In vitro interaction between a chloroplast transit peptide and chloroplast outer envelope lipids is sequence-specific and lipid class-dependent.

Interaction of artificial lipid bilayers (liposomes) with the purified transit peptide (SS-tp) of the precursor form of the small subunit for ribulose-2,5-bisphosphate carboxylase/oxygenase (prSSU) has been studied using a vesicle-disruption assay (calcein dye release) and electron microscopy. Employing purified forms of Escherichia coli-expressed prSSU, mature small subunit, glutathione S-transferase-transit peptide fusion protein, and SS-tp in dye release studies demonstrated that lipid interaction is mediated primarily through the transit peptide. Using chemically synthesized peptides (20-mers), the lipid-interacting domain of the transit peptide was partially mapped to the C-terminal 20 amino acids of the transit peptide. Peptides corresponding to other regions of the transit peptide and control peptides promoted significantly less calcein release. Interaction between the transit peptide and the bilayer was very rapid and could not be resolved by stopped-flow fluorometry with a mixing time of <50 ms. Interaction between the peptides and bilayer was also lipid class-dependent. Disruption occurred only when the bilayer contained the galactolipid monogalactosyldiacylglycerol (MGDG). The extent of bilayer disruption directly correlated with the relative concentration of MGDG in the liposome, with maximum calcein release occurring in 20 mol % MGDG liposomes. Lipid bilayers with greater than 20 mol % MGDG could not be achieved as determined by calcein entrapment. Electron microscopy of the liposomes before and after addition of the transit peptide suggested that the transit peptide induced a dramatic reorganization of lipids. These results are discussed in light of a possible mechanism for the early steps in protein transport that may involve polymorphic changes in the envelope membrane organization to include localized non-bilayer HII structures.

Interaction of artificial lipid bilayers (liposomes) with the purified transit peptide (SS-tp) of the precursor form of the small subunit for ribulose-2,5-bisphosphate carboxylase/oxygenase (prSSU) has been studied using a vesicle-disruption assay (calcein dye release) and electron microscopy. Employing purified forms of Escherichia coli-expressed prSSU, mature small subunit, glutathione S-transferase-transit peptide fusion protein, and SS-tp in dye release studies demonstrated that lipid interaction is mediated primarily through the transit peptide. Using chemically synthesized peptides (20mers), the lipid-interacting domain of the transit peptide was partially mapped to the C-terminal 20 amino acids of the transit peptide. Peptides corresponding to other regions of the transit peptide and control peptides promoted significantly less calcein release. Interaction between the transit peptide and the bilayer was very rapid and could not be resolved by stopped-flow fluorometry with a mixing time of <50 ms. Interaction between the peptides and bilayer was also lipid class-dependent. Disruption occurred only when the bilayer contained the galactolipid monogalactosyldiacylglycerol (MGDG). The extent of bilayer disruption directly correlated with the relative concentration of MGDG in the liposome, with maximum calcein release occurring in 20 mol % MGDG liposomes. Lipid bilayers with greater than 20 mol % MGDG could not be achieved as determined by calcein entrapment. Electron microscopy of the liposomes before and after addition of the transit peptide suggested that the transit peptide induced a dramatic reorganization of lipids. These results are discussed in light of a possible mechanism for the early steps in protein transport that may involve polymorphic changes in the envelope membrane organization to include localized non-bilayer H II structures.
Targeting and transport of cytoplasmically synthesized proteins to eukaryotic organelles involves a diverse array of topogenic sequences. Chloroplasts and mitochondria are unique among organelles in that they are capable of semi-autonomous protein synthesis. Nonetheless, most of the chloroplastic and the mitochondrial proteins are synthesized as precursor proteins in the cytoplasm on free 80 S ribosomes. These precursors contain an N-terminal extension and are imported post-trans-lationally into the organelle. Upon translocation of the precursor across the target membrane(s), the N-terminal extension is removed by a specific peptidase. Examples of these N-terminal extensions include the bacterial signal sequence (1,2), the endoplasmic reticulum signal peptide (3), the mitochondrial presequence (4,5), and the chloroplast transit peptide (6,7). These peptide extensions contain the information that is both necessary and sufficient to target the precursor to the correct membrane and to promote binding and/or translocation (8).
In the past few years, several hundred chloroplast transit peptides have been identified and sequenced, and the information deposited into a data base called CHLPEP (9). All transit peptides are believed to share the common role of targeting cytoplasmically synthesized proteins to and across the chloroplast envelope, yet the data base reveals no consensus sequence. Chloroplast transit peptides vary from 25 to 125 amino acids in length and are only loosely similar due to a modular design. As a group, they contain three domains: an uncharged N terminus of 10 -12 amino acids; a central region, which is not amphiphilic, contains hydroxylated and charged amino acids, and accounts for the variable length of the transit peptide; and a C-terminal domain of 10 -12 amino acids, which lacks leucine and lysine and is predicted to form an amphiphilic ␤-strand (6,7).
