Inhibition of fatty acid synthesis by synthesized fatty acids

While FA synthesis, we found formation of tiny aggregates (Fig. 1e). The aggregates were precipitated by centrifugation and analyzed by SDS-PAGE, revealing that those were mostly FabZ (Fig. 1f). The aggregates did not appear when the essential enzyme for FA synthesis (e.g., ACP) was omitted (Fig. 2e, f, Supplementary Fig. 4) where FA synthesis did not occur. This implies the synthesized FAs induced the abnormal structural formation of FabZ.

Open in a separate window

Fig. 2

Phospholipid synthesis by cell-free synthesized acyltransferases.

a Schematic of the phospholipid synthesis pathway in a cell-free system. b SDS-PAGE image of cell-free synthesized PlsX and PlsY. Confocal microscopy images of c GFP-PlsX and d PlsY-GFP synthesized inside giant unilamellar vesicles (GUVs) in the presence (bottom) or absence (top) of Ficoll. The percentage of POPG (mol%) in the lipid composition of the GUV is described above the images. e Synthesis of lysophosphatidic acids (LPAs) by PlsX and PlsY synthesized in a bulk cell-free system. f SDS-PAGE image of cell-free synthesized PlsX, PlsY, and PlsC. g Synthesis of phosphatidic acids (PAs) by cell-free synthesized PlsX, PlsY, and PlsC. At 30min, the resources [acetyl-CoA (Ace-CoA), malonyl-CoA (Mal-CoA), and NAD(P)H] (red) or buffer (black) were additionally supplied to the reaction mixture. h Increase of PA synthesis by supplying the resources at 30 and 60min (blue) or only 30min (red). The slopes indicate the transition in PA productions between 60 to 120min. i Effect of the various concentrations of the resources supplied from the beginning of PA synthesis reaction. The concentration of 1.0× [Ace-CoA, Mal-CoA] is 2 and 4mM, respectively. Each dot on the graphs represents individual experimental data. Double asterisk, single asterisk, and n.s. indicate that P values are less than 0.01, P values are less than 0.05, and P values are more than 0.05, respectively. Error bars indicate the standard deviation of at least triplicate measurements.

Given the results of the reaction arrest and the enzyme aggregations, we thought that the synthesized FAs impede the FA synthesis system. To verify this, we tried to synthesize FA in the presence of FA (C18:1, oleic acid) at various concentrations. The yield of FA synthesis drastically decreased as the concentration of oleic acid increased; the yield reduced to 25% when more than 200µM of oleic acid was added (Fig. 1g). Conversely, the yield increased up to 1.5-fold when 150µg/mL of liposomes (small size membrane vesicles) were added during the FA synthesis (Fig. 1h). The addition of liposomes also tended to suppress the formation of aggregates even after they formed (Supplementary Fig. 5). Therefore, we envisaged that liposomes may work as a reservoir for the synthesized FAs and reduce the concentration of free or micelle FAs. A series of results obtained indicate that the synthesized FAs induce protein aggregation and disrupt their functional activities. This may be the reason why the reconstructed FA synthesis system exhibited limited productivity in the previous study²². Such unfavorable effect of FAs has been reported also in other proteins²⁴. The addition of liposomes may lead to the extension of the enzyme’s lifetime, thus resulting in more FA synthesis.

If the same amounts of FAs were synthesized within a space of membrane vesicle (e.g., diameter 30µm) as the in vitro reaction (300µM), this brings that only 15% of the lipids composing the vesicle are produced (Supplementary Text 1). Even worse, not all of the FAs integrate into the vesicle membrane, because FAs have a high critical micelle concentration. Therefore, we concluded that, in order to build self-growing artificial cells, it is necessary to reconstruct the mechanism for producing more stable and inoffensive lipids and integrating them into the membrane.

Design of artificial phospholipid synthesis system combining in vitro fatty acid synthesis and cell-free system

So far, the synthesized FAs were released from ACP by TesA, thus the resulting free FAs in the aqueous reaction solution led to the inhibition of Fab enzymes. To avoid this, we designed the system to drive the reaction towards phospholipid synthesis. In bacteria, TesA does not exist in the cytosol, it locates in periplasm space. Newly synthesized FAs are directly delivered to the cell membrane in the form of FA-ACP (acyl-ACP). On the surface of the cell membrane, acyl-ACP is converted into acyl-PO₄ by PlsX, then acyl-PO₄ binds to a glycerol-3-phosphate by PlsY to form LPA^25,26 (Fig. 2a). If PlsX and PlsY are added to the reconstructed FA synthesis system, it is expected that the synthesized FAs are sequentially converted into LPAs, avoiding the risks of enzyme inactivation and poor membrane localization. However, these enzymes are membrane proteins that require the use of detergent during purification. Detergents melt vesicles. Therefore, we decided to generate PlsX and PlsY from their corresponding genes inside the membrane vesicles to synthesize phospholipids in cooperation with FA synthesis.

