Anthony P. Pugsley. E-mail firstname.lastname@example.org; Tel. (+33) 1 45688494; Fax (+33) 1 45688960.
We report a novel strategy for selecting mutations that mislocalize lipoproteins within the Escherichia coli cell envelope and describe the mutants obtained. A strain carrying a deletion of the chromosomal malE gene, coding for the periplasmic maltose-binding protein (MalE), cannot use maltose unless a wild-type copy of malE is present in trans. Replacement of the natural signal peptide of preMalE by the signal peptide and the first four amino acids of a cytoplasmic membrane-anchored lipoprotein resulted in N-terminal fatty acylation of MalE (lipoMalE) and anchoring to the periplasmic face of the cytoplasmic membrane, where it could still function. When the aspartate at position +2 of this protein was replaced by a serine, lipoMalE was sorted to the outer membrane, where it could not function. Chemical mutagenesis followed by selection for maltose-using mutants resulted in the identification of two classes of mutations. The single class I mutant carried a plasmid-borne mutation that replaced the serine at position +2 by phenylalanine. Systematic substitutions of the amino acid at position +2 revealed that, besides phenylalanine, tryptophan, tyrosine, glycine and proline could all replace classical cytoplasmic membrane lipoprotein sorting signal (aspartate +2). Analysis of known and putative lipoproteins encoded by the E. coli K-12 genome indicated that these amino acids are rarely found at position +2. In the class II mutants, a chromosomal mutation caused small and variable amounts of lipoMalE to remain associated with the cytoplasmic membrane. Similar amounts of another, endogenous outer membrane lipoprotein, NlpD, were also present in the cytoplasmic membrane in these mutants, indicating a minor, general defect in the sorting of outer membrane lipoproteins. Four representative class II mutants analysed were shown not to carry mutations in the lolA or lolB genes, known to be involved in the sorting of lipoproteins to the outer membrane.
Bacterial lipoproteins form a subclass of exported proteins that are processed by a specific signal peptidase (lipoprotein signal peptidase) and whose N-terminal cysteine residue is modified by the addition of fatty acids that anchor the protein in the membrane (Wu, 1996). The sequence of the amino acids around this cysteine residue in lipoprotein precursors is relatively well-conserved, and their signal peptides have several features that distinguish them from precursors of other bacterial-exported proteins (von Heijne, 1989; Wu, 1996). Thus, putative lipoproteins can be identified with relatively high precision in predicted protein sequences from genomic databases. In the Escherichia coli K-12 genome database, for example, we identified 94 putative or known lipoproteins, many more than can be identified by labelling E. coli lipoproteins with radioactive fatty acids (Ichihara et al., 1981).
In E. coli, and presumably also in other Gram-negative bacteria, lipoproteins are usually found in one of two quite distinct locations: the periplasmic face of the outer membrane and the outer face of the cytoplasmic (inner) membrane. In some Gram-negative bacteria, lipoproteins are also found totally or partially exposed on the cell surface. In some cases, such as pullulanase (PulA) produced by Klebsiella oxytoca (d'Enfert et al., 1987), cell surface localization depends on a complex machinery that, in this particular case, translocates the lipoprotein to the outside of the cell.
A first seminal step towards understanding the molecular mechanisms underlying the sorting of lipoproteins between the periplasmic faces of the cytoplasmic and outer membranes was made by Inouye and collaborators (Yamaguchi et al., 1988). They observed that two lipoproteins located in the cytoplasmic membrane, NlpA (Yamaguchi et al., 1988) and PulA, in strains in which secretion was blocked (Pugsley et al., 1990), had an aspartate residue (D+2) immediately after the fatty-acylated cysteine, whereas outer membrane lipoproteins consistently had other amino acids at this position. The idea that D+2 could specify cytoplasmic membrane retention of lipoproteins was confirmed by elegant studies in which replacement of residue +2 in an outer membrane lipoprotein by an aspartate causes the protein to remain anchored to the cytoplasmic membrane, whereas substitution of D+2 in NlpA by other amino acids (e.g. S+2 frequently found in outer membrane lipoproteins) caused it to become associated with the outer membrane. Subsequent studies from the same group (Gennity and Inouye, 1991; Gennity et al., 1992) confirmed the pivotal role of D+2 and led to the general acceptance of the ‘+2 rule’ for lipoprotein sorting.