The apparent lack of primary sequence similarity among chloroplast transit peptides suggests that a common secondary or tertiary structure may provide specificity for the chloroplast. However, at present neither the nature nor the position of any secondary or tertiary structure is known. One hypothesis suggests that chloroplast transit peptides have evolved to be rather flexible peptides with a minimum content of ␣-helixes or ␤-sheets, and behave as ideal random coils (7). An alternative possibility is that a specific secondary structure forms only after the transit peptide interacts with the chloroplast's lipid bilayer. Such membrane-induced secondary and/or tertiary structures could define a recognition element for the import machinery. An attractive feature of the latter hypothesis is that a membrane-induced conformation could be shared by many different transit peptides and enable a single import receptor such as the newly identified import receptor, OM-86, to bind and facilitate transport of potentially hundreds of different precursors (10 -12).
Currently most models of protein transport suggest that a "protein translocation complex" in the organelle membrane contains components that recognize the precursor and act as a receptor, while other transmembrane components function as a protein translocase. Indeed, proteinaceous components of the protein translocation complex have been identified recently in mitochondria (13)(14)(15)(16)(17) and chloroplasts (10 -12, 18).
However, these current models of protein transport across membranes largely ignore the role of lipids that make up the bilayer. The chloroplast is enclosed by an inner and outer membrane, which together constitute the chloroplast envelope. Both membranes have an unusual lipid content, containing monogalactosyldiacylglycerol (MGDG), 1 digalactosyldiacylglycerol (DGDG), sulfolipid, and the negatively charged phosphatidylglycerol (PG) (19). In fact the chloroplast outer envelope membrane is the only cytoplasmically exposed membrane in the plant cell that contains the galactolipids, MGDG and DGDG. Moreover, the outer membrane has a very high lipid/ protein ratio of 3.0, indicating that the lipid domains are largely exposed to the cytoplasm (20). It has been suggested that the unique lipids in the chloroplast envelope may play a direct or indirect role in the protein transport process (21). In fact, it has been shown that treatments which alter the lipid content of the outer envelope can cause a significant change in the protein transport activity (22). The potential involvement of lipids in protein transport is further suggested by the pronounced "membrane-active" nature of targeting sequences in general. It has been shown by a variety of techniques that bacterial and endoplasmic reticulum signal peptides (1,(23)(24)(25), mitochondrial presequences (26 -29), and chloroplast transit peptides (30 -32) are capable of interacting with artificial bilayers and monolayers which are devoid of protein components.
In this study we investigated the interactions between SS-tp and artificial bilayers whose composition mimics the chloroplast outer membrane. We utilized purified forms of prSSU, mSSU (the precursor and mature forms of ribulose bisphosphate carboxylase, respectively), SS-tp (the purified full-length transit peptide for prSSU), as well as synthetic peptides which correspond to four domains of the transit peptide. We found that the interaction between the peptides and the bilayer was sequence-and lipid class-dependent. The results are discussed in terms of possible early steps in chloroplast protein import as well as a general mechanism for peptide/lipid interactions.

EXPERIMENTAL PROCEDURES
Materials-Plant MGDG and plant DGDG were purchased from Matreya, Inc. PC and PG were purchased from Boehringer Mannheim. Synthetic peptides corresponding to regions of the prSSU transit peptide were obtained from Multiple Peptide Systems and the purity was shown to be between 60 and 90% as determined by high performance liquid chromatography (data not shown). Glutathione-agarose, S-peptide, and reduced, carboxymethylated lactalbumin, were purchased from Sigma. All other chemicals were analytical grade.
Preparation of Precursor Proteins and the Transit Peptide-Overexpressed prSSU and mSSU proteins were isolated from BL21 strain of Escherichia coli as described by Klein and Salvucci (33). Radiolabeled prSSU was obtained by growing the cells for 3 h in methionine/cysteinedeficient media, followed by incubation with Tran 35 S-label metabolic labeling reagent from ICN. We routinely solubilize the inclusion bodies in the presence of a chemical denaturant such as 6 M guanidine HCl or 8 M urea.
GST-tp, GST, and SS-tp-The proteins/peptides were isolated from GST-gene fusion system. 2 The protein of interest was fused to the C terminus of glutathione-S-transferase from Schistosoma japonicum (34) and purified from bacterial lysates by affinity chromatography using glutathione-Sepharose 4B. Fusion protein (GST-tp) was isolated under mild conditions by elution with 5 mM glutathione in phosphate-buffered saline. Transit peptide expressed in the vector pGEX-2T was obtained by cleavage with thrombin. GST was isolated by thrombin cleavage followed by glutathione elution.