For synthesizing PlsX and PlsY, we used a reconstructed cell-free system, the PURE system³. Both proteins were successfully co-synthesized in a bulk (test tube) condition (Fig. 2b), resulting in about 1µM protein production (Supplementary Fig. 6). Note that the DnaKJE chaperone was supplied to the PURE system because the solubility of PlsX is highly dependent on DnaKJE²⁷. The DnaKJE is the mixture of DnaK, DnaJ, and GrpE chaperons of E. coli. When liposomes were supplied as localization place during the protein synthesis, 70% of the synthesized PlsX and PlsY were collected in the supernatant fraction after centrifugation, suggesting a large part of the proteins spontaneously localized to the liposome membrane (Supplementary Fig. 7).

Membrane localization of acyltransferases synthesized in artificial cells

To visualize the membrane localization, we synthesized PlsX and PlsY as GFP-fusion proteins inside artificial cells. We encapsulated the PURE system with the corresponding genes inside vesicles, then incubated so prepared artificial cells. GFP-PlsX was synthesized well, however, they did not localize on the vesicle membrane (Fig. 2c and Supplementary Fig. 8). Since PlsX is a peripheral membrane protein that requires a negative charge on the membrane surface for the localization^28,29, we added 1-palmitoyl-2-oleyl-sn-glycero-3-phospho-rac-(1-glycerol) (POPG) into the lipid composition of the artificial cell membrane in the range of 0–50mol%. We found a weak tendency of membrane localization at 25mol% POPG, and a large part of the GFP-PlsX localized to the membrane at 50mol% POPG (Fig. 2c and Supplementary Fig. 8). This tendency became more significant when Ficoll was together encapsulated inside the vesicles. Ficoll is a bulky polymer that is often used for inducing a molecular crowding effect inside vesicles or water-in-oil emulsions³⁰. The effect of Ficoll for the membrane localization of peripheral proteins has been observed also in the case of FtsZ⁶. When Ficoll was added, fluorescence dots were observed even at 0 or 10mol% POPG (Fig. 2c and Supplementary Fig. 8), indicating non-regulated polymerization of GFP-PlsX. Conversely, at 25mol% POPG, most proteins localized onto the membrane. And, at 50mol% POPG, visually all proteins localized onto the membrane. These results showed that the degree of membrane localization of PlsX is substantially enhanced both by the negatively charged membrane surface and the molecular crowding effect. In contrast to PlsX, PlsY localized onto the membrane due to the hydrophobic interaction. We observed the efficient membrane localization of PlsY-GFP that was synthesized in the presence of Ficoll inside giant vesicles containing 50mol% POPG (Fig. 2d and Supplementary Fig. 9).

Lysophospholipid synthesis by coupling fatty acid synthesis and acyltransferases

To synthesize LPA, we combined the FA synthesis system and the cell-free system synthesizing acyltransferases, PlsX and PlsY. Both acyltransferases were first synthesized in a bulk PURE system, supplemented with liposomes consisting of POPC and POPG (50:50mol%), then mixed with Fab enzymes. The LPA synthesis was initiated by the addition of acetyl-CoA, malonyl-CoA, and NAD(P)H. We detected the synthesis of LPAs containing several types of hydrocarbon chains (Fig. 2e). The concentration of total LPAs reached 120µM. This means that about 65% of the malonyl-CoA and 20% of the acetyl-CoA input were converted into LPA (as LPA 18:0). We also succeeded in synthesizing PlsX and PlsY in the presence of Fab enzymes, followed by the addition of the substrates to initiate LPA synthesis. The obtained data support that the cell-free synthesized acyltransferases could maintain the LPA synthesis activity on the lipid membrane in collaboration with FA synthesis. However, unfortunately, the yield of LPA (Fig. 2e) was lower than that of FA synthesis (Fig. 1c, d, g, h). We supposed that this is due to the functional regulation of PlsY by the synthesized LPA³¹ to prevent the accumulation of LPA at the cell membrane. To avoid this regulation and increase productivity, we next proceeded to the synthesis of phospholipids, which are more stable and harmless lipids.

Phospholipid synthesis by coupling fatty acid synthesis and acyltransferases

The first phospholipid, phosphatidic acid (PA), is synthesized from LPA and acyl-ACP by PlsC (Fig. 2a), thus we additionally added the gene of PlsC in the cell-free mixture to synthesize PA. When PlsX, PlsY, and PlsC were co-synthesized (Fig. 2f), 150µM PA was synthesized at 30min (Fig. 2g) but the yield did not increase after that. This is supposedly due to the depletion of the resources (acetyl-CoA, malonyl-CoA, and NAD(P)H) because the amount of PA synthesized is close to the theoretical maximum that is estimated by calculation of substrate consumption. When we supplemented additional resources at the 30min as same as the initial concentrations, the PA yield increased up to 250µM in the next 30min (Fig. 2g). Additionally, when the resources were again supplemented at the 60min, the PA yield further increased to 350–400µM at 90–120min (Fig. 2h). Note that one PA contains two acyl-chains, thus the maximal yield of the synthesized PA is equivalent to 700–800µM FA or LPA. This is a remarkable enhancement compared to that of FA or LPA synthesis in the previous scheme. As the PA synthesis was still increasing even after the second dose, the productivity of phospholipid synthesis in the constructed system may be governed by the concentration of acetyl-CoA and malonyl-CoA. In this experiment, 2.1mM phospholipids, which are forming 1.3nM liposomes, were added in the reaction mixture, thus the synthesized 400µM PA corresponds to approximately 20% of the liposome-forming lipids. Interestingly, when we added 1.5 or 2 times more resources at the beginning of the reaction, the increase in PA synthesis was not observed and even slightly decreased (Fig. 2i). This result indicates that maintaining the optimum concentration of the substrates is important to maximize the yield and the excess substrate negatively affects lipid synthesis. We concluded that the smooth flux of sequential reaction starting from FA synthesis to phospholipid synthesis can avoid the accumulation of synthetic intermediates and the reaction suppression by those negative feedback, thus maximizing the yield of lipid synthesis.

