The cytoplasmic membrane proteins of bacterial binding protein‐dependent transporters belong to the superfamily of ABC transporters. The hydrophobic proteins display a conserved, at least 20 amino acid EAA‐‐‐G‐‐‐‐‐‐‐‐‐I‐LP region exposed in the cytosol, the EAA region. We mutagenized the EAA regions of MalF and MalG proteins of the Escherichia coli maltose transport system. Substitutions at the same positions in MalF and MalG have different phenotypes, indicating that EAA regions do not act symmetrically. Mutations in malG or malF that slightly affect or do not affect transport, determine a completely defective phenotype when present together. This suggests that EAA regions of MalF and MalG may interact during transport. Maltose‐negative mutants fall into two categories with respect to the cellular localization of the MalK ATPase: in the first, MalK is membrane‐bound, as in wild‐type strains, while in the second, it is cytosolic, as in strains deleted in the malF and malG genes. From maltose‐negative mutants of the two categories, we isolated suppressor mutations within malK that restore transport. They map mainly in the putative helical domain of MalK, suggesting that EAA regions may constitute a recognition site for the ABC ATPase helical domain.
In bacteria, high affinity uptake of nutrients is achieved by complex substrate binding protein‐dependent transport systems. These multi‐component permeases consist of one periplasmic substrate‐binding protein, two hydrophobic cytoplasmic membrane proteins and two subunits of a peripheral cytoplasmic membrane protein that displays ATP‐binding motifs. The cytoplasmic membrane proteins form a complex with a stoichiometry of two ATP‐binding subunits and two integral membrane subunits that mediates the ATP‐dependent translocation of the substrates into the cytoplasm (Bishop et al., 1989; Davidson and Nikaido, 1991). Such a characteristic organization is an intrinsic property of the superfamily of ABC (ATP Binding Cassette) transporters (Higgins, 1992) or Traffic ATPases (Ames and Lecar, 1992) widely distributed among living organisms.
The maltose transport system of Escherichia coli is one of the most studied bacterial ABC transporters and its functional mechanism is known to some extent (see Boos et al., 1996 for a review). Genetic, and more recently biochemical approaches, have led to a functional model that is summarized below. Maltose and maltodextrins enter the periplasm by facilitated diffusion through a specific outer membrane porin coded for by the lamB gene, which is specifically required for maltose transport at sub‐micromolar concentrations and for maltodextrins at all concentrations (Szmelcman and Hofnung, 1975). In the periplasm, maltose‐binding protein (MBP) binds substrates at high affinity (KD = 1 μM). Upon binding of substrates, MBP undergoes a conformational change (Szmelcman et al., 1976) and interacts with a cytoplasmic membrane complex made of MalF, MalG and two subunits of MalK (MalFGK2), and very likely with hydrophobic membrane proteins MalF and MalG (Hor and Shuman, 1993) that carry also a substrate binding site(s) (Treptow and Shuman, 1988). Recent results obtained with a proteoliposome‐reconstituted transport system suggest that MBP transmits through MalF and MalG a signal to MalK allowing it to hydrolyze ATP (Davidson et al., 1992).
An important step in understanding maltose transport mechanism is the analysis of protein–protein interactions. The existence of mutations allowing maltose transport in the absence of MBP led to the identification of sites on MalF and MalG thought to be important for the recognition of MBP (Covitz et al., 1994). We are interested in the identification of regions of hydrophobic membrane proteins involved in the interaction with the MalK ATPase. By comparing the sequences of hydrophobic membrane proteins from several binding protein‐dependent transporters, we identified a conserved sequence (EAA‐‐‐G‐‐‐‐‐‐‐‐‐I‐LP) located at a distance of ∼100 residues of their C‐terminal ends (Dassa and Hofnung, 1985b). This 20 amino acid sequence (called thereafter EAA region) is hydrophilic and lies in a loop facing the cytosol in all proteins of known topology (Saurin et al., 1994). It has been proposed that it might constitute a recognition site for a ligand or a partner common to binding protein‐dependent permeases (Dassa and Hofnung, 1985a), which could be the most conserved component in such permeases, namely the ATP‐binding proteins (Kerppola and Ames, 1992). This EAA region is probably of functional importance since several mutations have been characterized within that reduce or abolish transport in the maltose (Dassa, 1990, 1993), the iron (iii) hydroxamate (Köster and Böhm, 1992) and the phosphate (Webb et al., 1992) uptake systems.
To investigate the physiological relevance of the conserved EAA region, we generated substitution mutations in the corresponding regions of MalF and MalG. The most conserved residues were modified and we analyzed the consequences of these changes on maltose and maltodextrins uptake, and on the expression of the proteins. The membrane association of MalK was altered in some uptake‐defective mutants. We isolated suppressor mutations in malK that restore maltose uptake and MalK membrane association in these mutants. We discussed a model suggesting that EAA regions constitute a recognition site for the MalK subunits.
