Distinct “Assisted” and “Spontaneous” Mechanisms for the Insertion of Polytopic Chlorophyll-binding Proteins into the Thylakoid Membrane*

The biogenesis of several bacterial polytopic membrane proteins has been shown to require signal recognition particle (SRP) and protein transport machinery, and one such protein, the major light-harvesting chlorophyll-binding protein (LHCP) exhibits these requirements in chloroplasts. In this report we have used in vitro insertion assays to analyze four additional members of the chlorophyll-a/b-binding protein family. We show that two members, Lhca1 and Lhcb5, display an absolute requirement for stroma, nucleoside triphosphates, and protein transport apparatus, indicating an “assisted” pathway that probably resembles that of LHCP. Two other members, however, namely an early light-inducible protein 2 (Elip2) and photosystem II subunit S (PsbS), can insert efficiently in the complete absence of SRP, SecA activity, nucleoside triphosphates, or a functional Sec system. The data suggest a possibly spontaneous insertion mechanism that, to date, has been characterized only for simple single-span proteins. Of the membrane proteins whose insertion into thylakoids has been analyzed, five have now been shown to insert by a SRP/Sec-independent mechanism, suggesting that this is a mainstream form of targeting pathway. We also show that PsbS and Elip2 molecules are capable of following either “unassisted” or assisted pathways, and we discuss the implications for the mechanism and role of SRP in chloroplasts.

The insertion of membrane proteins is a complex process in which two major obstacles must be overcome: the transfer of hydrophobic regions into the bilayer with the correct final topology and the efficient translocation of hydrophilic domains to the trans side of the bilayer. Many such proteins are believed to insert post-translationally into bacterial, mitochondrial, and chloroplast membranes, and a variety of in vivo and in vitro approaches have been used to study the underlying mechanisms. In several cases it has been shown that the proteins follow an "assisted" pathway that involves protein transport machinery in both the soluble phase and the target membrane. Escherichia coli has been a popular model system, and a range of membrane proteins has been shown to rely on the cytosolic signal recognition particle (SRP), 1 which is a complex of 4.5 S RNA molecule and Ffh protein, a homolog of the 54-kDa component of eukaryotic SRPs that target proteins to the endoplasmic reticulum (1)(2)(3)(4)(5). It is generally believed that this factor docks with the SecYEG translocon in the inner membrane, and this has been confirmed recently for at least one SRP substrate (6 -8). A further component, FtsY, has been suggested to serve as a soluble SRP receptor that transfers SRP substrates to the SecYEG complex (6,9).
Other membrane proteins appear to use simpler insertion mechanisms that require none of the known protein transport machinery characterized to date for the targeting of either hydrophobic or soluble proteins. In E. coli, two single-span proteins, the coat proteins of the M13 and pf3 phages, have been shown to insert into the inner membrane without the aid of SRP, SecA, or the membrane-bound Sec apparatus (10,11). These proteins may thus insert spontaneously into the bilayer. A similar mechanism may apply to other bacterial inner membrane proteins, but none has been characterized in vitro, and the in vivo analyses have tended to offer a less direct mode of analysis.
The chloroplast thylakoid membrane has also been employed as a model system for the study of membrane protein biogenesis, in part because the component photosynthetic proteins have been characterized intensively and partly because the membrane itself has been shown to insert or import a wide range of proteins in vitro (for review see Ref. 12). This may reflect the relative ease with which these membranes can be isolated in a purified form. The chloroplast is prokaryotic-like in many respects, having probably evolved from an endosymbiotic cyanobacterium, and it contains a Sec system for the transport of proteins into the lumen (13)(14)(15) as well as a stromal SRP molecule containing a homolog of the 54-kDa protein (cpSRP54 (16)). As with bacteria, two distinct pathways have been identified for the insertion of membrane proteins, one of which is assisted, and the other of which is SRP/Sec/⌬pHindependent. One imported multispanning membrane protein, the major light-harvesting chlorophyll-binding protein of photosystem II (Lhcb1, but usually termed LHCII or LHCP) has been shown to require SRP together with GTP for insertion into the membrane (17), and proteolysis of thylakoids blocks this insertion process (18,19) indicating the involvement of proteintargeting apparatus (probably the Sec apparatus, but this remains to be confirmed). In broad terms this targeting pathway therefore resembles the bacterial SRP-dependent pathway for membrane proteins, but significant mechanistic differences may exist because no RNA molecule has been identified in stromal SRP, and the SRP54 subunit instead forms a complex with a novel 43-kDa subunit (20).
