Roles of the Hydrophobic Cavity and Lid of LolA in the Lipoprotein Transfer Reaction in Escherichia coli*

LolA, a periplasmic chaperone, binds to outer membrane-specific lipoproteins released from the inner membrane through the action of an ATP-binding cassette transporter, LolCDE and then transfers them to the outer membrane receptor LolB, thereby mediating the inner to outer membrane transport of lipoproteins. The crystal structure of free LolA revealed that it has an internal hydrophobic cavity, which is surrounded by hydrophobic residues and closed by a lid comprising α-helices. The hydrophobic cavity most likely represents the binding site for the lipid moiety of a lipoprotein. It is speculated that the lid undergoes opening and closing upon the binding and transfer of lipoproteins, respectively. To determine the functions of the hydrophobic cavity and lid in detail, 14 residues involved in the formation of these structures were subjected to random mutagenesis. Among the obtained 21 LolA derivatives that did not support growth, 14 were active as to the binding of lipoproteins but defective in the transfer of lipoproteins to LolB, causing the periplasmic accumulation of a lipoprotein as a complex with a LolA derivative. A LolA derivative, I93G, bound lipoproteins faster than wild-type LolA did, whereas it did not transfer associated lipoproteins to LolB. When I93G and wild type LolA co-existed, lipoproteins were bound only to I93G; which therefore exhibited a dominant negative property. Another derivative, L59R, was also defective in the transfer of lipoproteins to LolB but did not exhibit a dominant negative property. Taken together, these results indicate that both the hydrophobic cavity and the lid are critically important for not only the binding of lipoproteins but also their transfer.

Escherichia coli has at least 90 lipoproteins (1)(2)(3), which are anchored through N-terminal lipids to the periplasmic side of either the inner or outer membrane. Lipoproteins synthesized as precursors are translocated to the periplasmic side of the inner membrane and then processed to mature forms (4,5). The N-terminal Cys of mature lipoproteins is modified by thioether-linked diacylglycerol and amino-linked acyl chains (6). When lipoproteins have Asp at position 2 (7), they are retained in the inner membrane, presumably through an interaction between the Asp at position 2 and phospholipids (8). Residues other than Asp at position 2 cause outer membrane localization of lipoproteins (9,10).
The sorting and outer membrane localization of lipoproteins depend on the Lol system comprising five Lol proteins. The LolCDE complex belongs to an ATP-binding cassette transporter superfamily and releases outer membrane-directed lipoproteins from the inner membrane in an ATP-dependent manner (11), leading to the formation of a water-soluble complex comprising one molecule each of a lipoprotein and LolA in the periplasm (12). LolA then transfers the associated lipoprotein to LolB in the outer membrane in an energy-independent manner. LolB is itself a lipoprotein anchored to the outer membrane and catalyzes the incorporation of lipoproteins into the outer membrane (13).
The structures of free LolA and LolB are strikingly similar to each other, whereas their amino acid sequences are dissimilar (14). Both proteins are characterized as an incomplete ␤-barrel having an internal space surrounded by hydrophobic residues and covered by a lid composed of three ␣-helices. This hydrophobic cavity most likely represents the binding site for the lipid moiety of a lipoprotein. More than 10 aromatic and aliphatic residues surround the hydrophobic cavity of LolA (Fig. 1A, gray globules) (14). The LolA lid closes due to the hydrogen bonds between Arg at position 43, which is located in the loop connecting the ␤2 and ␤3-strands, and residues in the ␣-helices (Fig.  1B). On the other hand, the LolB lid is always open, and one crystal form of LolB contained polyethylene glycol monomethyl ether, which was used in the crystallization process, in its hydrophobic cavity (14). These results suggest that opening and closing of the LolA lid upon the binding and transfer of lipoproteins, respectively, is important for efficient unidirectional lipoprotein transfer between the two similar structures. Indeed, substitution of Arg 43 with Leu generated a defective LolA mutant, R43L, which can accept a lipoprotein but cannot transfer it to LolB (15). However, recent analysis revealed that any residue other than Leu could replace Arg 43 without inhibiting growth (16), suggesting that residues other than Arg 43 also contribute to the closing of the lid or, alternatively, the lid closing is not absolutely essential for the LolA function, although unliganded LolA only exists in the closed form (14). On the other hand, the efficiency of lipoprotein transfer to LolB was variously affected depending on the species of residue introduced at position 43 (16). From these results, it was proposed that the interaction between LolA and lipoproteins should be weak for efficient transfer of lipoproteins to LolB.