In Vitro Protein Import/Competition Assays-Dwarf pea seedlings were grown in a EGC TC-30 growth chamber at 17.5°C with 165 mE/m 2 /s of cool white incandescent light (Sylvania). Intact chloroplasts were isolated via continuous Percoll. Import assays were performed as described previously (35). Briefly, 35 S-labeled prSSU (5 ϫ 10 6 cpm/g) was incubated with freshly prepared chloroplasts (1 mg/ml chlorophyll) in the presence of 3 mM MgATP for 30 min at room temperature. In competition experiments, 20 -30 molar excess of the competing protein/ peptide was added just before the radiolabeled precursor. Intact chloroplasts were pelleted over a 40% Percoll cushion. Chloroplast pellets were dissolved in 1 ϫ SDS sample buffer and subjected to SDS-PAGE, followed by autoradiography.
Thermolysin Treatment-Post-imported chloroplasts were incubated at 4°C for 15 min with 200 mg/ml thermolysin in import buffer (35). Enzyme reactions were terminated by adding EDTA to 10 mM. Intact chloroplasts were reisolated over a 40% Percoll cushion. Chloroplast pellets were dissolved in SDS sample buffer and subjected to SDS-PAGE, followed by autoradiography (35).
Liposome Preparation-Lipid vesicles were prepared by mixing appropriate mixtures of diacyl lipids (10 mM) in chloroform. Solvent was then evaporated under a stream of N 2 , and the samples were vacuum desiccated for no less than 3 h. The dried lipid film was then hydrated overnight in phosphate-buffered saline, pH 7.8, containing 1 mM EGTA, 0.02% azide, and 50 mM calcein. Small unilamellar vesicles were prepared by vortexing the lipid mixture and then sonicating in a bath sonicator (Laboratory Supplies Inc.) for 5 min. Two additional sonication cycles were performed with a 6 -12-h interval. Free unincorporated calcein was removed by chromatography on a Bio-Gel A-0.5m column equilibrated in phosphate-buffered saline/EGTA/azide buffer. Liposomes eluted in the void volume fractions showed a calcein fluorescence quenching of Ͼ70 -80% (see below). Liposomes were used at a lipid concentration of 1-2 M.
Fluorescence Quenching Measurements-Fluorescence of liposomes was performed as described (36) using a Perkin-Elmer LS 50 Spectrofluorometer. Fluorescence excitation was at 490 nm. Fluorescence quenching was calculated from the formula below.
F 0 and F t are the fluorescence of the liposome samples before and after addition of 0.1% Triton X-100, respectively. Protein/Peptide-induced Release of Calcein from Liposomes-Liposomes containing entrapped calcein were incubated at room temperature for 5-10 min with various concentrations of different proteins/ peptides as indicated in the figure legends. The fluorescence measurements were performed as described above. The percentage of calcein release was calculated using the formula below.
F 0 and F are the calcein fluorescence before and after addition of the protein/peptide, respectively, and F t is the total fluorescence after addition of 0.1% Triton X-100. All measurements were done in triplicate and the variations were within 10% of the mean. Stopped-flow Fluorescence Measurements-Measurements of the kinetics of transit peptide-induced calcein release were made on a Biologic SFM-3 stopped-flow device (Molecular Kinetics). Excitation was at 490 nm, while the fluorescence was measured at a right angle through a 515 LWP filter (Oriel, Inc). The mixing time of this instrument is Ͻ50 ms. The stopped-flow device was thermostated using a circulating water bath, and measurements were made at 23°C.
Electron Microscopy-Liposomes consisting of PC or MGDG/DGDG/ PC/PG were negatively stained with 0.1% uranyl acetate and viewed in a Hitachi 600 electron microscope operating at 75 kV. Liposomes were treated with SS-tp at room temperature for 10 min prior to staining with uranyl acetate.

RESULTS
Production and in Vitro Activity of E. coli Expressed prSSU, mSSU, GST-tp, GST, and SS-tp-To facilitate the in vitro studies described above, we utilized E. coli expression vectors that allow production of biochemical amounts of the purified 1 The abbreviations used are: MGDG, monogalactosyldiacylglycerol; DGDG, digalactosyldiacylglycerol; prSSU, precursor for small subunit of ribulose-2,5-bisphosphate carboxylase/oxygenase; mSSU, mature small subunit of ribulose-2,5-bisphosphate carboxylase/oxygenase; SStp, transit peptide for prSSU; GST, glutathione S-transferase; PC, phosphatidylcholine; PG, phosphatidylglycerol; CL, cardiolipin; PE, phosphatidylethanolamine; I.B., chloroplast import buffer; DTT, dithiothreitol; S-peptide, ribonuclease-S-peptide from bovine pancreas; Lkt A 18 , synthetic peptide from P. hemolytica leucotoxin A (amino acids 793-811); Lkt A 16 , synthetic peptide from P. hemolytica leucotoxin A (amino acids 780 -796); F H -C-term 16  proteins and peptides. Purity of the peptides/proteins was determined by SDS-PAGE, followed by Coomassie Brilliant Blue staining (Fig. 1A). Both prSSU and mSSU (tobacco) were expressed in E. coli using the pET expression system (Novagen), as described in Ref. 33. The full-length transit peptide of pea prSSU was expressed in E. coli as a C-terminal fusion protein to GST using the pGEX (Pharmacia) expression system. 2 The transit peptide was removed from GST by cleavage with thrombin. These proteins were judged near homogeneous, based on Coomassie Brilliant Blue staining.