The second seminal step in studies of lipoprotein sorting was the discovery of two E. coli cell envelope proteins, the periplasmic lipoprotein shuttle protein LolA and the outer membrane lipoprotein transit receptor LolB. These proteins mediate the transport of lipoproteins other than those with the D+2 signal to the outer membrane (Matsuyama et al., 1995; 1997). Thus, D+2 is a ‘default’ cytoplasmic membrane retention signal; lipoproteins with other amino acids at position +2 are actively displaced to the outer membrane by a machinery that includes LolA and LolB and a putative ATPase (Matsuyama et al., 1995; 1997; Yakushi et al., 1998).
The structural genes for both LolA and LolB have been cloned, and lolA has been shown to be essential (Tajima et al., 1998). The toxic effects of Lpp accumulation in the cytoplasmic membrane caused by the replacement of the S+2 by D+2 can be overcome by replacing the C-terminal lysine residue of Lpp that is normally cross-linked to the peptidoglycan (Yakushi et al., 1997). However, LolA depletion is toxic even in strains lacking Lpp (Tajima et al., 1998), indicating that at least one other outer membrane lipoprotein is toxic to cells if it is retained in the cytoplasmic membrane.
LolA and LolB were discovered using a biochemical approach. The sorting of lipoproteins should also be amenable to genetic analysis. Unfortunately, most of the characterized E. coli K-12 lipoproteins do not have a known function. NlpA, the archetype cytoplasmic membrane lipoprotein, is non-essential, and its relocalization to the outer membrane does not generate a new phenotype. The extracellular secretion of PulA in E. coli is reduced when its D+2 residue is replaced by S+2, N+2 or E+2, all of which are thought to cause sorting of the secretion intermediate to the periplasmic face of the outer membrane, but the secretion defect is not sufficient to allow it to be used to select sorting-defective mutants (Pugsley and Kornacker, 1991). An outer membrane lipoprotein component of the pullulanase secretion machinery, PulS, is non-functional if it is relocated to the cytoplasmic membrane by the introduction of a D+2 residue, resulting in a total defect in pullulanase secretion (Hardie et al., 1996). Although pullulanase secretion can be used as a selective tool (d'Enfert et al., 1987), it is unlikely that restoration of secretion in strains with cytoplasmic membrane-anchored PulS could be used as a positive selection for lipoprotein sorting defects, because such mutations would also interfere with pullulanase sorting. Therefore, we sought a completely different selection procedure for lipoprotein sorting defects that is based on artificial lipoproteins, which have been shown to follow the +2 rule (Ghrayeb and Inouye, 1984).
Maltose-binding protein (MalE), which is absolutely required in E. coli for uptake of and growth on maltose (Boos and Shuman, 1998), is made as a precursor whose signal peptide is cleaved by leader peptidase to release the mature protein into the periplasm. The selection procedure we developed is based on the observation that the abolition of signal peptide processing caused MalE to remain anchored in the cytoplasmic membrane without abolishing its ability to function in maltose transport (Fikes and Bassford, 1987; Dean et al., 1989; Fikes et al., 1990). We predicted that converting MalE into a cytoplasmic membrane lipoprotein would also allow it to function in maltose transport, whereas converting it into an outer membrane lipoprotein might render it non-functional. Both predictions were found to be correct. Therefore, we were able to select for restoration of growth on maltose in an E. coli K-12 strain in which MalE was tethered to the outer membrane by a fatty-acylated N-terminus. This paper reports the characterization of some of the mutants obtained.
Construction of fatty-acylated MalE
The lipoproteins used are encoded by pulA′–malE gene fusions whose product comprises the signal peptide and the first four amino acids of PulA (CDNS) followed by the mature region of MalE (KIEE). This protein (CD-MalE) is predicted to be fatty acylated and processed by lipoprotein signal peptidase. SDS–PAGE and immunoblotting of E. coli K-12 cells producing this protein with anti-MalE serum revealed that it migrated as a 47 kDa band (Fig. 1). The protein could be specifically labelled with radioactive palmitate (Fig. 1), as expected for a lipoprotein. As predicted (see Introduction), the presence of CD-MalE restored the ability of strains with chromosomal malE deletions to produce red colonies on MacConkey maltose medium and to grow on minimal medium with maltose as the sole carbon source. Therefore, CD-MalE is a functional variant of MalE.
A further malE gene fusion coding for CS-MalE was constructed in exactly the same way, except that the D+2 residue derived from PulA was converted into a serine. A pUC18-type plasmid bearing this gene fusion also complemented the chromosomal malE deletion after IPTG induction, indicating that CS-MalE is also a functional variant of MalE. CS-MalE was found to be fatty acylated and appeared as a 47 kDa band when examined by SDS–PAGE and immunoblotting with anti-MalE (Fig. 1).
