TGD4 involved in endoplasmic reticulum-to-chloroplast lipid trafficking is a phosphatidic acid binding protein

Authors


(fax +1 517 353 9334; e-mail benning@msu.edu).

Summary

The synthesis of galactoglycerolipids, which are prevalent in photosynthetic membranes, involves enzymes at the endoplasmic reticulum (ER) and the chloroplast envelope membranes. Genetic analysis of trigalactosyldiacylglycerol (TGD) proteins in Arabidopsis has demonstrated their role in polar lipid transfer from the ER to the chloroplast. The TGD1, 2, and 3 proteins resemble components of a bacterial-type ATP-binding cassette (ABC) transporter, with TGD1 representing the permease, TGD2 the substrate binding protein, and TGD3 the ATPase. However, the function of the TGD4 protein in this process is less clear and its location in plant cells remains to be firmly determined. The predicted C-terminal β-barrel structure of TGD4 is weakly similar to proteins of the outer cell membrane of Gram-negative bacteria. Here, we show that, like TGD2, the TGD4 protein when fused to DsRED specifically binds phosphatidic acid (PtdOH). As previously shown for tgd1 mutants, tgd4 mutants have elevated PtdOH content, probably in extraplastidic membranes. Using highly purified and specific antibodies to probe different cell fractions, we demonstrated that the TGD4 protein was present in the outer envelope membrane of chloroplasts, where it appeared to be deeply buried within the membrane except for the N-terminus, which was found to be exposed to the cytosol. It is proposed that TGD4 is either directly involved in the transfer of polar lipids, possibly PtdOH, from the ER to the outer chloroplast envelope membrane or in the transfer of PtdOH through the outer envelope membrane.

Introduction

Plant chloroplasts, the unique organelles of plant cells, harness solar energy and convert it into chemical energy by conducting photosynthesis, thereby providing food and oxygen for most of the living organisms on earth. The thylakoid membranes in the mature chloroplast represent an extensive and intricate membrane system that harbors the photosynthetic apparatus. The thylakoid lipids provide the structural matrix for the photosynthetic membrane into which the electron transport chain components are embedded. Thylakoid lipids have been observed in the crystal structures of both photosystem I and II (Jordan et al., 2001; Guskov et al., 2009), a location that is consistent with their possible roles in the proper assembly or function of photosynthetic complexes.

Unlike extraplastidic membranes, such as the endoplasmic reticulum (ER) or the plasma membrane, in which phosphoglycerolipids predominate, chloroplast membranes contain primarily galactoglycerolipids, which can account for approximately 70% of total lipids in leaf tissue (Dörmann and Benning, 2002). Of the galactoglycerolipids, monogalactosyldiacylglycerol (MGDG) and digalactosyldiacylglycerol (DGDG) represent the two most abundant classes. The amount of DGDG increases further during phosphate deprivation in leaves in order to substitute for the loss of phospholipids in extraplastidic membranes (Härtel et al., 2000).

Galactolipids are synthesized at the chloroplast envelope membranes (Benning and Ohta, 2005). The MGDG synthase encoded by MGD1 in Arabidopsis transfers a galactosyl residue from UDP-Gal to diacylglycerol (DAG) to generate an MGDG (Jarvis et al., 2000; Awai et al., 2001). MGD1 is localized at the inner envelope and faces the intermembrane space (Xu et al., 2005). Transfer of a second galactosyl residue from UDP-Gal to MGDG is catalyzed by the DGDG synthase encoded by DGD1 (Dörmann et al., 1999), which is localized at the outer envelope of the chloroplast and faces the cytosol (Froehlich et al., 2001). There are two pathways that contribute to the DAG precursor pool for galactoglycerolipid synthesis (Benning, 2009). In the ‘prokaryotic pathway’, DAG assembly from de novo synthesized fatty acids takes place within the chloroplast. In the ‘eukaryotic pathway’, acyl groups are exported from the plastid to be available for polar lipid assembly at the ER where most of the extraplastidic phosphoglycerolipids are synthesized. DAG moieties transferred from the ER to the chloroplast serve as precursors in the synthesis of galactoglycerolipids. Thylakoid lipids derived from the prokaryotic pathway carry a 16-carbon acyl chain at the sn-2 position of the glycerol backbone, the lipids derived from the eukaryotic pathway an 18-carbon acyl chain at the same position (Heinz and Roughan, 1983).

Four genes, TGD1, 2, 3, and 4, identified in a genetic mutant screen, encode proteins that are involved in ER-to-chloroplast lipid transfer in Arabidopsis (Xu et al., 2003, 2008; Awai et al., 2006; Lu et al., 2007). The respective tgd mutants accumulate abnormal oligogalactolipids, most prominently trigalactosyldiacylglycerol (TGDG), which gives rise to the mutant and protein designation, tgd and TGD, respectively. The tgd mutants also have fewer thylakoid lipids derived from the eukaryotic pathway. The accumulation of oligogalactolipids in these mutants results from the activation of a processive galactosyltransferase, likely SFR2 (Moellering et al., 2010). The TGD1, 2, and 3 proteins resemble the components of a bacterial-type ATP binding cassette (ABC) transporter complex that is probably associated with the inner envelope membrane (Benning, 2009). TGD1 contains multiple transmembrane domains and is proposed to be the permease of the complex (Xu et al., 2003). TGD2 is similar to the substrate binding protein and binds phosphatidic acid (PtdOH) specifically (Awai et al., 2006; Lu and Benning, 2009; Roston et al., 2011). TGD3 is an ATPase localized in the chloroplast stroma (Lu et al., 2007).

