The PtMAC2 fragment (NCBI accession No., LC768982; 1.7 kb) was amplified via PCR using the cDNA of P. tsukubaensis NBRC1940 as a template and the following set of oligonucleotide primers: 5′-CGC​GCG​GCA​GCC​ATA​TGC​TAG​GAG​ATC​AAG​TTT​GGA​AGG​AG-3′ (forward) and 5′-GTT​AGC​AGC​CGG​ATC​CTC​GAG​CTA​AAG​CTT​GGC​CTC​AGG​AG-3′ (reverse). We inserted the 1.7-kb PtMAC2 fragment into NdeI- and XhoI-digested pET15b using an In-Fusion Cloning Kit (TaKaRa Bio, Shiga, Japan) according to the manufacturer’s instructions, which yielded pET-NBRCMAC2. The ligated gene fragments were verified by DNA sequencing. The plasmid pET-NBRCMAC2_H158A for mutant PtMAC2p (H158A) expression was constructed by PCR using pET-NBRCMAC2 as a template, KOD one (Toyobo, Osaka, Japan) as polymerase, and the following set of oligonucleotide primers:5′-GGTCATGAATGCTTCTGCTACGCTAAATGGTCACCGCATG-3’ (forward) and 5′-CAT​GCG​GTG​ACC​ATT​TAG​CGTAGCAGA​AGC​ATT​CAT​GAC​C-3’ (reverse), including a mutation site (under lines). PCR conditions were as follows: 98°C for 10 s, 54°C for 5 s, and 68°C for 30 s with a total 30 cycles. The PCR mixture was then treated by DpnI at 37°C for 1 h. Subsequently, the PCR product was transformed into DH5α competent cells (Toyobo). The plasmid pET-NBRCMAC2_G162D for mutant PtMAC2p (G162D) expression was also constructed by PCR using pET-NBRCMAC2 as a template, KOD one (Toyobo, Osaka, Japan) as polymerase, and the following set of oligonucleotide primers:5′-CTAAATGATCACCGCATGCTCTTCCAAGGTTC-3’ (forward) and 5′-GCG​GTGATCATT​TAG​CGT​ATG​AGA​AGC​ATT​CAT​GAC​CA-3’ (reverse), including a mutation site (under lines). PCR conditions were as follows: 98°C for 10 s, 54°C for 5 s, and 68°C for 30 s with a total 30 cycles. The PCR mixture was then ligated using an In-Fusion Cloning Kit (TaKaRa Bio) according to the manufacturer’s instructions. The resultant plasmids pET-NBRCMAC2_H158A and pET-NBRCMAC2_G162D were amplified in DH5α cells and extracted using QIAprep Miniprep kit (QIAGEN, Venlo, Netherlands) according to the manufacturer’s instructions. The mutation was verified by DNA sequencing.

Escherichia coli BL21(DE3) cells harboring pET-NBRCMAC2, pET-NBRCMAC2_H158A, and pET-NBRCMAC2_G162D were cultured respectively at 37°C in 1 L of LB medium (Nacalai Tesque, Kyoto, Japan) supplemented with 100 μg mL⁻¹ sodium ampicillin in 3-L Erlenmeyer Flasks with baffles. When the turbidity (O.D. value) at 600 nm reached 0.4–0.6, the cells were cooled with ice water for 10 min and isopropyl d-thiogalactopyranoside (IPTG) was added to the culture to a final concentration of 0.1 mM. The cells were then cultured at 18 °C for 20 h. Subsequently, the cells were collected by centrifugation at 6,000 × g for 10 min at 4°C and washed with phosphate-buffered saline (PBS), which is composed of 137 mM NaCl, 8.1 mM Na₂HPO₄, 2.68 mM KCl, and 14.7 mM KH₂PO₄, at pH 7.2. The cells were then resuspended in buffer A (PBS supplemented with 150 mM NaCl). The cells were ultrasonically disrupted in ice water using an ultrasonic disruptor UD-211 (Tomy, Tokyo, Japan). The supernatant was obtained by centrifugation at 30,000 × g for 30 min at 4°C.

