C. Robinson, Department of Biological Sciences, University of Warwick, Coventry CV4 7AL, UK, Fax: + 44 2476523701, Tel.: + 44 2476523557, E-mail: Crobinson@bio.warwick.ac.uk
A series of thylakoid membrane proteins, including PsbX, PsbY and PsbW, are synthesized with cleavable signal peptides yet inserted using none of the known Sec/SRP/Tat/Oxa1-type insertion machineries. Here, we show that, although superficially similar to Sec-type signal peptides, these thylakoidal signal peptides contain very different determinants. First, we show that basic residues in the N-terminal domain are not important, ruling out electrostatic interactions as an essential element of the insertion mechanism, and implying a fundamentally different targeting mechanism when compared with the structurally similar M13 procoat. Second, we show that acidic residues in the C-domain are essential for the efficient maturation of the PsbX and PsbY-A1 peptides, and that even a single substitution of the −5 Glu by Val in the PsbX signal peptide abolishes maturation in the thylakoid. Processing efficiency is restored to an extent, but not completely, by the highly hydrophilic Asn, implying that this domain is required to be hydrophilic, but preferably negatively charged, in order to present the cleavage site in an optimal manner. We show that substitution of the PsbX C-domain Glu residues by Val leads to a burial of the cleavage site within the bilayer although insertion is unaffected. Finally, we show that substitution of the Glu residues in the lumenal A2 loop of the PsbY polyprotein leads to a block in cleavage on the stromal side of the membrane, and present evidence that the PsbY-A2 signal peptide is required to be relatively hydrophilic and unable to adopt a transmembrane conformation on its own. These data indicate that, rather than being merely additional hydrophobic regions to promote insertion, the signal peptides of these thylakoid proteins are complex domains with uniquely stringent requirements in the C-domain and/or translocated loop regions.
Most chloroplast thylakoid proteins are nuclear-encoded in plants and are therefore inserted into the membrane post-translationally after import from the cytosol (reviewed in [1,2]). Many of these proteins are synthesized with an N-terminal ‘transit’ peptide that mediates interaction with the envelope-localized import machinery and transport into the stroma, after which this presequence is removed by the stromal processing peptidase. In these cases, insertion into the thylakoid membrane involves targeting determinants located in the mature protein, and this has been experimentally confirmed for one thylakoid protein, the major light-harvesting chlorophyll-binding protein, Lhcb1 [3,4]. Further studies on this protein have shown it to integrate using a complex pathway involving stromal signal recognition particle (SRP), FtsY, GTP and membrane-bound translocation machinery that includes the protein Albino3 [5–9]. The Alb3 protein is related to the Oxa1p and YidC proteins that play important roles in the insertion of proteins into the bacterial plasma membrane and mitochondrial inner membrane, respectively [10,11]. In general, this insertion pathway resembles that used by at least some plasma membrane proteins in bacteria, which also involves SRP and FtsY (reviewed in ). This is perhaps unsurprising as chloroplasts are widely accepted to have evolved from endosymbiotic cyanobacteria.
Other thylakoid membrane proteins are inserted by a different pathway that contrasts markedly with the highly complex SRP-dependent pathway. The proteins PsbX, PsbW, CFoII and PsbY are synthesized with N-terminal bipartite presequences in which the first domain specifies import from the cytoplasm across the chloroplast envelope, after which it is removed by a stromal processing peptidase. The second domain resembles typical ‘signal’ peptides, containing three distinct domains: an N-terminal charged region (N-domain), hydrophobic core region (H-domain) and more polar carboxyterminal region (C-domain) ending with an Ala-Xaa-Ala consensus region. Signal peptides usually specify translocation by Sec-type translocation systems in the endoplasmic reticulum, bacterial plasma membrane or thylakoid membrane, and have been studied in detail in these systems (reviewed in [13,14]). However, these thylakoid proteins have been shown to insert into the thylakoid membrane in the absence of SRP, SecA, nucleoside triphosphates or (δpH, thereby excluding all the ‘assisted’ modes of insertion into thylakoids reported to date [15–17]. Furthermore, proteolysis of thylakoids destroys the membrane-bound Sec and twin-arginine translocase machineries but has no effect on the insertion of these proteins . In the absence of identifiable translocation factors it has been suggested that these proteins insert spontaneously into the thylakoid membrane. A similar mechanism was originally proposed for M13 procoat, which is also synthesized with a cleavable signal peptide and which inserts into the Escherichia coli plasma membrane by an SRP/Sec-independent pathway. However, this protein is now known to be highly dependent on YidC for insertion , whereas recent studies have shown that PsbX, PsbW and PsbY do not rely at all on the thylakoidal YidC homolog, Alb3 . Because none of the known thylakoidal protein transport machinery is required for the insertion of these proteins, it has been suggested that they may insert spontaneously into the thylakoid membrane.
