed at 277 K. The protein concentration was determined according to the method of Bradford (1976 ▸) using BSA as the standard.

2.2. Crystallization, data collection and structure determination  

The crystallization conditions for HvACT were determined via the hanging-drop vapour-diffusion method with screening using commercially available kits from Hampton Research and Rigaku Reagents. The protein was subjected to crystallization immediately after purification; finally, plate-shaped crystals of ∼140 × 110 × 10 µm in size were obtained after incubation at 293 K for 1–2 days. The crystallization conditions are presented in Table 2 ▸. The crystals were flash-cooled with liquid nitrogen in nylon loops without cryoprotectant solution, and diffraction data were collected on beamline BL-5A equipped with a PILATUS3 S6M detector at Photon Factory (PF), Japan. Indexing, integration, scaling and merging were performed by a program pipeline at the beamline using the XDS (Kabsch, 2010 ▸), POINTLESS (Evans, 2006 ▸) and AIMLESS (Evans & Murshudov, 2013 ▸) software. The initial model was obtained by molecular replacement using MoRDa (http://www.biomexsolutions.co.uk/morda), which automatically used PDB entry 6dd2 (HCT from Selaginella moellendorffii; Chiang et al., 2018 ▸) and part of PDB entry 2hnu (bovine oxytocin-related neurophysin from Bos taurus; residues 1–14 and 214–223; Li et al., 2007 ▸) as search models. Further model building was performed using ARP/wARP (Langer et al., 2008 ▸) followed by REFMAC5 (Murshudov et al., 2011 ▸) and Coot (Emsley et al., 2010 ▸). The atomic coordinates and structure factors have been deposited in the Protein Data Bank under accession code 7cys.

Table 2. Crystallization conditions for diffracting crystals.

Method	Hanging-drop vapour diffusion
Temperature (K)	293
Protein concentration (mg ml−1)	7.7
Buffer composition of protein solution	50 mM HEPES pH 6.5, 5 mM DTT, 100 mM or 50 mM NaCl
Composition of reservoir solution	0.08 M sodium cacodylate pH 6.5, 0.16 M magnesium acetate, 16%(v/v) polyethylene glycol 8000, 20%(v/v) glycerol
Volume and ratio of drop	2 µl, 1:1 protein:reservoir solution
Volume of reservoir (µl)	150

2.3. Generation of the H152A mutant and evaluation of the enzyme activity  

To construct the HvACT H152A mutant, inverse PCR was performed using pET-28a containing the HvACT gene as a template with the following primers: forward, 5′-CAGGCTATCGTGTCCGACGGC-3′; reverse, 5′-CGTGGTGAACCCCACGACGAG-3′. After self-ligation and sequence confirmation, the enzyme was expressed and purified as described in Section 2.1, except that the E. coli BL21(DE3)pLysS strain was used for protein expression. To evaluate the activity of the mutant, p-coumaroyl-CoA was used as the acyl donor and agmatine was used as the acyl acceptor. p-Coumaroyl-CoA was synthesized enzymatically using recombinant tobacco 4-coumarate:coenzyme A ligase as described by Beuerle & Pichersky (2002 ▸). Enzyme activity was assessed in 1 ml 50 mM Tris–HCl pH 7.5 at room temperature. p-Coumaroyl-CoA (1.0 µl), HvACT (1.0–5.0 µl) and 20 mM agmatine (10.0 µl) were added in sequence. The concentrations of the acyl donor and enzyme were varied as appropriate. HvACT activity was evaluated spectrophotometrically by monitoring the decrease in absorbance at 333 nm.