The TGD4 protein encoded by At3g06960.1 does not contain any known functional domains but is conserved in evolution from green algae to higher plants (Xu et al., 2008). It is also distantly related to the bacterial LptD protein, which is an outer membrane β-barrel protein in Escherichia coli and is involved in lipid A transport (Haarmann et al., 2010). Conflicting evidence has arisen with regard to the cellular localization of TGD4. Results following overexpression of functional TGD4 with the N-terminus fused to green fluorescent protein (GFP) were consistent with TGD4 localization at the ER. However, chloroplast proteomic studies suggested a chloroplast localization of TGD4 (Ferro et al., 2003; Zybailov et al., 2008). The goals of this study were to determine the molecular function of TGD4 and to resolve the conflicting data on the cellular localization of the TGD4 protein.

Results

TGD4 binds PtdOH in vitro

If TGD4 is involved in the transfer of lipids from the ER to the plastid as suggested by the mutant phenotype, it might bind membrane lipids specifically. Therefore, the lipid-binding properties of TGD4 were investigated by production of TGD4 fused to DsRED. The DsRED protein is a red fluorescent protein derived from the coral Discosoma sp. (Gross et al., 2000) and fusions with TGD2 have been used successfully to produce a soluble protein that is functional in lipid-binding assays (Lu and Benning, 2009; Roston et al., 2011). Initially, the DsRED protein was fused to the N-terminus of the full-length TGD4 protein with a C-terminal His-tag (DsRED–TGD4–His) and gave rise to a fusion protein that was membrane associated. The DsRED–TGD4–His protein and later its derivatives were solubilized and purified on a nickel-chelate column in the presence of the zwitter-ionic detergent foscholine-12. Removal of detergent from the DsRED–TGD4–His protein preparation resulted in protein precipitation, which was found to be minimized in the presence of choline chloride. Of all compounds tested, choline chloride was found to be a suitable stabilizer (Figure S1) and was routinely added to the purified protein.

In protein–lipid overlay assays probing lipids on commercially available membranes (Figure 1a), DsRED–TGD4–His was found to bind specifically to PtdOH, but not to any other phospholipids tested. Moreover, when probing different chloroplast lipids manually spotted onto membranes, DsRED–TGD4–His did not bind to any other lipids except PtdOH (Figure 1b). The DsRED–His protein itself did not bind to any of the lipids on both membranes.

Figure 1.

 The TGD4 protein binds phosphatidic acid (PtdOH) in vitro.
(a) Of all phospholipids tested, the DsRED–TGD4–His protein binds specifically to PtdOH in the lipid overlay assay. An immunoblot is shown with proteins detected using His-tag antibodies. LPtdOH, lysophosphatidic acid; LPtdCho, lysophosphatidylcholine; PtdIns, phosphatidylinositol; PtdIns(3)P, phosphatidylinositol 3-phosphate; PtdIns(4)P, phosphatidylinositol 4-phosphate; PtdIns(5)P, phosphatidylinositol 5-phosphate; PtdEtn, phosphatidylethanolamine; PtdCho, phosphatidylcholine; S1P, sphingosine 1-phosphate; PtdIns(3,4)P2, phosphatidylinositol 3,4-bisphosphate; PtdIns(3,5)P2, phosphatidylinositol 3,5-bisphosphate; PtdIns(4,5)P2, phosphatidylinositol 4,5-bisphosphate; PtdIns(3,4,5)P3, phosphatidylinositol 3,4,5-triphosphate; PtdOH, phosphatidic acid; PtdSer, phosphatidylserine.
(b) Of all plant lipids tested, DsRED–TGD4–His protein binds to PtdOH in the lipid overlay assay. An immunoblot is shown with proteins detected using His-tag antibodies. DAG, diacylglycerol; TAG, triacylglycerol; MGDG, monogalactosyldiacylglycerol; DGDG, digalactosyldiacylglycerol; SQDG, sulfoquinovosyldiacylglycerol; PtdGro, phosphatidylglycerol.
(c) Effect of PtdOH fatty acyl chain length on DsRED–TGD4–His binding affinity in the liposome association assay. Liposomes contained 40 mol% PtdOH and 60 mol% PtdCho. Protein gels stained with Coomassie Brilliant Blue (CBB) are shown. M, protein marker; L, loading control (total protein issued in the assay); N, no liposome control (represents protein precipitation and not lipid binding). Fatty acids are indicated with their number of carbons: number of double bonds.
(d) Effect of PtdOH fatty acyl desaturation level on the DsRED–TGD4–His binding affinity in the liposome association assay. Liposomes contained 40 mol% PtdOH and 60 mol% PtdCho. Protein gels stained with CBB are shown. M, protein marker; L, loading control; N, no liposome control; 4ME 16:0, diphytanoyl phosphatidic acid; NBD-PtdOH, fluorescent NBD group-labeled phosphatidic acid.
(e) Effect of pH on PtdOH binding to DsRED–TGD4–His. Liposomes contained 40 mol% PtdOH and 60 mol% PtdCho. Protein gels stained with CBB are shown. M, protein marker; L, loading control.