The initial stages of this spontaneous insertion mechanism involve binding of the intermediate-size protein to the membrane, after which both hydrophobic domains (one in the signal peptide, the other in the mature protein) insert into the membrane and form a transmembrane loop intermediate . As a result, the hydrophilic, negatively charged domain is translocated across the membrane to the lumen where further processing by a thylakoid processing peptidase (TPP) removes the remaining presequence leaving the mature protein inserted in the membrane with a lumenal N-terminus and a stromal C-terminus. The presence of the hydrophobic signal peptide has been shown to be essential in the case of CFoII  but the important features within this class of signal peptide have not been studied in any systematic manner and, because Sec- and Tat-type signal peptides interact with proteinaceous binding sites, it is possible that these thylakoid signal peptides may possess unique characteristics that are essential for their correct functioning. In this study we have analyzed the importance of charged residues in the insertion and proteolytic processing of PsbX, PsbW and PsbY. We show that basic residues in the N-domain are not important for either process whereas acidic residues in the C-domains of several of the signal peptides play important roles in the processing of precursor forms to the mature size. These requirements are completely unlike those of M13 procoat, which also bears a signal peptide, or the Sec-type signal peptides of translocated proteins.
Materials and methods
Construction and expression of truncated pre-PsbW proteins
A cDNA clone encoding the precursor form of Arabidopsis PsbW, pPsbW  was amplified using inverse PCR to generate intermediate-size and ‘short’ (see Results section) versions truncated at the N-terminus (iPsbW and sPsbW). For iPsbW, the forward and reverse primers were ATGGG TAAGAAGAAGGGAGGA and TCTCTTTGCTCGGA CGCG, respectively. For sPsbW, the forward and reverse primers were ATGGAGACAAAGCAAGGAAAC and TCTCTATTTGCTCGGACGCG. All constructs were synthesized in vitro by transcription of cDNAs followed by translation in a wheat germ lysate (Promega Biotech) in the presence of [35S]methionine.
Mutagenesis of pPsbX and pPsbY
cDNA clones encoding Arabidopsis pPsbX  and pPsbY  were subjected to site-specific mutagenesis using the Stratagene Quikchange™ kit according to the manufacturer's instructions. All mutants were fully sequenced to verify the mutagenesis results, and the precursor proteins were synthesized as described above, except that the PsbX data were obtained using [3H]leucine, as were some of the pPsbY data (see text).
Chloroplast import reactions were carried out using intact pea chloroplasts from 8- to 9-day-old-seedlings as described previously [22,23]. Urea washes were carried out as according to  using a method modified from that detailed in . For time course analysis, precursor proteins were imported into chloroplasts for 10 min, after which the organelles were centrifuged (1 min in a microcentrifuge), and the pellet resuspended in 1 mL of import buffer (50 mm Hepes/KOH, 330 mm sorbitol) and further incubated. Sonication studies were carried out using a Branson 1210 water bath sonicator at 0 °C. Thylakoid import reactions were carried out as in , after which samples were analyzed immediately or after washing twice with 1 mL 20 mm Hepes/KOH, 5 mm MgCl2.
Electrostatic interactions are not essential in the early stages of the PsbW insertion process
Studies on M13 procoat (reviewed in ) have demonstrated the importance of electrostatic interactions between basic residues in the protein and the negatively charged membrane surface. These data indicated that basic residues were essential in both the extreme N-terminal region of the signal peptide and the C-terminal region of the mature protein. Removal of either set of charges led to a block in insertion, strongly indicating that the electrostatic interactions were required during the early stages of the insertion process, probably to bind the protein stably to the membrane surface. pPsbW resembles procoat in several respects, as detailed above, and similarly forms a loop intermediate during insertion  but the early stages of the insertion process are poorly understood and it is unclear how this protein binds to the thylakoid membrane prior to insertion. The thylakoid membrane is also negatively charged due to the presence of sulfolipids and an electrostatic interaction seemed possible. However, although the N-terminal region of the pPsbW signal peptide is positively charged, the C-terminal region of mature PsbW is devoid of basic residues, ruling out an electrostatic interaction with the membrane surface. This region is in fact highly negatively charged due to the presence of a series of five acidic residues (see Fig. 1A). This protein is thus an attractive model system in which to address this issue because we needed only to test the importance of the basic residues in the N-terminal region. This was achieved by simply truncating the protein to remove increasing numbers of basic residues.