The replacement of Phe at position 47 with Glu gave a dominant negative LolA derivative, which cannot accept lipoproteins and remains tightly associated with LolCDE (17). Since Phe 47 is one of the residues forming the hydrophobic cavity (Fig. 1A), Glu introduced at this position is speculated to cause additional hydrogen bonding, thereby rendering lid opening more difficult (16).
Taken together, these results suggest the importance of the hydrophobic cavity and lid closure for the ability of LolA to bind and transfer lipoproteins. Here we examined the functions of LolA derivatives having mutations of the residues forming the hydrophobic cavity or the lid. a hexahistidine tag (20) were raised in rabbits as described. Antibodies against maltose-binding protein were purchased from New England Biolabs.  (21), and TT016 (JE5505 lpp ϩ lacPO-lolA) (15) were used. The last strain carries the chromosomal lolA gene under the control of the lactose promoteroperator. Expression of the chromosomal lolA gene was induced with 0.1 mM IPTG 2 unless otherwise noted. L-broth was used as the standard medium. Labeling experiments were carried out in M63 minimal medium (22) supplemented with 0.5% NaCl, 0.2% maltose, 20 g/ml thiamine, 20 g/ml thymine, and 40 g/ml all amino acids except methionine and cysteine. When required, chloramphenicol was added at a concentration of 25 g/ml.
Construction of LolA Mutants-As reported for the construction of LolA derivatives R43L (15) and F47E (17), random mutagenesis of target amino acid residues was performed by means of PCR with a QuikChange site-directed mutagenesis kit (Stratagene) according to the manufacturer's protocol, with pAM201 carrying lolA-His under the control of P BAD (15) as a template and a pair of oligonucleotides (supplemental Table S1) as primers. The plasmid DNAs were amplified in the XL1-Blue strain lacking functional endonuclease R and then transformed into TT016. Chloramphenicol-resistant transformants were selected on L-broth plates supplemented with 0.1 mM IPTG, and then their growth in the absence of IPTG was examined on plates containing 0.2% arabinose. Plasmids carrying defective lolA were isolated from the transformants that did not grow on these plates.
The mutations were confirmed by sequencing of both strands of DNA. Where specified, the growth of TT016 cells harboring the specified plasmids was examined in the presence of both 0.1 mM IPTG and the indicated concentrations of arabinose.
To construct pSW77, which carries a gene encoding wild-type LolA with a FLAG tag at its C terminus, PCR was performed with a QuikChange site-directed mutagenesis kit, with pAM201 as a template and a pair of oligonucleotides (supplemental data).
Purification of His-tagged or FLAG-tagged LolA Proteins-TT016 cells harboring pAM201 or a derivative of it encoding His-tagged LolA, I93G, I93E, I93N, L59R, or R43L were grown on L-broth supplemented with 1 mM IPTG and 25 g/ml chloramphenicol at 37°C. At A 660 of 0.6, the cells were induced with 0.2% arabinose for 1.5 h and then converted to spheroplasts as described (23). A periplasmic fraction was obtained as a spheroplast supernatant and dialyzed against 20 mM Tris-HCl (pH 8.0), 300 mM NaCl, 10 mM MgCl 2 overnight at 4°C. His-tagged LolA and derivatives were purified on a metal affinity column packed with TALON (Clontech) resin as reported (15). Some LolA derivatives contained lipoproteins, as shown below. To remove associated lipoproteins, all His-tagged LolA proteins adsorbed to the column were washed three times with 20 mM Tris-HCl (pH 8.0) containing 300 mM NaCl and 0.2% ␤-D-fructopyranosyl-␣-D-glucopyranoside monodecanoate. The column was extensively washed to remove the detergent and then eluted with 20 mM Tris-HCl (pH 8.0) containing 300 mM NaCl and 250 mM imidazole.