The experimental system
We constructed three plasmids to generate easily site‐directed mutations within EAA regions. Plasmids pTAZFQ and pTAZGQ carry the malF and malG genes respectively under the control of the tac promoter. Plasmid pTAZFGQ carrying these two genes, was constructed by inserting the BamHI–HindIII fragment from pTAZGQ into pTAZFQ. Each plasmid contains a copy of the lacIQ gene in order to control the expression of cloned genes (Figure 1). pTAZGQ, pTAZFQ and pTAZFGQ were able to complement mutations in PMED34 (MalG−), DHB4 (MalF−) and ED170 (MalF−G−) respectively. To evaluate the expression of cloned genes, particulate fractions were prepared from strain MM140 (ED170 transformed with pTAZFGQ), analyzed by SDS–PAGE, and proteins were revealed by immunoblotting. We compared this amount of protein with the level of expression of MalF and MalG in a total extract of the fully induced maltose transport positive strain MC4100. We found that a concentration of 10 μM isopropyl β‐d‐thiogalactopyranoside (IPTG) was enough to ensure the production of 110 and 78% of the induced chromosomal levels for the MalF and MalG proteins respectively, as judged from the scanning of Western blots. We analyzed the cellular localization of MalF and MalG (Figure 2). The proteins partitioned in a particulate fraction, from which they could be partially extracted by a buffer containing 1% Triton X‐100. As shown previously, a substantial amount of MalF and MalG is still not extracted by the detergent (Dassa, 1990; Panagiotidis et al., 1993). Triton X‐100 is a non‐ionic detergent known to specifically solubilize cytoplasmic membrane proteins (Diedrich et al., 1977), and has been used to show the membrane localization of the MalFGK2 complex. Thus, we interpret this result to indicate that, under our expression conditions, the bulk of the proteins is correctly assembled in the membrane.
Site‐directed mutagenesis of conserved residues in the EAA region
Starting from the assumption that residues are conserved because they play an important functional role, we mutagenized the most conserved residues in the EAA regions of MalF and MalG. We modified residues located at positions 1 (E), 3 (A or S), 7 (G) and 20 (P) of the consensus of EAA motifs (Saurin et al., 1994). In the absence of structural information on MalF and MalG, we made several changes at each position, modifying the size, the charge and the side chain mobility of the amino acids (Figure 3). The mutations in malF and malG were generated on plasmids pTAZFQ and pTAZGQ respectively, sequenced and eventually recombined on plasmid pTAZFGQ as described above. Each mutant was identified by a set of two symbols separated by a slash. The first symbol describes the changes effected in malF and the second those in malG. A non‐mutated gene was represented by a hyphen. Numbers between letters point to the relative position of the residue in the consensus sequence. For example, E1A/A3L means that the E residue in the EAA region of MalF was changed to A and that the A residue in position 3 of MalG was changed to L.
Twenty three plasmids were constructed, each carrying a substitution in MalG (nine mutants), in MalF (nine mutants) or in both proteins (five mutants). The growth rate of mutants induced by 10 μM IPTG in glycerol‐supplemented minimal medium and in ML medium (data not shown) was similar to that of strains MM141 (MalF−G−) and MM140 (MalF+G+). Hence, the expression of mutated proteins does not determine major growth defects in bacteria. Particulate fractions of mutants were separated on SDS–polyacrylamide gels and analyzed by immunoblotting (Figure 4 and Table I). Mutated proteins have electrophoretic mobilities similar to that of the wild‐type. There are variations in protein amounts with a 2‐ to 3‐fold difference between extreme values, with the exception of the MalG level in E1K/A3L, E1L/E1L and S3K/A3D mutants, where there is a 4‐fold increase as compared to MM140. This higher level of expression was not investigated further. Each mutant protein was correctly inserted in the cytoplasmic membrane, as suggested by the fact that it was extracted from the particulate fraction by Triton X‐100 as efficiently as wild‐type proteins (data not shown).
Phenotypes of mutants
Table I shows the characteristics of mutants. Their phenotypes were scored on MacConkey plates supplemented with maltose or maltodextrins. No Mal+ Dex− or Dex+ Mal− mutants were found. Phenotypes were characterized by measuring growth rates on maltose minimal medium and by determining maltose uptake rates at a substrate concentration of 4 μM. Mutants were classified according to decreasing maltose uptake rates. Class I mutants have initial velocities ranging from 240 to 150, doubling times ranging from 220 to 240 min and they form deep red colonies on MacConkey plates. Class II mutants have initial velocities ranging from 60 to 145, doubling times comprised between 230 and 300 min and they form red to pink colonies. Class III mutants form white colonies, do not grow on minimum maltose medium and do not transport maltose. These three parameters are in general well correlated, with few exceptions that could be explained by the difference of maltose concentration used in phenotypic determinations, in growth rate measurements and in transport assays. Since the mutated proteins are not found in significantly lesser amounts than the wild‐type proteins, the defects in maltose‐uptake‐deficient mutants are most probably not due to a reduced protein expression, nor to a defective localization.
