Chaperonin-promoted post-translational membrane insertion of a multispanning membrane protein lactose permease.

Using an in vitro membrane-free translation system from Escherichia coli, it is shown that chaperonin GroEL added cotranslationally interacts with newly synthesized lactose permease (LacY), a polytopic membrane protein, thereby preventing aggregation. Subsequently, when the isolated GroEL-LacY complex is incubated with inverted membrane vesicles, the permease is inserted into the membrane in a MgATP-dependent manner. Post-translational membrane insertion is also observed when aggregation of newly synthesized LacY is prevented by addition of the nonionic detergent n-dodecyl-beta,D-maltoside during translation in place of GroEL. No membrane integration occurs with right-side-out vesicles, indicating that LacY interacts specifically only with the cytosolic face of the membrane. Ligand thiodigalactoside protection against alkylation of the Cys-148 residue in the permease shows proper post-translational insertion. Moreover, limited proteolysis of soluble LacY either complexed with GroEL or in detergent indicates that the newly synthesized protein assumes a conformation that is comparable to that of native, membrane-embedded permease prior to insertion into the membrane.

The Escherichia coli lactose permease (LacY) 1 catalyzes the coupled translocation of ␤-galactosides and H ϩ across the cytoplasmic membrane. Within the framework of current interest in the biosynthesis and assembly of membrane proteins, LacY provides a particularly attractive model, since this hydrophobic polytopic plasma membrane protein is well characterized (1,2). The functional unit of LacY is a monomer of 46.5-kDa containing 12 membrane-spanning hydrophobic ␣-helices (1) that traverse the membrane in a zig-zag fashion packed into a particle of about 50 ϫ 40 Å (3). Although LacY represents one of the most extensively characterized transport proteins, it is not known how this protein or other integral membrane proteins insert into membranes. As shown previously, LacY can be synthesized in vitro (4,5) and inserted cotranslationally into INV (5). Unlike secretory proteins, there is no evidence for post-translational insertion of LacY or other polytopic membrane proteins in E. coli, and, generally, little is known about biogenesis of polytopic membrane proteins. Membrane proteins are very hydrophobic and tend to aggregate in membrane-free translation systems (5). In this respect, molecular chaperones may act to prevent aggregation. It has been demonstrated by Bochkareva et al. (6) that chaperonin GroEL maintains the translocation-competent unfolded state of newly synthesized pre-␤-lactamase and that translocation across the membrane requires MgATP-promoted discharge of the protein from GroEL. The physiological role of molecular chaperones, including GroEL, is to interact with transiently exposed hydrophobic patches and to promote folding, assembly, or secretion of various proteins mainly by preventing aggregation (7)(8)(9). However, the possible role of GroEL or other molecular chaperones in the targeting and insertion of polytopic membrane proteins into the membrane is an open question.
S-30 crude extract from the E. coli MRE600 was prepared as described (14) and purified from membranes by centrifugation. For this, 0.3 ml of the extract were loaded over 4.7 ml of 22% sucrose in buffer A and centrifuged at 4°C for 110 min at 45,000 rpm (SW50.1 rotor, Beckman L5-65 centrifuge). The top 3.5 ml were collected and diluted 3 times with the same buffer. The samples were then concentrated to 10.5 mg of protein per ml using Centriprep 10 (Amicon) and stored at Ϫ80°C. INV were prepared from E. coli MC4100 cells by a low pressure French press (14) in buffer containing 1 mM EDTA, 1 mM DTT, 50 mM TEA-HCl, pH 7.5. The final concentration of sucrose gradient-purified INV was 6 mg of protein per ml as determined according to Ref. 15 using BSA as a standard. 35 S-Labeled cells (10 A 600 per ml in the same buffer) were sonicated on ice by Microson (Heat Systems Inc.) for 4 ϫ 15 s, and membranes were purified by two cycles of centrifugation at 4°C for 45 min at 75,000 rpm (TLA100.1 rotor, Beckman TL100 centrifuge) through the same buffer containing 20% sucrose. SDS-Tricine PAGE was carried out according to Ref. 16  Where indicated, 4.5-6.0 g of GroEL or 0.1% DM was included. After a 15-min incubation at 37°C, cold Met was added (final concentration 0.4 mM), and the translation was stopped by 1 mM puromycin and * 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.