To independently verify PtdOH binding in a different assay and to test whether the protein showed preferences for different molecular species of PtdOH with regard to the acyl composition of the DAG moiety, a liposome-binding assay was developed in which binding of the protein to liposomes that contained different species of PtdOH could be tested (Figure 1c,d). In this assay, liposomes of defined lipid composition were incubated with the DsRED–TGD4–His protein. Following centrifugation, protein bound to the liposomes was detected in the pellet whilst unbound protein remained in the supernatant. A prerequisite for the liposome binding assay to work was the exclusion of detergent, while at the same time stabilization of the DsRED–TGD4–His fusion protein by addition of choline chloride was needed. Using this assay, DsRED–TGD4–His was found to bind to dipalmitoyl PtdOH and distearoyl PtdOH although the binding of distearoyl PtdOH appeared to be stronger. For PtdOH species of the same acyl chain length but different desaturation levels, DsRED–TGD4–His showed a higher affinity for PtdOH with an increasing number of double bonds. Interestingly, DsRED–TGD4–His appeared to have an even higher affinity to diphytanoyl PtdOH, which carries branched acyl chains with four methyl groups. However, DsRED–TGD4–His did not bind PtdOH that carried fluorescently labeled acyl substituents. The secondary band visible for the DsRED fusion proteins on the gels (Figures 1c–e and 2) was a result of DsRED self-cleavage during denaturation prior to electrophoresis (Gross et al., 2000). Because pH affects protonation of PtdOH and, in some instances, also PtdOH binding to proteins (Young et al., 2010), the effect of pH was tested. However, the binding of DsRED–TGD4–His to PtdOH was not affected over a pH range of 6.4–7.8 (Figure 1e).

Figure 2.

 Phosphatidic acid (PtdOH) binds to the N-terminal domain of DsRED–TGD4–His.
(a) The primary structure of the TGD4 protein, DsRED–His, DsRED–TGD4–His and truncation mutants. HR, hydrophobic region (cross-hatched bar); solid bar, TGD4; open bar, DsRED–His; gray bar, His-tag; dashed line, deletion. The numbers refer to amino acids.
(b) PtdOH binding affinity of DsRED–TGD4–His derivatives in the liposome binding assay. Protein gels stained with CBB are shown. Liposomes were made up of dioleoyl-PtdOH and dioleoyl-PtdCho. The weight percentage of PtdOH in the liposome varied from 0 to 80%. M, protein marker; L, loading control; N, no liposome control.

PtdOH binding is primarily a function of the N-terminus half of TGD4

To begin to determine the possible location of a PtdOH binding site in TGD4, a series of DsRED–TGD4–His truncation mutants was constructed as shown in Figure 2 (a). TGD4 contains a hydrophobic region of 23 amino acids (287D–309F) predicted by Aramemnon (Schwacke et al., 2003). To test whether this region is involved, it was deleted in the DsRED–ΔTGD4–His protein (Figure 2a). The N-terminus portion of TGD4 up to the mentioned hydrophobic region was fused to the N-terminus of DsRED and gave rise to TGD4N-DsRED–His (Figure 2a). The TGD4 part C-terminus of the hydrophobic region was fused to the C-terminus of DsRED giving rise to DsRED–TGD4C–His (Figure 2a). Except for DsRED–His alone, all tested recombinant fusion proteins were found to bind to PtdOH-containing liposomes, and more so as the fraction of PtdOH in the liposomes increased. Relative to the respective loading control (L) and representative of the total amount of protein in the assay, the TGD4N-DsRED–His protein showed an affinity to PtdOH liposomes similar to the full-length protein DsRED–TGD4–His (full binding at 40% PtdOH), which suggested that a major PtdOH binding region resides within the N-terminus part of TGD4. In contrast, the DsRED–TGD4C–His protein had much lower affinity compared with the wild-type protein DsRED–TGD4–His but still bound PtdOH, because only a fraction of the protein in the assay (compared with the loading control L which represents total protein) was present in the liposome pellet at all concentrations tested. Thus we concluded that PtdOH binding activity does not require the central hydrophobic region of TGD4 and resides primarily, although not exclusively, in the N-terminal portion of TGD4.

