Sec-independent Insertion of Thylakoid Membrane Proteins

A group of membrane proteins are synthesized with cleavable signal sequences but inserted into the thylakoid membrane by an unusual Sec/SRP-independent mechanism. In this report we describe a key intermediate in the insertion of one such protein, photosystem II subunit W (PSII-W). A single mutation in the terminal cleavage site partially blocks processing and leads to the formation of an intermediate-size protein in the thylakoid membrane during chloroplast import assays. This protein is in the form of a loop structure: the N and C termini are exposed on the stromal face, whereas the cleavage site has been translocated into the lumen. In this respect the insertion of this protein resembles that of M13 procoat, which also adopts a loop structure during insertion, and we present preliminary evidence that a similar mechanism is used by another thylakoid protein, PSII-X. However, whereas the negatively charged region of procoat is translocated by an apparently electrophoretic mechanism using the ΔμH +, the corresponding region of PSII-W is equally acidic but insertion is ΔμH +independent. We furthermore show that neutralization of this region has no apparent effect on the insertion process. We propose that a central element in this insertion mechanism is a loop structure whose formation is driven by hydrophobic interactions.

The biogenesis of biological membranes requires the synthesis of numerous hydrophobic proteins and their insertion into membrane bilayers, and intensive efforts have been made to understand the mechanisms involved. The majority of bacterial, chloroplast, and mitochondrial membrane proteins are inserted post-translationally, and interest has centered on the mechanisms by which these proteins are initially maintained in a soluble form, and how the hydrophobic regions are then transferred from an aqueous milieu into the apolar regions of the membrane bilayer. Some elements of the insertion events have been characterized in detail, one example being the influence of positive charges on overall topology, since these residues have a strong tendency to remain on the cis side of the membrane (1). It has also proved possible to recognize two broad categories of insertion mechanism, which can be regarded as "assisted" and "unassisted" according to whether protein translocation machinery is relied upon. In bacteria, for example, the secretion (Sec) apparatus used for the translocation of periplasmic proteins is also required for the insertion of some membrane proteins (2)(3)(4)(5). In addition, another element of the export machinery, signal recognition particle (SRP), 1 is involved in the insertion of a range of membrane proteins (6 -8). Other proteins, however, are integrated by Sec-independent mechanisms and these may well insert spontaneously into the plasma membrane (9 -12). However, whereas the overall requirements have been detailed for several membrane proteins, the actual insertion mechanisms are poorly understood in most cases. In vitro reconstitution assays favored for the study of protein translocation have been of limited use because most hydrophobic proteins are inherently unstable in aqueous solution. Few proteins have been examined in genuine mechanistic detail, although the coat proteins of the M13 and Pf3 phages are notable exceptions that have been shown to integrate into the Escherichia coli membrane by Sec-independent, possibly spontaneous mechanisms (reviewed in Ref. 9).
Our knowledge of membrane protein insertion is particularly vague in the case of thylakoid proteins. Most thylakoid proteins are imported after synthesis in the cytosol, and at least four distinct pathways have been identified for their subsequent targeting into and across the thylakoid membrane (reviewed in Refs. 13 and 14). Two of the pathways apply to thylakoid lumen proteins, which are imported by means of bipartite presequences containing envelope transit and thylakoid transfer signals in tandem. After import into the stroma, the transfer signal directs transport across the thylakoid membrane by either an ATP-dependent, Sec-related mechanism, or a very different mechanism that relies on the thylakoidal ⌬pH. After translocation, the stromal intermediates are processed to the mature forms by a lumen-facing thylakoidal processing peptidase, TPP (15). The thylakoid transfer signals of these proteins are all similar in overall structure to bacterial signal peptides, but specify translocation by only one of these mechanisms (16 -18).
Two further pathways have been identified for integral thylakoid membrane proteins. The multispanning membrane protein, LHCP, is synthesized with an envelope transit signal only, and hence integrates into the thylakoid membrane by means of information contained in the mature protein (19,20). The integration process requires GTP and a stromal homologue of the 54-kDa protein of signal recognition particles (21), and is thus likely to be similar in certain respects to the SRP-dependent pathway identified in bacteria. However, a very different mechanism has been demonstrated for three other thylakoid proteins: subunit II of the integral CF o component of the thylakoidal ATP synthase (CF o II) and subunits X and W of photosystem II (PSII-X and PSII-W). These proteins are synthesized with bipartite presequences that strongly resemble typical Sec-type signal peptides (22)(23)(24), yet they integrate into thylakoid membranes in the absence of stromal factors, nucle-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed. oside triphosphates, or a ⌬pH (25,26). Furthermore, mild proteolysis of thylakoids blocks the Sec-, SRP-, and ⌬pH-dependent mechanisms but has no effect on the insertion of CF o II, PSII-W, and PSII-X (26,27). It has therefore been proposed that they insert spontaneously into the thylakoid membrane. This particular insertion mechanism is highly unusual in that no other mainstream insertion process studied to date involves the use of cleavable signal peptides for "unassisted" membrane insertion, and M13 procoat is in fact the only other protein known to use this type of mechanism (see below). In this report we have analyzed this integration pathway in greater detail in order to elucidate mechanistic details of the membrane-insertion events. We describe a loop intermediate that is central to the insertion process and we present evidence that hydrophobic interactions are the driving force for insertion.