The functionality of CS-MalE is plasmid copy number dependent
In some experiments, CS-MalE was able to replace normal MalE in maltose transport at 37°C, even without induction, which was contrary to our expectations (see Introduction). However, when the copy number of the gene fusions was reduced, by using strain LH1018 as host for pUC-based plasmids at 37°C (the pcn mutation reduces the copy number of pUC-type plasmids; Lopilato et al., 1986), by subcloning the gene fusions into lower copy number vectors such as pHSG575 (Takeshita et al., 1987) or by lowering the incubation temperature from 37°C to 30°C, complementation of ΔmalE444 was obtained with CS-MalE only upon IPTG induction. On the contrary, the gene fusion encoding CD-MalE complemented ΔmalE444 even when carried by low-copy-number vectors and uninduced. This differential phenotype, which was not the result of differences in the levels of the two proteins (not shown), permitted us to select mutations that allowed cells producing low levels of CS-MalE to grow on minimal maltose medium (see below).
Localization of fatty-acylated MalE variants
The subcellular location of CD-MalE and CS-MalE was determined by floating membranes obtained from French press-disrupted sphaeroplasts through a centrifuged sucrose gradient (see Experimental procedures). CD-MalE was found to float to near the top of these gradients, together with cytoplasmic membrane marker proteins, whereas CS-MalE was found in denser fractions closer to the bottom of the tubes, together with outer membrane marker proteins (Fig. 2). Thus, the fatty-acylated proteins fractionate according to the ‘+2 rule’ derived from studies of other lipoproteins (see Introduction). The same result was obtained independently of copy number and induction by IPTG (which caused an ≈ 10-fold increase in yields). Therefore, the processes of cytoplasmic membrane retention and outer membrane sorting are not readily saturable, at least within the limits of production of these proteins (estimated to reach a maximum of 1% of total cell proteins on the basis of the amounts detected by Coomassie blue staining of SDS–PAGE gels; not shown).
These fractionation data are at variance with previous studies from our laboratory suggesting that D+2 lipoproteins might be located in a unique domain of the cell envelope with a density intermediate between those of the cytoplasmic and outer membranes (Bouvier et al., 1991; Pugsley and Kornacker, 1991; Poquet et al., 1993). In most of the experiments on which this interpretation was based, full-length PulA was used as a ‘typical’ D+2 lipoprotein. To determine whether the choice of reporter protein or changes in fractionation procedure between the previous and the present experiments accounted for the different results, membranes from cells producing full-length PulA encoded by pCHAP3013 were fractionated in the same way as in Fig. 2. Several differences were noted between the results obtained with CD-MalE and PulA. First, a large amount (> 50%) of PulA, but no CD-MalE, was recovered in the soluble fraction. Secondly, very little of the ‘membrane-associated’ PulA floated up the gradient with cytoplasmic membrane-derived vesicles. Instead, it formed a broad peak covering almost the entire gradient, with a maximum concentration close to the position in the gradient at which the outer membranes were found. This result is similar to that repeatedly obtained in previous studies (Pugsley and Kornacker, 1991; Poquet et al., 1993). The difference between CD-MalE and PulA could be that the large size of PulA causes it to slough off from the membrane during fractionation.
Two additional methods were used to determine whether the sucrose gradient fractionation data accurately reflected the location of CD-MalE and CS-MalE. First, the MalE segment of the two proteins was replaced by CelZ, a secretable cellulase normally produced by Erwinia chrysanthemi (Py et al., 1991a). CD-CelZ and CS-CelZ both followed the +2 rule for lipoprotein sorting according to sucrose gradient fractionation (not shown), indicating that CelZ, like MalE, is a neutral reporter protein. CelZ was chosen for these experiments because it binds avidly to its ligand, insoluble cellulose (Avicel) (Py et al., 1991b), even when associated with membrane vesicles or fragments. This enabled us to mix crude membrane preparations from cells producing CD-CelZ or CS-CelZ with Avicel and examine the other membrane-derived proteins that were bound by SDS–PAGE. The proteins bound together with CS-CelZ gave a typical outer membrane pattern, whereas the pattern obtained with CD-CelZ appeared to be a mixture of outer and cytoplasmic membrane proteins (Fig. 3). This is probably because over 80% of the cytoplasmic membrane vesicles prepared by the French press method used are in the inverted orientation (Poolman et al., 1983; Futai, 1984). CD-CelZ in these vesicles is inaccessible to Avicel because it is exposed on the luminal (periplasmic) face. Thus, recovery of cytoplasmic membrane-associated CD-CelZ is suboptimal, and the method overestimates the level of contamination of this membrane by the outer membrane. Other cell disruption methods increased membrane fusion and could not be used.
