The Phosphatidic Acid Binding Site of the Arabidopsis Trigalactosyldiacylglycerol 4 (TGD4) Protein Required for Lipid Import into Chloroplasts*

Background: The phosphatidic acid (PtdOH) binding site of TGD4 is not yet known. Results: Amino acids 1–80 and 110–145 of TGD4 represent two PtdOH interacting sequences and TGD4 forms a homodimer in vitro and in vivo. Conclusion: TGD4 N terminus binds PtdOH while its C terminus interacts with a second TGD4 protein. Significance: This work reveals the structural functional relationship of an essential chloroplast β-barrel lipid transfer protein. Chloroplast membrane lipid synthesis relies on the import of glycerolipids from the ER. The TGD (TriGalactosylDiacylglycerol) proteins are required for this lipid transfer process. The TGD1, -2, and -3 proteins form a putative ABC (ATP-binding cassette) transporter transporting ER-derived lipids through the inner envelope membrane of the chloroplast, while TGD4 binds phosphatidic acid (PtdOH) and resides in the outer chloroplast envelope. We identified two sequences in TGD4, amino acids 1–80 and 110–145, which are necessary and sufficient for PtdOH binding. Deletion of both sequences abolished PtdOH binding activity. We also found that TGD4 from 18:3 plants bound specifically and with increased affinity PtdOH. TGD4 did not interact with other proteins and formed a homodimer both in vitro and in vivo. Our results suggest that TGD4 is an integral dimeric β-barrel lipid transfer protein that binds PtdOH with its N terminus and contains dimerization domains at its C terminus.

In land plants, glycolipids, such as monogalactosyldiacylglycerol (MGDG), digalactosyldiacylglycerol (DGDG), and sulfoquinovosyldiacylglycerol (SQDG) comprise ϳ64% of total membrane lipids (1). Glycolipids are synthesized through two independent pathways: the prokaryotic and the eukaryotic pathways (2). Plastid specific acyltransferases of the prokaryotic pathway are associated with the inner plastid envelope membrane and transfer the acyl groups, which are de novo synthesized in the chloroplast and attached to acyl carrier proteins (ACPs), to glycerol-3-phosphate producing PtdOH. This PtdOH is then dephosphorylated to diacylglycerol (DAG), 3 the direct substrate for glycolipid synthesis. In the eukaryotic pathway, endoplasmic reticulum (ER)-associated acyltransferases use acyl-CoAs derived from plastid-exported acyl groups and glycerol-3-phosphate as substrates, providing PtdOH for the synthesis of extraplastidic phospholipids, such as phosphatidylcholine (PtdCho) and phosphatidylethanolamine (PtdEtn). A certain portion of lipid precursors returns to the chloroplast, where ER-assembled DAG moieties are incorporated into glycolipids (3). Glycolipids originating from the prokaryotic pathway have a 16-carbon acyl chain in the sn-2 position of the glyceryl backbone, while glycolipids derived from the eukaryotic pathway have an acyl chain of 18 carbons in the same position. This difference has been attributed to the different substrate preferences of acyltransferases in the chloroplast and the ER (4,5). Plants that utilize both pathways, such as Arabidopsis thaliana, have 16:3 fatty acids in the sn-2 position of their thylakoid lipids, and thus are referred to as 16:3 plants. Other plants, such as maize or castor bean, rely exclusively on the eukaryotic pathway and have mostly 18:3 fatty acids in the sn-2 position of their thylakoid lipids; these plants are designated 18:3 plants (6).
Four Arabidopsis proteins, TGD1-4, are currently known to be involved in the ER-to-chloroplast lipid trafficking process (7)(8)(9)(10). Plants with mutations in any of the TGD genes are impaired in the eukaryotic pathway and accumulate an additional lipid, trigalactosyldiacylglycerol (TGDG). The TGD1, 2, and 3 proteins are proposed to form a bacterial-type ABC transporter complex (11), in which TGD1 is a membrane permease, TGD2 a substrate-binding protein, which specifically binds to PtdOH, and TGD3 an ATPase. TGD4 is a predicted ␤-barrel membrane protein (12,13) embedded in the chloroplast outer envelope and, like TGD2, specifically binds PtdOH (14). Based on the current findings, it seems likely that the TGD proteins form a lipid transfer conduit to import PtdOH from the ER through the plastid envelopes. Here we describe the identification of two PtdOH binding sequences of TGD4, the lipid bind- ing properties of TGD4 proteins in 18:3 plants, and the composition of the TGD4 complex.