A very different SRP/Sec-independent pathway has been identified for a subset of thylakoid membrane proteins: CF 0 II and the X and W subunits of photosystem II (PsbX, PsbW). Each of the mature proteins contains a single transmembrane span, and each is synthesized in the cytosol with a bipartite presequence in which a typical "envelope transit" signal is followed by a cleavable signal peptide. These proteins insert into thylakoids in the absence of SRP, SecA, nucleoside triphosphates (NTPs) or a ⌬pH (21)(22)(23), and there is strong evidence that the Sec apparatus is likewise not involved: proteolysis of thylakoids abolishes Sec-dependent transport of lumenal proteins yet has no effect on the insertion of these membrane proteins (19). It has been suggested that these proteins may insert spontaneously into the membrane, and the signal peptides probably serve an unusual function by simply providing an additional hydrophobic region that assists insertion through the formation of a loop intermediate (24). In this respect the insertion of these proteins may resemble that of M13 coat protein, which is the only other membrane protein known to be synthesized with a signal peptide but inserted by a Sec-independent process (10).
To date, Sec/SRP-independent insertion processes have only been characterized for simple single-span membrane proteins, and the SRP-dependent process has been characterized only for LHCP. In this study we have analyzed the insertion of four additional members of the extended light-harvesting chlorophyll a/b-binding protein (CAB) family. We show that two members resemble LHCP in that stromal factors, NTPs, and protein transport machinery are all absolutely essential for integration. In contrast, two other members, early light-inducible protein 2 (Elip2) and photosystem II subunit S (PsbS), can insert efficiently in the complete absence of these factors, suggesting a possible spontaneous insertion mechanism. A portion of the population of Elip2 and PsbS molecules does, however, appear to utilize the assisted pathway described above, indicating that some thylakoid proteins are capable of following parallel insertion pathways.

Synthesis of Precursor
Proteins-EST cDNA clones encoding Arabidopsis CAB proteins were obtained from the Arabidopsis stock center at Ohio State University. Full-length cDNAs were characterized, and both strands of each cDNA were fully sequenced; the full sequence details of these clones together with those encoding other CAB proteins will appear elsewhere. 2 The stock numbers for the clones are 93I7T7, 37A1T7, VCVCD09, and 137M5T7 for cDNAs encoding Lhca1, Lhcb5, Elip2, and PsbS, respectively. The precursor proteins were prepared in vitro by transcription of the cDNAs using T7 or T3 RNA polymerase followed by translation in a wheat germ lysate in the presence of [ 35 S]methionine. Other precursors were synthesized as detailed in Ref. 19, and a cDNA encoding petunia pLHCP was kindly provided by Dr. Paul Viitanen.
Integration Assays-Isolated thylakoid membranes were prepared from intact pea chloroplasts, as described in Ref. 25. Thylakoid membranes were washed twice with ice-cold 10 mM Hepes-KOH, pH 8.0, and 5 mM MgCl 2 (HM) and were then resuspended in HM or stromal extract to a concentration of 0.5 mg ml Ϫ1 chlorophyll. Stromal extract was prepared by lysing intact chloroplasts in HM at 1.0 mg ml Ϫ1 chlorophyll. Integration assays (50 l) contained 20 g of chlorophyll and 5 l of in vitro-translated precursor protein mixture (treated with 0.1 mg ml Ϫ1 final concentration of puromycin). Where appropriate, the thylakoids and translation mixture were preincubated on ice in the presence of inhibitors. Apyrase was used to deplete the assay of NTPs as detailed in Hulford et al. (26). Assays were performed in an illuminated water bath (intensity 150 mol photons m Ϫ2 s Ϫ1 , 26°C) for 20 min and were terminated by the addition of 1 ml of ice-cold HM followed by reisolation of the thylakoid membranes for direct analysis, treatment by protease, or urea extraction. All samples were analyzed by SDS-polyacrylamide gel electrophoresis and fluorography. Insertion efficiencies were quantitated by measurement of the protease-resistant degradation product using a PhosphorImager and are shown quantitated as the percentage of available precursor in the insertion reaction.
