Cell-Free Phospholipid Biosynthesis by Gene-Encoded Enzymes Reconstituted in Liposomes

The goal of bottom-up synthetic biology culminates in the assembly of an entire cell from separate biological building blocks. One major challenge resides in the in vitro production and implementation of complex genetic and metabolic pathways that can support essential cellular functions. Here, we show that phospholipid biosynthesis, a multiple-step process involved in cell membrane homeostasis, can be reconstituted starting from the genes encoding for all necessary proteins. A total of eight E. coli enzymes for acyl transfer and headgroup modifications were produced in a cell-free gene expression system and were co-translationally reconstituted in liposomes. Acyl-coenzyme A and glycerol-3-phosphate were used as canonical precursors to generate a variety of important bacterial lipids. Moreover, this study demonstrates that two-step acyl transfer can occur from enzymes synthesized inside vesicles. Besides clear implications for growth and potentially division of a synthetic cell, we postulate that gene-based lipid biosynthesis can become instrumental for ex vivo and protein purification-free production of natural and non-natural lipids.


SUPPLEMENTARY TABLE A
, we estimated this fraction to represent ~15% of the total DPPA synthesized, which after correcting for the fact that only ~52% of internal standard DOPG is recovered leads to ~30%.
The mean and standard deviation of the normalized counts for each data point were calculated.
Concentrations were obtained by multiplying the normalized counts by the average concentration of DOPG and solving the line equation for concentration: where b is the slope. Concentration errors were calculated by using the variance formula: An alternative method was used to convert MS counts into lipid concentrations. First, a linear regression of the count ratios DPPA/DOPG to the known concentrations of DPPA was made: The slope b and its error σ b were extracted. We then calculated the actual value of the concentrations in a sample as: The error of DPPA counts was calculated as: Similar results were obtained as with the first method.

Calculation of increase surface area of liposomes upon incorporation of synthesized DPPA
To estimate the vesicle growth through synthesis and membrane incorporation of DPPA lipids, the following calculation was made. First the initial surface area, A, of a 400-nm-sized vesicle was calculated: Then the total number of lipids per vesicles, N lip./ves. , was calculated taking into account the two leaflets of the membrane. A cross-sectional area of 72.1 Å 2 corresponding to that of a DOPC molecule was assumed.
The concentration of vesicles is derived from the initial concentration of lipids, C lip. = 508 µM, and the number of lipids per liposome as: The concentration of synthesized DPPA was determined by kinetics experiments with both GPAT and LPAAT enzymes and a value of 26 µM was found (Figure 3e). Given that about 28% of total synthesized DPPA integrated in liposome membrane (Figure 4d), the concentration of membraneinserted DPPA, C PA , is 7 µM. Therefore, assuming homogenous partitioning of DPPA lipids between vesicles, the number of DPPA molecules per liposome is: Using a cross-sectional area per DPPA lipid of 50 Å 2 it is possible to calculate the total additional surface area as: A PA/ves. = 50 × (1e -10 ) 2 × N PA/ves. / 2.
The percentage area increase was calculated as A PA/ves. / A × 100 and a value of ~1% was found.

Estimation of the number of synthesized membrane proteins per liposome
Based on previous PUREfrex-based IVTT experiments we estimate the concentration of synthesized proteins to be in the order of 0.5 µM 1,2 . The precise concentration may vary from one protein to the other, the fraction of active proteins too. An approximation of the number of synthesized proteins and the number of reconstituted pathways per vesicle can be found assuming that (i) an amount of 500 nM protein is produced from a single-gene expression reaction, hence coexpression of 5 genes will generate ~100 nM of each specific protein (though a nonlinear relationship is expected but difficult to predict), (ii) 1/3 of proteins incorporates inside liposomes (see Figure 1c where the fraction of GPAT and LPAAT proteins co-purified with vesicles can be evaluated), (iii) proteins randomly insert into liposomes, such that the surface density of proteins is the same for every liposome, (iv) all membrane-bound proteins are active and (v) proteins can insert into the bilayer in a bi-directional manner (this might not be the case for PssA) limiting the fraction of active enzymes to 0.5. Under these conditions one liposome contains ~225 membrane proteins.
This means that every liposome potentially contains ~45 copies of the 5-protein pathways. Further quantitative investigations are definitely needed to refine these numbers, a major challenge being to determine the fraction of active proteins.