E at position 1 is conserved in 50% of EAA sequences and is substituted by D in 12% sequences. Position 3 (A) is conserved in 75% of EAA sequences or is substituted by S, T or C in 20% sequences (Saurin et al., 1994). Our data show that any change affecting the E residue in MalF or in MalG leads to a reduction by 2‐ to 3‐fold in transport rates. Changes made to the third residue of the region (A or S) promote a reduction in transport rates that is related to the nature of the side chain of the replacing residue. A change to a charged residue (K or D) is more detrimental for transport than a change to a short side‐chain polar (S) residue. The G residue, invariant in EAA motifs, could be changed to A without incidence on transport. This is in agreement with earlier observations made on the FhuB protein (Köster and Böhm, 1992) and establishes that conserved EAA motifs play similar roles in binding protein‐dependent transport systems. The highly conserved P residue at position 20 is not essential for transport. This residue is predicted to be in a transmembrane segment in MalF and in MalG (Boyd and Beckwith, 1989; Dassa and Muir, 1993). It may be concluded that transport is not dependent on a peculiar property of this residue such as a cis–trans isomerization.
Substitutions made at the same positions in malG or in malF affect MalG more severely than MalF. At position 3, a substitution to D has a more pronounced effect in MalG (transport defective) than in MalF (reduced transport rate). At position 7, a change to P leads to a completely defective MalG protein while MalF is unaffected. All double mutants having a mutation in malF and in malG, except G7A/G7A, are unable to carry out transport (Class III). However, the single mutants from which the double mutants are made have a transport activity of ∼50–30% of the wild‐type (Class II).
Cytoplasmic localization of MalK in maltose–uptake‐defective mutants
The EAA region is located in a cytoplasmic hydrophilic loop in all hydrophobic membrane proteins with a known topology (Saurin et al., 1994). This location makes it a candidate for an interaction site with the ATPase subunit MalK. We reasoned that the phenotypes of our mutants might be explained by a defective or a non‐productive interaction with MalK. To test this hypothesis, we analyzed the cellular localization of MalK in Class II and III mutants. As in MM140 and its mutant derivatives the malK gene is chromosomal and under the control of the pmal promoter, maltose‐uptake‐negative mutants would have reduced to undetectable levels of protein MalK. To overcome this problem, we constructed plasmid pACYK that carries the malK gene under the control of the trc promoter. We found that in MM142 (ED170 transformed with pTAZFGQ and pACYK), a 10 μM IPTG concentration determines an amount of MalK 77% of that in maltose‐induced MC4100. In these conditions the amount of MalF is 90% of that in 10 μM IPTG‐induced MM140. The presence of pACYK does not alter the phenotype of MM142, since maltose uptake rates at 4 μM maltose and growth rates on maltose minimal medium are identical to those of MM140 (data not shown).
We transformed with pACYK a selected set of Class II and Class III mutants and we used strains MM142 and MM143 as controls. We found that particulate fractions of mutants and controls have relative amounts of MalF and MalG as described in Table II. These relative amounts are somewhat different to what was observed without pACYK but remain in the same range of a 2‐ to 3‐fold difference between extreme values.
The cellular localization of MalK was determined as described in Materials and methods (Figure 5). In MM142 (MalF+G+), 80% of MalK is found in the particulate fraction and it is efficiently extracted by Triton X‐100. In MM143 (MalF−G−), ∼60% of MalK is found in a soluble fraction. The residual MalK protein present in the particulate fraction is hardly extracted by Triton X‐100, but the Triton‐insoluble pellet could be solubilized by using 6 M urea. By comparison, the Triton‐insoluble fraction of MalG is insoluble in urea, consistent with its membrane association (data not shown). This suggests that the Triton‐insoluble MalK fraction might constitute aggregates that sediment artefactually with the membranes as has been previously observed by other groups (Reyes and Shuman, 1988; Walter et al., 1992a). Thus we interpret our results as follows: the amount of MalK found in the Triton X‐100 soluble fraction (Figure 5, Ts) represents the fraction correctly associated in the membrane, whereas the amount of protein found in the soluble fraction (Figure 5C) and in the Triton insoluble–urea soluble fraction (Figure 5, Us) represents the fraction of MalK which is not associated with the membrane and is consequently located in the cytoplasm. The Triton insoluble–urea insoluble fraction (Figure 5, Ui) contains membrane proteins not extractable by Triton X‐100 and most notably outer membrane proteins (data not shown).