Isolation of GroEL-LacY Complex and Transfer of LacY from the Complex into INV-GroEL-[ 35 S]LacY complex was isolated by centrifugation of the translation sample through 1.4 ml of a 5-20% sucrose gradient in buffer containing 50 mM KCl, 10 MgCl 2 , 0.1 mM EDTA, 2 mM DTT, 40 mM TEA acetate, pH 7.5 (140 min at 54,000 rpm, TLS 55 rotor, Beckman TL100 centrifuge). 100-l fractions were collected from the top, and fractions 7-9 containing GroEL were diluted 3 times with the same buffer and concentrated to a final volume of ϳ80 -100 l using Centricon 30 (Amicon). Transfer of LacY was carried out in 50 l of buffer (5 mM potassium succinate, 10 mM MgCl 2 , 35 mM potassium acetate, 2 mM DTT, 50 mM TEA acetate, pH 7.5) containing 10 l of GroEL-[ 35 S]LacY complex (ϳ1.5 ϫ 10 6 cpm/1.7 g of GroEL), 48 g of INV, 3 g of creatine kinase, 0.5 mol of creatine phosphate and, also, where indicated, 1 g of GroES and/or 3.5 mM ATP. After a 30-min incubation at 37°C, EDTA (5 l of 0.2 M) was added, and the mixture was incubated for 10 min at 25°C and cooled on ice. The following isolation of INV carrying [ 35 S]LacY is described in the legend to Fig. 2.  Table I). The reaction was quenched by addition of 2 mM DTT. The samples were diluted 4 times with buffer C, and TDG was removed by centrifugation (40 min at 75,000 rpm in a TLA100.1 rotor, Beckman TL100 centrifuge). The vesicles were washed 3 times with the same buffer (without MgCl 2 ). The washed INV were dissolved in 40 l of buffer C containing 0.05% DM, incubated with 0.4 mM NEM-biotin for 10 min at 25°C, and the reaction was quenched by the addition of 10 mM DTT. Membranes were then incubated in alkaline pH as described above, and the excess of NEM-biotin was removed by centrifugation and 3 cycles of washing with buffer B. Pellets were dissolved in 100 l of DM buffer (1% DM, 150 mM NaCl, 0.2 mM EDTA, 50 mM TEA acetate, pH 7.5), and the DM extracts (3-5 ϫ 10 5 cpm) were incubated with 80 l of 50% suspension of avidin-agarose in the same buffer for 1 h at room temperature with continuous rotation. After extensive washing with DM buffer, the agarose beads were assayed by liquid scintillation spectrometry.
GroEL-promoted Interaction of LacY with Right-side-out Membrane Vesicles Prior to and after Sonication of Them-Right-side-out membrane vesicles were prepared as described by Kaback (18) and converted into INV by sonication as described above for the whole cells in a buffer containing 0.1 mM DTT, 50 mM TEA acetate, pH 7.5. Sonicated vesicles were utilized without additional purification. Transfer of LacY from the GroEL-[ 35 S]LacY complex into the right-side-out vesicles prior to or after sonication was carried out under identical conditions as described above. Then, the vesicles were isolated by centrifugation at 4°C for 45 min either at 14,000 rpm in an Eppendorf centrifuge 5415 C (for the right-side-out vesicles) or at 75,000 rpm through 20% sucrose layer in buffer B in a TLA100.1 rotor, Beckman TL100 centrifuge (for the sonicated vesicles). The membrane pellets were suspended in buffer A and treated in alkaline pH as described above. were collected from the top, and the membrane pellet was dissolved in 30 -60 l of buffer A by an occasional vortexing for 2 h at 4°C. Aggregates were pelleted by a 3-min centrifugation at 10,000 rpm, and the solubilized membranes were analyzed by SDS-Tricine-PAGE.