EXPERIMENTAL PROCEDURES
Expression and Purification of DsRED-TGD4 Proteins-The truncation mutants T1-T21 were generated starting with construct pLW01/DsRED-TGD4 as the template for polymerase chain reaction (PCR) (14). The PCR products were subsequently inserted into the pLW01 vector using SacI and NotI restriction sites. M1-M9 point mutations were generated by site-directed mutagenesis (Stratagene, La Jolla, CA) using pLW01/DsRED-TGD4 T14 (AA 110 -145) as the template and M10-M11 were commercially synthesized (IDT, San Jose, CA). To isolate TGD4 homologs from Ricinus communis (RcTGD4) and Zea mays (ZmTGD4), cDNAs obtained by reverse transcription from isolated total RNA were used as the template for PCR. RcTGD4 and ZmTGD4 were inserted into the BamHI/ NotI and EcoRI/NotI restriction sites of the pLW01/DsRED-His vector, respectively. Primer sequences used in this study are summarized in supplemental Table S1. Sequencing was performed by the Research Technology Support Facility (RTSF) at Michigan State University to verify the accuracy of cloned fragments. Recombinant proteins were produced in Escherichia coli strain BL21 (DE3) and purified as previously described (14). Briefly, all proteins were produced at 16°C overnight after adding 100 M isopropyl ␤-D-1-thiogalactopyranoside (IPTG) to the cell culture at an A 600 ϭ 0.6 except for T1, T3, T17-T20, M10, M11, which were produced by incubation at 28°C for 3 h. Cells were resuspended in lysis buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 10 mM imidazole, protease inhibitor mixture (Roche, Indianapolis, IN), 1% forskolin-12 (Affymetrix, Santa Clara, CA), pH 8.0) and lysed by incubating with lysozyme (Sigma) at a final concentration of 0.2 mg/ml for 30 min followed by sonication using the sonicator 3000 with a microprobe (Misonix, Farmingdale, NY) for 2 min with 10 s pulses. Recombinant proteins were purified to homogeneity by Ni 2ϩ -nitrilotriacetic acid (NTA)-agarose (Qiagen, Valencia, CA) in the presence of 0.1% foscholine-12 according to the manufacturer's instruction. The purified proteins were dialyzed with Tris-buffered saline (TBS: 50 mM Tris-HCl, pH 7.0, 0.1 M NaCl) with 2 M choline chloride using Amicon centrifugal filter devices with 10 kDa cutoff (Millipore, Billerica, MA). Purified proteins were quantified by RC DC protein assay (Bio-Rad, Hercules, CA).
Liposome Association Assay and Data Quantification-Multi-lamellar liposomes composed of dioleoyl-PtdCho and dioleoyl-PtdOH with different w/w ratios as mentioned in the figure legends were prepared as previously described (14). Briefly, 250 g of total lipids (Avanti, Alabaster, AL; Larodan, Sweden) dried under a stream of nitrogen were hydrated by adding 500 l of TBS with 0.2 M choline chloride and incubating at 37°C for 1 h. Multi-lamellar liposomes were obtained by vortexing for 2 min at the highest speed. Purified proteins (0.25 M) were incubated with liposomes on ice for 30 min. Unbound proteins were separated from liposome-bound proteins by centrifugation at 13,000 ϫ g for 10 min at 4°C followed by two washes with TBS containing 0.2 M choline chloride. The liposome-protein pellet was examined on a SDS-PAGE gel (15) followed by Coomassie Brilliant Blue R-250 (Sigma) staining.