We analyzed the cellular localization of MalK in strains MM144 (MalF−G+) and MM145 (MalF+G−). In both cases, the bulk of MalK was present in the soluble and in the Triton insoluble–urea soluble membrane fraction. The roles of MalF and MalG in the membrane association of MalK are controversial. Earlier observations showed that MalK was found in a cytoplasmic fraction in a mutant defective in the malG gene (Shuman et al., 1980). More recently, the same group showed that MalF alone was able to direct MalK to the cytoplasmic membrane (Panagiotidis et al., 1993). Our results support the idea that both MalF and MalG are needed for the correct membrane association of MalK.
The cellular localization of MalK in mutants was also evaluated. Two of the five Class III mutants tested, −/A3D and −/G7P, behaved as MalF−G− deficient strains although they expressed these proteins. One Class III mutant, E1K/A3L, and most of the Class II mutants, E1K/−, E1L/− and −/E1L behaved as the wild‐type strain. In mutants E1L/E1L, −/A3L of Class III and Class II respectively, MalK had an intermediate distribution. The fact that one of the Class III mutants behaved as the wild‐type strain suggests that there are at least two different types of defects introduced by mutations in the EAA region, one that results in the cytoplasmic localization of MalK and one that maintains the correct membrane association of MalK.
It is noticeable that the two mutations that completely dislocate MalK are substitutions in the EAA region of MalG. This is consistent with the observation that MalG is important for the constitution of the membrane complex. We asked whether substitutions located outside the EAA region might have the same effects. We analyzed the cellular localization of MalK in a strain carrying a derivative of pTAZFGQ, having a substitution D262K affecting a residue located in the last transmembrane segment of MalG (P.Lambert and E.Dassa, unpublished). Despite the fact that this strain is completely defective for maltose uptake, the cellular localization of MalK is not affected.
Mutations in the malK gene restore transport in maltose‐negative mutants
We then asked whether mutations of the malK gene would restore maltose uptake. Plasmid pACYK was mutagenized in vitro by hydroxylamine and was used to transform ED170 harboring maltose‐negative mutants −/G7P, E1K/A3L and E1L/E1L. These strains carry a chromosomal wild‐type copy of malK. We decided to isolate mutants in such a background for the following reasons. First, mutant −/G7P has a wild‐type allele of malF that might not function in the presence of a mutated allele of malK able to correct the defect present in malG. Second, as observed above, mutations in malF produce proteins able to support maltose transport to some extent by contrast with malG mutations that lead to completely defective proteins. We therefore suspected that in double mutants E1K/A3L and E1L/E1L, mutated MalF proteins would not accommodate malK mutations able to correct defects of malG mutations.
Nineteen maltose‐positive colonies, scored on MacConkey maltose medium, appeared after 2 days incubation on strains −/G7P, E1K/A3L and E1L/E1L. pACYK DNA was purified from these clones and transformed again into strain ED170 carrying the respective pTAZFGQ mutated plasmids. Eleven transformants became maltose positive in these conditions, indicating that the phenotype was indeed linked to the mutagenized pACYK plasmid. We replaced the wild‐type gene of pACYK by malK genes from the mutated plasmids and transformed the corresponding pTAZFGQ mutants with these constructs. Eight subclones were maltose positive, thereby demonstrating that a mutation(s) in malK was responsible for the phenotypic change. The three remaining subclones malK301, malK306 and malK308, all isolated in a E1L/E1L background, formed white colonies on MacConkey maltose plates. This is probably due to the presence of a mutation outside of malK in the original plasmid. To assess if such a mutation could be sufficient to restore maltose transport, we replaced the malK allele of these plasmids by a wild‐type copy from pACYK. The resulting phenotype was white. These experiments strongly suggest that a mutation in malK and an additional mutation(s) elsewhere on the plasmid are needed to determine a maltose‐positive phenotype in these clones. Such a mutation would probably raise the cellular level of the mutated MalK proteins, either by increasing the transcriptional level of the gene or the plasmid copy number. Indeed, we found that the expression of MalK in these three alleles was at least twice the normal level (Table III). DNA sequencing of the mutants (see below) revealed that no mutation was found in the ptrc promoter of malK.
We analyzed the cellular localization of MalK from revertants malK104, malK106, malK301, malK303, malK306 and malK308, selected with mutants −/G7P and E1L/E1L, and for which an aberrant localization of MalK was observed. In the presence of mutated MalF and MalG, the mutated MalK proteins were correctly associated to the membrane (data not shown).
