Structural insights into phosphatidylethanolamine formation in bacterial membrane biogenesis

Phosphatidylethanolamine (PE), a major component of the cellular membrane across all domains of life, is synthesized exclusively by membrane-anchored phosphatidylserine decarboxylase (PSD) in most bacteria. The enzyme undergoes auto-cleavage for activation and utilizes the pyruvoyl moiety to form a Schiff base intermediate with PS to facilitate decarboxylation. However, the structural basis for self-maturation, PS binding, and decarboxylation processes directed by PSD remain unclear. Here, we present X-ray crystal structures of PSD from Escherichia coli, representing an apo form and a PE-bound complex, in which the phospholipid is chemically conjugated to the essential pyruvoyl residue, mimicking the Schiff base intermediate. The high-resolution structures of PE-complexed PSD clearly illustrate extensive hydrophobic interactions with the fatty acyl chains of the phospholipid, providing insights into the broad specificity of the enzyme over a wide range of cellular PS. Furthermore, these structures strongly advocate the unique topology of the enzyme in a lipid bilayer environment, where the enzyme associates with cell membranes in a monotopic fashion via the N-terminal domain composed of three amphipathic helices. Lastly, mutagenesis analyses reveal that E. coli PSD primarily employs D90/D142–H144–S254 to achieve auto-cleavage for the proenzyme maturation, where D90 and D142 act in complementary to each other.

of the N-terminus serine to pyruvoyl in the α-subunit, which is an indispensable prosthetic group for decarboxylation activity 20 . PSD is a serine protease that employs a classic catalytic triad composed of Asp-His-Ser, where nucleophilic serine attacks the scissile peptide bond to produce α-and β-chains 21 . Through mutagenesis experiments, D139-H198-S308 and D210-H345-S463 were identified as the catalytic triads for PSD in Plasmodium knowlesi 22 and S. cerevisiae 23 , respectively. After the proenzyme undergoes auto-proteolysis, an active complex is formed between the α-and β-chains, which are mainly responsible for catalytic activity and membrane association, respectively. The reaction mechanism for PSD was proposed based on that of pyruvoyl-dependent histidine decarboxylase 24 , where the amine group of PS and the α-carbonyl carbon of pyruvoyl form a Schiff base intermediate, followed by decarboxylation 24 .
The first crystal structures of PSD from E. coli in apo-and lipid-bound forms have been reported recently, which offer valuable information regarding the membrane-associated mechanism, substrate binding, and determinants critical for catalytic activity 25 . The lipid-bound structure displays covalent conjugation of PE to the pyruvoyl residue; however, a rather low resolution (3.60 Å) limits the identification of acyl chain moieties of the bound lipid molecule. Here, we report two crystal structures of E. coli PSD, representing apo states at resolutions of 1.90 and 2.63 Å, along with two PE-bound structures of PSD at resolutions of 2.12 and 2.70 Å, which mimic the Schiff base intermediate formed between the pyruvoyl group and a phospholipid. Our high-resolution structures allow the identification of the exact locations where intermolecular interactions occur between the enzyme and the bound phospholipid molecule via diacyl chains in particular. Structure-guided mutagenesis analyses confirmed the key residues involved in phospholipid recognition, decarboxylation of PS, and maturation of PSD. In particular, we identified the crucial residues required for the activation of proenzymes, which have not been characterized for E. coli PSD.

Results
To facilitate crystallization and structure determination, we employed a recombinant E. coli PSD with the last 35 residues removed (PSD  ). The amino acid sequences of this region are missing in most bacteria, except for certain species of Enterobacterales ( Supplementary Fig. S1). The in vitro activity of PSD  was tested using soy PS as a substrate, and five major phospholipid species were quantitatively analyzed via LC-MS ( Fig. 1B and Supplementary Fig. S2A). Under our assay conditions, it is estimated that over 95% of PS is enzymatically converted to PE. This result is comparable to that of the full-length wild-type enzyme ( Supplementary Fig. S2B), indicating that truncation of the C-terminal residues does not significantly contribute to decarboxylation activity.