FLAG-tagged wild-type LolA was expressed in TT016 cells harboring pSW77 (P BAD -lolA-FLAG) with 0.2% arabinose. Periplasmic fractions containing FLAG-tagged LolA were adsorbed to an anti-FLAG M2 affinity column and eluted with 20 mM Tris-HCl (pH 8.0) containing 100 g/ml FLAG peptide. Purified His-tagged or FLAG-tagged LolA proteins were dialyzed against 20 mM Tris-HCl (pH 8.0) overnight at 4°C and then kept frozen at Ϫ80°C.
Release of L10P from Spheroplasts-The reported method (12) was slightly modified to examine the release reaction after the completion of L10P maturation. Briefly, E. coli DLP79-36 cells harboring pJYL10P (P BAD -L10P) (10) were grown on M63 (0.5% NaCl), 0.2% maltose minimal medium at 37°C. At A 660 of 1.0, the cells were induced with 0.2% arabinose for 5 min and then converted to spheroplasts. The spheroplasts were stabilized by the addition of 20 mM MgCl 2 , collected by centrifugation at 16,000 ϫ g for 2 min, and then resuspended in M63 (0.5% NaCl), 0.2% maltose minimal medium supplemented with 250 mM sucrose and 10 mM MgCl 2 . The suspension (300 l) containing 5 ϫ 10 8 spheroplasts was mixed with 750 l of M63 (0.5% NaCl), 0.2% maltose minimal medium supplemented with 250 mM sucrose, 10 mM MgCl 2 , and 10 Ci of Tran 35 S-label, followed by 1-min pulse-labeling and a 10-min chase with 12 mM nonradioactive Met and Cys. Histagged LolA proteins (2 g/ml) were then added to induce the release of L10P. To terminate the reaction, the mixture was chilled in ice water, followed by centrifugation at 16,000 ϫ g for 2 min. The spheroplasts and supernatant thus obtained were subjected to trichloroacetic acid precipitation and immunoprecipitation with anti-Lpp antibodies as reported (12). 35 S-Labeled L10P was analyzed by SDS-PAGE and fluorography.
Inhibition of LolA-dependent Release by Derivatives-Spheroplasts prepared as described above were pulse-labeled with 10 Ci of Tran 35 Slabel for 1 min in the presence of FLAG-tagged wild-type LolA (2 g/ml) and His-tagged wild-type LolA, I93G, L59R, or R43L (0 -2 g/ml) in a total volume of 1,050 l, followed by a 5-min chase with nonradioactive Met and Cys. The release reaction was terminated, and a spheroplast supernatant was obtained as described above. The supernatant was diluted 10-fold with fresh M63 minimal medium supplemented with 0.5% NaCl, 0.2% maltose, and 250 mM sucrose. An equal volume of the diluted supernatant was then applied to affinity columns for FLAG-tagged proteins and His-tagged proteins. Proteins adsorbed to the columns were eluted with FLAG peptide or imidazole, as described for the purification of tagged LolA proteins, and then subjected to trichloroacetic acid precipitation and immunoprecipitation with anti-Lpp antibodies, followed by SDS-PAGE and fluorography.