Urea Extraction-The technique was adapted from Breyton et al. (27). Thylakoid membranes (equivalent to 10 or 20 g of chlorophyll) were washed in 1 ml of 20 mM Tricine-NaOH, pH 8.0, centrifuged at 17,000 ϫ g for 5 min at 4°C and the supernatant carefully removed. The pellet was resuspended in 100 l of freshly made 6.8 M urea and 20 mM Tricine-NaOH, pH 8.0, and incubated for 10 min on the bench (22-24°C). The sample was then subjected to two freeze-thaw cycles (solid CO 2 /room temperature) before being centrifuged at 50,000 rpm (135,000 ϫ g) for 15 min at 4°C in a Beckman TL100 benchtop ultracentrifuge, using the TLA100.3 rotor. The first 80-l supernatant was withdrawn carefully to avoid contamination with the pellet, and the remaining supernatant was discarded (this inevitably resulted in the loss of a very small quantity of the pellet). The pellet was resuspended once more in 100 l of 6.8 M urea and 20 mM Tricine-NaOH, pH 8.0, and the extraction process was repeated. Equivalent amounts of the first supernatant and the second pellet were analyzed by SDS-polyacrylamide gel electrophoresis and fluorography. The second round of extraction rarely resulted in the removal of any extra material from the thylakoid membranes.
Trypsin Treatment of Thylakoid Membranes Before Integration Assays-The protocol was as described in Ref. 19, with the following modifications. After activation of the chloroplast ATPase by illuminating intact chloroplasts in the presence of dithiothreitol and Ca 2ϩ , isolated thylakoids in HM were incubated with or without 60 g ml Ϫ1 trypsin for 10 min on ice. Both samples were washed with HM buffer containing 120 g ml Ϫ1 trypsin inhibitor, and the membranes were reisolated by centrifugation at 50,000 rpm (135,000 ϫ g) for 6 min at 4°C in a Beckman TL100 ultracentrifuge, using the TLA100.3 rotor. The thylakoids were washed twice further with HM and 60 g ml Ϫ1 trypsin inhibitor and similarly reisolated before being resuspended in stromal extract (from chloroplasts lysed at 1.0 mg ml Ϫ1 chlorophyll) or HM buffer. Each assay contained (in 50 l) thylakoids equivalent to 20 g of chlorophyll in stromal extract or HM, 60 g ml Ϫ1 trypsin inhibitor, 0.5 mM MgATP, and 10 l of puromycin-treated in vitro translation mixture (which also contains ATP at 1.2 mM). The assay was set up in a darkroom under a green safelight, and the incubation was carried out for 30 min at 26 -27°C. Incubations were terminated by the addition of 1 ml of ice-cold HM followed by centrifugation at 17,000 ϫ g for 15 min at 4°C. There were no discernible differences in the recovery of membranes between trypsin-treated and control samples. One half of each sample was analyzed directly, the other half by thermolysin treatment (0.2 mg ml Ϫ1 thermolysin, 2.5 mM CaCl 2 , 40 min on ice). Samples were analyzed by SDS-polyacrylamide gel electrophoresis and fluorography.

Structures of CAB Family
Proteins-CAB proteins form a diverse family of pigment-binding proteins that apparently originated through internal gene duplication from an ancestral single-span protein (for review, see Ref. 28). Most plant CAB proteins have three transmembrane spans, and the structure of the most abundant member (LHCP) has been solved by electron crystallography (29). In this report we have studied the insertion of four Arabidopsis thaliana CAB proteins; these are the Lhca1 protein (Lhca corresponds to the light-harvesting complex of photosystem I), Lhcb5 (Lhcb ϭ light-harvesting complex of photosystem II), PsbS, and Elip2. Elips are early light-inducible proteins that appear very rapidly after the onset of illumination and then decline rapidly in abundance. Their function is unclear, but roles in transient binding of chlorophyll molecules or in energy dissipation have both been suggested (28). All of the proteins exhibit diagnostic conserved transmembrane regions (the first and third regions in the Lhca1, Lhcb5, and Elip2 sequences (28)). PsbS is an unusual component of the photosystem II core complex which is believed to contain four transmembrane spans (30,31). The full sequence details will be presented elsewhere. 