To verify the biological activity of the proteins produced in E. coli, we performed in vitro chloroplast import assays as described in Ref. 35. These assays test the import competence of prSSU as well as the ability of other proteins/peptides to competitively inhibit the import of 35 S-labeled prSSU. Fig. 1B shows the data from an import assay using E. coli expressed 35 S-prSSU in the presence or absence of a competing protein/ peptide. A 20 M excess of cold prSSU completely blocked the import and processing of 35 S-prSSU (Fig. 1B, lane 2). As expected, mSSU, which lacks the SS-tp, had no effect on the import and processing of prSSU. SS-tp competed with prSSU for import (translocation); however, the inhibition was not complete at the concentration used in this experiment. Interest-ingly, prSSU appears to accumulate as a membrane-bound form in the presence of competing SS-tp, suggesting that SS-tp competitively blocks import and processing of the prSSU protein after binding to the chloroplast.
To confirm that the processed form of prSSU is imported and authentically processed in the chloroplast stroma, intact chloroplasts were treated with thermolysin to digest surface-exposed forms of prSSU or mSSU (Fig. 1, lanes 7-12). Thermolysin is unable to cross the outer envelope membrane (37) and, therefore, should digest only the proteins exposed or bound to the surface of the chloroplast. The majority of processed mSSU is resistant to thermolysin, indicating that mSSU is localized in the stroma (lanes 7-12). In the case where import was blocked with the SS-tp (Fig. 1A, lane 6), the form of prSSU that accumulated was still associated with the exterior of the chloroplast since it was removed by thermolysin treatment (Fig. 1A, lane  12). However, SS-tp does not block prSSU from binding to the chloroplast (lanes 6 and 12). These data indicate that SS-tp blocks prSSU translocation across the envelope membrane without affecting its ability to initially bind to the transport apparatus. Taken together, these results confirm the biological activity of the proteins/peptides used in this study.
Interaction between Chloroplast Envelope Lipids and Chloroplast Precursors-As shown in Table I, the chloroplast outer envelope membrane has a unique lipid composition in that it is high in galactolipids, specifically MGDG and DGDG. The ratio of the major components of the outer envelope is approximately MGDG:DGDG:PC:PG (2:3:3:1), with sulfolipids and phosphatidylinositol occurring as minor components (19). To mimic this outer membrane, we prepared artificial bilayers (liposomes) with phospholipids and galacto-lipids at the following molar ratio: MGDG:DGDG:PC:PG ϭ 2:3:4:1 (hereafter called OM liposomes). Calcein was encapsulated in the liposomes at a concentration of 50 mM as described in (36). At this concentration calcein is self-quenching and displays very low fluorescence emission. However, when the liposomes are disrupted by detergent solubilization, the calcein is released into the surrounding buffer and regains its characteristic fluorescent spectra. Using this fluorescence quenching assay, we demonstrated that over 70% of the calcein associated with the liposomes was entrapped (data not shown). The liposomes were stable for several months in buffer containing azide. The fluorescence emission spectra of the liposomes in the presence and absence of Triton X-100 is shown in Fig. 2A. Liposome destabilization was measured by an increase in calcein fluorescence at 515 nm as dye is released from liposomes. The effect on the fluorescence emission spectra upon interaction of the liposomes with prSSU and other proteins is shown in Fig. 2 (A and B). The data indicate that prSSU interacts with the liposomes, resulting in increasing dye release with increasing prSSU concentration. The mature form of ribulose-2,5-bisphosphate carboxylase/ox- competition of import by the overexpressed proteins/peptides. Import assay was performed using 35 S-prSSU as described in Ref. 35. After import chloroplasts were reisolated over Percoll. In lanes 6 -12, the chloroplasts were treated were thermolysin (100 g/ml) for 30 min on ice. Chloroplasts were pelleted and boiled in sample buffer containing DTT and subjected to 10 -20% gradient SDS-PAGE followed by autoradiography. Lanes 1-6, 10 g of protein loaded; lanes 7-12, 5 g of protein loaded. ygenase (mSSU) did not promote any significant calcein release even at the highest concentration tested, indicating that only the precursor form of the protein interacts sufficiently with