Overall feature of apo structures. Using native and selenomethionine-derivatized recombinant PSD 1-287 , the crystal structures of apo PSD 1-287 were determined from two crystal forms at resolutions of 1.90 and 2.63 Å, which are described as Apo-PSD1 and Apo-PSD2, respectively. The crystallographic statistics are summarized in Table 1. The asymmetric unit of Apo-PSD1 contains two αβ-heterodimers, whereas that of Apo-PSD2 contains four ( Supplementary Fig. S3). The biological assembly of PSD 1-287 appears to be a dimer of αβ-heterodimers, (αβ) 2 , which is commonly identified in all four structures presented herein ( Fig. 2A). The heterotetrameric form of PSD  in solution was consistent with the results of the SEC-MALS analysis ( Supplementary Fig. S4). Two out of three molecules of N-dodecyl-b-maltoside (DDM), a detergent used to solubilize the enzyme, were located at the interface between two adjacent β-subunits, augmenting the dimerization of αβ-heterodimers in Apo-PSD1. No detergent molecules were modeled in Apo-PSD2 structure. The α and β subunits form a tight 1:1 complex in each heterodimer with an average interface area of 1790 Å 2 , where 29 out of 34 residues from the α subunit participate in forming the dimerization interface. Using the N-terminal sheet composed of residues from Pvl-254 to Ala-261, the α-subunit was integrated between two anti-parallel sheets of the β-subunit  www.nature.com/scientificreports/ (Fig. 2B). The N-terminal pyruvoyl residue (Pvl-254) of the α-subunit was unambiguously identified in both apo structures (Fig. 2C). Gly-253 of the β subunit, which is no longer connected to Pvl-254 after auto-cleavage, is approximately 12.6 Å apart when the distance between two C α s is measured, and it forms a salt bridge with the side chain of Arg-221. Consequently, Pvl-254, which is located at the base of the funnel-shaped active site, is open for substrate binding (Fig. 2D).
Identification of the membrane association domain. The three N-terminal helices contributed by each β-subunit lie in a plane, where hydrophobic residues are aligned on one side of the helices (Fig. 3A). Therefore, six N-terminal helices of the heterotetramer are ideally shaped to make contact with the coplanar lipid layers of the inner membrane of E. coli. To investigate the role of the three helices in membrane association, the cellular location of the wild-type protein and a series of truncated PSD in the N-terminal domain were examined via immunoblotting of soluble, membrane, and insoluble fractions after cell lysis (Fig. 3B). The mutant ΔH1 lacks only the N-terminal helix 1, whereas ΔH2 lacks helices 1 and 2, and ΔH3 is devoid of helices 1-3. Most wild-type proteins were detected in the membrane fraction, whereas ΔH1 exhibited enrichment in the soluble fraction, as well as in the membrane fraction. Additional truncations led to incomplete auto-cleavage of the proenzyme in Table 1. Crystallographic data collection and refinement statistics. Each dataset was collected from a single crystal. Values in parentheses are for highest-resolution shell. www.nature.com/scientificreports/ ΔH2 and ΔH3, and the level of mature PSD was low overall. Notably, uncleaved ΔH3 was enriched in the soluble fraction, and the cleaved form was not detected at a significant level in all fractions. The concentration of hydrophobic residues on the surface of the N-terminal helices further supports that the N-terminal domain is a hot spot for membrane association ( Fig. 3C and Supplementary Fig. S5). Interestingly, DDM molecules identified in the structure of Apo-PSD1 interact with the positively charged surface of PSD through the relatively polar disaccharide moiety, whereas the dodecyl tail of the detergent interacts with the hydrophobic surface of the enzyme encompassing the N-terminal helices ( Supplementary Fig. S6). These molecular interactions may mimic those occurring on the cellular membranes, where PSD is partially embedded within a single layer of phospholipids using the N-terminal helical domain.

Phospholipid-bound structures.