2 Assisted Insertion Mechanisms for Lhca1 and Lhcb5-In this study we used assays for the insertion of proteins into isolated thylakoids, and as criteria for correct insertion we used protease protection assays and urea washing of the membranes. To characterize the protease resistance of the authentic mature proteins we used thylakoids containing radiolabeled mature protein that had been imported into intact chloroplasts. Once inside the organelles, many membrane proteins have been found to insert efficiently and correctly into the thylakoid membrane, and incorrect insertion has not been documented to date. Fig. 1 shows chloroplast import assays using the precursors of Lhca1 and Lhcb5. The precursors are imported and sorted efficiently to the thylakoids (lanes T); incubation of these membranes with thermolysin (lanes Tϩ) reveals that Lhca1 is wholly resistant to digestion, whereas Lhcb5 is clipped to a slightly smaller size. The translation products are quantitatively digested to low molecular mass peptides by the same concentrations of protease (lanes Trϩ), confirming that the observed protease resistance of the mature proteins is caused by their transmembrane configurations. Fig. 2 shows assays for the insertion for these two proteins into thylakoids, in which we tested the requirement for added stromal extract or the effects of pretreating the incubations with apyrase. Apyrase hydrolyzes all available NTPs and totally blocks targeting by the SRP-and SecA-dependent pathways, both of which depend entirely on NTPs (26,32,33). Incubation of both pre-Lhca1 and pre-Lhcb5 with thylakoids and stromal extract leads to the generation of mature protein because of the action of stromal-processing peptidase, which removes the envelope transit domain. Under these conditions, a significant portion of the population of mature protein molecules is resistant to further digestion. The appearance of the protease-resistant Lhca1 and Lhcb5 depends almost entirely on the presence of stromal extract and is totally abolished by preincubation with apyrase. These data indicate strongly that the protease-resistant protein represents inserted protein and that insertion requires stromal factors and NTPs. Only one point remains unclear; mature Lhcb5 generated in chloroplast import assays is digested slightly by thermolysin, whereas the same concentration of protease does not digest the inserted protein in thylakoid insertion assays. One possibility is that this difference reflects differing states of assembly of the protein in the two types of assay, but further work will be required to test this. Other tests have confirmed that the proteaseresistant Lhca1 and Lhcb5 are indeed inserted into the membrane; the proteins are recovered only in the pellet fraction after extensive urea washes as used below for PsbS and Elip2 (not shown). In general, therefore, Lhca1 and Lhcb5 appear to resemble LHCP in that both stromal factors and NTPs are required for insertion into thylakoids. Future studies should reveal whether the same stromal factors are involved in each case.
PsbS and Elip2 Are Able to Insert by an SRP/Sec-independent Mechanism-Our primary aim in this study was to determine whether all of the CAB family members use similar insertion pathways, and PsbS and Elip2 were found to be good test subjects because of their ability to insert efficiently into isolated thylakoids. Again, acquisition of protease resistance was used as a criterion for membrane insertion. Fig. 3A shows chloroplast imports of pre-Elip2 in the absence or presence of nigericin, a proton ionophore that dissipates the thylakoidal ⌬pH. This compound markedly inhibits the insertion of LHCP leading to its appearance in the stroma (32). However, no effect is observed on the import characteristics of Elip2, and the protein is found almost entirely in the thylakoid fraction in both panels. Incubation of the thylakoids with 50 g/ml proteinase K (lanes Tϩ) leads to the appearance of two degradation products (DP1 and DP2). Assays for the import of Elip2 into thylakoids (Fig. 3B) show that the thylakoid-associated protein is converted to the same two products by proteinase K, and comparison with authentic mature Elip2 from a chloroplast import shows that the DP1:DP2 ratio is identical when generated by even 100 or 150 g/ml proteinase K. Both concentrations of proteinase K digest the translation product essentially to completion. These data indicate that DP1 and DP2 are diagnostic of correct insertion and that Elip2 has inserted correctly into the isolated thylakoids. Insertion is more efficient in the presence of stromal extract but nevertheless occurs efficiently in the complete absence of stroma.