The overall structure of pPsbW is shown in Fig. 1. The first ‘envelope transit’ domain has been omitted from the pPsbW sequence which starts at residue 31. The precise site of cleavage by the stromal processing peptidase has yet to be identified but the likelihood is that it lies just before or after (or within) the KKK sequence in the N-terminal region of the signal peptide. Irrespective of the precise site of cleavage, the N-terminal domain is positively charged and we reduced the overall charge in this region by synthesizing an intermediate-size protein (iPsbW) that lacks the envelope transit domain and a smaller protein (sPsbW) that lacks any basic residues in the signal peptide.
The effects of the truncations were tested by assaying for the insertion of in vitro synthesized proteins into isolated thylakoids. Because the TPP active site is on the lumenal face , maturation is clear evidence of insertion and Fig. 1B shows that all of the proteins insert into pea thylakoids and become processed to the mature size. The truncated proteins insert with slightly lower efficiencies (sPsbW insertion efficiency is down to 45% of that of wild-type protein) but the truncations, clearly, by no means block insertion. It should be noted that even the sPsbW form still carries a single positive charge at its N-terminus, due to the protonated amino group. Nevertheless, we conclude that electrostatic interactions are not as important for PsbW insertion as for procoat insertion.
The translocated loop regions of spontaneously-inserting proteins contain negative charges in either the mature protein or the signal peptide
Four thylakoid membrane proteins have been shown to be synthesized with cleavable signal peptides but inserted by spontaneous mechanisms, and comparison of the translocated loop regions shows that they are all negatively charged (Fig. 2A). In the cases of PsbW and CFoII, the charges lie in the N-terminal region of the mature protein, but PsbX differs in that two Glu residues are located in the C-terminal region of the signal peptide. The polyprotein, PsbY, also contains acidic residues in this region of each signal peptide. Other types of signal peptide (e.g. those specifying Sec- or Tat-dependent translocation) rarely contain negative charges in the C-terminal region and we considered it possible that this feature may have evolved in the PsbX/PsbY signal peptides in order to enhance the overall insertion process. Constraints on the functions of the mature proteins may have precluded the presence of acidic residues in the N-terminal regions of the mature proteins. Accordingly, we sought to test whether this characteristic is important in the spontaneous insertion process by making site-specific mutations in the signal peptides, focusing primarily on PsbX as a simple model system but then extending the studies to encompass PsbY. The mutations are shown in Fig. 2B. In brief, the Glu residues were substituted by hydrophobic residues such as Val, or by highly hydrophilic but neutral residues such as Asn (attempts to replace one of the Glu residues by Gln were unsuccessful, for unknown reasons). The importance of net charge in the loop region was also tested.
Acidic residues in the signal peptide are important for efficient maturation of pPsbX.
As an initial test for the importance of the two Glu residues (positions −5 and −2, relative to the TPP cleavage site) we made a mutant in which both were substituted by Val. The import and sorting characteristics of this mutant, PsbX/VV, and wild-type PsbX were analyzed by incubating the precursor proteins (PsbX/VV and pPsbX, respectively) with intact chloroplasts and subsequently determining the intraorganellar locations and cleavage products (shown in Fig. 3). Wild-type pPsbX is efficiently imported, targeted to the thylakoids (lane T) and processed primarily to the mature size, as found previously . PsbX/VV is imported with similar efficiency and the protein is likewise targeted to the thylakoids, but only the intermediate form (iPsbX/VV) is found within the organelles. This intermediate is of precisely the same size as a mutant analyzed previously  in which the terminal processing site was altered to prevent cleavage by TPP. Clearly, the thylakoid-associated PsbX/VV corresponds to the product generated by the stromal processing peptidase. These data demonstrate that the intermediate form is either unable to insert into the membrane, or that it does insert but can not be processed by TPP.
We then carried out single-residue substitutions to determine whether either of the two Glu residues is more important in this context. Accordingly, we analyzed mutants in which only one of the Glu residues was substituted by Val. The results (Fig. 3, lower panels) show that substitution of the −2 Glu by Val (PsbX/EV) again affects maturation but to a lower extent. In the total chloroplast fraction (lane C) the intermediate- and mature-size bands are of approximately equal intensity, whereas the mature-size PsbX protein predominates in the thylakoid fraction (lane T). These findings suggest that the protein is gradually converted to the mature size during the import/fractionation procedure (the chloroplast fraction is removed and processed for electrophoresis well before the other samples, which require protease treatment and, in the case of the stroma/thylakoid samples, fractionation after lysis). This was confirmed by time-course analyses, which show gradual conversion to the mature size (data not shown; similar examples are shown below). The presence of Val at the −2 position thus slows down maturation to a considerable extent, but does not block it. In contrast, the final panel in Fig. 3 shows that substitution of the −5 Glu by Val (PsbX/VE) completely blocks maturation as found with the double Val mutant.