To identify the molecular determinants required for substrate recognition and binding, we first attempted to form a stable covalent linkage between PSD and a phospholipid via sodium cyanoborohydride (NaCNBH 3 )-dependent reduction of a Schiff base intermediate (Fig. 4A) 19 . Conjugation was tested using 8:0/8:0 PE (8PE), 10:0/10:0 PS (10PS), and 14:0/14:0 PS (14PS) with full-length and truncated PSD, and the results were analyzed using matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry. Incubation of PSD with a phospholipid increased the mass-to-charge ratio (m/z) of the α-subunit by 450.5, 506.7, and 618.8 for 8PE, 10PS, and 14PS, respectively, which is consistent with the formation of conjugates with the corresponding PE (Fig. 4B). The results indicate that the reduction of the Schiff base occurred after decarboxylation. MS data confirmed the successful modification of the protein with an efficiency greater than 94%: 98%, 97%, and 94% conjugation for 8PE, 10PE, and 14PE, respectively. Among these samples, 8PE-and 10PE-linked PSD 1-287 yielded crystals, which diffracted to 2.12 Å and 2.70 Å resolution, respectively. These X-ray diffraction data were used to determine the structures shown in Fig. 4C; they www.nature.com/scientificreports/ are further denoted as 8PE-PSD and 10PE-PSD, respectively. The asymmetric unit of 8PE-PSD contains two αβ-heterodimers, whereas that of 10PE-PSD contains four. The overall conformation of PSD in both PE-bound structures is highly homologous to that in apo structures (r.m.s.d. of C α s = 0.218 Å for 8PE-PSD; 0.399 Å and 0.390 Å for the 10PE-PSD αβ-heterotetramer). Similar to Apo-PSD1, one of the two DDM molecules was modeled at the interface between the two β-subunits in 8PE-PSD. Charge distribution on the enzyme surface demonstrates that the substrate-binding site creates a predominantly hydrophobic environment optimal for accommodating fatty acyl chains of glycerophospholipids (Fig. 4D). Solid electron densities were observed in all subunits of PE connected to Pvl-254. However, the quality of local electron densities around fatty acyl chains varies from one subunit to another. For example, in one αβ-heterodimer of 8PE-PSD, 12 out of 16 carbon atoms could be modeled in the fatty acyl group of 8PE, whereas only 9 were traceable in the other. Although we were able to observe the electron densities of two fatty acyl chains connected to the glycerol backbone, the sn-1 and sn-2 positions on the glycerol backbone could not be determined unambiguously with the present data even at a resolution of 2.12 Å. The assignment was based on the electron density in chain B of the 8PE-PSD structure, which was considered the best. The sn-2 acyl chains were modeled in close proximity to the hydrophobic protein surface, which was composed of Val-37, Phe-41, Phe-63, Phe-67, and Leu-252, with sn-1 acyl chains extending near the surface defined by Leu-13, Leu-18, and Thr-204. In general, the fatty acyl chain at the sn-2 position is more ordered than that connected to sn-1 in the present structure. For instance, the entire sn-2 acyl chains could be modeled in 10PE-PSD, whereas most sn-1 acyl chains were disordered because of the relatively shallow binding pocket. In both structures, the sn-2 acyl chains extend to the hydrophobic surface of the N-terminal helical domain, where the membrane association is presumed to occur, as described earlier. The sn-2 chain-binding site, which is composed of hydrophobic residues, allows for non-specific binding To examine whether these amino acid residues are critical for decarboxylation activity, we performed in vitro assays using PSD 1-322 variants containing a site-specific mutation, followed by LC-MS analysis (Fig. 5B). When Ser-166 was substituted with alanine, however, no significant change from the wild-type protein was observed in the in vitro assay. Meanwhile, Y137F was able to decarboxylate PS at a reduced rate (~ 50%). Tyr-137 is highly conserved among the PSD family (85%) or occasionally replaced by arginine (15%). Interestingly, when both Tyr-137 and Ser-166 were mutated, the double mutant protein Y137F/S166A exhibited a substantially lower activity compared to the wild-type protein, demonstrating the combinatorial effects of these residues in interacting with the phosphate moiety of the lipid substrate. Mutation of the absolutely conserved His-144 completely  Mutagenesis studies on the mechanism of auto-cleavage. Because PSD is a member of the serine protease family and is known to require a catalytic triad for auto-cleavage 22,23 , we investigated the amino acid residues that are responsible for the maturation of the proenzyme on the basis of genetic and structural informa- Candidate residues for the maturation of E. coli PSD were selected for mutation, and the auto-cleavage of the mutant proteins was analyzed from cell lysates after a 4 h induction period (Fig. 6A). Additionally, we purified the recombinant proteins from cell cultures incubated overnight for induction and analyzed the result of the cleavage, which occurred approximately 24 h after the induction of protein expression had begun. Alanine mutation of Ser-254 completely abolished auto-cleavage, which is consistent with a previous report 25 . Similarly, a single mutation of His-144 to alanine or asparagine appeared initially to hamper the auto-proteolysis of the proenzyme; however, both mutants were identified as substantially cleaved in the purified form. An unknown cleavage product with a size slightly smaller than that