Another method used to determine the location of CD- and CS-lipoproteins was immunoelectron microscopy. As shown in Fig. 4, only CS-MalE was found associated with the outer membrane of plasmolysed cells in which the outer and cytoplasmic membranes are separated, whereas CD-MalE was located in the cytoplasmic membrane. These data confirm that the sucrose gradient fractionation accurately indicates the positions of CS-MalE and CD-MalE in the cell envelope.
Mal+ mutants of strains producing CS-MalE
To select for Mal+ mutants, E. coli K-12 strain HS2019F′ producing low levels of CS-MalE (pCHAP4544) was mutagenized with nitrosoguanidine and then plated at 30°C on minimal maltose agar, which does not normally support growth of this strain. Cells in colonies that grew were re-isolated without selection and then retested. Cells from clones that continued to give the Mal+ phenotype on minimal maltose agar were tested for their ability to give Mal+ (red) colonies on MacConkey maltose medium, for the presence of unaltered levels of MalE of the same size as CS-MalE and for the presence of unaltered plasmid copy number. All clones that did not conform to one or more of these criteria were eliminated.
In the next step, plasmids extracted from the clones were used to retransform unmutagenized strain HS2019F′. This step revealed the existence of two classes of mutants. The plasmid DNA from the single class I mutant was able to complement the ΔmalE444 mutation, whereas plasmid DNA from 20 class II clones failed to complement this mutation. Therefore, we presumed that the class I mutation is an intragenic suppressor of the mutation introduced to convert CD into CS, whereas class II mutants carry chromosomal mutations. To check the latter, the plasmid was ‘chased’ from all original clones by transformation with pGP1-2, which confers spectinomycin (Sp) resistance and is incompatible with pCHAP4544, which confers chloramphenicol (Cm) resistance. The SpR CmS clones obtained, which were all Mal−, were then transformed with a fresh, unmutagenized stock of pCHAP-4544. CmR SpS clones derived from the class II mutants that had acquired pCHAP4544 at the expense of pGP1-2 were Mal+. This confirmed that class II mutations are chromosomal.
Location of MalE in the mutant strains
Sucrose gradient fractionation was used to determine the location of MalE in the different mutants. In the class I mutant, which was examined several times, 50–90% of MalE was located in the cytoplasmic membrane (Fig. 2). In contrast, the majority of MalE was found to remain associated with the outer membrane in the class II mutants, with only small and variable amounts of MalE in the cytoplasmic membrane (see example in Fig. 2). To check for possible specific effects on MalE, immunoblots of samples from the gradients were probed with antibodies against the outer membrane lipoprotein NlpD (Lange and Hengge-Aronis, 1994). These experiments revealed that the proportion of NlpD associated with the outer membrane did not change in some of the mutants, which were therefore not examined further. In other mutants, the proportion of NlpD recovered in the cytoplasmic membrane fraction was similar to the proportion of CS-MalE that was displaced to this location. Similar results were obtained for another outer membrane protein, Pal (not shown), but the distribution of non-lipoprotein markers (outer membrane proteins OmpA, OmpF, OmpC and LamB) was not affected.
The class I mutant is not a revertant
Sequencing of plasmid DNA from the class I mutant revealed that the S+2 codon (TCT) had been converted into a phenylalanine codon (TTT). No other mutation was detected in the rest of the malE gene, indicating that F+2 alone causes sorting of lipoproteins to the cytoplasmic membrane. To check the verity of this conclusion, we used polymerase chain reaction (PCR) to construct a new clone of CF-MalE. This clone also complemented the ΔmalE444 mutation, and the CF-MalE protein was found in high amounts in the cytoplasmic membrane (Fig. 5).