The effects of apyrase and the presence/absence of stromal extract are shown in Fig. 4, and the efficiency of insertion was quantitated by measurement of the [ 35 S]methionine in the degradation products. Insertion efficiencies are calculated on the assumption that the degradation products contain the same number of labeled residues as mature size protein (because the sites of cleavage are not known). The figures given are therefore certain to be underestimates because analysis of the primary sequence shows that methionine residues must be lost when Elip2 is converted to this size of degradation product by cleavage at either the NH 2 or COOH terminus (not shown). In the absence of stroma, insertion of Elip2 is again observed as demonstrated by the appearance of DP1 and DP2. The presence of stroma again enhances the insertion efficiency approximately 2-fold, and the presence of apyrase reduces import efficiency to about the same level as observed in the absence of stroma. These data strongly suggest that a proportion of the molecules use an assisted pathway that depends on stromal factors and NTPs, whereas just over half of the inserting molecules do so in the complete absence of stroma/NTPs. The lower panel of Fig. 4 is a control to verify the effectiveness of apyrase in this particular experiment. i33K is an intermediate size construct that is imported into the thylakoid lumen by the ATP-dependent Sec system (26), and the data show that the appearance of protease-protected, mature 33-kDa protein is stimulated by the presence of stromal extract but completely blocked by apyrase treatment. Similar tests were carried out on PsbS. Fig. 5 shows a chloroplast import assay in which the precursor protein (pPsbS) is efficiently imported, processed to the mature size, and localized in the thylakoid membrane (lane T). Incubation of the thylakoids with 50 g/ml proteinase K converts the 24-kDa mature protein to a series of degradation products with masses of between 12 and 15 kDa. The larger product (DP1) becomes less prominent at higher concentrations of proteinase K (100 and 150 g/ml), but the remaining products (DP2-4) are relatively resistant even to these very high concentrations of protease. The translation product is completely digested by all three concentrations of proteinase K, and labeled digestion products are only observed comigrating with the dye front (df). Incubation of pPsbS with isolated thylakoids in either the presence or absence of stromal extract leads to the appearance of the same four DPs (Fig. 5, lower panel), and the relative intensities of the four bands are essentially identical when thylakoids are digested from chloroplast or thylakoid import assays. This finding represents strong evidence that PsbS is able to insert into thylakoids. Calculation of the insertion efficiency is made more difficult by the presence of only a single methionine in the mature protein (the presequence contains two), hence some degradation products may not be apparent in this type of assay. However, even taking this into account, our experience is that PsbS inserts into thylakoids with relatively high efficiency (e.g. slightly more efficiently than Elip2 or pLHCP).
The effects of apyrase are shown in Fig. 6. As with Elip2, insertion is stimulated by the presence of stroma, and apyrase reduces this enhanced level of insertion to the figure observed in the absence of stroma. The presence of stromal extract leads to the processing of the precursor protein to the mature size but insertion proceeds in the absence of any apparent processing in the "Ϫstromal extract" incubation, indicating that the full precursor is competent for insertion. The data indicate that, in this FIG. 2. Insertion of Lhcb5 and Lhca1 into thylakoids requires stromal extract and NTPs. pLhca1 and pL-hcb5 were incubated with isolated, washed pea thylakoids (as detailed under "Materials and Methods") in the presence or absence of added stroma as indicated. Other samples were preincubated with apyrase. After incubation, samples were washed once and analyzed immediately or after treatment with thermolysin (therm) under the conditions used in Fig.  1. Lanes Trϩ, thermolysin-treated translation mixture as in Fig. 1.   FIG. 3. Import of Elip2 into intact chloroplasts and insertion into isolated thylakoids. Panel A, the precursor of Elip2 (pre-Elip2) was incubated with intact chloroplasts and samples subsequently fractionated and analyzed as detailed for Lhcb5 and Lhca1 in Fig. 1. Lanes Tϩ, thylakoid membranes were incubated with 50 g/ml proteinase K for 30 min on ice. Import incubations were carried out under control conditions or in the presence of 2 M nigericin. Panel B, pre-Elip2 translation product (Tr) was incubated with isolated pea thylakoids in the absence or presence of stroma as indicated (Ϫstroma, ϩstroma). After incubation, samples of washed thylakoids (lanes T) were treated with 100 or 150 g/ml proteinase K as indicated above the lanes. The same proteinase K treatments were carried out on thylakoids (T) from a chloroplast import reaction (chloro.) carried out as in panel A and on translation mixtures (transl.) incubated with stroma.

FIG. 4. Elip2 can insert into thylakoids in the absence of stromal extract or NTPs.
Pre-Elip2 and i33K (lanes Tr) were incubated with thylakoids as in Fig. 3 in the absence or presence of stromal extract (SE) as indicated. Incubations were also carried out in the presence of stromal extract after pretreatment with boiled apyrase (BAp) or active apyrase (Ap) as indicated above the lanes. After incubation, samples of washed thylakoids were analyzed directly or after incubation (Ϫ or ϩ, respectively) with either 50 g/ml proteinase K for 30 min on ice (for Elip2) or 200 g/ml thermolysin under the same conditions (for i33K imports). Lane Trϩ, Elip2 translation product incubated with the proteinase K under the same conditions. The symbols are as in Fig. 3. Insertion efficiencies are shown quantitated as detailed under "Materials and Methods." experiment, just over half of the PsbS inserts by an assisted mechanism with the remainder able to insert in the absence of NTPs or stromal factors.
