Molecular cloning, heterologous expression, and enzymatic characterization of lysoplasmalogen‐specific phospholipase D from Thermocrispum sp.

Lysoplasmalogen (LyPls)‐specific phospholipase D (LyPls‐PLD) is an enzyme that catalyses the hydrolytic cleavage of the phosphoester bond of LyPls, releasing ethanolamine or choline, and 1‐(1‐alkenyl)‐sn‐glycero‐3‐phosphate (lysoplasmenic acid). Little is known about LyPls‐PLD and metabolic pathways of plasmalogen (Pls). Reportedly, Pls levels in human serum/plasma correlate with several diseases such as Alzheimer's disease and arteriosclerosis as well as a variety of biological processes including apoptosis and cell signaling. We identified a LyPls‐PLD from Thermocrispum sp. strain RD004668, and the enzyme was purified, characterized, cloned, and expressed using pET24a(+)/Escherichia coli with a His tag. The enzyme's preferred substrate was choline LyPls (LyPlsCho), with only modest activity toward ethanolamine LyPls. Under optimum conditions (pH 8.0 and 50 °C), steady‐state kinetic analysis for LyPlsCho yielded K m and k cat values of 13.2 μm and 70.6 s−1, respectively. The ORF of LyPls‐PLD gene consisted of 1005 bp coding a 334‐amino‐acid (aa) protein. The deduced aa sequence of LyPls‐PLD showed high similarity to those of glycerophosphodiester phosphodiesterases (GDPDs); however, the substrate specificity differed completely from those of GDPDs and general phospholipase Ds (PLDs). Structural homology modeling showed that two putative catalytic residues (His46, His88) of LyPls‐PLD were highly conserved to GDPDs. Mutational and kinetic analyses suggested that Ala55, Asn56, and Phe211 in the active site of LyPls‐PLD may participate in the substrate recognition. These findings will help to elucidate differences among LyPls‐PLD, PLD, and GDPD with regard to function, substrate recognition mechanism, and biochemical roles. Data Accessibility Thermocrispum sp. strain RD004668 and its 16S rDNA sequence were deposited in the NITE Patent Microorganisms Depositary (NPMD; Chiba, Japan) as NITE BP‐01628 and in the DDBJ database under the accession number AB873024. The nucleotide sequences of the 16S rDNA of strain RD004668 and the LyPls‐PLD gene were deposited in the DDBJ database under the accession numbers AB873024 and AB874601, respectively. Enzyme EC number EC 3.1.4.4


Data Accessibility
Thermocrispum sp. strain RD004668 and its 16S rDNA sequence were deposited in the NITE Patent Microorganisms Depositary (NPMD; Chiba, Japan) as NITE BP-01628 and in the DDBJ database under the accession number AB873024. The nucleotide sequences of the 16S rDNA of strain RD004668 and the LyPls-PLD gene were deposited in the DDBJ database under the accession numbers AB873024 and AB874601, respectively. Enzyme EC number EC 3. 1.4.4 Plasmalogens (Pls) are glycerophospholipids with a vinyl ether bond at the sn-1 position and an ester bond at the sn-2 position. Pls are broadly found in organisms from anaerobic bacteria to invertebrate and vertebrate animals [1]. Pls as well as common diacylglycerophospholipids (DAGPLs) are important components of cell membrane. In mammals, Pls constitute 4-32% of the total phospholipid mass, and are particularly enriched in neural tissue, heart, lung, and circulating immune cells [2,3]. Although the functions of Pls have not been fully elucidated, choline Pls (PlsCho) and ethanolamine Pls (PlsEtn) levels in human serum and plasma correlate with a variety of biological processes including apoptosis, cell signaling [4], and various diseases [5][6][7]. Recently, PlsEtn and PlsCho have been implicated in several diseases such as Alzheimer's disease and arteriosclerosis [5], respectively. More recently, Maeba et al. reported that PlsChos, particularly those containing oleic acid (18 : 1) in the sn-2 position, were strongly associated with a wide range of risk factors for metabolic syndrome and atherosclerosis and may therefore be useful biomarkers [7]. As interest grows in the development of diagnostic reagent kits for early stage diseases, the Pls-specific enzymes are receiving increasing attention.
There are several reports on enzymes involved in Pls metabolism. In mammals, LyPls are formed from Pls by the action of phospholipase A 2 (PLA 2 ), which cleaves the sn-2 acyl bond of DAGPLs to release a free fatty acid (FFA) [3]. Recently, we also found that phospholipase A 1 (PLA 1 ), which cleaves the sn-1 acyl bond of DAGPLs to release FFA, of Streptomyces albidoflavus NA297 (SaPLA 1 ) can hydrolyze PlsCho and PlsEtn to yield choline lysoplasmalogen (LyPlsCho) and ethanolamine lysoplasmalogen (LyPlsEtn), respectively [8]. In mammals, it is important that removal of the acyl moiety from sn-2 of Pls by the action of PLA 2 is the only known pathway for generation of lysoplasmalogens (LyPls). In contrast with Pls, LyPls are amphiphilic and are normally maintained at very low levels in cell membranes [3]. Wu et al. reported lysoplasmalogenase (LyPlsase) identified as an integral membrane protein TMEM86B (EC 3.3.2.2 and EC 3.3.2.5) that catalyses the hydrolytic cleavage of the vinyl ether bond at the sn-1 position of LyPls, forming fatty aldehyde and sn-glycero-3-phosphoethanolamine (GPE) or sn-glycero-3phosphocholine (GPC) [1,9]. They also purified and characterized LyPlsase from Legionella pneumophila and cloned its gene [3]. Moreover, they described that LyPlsase (TMEM86B) is member of the larger YhhN family of proteins which are present in 138 species of eukaryotes and 1205 of bacteria. LyPls can be converted back into Pls by a transacylase [10]. LyPls can be further broken down by phospholipase C (PLC) and phospholipase D (PLD) [11][12][13]. Wolf and Gross reported that PLC in the cytosolic fraction of canine myocardium can hydrolyze Pls, releasing the head groups, phosphoethanolamine or phosphocholine [14]. Wykle and Schremmer reported that lysophospholipase D in brain microsomes can hydrolyze LyPls or 1-O-1 0 -alkyl-2hydroxylglycerophospholipid, releasing ethanolamine (Etn) or choline (Cho) [11]; however, this enzyme has not been purified and its gene has not been identified. We recently found that PLD (PLD 684 ) from Streptomyces sp. NA684 can hydrolyze PlsCho in the presence of 0.1-0.2% (w/v) Triton X-100 [15]. To the best of our knowledge, however, no other information is available concerning LyPls-or Pls-specific PLD. Also, little is known about Pls metabolic pathways. Understandably, their enzymes also remain to be characterized.
In the present study, we describe a novel enzyme, LyPls-specific PLD (LyPLs-PLD), capable of hydrolyzing LyPls to release the corresponding Cho or Etn (Fig. 1). Here, we report the purification, characterization, molecular cloning of LyPLs-PLD from Thermocrispum sp. strain RD004668, as well as its efficient heterologous production using Escherichia coli. The identification of LyPls-PLD reveals a new pathway for Pls metabolism, in particular, in microorganisms. The characterization of LyPls-PLD demonstrates differences between PLD, the phosphodiesterase family, and glycerophosphodiester phosphodiesterase (GDPD) [EC 3.1.4.46] with regard to function, substrate recognition mechanism, and biochemical roles.