DM-promoted Insertion of LacY into INV -The
Proteolysis of LacY -40-l samples contained 0.5-2.0 ϫ 10 5 cpm of various LacY in a pH 7.5 buffer (8.5 mM MgCl 2 , 35 mM KCl, 1 mM DTT, 3.75 mM CaCl 2 (for the samples treated with thermolysin), 50 mM TEA acetate). BSA was also added in order to achieve approximately 20 g of total protein in the samples. 8-l aliquots from each sample were then mixed with 2 l of thermolysin or trypsin (final concentrations of the proteases are given in Fig. 4) and incubated at 25°C for 20 min or 10 min, respectively. Proteolysis was stopped by a 10-min incubation at 25°C with 3 l (1.5 mg/ml) of trypsin inhibitor from beef pancreas (type I-P, Sigma) in the case of trypsin, or with 3 l of stop-solution containing 60 mM EDTA, 75 mM EGTA, and a mixture of protease inhibitors (19) in the case of thermolysin. After addition of 4.5 l of 5% SDSsample buffer, the samples were incubated for 30 min at 37°C and analyzed by SDS-Tricine-PAGE. Note: according to a Coomassie staining of the gels, GroEL was practically stable to the action of thermolysin or trypsin under described conditions (data not shown).

RESULTS
In this study, we examine the effect of GroEL on the in vitro, membrane-free translation of LacY. As shown in Fig. 1, the membrane-free biosynthesis of [ 35 S]Met-labeled LacY leads to almost complete aggregation. In contrast, if exogenous GroEL is added to the translation mixture, about 30% of the LacY synthesized remains soluble, co-sedimenting with GroEL. The finding indicates that GroEL interacts with newly synthesized LacY, thereby preventing aggregation. In order to determine whether or not the LacY bound to GroEL is able to interact with membranes, sucrose gradient fractions containing the GroEL-LacY complex were incubated with E. coli INV at 37°C. The membranes were then separated from the GroEL-LacY complex ( Fig. 2A, fractions 8 and 2, respectively) by centrifu- gation through a 7.5-20% sucrose gradient layered on 40% sucrose. As shown in Fig. 2B, a significant portion of LacY becomes associated with the INV only in the presence of Mg 2ϩ and ATP. The co-chaperonin GroES (a GroEL helper (7-9)) further increases the transfer of LacY from GroEL into the INV, and, in the presence of both MgATP and GroES, about 30% of the LacY bound to GroEL becomes associated with the membranes. Since MgATP and GroES are required for dissociation of GroEL complexes with various proteins (7-9), the observations suggest that discharge of LacY from GroEL is a prerequisite for the association of LacY with membranes.
Integral membrane proteins are solubilized only under conditions that dissolve the membrane (i.e. in the presence of detergents) and remain membrane-bound under conditions that do not destroy the integrity of membrane (e.g. treatment with EDTA, alkaline pH, or high concentrations of chaotropes). As shown in Fig. 2C, treatment of INV-LacY under the conditions described reveals that LacY is solubilized only by the detergent DM which solubilizes the membrane. Extraction at high pH or with 6 M urea, which is known to release peripheral membrane proteins (20 -22), is not effective. The results support the conclusion that soluble, GroEL-bound LacY can insert into the plasma membrane.
Since interaction with GroEL maintains newly synthesized LacY in a soluble state, we reasoned that a mild detergent such as DM might mimic the chaperonin. In control experiments (not shown), it was demonstrated that the in vitro synthesis of is lower than the critical micelle concentration (0.6 mM (23) or 0.16 mM (24)). As shown in Fig. 3B, about half of the soluble LacY becomes associated with the membrane in a manner that is dependent on the presence of Mg 2ϩ but independent of MgATP. Furthermore, the membrane-associated LacY is extracted only with high concentrations of DM (1%) and not with alkali or 6 M urea (Fig. 3C). Thus, it seems likely that newly synthesized DM-soluble LacY is inserted into the membrane and that GroEL or DM interacts with LacY synthesized in vitro in such a manner as to promote post-translational insertion.
To study the specificity of the transfer of LacY from GroEL-LacY complex into membranes, we tested whether this productive interaction mimics the in vivo situation where insertion occurs from the cytosolic face of the membrane. For this purpose, the same experiment was carried out with right-side-out membrane vesicles prior to or after conversion of these membranes into INV by sonication. As shown in Fig. 2D, the interaction of LacY with the membranes is enhanced by sonication (lanes 2 and 3) that is a priori not surprising since after sonication the surface area of the vesicles increases. However, treatment under alkaline pH removes comparable quantities of LacY from the membrane in both cases (lanes 4 and 5). At the same time, the sonicated membranes contain a significantly higher quantity of LacY that is resistant to alkaline treatment and extractable only by detergents such as SDS (lanes 6 and 7) or DM (lanes 8 and 9). Thus, conversion of the right-side-out vesicles into INV clearly promotes insertion of LacY. This observation shows that, after release from ribosomes, LacY is able to recognize and target specifically to the cytosolic face of the membranes.