of the proenzyme was detected from the lysates for both H144A and H144N. However, this unknown fragment was insoluble and not observed after purification. For the H147N and H147A mutants, a significant cleavage occurred after 4 h of induction, which became nearly complete after the purification steps. Similarly, in single mutant proteins of Asp-90, most of the proenzyme was cleaved initially, and the completely processed PSD was identified after purification. However, purified D90N or D90A mutant proteins did not compromise PS decarboxylation activity, unlike His-144 mutants (Fig. 6B). Since the side chains of Asp-90 and His-144 are ~ 9 Å apart in the crystal structures, Asp-90 may not be optimally positioned to depolarize His-144. In the vicinity of the histidine residue, Asp-142 is located at ~ 6 Å; Asp-142 has not been considered as a component of the catalytic triad of PSD in the past. We tested this residue for auto-cleavage by mutating it to asparagine or alanine; however, the results were nearly identical for these mutants and Asp-90 mutants. The results of these single mutants in the pro-enzyme processing suggest that His-144 and His-147 may act as complementary components of the D-H-S triad for E. coli PSD, as may Asp-90 and Asp-142. To investigate this possibility, we constructed and analyzed double mutant proteins of these pairs, that is, H144A/H147A and D90A/D142A. Surprisingly, the D90A/D142A double mutant did not undergo auto-cleavage and remained as a proenzyme in the lysates, although became cleaved in the purified form. Meanwhile, the result from H144A/ H147A mutant was quite similar to that of H144A; the double mutant was not able to auto-cleave initially, yet displayed significant cleavage products after purification steps. Therefore, our single-and double-mutation experiments support the hypothesis that His-144 plays a major role in activating Ser-254, whereas Asp-90 and Asp-142 can function in complementary to each other in the proenzyme maturation process. Lastly, we introduced mutations at Gly-253 and Thr-255, which are highly conserved amino acids located next to the nucleophilic Ser-254. Together, these residues comprise the LGST motif, the consensus sequence conserved among the PSD family on which auto-cleavage occurs. G253A, G253P, and T255A could not form auto-cleaved www.nature.com/scientificreports/ products after 4 h of induction; however, G253A and T255A displayed approximately 70% cleavage after a longer period, as identified in the purified form. G253P remained uncleaved in its purified form, similar to S254A.

Discussion
Previous studies have shown that substrate analogs lacking fatty acyl chains, such as serine, phosphoserine, or glycerol phosphoserine were not decarboxylated by PSD, underscoring the importance of essential hydrophobic interactions with fatty acyl chains for substrate binding and catalysis 26 . Our high-resolution structures of PE-bound PSD reveal detailed molecular interactions between the protein and the phospholipid substrate; one fatty acyl chain binds to the larger hydrophobic protein surface defined by Val-37, Phe-41, Phe-63, Phe-67, and Leu-252, whereas the other acyl chain is found on the relatively smaller surface composed of Leu-13, Leu-18, and Thr-204. It is not feasible to unambiguously discriminate the sn-isomers of the bound phospholipid, and our structural data suggest that both binding modes may be plausible; that is, the sn-1 or sn-2 acyl chain is not restricted to binding at a particular site of the enzyme. Although fatty acyl chains are an essential component for effective binding to PSD, it appears that the enzyme mainly recognizes one of the fatty acyl chains via non-specific hydrophobic interactions, whether it is an sn-1 or sn-2 acyl chain. Phospholipids are highly diverse in length and degree of saturation of fatty acyl chains, where 258 different species have been experimentally identified in E. coli to date 27 . As the sole enzyme for synthesizing PE in bacteria, PSD must act promiscuously on a wide range of PS. Current structural data provide insights into the broad specificity of PSD, where the lipid-binding surface permits non-specific hydrophobic interactions with various types of fatty acyl chains of the phospholipid. Additional molecular determinants for PS binding are Tyr-137, His-144, and Ser-166, which interact with phosphate and carboxyl groups. Our structural and biochemical studies clearly support the formation of Schiff base intermediates on the reaction coordinate and suggest that His-144 is likely to interact with the departing carboxyl group of PS, which was also shown to be critical in the formation of a stable Schiff base intermediate (Fig. 7A). Another valuable feature highlighted in the present structures is that PSD is a monotopic membrane protein that is embedded into a single face of the membrane. Structures of monotopic membrane proteins are extremely rare, accounting for only ~ 0.06% of the total non-redundant structures 28 . We propose that heterotetrameric PSD employs coplanar amphipathic helices in the N-terminus of beta-subunits to associate with the cell membrane, as illustrated in Fig. 7B. Estimation of the protein-membrane border in bacterial cells is further aided by cocrystallized DDM, the length of which roughly encompasses a single membrane leaflet (Supplementary Fig. S8). In this model, the active site is in proximity to the membrane interface, similar to phosphoglycosyl transferase, PglC, another example of monotopic membrane protein 29 .