The imported PsbX mutants described in Fig. 3 are found exclusively in the thylakoid fraction which suggests that insertion has taken place. However, to confirm this point we carried out urea washes of the thylakoids because this procedure is highly effective at removing extrinsic membrane-associated proteins [19,24]. Figure 4 shows that this procedure is sufficiently harsh to remove even some of the fully inserted mature size wild-type PsbX, because some is found in the supernatant fraction (lane Sn) after the extraction procedure. This is apparently because single-span proteins are more easily removed from the thylakoids by urea . However, most of the mature-size PsbX is found in the membrane pellet fraction (lane pel) and the same applies to the intermediate size iPsbX/VV, which is equally resistant to extraction. As with the double Val mutant, urea washes confirmed that the imported mature-size single-Val mutants are fully integrated into the thylakoid membrane (data not shown). Accordingly, we propose that the protein cannot be cleaved by TPP, and this could be due to one of two reasons: first, the processing site may have been altered such that TPP can no longer recognize the cleavage site, or secondly, the processing site may be intact but TPP may be unable to reach it.
Hydrophilic Asn residues can partially compensate for loss of negative charge in the C-domain of PsbX
The above data indicate that substitution of the −5 Glu has far more dramatic consequences than alteration of the −2 residue, which suggests that the −5 Glu is significant either because a negative charge is important in this region, and/or because the presence of a very hydrophilic residue is important for maturation. The −5 Glu effectively caps the H-domain and the VE mutant thus contains a longer hydrophobic region that now extends to the −2 Glu (see Fig. 2). We investigated these possibilities by substituting the −5 and −2 glutamates with asparagine, which is highly hydrophilic but uncharged. According to the Kyte–Doolittle and GES hydrophobicity scales, Asn is almost as hydrophilic as Glu [26,27]. The −5 and −2 substitutions (see Fig. 2) are termed PsbX/NE and PsbX/EN, respectively, according to whether the first or second Glu is substituted by Asn, and the double mutant is PsbX/NN. Import assays using the EN and NE single mutants are shown in the upper panel of Fig. 5. As with the other mutants analyzed in this study, these proteins are efficiently imported and targeted to the thylakoid membrane. No stromal intermediates are present and the thylakoid-associated proteins are as resistant to urea-extraction as authentic PsbX (not shown). These data indicate that the mutations have no detectable effect on insertion efficiency. Both mutants are also processed to the mature size but it is notable that maturation is not as efficient as for the wild-type protein. Whereas PsbX is invariably found almost exclusively as the mature form after import into chloroplasts, the intermediate-size forms of both single-Asn mutants are apparent in the thylakoid fractions indicating an inhibitory effect on maturation by TPP.
This effect is exacerbated in the PsbX/NN mutant that contains Asn at both the −5 and −2 positions. In this case, a much greater proportion of the imported protein is present as the intermediate form (iPsbX/NN) at the end of the import/fractionation procedure. In this experiment we also carried out a time-course analysis in which the PsbX/NN mutant and wild-type pPsbX were imported for 10 min, after which the chloroplasts were washed to remove nonimported protein and samples were analyzed at various times thereafter to follow the maturation of the imported protein (lower panels of Fig. 5). Repeat tests using wild-type pPsbX showed that protein is found only as the mature protein at even early time-points (see bottom panel). In contrast, PsbX/NN is found primarily as the intermediate form at early time points and this form is only gradually converted to the mature size during the subsequent 60 min All of the imported protein was found to be inserted in the thylakoid fraction at each time-point (not shown), demonstrating that the double Asn mutations slow down processing by TPP but do not block this process. We conclude from these experiments that hydrophilic Asn residues at the −5 and −2 positions enable processing by TPP to occur, but with less efficiency than when Glu residues are present.