Recycling of CoA towards sustainable phospholipid synthesis

The constructed cell-free phospholipid synthesis system uses acetyl-CoA and malonyl-CoA as hydrocarbon sources. This system also produces, however, CoAs as the by-products while phospholipid synthesis: for example, 18 CoAs will be produced per one PA (18:0 or 18:1) synthesis. This is not a negligible problem when we try to synthesize sufficient phospholipids in artificial cells for reproducing the self-reproduction. A considerable amount of CoA will be accumulated. To avoid this problem and make the system more sustainable, we implemented an additional system that recycles CoA for the production of acetyl-CoA and malonyl-CoA in the constructed cell-free system. We purified acetyl-CoA synthetase (ACS) and acetyl-CoA carboxylase (Acc) ABCD (Fig. 3a). ACS synthesizes acetyl-CoA from acetic acid and CoA using ATP, and AccABCD synthesizes malonyl-CoA from acetyl-CoA and bicarbonate. The specific activities of ACS and AccDA (in the presence of 10µM AccBC) were 425±79 and 300±65µmol/min/mg, respectively (Supplementary Fig. 10).

Open in a separate window

Fig. 3

Phosphatidic acid synthesis by synthesizing acetyl-CoA or/and malonyl-CoA.

a An SDS-PAGE image of the purified acetyl-CoA carboxylase DA (AccDA), acetyl-CoA carboxylase BC (AccBC), and acetyl-CoA synthase (ACS). The positions of molecular size are described on the left side of the gel image. The purified enzymes were introduced into the cell-free system synthesizing acyltransferases and fatty acids. b Phosphatidic acid synthesis through the synthesis of acetyl-CoA by ACS. The reaction was initiated by the addition of CoA and various concentrations of malonyl-CoA. c Phosphatidic acid synthesis through the synthesis of malonyl-CoA by AccABCD. The reaction was initiated by the addition of various concentrations of acetyl-Co, which was converted to malonyl-CoA or directly consumed as the substrate of FA. d Phosphatidic acid synthesis through the synthesis of acetyl-CoA and malonyl-CoA by the addition of ACS and AccABCD. The reactions were initiated by the addition of various concentrations of CoA. Each dot on the graphs represents individual experimental data. Error bars indicate the standard deviation of at least triplicate measurements. FA fatty acid, PA phosphatidic acid, Ace-CoA acetyl-CoA, Mal-CoA malonyl-CoA.

When we introduced the ACS in the cell-free system and supplied CoA instead of acetyl-CoA, we detected successful synthesis of PA (Fig. 3b). This means that acetyl-CoA was synthesized from CoA by ACS and sequentially consumed as the substrate for FA synthesis. We found that the whole steps (acetyl-CoA synthesis, FA synthesis, and phospholipid synthesis) were completed within 10min and the maximum yield was obtained when 4mM malonyl-CoA was supplied. We also found a decrease in PA products after 10min when 4 or 8mM malonyl-CoA was supplied. Although the reason is unknown, this is probably due to the reverse activity of the acyltransferase. When we introduced AccABCD and supplied acetyl-CoA instead of malonyl-CoA, we also detected the PA synthesis (Fig. 3c). In this condition, a part of acetyl-CoA was converted to malonyl-CoA and consumed as the substrate for FA synthesis, together with remaining another part of acetyl-CoA. We found that the maximum yield was obtained when 8mM acetyl-CoA was supplied. The rate of PA synthesis was not as fast as that of the reaction synthesizing acetyl-CoA, suggesting that AccABCD may be the rate-determining factor. Finally, successful PA synthesis was detected when both ACS and AccABCD were added to the cell-free system. CoA was supplied instead of acetyl-COA and malonyl-CoA (Fig. 3d). The yield and rate of PA synthesis showed a pattern similar to that of the reaction synthesizing malonyl-CoA. No PA peaks were detected by LC/MS analysis when ACS, AccDA, or AccBC was removed from the cell-free system (Supplementary Fig. 11).

We did not add acetic acid as carbon source because the cell-free system (PURE system) contains enough acetate as a counter ion of the magnesium³. The PURE system also contains ATP and an energy regeneration system³, but the supplement of ATP increased the PA productivity (Supplementary Fig. 12). PA synthesis was observed even without the addition of HCO₃⁻ (Supplementary Fig. 12). This is because CO₂ dissolved in the reaction solution was being converted to HCO₃⁻ and used for malonyl-CoA synthesis. In total, the net amount of CO₂ (HCO3^-) remains constant although it was consumed by AccC³², because CO₂ was generated from the FA elongation step by FabB or FabH (Supplementary Fig. 13).