Although protease resistance is used widely as an indication that membrane insertion has taken place, we sought an alternative criterion, and urea washing has proven to be very useful. Breyton et al. (27) have shown that this form of washing is highly effective at removing extrinsic, non-inserted proteins from thylakoid membranes, and we have found the same to be true in tests on newly inserted thylakoid membrane proteins (34). Fig. 7A shows a Coomassie-stained gel that illustrates the effects of urea washing on some of the major thylakoid proteins. Known extrinsic proteins such as the 33-kDa and 23-kDa oxygen-evolving complex proteins are quantitatively removed and recovered in the supernatant, whereas LHCP, the major visible staining band, is resistant and hence found in the pellet fraction. As a control for thylakoid insertion assays we used LHCP as shown in Fig. 7B. Incubation of the petunia precursor protein (pLHCP) with thylakoids and stromal extract leads to the appearance of mature LHCP (lane T), and digestion of the membranes with thermolysin (Tϩ) yields resistant mature-size protein that is diagnostic of correct insertion (19,35). Urea washing of undigested membranes generates pellet and supernatant fractions, each of which contains both pLHCP and mature LHCP. When the import reaction is preincubated with apyrase, insertion is completely blocked, and no protease-resistant LHCP is apparent in the Tϩ lane, as found previously (19). Notably, the thylakoid-associated LHCP is almost entirely urea-extractable, and virtually no protein is found in the pellet fraction. This result demonstrates that urea washing is highly effective at discriminating between inserted and non-inserted LHCP.
Similar tests on Elip2 and PsbS are shown in panels C and D. In both cases, a significant proportion of the population of thylakoid-associated protein is found to be resistant to urea Figs. 1 and 3. Samples of either the translation product or thylakoids from an import reaction (as indicated above lanes) were incubated with proteinase K for 30 min on ice at concentrations indicated above the lanes in g/ml. df, dye front. Lower panel, pPsbS was incubated with isolated thylakoids in the presence or absence of stroma, and samples of the washed thylakoids (T) were incubated with proteinase K at 50 or 100 g/ml. Samples of thylakoids from a chloroplast import reaction (chloro.) were treated in the same manner. Symbols are as in the top panel.

FIG. 5. Import of PsbS into chloroplasts and insertion into thylakoids. Top panel, pPsbS (lane Tr) was incubated with pea chloroplasts, and samples were analyzed and fractionated as detailed in
FIG. 6. Effects of apyrase on the insertion of pPsbS into thylakoids. pPsbS (lane Tr) was incubated with thylakoids in the absence or presence of stromal extract (SE). Further incubations carried out in the presence of stromal extract were preincubated with boiled apyrase (BAp) or active apyrase (Ap) as indicated. Samples of thylakoids were washed once and analyzed immediately (Ϫ) or after incubation with 100 g/ml proteinase K (ϩ). Symbols are as in Fig. 5.

FIG. 7. Use of urea washing to identify inserted CAB proteins.
Panel A, pea thylakoids (T) were washed with urea, and samples of the pellet (P) and supernatant (S) fractions were analyzed. Indicated on the Coomassie-stained gel are the ␣ and ␤ subunits of the CF 1 CF 0 -ATPase, the lumenal 33-kDa and 23-kDa proteins of the oxygen-evolving complex and LHCP. Panel B, petunia pLHCP was incubated with pea thylakoids under control conditions or after pretreatment with apyrase. After incubation, samples of washed thylakoids (T), thylakoids treated with 200 g/ml thermolysin (Tϩ), and samples of the pellet (Pel) and supernatant (Sn) fractions were analyzed after washing of non-protease-treated thylakoids with urea. Panel C: left, experiment identical to that in panel B except that pElip2 was used as precursor; right, samples of thylakoids after the insertion reaction (T), the same thylakoids after incubation with 100 g/ml proteinase K (Tϩ), and the pellet/supernatant fractions after urea washing of protease-treated thylakoids. Panel D, as in panel C, left, except using pPsbS as substrate. Lanes Trϩ show protease-treated translation products in all cases. extraction and hence inserted. In the Elip2 experiment we also treated an aliquot of the membranes with proteinase K after the import incubation (as described in Fig. 4), and the right panel shows that the two DPs are completely resistant to urea extraction, confirming that they are fully integrated into the membrane. Import incubations carried out after apyrase treatment show a slightly greater proportion of protein in the supernatant fraction, consistent with the lowering of insertion efficiency observed above.