Valine substitutions at the −5 and −2 positions may lead to burial of the cleavage site within the membrane
Several of the PsbX mutants shown above exhibit slow maturation kinetics within the chloroplast, despite being inserted into the thylakoid membrane. We propose that this stems from an inability of TPP to actually access the cleavage site, rather than an alteration of the site such that TPP can reach the site but not carry out cleavage. Two points should be emphasized. Firstly, the identity of the −5 and −2 residues varies enormously among thylakoidal signal peptides, and many different residues are found at these two positions. Asparagine, in particular, is common in the C-domain and is often found at the −5 and −2 positions. Almost any residue appears to be tolerated at the −2 position and it is unlikely in the extreme that valine should pose a problem. In general, the important determinants for TPP cleavage appear to be short-chain residues at the −3 and −1 residues , and a helix-breaking residue is also commonly found in the region of −4 to −6. Other signal peptidases exhibit broadly similar preferences .
In a second line of investigation, we analyzed the positioning of the translocated loop regions of several PsbX derivatives, by comparing their sensitivities to digestion by elastase (Fig. 6). The experiment involved importing wild-type PsbX (which is cleaved exclusively to the mature size) and three mutant forms. The first mutant (PsbX/A74T) contains threonine at the −1 position in place of alanine and previous studies on this mutant  showed that this mutation has no effect on insertion efficiency but cleavage by TPP is blocked, leading to the formation of a loop intermediate with the TPP site exposed on the lumenal side of the membrane. The other two mutants analyzed were PsbX/NN, which has a reduced rate of maturation and PsbX/VV, which is completely blocked in maturation. The aim here was to determine whether this block was due to alteration of the TPP site, such that the peptidase can no longer cleave, or inaccessibility of the site. Control tests (Fig. 6A) confirmed that all of these proteins are sensitive to elastase when not inserted into membranes; elastase cleaves pPsbX to yield a primary degradation product (denoted by asterisk) that is in fact slightly larger than mature PsbX. Importantly, all of the PsbX forms are cleaved to the same products and the mutants are not cleaved with lower efficiency. The PsbX/VV mutant, which is of particular interest in this experiment, is indeed cleaved with higher efficiency than wild-type pPsbX.
After import of the proteins into chloroplasts, samples of the thylakoids were analyzed without further treatment (lanes T), after incubation with elastase (lanes T-el) or after incubation with elastase and concommitant sonication in a water bath, such that the protease can enter the lumenal space (lanes T-son). The data using wild-type PsbX show that the mature-size protein is not cleaved in any of the samples, as expected because this small protein is essentially buried within the bilayer with only short domains exposed on either face. With PsbX/A74T, the thylakoid sample contains primarily intermediate-size protein, and elastase treatment without sonication has little effect. Again, this is unsurprising because the regions exposed to the stromal face (C-terminus of the mature protein and N-terminus of the signal peptide) are short. However, when sonicated the protease efficiently cleaves the intermediate to a product of similar mobility to the mature protein, indicating that the loop region is exposed to the lumen. The PsbX/NN mutant behaves similarly; most of the protein is of mature-size by the end of the experiment but the intermediate is again resistant to proteolysis in the absence of sonication but sensitive when sonicated. In contrast, the PsbX/VV mutant is almost totally resistant to digestion under all conditions and only a very minor proportion is cleaved when the sample is sonicated. The loop region is thus inaccessible to elastase on either side of the membrane and must therefore be buried in the bilayer to a much greater extent than is the case with the wild-type protein.
Acidic residues are important for efficient processing of the PsbY polyprotein
PsbY is an unusual protein that is synthesized with two cleavable signal peptides . After insertion into the thylakoid membrane, TPP cleaves twice on the lumenal face to release the two signal peptides and an unidentified protease cleaves on the stromal face of the membrane to complete the process and generate the two single-span mature proteins, denoted A1 and A2 . Mutagenesis studies  have clearly demonstrated that the cleavage on the stromal face occurs at a late stage in the overall insertion process. As shown in Fig. 7, both of the signal peptides (SP1 and SP2) contain acidic residues in the C-domain, and the A2 loop also contains Glu at the +3 residue, relative to the cleavage site. We tested the importance of these residues by substituting the A1 Glu with Val (PsbY-A1/V) and both A2 Glu residues with Val (PsbY-A2/VV); the precise structures of these mutants are shown in Fig. 7B.