PtMAC2p purification from the supernatants was achieved through a 0.22-μm polyethersulfone membrane. The filtered samples were applied to a HisTrap HP column (5 mL; Cytiva, Tokyo, Japan) that had been equilibrated with Buffer A. The HisTrap HP column was washed with 40 mL of buffer A supplemented with 10 mM imidazole and 40 mL of buffer A with 40 mM imidazole. PtMAC2p was eluted with 20 mL of buffer A containing 300 mM imidazole. The sample was desalted and equilibrated with a buffer containing 20 mM potassium phosphate (KPi) at pH 7.2 by ultrafiltration using Vivaspin 20-5K (Sartorius, Göttingen, Germany). Subsequently, the sample was applied to a HiTrap Q HP anion exchange column (5 mL; Cytiva) that had been equilibrated with the same buffer. The HiTrap Q HP column was washed with 20 mL of 20 mM KPi (pH 7.2) supplemented with 100 mM NaCl and PtMAC2p was eluted with 15 mL of 20 mM KPi (pH 7.2) with 200 mM NaCl. PtMAC2p H158A was purified in the same manner as the wild-type enzyme. The concentration of the purified PtMAC2p was determined by measuring the absorbance at 280 nm. The extinction coefficient at 280 nm was calculated based on the amino acid sequence 58,330 M^–1 cm^–1 (A ₂₈₀ was 0.964 when 1 mg mL^–1 of PtMAC2p was in solution).

PtMAC2p activity was assayed using the 5,5-dithio-bis-(2-nitrobenzoic acid) reagent (Ellman’s reagent). The assay mixture containing 50 mM sodium phosphate buffer (pH 7.2), 0.1 mM lauroyl-CoA, 1 mM mono-acylated MEL-D produced from P. tsukubaensis strain ΔPtMAC2 (Saika et al., 2018), 1 mM DTNB, and PtMAC2p (0.2 µg of wild-type enzyme, 2 µg of H158A, or 7 µg of G162D) was incubated at 25°C. The increase in the reaction product was measured by the increase of absorbance at 412 nm of 2-nitro-5-thiobenzoic acid, the reduced form of DTNB. The product concentration was calculated using a molar absorption coefficient at 412 nm as 13,600 M^–1 cm^–1. All assays were performed in triplicate.

For crystallization, PtMAC2p was concentrated to 5 mg mL^–1 by ultrafiltration using a Vivaspin 20-5K. PtMAC2p crystals were grown at 15°C using hanging drop vapor diffusion. The purified PtMAC2p solution was mixed with equal volume of reservoir solution 1, containing 19% polyethylene glycol (PEG) 3350 and 200 mM lithium acetate. Also, a PtMAC2p solution supplemented with 5% glycerol was mixed with equal volume of reservoir solution 2 containing 21% PEG 3350 and 300 mM lithium chloride. Crystals were grown at 15°C for a week. The crystals prepared from reservoir solutions 1 and 2 were designated as types A and B, respectively. Native crystals types A and B were soaked in the reservoir solutions supplemented with 25% and 20% glycerol as cryoprotectants, respectively, and then flash cooled in liquid nitrogen. For phasing, type A crystals were soaked in reservoir solution 1 supplemented with 1 M potassium iodide (KI) for a few seconds and then flash-cooled in liquid nitrogen.