Synthesis of Precursor Proteins-Precursors of Arabidopsis thaliana
PSII-W and PSII-X were synthesized by transcription in vitro of cDNA clones followed by translation of capped transcripts in the presence of [ 3 H]leucine as described (24,26). Mutated forms were prepared by oligonucleotide-directed site-specific mutagenesis of cDNAs using an Amersham International kit according to the manufacturer's instructions.
Import Assays and Topological Analysis-Intact chloroplasts were prepared from leaves of 8 -9-day-old pea seedlings (Pisum sativum, var. Feltham First) by Percoll pad centrifugation and used in import assays as described (24 -26). Proteolysis of thylakoid membranes was carried out in 10 mM Hepes, pH 8.0, 5 mM MgCl 2 (HM buffer) unless otherwise specified in the figure legends, using thylakoid suspensions at 1 mg/ml chlorophyll. Sonication involved the use of a sonicating water bath at 0°C. Carbonate washing of thylakoids was carried out as detailed by Michl et al. (25).
Sequence Analysis-The sequences of PSII-X proteins were extracted from data base entries for the plastid genomes of Odontella sinensis, Cyanophora paradoxa, and Porphyra purpurea (accession numbers Z67753, U30821, and U38804, respectively).

RESULTS
Cleavage by TPP Is Not Required for the Insertion of PSII-W-Pre-PSII-W and pre-PSII-X share overall structural similarities with M13 procoat and appear likewise to insert into their target membrane without the aid of translocation machinery. However, there are significant differences in terms of insertion requirements (see below) and we have set out to investigate the thylakoid proteins in greater detail. Our first aim was to determine whether the cleavable signal-type peptides of the thylakoid proteins are used for a similar purpose: the formation of a loop intermediate in which the hydrophobic region (H-domain) spans the membrane. This is probably the most important role of the M13 signal peptide and our priority was to determine whether such an intermediate is a core feature in the biogenesis of PSII-W and/or PSII-X. We addressed this question by inhibiting the cleavage step catalyzed by TPP (whose active site is in the thylakoid lumen; Ref. 15), in order to probe the location and topology of this protein immediately prior to the final maturation step. Fig. 1 shows the structures of the relevant sections of both precursors, including the signal peptide and the mature region of the mature protein. TPP is known to depend on the presence of short-chain residues at the Ϫ3 to Ϫ1 positions in the substrate, relative to the processing site, and Ala at the Ϫ1 position is essential with even Ser and Gly unable to support efficient cleavage (28). We therefore introduced Thr at this position in both pre-PSII-W and pre-PSII-X, in the expectation that processing would be drastically affected without substantially altering the characteristics of the translocated region.
The import characteristics of the PSII-W mutant (W/A78T) are shown in Fig. 2. The wild-type precursor protein is imported into the chloroplasts, processed to the mature size, and fractionation tests confirm that this polypeptide is located exclusively in the membrane fraction (lane T). Protease treatment of the thylakoid membranes (lane Tϩ) results in digestion of the exposed C-terminal region (as found in previous studies on PSII-W (29)), and the production of a slightly smaller degradation product. The mutant protein, on the other hand, is converted to a mixture of mature-and intermediatesize proteins, the latter form presumably resulting from the action of stromal processing peptidase, which cleaves the envelope transit domains from most bipartite presequences. Importantly, this iPSII-W form is also found only in the thylakoid fraction and further tests show that it is completely resistant to carbonate extraction, because after this treatment both the intermediate and mature bands are recovered only in the pellet fraction (lane P). Since this procedure effectively removes extrinsic proteins from the thylakoid membrane (25), we can exclude the possibility that the action of TPP is required for the insertion of this protein. Thermolysin treatment of the iPSII-W results in a mobility shift that is comparable to that of mature PSII-W, confirming that at least one section of the intermediate is exposed on the stromal face of the thylakoid membrane.