During the preparation of this manuscript, Watanabe et al. reported two crystal structures of E. coli PSD in apo and PE-bound states at resolutions of 2.6 and 3.6 Å, respectively 25 . The conclusions drawn by the authors are largely consistent with ours, such as the biological assembly of the enzyme, membrane association of the enzyme, and biochemical properties of key residues contributing to phospholipid recognition. Our high-resolution www.nature.com/scientificreports/ structures provide more detailed insights into the protein-lipid interactions, indicating the precise active sites on the enzyme where the acyl chains of phospholipids bind. Furthermore, our in-depth analyses of the auto-cleavage of proenzymes required for maturation identified critical residues comprising a D-H-S catalytic triad for E. coli PSD; His-144 appears to be the major player in enhancing the nucleophilicity of Ser-254, whereas Asp-90 and Asp-142 can functionally complement to each other during the maturation process. Despite H144A, H144N, H144A/H147, and D90A/D142A were not able to auto-cleave initially, these mutant proteins were eventually cleaved to a certain degree after a longer period, suggesting an alternative mechanism may exist for the activation of the proenzyme. In summary, our data support that the auto-cleavage of the E. coli PSD is mainly achieved by D90/D142-H144-S254, analogous to a conventional D-H-S catalytic triad. The LGST motif, a signature cleavage site conserved among bacterial PSD, is predicted to form a short loop connecting two beta sheets in the proenzyme. It has been proposed that the relief of conformational strain across the cleavage site in proenzymes may be a driving force for auto-cleavage in other pyruvoyl-dependent decarboxylases, including S-adenosylmethionine decarboxylase 30 , aspartate decarboxylase 31,32 , and histidine decarboxylase 33 . Our mutagenesis studies showed that G253A, G253P, and T255A exhibited significantly reduced auto-cleavage, suggesting the importance of the correct conformation of the loop and positioning of the serine hydroxyl group for proenzyme maturation. Because of the apparent structural reorganization of the processed α-and β-subunits in the current structures, it is challenging to precisely locate and orient the key residues in the 3-D space prior to the auto-cleavage event. The determination of the pro-PSD structure will provide critical insights regarding the activation process.