All diffraction data were collected at the BL44XU station at SPring-8 (Hyogo, Japan). Native datasets were collected at λ = 0.9 Å, while an anomalous dataset of KI-derivative crystal was collected at λ = 1.6 Å. All the datasets were processed and scaled using XDS version 10 Jan 2022 (Kabsch, 2010). The initial phase was solved by single anomalous dispersion (SAD) using the KI-derivative dataset with Autosol from Phenix 1.20.1 (Terwilliger et al., 2009; Liebschner et al., 2019). Subsequently, the model was manually completed using Coot (Emsley et al., 2010) and refined using Phenix.refine (Afonine et al., 2012). The obtained model was used as a search model for the native datasets. The phases of the native crystals were solved by molecular replacement using the structure of KI-derivative crystal as a search model with Phaser (McCoy et al., 2007). The coordinates were refined using Phenix.refine. After each refinement cycle, the models were manually adjusted using Coot software. Structural models were generated using Pymol 2.5.0 (Schrödinger, LLC). The quality of the refined model was verified using MolProbity 4.5.2 (Williams et al., 2017). The root mean square deviations (RMSD) value of Cα-atoms between structures in crystals types A and B was calculated using LSQKAB (Kabsch, 1976). The volume of the protein cavity was calculated by CASTp (Tian et al., 2018). The atomic coordinates and structure factors were deposited in the Protein Data Bank (PDB) under accession codes 8JOR (type A crystal) and 8JOS (type B crystal).

Recombinant PtMAC2p with a polyhistidine tag at the N-terminus was expressed in E. coli and was homogeneously purified (Supplementary Figure S1). Purified PtMAC2p catalyzes the transfer of a fatty acid of lauroyl-CoA (C12) to mono-acylated MEL-D in vitro. Two types of PtMAC2p crystals (A and B) were obtained using the purified enzyme (Supplementary Figure S2). Both crystals belong to the orthorhombic space group P2₁2₁2₁, while their b axes differ by about 10 Å (Table 1).

The initial structure of PtMAC2p was determined by the SAD-phase method using a type A KI-derivative crystal (Table 1; Supplementary Figure S3). Native crystals in types A and B were diffracted to 1.45 and 1.59 Å resolutions, respectively (Table 1). PtMAC2p is a monomer and forms α/β-folds which are composed of 18 α-helices (α1–α18) and 24 β-strands (β1–β24) (Figures 2, 3). In the type A crystal, residues Met1 to Ser539 and residues numbered −7 to 0 (GLVPRGSH), which are from the DNA sequence included in the pET15b vector, were modeled using a 2Fo-Fc map at 1.0 σ (Figure 3). In contrast, Met1 to Leu549, Leu549 is the C-terminal residue of PtMAC2p, were modeled in crystal type B (Figure 2A; Figure 3). Residues 540–549 can be modeled in crystal type B because these residues interacted with a symmetric molecule in the crystal owing to crystal packing, although these residues in type A were disordered, suggesting that the C-terminus of PtMAC2p is a flexible region (the yellow region in Figure 2A). The RMSD value of 539 Cα atoms (residues 1–539) between the two PtMAC2p models was 0.643 Å, indicating that the overall structures were almost identical. PtMAC2p possesses a cleft at the center of the molecule (indicated by the arrow in Figure 2A), suggesting that this cleft is a putative active site.

Proteins structurally homologous to PtMAC2p were searched using the DALI server (Holm, 2020). This search revealed that trichothecene 15-O-acetyltransferase (TRI3), which belongs to the BAHD acyltransferase superfamily, showed the highest similarity to PtMAC2p with a Z-score of 29.4, despite the amino acid-sequence identity between PtMAC2p and TRI3 being low (<15%) (Table 2) (Garvey et al., 2009). Other homologous proteins had relatively low Z-scores (<22.7).

To investigate the active site of PtMAC2p, we tried to obtain the PtMAC2 crystal with substrates, but could not. Thus, we compared its structure with a BAHD acyltransferases, TRI3, which has the highest structural similarity to PtMAC2p. The PtMAC2p model was also compared with trichothecene 3-O-acetyltransferase (TRI101) because the crystal structure of TRI101 in complex with an acyl donor (deoxynivalenol) and an acyl-acceptor analog (CoA) has been determined (Garvey et al., 2008) (Figure 4; Supplementary Figure S4).