PtdOH is increased in the tgd4 mutants

Previous lipid profiling of the tgd4 mutant plants did not extend to PtdOH (Xu et al., 2008). If TGD4 is involved in the transfer of PtdOH from the ER to the plastid, one would expect an increase in PtdOH content in the tgd4 mutant, as was previously observed for tgd1 (Xu et al., 2005). The tgd4-1 allele carries a one-amino acid substitution (P20L) while tgd4-2 and tgd4-3 are T-DNA knock-out lines (Xu et al., 2008). Total lipid extracts from the wild type and the different tgd4 mutant alleles were separated by two-dimensional thin-layer chromatography (TLC), which allowed clean isolation of PtdOH (Figure 3a), and subsequent quantification (Figure 3b). All tgd4 mutant alleles showed increased relative amounts of PtdOH, and approximately doubled in the weak tgd4-1 point mutant allele and tripled in the strong tgd4-2 allele (Figure 3b) compared with wild type. The probing of lipids in chloroplasts isolated from the wild type and the weaker tgd4-1 mutant allele (which was not possible for the stronger T-DNA alleles due to the limited availability of material) did not reveal PtdOH in mutant chloroplasts (Figure S2). Thus it is likely that the additional PtdOH observed in the tgd4-1 mutant is associated with extraplastidic membranes. Analysis of the fatty acid composition of PtdOH in the tgd4-2 mutant revealed an elevated 18:1 (number of carbons:number of double bonds) and decreased 18:3 acyl group content, this finding was similar to observations previously made for the tgd1 mutant (Xu et al., 2005).

Figure 3.

 The tgd4 mutants accumulate phosphatidic acid (PtdOH) in vivo.
(a) PtdOH separated by two-dimensional TLC. Wild type (WT) and tgd4-3 were compared. Lipids were visualized by charring using 50% sulfuric acid. Abbreviations of lipids are as in Figure 1. TGDG, trigalactosyldiacylglycerol.
(b) Quantification of PtdOH by gas-liquid chromatography. Values represent the molar ratio of PtdOH to total lipids. Error bars indicate the standard deviation of three biological repeats.
(c) PtdOH fatty acid profile of wild type (WT) and tgd4-2 mutants. Fatty acid species are designated with numbers of carbons:number of double bonds. Error bars represent the standard deviation of three biological repeats.

TGD4 protein is localized in the outer chloroplast envelope membrane

To investigate the location of the native TGD4 protein, a polyclonal TGD4 antiserum was produced in rabbits using purified DsRED–ΔTGD4–His as antigen. TGD4 antibodies were highly purified from the crude serum. Using immunoblotting, a signal that corresponds to the TGD4 protein with a calculated molecular weight of 52.8 kDa, was detected in leaf extract of the wild type but not of the tgd4-1 mutants (Figure 4a). It is interesting to note that the TGD4 protein was not detectable in this point mutant, which suggested that the respective mutation affects the abundance of TGD4 in vivo.

Figure 4.

 TGD4 is localized to the chloroplast. Immunoblots are shown for TGD4, TOC75, and BIP. RuBisCO was visualized by CBB staining.
(a) Purified polyclonal antibody raised against DsRED–ΔTGD4–His specifically detects TGD4 in wild type (WT) but not in the tgd4-1 point mutant line. Numbers on the left indicate the molecular weights of protein markers in kDa.
(b) TGD4 is enriched in chloroplast preparations compared with total leaf extracts.TOC75, chloroplast outer envelope marker; BIP, endoplasmic reticulum (ER) marker; RuBisCO, loading control.
(c) TGD4 does not co-fractionate with ER markers on a sucrose gradient. TIC110, chloroplast inner envelope marker. Chlorophyll content serves as a thylakoid marker.

Cell fractionation in combination with protein immunoblotting and detection with the purified TGD4 antibody was employed to localize TGD4. The TGD4 protein was enriched in isolated chloroplasts in wild-type plants (Figure 4b) in parallel with the chloroplast outer envelope marker TOC75 (Tranel et al., 1995). However, the ER luminal binding protein marker (BIP) (Oliver et al., 1995), was absent from the isolated chloroplasts. To determine whether TGD4 might also be present in the ER, an Arabidopsis wild-type microsomal preparation was fractionated by a continuous sucrose gradient to separate ER from other membranes (Figure 4c). ER microsomes represented by BIP and SMT1, an ER membrane protein (Boutte and Grebe, 2009), were present in the denser fractions, which also contained thylakoid membrane fragments as indicated by the presence of chlorophyll. TOC75 was enriched in the medium dense fractions while TIC110, an inner envelope marker (Inaba et al., 2005), was found in both medium dense and dense fractions. The fractionation profile for TGD4 was most similar to that of TOC75, which suggests that TGD4 is primarily associated with the chloroplast.

To further refine the localization of the native TGD4 protein, chloroplasts isolated from the wild-type leaves were subjected to protease digestion. Thermolysin does not penetrate the chloroplast outer envelope membrane and, therefore, it only digests proteins of the outer envelope membrane exposed to the cytosol. On the other hand trypsin, which is smaller in size, is able to penetrate the outer envelope membrane but not the inner envelope membrane and digests proteins associated with the inner envelope membrane and that face the intermembrane space (Joyard et al., 1983). As shown in Figure 5(a,b), TGD4 protein was susceptible to thermolysin and trypsin digestion, as was TOC159 an outer envelope membrane protein (Hiltbrunner et al., 2001), while the stroma protein RuBisCO was resistant to both. The addition of TritonX-100 disrupts chloroplast envelopes and allowed complete accessibility by both proteases.