Quantitative analysis was based on densitometry using ImageJ software (16) with three technical repeats. The amount of liposome-bound full-length DsRED-TGD4 was set at 100% and the final data were normalized to individual loading controls. The results in Fig. 6C were fitted to Hill's equation (Equation 1) using Origin software (OriginLab, Northampton, MA), where "n " is Hill's number, "L" is ligand concentration, and K d is the dissociation constant.
Protein Lipid Overlay Assay-Lipid strips were prepared as previously described (14). Briefly, 1 nmol of lipids in 1 l of buffer (25% v/v chloroform, 50% v/v methanol, 10 M HCl, 1% Ponseau S) were spotted onto a nitrocellulose membrane (GE Healthcare, Piscataway, NJ) and dried in a fume hood for 1 h. Finished membranes were blocked in TBST (TBS with 0.1% Tween-20) buffer plus 3% Bovine Serum Albumin (BSA) for 2 h at room temperature. Purified proteins were added to the blocking solution at 1 g/ml final concentration and incubated with lipid strips at 4°C overnight followed by washing in TBST buffer 3 ϫ 10 min and immunoblotting with anti-His (C-term) antibody (Invitrogen, Grand Island, NY) at a dilution of 1:5000. Secondary antibody anti-mouse horseradish peroxidase (Bio-Rad) (dilution: 1:20,000) was added for 0.5 h followed by washing in TBST buffer 6 ϫ 10 min. Signal was detected using a chemiluminescence kit from Sigma.
Chloroplast Preparation-Arabidopsis chloroplasts were prepared as previously described with modifications (19). Briefly, 4-week-old seedlings harvested from MS solid medium were ground in grinding buffer (330 mM sorbitol, 50 mM HEPES-KOH, pH 8.0, 2 mM EGTA, 0.1% w/v BSA with 5 mM MgCl 2 ) and separated on a 40 and 80% discontinuous Percoll (Sigma) gradient by centrifugation at 4,000 ϫ rpm for 10 min. The intact chloroplasts collecting at the interface were isolated and washed once with buffer (330 mM Sorbitol, 50 mM HEPES-KOH, pH 8.0, 5 mM MgCl 2 ) followed by centrifugation at 700 ϫ g for 5 min.
Accession Number-The sequence of Arabidopsis thaliana TGD4 can be found in The Arabidopsis Information Resource under the name At3g06960. The sequence of Ricinus communis and Zea mays TGD4 can be found in GenBank TM under the accession numbers XM_002519286.1 and XM_002519286.1, respectively.

Amino acids 110 -145 Are Necessary and Sufficient for
PtdOH Binding by TGD4-To locate the PtdOH binding site(s) within the TGD4 primary sequence, partially overlapping trun-cation mutants T1-T5 were constructed (Fig. 1A). The N termini of the respective truncations were fused to DsRED to enhance solubility, and the resulting recombinant proteins were produced in E. coli and purified to homogeneity using a C-terminal 6ϫ His tag. The proteins were then tested for PtdOH binding in a liposome association assay, wherein proteins interacting with the PtdOH-containing liposome precip-FIGURE 1. Coarse mapping of the TGD4 PtdOH binding domain. A, schematic representation of TGD4 truncation mutants T1-T5 fused with DsRED on the N terminus. White bar, DsRED; black bar, TGD4 fragment; gray bar, 6ϫ His tag; numbers below indicate amino acid positions in the full-length, wild-type sequence. B, multi-lameller liposomes consisting of 40% (w/w) PtdOH and 60% (w/w) PtdCho were incubated with purified proteins. Proteins bound to the liposome co-precipitated after centrifugation at 13,000 ϫ g. Protein content of the pellet was analyzed by SDS-PAGE and stained by Coomassie Brilliant Blue. Numbers indicate molecular masses in kDa. T, total protein used in this assay. C, images were quantified by densitometry. Data were normalized with bound DsRED-TGD4 as 100%. Averages and standard deviations were calculated from three repeats. White bar, DsRED; black bar, TGD4 fragment; gray bar: 6ϫ His tag; numbers below indicate amino acids. B, same liposome binding assay was performed for truncation mutants T7-T12 as in described in the legend of Fig. 1B. Numbers indicate molecular masses in kDa. T, total protein used in this assay. Images were quantified and data were normalized with liposome-bound T2 protein as 100%. Averages and standard deviations were calculated from 3 repeats. C, T14 (AA 100 -150) secondary structure predicted by PROF (Predict-Protein). Y-axis represents the likelihood of each structural signature. Note that amino acids 110 -145 formed a putative soluble loop.