Purification and characterization of the wild-type enzyme
The purified wild-type (WT) enzyme (30.7 lg protein) with specific activity of 50.6 UÁmg À1 protein was obtained from the 1.8-L culture supernatant (Table 1, Fig. S1A). Next-generation sequencing (NGS) revealed a predicted ORF of 67 013 base pairs (bp) ( Table S1). The peptide sequence of WT enzyme obtained with LC-MS was utilized as a probe, and the enzyme gene (lplspld) was identified in the predicted ORF database as 1005 bp encoding a 334-amino acid (aa)-long protein (Fig. 2). A possible ribosome-binding site (atga) was identified 40 nucleotides upstream of the start codon, gtg. Three possible terminator regions (ggccatggcg, ccggacgaggtccgg, and ggtccgggcc) were found 6-39 nucleotides downstream of the stop codon. Three possible promoter regions were found~10-40 nucleotides upstream of the start codon: À35, atgtct, tggtcc, and gtgcta; À10, taaagt and gatgat. Based on SignalP prediction, the N-terminal sequence of the active form of the enzyme starts at Thr28 of the deduced aa sequence, indicating that the preceding 27-aa residues represent a Sec signal peptide sequence containing a signal peptidase recognition site (Ala-Xxx-Ala) that is required for secretion (Fig. 2). The molecular mass of the gene product (i.e., the 307-aa protein) without the signal sequence is calculated to be 33 452 Da in agreement with that of the purified enzyme estimated by SDS/PAGE. The isoelectric point (pI) of the mature enzyme was calculated as 5.55 using GENETYX-MAC version 16.0.8. SOSUI system analysis [16] indicated that LyPls-PLD may be a membrane protein with a single transmembrane helix ( 218 K to L 240 ) and an average hydrophobicity value (À0.227).

Expression and purification of rLyPls-PLD
High-efficiency production of the recombinant enzyme (rLyPls-PLD) was successfully achieved in E. coli BL21 (DE3) and Shuffle T7 cells transformed with pET24a/ lpls-pld mat_his as well as pET24a/lpls-pld mat. High specific activity (345 UÁmg À1 protein) and a large amount (~1 mg protein) of pure rLyPls-PLD were purified to electrophoretic homogeneity from 200 mL Shuffle T7 E. coli cells culture carrying pET24a/lpls-pld mat_his ( Fig. S1B and Table 2). Further high-efficiency extracellular production of rLyPls-PLD was successfully achieved in S. lividans cells transformed with an expression vector pUC702 [15] (data not shown).