In Vitro Outer Membrane Incorporation of L10P-LolB-dependent incorporation of L10P into outer membranes was examined using 35 Slabeled L10P released from spheroplasts, as described (13). Briefly, a spheroplast supernatant (1 ml) containing the LolA-L10P complex was centrifuged at 100,000 ϫ g for 30 min to remove insoluble materials. The supernatant was then incubated at 30°C for the indicated times in the presence of 0.2 mg/ml outer membranes. The outer membrane incorporation of L10P was terminated by chilling of the reaction mixture (200 l) in ice water and analyzed after fractionation into a supernatant and pellet by centrifugation at 100,000 ϫ g for 30 min. The supernatant and pellet fractions thus obtained were analyzed by SDS-PAGE and fluorography after trichloroacetic acid precipitation.
Other Methods-Outer membranes were prepared from E. coli JE5505 cells as reported (13). Immunoprecipitation was performed as described (24). SDS-PAGE was carried out as described by Laemmli (25) and Hussain et al. (19). Densitometric quantification was performed with an ATTO Densitograph.

RESULTS
Hydrophobic Cavity and Lid of LolA-Residues forming the hydrophobic cavity (Fig. 1A) are conserved among LolA homologs (italicized boldface letters in Fig. 1C). Arg 43 forms hydrogen bonds with the main chain of Leu 10 in the ␣1-helix, Val 13 in the loop between the ␣1-helix and the ␤1-strand, and Ile 93 and Ala 94 in the ␣2-helix, thereby causing the tight fixation of the ␣-helices to the ␤-strands like a lid (Fig. 1B). The residues forming hydrogen bonds with Arg 43 are aliphatic ones and are conserved among LolA homologues (boldface letters in Fig. 1C). The lid of LolA, comprising three ␣-helices, is expected to undergo opening and closing upon the binding and transfer of lipoproteins, respectively (14).
To understand how the hydrophobic cavity and lid status affects the LolA function, 14 residues involved in the formation of the hydrophobic cavity and closing of the lid were subjected to random mutagenesis. LolA mutants that could not support the growth of E. coli were then screened for. Among 21 mutants thus isolated, five exhibited a dominant negative phenotype (i.e. expression of the mutant inhibited growth even in the presence of wild-type LolA) (    1IWL) (14), suggesting that the hydrogen bonding between these two residues contributes little to the stabilization of the closed form of LolA. Among 10 residues surrounding the hydrophobic space, Tyr 152 is located most distantly from the hydrophobic cavity (Fig. 1A). This may be the reason why no defective mutant was obtained for Tyr 152 . Periplasmic Accumulation of Lipoproteins-Since the outer membrane localization of lipoproteins is very rapid, they are not detected in the periplasm under normal conditions (15). On the other hand, many residues introduced in place of Arg at position 43 were recently found to cause the periplasmic accumulation of lipoproteins as complexes with LolA derivatives (16). These mutations were found to increase the strength of the hydrophobic interaction between lipoproteins and LolA derivatives, thereby reducing the efficiency of lipoprotein transfer to LolB (16). Since we introduced mutations at residues involved in formation of the hydrophobic cavity or stabilization of the LolA lid, periplasmic accumulation of outer membrane-specific lipoproteins, Lpp and Pal, was examined in TT016 cells expressing the respective LolA derivatives in the presence of wild-type LolA (Fig. 2).
When wild-type LolA was expressed from a plasmid, neither Lpp nor Pal was detected in the periplasm (Fig. 2, left lane). Maltose-binding protein (26) was examined as a control to confirm that similar amounts of periplasmic proteins were examined for the respective cultures. Among the defective LolA derivatives examined, 14 were found to cause the periplasmic accumulation of lipoproteins to various extents ( Fig. 2 and Table 1), suggesting that these derivatives are active as to lipoprotein binding but defective in lipoprotein transfer. The properties of the I93G derivative were the most interesting among the defective derivatives, because it caused the most significant periplasmic accumulation of lipoproteins and exhibited a dominant negative phenotype, whereas two other defective derivatives, I93E and I93N, had neither property. Ile 93 forms a hydrogen bond with Arg 43 and undergoes a hydrophobic interaction with Phe 140 (Fig. 1B), which is located near the hydrophobic cavity (Fig. 1A). Ile 93 is therefore involved in closing of the LolA lid and, in addition, may affect the status of the hydrophobic cavity. Disruption of the hydrophobic interaction between Ile 93 and Phe 140 is most likely the reason why I93E and I93N are defective. Indeed, when the same mutations were introduced at Phe 140 , the resultant derivatives, F140E and F140N, were also defective. 3 We then asked why I93G exhibits a dominant negative phenotype.