The present structural information will be highly valuable in developing a novel class of antibiotics because numerous pathogens are known to require PSD for viability. Additionally, the activity of a mammalian homolog has been reported to be important. In an animal model, mouse embryos lacking psd1 did not survive past 9 days of development 34 . Human PISD has been associated with various diseases 35 including cancer [36][37][38] , Parkinson's disease 39 , Alzheimer's disease 40 , liver disease 41 , candidiasis [42][43][44][45] , and malaria 46,47 . The physiological significance of PISD has been demonstrated in tumor-initiating cells, where overexpression of PISD downregulates mitochondrial function and inhibits tumor growth 37 . Notably, recent studies demonstrated an emerging role of PISD in tumor regulation, where the tumor repressor LACTB downregulates PISD levels, leading to the alteration of mitochondrial lipid metabolism and differentiation of certain cancer cells 35 . Therefore, the use of therapeutic compounds targeting human PISD may serve as an effective strategy for treating related diseases, including cancer. The structural and functional information from E. coli PSD can be extended to understand and predict the biological behaviors of mammalian homologs. Cloning. Plasmid encoding wild type E. coli psd gene was purchased from NBRP, Shizuoka, Japan (Resource No. JW4121-AM) 48 . To generate full-length PSD expressing vector, pLATE31-PSD 1-322 , the gene was amplified by the following primers: PSD_ECOLI_1-322_pLATE31 and PSD_ECOLI_1-322_pLATE31 and cloned into pLATE31 vector with a C-terminal His 6 tag using aLICator LIC Cloning and Expression Kit (Thermo Scientific, Massachusetts). To generate pLATE31-PSD 1-287 encoding C-terminally truncated psd gene with 1-287 residues, DNA fragments were amplified from the pLATE31_PSD 1-322 by PSD_ECOLI_1-322_pLATE31_For and PSD_ECOLI_1-287_pLATE31_Rev and cloned into the pLATE31 vector. The sequences were verified by DNA sequencing at the Macrogen (Seoul, South Korea). All site-directed point mutations were introduced to full-length psd by QuikChange II XL Site-Directed Mutagenesis Kit (Agilent Technologies, California). Primers used in this study are listed in Supplementary table S1. www.nature.com/scientificreports/ eluted PSD was concentrated using a 3 K MWCO 15 mL or 0.5 mL Amicon Ultra centrifugal filter (Millipore, Massachusetts), flash-frozen in LN2, and stored at -86ºC until used. The wild type and mutants PSD 1-322 were expressed and purified identically. For selenomethionine (SeMet) substitution, the E. coli BL21(DE3) transformed with pLATE31-PSD 1-287 vector was grown in M9 media supplemented with 100 μg/mL ampicillin at 37 °C to OD 600 = 0.2. Methionine biosynthesis was inhibited by the addition of lysine, phenylalanine, and threonine at 100 mg/L, isoleucine, leucine, valine, and SeMet at 50 mg/L 49 . Expression was induced by 0.1 mM IPTG 15 min after the addition of the amino acids. The cells were incubated and harvested as that of the native protein. The SeMet substituted PSD was purified similarly to that of the native protein with minor changes. Briefly, all buffers used in purification were degassed. The cells were lysed by sonication in buffer B except DTT replaced by 5 mM tris(2-carboxyethyl) phosphine (TCEP, Sigma-Aldrich). On the affinity purification and SEC, buffers were additionally supplemented with 0.2 mM TCEP.

Expression and purification.
Ligand conjugation by reduction for structural characterization. Schiff base reduction was performed as described by Li and Dowhan 19 with minor changes. Prior to the reduction, purified PSD 1- All crystals were cryoprotected by mother liquor supplemented with 20% (v/v) glycerol and flash-frozen in liquid nitrogen. X-ray diffraction data were collected under cryogenic conditions using the ADSC Q270 detector at Pohang Accelerator Laboratory (PAL; Pohang, South Korea) beamline 7A (SeMet, Apo-PSD1 and 2) or using Eiger X 9 M detector at PAL beamline 5C (8PE-and 10PE-bound PSD 1-287 ) (Table S1). SeMet data were indexed, integrated, scaled by HKL2000 50 . All native data were processed by XDS 51 and merged by Aimless 52 in CCP4 suite 53 . To solve the phase problem, the processed SeMet data were used for Single-wavelength Anomalous Dispersion (SAD) phasing by AutoSol 54 in Phenix suite 55 . Apo-PSD1 structure was built by iterative runs of real space model building using Coot 56 , and data refinement by REFMAC5 57 , or Phenix Refine 58 . The Apo-PSD1 structure was used as a search model of molecular replacement for determining other structures. After the main chains were built, pyruvoyl residues were modeled. Newly introduced pyruvate residues are linked with Thr-255 using JLigand 59 . To model 8PE-and 10PE-conjugated pyruvoyl residue, ACEDRG 60 was used to generate a molecular model and define structural restraints of lipid-conjugated pyruvate. PyMOL 61 was used to visualize the structures.

Multi-angle light scattering (MALS) analysis.