Each malK allele was co‐transformed with pTAZFGQ plasmids bearing different uptake‐defective malF and/or malG mutants in strain ED170. Phenotypes were recorded on MacConkey maltose plates (Figure 6). All malF and malG mutants tested cannot ferment maltose in the presence of wild‐type malK. By contrast, all malK suppressors are able to function with wild‐type copies of malF and malG. There is no strict allele specificity but malK suppressor isolated in a double malF and malG mutant background is not able to suppress single mutant −/G7P and vice versa. To evaluate the influence of the chromosomal copy of malK in the ED170 background, we constructed strain ED169 that harbors a large malBΔ107 deletion. In this context, the malE gene was provided by plasmid pTLCME. This strain was co‐transformed with the same plasmid combinations as ED170. Phenotypes were recorded on MacConkey maltose plates. The phenotypes observed on ED169 are similar to those seen on ED170 (data not shown). This demonstrates that the chromosomal malK copy of ED170 is dispensable for the restoration of maltose uptake by malK suppressors and is consistent with the observation that all suppressors could function with wild‐type MalF and MalG. No suppressor can restore a maltose‐positive phenotype in the absence of pTLCME (data not shown). Consequently, these suppressors do not display an MBP‐independent phenotype suggesting that they do not constitutively hydrolyze ATP (Davidson et al., 1992).
Most MalK mutants able to relieve defects in EAA fall into the so‐called helical domain of ABC transporters
The 11 mutated malK genes were completely sequenced to determine if specific regions of MalK were important for suppressing the defects in MalF or MalG. The results of these experiments were reported in Table III. The −/G7P suppressors malK104 and malK106 carry a V149M and V149I substitution respectively. The E1K/A3L suppressors malK203, malK204 and malK205 have the same V117M substitution. malK201 carries three substitutions. We subcloned the mutations and we found that the phenotype was exclusively dependent on the N‐proximal mutation V117M. E1L/E1L suppressor malK303 has also a V117M substitution. Suppressors malK301, malK306 and malK308 were also sequenced, even if these changes in malK needed higher levels of expression to determine the phenotype. We identified two new suppressors V154I and M187I from malK301 and malK308 respectively. malK306 carries four substitutions and we were not able to subclone the different mutations because of the lack of appropriate restriction enzymes sites. All substitutions correspond to CG→TA transitions known to be specifically induced by hydroxylamine (Lawley, 1967). Mutations are located between codons 85 and 210 of MalK, and the same V117M mutation was found in five suppressors isolated from the E1K/A3L and the E1L/E1L backgrounds. This distribution argues against a random localization of reversion mutations, indicating that alteration of specific regions of MalK suppresses the defects in MalF−G− mutants.
Most mutations are located between codons 85 and 154 of malK. According to current secondary structure models of ABC ATPases, this area corresponds to Loop 2–3 (Hyde et al., 1990) or to the so‐called helical domain (Mimura et al., 1991) of ABC ATPases. This region constitutes the main difference between the predicted secondary structure models of ABC domains and the known structure of adenylate kinase. It has been proposed to be involved in the coupling of ATP hydrolysis to transport (Hyde et al., 1990) or to constitute a site of interaction between ABC ATPases and hydrophobic membrane proteins (Ames et al., 1990).
We analyzed by a genetic approach the functional role of EAA motifs, present in all hydrophobic membrane proteins of known binding protein‐dependent permeases. Twenty‐three mutants were constructed by site‐directed mutagenesis in MalF and in MalG. No change obviously affects the stability and the membrane location of mutated proteins. Five mutations severely impairing transport were characterized. Hence, this region appears to play an important physiological role.
Two single substitutions (A3D and G7P) in MalG completely abolish transport while identical changes made in MalF have no effect. This suggests that EAA regions, and perhaps the MalG and MalF proteins, do not function symmetrically whereas it is generally admitted that the two proteins play similar roles in the constitution of a substrate binding site (Shuman and Panagiotidis, 1993). We characterized mutants carrying mutations in both MalF and MalG that are completely defective in transport. The reduction of initial uptake rates in such double mutants is greater than that determined by the corresponding single mutations. This suggests that the EAA regions of MalF and MalG might cooperate or interact together, directly or indirectly, in the constitution of a functionally important site. A cooperative effect was observed in double EAA mutants in FhuB, a hydrophobic iron‐hydroxamate uptake protein having two EAA motifs (Köster and Böhm, 1992). Recently, cooperativity in ATP hydrolysis was demonstrated in the MalFGK2 complex (Davidson et al., 1996) and was interpreted as the consequence of interactions between MalK subunits. This effect was not detected on purified MalK (Walter et al., 1992a). We speculate that it might be mediated by MalF and MalG.