Comparative sequence analysis of LyPls-PLD
A homology search performed using the BLAST algorithm demonstrated that the aa sequence of the mature LyPls-PLD shared 68% and 31% identity with GDPDs from Stackebrandtia nassauensis (snGDPD; UniProt accession no. D3Q1U5) [17] and Thermoanaerobacter tengcongensis (ttGDPD; UniProt accession no. Q8RB32) [18], and showed high similarity with other GDPDs as well (Fig. 6). However, most of them remain to be biochemically and functionally characterized. The deduced aa sequence of LyPls-PLD shared no similarity to those of PLDs and contained no HKD motifs conserved in many PLDs [19]. A distance-based phylogenetic analysis revealed that LyPls-PLD belonged to a member of the GDPD family and was clearly separated from a group containing PLD (Fig. 7). Intriguingly, this suggests that GDPD might be an ancestor protein of PLD. Moreover, LyPls-PLD was clearly separated from at a group containing ptGDPD, pmGDPD, and tmGDPD. Pfam analysis [20] showed that the N terminus (residues 19-271) of LyPls-PLD was assigned as a GDPD family domain and shared structural similarity with ttGDPD (Protein Data Bank code: 2PZ0) and GDPD from Parabacteroides distasonis (pdGDPD; Protein Data Bank code: 3NO3). 84.5 AE 0.7 EDTA 0 a rLyPls-PLD activity was assayed under standard assay condition I.
The relative activity was determined by defining the activity without chemicals as 100%. Data represent the means and standard deviations of experiments performed in triplicate.

Homology modeling analysis and molecular docking
VERIFY3D showed that the modeled structure exhibited 116.6 of 3D-1D total score (i.e., high modeling quality), suggesting that it is sufficient for comparative structural discussion. The modeled structure of LyPls-PLD showed high similarity to the crystal structures of GDPDs such as Protein Data Bank code: 3L12, 4R7O, 2OOG, 3QVQ, and 3KS6 and adopted a TIM barrel fold (Fig. 8). ttGDPD is one of the few GDPDs that have been well studied and characterized. Thus, we compared the structural feature of LyPls-PLD and ttGDPD. Twin arm structure and deep hydrophobic pocket were observed in LyPls-PLD, but not in ttGDPD (Fig. 8C,D). The H46, E73, D75, H88, E175, F211, and W282 residues in LyPls-PLD were highly conserved with ttGDPD (Figs 6 and 8G). We speculated that E73, D75, and E175 are probably calciumion (Ca 2+ )-binding sites. The docking simulation indicated that the calculated free energy of LyPlsCho analog binding was À4.2 kJÁmol À1 . It also suggested that H46 and H88 of the putative catalytic residues were located near phosphate group of the substrate analog, and E50, A55, and N56 were observed near the head group of the substrate analog (Fig. 8G).

Mutational analysis
Except for A55N/D, N56D/Q, E175A, and F211A, we were unable to successfully express other mutant enzymes. The E175A variant was inactive, suggesting that the residue may be important for enzyme activity. For LyPlsCho, the enzyme activities of mutants A55N/ D, N56D/Q, and F211A were remarkably decreased compared with that of WT enzyme, especially N56D and F211A (Table 5). For LyPlsEtn, the enzyme activities of A55N/D and N56D variants were also markedly declined compared with that of WT enzyme; however, in the N56Q and F211A variants, there were slight decrease in the activities. Interestingly, the F211A variant exhibited higher activity toward PlsEtn than WT enzyme, but the enzyme activity of the N56Q variant markedly declined. Also, the other mutations had almost no effect on PlsEtn hydrolysis.

Enzyme kinetics
Kinetic parameters were summarized in Table 6. WT enzyme exhibited much higher affinity toward for LyPlsCho and catalytic efficiency for LyPlsCho hydrolysis than LyPlsEtn. The V max value was 123 lmolÁmin À1 Ámg À1 protein (62.5 lMÁmin À1 ). The apparent K m and k cat /K m values were 13.2 lM and 5345 s À1 ÁmM À1 , respectively (Fig. S2). The catalytic efficiency (k cat /K m ) of F211A mutant for LyPlsCho was remarkably decreased as compared with LyPlsEtn.