the liposomes to cause vesicle disruption. The control protein GST also failed to induce a significant release of dye from the liposomes. Both prSSU and mSSU were synthesized in E. coli as inclusion bodies and solubilization in 8 M urea or 6 M guanadinium HCl yielding an unfolded form of the protein. To test whether other unfolded proteins could interact with the bilayer and produce a calcein release, we incubated the liposomes with reduced carboxymethylated ␣-lactalbumin, a protein that is permanently unfolded. This protein had little effect on the liposomes even at the highest concentrations tested. We also confirmed that urea alone had no effect on calcein release from the liposomes. These results indicate that only the transit peptide-containing prSSU interacts with OM liposomes in such a way that promotes calcein release. Fig. 3 shows the amino acid sequence of the peptides used in the following experiments. The transit peptide of prSSU, which consists of 58 amino acids, is both necessary and sufficient for protein transport into chloroplasts (38). As shown in Fig. 4, SS-tp induced a rapid dye release from OM liposomes. Furthermore, liposome lysis was concentration-dependent, resulting in a maximum of ϳ50% dye release at 6 -8 M peptide. Taken together with data in the previous section, these data indicate that the precursor-lipid interaction is mediated primarily through SS-tp. Furthermore, this peptideliposome interaction was specific, since the control peptides (S-peptide, Lkt A 18 , Lkt A 16 , and F H C-term 16 ) whose sequence is shown in Fig. 3, generated no significant dye release from the OM liposomes.

Precursor/Lipid Interaction Is Mediated through the C Terminus of SS-tp-
Next, we investigated the interaction between OM liposomes and four 20-mer peptides which correspond to different regions of SS-tp (Fig. 3). As indicated in Fig. 4, the C terminus of SS-tp (SS-tp-(41-60)) induced the greatest degree of liposome lysis. The N-terminal 20 amino acids also caused a significant dye release. In contrast, the more hydrophobic middle region of the transit peptide, SS-tp-(21-40) did not induce liposome destabilization since the amount of dye released was insignificant. Table II summarizes the data discussed thus far in order of decreasing activity: prSSU Ͼ SS-tp Ն SS-tp- (41- The finding that SS-tp-(41-60) was more disruptive to the OM liposomes than SS-tp- (31)(32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45)(46)(47)(48)(49)(50) suggests that the C-terminal 10 amino acids of the SS-tp are the most "membrane-active" region of the 60 amino acid SS-tp. Therefore, the precursor/ lipid interaction was mediated primarily via the extreme C terminus of SS-tp. However, the fusion protein GST-tp caused very little vesicle disruption (Fig. 2B). This construct had no more vesicle disruption activity than GST alone. This result suggests that the transit peptide must exist either as a free peptide or as an N-terminal extension of prSSU to engage the bilayer. In both cases, the N-terminal domain would be free.
Precursor/Lipid Interaction Is Lipid Class-dependent-The chloroplast outer envelope membrane contains ϳ20% MGDG, the only H II phase-forming lipid present in this membrane (19). To investigate whether the SS-tp/lipid interaction is mediated via this non-bilayer-forming lipid, we made calcein-encapsulated liposomes with varying lipid compositions. As shown in Fig. 5A, SS-tp interacted only with liposomes containing MGDG. Liposomes composed of 100% PC, PC/DGDG (7:3 molar ratio), or PC/PG (9:1 molar ratio) did not produce a significant calcein release when incubated with SS-tp. As expected (Fig.  5B), mSSU failed to promote dye release from any of the liposome compositions tested. Similar to what was seen with SS-tp, prSSU promoted dye release only from liposomes that contain MGDG.
To confirm that liposome destabilization, as measured by calcein release, was truly dependent on the presence of MGDG, we made liposomes containing varying molar percentages of MGDG. Mol % of PC was varied to compensate for the difference in MGDG. DGDG and PG concentrations were kept constant (30 and 10 mol %, respectively). As shown in Fig. 6, increasing MGDG content resulted in increased lysis of liposomes in the presence of increasing amounts of SS-tp. As the mole percentage of MGDG was increased, the percentage of calcein release also increased. Even at the maximum (20% MGDG), however, the degree of calcein release was peptide-dependent and increased linearly with SS-tp concentration.