The import data using the PsbY-A1/V mutant are shown in Fig. 8A. In the control import using wild-type pPsbY (left hand panel), the precursor protein is imported and converted to a close doublet of mature A1 and A2 proteins, as found previously [23,31]. The PsbY-A1/V mutant is also targeted to the thylakoid membrane and the appearance of the A2 protein is unaffected. However, the A1 protein is apparent in the thylakoid sample analyzed at the end of the import/fractionation procedure (lane T) but is present in low amounts in the chloroplast samples (lanes C and C+). Instead, two larger intermediates are present (denoted ‘ints’). These polypeptides were found to accumulate when the processing of A1 was blocked in an earlier study  in which alteration of the −1 residue was shown to block cleavage by TPP. This suggests that the presence of the valine in PsbY-A1/V has likewise slowed down processing by TPP. We suspected that the near-absence of the intermediate bands in the thylakoid sample (lane T) may result from slow but continuing cleavage by TPP during the course of the experiment (as described earlier). This is confirmed in Fig. 8B, which shows time-course analyses similar to that described above for the PsbX/NN mutant. After a 15-min import incubation the chloroplasts were washed to remove nonimported protein and the organelles were further incubated for the times (in min) shown above the lanes. The import reaction using wild-type pPsbY shows that essentially all of the imported protein is present as mature A1 and A2 at the earliest time-points. However, after import of pPsbY-A1/V the A1 protein is virtually absent at the initial time-point and the two intermediate bands are instead present. These decline over the subsequent 10–20 min and the A1 protein appears. These data show that the presence of the valine at the −5 position, relative to the TPP cleavage site, leads to a substantial inhibition of processing at the A1 site.
The data obtained using the double Val substitution in the A2 cleavage region are shown in Fig. 9. Here, the substitutions lead to very different effects. The import of PsbY-A2/VV is shown in Fig. 9A, together with a thylakoid sample from a control import (lane Con) using wild-type pPsbY. This mutant is imported and targeted to the thylakoid membrane, but surprisingly the appearance of the A2 protein is unaffected and it is the A1 protein which is absent. A larger intermediate is instead apparent, which was assumed to contain the A1 protein (labeled A1int). However, it was deemed important to verify this point, firstly because this result was completely unexpected, but also because we considered it possible that both the A1 and A2 bands may have shifted in the gel due to complex effects arising from the mutations. We therefore used alternative means to identify the A1- and A2-containing proteins unambiguously. Analysis of the protein sequence (see Fig. 7) reveals that the A1 and A2 mature proteins each contain a single methionine towards the C-terminus of the peptide (shown underlined and italicized). The methionine at the end of A2 was substituted with leucine in both wild-type pPsbY (‘PsbY-A2-met’) and the PsbY-A2/VV mutant (mutant PsbY-A2/VV-met). The proteins were synthesized in the presence of either [3H]leucine or [35S]methionine and the import of these proteins is shown in Fig. 9B. Translation products (Tr) and the thylakoid samples from the import reactions (lanes imp) are shown above the lanes, together with the radiolabeled amino acid used in the translation (leu or met). In the control import with PsbY-A2-met (panel A2-met) the A1 and A2 proteins are both apparent as expected when the protein is synthesized in the presence of [3H]leucine, but the A2 protein is now absent when a [35S]methionine-labeled translation product is used (lane imp met). This confirms that the A2 protein does indeed contain a single methionine residue as predicted from the sequence. In the case of the PsbY-A2/VV-met mutant, the two polypeptides observed in Fig. 9 A are again observed in the [3H]leucine-labeled sample and the identity of the lower band as A2 protein is again confirmed by the finding that the band is absent in the [35S]methionine-labeled sample, while the A1int band is still present. This result confirms that the double valine substitution in the A2 cleavage site region does not actually affect cleavage of A2, but instead leads to a complete block in the processing of A1 to the mature size. The A1int polypeptide is too small to contain three transmembrane spans (the three-span intermediates are characterized in ) and, because cleavage on the stromal surface is known to occur last and the A1 TPP cleavage site is completely unaffected, this polypeptide almost certainly comprises A1 plus the A2 signal peptide.
Previous studies have shown that a series of thylakoid membrane proteins are synthesized with cleavable signal peptides, yet are inserted by mechanisms that do not rely on any of the known translocation machinery, either in the soluble phase or at the membrane surface. It has been suggested that these signal peptides provide an additional hydrophobic region that helps to drive the insertion process, perhaps through the formation of a ‘helical hairpin’ that might provide the required driving force to flip the N-terminus of the mature protein across the thylakoid membrane. Intriguingly, these signal peptides resemble those of Sec-dependent lumenal proteins to a marked degree, and one of this class of signals can even function as a Sec-type signal for a lumenal passenger protein in chloroplasts . However, the data shown here point to defining features in some of these peptides that are essential for their correct functioning and which are not apparent in other forms of signal peptide. Our data also lead to new ideas on the biogenesis of the unique PsbY polyprotein.