Discussion
There is limited information on Pls metabolic pathways and its key enzymes in living organisms. We identified and characterized a novel LyPls-PLD involved in Pls metabolism, in particular, in microorganisms ( Fig. 9). Moreover, the ORF of lpls-pld was identified with NGS, and the peptide was sequenced with LC-MS. We also efficiently produced rLyPls-PLD using E. coli. The putative gtg translational start codon of LyPls-PLD was similar to ttg in Streptoverticillium cinnamoneum PLD (StvPLD) [21]. The signal peptide of LyPls-PLD had a typical Sec signal sequence, suggesting that it should be secreted via the secretory pathway (Sec-system secretion) [22]. Most Streptomyces phospholipases are secreted via the Sec-system pathway [15,[23][24][25][26] except for PLC [23] and C-type enzymes such as glycerophosphocholine cholinephosphodiesterase (GPC-CP) [27] and glycerophosphoethanolamine ethanolaminephosphodiesterase (GPE-EP) [28] that are secreted via a Tat-system pathway. The deduced aa sequence of LyPls-PLD showed high similarity to those of other bacterial GDPDs such as snGDPD [17] and ttGDPD, but no similarity to those of known PLDs. The catalytic reaction of LyPls-PLD is identical to those of GDPD and PLD in terms of hydrolysis of the same phosphoester bond; however, the substrate specificity differs completely from those of GDPD and PLD. In general, GDPDs display broad specificity for glycerophosphodiesters; GPC, GPE, glycerophosphoglycerol, and bis(glycerophosphoglycerol) [18,29]. Conversely, LyPls-PLD exhibited no activity toward GPC, DAGPLs, or Pls but high specificity to LyPlsCho. Moreover, the phylogenetic analysis revealed that LyPls-PLD belongs to GDPD superfamily because of clear separation from a group containing one other GDPD group as well as PLD. It also indicates that the GDPD gene could be an ancestral gene of PLD. According to Pfam, the GDPD family is a member of the clan PLC (CL0384), which contains two GDPD families and two phosphoinositide (PI)-PLC families. The GDPD family appears to have weak similarity to mammalian PI-PLCs, suggesting that the family may adopt a TIM barrel fold (PF00388).
Liang et al. proposed that ttGDPD binds substrate via the coordination effect of a metal ion, and a possible role for the ion in the catalytic reaction would be acting as an electrophile to stabilize the reaction intermediate [18]. This is similar to the coordination effect of metal ion on PI-PLC. Almost all Streptomyces PLDs exhibit Ca 2+ -independent activity [19]; however, some PLDs such as from S. racemochromogenes [33], S. olivochromogenes [36], and S. tendae [32] are Ca 2+ -dependent enzymes, and the iron-containing enzyme ScPLD is activated by Ca 2+ [19,37]. Additionally, both Ca 2+ -dependent and -independent PLDs have been found in mammals, yeasts, bacteria, and nearly all plants; these PLDs also require a micro-to millimolar Ca 2+ concentration to stimulate activity [38]. Because LyPls-PLD is a Ca 2+ -dependent enzyme, the catalytic mechanism of LyPls-PLD might be similar to that of Ca 2+ -dependent PLD or GDPDs. In fact, H46 and H88 in LyPls-PLD were highly conserved with those in the catalytic residues of GDPDs as well as PLD. The X-ray crystal structure analysis of ttGDPD revealed that E44, D46, E119, two water molecules, and the OH group of glycerol compose an octahedral arrangement that exhibits tetragonal bipyramidal coordination with a glycerol molecule binding at one Ca 2+ ion; however, native ttGDPD activity requires Mg 2+ but not Ca 2+ as a cofactor [18]. It is well known that hard acids such as Ca 2+ and Mg 2+ form a complex with hard bases such as water molecules, OH À , COO À , phosphate (PO 4À 3 ), and RNH 2 . Thus, most phosphodiesterases such as acid     [18]. The reason underlying the discrepant metal ion-dependency between ttGDPD and LyPls-PLD remains unclear. With regard to LyPls-PLD, calcium ion may play a key role in stabilizing phosphate group of the substrate and the reaction product, lysoplasmenic acid (LyPlsA; 1-(1-alkenyl)-sn-glycero-3-phosphate) as well as role as electrophile (Fig. S3). The LyPls-PLD hydrolytic activity was also enhanced~2-fold in the presence of 0.1-1 mM AlCl 3. The effect of Al(III) on LyPls-PLD activity seems be the intrinsic property. Likewise, PLD hydrolytic activity from Actinomadura sp. No. 362 was also stimulated 1.4-fold by 1 mM AlCl 3. [40]. Ogino et al. reported that the transphosphatidylation activity for PC and PE of StvPLD was enhanced by Al(III) (~2.5-fold); however, the hydrolytic activity was unaffected with up to 5 mM Al(III) and was completely inhibited by > 5 mM Al(III) [21]. We hypothesize that the physical condition of the substrate micelles (e.g., size and form) was changed with the concentration of Al(III), thus affecting activity. We previously reported a similar result upon enhancement of sphingomyelinase activity by Mg 2+ ions [41].
We also assessed the effect of Triton X-100 on LyPls-PLD activity. The hydrolytic activity of Streptomyces phospholipases are generally stimulated by Triton X-100 [25,26,32,[42][43][44] because their substrate recognition mechanisms appear to depend on substrate form such as emulsified substrate or mixed micelle substrate with Triton X-100 [15]. In contrast, LyPls-PLD activity declined with increasing Triton X-100 concentration in the reaction mixture. We recently reported that the substrate recognition mechanism of PLD 684 depended on substrate forms and preferred mixed micelle substrates to liposomal substrates [15]. Likewise, LyPls-PLD preferred the micelle-formed LyPlsCho substrate to the liposomal LyPlsCho composed of 1 mM POPC/0.4 mM LyPlsCho with modest activity (51.6% activity found using micellar LyPlsCho). This indicates that LyPls-PLD may prefer unimolecular micelle or liposomal substrate but not Triton X-100/LyPlsCho-mixed micelle.
The structure modeling analysis suggested that LyPls-PLD would be similar to the crystal structure of ttGDPD as well as other GDPDs but not those of bacterial PLDs [19,45]. Also, the model structure of LyPls-PLD appeared to adopt a TIM barrel fold. Surprisingly, conjugated polyketone reductase C2 (Protein Data Bank code: 4H8N) from Candida parapsilosis (NADPH-dependent ketopantoyl lactone reductase) also adopts a TIM barrel fold, and its active site interacts with the two phosphate groups of NADPH [46]. Thus, the TIM barrel fold could play a key role in the interaction with the phosphate group. Shi et al. [18] discussed the catalytic mechanism of ttGDPD using its crystal structure. Based on their findings, we speculated that the catalytic mechanism of LyPls-PLD would be similar to those of ttGDPD and PLD [47]. Namely, LyPls-PLD would bind the substrate via the coordination effect of Ca 2+ , and two histidines (H46 and H88) certainly play a role as acid-base catalyst like in ttGDPD and PLD (Fig. S3). With regard to the  catalytic residue, two histidines were conserved between LyPls-PLD and ttGDPD, as well as in other GDPDs and PLDs. Interestingly, GPC-CP and GPE-EP (i.e., a PLC-like enzyme), which cleave GPC or GPE into glycerol and phosphocholine or phosphoethanolamine, utilize two histidines to bind the substrate's phosphate [28]. However, the substrate specificity of LyPls-PLD completely differs from those of GDPD and PLD. Mutational analysis demonstrated that A55 and N56 might be involved in the head group recognition. The N56D mutant, which changes from amide group of Asn to the negative charge of carboxyl group of Asp residue at reaction pH 8, unexpectedly exhibited a remarkable decrease of LyPlsCho hydrolysis activity but had a minimal effect on PlsEtn hydrolysis activity. However, the same mutation had a small effect on LyPlsEtn hydrolysis activity compared with LyPlsCho. Most interestingly, the N56Q variant, in spite of locating distant from sn-2 acyl chain of PlsEtn, exhibited 40% activity for PlsEtn compared with the activity of the WT enzyme, whereas the other variants, except for F211A, had a modest and almost no effect on PlsEtn hydrolysis. Based on these results, we reasoned that replacement of Asn56 with Gln is a mutation that permits LyPlsEtn hydrolysis but not PlsEtn, while a longer chain might interrupt LyPlsCho binding due to a bulky Cho rather than an Etn head group. Moreover, the kinetic analysis demonstrated that the affinity of LyPlsCho to the enzyme markedly decreased in F211A mutant due to the replacement of Phe, which has a bulky and hydrophobic side chain, by Ala with a small methyl group. This suggests that the Cho head group might interact hydrophobically with Phe group of F211 when the enzyme incorporates the substrate. However, F211 and W282 were located in close proximity to sn-1 ether bond and its alkyl chain of the substrate analog, but not near the Cho head group. Thus, we considered that F211 and W282 residues could interact with the sn-1 ether bond and its alkyl chain of LyPls substrate. Further investigation into the protein's structure is required to elucidate the mechanisms responsible for LyPls-PLD substrate recognition. Crystal structure determination and more detailed mutation analyses are currently in progress.  The inhibition study results suggested that compounds such as GPC and CDP-choline with no acyl chain and ether bond were not LyPls-PLD substrates and were unable to enter enzyme's active center. Compared with LPC and LPE, the inhibitory effect of LPA was weak; however, LPC and LPE can be LyPls-PLD substrates. Additionally, LyPls-PLD preferred the Cho head group over Etn of lysophospholipids as well as LyPls. Wu et al. [1] reported that 50% of LyPlsase activity was inhibited with 50 lM LPA. As for LyPls-PLD inhibition, a concentration of 4 mM analog (10fold higher concentration toward LyPlsCho substrate) was required to inhibit enzyme activity (Table 4). Substrate specificity profiling of LyPls-PLD showed that the enzyme recognizes a vinyl ether bond at the sn-1 position and the head group (Fig. 5, Tables 5 and 6). Taken together, we concluded that LyPls-PLD would preferentially recognize the sn-1 vinyl ether bond and then the head group. Yet, LyPls-PLD likely prefers vinyl ether bond over acyl ester bond at the sn-1 position.
Finally, we concluded that LyPls-PLD is a novel LyPlsCho-specific enzyme that is clearly different from any known PLD as well as GDPD; however, the catalytic mechanism of LyPls-PLD seems similar to that of ttGDPD rather than PLD (Fig. S3). The identification of LyPls-PLD reveals a new pathway for Pls metabolism, in particular, in microorganisms. We showed that LyPls-PLD is obviously different from PLD and the phosphodiesterase families containing GDPD with regard to the catalytic function, substrate recognition mechanism, and biochemical roles. More detailed mutation analyses are currently in progress to elucidate the substrate recognition mechanism of LyPls-PLD. Further advances in this area could lead to the development of diagnostic reagent kits for early stages of diseases such as Alzheimer's disease and arteriosclerosis.  (Toronto, ON, Canada). SM and Etn hydrochloride were purchased from Sigma-Aldrich Japan Co., LLC (Tokyo, Japan). GPC was purchased from Bachem AG (Torrance, CA, USA). CDP-choline and Cho hydrochloride were purchased from Wako Pure Chemical Industries Ltd. (Osaka, Japan). Bacto-peptone and Bactomalt extract were purchased from Becton, Dickinson and Company (Franklin Lakes, NJ, USA). Toyopearl Giga Cap Q-650M and Toyopearl PPG-600M were purchased from Tosoh (Tokyo, Japan). RESOURCE Q, RESOURCE ISO, Mono Q, Superdex 200 10/300 GL, and HisTrap HP columns were purchased from GE Healthcare Japan (Tokyo, Japan). Choline oxidase (COD) from Arthrobacter globiformis was from Asahi Kasei Pharma (Tokyo, Japan). Peroxidase (POD) and yeast extract BSP-B were purchased from Oriental Yeast Co., Ltd. (Tokyo, Japan). 4-Aminoantipyrine (4-AA) was purchased from Nacalai Tesque Inc. (Kyoto, Japan). N,N-Bis(4-sulfobutyl)-3-methylaniline, disodium salt (TODB) was purchased from Dojindo Laboratories (Kumamoto, Japan). His6-tagged recombinant amine oxidase (rSrAOX) from Syncephalastrum racemosum was produced using pET24a(+)/E. coli and purified using affinity column chromatography with HisTrap HP [48]. All other chemicals were of the highest or analytical grade.