itate with the liposomes following centrifugation. The protein content of the liposome pellet was then examined by SDS-PAGE, and the protein signals were quantified by densitometric scanning. As shown in Fig. 1, B and C, the N-terminal 1-300 amino acids accounted for the PtdOH binding activity of TGD4, while the C-terminal portion of the protein showed greatly reduced PtdOH binding.
Additional mutants were constructed to test this hypothesis: T13 (AA 125-145), T14 (AA 110 -145), and T15, which is a truncated version of TGD4 that contains the entire TGD4 protein sequence, except that the loop was substituted with a 6ϫ His linker sequence (Fig. 3A). While T13 only retained partial PtdOH binding activity, T14 still showed full PtdOH binding (Fig. 3B). To test if these truncations were still specific to PtdOH, a lipid overlay assay was used. The purified proteins were incubated with a hydrophobic membrane onto which different plant lipids were spotted. The proteins were then detected with an anti-His antibody (Fig. 3C). Both T14 and full-length TGD4 bound exclusively to PtdOH, and with similar strength. T15 bound specifically to PtdOH as well, but the signal was attenuated suggesting weakened PtdOH binding. In the liposome association assay, T15 lost about 50% of the total PtdOH binding activity, suggesting the existence of an additional PtdOH binding sequence (Fig. 3D). As positively charged amino acids are usually essential for PtdOH binding through electrostatic interactions with the negatively charged PtdOH below indicate amino acids. B, same liposome binding assay was performed for truncation mutants T13-T15 as in Fig. 1B. Numbers indicate molecular masses in kDa. T, total protein. Images were similarly quantified as in Fig. 2B and normalized with liposome-bound T11 as 100%. C, membranes spotted with various lipids were incubated with purified proteins. Proteins bound to lipids were detected using anti-His antibody. DAG, dioleoyl-diacylglycerol; DGDG, digalactosyldiacylglycerol; MGDG, monogalactosyldiacylglycerol; PtdCho, dioleoyl-phosphatidylcholine; PtdEtn, dioleoyl-phosphatidylethanolamine; PtdGly, phosphatidylglycerol; PtdIns, dioleoyl-phosphatidylinositol; PtdOH, dioleoyl-phosphatidic acid; PtdSer, dioleoyl-phosphatidylserine; SQDG, sulfoquinovosyldiacylglycerol; TAG, trioleoyl-triacylglycerol. D, same liposome binding assay as in B, except that the PtdOH concentration was 30% (w/w). Data were normalized with liposome-bound DsRED-AtTGD4 as 100%. Numbers indicate molecular masses in kDa. T, total protein. E, alignment of TGD4 amino acids 110 -145 within three species. Black boxes, identical amino acids; gray boxes, similar amino acids; M1-M9, T14 point mutants with indicated amino acid changed to alanine. F, liposome binding assays with 30% (w/w) PtdOH were conducted with different point mutants. T, total proteins. Numbers indicate molecular masses in kDa. The data were quantified by densitometry with three repeats and normalized with liposome-bound T14 as 100%.
head group (25), we hypothesized that mutating those positive residues in the minimal PtdOH binding domain of TGD4 would affect PtdOH binding. Thus, we changed each positively charged amino acids between residues 110 and 145 to alanine (Fig. 3E) and tested the point mutants' PtdOH binding activity except for M6 (AA 110 -145, K140A), which we were unable to produce in E. coli. We also included a point mutant targeting a conserved serine, M1 (AA 110 -145, S128A). while no single point mutant abolished PtdOH binding, mutants M1-to-M4 retained only about 60% of PtdOH binding activity compared with T14 (AA 110 -145) (Fig. 3F).