Purified E. coli PSD 1-287 were buffer exchanged to buffer E (30 mM Tris-HCl pH 7.5, 150 mM NaCl, supplemented with 0.05% (w/v) DDM), concentrated to 2.2 mg/mL, flash-frozen in LN2, and stored at − 86 °C until used. The concentrated protein was separated by Superdex 200 10/300 GL equilibrated with the degassed buffer E with a flow rate of 0.5 mL/min. The eluents were analyzed by MALS detector (Dawn Heleos II, Wyatt) equipped with differential refractive index (dRI; Optilab T-Rex, Wyatt) and UV detector. The signals were analyzed by ASTRA 6 program. Theoretical molecular weight and extinction coefficient of the www.nature.com/scientificreports/ for 1 h. After the ultracentrifuge, the supernatant (soluble fraction) and the pellet (membrane fraction) were collected. The pellet was lysed in 25 mL of 1 × PBS. Each fraction was analyzed by 12% SDS-PAGE and visualized by western blotting using HRP Anti-6 × His tag antibody (cat. no. ab1187, Abcam, UK). Original images of the membrane containing blots are shown in Supplementary Fig. S9.
LC-MS analysis of functional activity. The enzymatic reaction was initiated by mixing PSD 1-287 and soy PS in buffer A supplemented with 0.05% (w/v) DDM to a final concentration of 10 μM and 1 mM, respectively, and incubated in a water bath at 37 °C for 30 min. The reaction mixtures were quenched by adding an equal volume of 1 M HCl. Subsequently, lipids in the reaction samples were extracted by Folch's method 62 . Briefly, one volume of the quenched mixtures was transferred to 5 volumes of the chloroform/methanol (2:1, v/v) in a glass tube. The tube vortexed and centrifuged briefly at 500 rcf, 4 °C for 10 min. By carefully discarding the upper phase, the lower phase was collected in a new glass tube and dried out. The extracted lipids were dissolved in methanol and subjected to LC-MS analysis. An aliquot of 20 μL of the sample was injected into a reversed phase HPLC column (ZORBAX RR Eclipse Plus C18, 95 Å, 4.6 × 100 mm, 3.5 µm; Agilent Technologies) attached to Agilent 1260 Infinity Quaternary LC system (Agilent Technologies). Chromatographic separation was performed by using Solvent A (10 mM ammonium acetate in acetonitrile:water 60:40 (v/v)), and solvent B [10 mM ammonium acetate in isopropanol:acetonitrile 90:10 (v/v)], using a gradient step with the flow rate of 0.8 mL/ min as following: (1) 70% B for 1 min (column equilibration); (2) a linear gradient from 70 to 100% B for 11 min; (3) a hold at 100% B for 1 min (column wash); (4) a linear gradient from 100 to 70% B for 1 min; (5) a hold at 70% B for 1 min (column equilibration) with additional hold for 5 min using Post time. Eluted fractions were loaded to the MS and MS/MS system (Agilent 6520 Q-TOF LC/MS; Agilent Technologies) and analyzed in the positive ion mode utilizing Dual ESI as an ionization source. Major phospholipids were identified and relatively quantified from MS and MS/MS spectra by the MassHunter program with the aid of LipidBLAST 63 . Results were visualized by GraphPad Prism 64 . Functional assay of the PSD 1-322 was performed in the same manner as the PSD 1-287 with final enzyme concentrations for the reaction reduced to 1 μM. For conversion rate analysis of wild type and mutants PSD 1-322 , 1 μM of the enzyme is incubated with 1 mM 16:0/18:1 PS in a 37 °C water bath for 0, 5, 10, 15, and 20 min. Following extraction of phospholipids and LC-MS analysis were performed as described above.
Ligand conjugation by reduction of His-144 mutants. Conjugation assay was performed similarly as used for the structure characterization described above with minor changes. Buffer exchanged PSD 1-322 variants (H144A and H144N) were concentrated to 0.431 mg/mL (H144A) or 0.901 mg/mL (H144N). The protein, 5 M NaCNBH 3 , and 10 mM 8PE/10PS/14PS were sequentially mixed with a volume ratio of 89:1:10 while vortexing. The mixture was incubated in a water bath at 37 °C for 15 h and analyzed by MALDI-ToF. www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.