A Functional Sec System Is Not Required for the Insertion of Elip2-Elip2 and PsbS can clearly insert into thylakoids in the absence of SRP, SecA or NTPs, but as a final test we sought to determine whether the membrane-bound transport machinery (probably SecYEG) is required for their insertion. This type of analysis is complicated by the finding that insertion of LHCP is highly dependent on the thylakoidal proton motive force (32), and in previous studies on petunia LHCP we found that almost no insertion took place in the presence of nigericin (19). Even mild proteolysis of thylakoids blocks their ability to generate any ⌬pH by photosynthetic electron transport (19), and tests for the involvement of surface-bound transport apparatus therefore require that a ⌬pH is generated by other means. We have shown that this can be achieved by driving the CF 1 CF 0 -ATP synthase in reverse in the dark through the addition of ATP because this protein is highly resistant to trypsin digestion (19). This procedure was used in the experiment shown in Fig. 8. In the control experiments, a ⌬pH of 2.1 was generated, and this is sufficient to support the insertion of pLHCP into the membrane. In other tests (not shown) it was found that these thylakoids were also able to import p23K with high efficiency; because this protein is targeted by the ⌬pH-dependent pathway this finding provides further evidence that a high ⌬pH was generated. The Sec substrate, i33K, is also imported and processed to the mature size under these conditions (Fig. 8). A ⌬pH of equal magnitude was generated by the trypsin-treated thylakoids, but the import of i33K is blocked indicating that the membrane-bound Sec machinery has been destroyed. Identical results were found in earlier tests on i33K (19). Similar results are obtained using pLHCP as substrate; no insertion is apparent using the trypsin-treated thylakoids, underlining the importance of translocation machinery as found previously (18,19). Although neither Elip2 nor PsbS requires a ⌬pH for insertion ( Fig. 3 and data not shown) they were imported under identical conditions, and the lower panel shows that Elip2 inserts in both the control and trypsin-treated thylakoids. Insertion efficiency is lowered by the trypsin treatment, consistent with the data shown above indicating that a proportion of the population of molecules (roughly half) is targeted by an assisted mechanism. Similar tests have been carried out on PsbS, but these tests have been complicated by the fact that pPsbS is exquisitely sensitive to proteolysis. Although the thylakoids are washed extensively after the trypsin treatment, the minute quantities that remain with the thylakoids are able to digest the pPsbS molecules to a significant extent. Insertion of pPsbS is nevertheless observed (not shown), but an alternative proteolysis regime will be required before a more quantitative assessment of insertion efficiency is possible.

DISCUSSION
Previous work on the insertion of thylakoid membrane proteins has focused primarily on a very small number of proteins: one single multispanning protein (LHCP) and a series of singlespan proteins that are synthesized with signal peptides. The data from these studies showed that the requirements of these proteins differ in significant respects, in that the insertion of LHCP is totally dependent on stromal factors (including cpSRP54 but also probably others), NTPs, and protein transport machinery in the thylakoid membrane (17,19,26,32), whereas CF 0 II, PsbX, and PsbW require none of these factors for efficient insertion (19,(21)(22)(23). It has remained unclear whether these requirements simply reflect either the structural differences between multi-and single-spanning proteins or that these particular single-span proteins are synthesized with signal peptides. In this study we have addressed this question through the analysis of a number of multispanning proteins, and the data indicate that such proteins can insert by both types of mechanism. Two of the proteins, Lhca1 and Lhcb5, appear to resemble LHCP in that they likewise depend absolutely on stroma, NTPs, and protein transport apparatus.
We have yet to determine whether SRP is involved in the insertion of these proteins, but this appears highly likely on the basis of these data.
Elip2 and PsbS have very different characteristics. These proteins can clearly insert in the absence of the above factors, and we therefore propose that these proteins insert spontaneously into the thylakoid membrane, as has been suggested for CF 0 II, PsbX, and PsbW. We would, however, stress that this notion is based on the lack of requirement for any of the known targeting factors, and further tests are certainly required to determine whether any unidentified factors are involved, whether soluble or membrane-bound. For example, the wheat germ translation system could conceivably contain soluble chaperone-type molecules, and proteolysis of thylakoids does not digest every thylakoid membrane protein.