The observation that maximum calcein release occurred at 20 mol % MGDG is interesting for two reasons. First, 20 mol % was the maximum amount of MGDG that we could incorporate into liposomes while maintaining calcein entrapment (data not shown). Second, the chloroplast outer envelope contains approximately 20 mol % MGDG in vivo. These observations suggest that 20 mol % is the maximum amount of this MGDG that can be maintained in this bilayer either in vitro or in vivo. Furthermore, our data indicate that this H II phase-forming lipid may mediate the in vitro interaction between the precursor/transit peptide and the lipid bilayer, leading to calcein release. Collectively, these findings argue that the chloroplast outer envelope resides in a meta-stable state, which is susceptible to additional destabilization by interacting with SS-tp alone or as part of prSSU.

Kinetic Analysis of Peptide/Lipid Interactions-The previous experiments have demonstrated the ability of prSSU and
SS-tp to interact with lipid bilayers in a way that causes loss of vesicle barrier function. However, these experiments reflect steady state measurements once the interaction has reached equilibrium. To investigate the rate of dye release from liposomes upon interaction with SS-tp, we performed rapid stopped-flow fluorometry. This technique allows us to rapidly mix the liposomes and peptides, and monitor the fluorescence increase as a function of time. The mixing time of our instrument was Ͻ50 ms. Nonetheless, the rate of interaction was too rapid to be resolved even by this stopped-flow device. The increase in calcein fluorescence had already reached its maximum within the mixing time of this instrument (Fig. 7). In fact, the peptide-induced release of calcein appeared to be slightly faster than the release due to solubilization with Triton X-100.
tides derived from the transit peptide share the ability to cause a rapid release of calcein from MGDG-containing bilayers. The precise mechanism leading to this calcein release is not known. These experiments are not able to distinguish whether calcein release is the result of the formation of discrete peptide-induced channels or, alternatively, whether these sequences cause structural changes in the lipid bilayer that permit calcein release. To attempt to resolve these two opposing models, we stained the liposomes with the negative stain, uranyl acetate, and visualized peptide-induced morphological changes by electron microscopy. We examined two different liposome preparations: 1) the OM liposomes that mimic the composition of the chloroplast outer envelope, and 2) liposomes containing PC alone, to serve as a control. The PC-liposomes did not disrupt when incubated with the transit peptide (Fig. 8, A and B). This observation is consistent with fluorescence data, which showed little effect of the transit peptide on release of calcein from this  . Excitation was at 490 nm while the fluorescence was measured at a right angle through a 515 LWP filter (Oriel, Inc). The mixing time of this stopped-flow chamber is Ͻ50 ms. The stopped-flow device was thermostated using a circulating water bath, and measurements were made at 23°C. liposome preparation (Fig. 5, A-C). In contrast, when the OM liposomes were treated with the transit peptide at 10 M, a marked structural change in vesicle morphology was apparent (Fig. 8, D-F). Although the kinetics of this transition have not been carefully studied, the morphological change occurred less than 2 min after peptide addition. This effect involved the conversion of spherical liposomes (Fig. 8C) into a stacked array of elliptical vesicles. Analysis of 456 particles indicated that Ͼ70% of the liposomes underwent a change in morphology from an initial spherical shape to an elliptical and/or stacked array after addition of the transit peptide (data not shown). Once formed, these elliptically shaped particles appeared to associate with one another and formed a stacked array of elliptical discs. Although the majority of the particles were found in stacked arrays with 2-5 discs, ϳ30% of the vesicles remained spherical that stained less electron dense than the aggregated vesicles. Although the significance of these morphological changes is not self-evident, they suggest that interaction with SS-tp may induce a dramatic local lipid reorganization.

DISCUSSION
In this study, we have investigated the interactions between SS-tp and artificial bilayers whose composition mimics the chloroplast outer membrane. We have utilized purified forms of prSSU, mSSU, SS-tp, and synthetic peptides which correspond to four domains of the transit peptide. We have demonstrated that the interaction of prSSU with artificial bilayers is mediated primarily via the transit peptide. The membrane interacting domain has been mapped to the C-terminal 10 amino acids of the transit peptide. The interaction is strongly dependent on the lipid content of the artificial bilayer, requiring the galactolipid, MGDG. Furthermore, the interaction is very rapid and leads to a drastic change in the morphology of the liposomes. This work represents only the second attempt to characterize the interaction of a full-length chloroplast transit peptide with artificial membranes and represents the longest transit peptide studied to date (31).