TT016 cells harboring a plasmid carrying a gene for the wild type LolA or an Ile 93 derivative were grown in the presence of 0.1 mM IPTG and various concentrations of arabinose (Fig. 3A). The chromosomeencoded wild-type LolA and plasmid-encoded His-tagged LolA proteins in the periplasmic fraction could be distinguished on SDS-PAGE (Fig. 3B). Overexpression of the wild-type LolA from both the chromosome and plasmid did not affect growth although the LolA level in the presence of 0.2% arabinose was significantly higher than that in its absence. Neither the I93E nor the I93N derivative inhibited growth, although both derivatives were fully expressed. In marked contrast, I93G inhibited growth with arabinose concentrations as low as 0.02% (Fig. 3A), at which the level of I93G was only slightly higher than that of wild-type LolA (Fig. 3). When I93G was fully induced with 0.2% arabinose, wild-type LolA became almost undetectable for an unknown reason, whereas expression of I93E or I93N hardly affected the level of the wild-type LolA. Since I93G inhibited growth with 0.02% arabinose (Fig.  3A), at which the wild-type LolA was stably expressed (Fig. 3B), disappearance of the wild-type LolA does not account for the dominant negative effect of I93G.
The levels of Pal and Lpp accumulated in the periplasm increased with an increase in the amount of I93G (Fig. 3B). In contrast, two other Ile 93 derivatives did not cause the periplasmic accumulation of lipoproteins even when they were expressed at maximum levels.

I93G Efficiently Binds Lipoproteins but Cannot Transfer Them to LolB-L10P is a derivative of the major outer membrane lipoprotein
Lpp and is suitable for kinetic analysis of the release reaction, since its release from spheroplasts absolutely depends on LolA and takes place even after mature L10P has been kept anchored to the inner membrane for a long time (27). L10P was labeled with 35 S and chased with nonradioactive amino acids in spheroplasts in the absence of LolA, and then its release was induced by the addition of LolA, I93G, or another Ile 93 derivative (Fig. 4). [ 35 S]L10P remained in the spheroplasts in the absence of LolA, whereas the addition of LolA caused the nearly complete release of L10P in 2 min (Fig. 4A). I93G also induced the L10P release into the supernatant, and little [ 35 S]L10P remained in the spheroplasts at 1 min after the addition of I93G. On the other hand, neither I93E nor I93N induced the release of L10P. Densitometric quantification of [ 35 S]L10P release revealed that I93G is considerably more active than the wild-type LolA as to the release reaction (Fig. 4). To determine whether or not this efficient release of lipoproteins from spheroplasts is characteristic of I93G, the release activity was examined with another derivative, in which a residue forming the hydrophobic cavity had been mutated. L59R possessed Arg in place of Leu at position 59, which is involved in the formation of the hydrophobic cavity (Fig.  1A). This derivative caused the periplasmic accumulation of lipoproteins but did not exhibit a dominant negative phenotype ( Table 1). The rate of L10P release by L59R (Fig. 4C) was essentially the same as that by wild-type LolA (Fig. 4).