A Loop Intermediate in the Insertion of PSII-W-The topology of the iPSII-W polypeptide was mapped by determining the accessibility of sites for trypsin and thermolysin cleavage. Fig.  1 shows that the PSII-W mature protein contains only a single basic residue which is located in the lumenal N-terminal region (Arg at position 5); the presequence contains many basic residues but these are all located prior to the H-domain in the signal peptide. In a loop structure these sites are predicted to be on the lumenal and stromal sides of the membrane, respectively. Fig. 3A shows that trypsin cleaves iPSII-W to a smaller form that remains significantly larger than mature-size PSII-W; low concentrations of trypsin (e.g. 10 g/ml) generate a mixture of iPSII-W and a smaller degradation product, whereas 25 g/ml is sufficient to fully convert the iPSII-W to the degradation product. Trypsin therefore cleaves the N terminus of the iPSII-W signal peptide on the stromal surface of the membrane. A loop intermediate would also contain the C terminus of the mature protein on this face of the membrane, and Fig. 3B shows that this is the case. The C terminus of mature PSII-W is exposed on the stromal face and the data in Fig. 2 showed that it is cleaved by thermolysin; Fig. 3B shows that thermolysin also cleaves a small fragment from iPSII-W, generating a degradation product similar in mobility to that produced by trypsin (compare lanes Th and Tp). Significantly, incubation of iPSII-W with both thermolysin and trypsin generates a much smaller degradation product (lane Th/Tp) indicating that trypsin and thermolysin cleave at different ends of the molecule (the N and C termini, respectively). Lane Tp/son of Fig. 3B provides final confirmation of a loop structure: thylakoids containing iPSII-W were incubated with trypsin and the thylakoids were sonicated to allow access of the protease to the lumenal side of the membrane. Both the mature PSII-W and the iPSII-W are almost quantitatively converted to a polypeptide that is smaller than mature PSII-W, indicating that cleavage has taken place after the Arg at position 5 in the mature protein, which must therefore be exposed on the lumenal side of the membrane.
Evidence Favoring a Similar Insertion Process for PSII-X-A similar approach was also taken to investigate the insertion of PSII-X because it is as yet unclear whether the three thylakoid proteins currently viewed as a group (CF o II, PSII-X, and PSII-W) actually insert by similar mechanisms. The cleavage site of pre-PSII-X was similarly mutated by the introduction of a Thr at the Ϫ1 position and the effects examined as for PSII-W. Fig. 4 shows that this mutation likewise inhibits processing by TPP although the effects are in fact more drastic; analysis of the thylakoid fraction following an import reaction shows that hardly any mature PSII-X is detectable, indicating that the action of TPP is almost totally blocked. Once again, an intermediate-size form (int) accumulates which is presumed to result from the action of stromal processing peptidase. This intermediate form is located exclusively in the thylakoid membrane showing that TPP action is not required for the insertion of PSII-X.
Topology studies were carried out on the iPSII-X polypeptide but the results were not as clear-cut as those obtained with PSII-W. Thylakoids containing the imported, mature-size wildtype protein were incubated with thermolysin, Staphylococcus V8 protease (V8), a mixture of thermolysin/V8, and trypsin, but no cleavage was apparent (left-hand panel of Fig. 5). Apparently, only a small region of mature PSII-X protrudes into the stroma. Unfortunately, the membrane-bound intermediate form is also resistant to proteolysis as shown in the right-hand panel; thermolysin and trypsin fail to generate defined cleavage products, and only V8 of the proteases tested is able to cleave the protein. However, this takes place only at very high concentrations (200 g/ml for 60 min) and some penetration into the lumen clearly occurs because the cleavage product (lane V) is similar in size to the mature protein. This presumably means that the V8 cleaves at one or both of the Glu residues in the C-terminal region of the signal peptide, just before the cleavage site (see Fig. 1). However, it is notable that this cleavage is much more efficient when the thylakoids are sonicated in the presence of V8 to allow complete access. Under these conditions the intermediate is quantitatively converted to the degradation product (lane Vs) providing evidence, albeit circumstantial, that this region is in the lumen. This would imply the formation of a loop intermediate as with PSII-W, but further tests will be required to confirm this point. . Lanes Tr, translation products. In the case of the W/A78T import, further samples of the thylakoids were subjected to carbonate washing (see "Experimental Procedures") and samples were analyzed of the pellet and supernatant fractions (P, S). Int, intermediate-size form of PSII-W; DG1 and DG2 denote proteolytic degradation products.

FIG. 3. A loop intermediate in the PSII-W insertion process.