As shown above, the addition of ACS and AccABCD expanded the cell-free phospholipid synthesis system allowing for the production of the substrates of FA by recycling CoA. The produced substrates were consumed for the FA synthesis and, sequentially, used for the PA synthesis by the cell-free synthesized acyltransferase (Supplementary Fig. S13). Although the yield of PA synthesis (Fig. 3d) was reduced compared to the reaction using acetyl-CoA and malonyl-CoA, we decided that the constructed cell-free system enables the synthesis of phospholipids in the closed space of artificial cells.

Overexpression and protein purification

All fatty acid binding (Fab) enzymes, acyl carrier protein (ACP) and thioesterase I (TesA) were overexpressed (Supplementary Table 3) and purified as previously reported by Yu et al.²² with some modifications. Briefly, for FabA/B/D/F/G/H/I/TesA, the transformed cells were cultivated in 1L LB medium supplied 50µg/mL kanamycin at 37˚C. At OD₆₀₀ 0.6–0.8, 0.1mM Isopropyl-β-D-thiogalactoside (IPTG) was introduced for induction, and the cultivations were continued for 3h. The cells were then collected by centrifugation 5500×g for 20min at 4˚C, washed with Buffer A (50mM Tris-HCl (pH 7.6), 2mM dithiothreitol (DTT), 10% glycerol), and stored at −80˚C until use. The frozen cells were dissolved in 35mL Buffer A on ice and disrupted by sonication (5s on, 3s off power 40%, total 5min). The unbroken cells and debris were removed by centrifuging 30k×g for 1h at 4˚C. The resulting supernatants were collected and injected into AKTA™ pure equipped HisTrap HP column (GE Healthcare), which was pre-equilibrated with Buffer A. The 6HisTagged proteins were eluted by a gradient of Buffer B (Buffer A with 1M Imidazole) from 3% to 50%. The target proteins were collected and desalted using Amicon (with their corresponding cutoff size) (Millipore) before subjecting to the HiTrap Q column (GE Healthcare). The injected proteins were washed and eluted by a gradient of Buffer C (Buffer A with 1M NaCl) from 0% to 100%. Finally, the proteins were concentrated by Amicon and their concentration was determined with Bradford (Takara Bio) or BCA (Thermo Fisher Scientific) protein assay kit. In general, the purified proteins are stable for at least a few months at −80˚C.

ACP was co-expressed with a 4′-phosphopantetheinyl transferase (SFP) because ACP needs post-translational modification to form Holo-type ACP (Holo-ACP). The source of the sfp gene is Bacillus subtilis, and the codon was optimized for the E. coli expression system when we ordered it as a synthetic gene (Supplementary Table 4). The gene was inserted between NcoI and BamHI sites of the pCDF-1b (Novagen) vector using the primers sfp-fw and sfp-rv (Supplementary Table 5), resulting in pCDF-sfp (Supplementary Table 3). The E. coli BL21 (DE3) cells harboring pTL14, which contains the acp gene, and pCDF-sfp were cultivated in TB (Terrific Broth) medium supplied kanamycin and streptomycin until OD₆₀₀ arrived at 0.6. After the induction with 1mM IPTG, cells were collected at OD₆₀₀ 1.9 and washed with 50mM HEPES-KOH (pH7.6). The obtained cell pellets were dissolved in Buffer A and disrupted by French Press at 500bar (7000psi) three times. After removing cells unbroken and debris by centrifugation, the supernatant was subjected HisTrap HP column as in the case of the Fab proteins, using the Buffer A and B containing 300mM NaCl. The target proteins were collected and desalted before subjecting MonoQ 4.6/100PE column (GE Healthcare) using Buffer A and C as shown in Fab proteins. The collected target proteins were desalted by nap-5 and concentrated by Amicon. Concentrations of the protein were determined by Protein Assay (BioRad). MALDI-TOF-MS was used to confirm the Holo-type of the purified ACP.

Because FabZ tends to aggregate, we modified the expression and purification method as follows. The expression plasmid pXY-FabZ was introduced to E. coli C43 (DE3) strain cells. The cells were cultivated at 37˚C in 1L LB medium supplied kanamycin until the OD₆₀₀ 0.6, then, 1mM IPTG was added followed further 2h of cultivation. The resulting cells were collected and washed with 50mM HEPES-KOH (pH 7.6), then stored at −80˚C until use. The cell pellets were resuspended in Buffer D (100mM potassium phosphate (pH 7.4), 300mM NaCl, 1mM DTT) and disrupted by French Press as same as ACP. After removing cells unbroken and debris, the supernatants were subjected HisTrap HP column with Buffer D and E (Buffer D with 1M Imidazole). The FabZ was eluted with a gradient of Buffer E from 10% to 100%. The fractions containing FabZ were collected and desalted by repeating dilution with Buffer D and concentrating by Amicon. Note that the purified FabZ should not be concentrated to a high concentration to prevent aggregation. We did not proceed to further purification because FabZ is unstable, and only HisTrap purification is enough for purity. When determination of the FabZ concentration by the Protein Assay is difficult, the concentration was defined by comparing the band intensity of FabZ on SDS-PAGE with the standard protein which concentration had already been determined. Interestingly, the aggregated FabZ was dissolved by mixing with ACP. This might be due to the complex formation of FabZ and ACP that has been reported by Zhang et al.³⁸. Therefore, FabZ should be stored as a mixture with ACP in the molar ratio of 1:30 (FabZ: ACP).