Bacterial strains, plasmids, and culture conditions
Approximately 200 actinomycetes strains were obtained from NBRC (the NITE Biological Resource Center, Chiba, Japan), and LyPls-PLD producers were screened with enzyme activity assays. Strain RD004668 (RD4668) showed the highest LyPls-PLD activity in 5 mL cultivation and was selected for further investigation. Strain RD4668 from excrement of Kanagawa, Japan was identified as Thermocrispum sp., a near relative of Thermocrispum municipal, based on morphological, physiological, biochemical characterizations, and 16S rDNA sequence analysis. The 16S rDNA sequence of strain RD4668 was deposited in the DDBJ database under accession number AB873024. Strain RD4668 was deposited as NITE BP-01628 in the NITE Patent Microorganisms Depositary (NPMD; Chiba, Japan). The optimum growth temperature of strain RD4668 is 45°C. Strain RD4668 was incubated in 5 mL ISP2 medium (1% malt extract, 0.4% yeast extract, 0.4% glucose, pH 7.3) supplemented with 0.41 mM Brij35 at 45°C for 48 h with shaking (160 strokes per min). Next, 1 mL of the 48-h culture was transferred into a 500-mL flask containing 100 mL ISP2 medium and incubated at 45°C for 84 h with shaking (180 rpm). was collected by centrifugation (18 800 g, 10 min) and suspended in buffer B (buffer A, pH 9.0), followed by dialyzing against buffer B. After removing insoluble materials by centrifugation (21 800 g, 10 min), the obtained supernatant was loaded onto a Toyopearl Giga Cap Q-650M column (2.5 9 4.0 cm) equilibrated with buffer B. After washing the column with three column volumes (3 CV), the bound proteins were eluted with a linear gradient (20 CV) of 0 to 1 M NaCl in buffer B at 2 cmÁmin À1 . Ammonium sulfate was added to the active fractions to 1.5 M followed by loading onto a Toyopearl PPG-600M column (2.5 9 4.0 cm) equilibrated with 1.5 M (NH 4 ) 2 SO 4 /buffer A. After washing, the proteins were eluted with a linear gradient (15 CV) of 1.5 to 0 M (NH 4 ) 2 SO 4 in buffer A and with buffer A (10 CV) at 2 cmÁmin À1 . The active fractions were pooled, and the buffer was replaced with buffer B using Vivaspin 20-10 000 MWCO (Sartorius AG, G€ oettingen, Germany) followed by loading onto a RESOURCE Q column (1 mL) equilibrated with buffer B. The proteins were eluted with a linear gradient (40 CV) of 0 to 0.5 M NaCl in buffer B at 6 cmÁmin À1 . The active fractions were concentrated using Vivaspin followed by loading onto a Superdex 200 column (24 mL) equilibrated with 0.15 M NaCl/buffer A. The proteins were eluted with the same buffer at 37.5 cmÁh À1 . The active fractions were exchanged to 1.5 M (NH 4 ) 2 SO 4 /buffer A using Vivaspin followed by loading onto a RESOURCE ISO column (1 mL) equilibrated with the same buffer. After washing, the proteins were eluted with a linear gradient (40 CV) of 1.5 to 0 M (NH 4 ) 2 SO 4 in buffer A at 6 cmÁmin À1 . The buffer of the active fractions was exchanged to buffer B using Vivaspin followed by loading onto a Mono Q column (1 mL) equilibrated with the same buffer. After washing, the proteins were eluted with a linear gradient (40 CV) of 0 to 0.4 M NaCl in buffer B at 6 cmÁmin À1 . Fractions exhibiting high specific LyPls-PLD activity and running as a single band on SDS/PAGE were pooled and used in subsequent investigations.