The C-terminal Domain of SS-tp Disrupts Bilayers-Fluorescence emission spectra of dye-filled liposomes in the presence of SS-tp indicate that the transit peptide region of the precursor was sufficient to destabilize membranes and promote calcein release, although the full-length precursor produced an even greater effect in the same assay. The magnitude of liposome destabilization by prSSU alone exceeded the additive effect of the transit peptide alone and mSSU alone (86% release versus 60% for SS-tp ϩ mSSU), suggesting that the mature domain enhances the vesicle disruption activity of the transit peptide in cis, when it is physically fused at its C terminus. Recently, an analogous study of the chloroplastic precursor of ferrodoxin (prFd) and its full-length transit peptide also demonstrated that the transit peptide was active in disrupting vesicles, but surprisingly, the full-length precursor was not (30). The authors suggest that the ferrodoxin mature domain somehow prevents vesicle lysis. Unknown structural differences between prSSU and prFd probably account for the dramatic difference in the ability of these two chloroplast-destined precursors to disrupt bilayers. Nonetheless, finding that the mature domains of ferredoxin and small subunit appear to modulate the membrane-interactive properties of their respective precursors (decreased for prFd, increased for prSSU) suggests that the junction between the transit peptide and mature protein may significantly influence the structure and function of the transit peptide.
We have shown that the transit peptide is able to disrupt vesicle barrier function as a free peptide, fused to the N terminus of mSSU, but not as a C-terminal domain to GST. Our data strongly suggest that the N terminus must be free in order to engage the lipid bilayer. This finding is supported by in vitro competition studies that show that neither GST-tp nor His tag -tp (a fusion protein with a His-tag and S-tag at the N terminus of SS-tp) are able to compete with prSSU for binding or import. 3 The rationale behind the requirement for a free N terminus in lipid bilayer interaction is not entirely clear; however, it could involve a mechanism similar to the "loop out mechanism" that was initially proposed to describe the interaction of signal peptides with their target membranes (39).
Recent reports indicate that prFd and its purified transit peptide both are capable of inducing a change in conductivity across the chloroplast membranes (42). The authors suggest that the increase in conductivity results from a precursormediated opening of the translocation pores. Like our study, the ferrodoxin report also implicates the transit peptide C terminus, since a mutant form of prFd (preFd-⌬7), which is deleted for the C-terminal 7 amino acids of the transit peptide, shows no ability to induce a change in conductivity, nor does the mutant precursor function as a competitive inhibitor of binding and import. Their findings are entirely consistent with our observation that the C terminus of the prSSU transit peptide is required to induce large changes in membrane permeability. Therefore, two independent laboratories have now established that the membrane-interacting activity associated with chloroplast transit peptides is mediated by a C-terminal domain that is only 7 amino acids long.
Extrapolation of these findings to the larger population of chloroplast-targeted proteins may have significant unifying implications for protein targeting to the chloroplast and its various compartments. For example, it was previously established that the stromal-targeting domain of plastocyanin does not interact with artificial bilayers that are composed of either PC/PG (molar ratio 9:1) or MG/DG/PC/PG (molar ratio 15:35: 40:10, representative of the chloroplast envelope), even at concentrations as high as 25 M (43). This result may be explained by fact that the plastocyanin transit peptide contains not only a 43-amino acid stromal-targeting domain but also an additional 23-amino acid C-terminal domain in vivo, which is proposed to function as the thylakoid-transfer domain. By analogy to our work and the findings of Bulychev and co-workers (42), the membrane-interactive region of the plastocyanin bipartite transit peptide is predicted to lie exclusively in the adjacent thylakoid-targeting domain. Consistent with this hypothesis, the putative thylakoid-targeting domain is rich in hydrophobic amino acids and is believed to resemble the signal peptide for secreted proteins (44). Furthermore, chimeric proteins containing only the 1-43 region of the plastocyanin transit peptide fused to dihydrofolate reductase failed to import into chloroplasts in vitro; however, when the chimeric protein contained at least residues 1-53 of plastocyanin, import was restored (45). Thus, with plastocyanin, the presence of a membraneactive domain as small as 10 amino acids may be sufficient to restore transport across the chloroplast envelope. These observations provide some insight into the minimal size of the membrane-interacting domain needed for successful binding and import into chloroplasts. They also suggest that the requirement for a membrane-interacting domain is a general feature of chloroplast transport and is probably independent of subcompartmentalization. Further experimentation will be required to determine if proteins localized to the inner membrane, outer membrane, and inter-membrane space of the envelope share this same targeting requirement.

SS-tp/Lipid Interaction Requires a Non-bilayer-forming
Lipid-Without exception, membranes that are active in protein translocation contain significant levels of lipids that strongly prefer to adopt a non-bilayer structure. Specific examples include phosphatidylethanolamine (PE) in the inner membrane of E. coli, cardiolipin (CL) in the endoplasmic reticulum and mitochondria, and MGDG in the chloroplast envelope and thylakoid (46). These unique lipids are believed to exhibit a wedgelike molecular shape and they prefer to form an H II phase when isolated. Several reports have suggested that these non-bilayer-forming lipids are required for protein translocation. Consistent with this notion, our work has demonstrated that the interaction between the transit peptide and the artificial bilayers is dependent on lipid composition. More specifically, this interaction is only observed when the liposomes contain MGDG. Previous monolayer insertion experiments also observed an MGDG dependence for the ferredoxin transit peptide (31) and a synthetic peptide corresponding to the last 20 amino acids of SS-tp (41).