Acetyl-CoA synthetase (ACS) was overexpressed from the synthetic gene (Supplementary Table 4) located between NdeI and XhoI sites of the pET-28a vector, resulted in pET28a-acs, introduced in E. coli C43 (DE3) cells, referring to a previous report³⁹. Cells were cultivated at 37 ˚C in LB medium supplied kanamycin until the A₆₀₀ 0.4, then, 0.4mM IPTG was added followed by additional three hours of cultivation. The resulting cells were collected and washed with 100mM potassium phosphate (pH 7.4), then stored at −80˚C until use. The frozen cell pellet was dissolved in Buffer D and disrupted by French Press as in the case of ACP. After removing cells unbroken and debris, the supernatants were subjected by AKTA Avant 25 (GE Healthcare) equipped HisTrap HP column with Buffer D and E. The ACS was eluted by the gradient of Buffer E from 10% to 50% for 20min by a flow rate 2.5ml/min. The fractions containing ACS were collected and desalted by repeating dilution with Buffer E and concentrating by Amicon 30kDa cut off. The resulting sample was further purified by MonoQ 4.6/100PE column (GE Healthcare) using Buffer F (50mM Tris-HCl (pH 8.0), 10% glycerol) and Buffer G (Buffer F with 1M NaCl). ACS was eluted by the gradient of Buffer G from 0% to 100% for 20min by a flow rate 1.0ml/min. The collected ACS was desalted and concentrated using Amicon. We did not cut off HisTag because the protein aggregates.

AccBC complex was purified as previous reports³², with slight modifications. The polycistronic genes for AccB and AccC were obtained as a synthetic gene (Supplementary Table 4) with codon optimization for the E. coli expression system (FASMAC), which inserted between NdeI and XhoI sites of the pET28a vector, resulted in pET28a-accBC. In this construct, a 6HisTag was inserted at the N-terminus of the accB gene. The gene of birA, biotin-[acetyl-CoA-carboxylase] ligase, is also codon-optimized and inserted in the pCDF-1b vector using NcoI and XhoI sites, resulted in pCDF-birA. Both pET28a-accBC and pCDF-birA were co-introduced into C43 (DE3) E. coli cells (Lucigen) and then cultivated at 37 ˚C in LB medium supplemented with kanamycin and streptomycin. At the OD₆₆₀ 0.4, 0.5mM IPTG and 2µM biotin were added, then, followed by further cultivation for four hours. Cells were collected and dissolved in Buffer H (20mM potassium phosphate (pH 8.0), 500mM NaCl, 2-mercaptoethanol, protease inhibitor cocktail), then disrupted by French Press as described above. After removing cells unbroken and debris, the lysate was applied to the HisTrap HP column with Buffer H and I (Buffer H with 1M imidazole). AccBC protein was eluted as complex by the gradient of Buffer I from 2% to 100% for 20min by 5.0ml/min flow rate. The fractions of AccBC were collected and desalted using Amicon (10kDa cut off) and Buffer J (50mM Tris-HCl (pH 8.0), 100mM NaCl, 10% glycerol). The desalted proteins were further purified by the MonoQ column with Buffer J and K (Buffer J with 1M NaCl). AccBC was eluted by the gradient of Buffer K from 0% to 100% for 20min by 1.5ml/min flow rate. The collected AccBC fractions were buffer exchanged for Buffer L (20mM Tris-HCl (8.0), 300mM NaCl, 10% glycerol) while being concentrated by Amicon (10 KDa cut off). The concentration was determined by the BCA protein assay kit. Finally, the efficiency of biotinylation of the purified AccB biotin carboxyl carrier protein of acetyl-CoA carboxylase was confirmed. One microgram of AccBC sample was dissolved in 25.5µL 50mM ammonium bicarbonate (pH 8.0), then 1.5µL fresh 500mM DTT was added. After boiling at 95˚C for five minutes, 3µL 500mM iodoacetamide was added, followed by a 20min incubation at room temperature. For digestion of the protein, 0.5µL 5ng/µL Gluc-C (Promega) was added and incubated for three hours at 37˚C, followed by additional supplying 0.5uL Gluc-C and incubation at 30˚C overnight. The digestion reaction was terminated by the addition of 3µL of 50% trifluoroacetic acid (TFA). The sample was then purified by a C18 spin column (Thermo Fisher) following the manufacturer’s protocol. The eluted sample was dried up and dissolved in 5% acetonitrile solution supplied 0.5% TFA, before analyzing by NanoLC-MS (Thermo Fisher). From the obtained data, the area intensity of the biotinylated fragment (AMkMMNQIE, m/z=661.29232, z=2, where k=biotinylated lysine residue) was compared with that of a non-biotinylated fragment (AMKMMNQIE, m/z=548.25360, z=2). Generally, more than 50% of the purified AccB was biotinylated.