A Second PtdOH Binding Domain Was Present in the First 80 Amino Acids of TGD4-As shown in Fig. 1C, T5 (AA 1-100) retained wild-type PtdOH binding activity despite lacking the aforementioned, PtdOH-binding sequence (AA 110 -145). We investigated which sequence in this most N-terminal portion of the protein represented by T5 (AA 1-100) is responsible for PtdOH binding. Additional truncation mutants T17-T19 were constructed (Fig. 4A) and their PtdOH binding activity was tested by liposome association assay. Only T19 (AA 1-80) bound to PtdOH while T18 (AA 1-50) and T17 (AA 1-25) did not, indicating that amino acids 50 -80 were necessary for PtdOH binding by this fragment (Fig. 4B). This region was also predicted to have a loop structure and was highly conserved among different species (Fig. 4, A and E). Additional mutants were tested, which included a truncation spanning residues 50 -80 (T20) and two point mutants, M10 and M11, in the 50 -80 AA region that had an alanine in place of a positively charged arginine (Fig. 4A). In contrast to T14 (AA 110 -145), T20 (AA 50 -80) was not sufficient for PtdOH binding. The truncation mutant T21 that lacked both PtdOH binding sequences (AA 110 -145 and AA 1-80) abolished the PtdOH binding activity, indicating that both sequences are necessary in order for TGD4 to bind PtdOH and that these two sequences are the only PtdOH binding sequences in the TGD4 protein.  binding assay with 40% (w/w) PtdOH was conducted for truncation mutants T17-T19. T: total proteins; Numbers indicate protein masses in kDa. Data were normalized and quantified as Fig. 1C. C, liposome binding assay with 30% (w/w) PtdOH was performed for T20, M10, M11, and T21. The results were quantified as above. Note that T20 was not sufficient for PtdOH binding. D, lipid overlay assay testing the lipid binding specificity of T19 (AA 1-80). Lipid abbreviations are the same as in Fig. 3C. E, T19 (AA 1-80) secondary structure predicted by PROF (PredictProtein). Y-axis represents the likelihood of each structural signature. FEBRUARY 15, 2013 • VOLUME 288 • NUMBER 7 plants (4,5) leading to the suggestion that DAG instead of PtdOH is the lipid transferred from the ER to the chloroplast in 18:3 plants.

TGD4 Phosphatidic Acid Binding Site
To test this hypothesis, cDNAs of TGD4 orthologs of the 18:3 plants Ricinus comunis (RcTGD4) and Zea mays (ZmTGD4) were isolated and expressed in E. coli to produce fusion proteins with DsRED at the TGD4 N terminus. AtTGD4 shares about 50% amino acid identity with RcTGD4 and ZmTGD4 (supplemental Fig. S1). The lipid overlay assay showed that, similar to AtTGD4, both RcTGD4 and ZmTGD4 bound only to PtdOH and not to DAG (Fig. 5A). To estimate the relative PtdOH binding affinities, the liposome association assay was employed. Both TGD4 orthologs from the 18:3 plants, especially RcTGD4, showed much stronger PtdOH binding than AtTGD4 (Fig. 5B).