To examine the transfer of lipoproteins to the LolB-containing outer membrane, L10P was released with LolA, I93G, or L59R from spheroplasts as in Fig. 4, and the spheroplast supernatant containing the LolA-L10P, I93G-L10P, or L59R-L10P complex was incubated with the outer membrane. L10P released as a complex with LolA was nearly completely transferred to the outer membrane in about 30 min (Fig. 5, A and  C). In contrast, most L10P molecules released as a complex with I93G (Fig. 5A) or L59R (Fig. 5C) remained in the supernatant even after a 60-min incubation with the LolB-containing outer membrane. Densitometric quantification revealed that the transfer activities of I93G and L59R were about 10 and 3% of that of LolA at 20 min after the start of the reaction, respectively (Fig. 5, B and D). Taken together, these results indicate that both I93G and L59R are defective in the transfer of associated lipoproteins to LolB. Moreover, it is suggested that the dominant negative phenotype of I93G is related to its very high activity to bind lipoproteins.
To determine whether or not LolA derivatives inhibits the binding of lipoproteins to wild-type LolA, we constructed and purified LolA pos-  sessing a FLAG tag at its C terminus. Expression of this LolA-FLAG alone supported the growth of TT016 in the absence of IPTG (data not shown). Moreover, LolA-FLAG released L10P from spheroplasts (Fig.  6A) and then transferred L10P to LolB (data not shown). These results indicate that LolA-FLAG is functional both in vivo and in vitro. The release of [ 35 S]L10P from spheroplasts was examined in the presence of a fixed amount (2 g) of LolA-FLAG and various amounts of His-tagged LolA, I93G, L59R, or R43L. R43L was used as a control, because this derivative causes the periplasmic accumulation of lipoproteins but does not exhibit the dominant negative property (15). L10P released as complexes with LolA proteins into the spheroplast supernatant was adsorbed to His tag and FLAG tag affinity columns and then eluted with imidazole and FLAG peptide, respectively. [ 35 S]L10P co-eluted with the respective LolA proteins was analyzed by fluorography (Fig. 6A). LolA-FLAG and LolA-His were found to be equivalent as to the release activity, and almost the same amounts of L10P were co-eluted with LolA-FLAG and LolA-His when the same amounts of the two LolA protein species were added. The activities of R43L and L59R were also nearly the same as that of the wild-type LolA. On the other hand, when I93G-His was added, the amount of L10P co-eluted with LolA-FLAG was significantly decreased. The results were plotted as a function of the amount of His-tagged LolA protein (Fig. 6, B and C). The amounts of L10P bound to 2 g of LolA-FLAG decreased to 50% upon the addition of 2 g of LolA-His, R43L-His, or L59R-His. On the other hand, 1 g of I93G-His completely inhibited L10P binding to 2 g of LolA-FLAG (Fig.  6B). The amounts of L10P bound to the His tag affinity column (Fig. 6C)  FEBRUARY 10, 2006 • VOLUME 281 • NUMBER 6  Fig. 4. Spheroplast supernatants containing the LolA-L10P, I93G-L10P, or L59R-L10P complex were incubated for the indicated times with 0.2 mg/ml outer membrane containing LolB as described under "Experimental Procedures." A and C, the reaction mixtures were fractionated into pellets (P) and supernatants (S) by centrifugation. L10P in each fraction was detected by SDS-PAGE and fluorography. B and D, the fluorograph shown in A and B was quantified as in Fig. 4, and the percentages of L10P incorporated into the outer membrane were plotted as a function of time. indicate that His-tagged LolA protein competitively inhibited the binding of L10P to FLAG tag LolA (Fig. 6B).

Hydrophobic Cavity and Its Lid of LolA
I93G Does Not Take Up L10P That Is Already in a Complex with LolA-The results presented above indicate that I93G has considerably higher affinity for lipoproteins than does the wild-type LolA. To determine whether or not I93G can take up L10P from the LolA-L10P complex, [ 35 S]L10P was released from spheroplasts with the wild-type LolA possessing no tag. The spheroplast supernatant containing the LolA-L10P complex was incubated with various amounts of His-tagged I93G and then applied to the affinity column for His-tagged proteins. No L10P molecule adsorbed to the His tag affinity column (Fig. 7), indicating that I93G does not take L10P that is already present as a complex with LolA. Therefore, inhibition of the LolA-L10P complex formation by I93G (Fig. 6) takes place at the step of L10P release from the inner membrane.