3. Results and discussion  

For crystallization, HvACT was expressed in E. coli as an N-terminally His-tagged protein. While gel filtration following affinity chromatography yielded an apparently homogenous enzyme on an SDS–PAGE gel, further purification by anion-exchange chromatography on Mono Q showed multiple minor peaks (data not shown). The purification step, as well as the removal of the N-terminal His tag, were essential for crystallization. Crystals suitable for data collection were obtained from a condition in which cacodylate was used as the buffer (0.08 M cacodylate pH 6.5). The structure of apo-form HvACT was determined at 1.81 Å resolution; data-collection and refinement statistics are summarized in Table 3 ▸. The HvACT crystal belonged to the monoclinic space group P2₁, with unit-cell parameters a = 57.6, b = 59.5, c = 73.6 Å, α = 90, β = 91.3, γ = 90°; the Matthews coefficient was calculated to be 2.49 ų Da⁻¹, which corresponded to one molecule per crystallographic asymmetric unit, with an estimated solvent content of 50.6%. The structures of the N-terminal region (the sequence derived from the expression vector), residues 211–214 and the three C-terminal residues could not be defined because of a lack of electron density. During refinement, blobs of electron density were observed at the tips of cysteine residues. Since the crystallization buffer contained cacodylate and dithiothreitol, we concluded that eight of the nine cysteines in HvACT were modified as S-(dimethylarsenic)cysteine, which was confirmed by the shape of the F _o − F _c OMIT map and strong anomalous difference densities at the positions of As atoms (Figs. 1 ▸ a and 1 ▸ b). SDS–PAGE analysis under reducing and nonreducing conditions showed that a reducing agent did not affect the electrophoretic profiles (Fig. 2 ▸ a). Furthermore, there were no significant differences between the chromatograms of gel filtration with and without DTT (Fig. 2 ▸ b). These results indicate that disulfide bonds are not easily formed during purification and the short storage period.

Table 3. Data-collection and refinement statistics.

Values in parentheses are for the outer shell.

Data collection
Beamline	BL-5A, PF
Wavelength (Å)	1.000
Temperature (K)	95
Space group	P21
a, b, c (Å)	57.6, 59.5, 73.6
α. β. γ (°)	90, 91.3, 90
No. of molecules in asymmetric unit	1
Resolution range (Å)	50–1.81 (1.85–1.81)
No. of unique reflections	44594 (2201)
Multiplicity	3.1 (2.0)
Completeness (%)	97.8 (82.5)
〈I/σ(I)〉	16.3 (2.2)
R merge †	0.036 (0.237)
CC1/2	0.999 (0.914)
Refinement
Resolution range (Å)	46.3–1.81 (1.86–1.81)
Completeness (%)	97.8 (83.2)
No. of reflections	42403 (2686)
R work/R free	0.191/0.203 (0.243/0.239)
No. of non-H atoms
Protein	3312
Water	226
R.m.s. deviations
Bond lengths (Å)	0.004
Angles (Å)	1.354
Average B factors (Å2)
Protein	25.2
Water	26.4
Ramachandran plot‡ (%)
Favoured	98.0
Allowed	2.0
Disallowed	0.0

^† R _merge = , where I_i(hkl) is the intensity of an observation and 〈I(hkl)〉 is the mean value for the unique reflection; summations are over all reflections.

^‡Ramachandran plots were calculated by RAMPAGE (Lovell et al., 2003 ▸) in the CCP4 suite (Winn et al., 2011 ▸).

Figure 1.

(a) Anomalous difference Fourier maps contoured at the 5σ level are shown on the HvACT structure. The eight of the nine cysteines in the enzyme that are modified with dimethylarsenic are shown as stick models. The numbers are the residue numbers of the cysteine residues. (b) An F _o − F _c OMIT map calculated by excluding Cys24 is shown in blue mesh contoured at 2σ and a difference anomalous Fourier map contoured at 5σ is shown in red.

Figure 2.

(a) SDS–PAGE of purified HvACT under reducing and nonreducing conditions. (b) Chromatograms of gel filtration on Superdex 200 Increase 10/300 GL using buffer with (top) and without (bottom) 5 mM DTT. Protein elution was monitored by measuring the absorbance at 280 nm. The highest peaks correspond to HvACT.