Enzyme activity assays
The WT enzyme activity for LyPlsCho was assayed by measuring the Cho formation rate as follows. Solutions of TODB, 4-AA, and POD were prepared in distilled water. The standard assay mixture (50 lL) containing 80 mM Tris/HCl (pH 8.0), 0.4 mM LyPlsCho, and 2 mM CaCl 2 was incubated at 37°C for 5 min. The assay was started by addition of 10% (v/v) enzyme solution and further incubation at 37°C for 10 min (standard assay condition I). A colorimetric solution (200 lL) containing 0.03% (w/v) 4-AA, 0.02% (w/v) TODB, 0.75 UÁmL À1 COD, 5 UÁmL À1 POD, and 10 mM EDTA was added to the enzyme reaction mixture to stop the reaction and determine the concentration of Cho ([Cho]) released by the enzyme reaction. Absorbance at 550 nm (A 550 ) based on a quinone dye generated by the coupling reaction of H 2 O 2 with 4-AA and TODB was measured using a Multiskan FC microplate reader (Thermo Fisher Scientific K.K., Kanagawa, Japan). The concentration of produced H 2 O 2 ([H 2 O 2 ]) was determined by measuring the absorbance at 550 nm (A 550 ) using a calibration curve generated from known [H 2 O 2 ]. One unit (U) of enzyme activity was defined as the amount of enzyme that released 1 lmol Cho from LyPlsCho per min. The enzyme activity for LyPlsEtn was assayed by measuring the rate of Etn formation. The concentration of Etn ([Etn]) released by the enzyme reaction with the standard assay mixture containing 0.4 mM LyPlsEtn was determined using 0.75 UÁmL À1 rSrAOX instead of COD, and the others were determined as described above.