Other amino acids at position +2 can cause cytoplasmic membrane retention of lipoprotein
The results described above prompted us to examine the effects of all possible amino acids at the +2 position in the model MalE lipoprotein. As shown in Fig. 5, clones with the amino acids D, E, F, G, H, K, N, P, R, W and Y in position +2 were able to complement ΔmalE444 to at least some extent, whereas clones with amino acids A, C, I, L, M, Q, S, T and V were not. The correlation between the presence of relatively large amounts of MalE in the cytoplasmic membrane and the maltose phenotype is strongest for amino acids D, F, P, W and Y. Indeed, very little CW-MalE was found in the outer membrane, indicating that W+2 is at least as good as D+2 at causing cytoplasmic membrane retention. On the other hand, CE-MalE was strongly Mal+ but barely detectable in the cytoplasmic membrane fraction. Likewise, CN-MalE, CK-MalE and CH-MalE caused a modest Mal+ phenotype despite their total absence from the cytoplasmic membrane.
CK-MalE and CR-MalE are not lipoproteins
One explanation for the retention of some of the MalE variants in the cytoplasmic membrane could be that they are not fatty acylated or processed. As lipoproteins must receive the thioester-linked diacylglyceride before they can be processed by lipoprotein signal peptidase, we checked fatty acylation of the model lipoproteins first. As shown in Fig. 6, all clones except CK-MalE and CR-MalE were fatty acylated, indicating that the first step in their processing has occurred normally. However, the incorporation of palmitate does not indicate that the signal peptide is processed. Therefore, we examined processing of these variants in the presence and absence of globomycin, which inhibits lipoprotein signal peptidase (Inukai et al., 1978). As shown in Fig. 7, globomycin caused the accumulation of precursor forms of CE-MalE, CF-MalE, CG-MalE, CH-MalE, CN-MalE, CP-MalE, CW-MalE and CY-MalE that were larger than the single polypeptide seen in untreated cultures. Therefore, retention of these proteins in the cytoplasmic membrane is not caused by failure to process the signal peptide. CP-MalE seems to be particularly sensitive to globomycin treatment, possibly because of reduced affinity for lipoprotein signal peptidase, whereas precursors of CG-MalE, CH-MalE and CN-MalE were barely detectable after globomycin treatment (Fig. 7).
The situation with CK-MalE and CR-MalE was unexpected, because they were both found associated with the outer membrane, which should not occur if processing of the signal peptide has not occurred (note that immunoblotting of CK-MalE and CR-MalE revealed the presence of near-normal levels of both proteins in the samples shown in Fig. 6). In line with the data showing minimal fatty acylation, two closely spaced bands were observed with CK-MalE and CR-MalE, irrespective of the presence of globomycin, indicating inefficient translocation or poor processing of these proteins by lipoprotein signal peptidase.
CK-MalE and CR-MalE were clearly not aggregated, as they floated up the sucrose gradients with the outer membranes. To determine whether they were tightly associated with the membranes, they were washed with 6 M urea (Matsuyama et al., 1995). Small amounts of both proteins were released from the membranes by the urea, but the proportion of these proteins released was no greater than with the control protein CS-MalE. We conclude that CK-MalE and CR-MalE are firmly associated with the outer membrane, despite their incomplete processing. Interestingly, previous studies (Gennity and Inouye, 1991) have also shown that K+2 was unable to cause cytoplasmic membrane retention of lipoproteins.
Amino acids at position +2 in E. coli K-12 lipoproteins
It is widely accepted that the only amino acid that can cause lipoprotein retention in the E. coli cytoplasmic membrane is aspartate. The data presented above show that this is not the case. Therefore, we screened the sequenced genome of E. coli K-12 for known or putative lipoproteins and examined the amino acid predicted to be present at position +2. As shown in Table 1, most of the amino acids that cause partial or nearly complete retention of lipoproteins in the cytoplasmic membrane except aspartate (F, G, P, W and Y) are rarely, if ever, found at position +2 in known or predicted lipoproteins. Likewise, R and K, which cause lipoprotein translocation/processing defects (Fig. 7), are not found at position +2 in known E. coli lipoproteins, although K+2 was identified in two putative lipoproteins. It is not possible to comment on the determined location of known lipoproteins because the methods used (usually selective detergent extraction) are different from those used here and are inherently less accurate.
Table 1. . Amino acids at position +2 in the mature part of known and putative lipoproteins encoded by the E. coli chromosome. a. Based on Fig. 5. OM, outer membrane; CM, cytoplasmic membranes; OM-CM, both membranes.