TGD4 Forms a Functional Homodimer in Vitro and in Vivo-Proteins in a cell often form large complexes with other proteins involved in the same pathway to efficiently carry out their biological functions. These proteins may not be identified through genetic screening due to gene redundancy or mutant lethality. To identify possible additional components involved in lipid transfer from the ER to the plastid as part of the eukaryotic pathway, we searched for TGD4-interacting proteins in vivo using immunoprecipitation followed by mass spectrometry. However, no new proteins were found to interact with TGD4, and only TGD4 was identified repeatedly with high confidence (supplemental Table S2). The recombinant DsRED-TGD4 proteins migrated as oligomers, mainly dimers and pentamers, on a Blue-Native PAGE gel, while DsRED alone behaved as a monomer (Fig. 6, A and B). This oligomerization activity was mainly attributable to the C-terminal portion of TGD4 (DsRED-TGD4C), the region that was not responsible for PtdOH binding (Fig. 6, A and B). Liposome association experiments with increasing percentages of PtdOH showed a sigmoidal binding curve reminiscent of positive cooperativity characteristic of allosteric soluble proteins with a Hill number equal to 3 (Fig. 6C). It should be noted that the TGD4 protein is in a membrane-bound complex accepting a lipohilic substrate, and that the increasing percentage of PtdOH may alter the physical properties of the multi-lamellar liposome, which is a scenario in which the Hill equation is not strictly applicable. However, while the quantitative relationship between substrate concentration and TGD4 complex behavior in the membrane remains unresolved, qualitatively the observed sigmoidal relationship would be consistent with a protein complex containing multiple substrate-binding sites that cooperatively interact upon ligand binding.
To study the TGD4 complex in its native environment, and to employ an approach independent of DsRED, which represents a relatively large tag that might affect oligomerization of the fusion protein even though an engineered monomeric version of DsRED was used, chloroplasts from HA-TGD4 transgenic plants were isolated and their protein content was analyzed by Blue-Native PAGE. The TGD4 complex was specifically detected by immunoblotting using an HA antibody. The size of the complex fell between protein markers of 66 and 132 kDa under both resting and lipid-importing conditions (Fig. 6D). Under resting condition, isolated chloroplasts were incubated in an iso-osmotic buffer while in the lipid-importing condition liposomes composed of PtdOH and PtdCho were added to the chloroplast to allow lipid import as was previously shown (17). Both conditions were chosen to account for the possibility that TGD4 may exist in different complexes depending on the conditions. To better estimate the size of the complex, protein marker sizes were plotted against their migration distances. The approximate size of the complex was 115 kDa, which is approximately twice the size of the TGD4 monomer of 53 kDa. Since no further interacting partners were identified by mass spectrometry, and TGD4 did not interact with other known TGD proteins by co-immunoprecipitation, we concluded that TGD4 forms homodimers in vivo.

DISCUSSION
TGD4 was identified as a protein involved in ER-to-chloroplast lipid trafficking in a genetic screen (9). However, its molecular function was unknown until recently. TGD4 is a PtdOH binding protein that forms a ␤-barrel and is embedded in the chloroplast outer envelope membrane (14). Here we provide a detailed molecular and biochemical analysis of TGD4, in particular its PtdOH binding activity. Among the best characterized PtdOH binding proteins in plants are ABI1 (abscisic acid insensitive-1), AtPDK1 (3Ј-phosphoinositide-dependent kinase-1), PEPC (phosphoenolpyruvate carboxylase), ACBP1 (acyl-coenzyme A-binding protein-1), TGD2, TGD4, and MGD1 (monogalactosyldiacylglycerol synthase 1), of which a few have a characterized PtdOH binding region (7, 26 -31). Comparing characterized peptide sequences involved in PtdOH binding from diverse organisms, a few generalized conclusions can be drawn: 1) PtdOH-interacting, primary amino acid sequences do not appear to be conserved and are not easily recognizable, requiring that each be mapped individually; 2) positively charged amino acids are essential for PtdOH binding; 3) most PtdOH interacting sequences are located in solventexposed peptide loops with helix or coiled structures; and 4) peptide sequences mediating PtdOH binding can be relatively short. The latter point is important for the interpretation of our data, because a prerequisite for the validity of the systematic deletion approach chosen here is that PtdOH binding is fairly localized within the protein sequence and that the engineered mutations do not fundamentally affect the overall protein fold. As we observed binding to relatively small peptides of TGD4 in our experiments, we believe that the chosen approach is valid. However, knowing the full extent of the PtdOH binding site(s) of TGD4 will ultimately require a crystal structure of the TGD4-lipid substrate complex.