Does proper targeting of LacY lead to its proper insertion into the membranes? The most unequivocal means of answering the question is by functional assays. However, cell-free translation systems are known to be relatively inefficient, and, in this case, the amount of LacY inserted into the INV is so low that lactose transport assays are prohibited. Therefore, in order to test the function of post-translationally inserted permease, a more convenient assay was used which is based on the ability of the ligand TDG to protect LacY against alkylation by thiol reagents (1). As shown earlier (25), Cys-148 is the critical residue alkylation of which by NEM inactivates LacY, and this effect is inhibited by TDG (apparent K D of about 1.0 mM in INV (17)). In the original protocol described by Frillingos and Kaback (17), membranes from cells expressing LacY(Cys148) mutant with a biotin acceptor domain in the middle cytoplasmic loop were incubated with radioactive NEM in the absence or presence of 10 mM TDG followed by analysis of the permease labeling by avidin-Sepharose chromatography. In the present study, we used a LacY(Cys148) mutant without a biotin acceptor domain, and the alkylation was carried out in two steps. Note: The behavior of the mutant was similar compared with the wild type regarding the in vitro biosynthesis, binding to GroEL, and membrane insertion. First, the INV containing [ 35 S]LacY(Cys148) inserted post-translationally were treated with NEM in the absence or presence of TDG. Then, after removal of TDG (and DTT-inactivated reagent) by centrifugation, the formerly ligand-protected thiol groups were treated with NEM-biotin followed by isolation of the biotinylated [ 35 S]LacY(Cys148) on avidin-agarose. During the first step, it is expected that all the Cys residues of LacY fragments or the improperly inserted LacY which are exposed regardless of the presence of TDG should be blocked by NEM. This blockade should lower, during the second step, the background level of the NEM-biotinylation. The results are shown in Table I. A reproducible and reliable TDG-dependent enhancement of the NEM-biotinylation of LacY by more than 30% is observed. Consistent with its apparent K D of about 15 mM (26), the low-affinity ligand lactose exhibited no protective effect when tested at the same concentration as TDG (10 mM, see also Ref. 27). These results indicate that post-translationally inserted LacY molecules (although apparently not all of them) acquire a functional conformation.
In addition, a topological approach employing proteolysis was utilized (20) in order to show the proper post-translational insertion of LacY into the membranes. As shown previously, limited proteolysis of LacY inserted in vivo (19, 28 -30) or in vitro in the presence of membranes (5) yields two major proteolytic fragments, one with a molecular mass of 12-13 kDa observed with each of the proteases tested and another with a mass of either 18 -20 kDa (after treatment with thermolysin or chymotrypsin) or 23-25 kDa (after treatment with trypsin or clostripain). In contrast, when membranes are solubilized first with SDS, LacY is digested into very small peptide fragments (28,29). In a similar fashion, thermolysin converts the aggregates of [ 35 S]LacY, synthesized in vitro in the absence of INV, GroEL, or DM (fraction 14 in Fig. 1) and solubilized in SDS, into short (2-3-kDa) peptides and not into 12-, 18-, or 23-kDa fragments (data not shown). Thus, proteolysis is a proper criterion for tertiary structure of LacY.
Results obtained after digestion of LacY with thermolysin are presented in Fig. 4A. The following preparations of  PAGE, and the appearance of 2 fragments a and b at ϳ12 kDa and 18 kDa, respectively, is observed in each sample. The results suggest that post-translationally inserted LacY has native or near-native conformation. Importantly, the same pattern is observed when the LacY samples maintained in a soluble state with GroEL or DM are exposed to thermolysin (Fig.  4A, panels IV and V, respectively). Furthermore, an additional set of experiments (Fig. 4B) demonstrates that trypsin digestion of the GroEL-bound LacY (panel II) produces a fragmentation pattern identical to that observed with membranes containing LacY synthesized in vivo (panel I). With trypsin, however, proteolysis leads to the appearance of two major bands a and c at ϳ12 kDa and ϳ23 kDa, respectively (see Ref. 19 in addition).