A similar lipid class dependence has been observed for interactions between mitochondrial presequences and artificial membranes (47,48). It was first observed that the presequence to the mitochondrial cytochrome oxidase subunit IV would only interact with liposomes containing CL (47). More recently, mitochondrial presequences have been shown to induce contact sites between large unilamellar vesicles (49) and also between monolayers and large unilamellar vesicles (48). In both cases, this interaction was dependent on the presence of CL. Again, it was believed that a unique structural aspect of CL was important, since other anionic phospholipids failed to promote the interaction. Although the molecular architecture of these contact sites is not known, it was proposed that a local non-bilayer intermediate formed which CL could stabilize due to its H II phase preference (46). In support of this model, contact sites isolated from both yeast mitochondria (50) and mouse liver mitochondria (51) are substantially enriched in CL.
Possibly the strongest evidence for the essential involvement of non-bilayer lipid structures in protein transport comes from the recent characterization an E. coli mutant that is devoid of PE (52). In this strain, growth is tightly coupled to increased synthesis of CL, which in the presence of specific divalent cations (Mg 2ϩ ) will adopt a preference for the non-bilayer phase (53). These cells grow only in the presence of specific cations, indicating that polymorphic regulation of membrane-lipid composition is essential for cell viability. In vitro studies with inverted membrane vesicles from this mutant further indicate that protein translocation is dependent on either the presence of the cations listed above or by the incorporation of the nonbilayer lipid, PE. The authors conclude that non-bilayer lipids are essential for efficient protein transport across the plasma membrane of E. coli, and that cation-induced conversion of CL to a non-bilayer-forming lipid permits protein secretion and thus growth of this mutant.
The fact that the SS-tp/liposome interaction observed in this study is specific for MGDG supports a possible involvement of non-bilayer structures in chloroplast protein transport. Interestingly, the only other studies that have addressed the membrane interacting activity of a full-length higher plant transit peptide demonstrated that the ferredoxin transit peptide/lipid interaction is strongest with the anionic lipids, PG and sulfoquinouosyldiacylglycerol (31). However, even with this transit peptide, a significant interaction occurs with monolayers that contain MGDG.
SS-tp/Lipid Interaction Is Very Rapid, Resulting in Altered Liposome Morphology-The mechanism by which prSSU and SS-tp promote dye release is not known. We have measured that calcein release is very rapid (Ͻ50 ms), in contrast to several other studies that have demonstrated that peptide/lipid interactions exhibit much slower kinetics, on the order of 2-10 min. Slow kinetics have been reported for mitochondrial presequences (29,47,49), signal peptides (24), and transit peptides (31). The basis of such disparate kinetics is unknown.
Electro-physiological studies on intact chloroplasts have also demonstrated a much slower change in envelope resistance, when mediated by either the ferredoxin precursor or its transit peptide (42). The mechanism for membrane/precursor interaction in intact chloroplasts, therefore, may either be distinct from, or considerably more complex than, the interaction that we observed with protein-free liposomes. Several obvious differences between the two experimental systems may explain the opposing rates of membrane destabilization, such as protein/lipid ratio, the potential buffering capacity of protein complexes found specifically in chloroplast membranes, or the fact that the integrity of intact chloroplast is enhanced by two distinct bilayers.
Finally, we have observed a pronounced change in liposome morphology upon interaction with SS-tp. Mitochondrial presequences (29) and bacterial signal peptides (54) have also been shown to cause membrane aggregation. However, we are not aware of any other example in the literature where the morphology of individual liposomes changes so dramatically. Our OM liposomes were initially spherical and fairly homogeneous. After exposure to the SS-tp, however, the liposomes transformed into a clearly heterogeneous population containing a new component of elliptical and stacked vesicles. Although this type of electron microscopy cannot provide a three-dimensional image, it is possible that these new arrays represent a stack of disc-shaped vesicles. This observation evokes the obvious comparison to the appressed membrane stacks found in the thylakoid grana.
In summary, the results of this study provide new evidence for the potential involvement of specific precursor/lipid interactions that could be intimately involved in the chloroplastic protein-import process. Our work also suggests that chloroplastic transit peptides may universally contain a membrane interacting domain that is restricted to its C terminus and may even span into the N terminus of the mature domain. The precursor/ lipid interaction reported in this paper is dependent on the presence of the non-bilayer-forming lipid, MGDG, providing additional evidence for the universal involvement of non-bilayer structures in the insertion and translocation of proteins across membranes. Clearly, a comprehensive model of protein translocation will require new experiments that simultaneously incorporate the role of membrane lipids with the recently identified proteinaceous components of the protein translocation apparatus.