Figure 5.

 TGD4 is a membrane protein of the outer chloroplast envelope.
Immunoblots are shown for TGD4 and TOC75. RuBisCO was visualized by CBB staining.
(a) Wild-type (WT) chloroplasts were treated with 0 to 4 mg ml−1 thermolysin. TX-100, TritonX-100; TOC 159, outer envelope marker. RuBisCO, stroma marker.
(b) Wild-type (WT) chloroplasts were treated with 0 to 0.8 mg ml−1 trypsin.
(c) Chloroplasts isolated from a transgenic line that produces an N-terminus haemagglutinin (HA)-tagged TGD4 protein in the tgd4-1 mutant background were treated with 0–4.0 mg ml−1 thermolysin. HA–TGD4 was detected by immunoblotting using a monoclonal HA antibody.
(d) Wild-type (WT) chloroplasts were treated with hypotonic buffer only, 2 m NaCl, 0.1 m Na2CO3 or 0.1 m NaOH followed by centrifugation. Chl, chloroplast; S, supernatant; P, pellet.
(e) A histogram of the likelihood of the secondary structure of TGD4 predicted by PROF (PredictProtein). Numbers represent the amino acids. WA, water accessibility.

To study the topology of TGD4, a transgenic line that produced an N-terminally HA-tagged TGD4 protein in the tgd4-1 mutant background was generated. It should be noted that this construct was able to rescue the lipid phenotype of the tgd4-1 mutant (Figure S3). When chloroplasts isolated from this HA-TGD4-producing line were treated with thermolysin, the N-terminal HA-tag was susceptible to the protease (Figure 5c). Therefore, we proposed that the N-terminus of TGD4 was presumably facing the cytosol. Based on these results it is concluded that TGD4 is located in the outer envelope membrane of the chloroplast with its N-terminal portion at least partially exposed to the cytosol.

To determine the strength of the interaction between TGD4 and the outer envelope, isolated wild-type chloroplasts were extracted with sodium chloride, sodium carbonate, or sodium hydroxide (Figure 5d). Peripheral or monotopic membrane proteins can be extracted by sodium chloride or sodium carbonate respectively, while transmembrane proteins are resistant to strongly basic sodium hydroxide (Miege et al., 1999). TGD4, like TOC75, which is a β-barrel protein, could not be extracted by any of the three reagents. In contrast, RuBisCO, most of which is peripheral to the thylakoid membrane (Irving and Robinson, 2006), was extracted by all three reagents. The secondary structure prediction of TGD4 by PROF (Rost et al., 2004) suggested that the TGD4 protein most likely forms multiple β-sheets, especially at the C-terminus, and corresponds well with regions not accessible to water that are indicative of a possible β-barrel conformation (Figure 5e). Taken together, we propose that TGD4 is a transmembrane protein, possibly a β-barrel protein, localized in the outer envelope membrane of the chloroplast with its N-terminus partially exposed to the cytosol.

Discussion

In seed plants, the biogenesis of thylakoid lipids requires the import of precursors from the ER. The identity of the transported lipid(s) remains unresolved, but a likely candidate is PtdOH as this lipid specifically binds to TGD2 (Awai et al., 2006; Lu and Benning, 2009), the substrate-binding protein associated with the proposed inner envelope TGD1,2,3 complex (Roston et al., 2011). However, while the TGD1,2,3 complex possibly transfers PtdOH from the intermembrane face of the outer envelope membrane to the stroma face of the inner envelope membrane, the lipid transported from the ER to the outer envelope membrane may be different, e.g. phosphatidylcholine (PtdCho) (Roughan and Slack, 1982). PtdCho could be converted at the outer envelope membrane to PtdOH by the activity of a phospholipase D and make PtdOH available for further transfer by the TGD1,2,3 complex (Benning, 2009). Based on results described here, it is suggested that PtdOH might also be the lipid transported to and through the outer envelope from the ER. The finding that TGD4 specifically binds PtdOH in vitro is one piece of evidence in favor of the transport of PtdOH all the way from the ER to the stroma face of the inner envelope membrane. The PtdOH-binding activity of TGD4 was mostly attributed to its N-terminal fragment (1–286 aa), which faces the cytosol and is, therefore, potentially able to contact the ER. The C-terminal fragment was predicted to adopt a secondary structure of hydrophobic β-sheets that possibly forms a β-barrel (Imai et al., 2011). As the ancestor of chloroplasts is thought to be a Gram-negative bacterium, it is not surprising that outer membrane transporters of chloroplasts are β-barrel proteins (Schleiff et al., 2003). In fact, TGD4 is similar to LptD/Imp, an E. coliβ-barrel outer membrane protein involved in the transfer of lipid A through the outer bacterial membrane (Sperandeo et al., 2008).

A second piece of evidence in favor of PtdOH transport from the ER to the outer envelope membrane derives from the observation that the different tgd4 mutant alleles have increased PtdOH content, likely associated with extraplastidic membranes, i.e. the ER. The tgd1-1 mutant also showed an increase in PtdOH content outside of the chloroplast (Xu et al., 2005). The PtdOH present in tgd1 and tgd4 mutants has a very similar fatty acid profile with increased 18:1 and decreased 18:3 fatty acids. The reason for this change in fatty acid profile is not yet understood.