Protein analyses
The protein concentration was determined with a bicinchoninic acid protein assay reagent kit (Thermo Fisher Scientific K.K.) with bovine serum albumin as the standard. SDS/PAGE analysis was carried out according to the Laemmli method [50]. Internal terminal aa sequencing of thẽ 35-kDa band was performed with nanoLC-MS/MS on a Xevo QTOF MS system (Waters Corp., Milford, MA, USA) as described previously [25]. The enzyme gene was identified on PROTEINLYNX GLOBAL SERVER, version 2.3 (Waters) using the ORF database based on the NGS results.

Nucleotide and peptide sequence accession numbers
The nucleotide sequences of the 16S rDNA of strain RD4668 and the LyPls-PLD gene, designated lpls-pld, were deposited in the DDBJ database under the accession numbers AB873024 and AB874601, respectively. vectors, pET24a/lpls-pld mat (without His tag) and pET24a/ lpls-pld mat_his (with His tag), were constructed to produce active LyPls-PLD as previously reported [28]. The recombinant vector pET24a/lpls-pld mat carried the gene for mature LyPls-PLD, lpls-pld mat, between the NdeI and EcoRI sites of pET24a(+), while pET24a/lpls-pld mat_his contained lplspld mat between the NdeI site and a C-terminal His6 tag. To construct pET24a/lpls-pld mat, PCR primers were designed based on the ORF of lpls-pld identified in the ORF database. The forward primer 5 0 -ttcatatgaccaccagaacgacagacaatc-3 0 containing the N-terminal codon (NdeI, single underline; Thr, italics) and the reverse primer 5 0ttgaattctcatccgcacggatcgacgccc-3 0 containing the stop codon and EcoRI site were used. One Shot BL21 (DE3) Chemically Competent E. coli (Thermo Fisher Scientific K.K.) cells were transformed with the recombinant expression vectors, after which the transformants were selected on Luria-Bertani (LB) agar plates containing 50 lgÁmL À1 ampicillin (LBA) and then confirmed by colony PCR. Each transformant was screened for LyPls-PLD activity during 5 mL cultivation in LBA broth, and those exhibiting the highest activity were selected. To construct pET24a/lpls-pld mat_his, His6 tag was joined to the C-terminal aa sequence of LyPls-PLD by inverse PCR as previously reported [48] using pET24a/lpls-pld mat as a template and the inverse primers 5 0 -caccaccaccaccaccactg-3 0 and 5 0 -tccgcacggatcgacgcccgg-3 0 . SHuffle â T7 Express Competent E. coli cells (New England Biolabs Japan, Tokyo, Japan) were transformed with the recombinant expression vectors, after which the transformants were selected on LB agar plates containing 30 lgÁmL À1 kanamycin (LBK). Finally, a transformant exhibiting the highest activity was selected as above. Shuffle T7 E. coli cells harboring pET24a/lpls-pld mat_his were inoculated into a test tube containing 5 mL LBK seed medium and cultivated overnight at 30°C with shaking (160 strokesÁmin À1 ). Thereafter, 1% (v/v) inocula were transferred to 500-mL flasks containing 100 mL of LBK fermentation medium and cultivated at 30°C with shaking (160 rpm). After 24 h, 0.4 mM isopropyl-b-D-thiogalactopyranoside was added, and the cultures were continued for an additional 4 h at 30°C. Two hundred milliliters of the recombinant E. coli cell cultures were then centrifuged (18 800 g, 20 min), and the resulting cell paste (~2 g wet weight) was washed twice and suspended in buffer A. The cells were disrupted using a sonicator UD-201 (Tomy Seiko Co., Ltd., Tokyo, Japan; 100 W, 20 kHz, 10 min, on ice) and centrifuged at 21 800 g for 20 min. The cell-free extracts (cfe) were then loaded onto a HisTrap HP column (5 mL) equilibrated with 0.5 M NaCl, 5 mM imidazole/ 50 mM buffer A. After washing, the proteins were eluted with a linear gradient (15 CV) of 5 to 500 mM imidazole in the same buffer at 1.5 cmÁmin À1 . Fractions exhibiting high specific activity with high purity were pooled followed by buffer exchange into buffer A as described above and used for subsequent investigation.