As we obtained two types of Class III mutants with regard to the membrane association of MalK, we tentatively propose that EAA regions participate in at least two functions. First, they might play a central role in the interactions that allow the membrane association of MalK and the constitution of a functional transporter since some EAA mutations result in a cytoplasmic localization of MalK. This result is strengthened by the fact that revertants were selected in malK, restoring maltose transport and membrane association. Recently, Böhm et al. (1996) concluded from a genetic study that the EAA region of protein FhuB could be involved in an interaction with another partner of the iron‐hydroxamate transport system. Second, EAA regions might participate to the constitution of a substrate binding site or a site that transduces a signal to MalK. The membrane association of MalK is normal in maltose‐uptake‐negative mutant E1K/A3L. As malK revertants can be selected from this mutant, we suspect that it is affected in a functional interaction with MalK.
These functions might be directly determined by EAA regions, but we cannot exclude indirect effects, since there is not a strict allele specificity in the restoration of transport which would be expected if mutated MalK was interacting separately with either MalF or MalG. Mutations in EAA loops might induce a conformational change in another region of the proteins more directly involved in the interaction with MalK. Alternatively, EAA mutations might affect the formation of the MalF–MalG heterodimer, resulting in the release of MalK in the cytoplasm. This would be consistent with the aggravated phenotypes of double mutants and in this hypothesis, malK suppressor mutations might induce a conformational change in mutated MalF and MalG proteins allowing them to interact properly. Eventually, EAA mutations might alter a functional site involving the three proteins MalF, MalG and MalK. Extensive genetic and biochemical investigations will be required to choose between these hypotheses. Nevertheless, we report here for the first time that in an ABC transporter, mutations in a cytoplasmic loop of hydrophobic membrane proteins are corrected by mutations in the ATPase.
We mapped malK suppressors on an alignment of some ABC ATPases for which functional information is available (Figure 7). Interestingly, reversion mutations located between codons 85 and 154 affect the so‐called helical domain of ABC ATPases, located between ATP‐binding motifs Walker A and Walker B (Mimura et al., 1991). Studies using a chimeric MalK–HisP protein showed that the MalK helical domain was critical for restoration of maltose uptake in strains lacking the malK gene (Schneider and Walter, 1991). Mutation A85M affects a region named ‘center’ containing a highly conserved Q residue and a predicted catalytic carboxylate residue (Yoshida and Amano, 1995). Mutations in this area in HisP and in MalK affect primarily transport while ATP‐binding is essentially unaffected (Shyamala et al., 1991; Walter et al., 1992b). This behaviour is consistent with a defective interaction between the ATPase and hydrophobic subunits. Mutation V117M falls in a variable region where an HlyB suppressor able to export a secretion‐defective HlyA mutant is located. Since most suppressors of such mutated HlyA protein have mutations in transmembrane segments of the HlyB exporter, it was concluded that mutations with similar phenotypes in the ATPase fall in regions interacting with transmembrane domains (Sheps et al., 1995). Mutations V149M, V149I and V154I are near the LSGGQ region, also called linker peptide. In this area, mutations altering the drug specificity of the Mdr transporter have been described and were interpreted as defective in an interaction between nucleotide‐binding domains and transmembrane domains (Beaudet and Gros, 1995). As most malK suppressors in our study map in the helical domain of MalK, our results give more direct evidence that this region is involved in interactions with the membrane embedded subunits.
Mutations of suppressors malK301 (M187I) and malK306 are more difficult to interpret since they need an increased level of protein expression to restore transport. This might indicate that the mutated regions, located outside the helical domain, are not directly involved in an interaction with the membrane components but are none the less undergoing conformation changes during these interactions. It has been proposed that in the RecA ATPase, the corresponding region is a ‘switch region’ involved in the propagation of a conformational change triggered by ATP hydrolysis (Yoshida and Amano, 1995). It is tempting to speculate that such a region in ABC transporters might also undergo conformational changes during the exchange of signals with hydrophobic membrane subunits.
In conclusion, our study provides evidence that cytoplasmic loops of the hydrophobic domains of an ABC transporter interact functionally with nucleotide binding domains, and particularly the so‐called helical domain. This finding might shed new light on the study of ABC transporters since such interactions have often been postulated. The fact that we obtained two types of negative mutants regarding the strength of association of MalK with mutated MalF–MalG suggests that these interactions may be important for the proper assembly and also for the correct function of the transporter.
Materials and methods
Strains were derivatives of E.coli K‐12 and are described in Table IV. Miscellaneous genetics techniques were performed as described (Miller, 1972). Strain PMED34 was constructed by introducing F′ lacIQ lacZ::Tn5 into strain PMED35. Strain ED169 was constructed by co‐transduction of ΔmalB107 with the metA+ gene of pop5218 into pop6952. Strain ED170 was constructed by transducing srl::Tn10recA of PMED35 into pop6484.