Chromosomal mutations causing low-level retention of CS-MalE in the cytoplasmic membrane do not map to lolA or lolB
In order to see whether representative class II mutants carried mutations in lolA or lolB, we attempted to transduce out the mutations (and thereby recover a Mal− phenotype) using P1 phage grown on E. coli strains with transposon Tn10 inserted close to lolA (strain CAG12094) or lolB (strains CAG12016, CAG12169 and CAG18478; Table 2). Only clones with a relatively strong Mal+ phenotype could be tested. In none of the four independent mutants tested were we able to obtain Mal− co-transductants with Tn10, indicating that the mutations in these four strains, at least, are not in lolA or lolB. Attempts to characterize these mutants further failed because we were not able to transduce the chromosomal mutation into other strain backgrounds.
Table 2. . E. coli K-12 strains.
The genetic selection reported here was designed to identify unknown factors in lipoprotein sorting in E. coli. The unexpected identification of phenylalanine as an alternative to aspartate for cytoplasmic membrane lipoprotein retention led us to reappraise the +2 rule for lipoprotein sorting in E. coli. The fact that phenylalanine and many of the other amino acids that cause detectable levels of cytoplasmic membrane retention (Fig. 5) share no structural resemblance with aspartate means that recognition by LolA, the periplasmic lipoprotein shuttle protein, must be based on features other than failure to recognize the fatty acid–cysteine–aspartate motif. It is presumably important that the bacteria do not use ‘ambiguous’ lipoprotein sorting signals such as G or Y (Fig. 5; Table 1[link]), unless the lipoproteins concerned must be located in both membranes simultaneously. Earlier studies suggested that the function of the cytoplasmic membrane retention signal D+2 was largely context independent (Gennity and Inouye, 1991; Gennity et al., 1992). This might not be the case with the alternative cytoplasmic membrane retention signals F, G, P, W and Y, whose function might be influenced by adjacent amino acids derived, in this case, from PulA and MalE. This question can again be addressed by genetic methods.
The maltose phenotype proved to be a sensitive indicator of MalE lipoprotein localization. However, the correlation between Mal+ phenotype and location in the cytoplasmic membrane is not absolute. The situation with R+2 and K+2 is confused by the fact that processing of these proteins was apparently abnormal. As their signal peptides are apparently retained, they should remain anchored in the cytoplasmic membrane, which is not the case. The fact that E+2, and to a lesser extent H+2 and N+2, produced a Mal+ phenotype, despite complete or almost complete sorting to the outer membrane (Fig. 5), is more puzzling. One possible explanation for the Mal+ phenotype of bacteria producing these variants is that they are subject to limited degradation that releases small amounts of a periplasmic form that can function in maltose transport. However, if this is the case, the periplasmic form is undetectable in the soluble cell fraction (not shown). It should be recalled, moreover, that the maltose phenotype is scored on plates, whereas the fractionation data were obtained with bacteria grown in rich liquid cultures. It is not possible to test whether growth conditions influence lipoprotein sorting using these assays. It is worth noting that E and N are the two amino acids with maximum resemblance to the canonical cytoplasmic membrane retention signal D and that both amino acids were previously found to cause outer membrane sorting of lipoproteins (Gennity and Inouye, 1991; Pugsley and Kornacker, 1991; Gennity et al., 1992).
Unfortunately, the potentially interesting chromosomal mutations in class II mutants caused only weak Mal+ phenotypes, presumably because only a small proportion of CS-MalE was retained in the cytoplasmic membrane (Fig. 2). Furthermore, we were unsuccessful in attempts to transduce these mutations into other strain backgrounds. Possible explanations are that the phenotype results from two or more mutations or that the Mal phenotype is sensitive to subtle variations in plasmid copy number, gene expression levels or other factors. The probable explanation for our failure to isolate tighter mutations is that they would cause considerable amounts of the major outer membrane lipoprotein, Lpp, to be retained in the cytoplasmic membrane, where it cannot function normally and is potentially lethal (Yakushi et al., 1997). Presumably, the small amounts of Lpp retained in the cytoplasmic membrane in the class II mutants are compatible with viability. If this selection were to be used again, it would have to include provision for the identification of conditional mutations and should be performed in strains lacking Lpp, which are leaky but viable (Tajima et al., 1998). Such an approach might be useful because at least one gene coding for a component of the lipoprotein sorting machinery, the putative cytoplasmic ATPase (Yakushi et al., 1998), has not been identified. Moreover, the use of different selection procedures might allow further ‘lipoprotein sorting’ genes to be identified and could resolve the unanswered question of whether lipoprotein retention in the cytoplasmic membrane occurs by default or involves a specific machinery separate from that encoded by the known lol genes. We are currently examining several novel natural lipoproteins that have known functions and that confer a selectable phenotype when anchored to the cytoplasmic or outer membranes in E. coli.