Phospholipid headgroup modification enzymes
Besides complex functions such as facilitating compartment division, regulating transport of molecules and participating to membrane signaling, the most basic requirement of the lipid membrane is the ability to form stable bilayers. Given the inverse cone-shape structure (ratio of the diameter of the headgroup to that of the tails is < 1) of PA glycerophospholipids, such as DPPA and DOPA used in this study, they do not pack in a flat lipid sheet and, thus, cannot form stable bilayers in regular conditions. As a comparison diacyl-PC lipids, which are most frequently used to form supported bilayers or liposomes, have a cylinder shape with headgroup-to-tails diameter ratio ~1.
Therefore, converting the PA lipids synthesized in the vesicles into phospholipids of larger head group is essential to support stability of growing liposome-based minimal cells.
It is widely reported that the composition of E. coli membranes is approximately 80% diacylphosphatidylethanolamine (XXPE) and 20% diacylphosphatidylglycerol (XXPG) with a small fraction of cardiolipin 3,4,5 . Hence, it is natural to convert PA lipids into PE and PG in the E.
coli-based PURE system to mimic a physiological membrane environment. The enzymes required for the synthesis of PE and PG lipids from PA are shown in Figure 6. Some of their structural and functional properties are summarized below.
CdsA. The first reaction of the pathway for both the synthesis of PE and PG lipids is where a phosphatidic acid, cytidine triphosphate (CTP) and a proton react to form a cytosine diphosphatediacylglycerol with the release of diphosphate. This reaction is catalyzed by the gene product of cdsA, a phosphatidate cytidylyltransferase. The molecular weight of the protein as predicted by the nucleotide sequence is ~31 kDa and the experimentally observed molecular weight is ~27 kDa 6 .
Sequence analysis predicts eight transmembrane helices and databases list it as localized to the inner membrane of E. coli 7 .
PssA. The first committed step of synthesizing PE from PA is the formation of phosphatidylserine. The gene product of pssA, the diacylphosphatidylserine synthase, ligates Lserine to CDP-diacylglycerol via a reaction mechanism releasing CMP and a proton 8,9 . The enzyme PssA is considered to be a peripheral membrane protein that spends time associated with membranes or free in the cytoplasm 10 . The predicted molecular weight from the nucleotide sequence is ~53 kDa 11 .
Psd. The final step in the formation of PE lipids is the decarboxylation of phosphatidylserine by the gene product of psd, the phosphatidylserine decarboxylase. In this reaction diacylphosphatidylserine (PS) and a proton react, and diacylphosphatidylethanolamine and carbon dioxide are released. The enzyme itself is a heterodimer, produced from a single polypeptide that self-cleaves posttranslationally 12,13 . The Psd enzyme is located at the inner membrane 14,15 . The predicted molecular weight from the nucleotide sequence is ~36 kDa 16 .
PgsA. The first committed step in the PG biosynthesis pathway starting from phosphatidic acid is the formation of phosphatidylglycerolphosphate (PGP). The gene product of pgsA, the phosphatidylglycerophosphate synthase, accepts CDP diacylglycerol and ligates the sn-glycerol-3phosphate to it releasing diacyl-phosphatidylglycerolphosphate, CMP and a proton. PgsA is an integral membrane protein 9 located at the inner membrane 13 . The enzyme has an absolute requirement for magnesium to function 17 (PUREfrex contains 14 mM Mg 2+ ). The predicted molecular weight from the nucleotide sequence is ~21 kDa 18 .

PgpA, PgpC (PgpB).
The final step in the formation of PG lipids is the dephosphorylation of PGP into diacyl-phosphatidylglycerol. There exist three enzymes, called phosphatidylglycerolphosphatases, that can perform this hydrolysis reaction 19 . They are the products of the pgpA, pgpB and pgpC genes. PgpA contains a single transmembrane segment and its active site faces the cytoplasm 20 . PgpB was originally thought to be an outer membrane phosphatase, but more recent results indicate an inner membrane location 21 . PgpC is predicted to have a single transmembrane domain with its active site facing the cytoplasm 19 . The molecular weight of PgpA from its nucleotide sequence is ~19 kDa 22 , that of PgpB is ~29 kDa 23 , (while from experiment it is ~28 kDa 22 ) and that of PgpC is ~24 kDa 24 . In our experiments we expressed either PgpA or PgpC, both resulting in synthesis of PG lipids.
To our knowledge we are the first to attempt to express and study the complete synthesis pathways for PE and PG lipids in vitro. Previous works focused on generating phosphatidylcholine with purified proteins 25,26 . Here we show that we can produce the membrane forming lipids PG and PE as well as their intermediates using cell-free synthesized enzymes and simple G3P and monoacyl-CoA precursors.