The W/A78T mutant described in Fig. 2 was imported into chloroplasts and the thylakoid fraction isolated. A, thylakoids were incubated with trypsin (at the concentrations indicated above the lanes) for 30 min on ice. DG denotes degradation product generated from the iPSII-W polypeptide. B, the thylakoids were incubated with 200 g/ml thermolysin (Th), 25 g/ml trypsin (Tp), or 200 g/ml thermolysin followed by 25 g/ml trypsin (Th ϩTr). In the latter case, the thylakoids were washed once in 1 ml of HM buffer between protease treatments. A further sample was incubated with 25 g/ml trypsin and sonicated for 5 min (Tp ϩ son). All samples were incubated for 30 min on ice.
Evidence That Hydrophobic Forces Drive the Insertion of Pre-PSII-W-Studies in bacteria have shown that two types of force can be instrumental in driving protein insertion. Hydrophobic forces appear to be important in all cases, reflecting the loss of free energy achieved by moving hydrophobic regions into the lipid bilayer, but electrophoretic forces are also vital in the insertion of both Pf3 coat protein and M13 procoat. In both cases, insertion depends on the ⌬H ϩ (positive outside) which is believed to drive the translocation of negatively charged residues across the membrane. The translocated regions of CF o II, pre-PSII-X, and pre-PSII-W are all similar to that of procoat in terms of charge: the N-terminal regions of mature coat protein, CF o II, and PSII-W contain net charges of Ϫ3, Ϫ3, and Ϫ2, respectively, and, although mature PSII-X contains no negative charges at the N terminus of the mature protein, the translocated region is nevertheless acidic due to the two Glu residues in the C-terminal domain of the signal peptide (see Fig. 1). Although the insertion of the thylakoid proteins is unaffected by dissipation of the thylakoidal ⌬H ϩ in each case (23,24), we deemed it important to determine whether insertion might be assisted by a smaller, standing charge difference across the thylakoid membrane that might be important in the insertion process even in the absence of a photosynthetically driven ⌬H ϩ . This possibility was tested by mutating acidic residues in the translocated region of pre-PSII-W as shown in Fig. 1. In one of the mutants (W/E88Q) the net charge of this region was reduced to Ϫ1 and the second mutant, W/D82N,E83Q contained a neutral translocated region. Fig. 6 shows that both proteins are imported and efficiently inserted into the thylakoid membrane; the protease-treated chloroplasts contain only mature-size PSII-W which is located exclusively in the membrane fraction. The mutations therefore have no detectable effect on the import characteristics and we conclude that electrophoretic forces play no role in the insertion of this protein. DISCUSSION CF o II, PSII-W, and PSII-X represent an unusual group of membrane proteins in that no other membrane protein, with the exception of M13 procoat, has been found to rely on a cleavable signal peptide for membrane insertion by a Sec-and SRP-independent mechanism. As shown in Fig. 1, the three thylakoid proteins share additional common features with procoat: each contains a single transmembrane span in the mature protein and the translocated regions are all negatively charged. It was thus unsurprising that initial models for the Sec-independent insertion of CF o II (25) were based on the procoat insertion model, and this study has confirmed that in one fundamental respect the insertion of PSII-W does indeed resemble that of procoat. Both proteins adopt a loop intermediate during the insertion process, strongly suggesting that the function of the PSII-W signal peptide is to provide a second hydrophobic domain which, in concert with the transmembrane segment in the mature protein, is then able to drive the translocation of the hydrophilic intervening region into the thylakoid lumen. Interestingly, signal-type peptides appear also to form loop structures when directing soluble proteins across membranes by means of protein translocation systems (30,31).