AccDA was expressed from pACS275³² introduced in C43 (DE3) E. coli cells. pACS275 contains the genes of accD, where a 6HisTag sequence was added at the C-terminus, and accA. The protein purification was carried out following a previous report³², with slight modifications. The cells were cultivated at 37 ˚C until the OD₆₆₀ reached 0.5, then 0.5mM IPTG was added, then cells were further cultivated at 25˚C for 4h. Cells were collected and washed with Buffer M (20mM Tris-HCl (pH 8.0), 500mM NaCl, 10% glycerol). For disruption of the cells, 2µL 250 U/µL Benzonase (Merck) and 8mg lysozyme were added into a 40mL cell solution and kept on ice for one hour. After three times freeze-and-thaw, unbroken cells and debris were removed by centrifugation and the supernatant was passed a 0.45µm filter. The resulting lysate was supplemented 25mM (final concentration) imidazole, then subjected by the HisTrap HP column with Buffer M and N (Buffer M with 1M imidazole). The sample was washed with 5% Buffer N for 10 column volume by flow rate 5.0mL/min, then eluted by the gradient of Buffer N from 5% to 25% for 20 column volume. AccDA fractions were collected and diluted five times with Buffer O (20mM Tris-HCl (pH 8.0) and 10% glycerol). The resulting sample was loaded to the HiTrap Capto Q column (Cytiva) and washed 10% Buffer P (Buffer O with 1M NaCl) for 10 column volume. AccDA sample was eluted by the gradient of Buffer P from 10% to 100% for 20 column volume. The collected fractions were concentrated and buffer exchanged with Buffer L using Amicon (10kDa cut off).

The compositions of all buffers used for protein purifications are shown in Supplementary Table 6.

Liposome preparation

Small size liposomes were used for in vitro phospholipid synthesis reaction. 40mg POPC powder or a mixture of 20mg POPC and 20mg POPG powder was dissolved in chloroform within a round-bottom 50mL flask and briefly processed by bath sonication. Then chloroform was removed by rotary evaporator and placed under low pressure within a desiccator for overnight to remove the residual chloroform completely. The resulting lipid film was dissolved with 1mL 50mM HEPES-KOH (pH 7.6) by vortexing and sonicating. The lipid solution was next processed by micro extruder using 1µm and 200nm pore size polycarbonate membrane in a stepwise fashion. The prepared liposome solution was divided into micro tubes and stored at −80˚C. The liposome sample is generally stable for at least years, but it should be treated by a bath sonication before use.

In vitro fatty acid synthesis

In vitro fatty acid synthesis reaction was prepared as described in Supplementary Table 7, based on the report by Yu et al.²². ¹³C-stable isotope-labeled acetyl-CoA and malonyl-CoA were used as substrates to distinguish the products from environmental fatty acids. The reaction was generally carried out at 30˚C for 30min. The resulting reaction mixtures were diluted to 20-fold with 50% MeOH and centrifuged at 20,000×g for 30min at 16˚C. The resulting supernatants were collected and placed in glass vials. Synthesized fatty acids were analyzed by Shimadzu LCMS-2020 system equipped with InertSustain^Ⓡ Phenylhexyl column (3µm, 2.1mm Cat. No. 5020-89128, GL Science, Japan) with a guard column (GL Science, Japan). Fatty acids were separated in 60% B isocratic elution mode using mobile phase A (20mM ammonium acetate) and B (100% acetonitrile). The column oven was 35˚C, and the flow rate was 0.14mL/min. Elution was analyzed by selected ion monitoring (SIM) mode monitoring m/z of 211.9 (C¹³12:0), 241.2 (C¹³14:0), 269.2 (C¹³16:1), 271.2 (C¹³16:0), 299.2 (C¹³18:1), and 301.3 (C¹³18:0) (Supplementary Table 8) in the negative-ion mode. For the quantification, the area counts obtained from the peaks were converted into fatty acid amounts using the standard curve made from the analysis of the commercially available standard fatty acid mixture with defined concentrations. The standard fatty acids samples were measured each time of the experiment. An example of the standard curve is shown in Supplementary Fig. 20.

Preparation of template DNA for cell-free protein synthesis

The genes expressed in the cell-free system (the PURE system) were summarized in Supplementary Table 9. All genes were prepared as plasmid DNA by In-Fusion cloning or linear DNA by PCR (Supplementary Table 10) using the designed primer sets (Supplementary Table 5) and template DNAs (Supplementary Table 4). The E. coli gene of plsX and plsY was obtained from K12 strain cells and inserted into the pET28a vector. Using the resulting plasmids, we introduced a gene for super-folder green fluorescence protein (sfgfp)⁷ to the upstream of plsX or the downstream of plsY by In-Fusion to generate a GFP-PlsX_wt or PlsY_wt-GFP plasmid, respectively (see Supplementary Table 9 #7&8). These were used for expression in GUV.