This is the first direct demonstration that such complex multispanning proteins can insert in the absence of SecA, SRP, and NTPs, and, although the sample number is still small, these data raise the possibility that this may be a mainstream pathway for multispanning proteins as well as for single-spanning proteins of the CF 0 II type. At the same time, a proportion of the population of Elip2 and PsbS molecules can be targeted by an alternative pathway when these factors are available. These molecules may of course be targeted by the SRP path- Pea chloroplasts were incubated in the light together with dithiothreitol and CaCl 2 to activate the CF 1 CF 0 -ATP synthase, after which thylakoids were prepared as detailed under "Materials and Methods." Samples were incubated on ice for 15 min Ϯ 60 g/ml trypsin after which the thylakoids were washed three times and used for import reactions in the dark in the presence of 0.5 mM ATP and stromal extract. Import incubations were carried out using wheat i33K, petunia pLHCP, and Elip2 as substrates. After incubation, samples were analyzed immediately or after protease treatment of thylakoids as detailed in previous figures.
way, and this possibility will be addressed in future studies. The precise conditions of our in vitro assays are different from those found in vivo, and it is difficult to predict from these data alone the extent to which Elip2 and PsbS insert via the "spontaneous" pathway in intact chloroplasts. However, A. thaliana knockout strains have recently been generated in which cpSRP54 is totally undetectable (36), and the insertion of LHCP and other CAB family members is seriously affected in these seedlings during the early stages of growth. Notably, the insertion of PsbS is completely unaffected, providing firm evidence that cpSRP is not required for the efficient insertion of this protein in vivo as well as in vitro and supporting the possibility that PsbS is spontaneously inserted in vivo.
These data have interesting implications for the role and mechanism of cpSRP. Cross-linking studies (5, 37) have suggested that SRPs in general bind preferentially to regions of particularly high hydrophobicity, and stromal SRP was found to bind tightly to the third, most hydrophobic region within pea LHCP but not to the other two spans (37). Accordingly, this was proposed to be the SRP binding signal, and the data suggest that the SRP pathway might be particularly important for highly hydrophobic proteins. At the very least, our data show that hydrophobicity does not correlate with an absolute requirement for SRP. Hydropathy analysis of the proteins analyzed in this study (not shown) indicates that transmembrane segment III of pea LHCP is indeed more hydrophobic than segments I or II in this protein, and this span reaches a figure of greater than 2.0 on the GES scale (38). However, the corresponding regions within Elip2 and PsbS (spans III and IV, respectively) are only slightly less hydrophobic, and it is notable that other regions within these proteins are very hydrophobic indeed, particularly the span II regions in both proteins (this region in PsbS is by far the most hydrophobic among any in the five proteins in this sample, according to this type of prediction). If, as seems likely, these regions can bind to cpSRP in vitro, our data argue that SRP cross-linking does not necessarily reflect an essential role for this targeting factor. At the same time, these may be the regions that provoke binding by SRP and which thereby initiate the targeting of some Elip2 and PsbS molecules by the assisted pathway.
In this study we have shown that two further members of the CAB family, Lhca1 and Lhcb5, are able to use a pathway that requires stroma, NTPs, and protein translocation machinery. If this involves SRP (as seems likely, but which remains to be addressed) the data for Lhca1 may have other implications for the SRP pathway. Lhca1 was chosen for analysis because all three transmembrane spans are of relatively low hydrophobicity, yet this protein is totally dependent on stroma and NTPs for integration. If cpSRP is indeed required for the targeting of this protein, this would indicate that the cpSRP binding signal is probably more complex than a simple region of high hydrophobicity.
Irrespective of the mechanism by which SRP selects its substrates, the striking observation is that Elip2 and PsbS can insert in the absence of any known targeting machinery, whereas three other members of the same family are completely reliant on each of these factors. In particular, they are totally unable to insert in the absence of stroma. A possible explanation for these data might be that three of the proteins (LHCP, Lhcb5, and Lhca1) are simply import-incompetent in the absence of cpSRP (perhaps being prone to aggregation), whereas Elip2 and PsbS are more stable in solution. However, this explanation may be far too simplistic, and an important future aim should be to determine whether the SRP requirement correlates with other structural features in this family of membrane proteins.
The spontaneous insertion pathway is of significant interest, and, given the operation of broadly similar SRP-and Sec-dependent pathway in bacteria, we predict that bacterial membrane proteins will emerge with similar insertion mechanisms. Further work on this topic is certainly merited to unravel the crucial early events in this particular insertion process, and it will be especially important to determine whether soluble factors other than SRP are involved in the soluble phase of the pathway. The lack of requirement for ATP or GTP precludes several obvious candidate "chaperone" proteins such as FtsY or members of the Hsp60 or Hsp70 family, and it is therefore possible that this type of protein can maintain insertion-competence without the aid of soluble factors. Very little is known about this element of the insertion pathway, or indeed about the actual insertion event, but proteins such as Elip2 and PsbS may be good subjects for this form of analysis.