The experimental evidence favoring a Sec/SRP-independent insertion mechanism (23-27) is consistent with the unusual evolution of the signal peptides. Genes encoding PSII-X and CF o II have been identified in cyanobacteria and also within the plastid genomes of several eukaryotic algae/diatoms, such as O. sinensis and C. paradoxa. None of these cyanobacterial or plastid-encoded proteins is synthesized with a signal-type peptide, strongly suggesting that the signal peptides were acquired after the transfer of these genes to the nucleus in higher plants. This situation contrasts starkly with "genuine" Sec-type signal peptides such as those found in the precursors of lumenal proteins, because the cyanobacterial/plastid-encoded counterparts of these proteins are invariably synthesized with classical signal peptides. We therefore believe that the signal peptides of these membrane proteins differ from Sec-type signal peptides in terms of both function and origin. However, we can detect no large-scale structural differences among the mature proteins that might explain why nuclear-encoded, imported proteins are now synthesized with signal peptides. Mature PSII-X proteins, for example, are structurally very similar in higher plants and cyanobacteria (see Fig. 1). Given these similarities, it may be simplistic to suggest that the signal peptide functions simply by providing a critical, additional hydrophobic segment, because the mature proteins in cyanobacteria and the above mentioned algae can clearly insert despite possessing only a single hydrophobic region. The only obvious difference concerns the charge distributions in the N-terminal regions; these regions are neutral in the cyanobacterial and plastid-encoded proteins whereas that of Arabidopsis PSII-X is basic due to the Lys at position 10. This could conceivably render the translocation process more difficult to the extent that the additional FIG. 5. Topological analysis of iPSII-X. PSII-X and the X/A74T mutant were imported into chloroplasts as described in the legend to Fig. 4 and the thylakoid fractions (denoted Ϫ) incubated with proteases as follows: 200 g/ml thermolysin (T), 200 g/ml Staphylococcus V8 (V), thermolysin followed by V8 (T/V), or 25 g/ml trypsin (Tp). Lane Vs, V8 was incubated with thylakoids and the vesicles were sonicated to allow access of the protease into the lumen. All protease treatments were for 30 min on ice, except the V8 incubations that were for 60 min on ice. Int, intermediate form.
FIG. 6. Electrophoretic forces are not required for the insertion of PSII-W. The pre-PSII-W mutants W/E88Q and W/D82N,E83Q (see Fig. 1) were imported into intact chloroplasts and the organelles fractionated as described in the legend to Fig. 2. Lanes Tϩ, the thylakoid fraction was treated with thermolysin as detailed in the legend to Fig. 2. hydrophobic region is required to aid insertion. However, it is also possible that this unusual insertion mechanism has been forced on these proteins as a consequence of their more complex insertion pathway; the proteins may insert co-translationally in cyanobacteria, in which case the protein may never be free in the aqueous phase and insertion may be more favorable. It is even possible that the signal peptide renders the precursor protein more stable in solution, perhaps by forming a "helical hairpin" in which the signal peptide partially masks the hydrophobic character of the transmembrane segment. Structural studies on this type of precursor protein should prove instrumental in resolving these points.
Although the insertion of these thylakoid proteins is reminiscent of M13 procoat insertion, there are equally significant differences in the mechanisms used. The first concerns the initial events in the insertion process. Gallusser and Kuhn (32) have shown that electrostatic interactions play a critical role in the insertion of procoat, in which basic residues in both the extreme N and C termini bind to the negatively charged membrane surface; removal of either set of basic residues renders procoat wholly insertion incompetent. Like procoat, pre-CF o II contains positively charged N and C termini and a similar mechanism was considered possible for this protein (25). However, the C-terminal region of pre-PSII-W is extremely negatively charged (a series of five acidic residues follows the transmembrane section), precluding any electrostatic binding between this region and the membrane surface. Further tests are required to determine whether the basic N-terminal region plays any role through an electrostatic binding to the thylakoid surface.
A second difference concerns the energetics of insertion. Most Sec-dependent bacterial proteins are at least partially dependent on the transmembrane potential for efficient insertion, but the best characterized Sec-independent proteins (M13 procoat and Pf3 coat) are entirely dependent on the ⌬H ϩ (10, 33) In both cases this has been attributed to an essential role of the membrane potential in driving the translocation of negatively charged regions through the E. coli plasma membrane by an electrophoretic mechanism (34). In contrast, dissipation of the thylakoidal ⌬H ϩ has essentially no effect on the insertion of CF o II, PSII-X, or PSII-W (23,25,26). In the absence of any apparent role for electrophoretic forces, we propose that hydrophobic interactions between the H-domains in pre-PSII-W and the membrane interior are the primary driving force for membrane insertion. It must be emphasized at this point that similar forces are clearly crucial in the procoat insertion process, because mutant forms containing positively charged translocated regions can be inserted even in the absence of a ⌬H ϩ (35). Nevertheless, our data show that the two systems have very different tolerances and it remains to be determined how such superficially similar translocated regions can be so potential dependent in E. coli but not in thylakoids. One possibility under consideration is that these findings reflect differences in the lipid compositions of the two membranes. The thylakoid membrane is most unusual in that galactolipids account for over 80% of membrane lipids, whereas phospholipids predominate in most membranes including those of E. coli. It has been proposed that one of the major thylakoid lipids, monogalactosyldiacylglycerol, has a pronounced ability to form non-bilayer structures (36) and we speculate that this property may facilitate translocation of polar regions across the thylakoid membrane. In general, the actual insertion events are poorly characterized for those proteins that insert without the benefit of proteinaceous translocation apparatus, and structural studies are clearly required for an understanding of the conformational changes taking place.