Enzymatic characterization
The purified rLyPls-PLD was used for enzymatic characterization. Except for LyPlsCho, LyPlsEtn, SM, and GPC, the insoluble substrates such as PlsCho, PlsEtn, and PLs were dispersed in distilled water by vortexing and sonication. The liposomal substrate composed of 1 mM POPC/0.4 mM LyPlsCho was prepared according to the hydration method and used as liposomal substrate [26,51]. The enzyme activity toward each substrate was assayed under standard assay condition I. The effect of metal ions on enzyme activity was investigated under standard assay condition I with the same buffer containing 2 mM metal ion, EDTA, or 1 mM inhibitor. Inhibitors assessed were DTT, 2ME, IAA, and PMSF. The effect of Triton X-100 concentration ([Triton X-100]) and Ca 2+ concentration ([Ca 2+ ]) on enzyme activity was investigated under standard assay condition I. The effect of substrate analogs on LyPls-PLD activity was examined under standard assay condition I containing 0.4 mM LyPlsCho and 4 mM substrate analog. All experiments were carried out three times independently.
Each buffer (sodium acetate, MES-NaOH, BisTris/HCl, Tris/HCl, and glycine-NaOH) was used to investigate the effect of pH on enzyme activity and stability. The optimum pH was examined under standard assay condition I (37°C, 1 min) with 0.4 mM LyPlsCho and 2 mM CaCl 2 in 80 mM of each buffer. To determine pH stability, the enzyme sample was incubated at 4°C for 4 h in 50 mM of each buffer. The residual activity was assayed under standard assay condition I (50°C, pH 8.0 for 1 min: standard assay condition II). The optimum temperature was determined by measuring enzyme activity at each temperature under standard assay condition II. To determine thermal stability, the enzyme sample was incubated at each temperature for 60 min in 20 mM Tris/HCl buffer (pH 8.0), and the residual activity was assayed under standard assay condition II. All experiments were independently carried out three times.  Homology modeling of LyPls-PLD and docking model with substrate analog Based on a template of five proteins (Protein Data Bank code, 3L12, 4R7O, 2OOG, 3QVQ, 3KS6), the homology model of LyPls-PLD was created using an HHPRED search (http://toolkit.tuebingen.mpg.de/hhpred) [52] and Modeller 9.16 (http://toolkit.tuebingen.mpg.de/modeller) [53]. VERIFY3D (http://services.mbi.ucla.edu/Verify_3D/) [54] was used to assess the quality of the predicted models, which were drawn using MOLFEAT Version 5.1.0.24 (FiatLux Corp., Tokyo, Japan). The docking model between LyPls-PLD and LyPlsCho analog with a short chain (C3)-alkenyl ether bond at sn-1 was simulated using AutoDock [55].

Site-directed mutagenesis
We speculated that the highly conserved residues of H46, E73, D75, H88, E175, F211, and W282 in LyPls-PLD are likely concerned with catalytic function. Moreover, based on the substrate analog docking simulation analysis, we considered that E50, A55, and N56 located near the nitrogen atom of choline in the substrate analog may be involved in recognizing the substrate head group. We then tried to generate the mutant enzymes of rLyPls-PLD, H46A/R, E50D, A55N/D/E, N56E/D/Q, E73A/R, D75A/ R, H88A/R, E175A/R, F211A/R, and W282A/R using pET24a/lpls-pld mat_his as the template and our previously described inverse PCR method [15]. The PCR product was treated with DpnI to digest the parental DNA template and select for mutation-containing synthesized DNA. The pET24 vector DNA incorporating the desired mutations was then transformed into Shuffle T7 E. coli cells. The mutants were selected and cultured in 5 mL LBK broth as described for rLyPls-PLD production. The variant enzymes were purified using the HisTrap column as above.

Supporting information
Additional Supporting Information may be found online in the supporting information tab for this article: Fig. S1. SDS/PAGE analysis of purified enzyme (A) and rLyPls-PLD produced using transformed E. coli (B). Fig. S2. Michaelis-Menten plot of steady-state kinetics. Fig. S3. Predicted catalytic mechanism of LyPls-PLD. Table S1. Contigs assembled by Velvet and ORF prediction by getorf.