TGD4 lacks a recognizable chloroplast transit peptide and it was tentatively localized to the ER by transiently over-production of a functional GFP–TGD4 fusion protein in tobacco (Xu et al., 2008). Because GFP fused to TGD4 may sequester or expose a signal peptide due to altered folding (Hanson and Kohler, 2001), or because overproduction of the recombinant protein could lead to saturation of the cellular protein-sorting machinery, mistargeting of the majority of the recombinant protein visible by fluorescence microscopy might be possible using this approach. Here we employed specific TGD4 antibodies and determined that the native TGD4 protein is primarily associated with the outer chloroplast envelope membrane. This result was in agreement with previous proteomics studies of chloroplast envelopes (Ferro et al., 2003; Zybailov et al., 2008). However, our new result does not exclude the possibility that a subfraction of TGD4 is also associated with the ER, as the microsome preparations inevitably contained microsomes derived from both the outer envelope membrane and the ER. Moreover, physical membrane contacts between the ER and the chloroplast have been visualized and suggested to be the sites of lipid trafficking between the ER and the chloroplast (Andersson et al., 2007). It was determined previously that isolated chloroplasts of the tgd4-1 mutant do not have a reduced number of ER fragments attached compared with wild-type chloroplasts, a finding that suggested that TGD4 might not be directly involved in the tethering of the two membranes (Xu et al., 2008). However, this result did not exclude the possibility that TGD4 is enriched in ER–outer envelope membrane contact sites. Naturally, localization of proteins in membrane contact sites is bound to be ambiguous. For example, the yeast protein Mmm1, which is an essential component of the tethering complex in ER–mitochondrion contact sites (Kornmann et al., 2009), was first localized to the outer envelope of mitochondria by cellular fractionation (Burgess et al., 1994). However, more recent evidence suggests that, without interaction partners, Mmm1 redistributes to the entire ER network (Kornmann et al., 2009).

Based on all data available at this time, we hypothesize that the N-terminal portion of TGD4 extracts PtdOH produced at the ER and transfers it through a β-barrel channel that consists of the C-terminal portion of the protein to the intermembrane face of the outer envelope membrane. Here TGD2, a second PtdOH-binding protein involved in the process accepts PtdOH and transfers it to the TGD1/TGD3 ABC transporter complex, which facilitates PtdOH transfer across the inner envelope membrane hydrolyzing ATP. On the stroma face of the inner envelope membrane PtdOH is dephosphorylated to DAG, the ER-derived substrate for thylakoid lipid synthesis by the eukaryotic pathway.

Experimental procedures

Plant materials and growth conditions

Arabidopsis thaliana ecotype Col 2 and tgd4 mutant plants were grown as previously described (Xu et al., 2005). Surface-sterilized seeds were germinated on 0.5% (w/v) agar-solidified MS medium (Murashige and Skoog, 1962) supplemented with 1% sucrose and transferred to soil after 10 days for propagation. Aerial parts of 4-week-old plants grown on agar-solidified MS medium were harvested for chloroplast isolation and lipid analysis.

Construction of transgenic lines

The HA–TGD4 producing transgenic line was generated in the tgd4-1 mutant background. Full-length TGD4 was amplified by PCR from wild type derived cDNA using a forward primer that encodes the HA tag (Table S1). The PCR product was cloned into a binary vector derived from pPZP211 (Hajdukiewicz et al., 1994) using restriction sites BamHI and SalI. The construct was introduced into the tgd4-1 mutant using the floral dip method (Clough and Bent, 1998). Transgenic plants were selected on MS medium supplemented with 100 μg ml−1 gentamycin. Genotyping of the tgd4-1 allele was performed using CAPs markers as previously described (Xu et al., 2008).