HvACT contained 13 α-helices and 18 β-strands and could be divided into two domains (Figs. 3 ▸ a and 3 ▸ b), as is the case for other BAHD family structures. Domains I (the N-terminus to 172) and II (225 to the C-terminus) are connected by a long crossover (173–224), part of which (191–207) interacts with α11 and β14 in domain II. Both domains adopted a two-layer αβ-sandwich architecture (chloramphenicol acetyltransferase topology) according to the CATH classification (Orengo et al., 1997 ▸; https://www.cathdb.info/). The loop between β15 and β16 overhangs domain I, and β16 participates in the core β-sheet of the domain. The overall folding was similar to those of other known BAHD family structures and, using PyMOL, could be superimposed on those of acyltransferases from sorghum (SbHCT; PDB entry 4kec; Walker et al., 2013 ▸) and Arabidopsis (AtHCT; PDB entry 5kjs; Levsh et al., 2016 ▸), which are the BAHD family clade V members that possess the highest sequence similarity to HvACT (∼32%) among BAHD family enzymes of known structure, with r.m.s.d.s of 1.44 Å for SbHCT (304 C^α atoms) and 1.35 Å for AtHCT (295 C^α atoms) (Fig. 3 ▸ c). The most notable structural difference was detected in the loop between β9 and β10 (Cys208–Glu224), which is part of the inter-domain crossover. The presence of a solvent channel between the domains that penetrates the entire enzyme also resembles other BAHD family enzymes (Ma et al., 2005 ▸; Walker et al., 2013 ▸; Fig. 3 ▸ b). Despite the low inter-clade sequence identity, there are two highly conserved motifs across all of the clades. One of them is the HXXXD motif containing the catalytic centre histidine residue, which is located near the centre of the protein and faces the solvent channel. In HvACT, this motif corresponds to ¹⁵²HIVSD¹⁵⁶. As in the other BAHD family enzymes, His152 is located on the surface of the solvent channel, and the substrates can access the enzyme from both sides. In the case of vinorine synthase and SbHCT, it was proposed that each substrate enters from its own side (Ma et al., 2005 ▸; Walker et al., 2013 ▸), and this system is also likely to apply to HvACT. The K _m and V _max values of HvACT were 1.12 µM and 950 nmol s⁻¹ mg⁻¹, respectively, when p-coumaroyl-CoA and agmatine were used as substrates (Supplementary Fig. S2). A His152Ala substitution reduced the activity to 0.03% of that of the the wild type, indicating that His152 acts as the catalytic centre. The O atoms in the side chain of Asp156 were located 2.9 Å apart from the N atoms in the Arg287 side chain, indicating salt-bridge formation (Fig. 3 ▸ d). As the corresponding salt bridges in the vinorine synthase and SbHCT acyltransferases have been shown to contribute to structural maintenance (Ma et al., 2005 ▸; Walker et al., 2013 ▸), this interaction would possess a similar function in HvACT. The other conserved motif is DFGWG, which is located between β15 and β16 in HvACT and is known to be located away from the catalytic centre. It is unlikely to be directly involved in the catalytic reaction or substrate recognition in terms of position; however, it has been reported to be involved in maintaining activity, and an Asp-to-Ala substitution causes a significant decrease in activity (Ma et al., 2005 ▸; Walker et al., 2013 ▸; Suzuki et al., 2003 ▸). This motif is also located away from the catalytic centre in HvACT, although its function remains unknown.

Figure 3.

(a) Ribbon diagram of the global structure of apo HvACT. Domains I and II and the crossover part are coloured blue, brown and yellow, respectively. (b) Surface diagram of apo HvACT. Domains I and II and the crossover part are coloured blue, brown and yellow, respectively. The black triangle indicates the solvent channel. The catalytic centre His152 is shown in red. (c) Superimposition of HvACT (blue), SbHCT (PDB entry 4kec; green) and AtHCT (PDB entry 5kjs; purple). The structures are shown as an underside view compared with that in (a). (d) Close-up view of the HXXXD motif. The residues in the motifs and Arg287 are shown in stick format. O and N atoms are shown in red and blue, respectively. The distance between the Asp156 and Arg287 side chains is indicated. These figures were prepared using UCSF Chimera (Pettersen et al., 2004 ▸).

In this study the structure of HvACT was elucidated, which is the first structure of an N-acyltransferase belonging to clade IV of the BAHD superfamily. The study provides valuable structural information on this class of enzymes and a basis for further comparison with O-acyltransferases, with the aim of a better understanding of the mechanisms of substrate recognition and catalysis in this superfamily. Previous studies have demonstrated the permissiveness of HvACT to acyl-group donors and its rigorous specificity for the amine acyl acceptor agmatine (Bird & Smith, 1983 ▸; Burhenne et al., 2003 ▸; Nomura et al., 2018 ▸). As the structure obtained in this study is of the apo form, we are currently investigating the enzyme complexed with ligands in order to explore the mechanism of its strict substrate specificity towards agmatine.

Supplementary Material

PDB reference: barley agmatine coumaroyltransferase, 7cys