Concurring with the general features of PtdOH binding sites summarized above, we propose that TGD4 likely binds PtdOH through its aqueous solvent-exposed loops. The ␤-barrel core is suggested to be embedded in the membrane, and might serve as the conduit for lipid transfer. In support of this hypothesis, both 110 -145 AA and 50 -80 AA PtdOH binding sequences of TGD4 were predicted by BOCTOPUS (32) to face the cytosol. This would place the PtdOH binding site(s) in a position able to accept PtdOH from an extraplastidic membrane such as the ER, as shown in Fig. 7.
It should be noted that binding of PtdOH to TGD4 does not necessarily imply that it is also the transferred lipid substrate. Rather, binding of PtdOH by cytosol-exposed loops of TGD4 otherwise embedded into the outer chloroplast envelope could establish contact to an adjacent membrane such as the ER to provide a possible lipid transfer conduit. In the absence of an assay to directly measure lipid transfer by TGD4, we tried to obtain indirect clues regarding the transported lipid species. Toward this end, we determined the lipid binding properties of TGD4 from two 18:3 plant species. Based on the observation that plants lacking the prokaryotic pathway show reduced PtdOH phosphatase activity associated with the plastid envelope membranes (4,5), one would assume that a different lipid than PtdOH, namely DAG, should be imported from the ER to serve as precursors for thylakoid lipid biosynthesis. However, just like TGD4 of Arabidopsis, the two TGD4 proteins from maize and castor bean specifically bound to PtdOH and not to DAG (Fig. 5). In fact, they appeared to bind PtdOH more read-  FEBRUARY 15, 2013 • VOLUME 288 • NUMBER 7 ily, or with higher affinity than TGD4 of Arabidopsis. Because 18:3 plants lack the prokaryotic pathway for glycolipid synthesis, there must be increased metabolite flux through the eukaryotic pathway and proteins involved might have higher efficiency in transport or turnover of the lipid precursor derived from the ER. Thus our observation on the TGD4 proteins from maize and castor bean would support the hypothesis that PtdOH is indeed the transported lipid species. At this time we only have an incomplete knowledge of the molecular nature of plastid PtdOH phosphatase(s) (33), and further analysis of this enzyme and its location within the chloroplast envelope membranes in 18:3 and 16:3 plants will be required to more clearly understand the molecular basis for the differences in thylakoid lipid biosynthesis in these two plant types.

TGD4 Phosphatidic Acid Binding Site
Probing for additional proteins that might interact directly with TGD4 in a possible complex we came to the conclusion that TGD4 likely forms a homodimer in vivo. In contrast, most studied ␤-barrel membrane proteins are monomers (34). Within the limitation of the experimental approach discussed above, homodimer formation could explain the apparent cooperativity of PtdOH binding indicating a dynamic interaction between the two subunits. The C-terminal portion of TGD4 was necessary for this dimerization, while the PtdOH binding activity resided in the N-terminal sequences of the TGD4 subunits. We did not observe a TGD4 complex containing also the proteins of the TGD1, 2, 3 transporter of the inner envelope membrane of the chloroplast. Thus one has to assume that interactions between the TGD4 and the TGD1, -2, -3 complexes are transient or that they work independently.
In summary, our current analysis of TGD protein function supports the hypothesis that a TGD4 homodimer binds to PtdOH in the ER through its N-terminal PtdOH binding sequences (Fig. 7) and presumably transfers it across the outer envelope. The TGD2 protein of the TGD1, -2, -3 transporter complex extracts PtdOH from the inner leaflet of the outer envelope and presents it to the TGD1 permease, which transports PtdOH across the inner envelope at the expense of ATP hydrolyzed by TGD3. This hypothesis predicts that PtdOH has to be converted to DAG on the stroma side of the inner envelope membrane as DAG is the direct substrate for the bulk synthesis of chloroplast glycolipids. Further corroboration of this hypothesis would be aided by crystal structures of lipid substrate complexes, reconstitution of functional lipid transfer complexes in vitro, and a better understanding of location and substrate specificity of the PtdOH phosphatases predicted to be involved in the process.