DISCUSSION
LolA has a hydrophobic cavity inside its incomplete ␤-barrel, which is covered by a lid comprising ␣-helices (14). Such a structure is similar to StAR-related lipid transfer (START) domains (28), which are lipidbinding domains involved in intracellular lipid transport, lipid metabolism, and cell signaling and found in an extensive protein family in eukaryotic cells (29). The prototype is StAR (steroidogenic acute regulatory protein), which transfers cholesterol to the inner membrane of mitochondria in steroid hormone-producing cells. Human phosphatidylcholine transfer protein is a START domain family protein. The crystal structure of phosphatidylcholine transfer protein containing lipid inside the hydrophobic cavity has been reported (30). The hydrophobic cavity of LolA most likely represents a binding site for the lipid moiety of lipoproteins, whereas no amino acid sequence similarity exists between START domains and LolA. The structure of the LolA-lipoprotein complex is not yet known. The details of the function of the hydrophobic cavity therefore remain to be determined. In contrast, recent biochemical analyses suggested that the strength of the hydrophobic interaction between LolA and lipoprotein is critical for efficient lipoprotein transfer to LolB and is affected by residues introduced at position 43 (16). To understand the functional importance of the hydrophobic cavity and closing of the LolA lid, amino acid residues expected to be involved in these structures were mutagenized.
Of 21 defective LolA mutants, 14 caused periplasmic accumulation of lipoproteins to various extents (Fig. 2), indicating that these 14 mutations did not abolish the lipoprotein-binding activity but impaired the lipoprotein transfer activity. Three mutations were of residues involved in closing of the lid, and the other 11 were of residues forming the hydrophobic cavity (Table 1). It should be noted that the residues involved in the lid closing also form the hydrophobic cavity (Fig. 1A). If the hydrophobic interaction between LolA and a lipoprotein is too strong, lipoprotein transfer from LolA to LolB is inhibited (16). Therefore, the 14 mutations might increase the strength of the hydrophobic interaction. Alternatively, the mutations impaired the interaction with LolB. Among these mutations, five exhibited a dominant negative property. These mutants are especially interesting because they are expected to inhibit the function of wild-type LolA. Moreover, dominant negative mutants exhibiting lipoprotein-releasing activity have not been isolated before. On the other hand, seven mutants neither exhibited a dominant negative property nor caused periplasmic accumulation of lipoproteins. It may be noteworthy that five of them were isolated as sole mutants as to the respective residues despite our exhaustive mutant isolation. It seems likely that these five residues are less important for the LolA from spheroplasts was carried out at 30°C for 5 min in the presence of 2 g LolA-FLAG and various amounts of either LolA-His, R43L-His, I93G-His, or L59R-His. The reaction mixture was fractionated into a supernatant and pellet by centrifugation. An equal volume of the supernatant was adsorbed to affinity columns for FLAG-tagged and Histagged proteins and then eluted from the columns as described under "Experimental Procedures." A, [ 35 S]L10P was released with 2 g of LolA-FLAG (F) and the indicated amounts of His-tagged (H) LolA, R43L, I93G, or L59R. L10P co-eluted with LolA proteins adsorbed to a FLAG tag affinity column (F), and a His tag affinity column (H) was immunoprecipitated with anti-Lpp antibodies and then detected by SDS-PAGE and fluorography. B and C, the amount of L10P released as a complex with LolA-FLAG (B) or the indicated His-tagged LolA protein (C) was determined and expressed as a percentage, taking the amount in the absence of His-tagged proteins as 100. Open circles, LolA-His; closed diamonds, R43L-His; closed circles, I93G-His; closed triangles, L59R. function, and only mutations causing significant structure alteration, such as the mutation to Pro, were isolated. Therefore, most mutants seemed to have completely lost the LolA function.