Because the productivity of PlsX in the PURE system was not so high, we reduced the GC-contents just after the initial codon by introducing silent mutations by PCR (see Supplementary Table 9 #1). We used this liner DNA for cell-free LPA or PA synthesis. In the same manner, we also prepared six histidine-tagged plsX for the analysis of western-blotting, using an anti-HisTag antibody (see Supplementary Table 9 #4). The E. coli gene of plsY for cell-free LPA or PA synthesis was obtained as a synthetic gene with codon optimization for the E. coli expression system (see Supplementary Table 9 #2). As same as plsX, a six histidine-tagged plsY was prepared by PCR (see Supplementary Table 9 #5). The E. coli gene of plsC was obtained as a previously prepared plasmid pBT302_plsC¹⁹ (see Supplementary Table 9 #3). This was used for cell-free PA synthesis and for preparing a six histidine-tagged plsC DNA (see Supplementary Table 9 #6).

Cell-Free Protein synthesis

Cell-Free gene expressions were performed using PUREfrex 2.0 (GeneFrontier, Japan) following the manufacturer’s protocol. In general, linear template DNA(s) of plsX, plsY, and plsC were introduced into the reaction mixture supplied liposomes, sucrose, and DnaKJE chaperon mixture (DnaKJE mix), which highly affects to the solubility of PlsX²⁷ (eSol: http://www.tanpaku.org/tp-esol/index.php?lang=en) (Supplementary Table 2). For the gene expression to LPA or PA synthesis, the custom Sol. I, which was adjusted to reduce the volume from 10uL to 6µL by GeneFrontier company, was used to make a space to add the additional components, such as glycerol-3-phosphate (G3P) and fatty acid enzymes mixture (FA mixture). The FA mixture was prepared as shown in Supplementary Table 2, of which 2µL was used to make 20µL of the reaction mixture. For the gene expression to synthesize PA in the CoA recycling system, ACS, AccBC, and AccDA were additionally introduced using the water space of the mixture. In all cases, gene expression reactions were carried out at 37˚C for 2h using a thermal cycler.

Solubility assay

The ratio of membrane localization of the synthesized PlsX and PlsY were assessed as in a previous report⁴⁰, using the plsX-6His and plsY-6His genes. Proteins were synthesized in the presence or absence of liposomes consisting of 50% POPC and 50% POPG. The proteins in the supernatant or precipitation fractions were analyzed by western blotting using an anti-HisTag monoclonal antibody-conjugating horseradish peroxidase.

In vitro phospholipid synthesis

After the gene expression, the reaction mixture was supplemented with NADH, NADPH, acetyl-CoA, malonyl-CoA, and potassium phosphate buffer as described in Supplementary Table 2. For LPA or PA quantification by LC/MS, we used [¹³C]-labeled Acetyl-CoA and Malonyl-CoA. LPA or PA synthesis was performed at 30 ˚C. The resulting mixture was pre-treated as same as a previous report¹⁸ before analyzing by LC/MS. Briefly, 5µL of sample synthesizing LPA or PA was diluted with 45µL or 495µL 100% MeOH solution including 2mM acetylacetone, respectively. The diluted sample was sonicated for 10min and then centrifuged at 16,000×g for 5min at 15˚C. The supernatant was passed a 0.2µm filter (Millex®-LG) and injected into LCMS-2020 (Shimadzu). An internal standard was added to assess lipid extraction efficiency, which contained 0.025ng/µL each of DPPA, POPA, and DOPA in MeOH.

LC/MS analysis for synthesized phospholipids

For separation of synthesized LPAs or PAs, a metal-free column L-column2 or L-column3 (CERI, JAPAN) was used to avoid adsorption of phospholipids on the metal surface of the column. The LC/MS analysis method is based on Blanken et al.¹⁸ with slight modifications. Mobile phase A (water with 0.05% ammonium hydroxide and 2mM acetylacetone) and B (80% 2-propanol, 20% acetonitrile, 0.05% ammonium hydroxide, and 2mM acetylacetone) were used for the isolation of LPA or PA. For LPA, column oven at 55˚C, flow rate 0.2mL/min, and %B 30 were applied. 5µL samples were injected, then eluted by a gradient from 30% to 100%B in 30min followed. Elution was directly sprayed into the MS to analyze the products in negative mode. The m/z values of 395.0 (LPA¹³C14:0), 393.2 (LPA¹³C14:1), 425.1 (LPA¹³C16:0), 423.1 (LPA¹³C16:1), 455.1 (LPA¹³C18:0), and 453.1 (LPA¹³C18:1) were monitored as ¹³C-labeled products in SIM mode as shown in Supplementary Table 8. For PAs, when L-colum2 was used, the column oven at 60˚C, the flow rate 0.2mL/min, and %B 50 was applied. When L-column3 was used, the column oven at 45˚C, the flow rate 0.15mL/min, and %B 50 was applied. In both cases, 5µL samples were injected, then eluted by a gradient from 50%B to 100%B in 30min followed. The m/z values of 679.5 (PA¹³C16:0/16:0), 677.5 (LPA¹³C16:0/16:1), 709.5 (LPA¹³C16:0/18:0), 707.5 (LPA¹³C16:0/18:1), 675.5 (LPA¹³C16:1/16:1), 705.5 (LPA¹³C16:1/18:1), 739.5 (LPA¹³C18:0/18:0), 737.5 (LPA¹³C18:0/18:1), and 735.5 (LPA¹³C18:1/18:1) were monitored as ¹³C-labeled products in negative SIM mode as shown in Supplementary Table 8.