The only E. coli lipoprotein whose fatty acid content has been determined accurately is the major outer membrane lipoprotein Lpp (Hantke and Braun, 1973). Outer membrane lipoproteins from other bacteria have the same diacylglyceride and N-acylation that were demonstrated in Lpp (Bouchon et al., 1997). Moreover, crude analyses of cytoplasmic membrane-derived lipoproteins suggest that their fatty acid components are similar to those found in Lpp (Pugsley et al., 1986; Yamaguchi et al., 1988). Nevertheless, it remains possible that cytoplasmic membrane retention results from a failure to incorporate the third, amine-linked fatty acid onto lipoproteins with a cytoplasmic membrane retention signal or to the incorporation of different fatty acids into one or more of the three possible positions. Interestingly, none of the putative lipoproteins in Gram-positive bacteria, which have only a cytoplasmic membrane, have an aspartate at position +2. Furthermore, analysis of the genome of the Gram-positive bacterium, Bacillus subtilis, does not reveal the presence of a gene homologous to that encoding the Salmonella typhimurium lipoprotein aminoacyl transferase (Gupta et al., 1993), raising the possibility that none of the lipoproteins in this bacterium have an N-linked fatty acid (J.-M. van Dijl, personal communication). However, at least one study has reported apparently complete fatty acylation in a lipoprotein from another Gram-positive organism, Staphylococcus aureus (Navarre et al., 1996). As expected, production in E. coli of lipoproteins normally present in Gram-positive bacteria results in their full fatty acylation and sorting to the outer membrane (Lai et al., 1981; Sarvas and Palva, 1983).
Plasmids, strains and growth conditions
Strains of E. coli K-12 are listed in Table 2, and plasmids carrying genes coding for model lipoproteins or PulA under the control of the lacZ promoter are listed in Table 3. Bacteria were grown at 30°C or 37°C in rich medium L, minimal medium M63 containing 0.4% maltose or MacConkey maltose medium. Media were solidified, where appropriate, with 1.6% agar (Miller, 1972). Antibiotics were used at the following concentrations: chloramphenicol, 25 μg ml−1; kanamycin, 50 μg ml−1 (for plasmids) or 20 μg ml−1 (for strains with F′lacI q::km); spectinomycin, 25 μg ml−1; tetracycline, 16 μg ml−1.
Table 3. . Plasmids (Sauvonnet et al., 1995). All cloned genes were under vector-derived lacZp control. Km, kanamycin; Cm, chloramphenicol; Sp, spectinomycin.
Fatty-acylated MalE and CelZ
In order to create model lipoproteins, DNA coding for the lipoprotein signal sequence and the first four codons of the pulA gene, together with 100 bp of upstream sequence but lacking a functional promoter, was amplified and cloned into a high-copy-number pUC18-type vector. For this purpose, an EcoR1 site was created upstream from the pulA start codon and XbaI or HindIII sites downstream from the fourth codon. Different 3′ primers varied at the codon for amino acid +2 and had an Eco47III site immediately after the fourth codon [e.g. for S+2 codon TCT (bold), non-coding strand: 5′-CATCGTCTAGATTATTAAGCGCTGTTAGAGCAGCCGCT AAGTA-3′ with XbaI and Eco47III sites in italics]. DNA fragments coding for the mature (periplasmic) parts of MalE or CelZ (Sauvonnet et al., 1995) were cloned in frame at the Eco47III site. The resulting gene fusions were subcloned as EcoRI–XbaI or EcoRI–HindIII fragments into different vectors (Table 3).
Mutagenesis was performed on cells of strain HS2019F′ carrying pCHAP4544 essentially as described previously (Sambrook et al., 1989) using 150 μg ml−1 1-methyl-3-nitro-nitrosoguanidine. The mutagenized bacteria were plated on minimal maltose medium containing chloramphenicol and then incubated at 30°C to allow colonies to develop. Isolated colonies were purified out on L agar containing chloramphenicol and then retested for growth on minimal maltose agar and for maltose fermentation on MacConkey maltose agar.
Transduction with bacteriophage P1 was performed essentially as described by Miller (1972). Transductants carrying the transposon Tn10 were selected on medium containing tetracycline and screened for loss of the Mal+ phenotype caused by the initial mutation.