The studies on the PsbW truncations focused on the role of basic residues in the N-domain, because previous work on M13 procoat and Sec-type signal peptides has shown that basic residues in the N-domain play essential roles in insertion/translocation [12,13,32]. In fact, our data indicate quite clearly that these play no important function in the insertion of PsbW, because their removal inhibits insertion to only a minor extent. When considered in conjunction with other data on this group of thylakoid proteins, it is now very interesting to compare and contrast their insertion mechanism with that of procoat. Initial models for the insertion of these thylakoid proteins were based heavily on that of procoat insertion. M13 procoat and pPsbW are very similar indeed in structural terms, in that they possess a single transmembrane span in the mature protein, are synthesized with rather similar cleavable signal peptides and the intervening loop regions (which are flipped across the membrane) are of similar lengths and overall charge. Both proteins form loop intermediates prior to cleavage by signal peptidase but it is now clear that their insertion requirements are completely different in almost every sense. Previous work has shown procoat to rely heavily on the protonmotive force (reviewed in ) whereas pPsbW is ΔµH+- independent, as are the other thylakoid proteins in this group [15,16]. Procoat is also totally dependent on YidC for efficient insertion  whereas the thylakoid proteins do not require the homologous Alb3 protein . We have now shown that these proteins differ in the means by which they initiate insertion; electrostatic forces play a central role in the early stages of the procoat insertion mechanism  whereas pPsbW contains no basic residues in the C-terminal region and our data show that basic residues in the N-terminal region are not important for insertion into thylakoids. pPsbW must therefore interact with the thylakoid membrane by other means. Basic residues in the N-domain are also highly important for the functioning of Sec-type signal peptides, possibly to promote interaction with anionic phospholipids or SecA [13,32,33], and it therefore appears that the signal peptides of these Sec-independent thylakoid proteins function in fundamentally different ways, despite the superficial similarities.
The other studies on PsbX and PsbY focused on the C-domain, prompted by the presence of acidic residues in this region. Acidic residues are not important in any of the domains within Sec-type signal peptides and are generally uncommon, especially in the C-domain which is generally five or six residues in length and polar but uncharged . In contrast, our results point to an important function for acidic residues in the translocated regions of these Sec/SRP/Alb3-independent thylakoid membrane proteins. In some cases (e.g. CFoII) the extreme N-terminus of the mature protein is highly negatively charged, and we believe that additional acidic residues in the signal peptide are probably unnecessary. In other cases (for example PsbX and PsbY-A1), acidic residues are not present in the N-terminus of the mature protein and in these cases the signal peptides contain conserved acidic residues in the C-domain. Our data indicate that these residues are very important for the correct maturation of the inserted protein. Substitution of the −5 and −2 Glu residues by Val leads to a complete block in the maturation of PsbX, although insertion appears not to be affected. The −5 Glu, in particular, appears to be important because the presence of Val at this position alone is equally detrimental. Processing efficiency is restored to some extent by Asn at the −5 position. For reasons that are presently unclear, the translocated loop region is cleaved most efficiently when carrying negative charges although other hydrophilic residues can substitute to some extent.
On the basis of these data we propose that the translocated region needs to bear a negative charge in order for the TPP cleavage site to be presented in an optimal manner. A model for the effects of these mutations on the PsbX insertion mechanism is as follows. Insertion of the wild-type protein leads to the formation of a loop intermediate  and the hydrophilic nature of the loop region is essential for correct presentation to TPP. We believe that the presence of negative charges close to the TPP site serves to distance the site from the membrane interior and enable processing to occur. The presence of Val at the −5 site leads to a lengthening of the hydrophobic region which then becomes buried in the membrane interior. This premise is supported by studies on the PsbY-A1V mutant, which contains no negative charges in the TPP cleavage site region. Processing of this mutant is again significantly impaired although not to the same extent as in some of the PsbX mutants.
These studies are reminiscent of some observations made with Sec-type signal peptides [34–36], where alteration of the C-domain or H/C boundary can also affect processing by signal peptidase. However, in the vast majority of these cases, processing was not blocked but rather occurred elsewhere, or the mutations made were far more drastic than those generated in PsbX. It should be emphasized that a near-complete block in processing occurred after only a single substitution (−5 Glu to Val) and processing is drastically affected in the PsbX/NN mutant despite the presence of a highly polar C-domain of the correct length. Overall, these mutations have far more drastic consequences than similar mutations made in Sec-type signal peptides, and we conclude that this may be due to one or both of the following reasons: (a) our studies are on membrane proteins rather than hydrophilic translocated proteins, and the cleavage site region may therefore be more highly constrained in the membrane because the mature protein is not pulled across the bilayer; and/or (b) the unusual lipid composition of the thylakoid membrane (primarily galactolipid rather than phospholipid ) may require that the translocated loop is more effectively presented to the signal peptidase when acidic residues are present, for unknown reasons.