Formation of giant vesicles by droplet transfer method

The formation of giant vesicles encapsulating the cell-free system was performed as described in Shimane et al.⁴¹. Briefly, a 50µL of lipid mixture dissolved in chloroform was transferred into a grass tube and dried up by blowing nitrogen gas during vortexing to obtain lipid film. For example, to make 80mol% POPC and 20mol% POPG vesicles, a lipid mixture containing 40mM POPC and 10mM POPG was used for the lipid film formation. The dried lipid film was dissolved in 500µL mineral oil by processing heating at 70˚C for 1min, followed by vigorous vortexing for 30s. The lipid-oil mixture was heated again for 1min and vortexed for 3min. So prepared lipid-oil mixture was used for making water-in-oil droplets by mixing with 20uL of inner solution and vigorous vortexing for 30min. The prepared droplets solution was transferred over a 300uL of the outer solution (the same composition as 1× PUREfrex2.0 buffer lacking tRNA mix, creatine phosphate, formyl doner, and NTPs; provided by GeneFrontier company) containing 300mM glucose in a 1.5mL microtube and, then, centrifuged at 10,000xg for 5min at room temperature. After removing the upper oil phase by pipetting, the fraction of the giant vesicles precipitated at the bottom of the tube was carefully collected for 20µL. To wash away the inner solution that leaked out from the droplets failing to form vesicles, the collected vesicle fraction was mixed with 180µL of outer solution and precipitated again. The quality and population of the formed giant vesicles were checked by inverted microscopy equipped with a phase-contrast unit (Nikon IX73).

Phospholipid synthesis inside giant vesicles

The internal reaction mixture was prepared as described in Supplementary Table 2 (see For PA synthesis inside GUVs), then mixed with 12% (w/v) Ficoll PM70. This was encapsulated inside giant vesicles as described above with the outer solution. The collected 20uL vesicle solution (i.e., artificial cells) was mixed with 6µL of a feeding solution; containing 1×PUREfrex2.0 buffer, 13mM ATP, 6mM glycerol-3-phosphate, 4.8mM CoA, 42ng/µL RNaseA, and 300mM glucose. Protein synthesis was performed at 37˚C for 3h. After the protein synthesis within the artificial cells, the internal phospholipid synthesis was initiated by the addition of 5µL of 48mM NADPH dissolved in the outer solution. Phospholipid synthesis was carried out at 37˚C for 1–9h. The resulting 30µL artificial cell sample (1µL was used for microscopy check) was mixed with 5µL of 170ng/µL GFP-Spo protein to stain the vesicle membrane and, then, incubated at 37˚C for 30min followed by an additional 30min incubation at room temperature (25˚C). The artificial cells were observed by confocal microscopy (Nikon A1R system).

Confocal microscopy observation

Artificial cells were observed by Nikon confocal microscopy system A1R. For the observation using GFP-Spo, a 488nm laser was used with the setting as HV(GaAsP): 35, offset: 0, Intensity: 3.5, scan size: 1024, and scan speed: 0.5 frame/sec (Pixcel Dwell: 0.97us). For the NADPH permeability test, a 561nm laser was used with the setting as HV(GaAsP): 200, offset: 0, Intensity: 80, scan size: 1024, and scan speed: 0.5 frame/sec (Pixcel Dwell: 0.97us). For the CoA permeability test, a 488nm laser was used with the setting as HV(GaAsP): 20, offset: 0, Intensity: 2.0, scan size: 1024, and scan speed: 0.5 frame/sec (Pixcel Dwell: 0.97us). In all cases, images were taken as a set with a differential interference contrast (DIC) image.

Statistics and reproducibility—Error bars indicate standard deviation of at least triplicate measurements. t test was performed by Excel and applied to show statistical significance using of at least triplicate measurements. Exact number of replicates are shown in Supplementary Data 1.

The expression plasmids for Fab enzymes, ACP, and TesA were kindly supplied by Prof. Chaitan Khosla (Stanford Univ.). The expression plasmid for AccDA was kindly supplied by Prof. John E. Cronan (Univ. of Illinois). Dr. Takashi Kanamori (GeneFrontier) supplied the PURE system. We thank Dr. Yoshi Kawai (Newcastle Univ.), Dr. Taku Oshima (Toyama Pref. Univ.), and Dr. Ken Takai (JAMSTEC) for valuable discussion, and also thank Mr. Gaku Sato and Dr. Shigeru Shimamura for assisting experiments. This work was supported by the Human Frontier Science Program (RPG0029/2020 to Y.K.), JST PRESTO (JPMJPR18K5 to Y.K.), JSPS KAKENHI (16H06156, 16KK0161, 16H00797, 26119704, 21H05156 to Y.K.), Astrobiology Center Project of the National Institutes of Natural Sciences (AB291017 to Y.K.), and an internal Grant-in-Aid from Earth-Life Science Institute (to Y.K.).