In summary, the present study proposes that (i) in a soluble state, LacY is capable of folding into a structure that approximates roughly the tertiary structure of the permease in the native membrane, (ii) in a folded state, LacY recognizes specifically the cytosolic surface of the membrane and can integrate properly into the membrane by a post-translational mechanism. DISCUSSION Current understanding of targeting and insertion of integral membrane proteins into lipid bilayer stems mainly from a generally accepted notion that the process is similar to the translocation of secretory proteins across the membrane (31)(32)(33). Accordingly, one possible mode of biogenesis of polytopic membrane proteins in E. coli is cotranslational, resembling the SRP system of eukaryotes. Alternatively, targeting and insertion may be post-translational, similar to the pathway for protein translocation across the cytoplasmic membrane in E. coli. Although both modes of biogenesis may operate simultaneously, this study deals specifically with questions related to post-translational membrane protein targeting and insertion. It seems evident that post-translational biogenesis of hydrophobic membrane proteins like LacY must require a specific mechanism to prevent newly synthesized molecules from aggregating. It has been suggested that folding and stability of membrane-and water-soluble proteins follow similar principles leading to comparable polarities on their interiors but different polarities on their surfaces to enable solubilization in the appropriate environments (34,35). Specifically, membrane proteins possess an increased concentration of apolar residues on their surfaces and tend to aggregate, therefore, in an aqueous environment. Based on the finding presented here, it is proposed that GroEL or DM prevent aggregation by interaction with exposed hydrophobic patches on newly synthesized LacY. Moreover, by shielding the exposed hydrophobic residues, GroEL or DM may stabilize the transiently formed ␣-helices and helical hairpins and also their association to generate the proper folding. Stabilization of newly synthesized LacY in a near-native state in the presence of DM resembles the ability of permease, synthesized in the cell, to maintain close to native conformation in DM after extraction from membranes (36,37). It could be suggested that the LacY surface is surrounded by the uniform belt of DM molecules which presumably stabilizes the hydrophobic, transmembrane region of the permease and mimics its interactions with lipids in the membrane (38,39). Correspondingly, the apical hydrophobic surface in the central channel of GroEL, responsible for protein binding (40), could mimic the detergent belt and interact with the newly synthesized LacY molecule, thus stabilizing its native-like structure. It should be noted here that, unlike cytoplasmic proteins which must be unfolded in order to interact with GroEL (i.e. exposing their apolar interior (7-9)), the highly hydrophobic surface of membrane proteins like LacY may allow them to interact with the chaperonin also in the folded state.
The second step in biogenesis of membrane proteins is insertion into the membrane. This process has been postulated to occur by two alternative models. According to one hypothesis (41)(42)(43), insertion is directed by interactions between discrete topogenic regions in a newly synthesized membrane protein and specific receptor proteins in the membrane which initiates insertion through a water-filled proteinaceous channel. In the alternative proposal (44 -46), thermodynamic considerations of lipid-protein interactions lead to a suggestion that protein integration into the membrane is a spontaneous process. In this case, the polypeptide chains move directly through a hydrophobic core of the lipid bilayer during insertion. Both concepts presume that the nascent polypeptide chain should be partially (or completely) unfolded in order to integrate into the membrane, and stabilization of hydrophobic ␣-helices or helical hairpins and their assembly into a native three-dimensional structure occur only within the membrane.
The data presented here suggest that: (i) under conditions where aggregation is prevented, LacY can fold into a state corresponding roughly to its tertiary structure within the membrane, and (ii) in the folded state, LacY can specifically recognize and interact with the cytosolic surface of the membrane followed by proper insertion (i.e. the folded state of LacY seems not a barrier for membrane insertion).
Finally, not much is known at present about the possible participation of GroEL in membrane protein biogenesis in intact cells, and to the best of our knowledge only one published preliminary work of Sato et al. (47) deals with this question. These authors observed that overproduced GroEL/ES protects an unstable, truncated form of the tetracycline/H ϩ antiporter in E. coli against degradation. Our results certainly predict that GroEL may play a physiological role in membrane protein biogenesis, especially under stress conditions such as heat shock when it is overexpressed.