Cells from exponentially growing cultures were lysed with or without prior conversion to sphaeroplasts (Brockman and Heppel, 1968) by passage through a French press at 500 or 1200 bar respectively. Membranes were collected by centrifugation and loaded on the bottom of sucrose step gradients essentially as described previously (Possot et al., 1999) and centrifuged for 36 h at 220 000 × g. Fractions collected from the gradients were analysed by SDS–PAGE and immunoblotting.
Affinity purification on Avicel
Crude membranes prepared as above were subjected to Avicel affinity chromatography essentially as described previously (Py et al., 1991b). Membranes were resuspended in 25 mM HEPES (pH 7.5) + 1 mM MgSO4 and mixed with Avicel for 1 h at 4°C. The cellulose was then washed extensively with the same buffer, and bound proteins were eluted with 2% SDS, concentrated by TCA precipitation and examined by SDS–PAGE.
Immunogold labelling and electron microscopy
Cells were first plasmolysed by brief exposure to 20% sucrose in buffer containing 25 mM Tris (pH 7.5) and 1 mM EDTA. Pelleted bacteria were fixed with 4% formaldehyde (freshly made from paraformaldehyde) in 0.1 M cacodylate buffer (pH 7.4) and 20% sucrose for 1 h at 4°C. The cell pellets were rinsed with cacodylate buffer and then treated with 0.5% aqueous uranyl acetate solution followed by a final rinse in distilled water. Bacteria were embedded in 2% agarose (type IX; Sigma). Small blocks of agarose were embedded in Lowicryl HM20 at −50°C by the progressive lowering temperature (PLT) method. Ultrathin sections were collected on Formvar-coated nickel grids. Sections were then incubated in the following solutions: PBS containing 50 mM NH4Cl (10 min); PBS containing 1% bovine serum albumin (BSA) and 1% normal goat serum (NGS) (10 min); rabbit polyclonal anti-MalE antiserum diluted 1:100 in PBS–BSA–NGS (1 h); PBS containing 0.1% BSA (two washes of 5 min each); anti-rabbit immunoglobulin gold conjugate (10 nm particles; British Biocell Laboratories) diluted 1:20 in PBS containing 0.01% fish skin gelatine (Sigma) for 45 min. Sections were then washed once in PBS and three times in distilled water, fixed for 2 min with 1% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.4) and finally rinsed with distilled water and dried. Optional counterstaining was performed by treating the sections with 2% aqueous uranyl acetate solution for 20 min followed by a 3 min incubation in Millonig's lead tartrate solution. Specimens were examined with a Philips CM12 electron microscope operating under standard conditions.
IPTG-induced bacteria producing the model lipoproteins were labelled with 20 μCi ml−1 [3H]-palmitate for 4 h (OD600 of 1.5) in L broth, resuspended in SDS–PAGE sample buffer and examined by SDS–PAGE. The gels were first fixed and stained with Coomassie blue, soaked in Amplify (Amersham), dried and exposed to X-ray film for 48 h.
Inhibition of lipoprotein signal peptidase with globomycin
Bacteria were grown to an OD600 of 0.8 in L broth before being induced with 2 mM IPTG to increase production of the MalE lipoproteins. Globomycin (50 μg ml−1) was added at the same time, and incubation continued for 45 min at 37°C. Bacteria were harvested by centrifugation, resuspended in SDS sample buffer and examined by SDS–PAGE and immunoblotting.
SDS–PAGE and immunoblotting
SDS–PAGE was carried out essentially as described previously (Sambrook et al., 1989). After electrophoresis, separated proteins were transferred onto nitrocellulose sheets (Amersham) that were incubated first with primary polyclonal antiserum and then with horseradish peroxidase-labelled donkey anti-rabbit IgG (Amersham). Immunoblots were developed by enhanced chemiluminescence (ECL; Amersham).
We are grateful to Howard Shuman and Olivera Francetic for strains carrying ΔmalE, Jean-Michel Betton for antibodies against MalE, Fred Barras for antibodies against CelZ, Regine Hengge-Aronis for antibodies against NlpD, David Clarke for antibodies against DjlA, Masatoshi Inukai for globomycin, Nathalie Sauvonnet for constructing the first malE and celZ gene fusions, Jan Marteen van Dijl for discussions on fatty acylation in Gram-positive bacteria and for help in identifying lipoproteins in genome data banks, Malcolm Page for his interest and to all members of the secretion laboratory for their constant support. This work was financed by a Hoffmann-La Roche studentship to A.S. and by the European Union (Training and Mobility in Research grant number FMRX-CT96-0004) and a French Research Ministry grant (Programme fondamentale en Microbiologie et Maladies infectieuses et parasitaires).