Expression and purification of DsRED–TGD4 fusion proteins

The TGD4 cDNA was initially cloned into the pMalc2x vector (New England Biolabs, www.neb.com). The pMalc2x/TGD4 construct was modified to give rise to pMalc2x/ΔTGD4 by deletion of the 859–924 nt (referring to coding sequence NM_111576) fragment that encodes the hydrophobic region using site-directed mutagenesis. pMalc2x/TGD4 and pMalc2x/ΔTGD4 were used as PCR templates for the amplification of TGD4 (SacI, NotI), TGD4N (NcoI) and ΔTGD4 (SacI, NotI), TGD4C (SacI, NotI) respectively. The restriction sites were included in the primers (Table S1). Following restriction digestion, the PCR fragments were ligated into the pLW01/DsRED–His vector (Lu and Benning, 2009). Sequence identities were confirmed by sequencing at the MSU Research Technology Support Facility. To express DsRED–TGD4–His proteins, constructs pLW01/DsRED–TGD4–His, pLW01/DsRED–ΔTGD4–His, pLW01/TGD4N–DsRED–His and pLW01/dsRED–TGD4C–His were transformed into E. coli strain BL21 (DE3) (www.emdchemicals.com). A 5 ml overnight culture was used to inoculate a 200 ml culture. When the cell density reached A600 = 0.6–0.8, isopropyl-β-d-thiogalactopyranoside was added at a final concentration of 0.1 mm to induce protein expression at 16°C overnight. The cells were centrifuged at 5000 g for 10 min, and resuspended in lysis buffer [50 mm NaH2PO4, 300 mm NaCl, 10 mm imidazole, pH 8.0, 1% (w/v) foscholine-12 and protease inhibitor cocktail (Roche, www.roche-applied-science.com)] with 0.2 mg ml−1 lysozyme (Sigma, www.sigmaaldrich.com). After incubation on ice for 30 min, cells were lysed by sonication followed by centrifugation at 10 000 g for 20 min. The supernatant was filtered through a 0.45-μm filter and was loaded onto a Ni-NTA column (Qiagen, www.qiagen.com). Protein purification was carried out in accordance with manufacturer’s instructions except for the addition of 0.1 % foscholine-12 to the wash and elution buffers. The purified proteins were concentrated with an Amicon centrifugal filter device (Millipore, www.millipore.com) and the buffer was changed to Tris-buffered saline (TBS; 10 mm Tris–HCl, pH 8.0, 150 mm NaCl) with 2 m choline chloride, which stabilizes DsRED–TGD4 proteins (Figure S1). Protein concentration was determined by Bradford assay and protein purity was assessed by SDS-PAGE. The fusion proteins were then frozen in 10-μl aliquots at −80°C.

Protein–lipid overlay assay

The protein–lipid overlay assay was modified from (Awai et al., 2006; Lu and Benning, 2009). PIP2 lipid strips were purchased from Echelon Biosciences (www.echelon-inc.com). The lipids spotted onto membranes were purchased from Avanti Polar Lipids (avantilipids.com) as well as Larodan Fine Chemicals (www.larodan.se). Lipids (10 nmol) were suspended in 20 μl spotting buffer [250 μl chloroform, 500 μl methanol, 200 μl 50 mm HCl, 2 μl 1% (w/v) Ponceau S (Sigma)] and spotted onto Amersham Hybond-C Extra membranes (GE Healthcare, www.gelifesciences.com) followed by drying for 1 h in a fume hood. The lipid membranes were then blocked in 3% (w/v) bovine serum albumin (BSA) in TBST buffer [TBS with 0.25% (v/v) Tween 20] for 2 h at room temperature. Purified DsRED–TGD4–His fusion proteins were added at 1 μg ml−1 final concentration and incubated overnight at 4°C followed by washing three times in TBST. Lipid membranes were then incubated with 1:2000 diluted His antibody (Sigma) in blocking buffer for 2 h at room temperature followed by two washes with TBST. The membranes were processed for immunoblotting as described below.

Liposome association assay

The liposome association assay was adapted from procedures (Awai et al., 2006; Lu and Benning, 2009) with minor modifications. Dioleoyl-PtdCho and PtdOH with different acyl chain lengths and desaturation levels or 1-palmitoyl-2-(12-((7-nitro-2-1,3-benzoxadiazol-4-yl)amino)dodecanoyl)-sn-glycero-3-phosphate (NBD-PtdOH; Avanti, avantilipids.com) were mixed at indicated ratios to give a total lipid amount of 250 μg. The lipids were dried under a stream of nitrogen, resuspended in 0.5 ml TBS buffer with 0.2 m choline chloride and hydrated at 37°C for 1 h followed by vigorous vortexing for 2 min. The resulting multi-lamellar vesicles were centrifuged at 13 000 g for 10 min and then washed once with TBS buffer that contained 0.2 m choline chloride. The liposomes were resuspended into 100 μl TBS buffer with 0.2 m choline chloride and incubated with 2 μg purified DsRED–TGD4–His protein and its derivatives. The protein liposome mixture was incubated on ice for 30 min followed by centrifugation at 13 000 g for 10 min and two washes with 500 μl TBS that contained 0.2 m choline chloride. The resulting protein–liposome pellet was resuspended in 20 μl 2× Laemmli buffer (Laemmli, 1970) and processed by SDS-PAGE (Shapiro and Maizel, 1969).

Lipid analysis by two-dimensional TLC and GC

Total lipids were extracted from 300 mg fresh weight seedlings as described (Wang and Benning, 2011) and separated on TLC silica gel plates (EMD Chemicals). The first-dimension solvent contained chloroform: methanol: 7 m ammonium hydroxide (65:30:4, v/v/v) and the second-dimension solvent contained chloroform: methanol:acetic acid:water (170:25:25:6, v/v/v/v). Lipids were visualized either by 50% sulfuric acid or by iodine vapor. The iodine-stained lipids were scraped from TLC plates and quantified as described previously (Wang and Benning, 2011).