Structural comparison of the three enzymes indicated that the large cleft found in PtMAC2p is the binding site for an acyl acceptor (the width of the cleft entrance, 12–18 Å; the volume of the cavity, 1533 Å³; solid arrows in Figure 2A; Figure 4A). The acyl acceptor binding site is closed in TRI3 (the volume of the cavity, 372 Å³; solid arrows in Figure 4B) or remarkably narrow in TRI101 (the width of the cleft entrance, 5.6–7.2 Å; the volume of the cavity, 549 Å³; solid arrows in Figure 4C), suggesting that PtMAC2p has a uniquely large and open cleft that allows the binding of large substrates and products with fatty acids (e.g., a substrate mono-acylated MEL and a product di-acylated MEL). In contrast, a structural comparison of PtMAC2p and TRI101 showed that the CoA residue of acyl CoA (acyl donor) binds to a site different from the cleft (dashed arrows in Figure 4A). This acyl donor binding site and the cleft (acyl acceptor binding site) are connected, resulting in a catalytic tunnel structure (Figure 5).

TRI3 and TRI101 contain the HXXXD motif, which is essential for the catalytic activity of enzymes in the BAHD acyltransferase superfamily (Molina and Kosma, 2014). The His residue in the motif is a critical catalytic residue responsible for the deprotonation of the acyl acceptor substrate, creating a nucleophile that attacks the carbonyl carbon of the acyl CoA substrate, and resulting in the release of CoASH and the formation of ester products (Suzuki et al., 2003; Lallemand et al., 2012; Walker et al., 2013). Notably, PtMAC2p possesses His158 at the same location as His in the HXXXD motifs of TRI3 and TRI101 (Figures 6A, B). We then evaluated the sequence conservation of the motif in Mac2p homologs, which were obtained from the protein BLAST, using the amino acid sequence of PtMAC2p. Among the Mac2p homologs, His was completely conserved in the motif, but Asp was not (Figure 7). Therefore, HXXXD is incompletely conserved in PtMAC2p as His158-Thr159-Leu160-Asn161-Gly162. This strongly suggests that His158 is the catalytic residue of PtMAC2p, although the Asp residue in the motif is replaced by Gly162.

To further investigate the role of His158 in the catalytic activity, site-directed mutagenesis was performed. Substitution of His158 in PtMAC2p with Ala (H158A) resulted in more than 1000-fold decrease in catalytic activity of PtMAC2p (from 152 ± 3 μmol min^–1 mg^–1 in wild-type to 0.0864 ± 0.0091 μmol min^–1 mg^–1 in H158A mutant). This result reveals that His158 is an important catalytic residue in PtMAC2p and corresponds to the catalytic His in BAHD acyltransferases, such as TRI3 and TRI101.

We further explored structural features related to substrate recognition. In the type A crystal, electron density map of a PEG molecule in the reservoir solution for crystallization is observed. The PEG is bound near His158 in the cleft of PtMAC2p (Figure 5; Figure 8A). Hydrophobic amino acid residues, mainly from α2 (Val22, Met25, Ile26, and Ala29) and α11 (Phe234, Leu237, and Ile241) helices, are concentrated around PEG, although a hydrophilic residue, Arg240, is also located near PEG. These residues form a hydrophobic surface in the cleft (Figure 8B). These findings suggest that the PEG mimics the binding position of the acyl group of mono-acylated MEL and/or acyl CoA and that this hydrophobic cleft may accommodate one or two acyl groups.

In contrast, a CoA residue is likely recognized on the other side of PtMAC2p (dashed arrows in Figure 4A). The putative CoA-binding site was compared to that of TRI101 (Figures 6C, D). Although the structure and amino acid residues of the CoA-binding site of TRI101 are not well conserved in PtMAC2p, the residues involved in the recognition of the CoA phosphate groups appear to be partly common. The phosphate groups of CoA interact with basic residues in TRI101, such as Lys245, Arg335, and Lys382 (Figure 6D). In PtMAC2p, His291, Lys295, Arg340, and Lys391 are candidates for electrostatic interactions with phosphate groups. In contrast, there is no residue in PtMAC2p corresponding to Phe258 in TRI101, which interacts with an adenine moiety via stacking. Therefore, this region in PtMAC2p is likely the CoA binding site; however, further studies are needed to identify the correct binding site for CoA.