Media, chemicals and culture conditions
Strains carrying plasmids were grown at 30°C in order to minimize possible undesirable effects due to the elevated expression of mutant membrane proteins. Media were described in Miller (1972). The synthetic medium 63 was used for growth rate determinations, transport assays and in vivo pulse labeling. It was supplemented with 0.4% glycerol or maltose (Merck), 0.5 μg/ml thiamine and 0.4% Bacto‐Casamino Acids (Difco) or with a 0.004% mix of amino acids and nucleotides (Sigma) without methionine and cysteine for labeling experiments. Ampicillin (100 μg/ml), chloramphenicol (25 μg/ml), kanamycin (30 μg/ml) and tetracycline (10 μg/ml) were added when required. Rich medium LB was used for plasmid preparation and for mutant protein identification. Maltose and maltodextrins utilization was determined on MacConkey medium containing Maltose (2%) or 1% of the ‘malto‐oligosaccharide mixture’ from Pfanstiehl Laboratories Inc. (USA) respectively.
Recombinant DNA methods
Restriction enzymes were used according to the instructions of the suppliers. Site‐directed mutagenesis was performed as described (Kunkel, 1985). Mutagenic oligonucleotides are described in Table V. Single strand templates from different plasmids were prepared after infection of CJ236 with the M13MKO7 helper phage. Mutagenized plasmids were introduced into strain PMPM5 as described (Chung et al., 1989). DNA from randomly selected clones was sequenced and plasmids bearing the desired mutation were transformed into strain ED170. Nucleotide sequencing was performed by the Sequenase method (USB). α‐33P‐ and 35S‐nucleotides and [14C]maltose were obtained from Amersham (UK).
Plasmids are described in Figure 1 and were constructed as described below.
Plasmid pTAZGQ. An EcoRI–SspI fragment carrying the malG gene downstream of the ptac promoter and an EcoRI fragment bearing the lacIQ gene, both isolated from plasmid pDG (Dassa and Muir, 1993), were inserted successively into phagemid pTZ18‐R (Pharmacia Biotech).
Plasmid pTAZFQ. A fragment containing the malF gene downstream of the ptac promoter was amplified from plasmid pTFG (E.Dassa, unpublished) using two primers. One of the primers was designed to introduce a BamHI site 3′ to malF. The amplified fragment was digested with EcoRI and BamHI and inserted in pTZ18‐R. Then the EcoRI fragment bearing the lacIQ gene was inserted in the EcoRI site.
Plasmid pTAZFGQ. A BamHI–HindIII fragment from pTAZGQ carrying the malG gene was inserted into pTAZFQ.
Plasmid pACYK. This was constructed in two steps. First the malK gene from pKN101, a plasmid harbouring the malK and the lamB genes, was amplified using a set of primers that introduced a NcoI site at the ATG codon of malK and a HindIII site 3′ to malK. The fragment was digested by these two enzymes and inserted into plasmid pTrc99A (Pharmacia Biotech) giving rise to plasmid pTrK, where malK was under the control of the trc promoter. Second, an SphI–ScaI fragment from pTrK containing the lacIQ gene, the trc promoter and malK was inserted into plasmid pACYC184, generating plasmid pACYK compatible with the pTAZ series plasmids.
Plasmid pTLCME. This was obtained by cloning the malE gene under the control of the ptac promoter (Jean‐Michel Betton, unpublished) in pWSK129, a plasmid carrying the pSC101 origin of replication (Wang and Kushner, 1991).
Plasmid pACYK was treated with 0.8 M hydroxylamine (Sigma) at 70°C for 1 h using a procedure modified from Humphrey et al. (1976). Briefly, 20 μl of plasmid (1 μg/μl) was incubated with an equal volume of freshly prepared hydroxylamine stock solution (1.6 M hydroxylamine, 1 mM EDTA, pH 5.8) at 70°C. After 1 h incubation, the reaction was stopped by precipitating the DNA (70% ethanol, 0.1 M sodium acetate). Precipitated DNA was dissolved in 50 μl H2O and stored at −20°C until use. We used 2 μl of mutagenized DNA to transform strain ED170. Mutants were screened as described in Results.
Transport and enzyme assays
Transport assays were performed essentially as described (Szmelcman and Hofnung, 1975) with the following changes. Strains were grown in medium 63 supplemented with 0.4% glycerol, 0.4% maltose, 100 μg/ml ampicillin, 25 μg/ml chloramphenicol and 0.2% Bacto‐Casamino acids. At A600 = 0.2, 10 μM IPTG was added to induce expression of the mal genes on plasmids. Cells were harvested at A600 = 0.6 (∼3×108 cells/ml), washed twice and resuspended in 0.2 volume of medium 63 containing 30 μg/ml kanamycin. A mixture of [14C]maltose and non‐labelled maltose was added to 500 μl of suspension (final concentration 4 μM). Samples of 150 μl were withdrawn several times and quickly filtered through HAWP Millipore membrane filters of 0.45 μm pore size. Filters were thoroughly washed with medium 63, dried and counted.