Production and purification of TGD4 antibodies

For the generation of polyclonal antibodies 100 μg purified DsRED–ΔTGD4–His was injected three times to immunize rabbits (Cocalico Biologicals, www.cocalicobiologicals.com). To purify the antibodies from the serum, DsRED–TGD4C–His was conjugated with Affi-Gel 15 (Bio-Rad, www.bio-rad.com) beads in 0.1 m HEPES, 8 m Urea in accordance with the manufacturer’s instruction. Anti-TGD4 crude serum was incubated with the antigen-coupled beads overnight at 4°C. After washing seven times with 5 ml phosphate buffered saline each, antibodies were eluted with 0.1 m glycine, pH 2.7 and were immediately neutralized with 1 m Tris–HCl, pH 9.0.

Immunoblot analysis

Arabidopsis total leaf extracts or isolated chloroplasts were dissolved in 2× Laemmli buffer and the proteins were separated on SDS-PAGE followed by transfer to the polyvinylidene fluoride (PVDF) membrane (Bio-Rad), which was then blocked with 5% (w/v) non-fat dried milk in TBST buffer at room temperature for 1 h. Primary antibodies were added to the blocking solution at various dilutions and incubation was continued overnight at 4°C. The PVDF membrane was then incubated with horseradish peroxidase (HRP)-conjugated goat anti-mouse (diluted 1:20 000; Bio-Rad) or goat anti-rabbit sera (diluted 1:75 000; Bio-Rad) for 30 min at room temperature followed by six washes with TBST and detection using a chemiluminescence kit (Sigma). The TGD4 antibodies were diluted 1:500. BIP antibodies (diluted 1:500) were purchased from Santa Cruz Biotechnology (www.scbt.com); a monoclonal His-tag antibody (diluted 1:2000) and HA-antibodies (diluted 1:5000) were purchased from Sigma. SMT1 antibodies (diluted 1:200) were purchased from Agrisera (www.agrisera.com). TOC75 (diluted 1:3000) and TIC110 (diluted 1:3000) antibodies were kindly provided by John Froehlich, Michigan State University while the TOC159 (diluted 1:2000) antibody was kindly provided by Masato Nakai, Osaka University.

Chloroplast isolation and proteinase digestion

Intact Arabidopsis chloroplasts were purified by discontinuous Percoll (Sigma) gradient (Aronsson and Jarvis, 2002). To perform thermolysin and trypsin digestions, 10 μg chlorophyll equivalent chloroplasts were incubated with 0–4 mg ml−1 thermolysin (Sigma) or 0–0.8 mg ml−1 trypsin (Sigma) in digestion buffer (330 mm sorbitol, 50 mm HEPES–KOH pH 8.0, 5 mm MgCl2) at 100 μl total volume on ice for 30 min. 1% (v/v) TritonX-100 was added to the sample that contained the lowest amount of either proteinase as the positive control. The digestion was terminated by addition of 50 μl 20 mm EDTA or 50 μl 0.2 mg ml−1 trypsin inhibitor. After passage through a 40% Percoll cushion and washing with digestion buffer once, proteinase-digested intact chloroplasts were dissolved in 10 μl 2× Laemmli buffer and processed for SDS-PAGE and immunoblotting.

To test the interaction strength between TGD4 and the outer envelope, 10 μg chlorophyll equivalent chloroplasts of the wild type were treated with hypotonic buffer (10 mm MOPS–NaOH, 4 mm MgCl2) or reagents as indicated in Figure 5(d) on ice for 30 min followed by centrifugation at 100 000 g for 1 h. The protein compositions of both the supernatant and the pellet were examined by SDS-PAGE.

Membrane fractionation

Arabidopsis ER enriched microsomes were isolated from 4-week-old seedlings as described (Chen et al., 2002). Briefly, seedlings were homogenized using a pre-chilled mortar and pestle in grinding buffer that contained 50 mm Tris–HCl, pH 8.2, 20% (v/v) glycerol, 5 mm MgCl2, 1 mm dithiothreitol, 2 mm EDTA and protease inhibitor cocktail (Roche). The homogenate was then filtered through Miracloth and centrifuged at 12 000 g for 15 min. The supernatant was centrifuged again at 100 000 g for 1 h. The resulting microsomes were resuspended in 0.5 ml buffer that contained 10 mm Tris–HCl, pH 7.5, 10% (w/v) sucrose, 5 mm MgCl2, 2 mm EDTA, 1 mm dithiothreitol and protease inhibitor cocktail. The microsome suspension was separated on a 20–50% (w/v) continuous sucrose gradient at 100 000 g for 16 h at 4°C. Fractions of 1 ml were collected and processed for SDS-PAGE and immunoblotting.

Accession number

Sequence data from this article can be found in The Arabidopsis Information Resource under the name At3g06960.

Acknowledgements

This work was supported in parts by grants from the US National Science Foundation, MCB 0741395, and the US Department of Energy, Basic Energy Sciences, DE-FG02-98ER20305 to Christoph Benning. We thank Dr Rebecca Roston at Michigan State University for comments and critical reading of the manuscript. We are indebted Dr John Froehlich at Michigan State University and Dr Masato Nakai at Osaka University for kind provision of TOC 75, TIC110 and TOC159 antibodies. We also thank Dr R. Michael Garavito at Michigan State University for the pLW01/DsRED–His vector.

Ancillary