Mutations of the residues involved in the lid closing are likely to affect the hydrophobic interaction between LolA and lipoproteins, since these residues form the hydrophobic cavity (Fig. 1A). Indeed, three mutants, I93G, A94I, and A94L, exhibited a dominant negative property and caused periplasmic lipoprotein accumulation, suggesting that these mutants have higher affinity for lipoproteins and inhibit the wild-type LolA. The increase in the hydropathy scale (31) upon mutation from Ala (ϩ1.8) at position 94 to Ile (ϩ4.5) or Leu (ϩ3.8) most likely caused the increase in the affinity for lipoproteins. On the other hand, another dominant negative mutation caused a hydropathy scale change from Ile (ϩ4.5) to Gly (Ϫ0.4) at position 93. Therefore, an increase in the hydrophobic interaction between I93G and lipoproteins is difficult to assume. Moreover, a further decrease in the hydropathy scale to Ϫ3.5 (Glu and Asn) at this position caused two nonfunctional mutants, I93E and I93N. We therefore examined the mechanism by which I93G exhibits the dominant negative property.
The release reaction catalyzed by I93G was faster than that catalyzed by the wild-type LolA (Fig. 4). In contrast, its ability to transfer the associated lipoprotein to LolB was significantly defective (Fig. 5), causing the periplasmic accumulation of the I93G-L10P complex even in the presence of the wild-type LolA. R43L and L59R also exhibited reduced ability to transfer lipoproteins to LolB (15), whereas the efficiency of complex formation with L10P was similar for R43L, L59R, and wild-type LolA (Fig. 6). In marked contrast, I93G was much more efficient than the wild-type LolA as to the formation of a complex with L10P (Fig. 6). These results indicate that I93G has higher affinity for LolCDE than does wild-type LolA, R43L, or L59R. I93G therefore inhibits the complex formation between the wild-type LolA and L10P and thereby perturbs the lipoprotein transfer. This is the reason why the expression of I93G, but not R43L or L59R, inhibits the growth of cells even in the presence of wild-type LolA (Fig. 3). Importantly, however, I93G is unable to take up L10P once the complex is formed between L10P and LolA (Fig. 7).
Because of the absence of a side chain, Gly is a unique amino acid and is known as a helix breaker (32). We speculate that Gly introduced in place of Ile located in the ␣2-helix causes a significant structural change such as the opening of the I93G lid and exposure of the hydrophobic cavity to the bulk solvent region. As a result, the affinity for LolCDE is higher with I93G than with wild-type LolA. On the other hand, we speculate that the mutation decreased the affinity for LolB, thereby inhibiting the transfer of lipoproteins from I93G to LolB. Taken together, the results obtained with I93G strongly suggest that the status of the LolA lid is crucial for the interaction with LolCDE, LolB, and lipoprotein.
As observed with F20R, L59R, M91S, I106D, I106Q, and I106R, mutation of hydrophobic residues forming the hydrophobic cavity to polar ones caused periplasmic accumulation of lipoproteins. It is not clear at present whether or not the interaction with lipoproteins is strengthened by these mutations. A more plausible explanation is that these muta-tions impair the interaction with LolB, thereby preventing the transfer of lipoproteins to LolB. In any event, these mutants exhibited an unexpected phenotype and are worthy of further analyses.
Properties of Leu and Arg differ significantly. It is not immediately clear why R43L and L59R exhibited the same phenotype (i.e. active in the lipoprotein binding but defective in the lipoprotein transfer) although directions of the amino acid substitutions were opposite between the two derivatives. L59R seems to be a useful derivative for further examination of the function of the hydrophobic cavity.