Total extracts and cell fractionation
Strains were grown in 5 ml LB. At A600 = 0.2, 10 μM IPTG was added, and cells were grown to A600 = 1.0, chilled and recovered by centrifugation. Total cellular extracts were prepared as follows: a culture volume equivalent to 0.5 ml culture at A600 = 1.5 was centrifuged in a bench‐top microcentrifuge and resuspended in 20 μl 50 mM Tris–HCl pH 8, 150 mM NaCl. Twenty μl of double strength SDS–PAGE loading buffer were added and the homogenized mixture was boiled for 5 min in a water bath. For cell fractionation, bacteria were resuspended in 10 mM Tris–HCl pH 7.5 containing 0.75 M sucrose and lysed by addition of egg‐white lysozyme (1/8 volume of a 2 mg/ml solution in 0.1 M EDTA pH 7). After 30 min incubation at 0°C, spheroplasts were recovered by a 30 min centrifugation. Five hundred μl of 50 mM Tris–HCl pH 8, 150 mM NaCl, 2 mM EDTA were added to the spheroplast pellet and lysis was completed by three cycles of freezing and thawing. After addition of DNase I and a 15 min incubation at room temperature, cell debris was removed by low speed centrifugation. Particulate fractions were recovered from the extract by a 1 h centrifugation at 200 000 g. Triton X‐100 extracts were performed essentially as described by Diedrich et al. (1977). Triton X‐100 insoluble pellets were solubilized with freshly prepared 50 mM Tris–HCl pH 8, 150 mM NaCl, 6 M urea for 1 h at room temperature. Urea insoluble fractions were recovered by a 1 h centrifugation at 200 000 g. Each fraction was collected in equal volumes, and a subfraction was taken at each step of the fractionation/solubilization process for SDS–PAGE analysis.
In vivo pulse‐labeling
Strains ED169 transformed with the mutated pACYK were grown in 5 ml M63 medium containing glycerol, nucleotides and amino acids without methionine and cysteine. At A600 = 0.2, 10 μM IPTG was added, and cells were grown for 1 h. Cultures were normalized to an A600 of 0.4 and 200 μl of culture was labeled with 8 μCi of TRAN35S‐LABEL (ICN) for 10 min at room temperature. Two hundred μl of double strength SDS–PAGE loading buffer were added and the homogenized mixture was boiled 5 min in a water bath. Samples were run on an SDS–polyacrylamide gel, the gel was fixed 30 min at room temperature in 10% acetic acid and 25% isopropanol, dried and subjected to autoradiography.
Polyacrylamide gel analysis and immune detection of proteins
We used the BCA protein kit (Pierce) to determine the protein concentration of samples. Membrane extracts were solubilized in a buffer containing 1% SDS. The calibration curve was made by using bovine serum albumin diluted in the same buffer.
Protein samples were separated on SDS–polyacrylamide slab gels prepared as described (Lugtenberg et al., 1975). Immunodetection of MalF, MalG and MalK was performed by immunoblotting (Towbin et al., 1979; Dassa, 1990). Antibodies were specific to MalG (Dassa, 1990), MalF (gift of Beth Traxler) and MalK (gift of Erwin Schneider). Electrophoretically separated acellular extracts were transferred onto nitrocellulose membrane filters and probed successively with the specific antibody and with a horseradish peroxidase anti‐rabbit immunoglobulin conjugate (Bio‐Rad). Immune complexes were revealed by ECL Western Blotting Detection Reagents (Amersham, UK).
For quantitative analyses, gels or autoradiography were scanned with a MasterScan Interpretive densitometer (Scanalytics).
Note Added in proof
During the processing of the manuscript, reported mutations allowing LacK, an ABC ATP‐ase homologous to MalK, to interact with MalF and MalG. These mutations map into the helical domain of LacK.
We are grateful to Jean‐Michel Betton for helpful discussions and for the hydroxylamine mutagenesis. We thank Beth Traxler and Erwin Schneider for the generous gift of MalF and MalK antibodies. The technical assistance of Patricia Lambert in several steps of this work was highly appreciated. M.M. is supported by a fellowship of the Direction de la Recherche et de la Technologie (DRET 91815‐42/A000) and by the Institut de Formation Supérieure Biomédicale (IFSBM).
- Copyright © 1997 European Molecular Biology Organization