The third aspect of this study concerned the PsbY-A2 signal peptide, but very different results were obtained in this case. Here, the substitution of two Glu residues in the translocated loop by Val does not block cleavage by TPP, indicating that the Glu in the C-domain of this signal peptide is not as important. Possibly, the presence of three helix-breaking proline residues upstream (see Fig. 7) is sufficient to maintain the TPP site away from the membrane, or other effects may operate in this case. However, these mutations do have dramatic effects and in this case it is cleavage at the A1-SP2 site on the stromal surface that is completely inhibited. In fact, the studies on this mutant fortuitously provide important information on the biogenesis of the PsbY polyprotein. In previous work on PsbY , we noted that blockage of the TPP cleavage reaction at either the A1 or A2 sites led to the accumulation of a three-membrane-span intermediate, indicating that cleavage on the stromal side of the membrane had failed to occur in each case. Inhibition of TPP cleavage at both sites led to the stable accumulation of a four-span intermediate. Clearly, cleavage at the stromal site occurs at a late stage prompting the question: why is this protease unable to recognize this site until both cleavages by TPP have occurred? The present study provides further information on this issue; lengthening the H-domain of the A2 signal peptide leads to the stable accumulation of a two-span intermediate containing A1 and the A2 signal peptide (SP2). Our interpretation is that the unidentified protease on the stromal surface can only cleave when SP2 is released from a transmembrane state to adopt a flexible orientation in the membrane. Our model for the overall process is as follows (see Fig. 10).
Stage 1. The PsbY polyprotein inserts into the membrane in the double loop formation shown in Fig. 10, and TPP cleaves at one of the two sites. Most probably, TPP can cleave at either site first but for simplicity it is shown as cleaving at the SP1-A1 site. This releases SP1 which is rapidly degraded.
Stage 2. TPP cleaves at the SP2-A2 site, releasing A2 as a single-span mature protein and generating the A1-SP2 intermediate (Stage 2).
Stage 3. SP2 is now far more flexible, either because it is no longer tethered at the lumenal face by charged residues or because it is not bound to its partner polypeptide region, A2. The stromal loop region is more accessible and cleavage in this loop can now occur.
One possibility is that this final cleavage can only occur when the A1 and SP2 regions are unconstrained by cognate partner polypeptide regions (A1-SP1, A2-SP2). First, the PsbY-A2V mutant can be cleaved at both positions by TPP but the A1-SP2 intermediate accumulates as a stable species (see lower panel of Fig. 10). In our view, this is most likely because the SP2 H-domain is now significantly longer and is effectively a true membrane-spanning region. The H-domains of the signal peptides of these thylakoid membrane proteins are much shorter and are generally less hydrophobic than true membrane-spanning regions and, we believe, can only adopt transmembrane conformations when tethered to genuine transmembrane spans.
Further evidence for this proposed model comes from considerations of the stabilities of SP1 and SP2. It is notable that the A1-SP2 intermediate is highly stable, as are the PsbX and PsbW loop intermediates generated in a previous study . Clearly, the signal peptides are completely resistant to proteolysis when bound to genuine transmembrane spans. In complete contrast, the signal peptides cannot be detected in even low amounts when released during normal insertion reactions, despite being as large as some of the mature proteins (e.g. PsbX and PsbW are only 4 and 6 kDa, respectively). Tricine gels readily resolve these small mature proteins but the complete absence of the cleaved signal peptides, even immediately after insertion  means that they are degraded very rapidly indeed. We propose that this is due solely to their low hydrophobicity, which precludes the maintenance of transmembrane configurations and instead leads to other positions in the membrane, or even release from the membrane , upon which they are degraded by proteases that perhaps specifically target peptides that are unable to adopt transmembrane conformations.
In summary, we have shown that the signal peptides of these spontaneously-inserting proteins have evolved with specific and unusual properties that are especially important for correct proteolytic cleavage following insertion. In the cases of PsbX and PsbY-A1, the hydrophobicity of the C-domain is critical for correct maturation and negative charges in particular appear to be favored. In the case of PsbY-A2, the negative charge in the translocated loop plays a key role in defining the hydrophobicity of the A2 signal peptide, which is of necessity low in order to facilitate the movements that allow the final cleavage on the stromal surface. In general, these signal peptides are not merely additional hydrophobic regions but are rather exquisitely structured extensions whose properties complement those of the N-terminal regions of the mature proteins.
This work was supported by Biotechnology and Biological Sciences Research Council grant C09633 to C. R.