A mutation which affects both the specificity of PtsG sugar transport and the regulation of ptsG expression by Mlc in Escherichia coli

The phosphoenolpyruvate (PEP)-dependent phosphotransferase system (PTS) is the major sugar transport system in bacteria. It consists of a series of proteins which transfer a phosphate from PEP to a sugar as it enters the cell. There are two cytoplasmic components, Enzyme I (EI) and HPr (encoded by the ptsI and ptsH genes) which are common to the transport of most sugars and which pass the phosphate to the sugar-specific membrane components called Enzyme IIs (EIIs). The EIIs are responsible for the phosphorylation of the sugar during its transport across the inner membrane. The EIIs are multidomain proteins where the different domains (named IIA, IIB, IIC) are situated on one or several polypeptides (reviewed by Postma et al., 1996 ). The IIC domain is an intrinsic membrane-spanning protein, whilst the IIA and IIB domains are soluble proteins which in some cases are anchored to the cytoplasmic side of the inner membrane by the IIC domain.

The EIIs can be specific for one sugar or capable of transporting several different sugars. For example, the only known sugar substrate of EII^Nag from Escherichia coli, encoded by nagE, is N-acetylglucosamine (GlcNAc), although EII^Nag does transport the antibiotic streptozotocin (Lengeler, 1980 ). GlcNAc is also transported by the mannose (Man) PTS, EIIABCD^Man, encoded by the manXYZ operon. This transporter has a wide substrate specificity, transporting a number of hexoses, including Man, glucose (Glc), fructose and the amino sugars, glucosamine (GlcN) and GlcNAc. On the other hand, the Glc PTS (EIICB^Glc), encoded by ptsG, has a more limited range of substrates; it is the major transporter for Glc but it is also capable of transporting Man and the non-metabolizable Glc analogue α-methylglucoside (Curtis & Epstein, 1975 ; Kornberg & Jones-Mortimer, 1975 ). PtsG is, however, normally incapable of transporting the amino sugars so that in a wild-type E. coli strain GlcN is only transported by the Man PTS.

A spontaneous mutation was selected which allowed an E. coli manXYZ (ptsM) mutant strain to grow on GlcN (Jones-Mortimer & Kornberg, 1980 ). This mutation, called umgC, mapped to the ptsG loci (ptsG was previously called umg, uptake of α-methylglucoside). It was believed that this mutation rendered the expression of ptsG constitutive and thus capable of transporting GlcN which was not itself capable of inducing ptsG expression, unlike Glc (Jones-Mortimer & Kornberg, 1980 ).

It has been shown that expression of ptsG is controlled by the mlc-encoded transcription factor. Growth on Glc induces ptsG expression (Erni & Zanolari, 1986 ; Kornberg & Reeves, 1972 ), presumably by relieving Mlc repression (Kimata et al., 1998 ; Plumbridge, 1998b , 1999 ). Mlc has also been shown to control the expression of the manXYZ transporter (Plumbridge, 1998a ) and the central genes of the PTS, ptsHI (Kim et al., 1999 ; Plumbridge, 1999 ; Tanaka et al., 1999 ). These experiments suggested there was a correlation between the activity of PtsG, in particular the state of phosphorylation of IIB^Glc, and the derepression of Mlc-controlled genes (Fig. 1). In the light of this information, a possible explanation for the effect of the umgC mutation was that it rendered PtsG expression constitutive by affecting Mlc regulation. In this work the umgC mutation has been characterized and its effect on Mlc regulation of ptsG investigated.

(12K): Fig. 1. A model for the regulation of Mlc-controlled promoters by the activity of PtsG. (a) In the absence of Glc, IICBGlc (PtsG) is predominately in its phosphorylated form and Mlc-controlled genes are repressed. (b) Glc is transported across the cytoplasmic membrane via IICGlc, it is concomitantly phosphorylated by IIBGlc and Mlc-controlled genes are derepressed. One hypothesis to explain the correlation between unphosphorylated IICBGlc and derepression of Mlc-controlled genes invokes a physical interaction between Mlc and membrane-associated IICBGlc. It is proposed that the UmgC mutant described in this work partially mimics the conformation normally adopted by IICBGlc during transport of Glc and hence leads to derepression of Mlc-controlled genes.

Bacteriological methods and verification of the umgC phenotype.
The bacterial strains used are listed in Table 1. The ptsG-lacZ, ptsH-lacZ and manX-lacZ fusions carried on λ bacteriophage have been described previously (Plumbridge, 1998a , b , 1999 ). The effect of the different mutations on expression of these fusions was tested by introducing the mutations into lysogenic strains by P1 transductions. The umgC mutation was isolated from strain JM2053 (Table 1). A P1 lysate grown on JM2053 was used to transduce the ptsG22 ptsM8 (GlcN^-) strain, IBPC522, selecting for GlcN⁺. There were two classes of GlcN⁺ bacteria as judged by their colour on MacConkey/GlcN plates and both were Glc⁺ Man⁺. The two classes corresponded to those which had become manXYZ⁺ or umgC (ptsG⁺). Those which had received the umgC mutation (pink-red on MacConkey/GlcN, e.g. IBPC703) remained GlcN⁺ when the manXYZ::Tn9 mutation was introduced, whilst those which had acquired a manXYZ⁺ allele (red on MacConkey/GlcN, e.g. IBPC704) lost the ability to grow on GlcN when the manXYZ::Tn9 mutation was introduced. The umgC (GlcN⁺) mutation was shown to be located near to ptsG since it was co-transducible (510%) with zcf229::Tn10 The work described here shows that the umgC mutation is in fact an allele of ptsG. However, for simplicity and clarity it will be referred to as umgC to distinguish it from other ptsG mutations. The zcf229::Tn10. transposon was used to introduce the umgC mutation into manXYZ::Tn9 strains, screening for GlcN⁺ bacteria. The manXYZ::Tn9 mutation was removed by selecting for zea3125::Tn10Km and screening for Cm^S. The ptsM8 (manXYZ) mutation was introduced using zdj225::Tn10 (Plumbridge, 1990 ). Bacteria were grown in minimal MOPS medium at 30 °C and ß-galactosidase activities were measured as described previously (Miller, 1972 ; Plumbridge, 1998a ).

Table 1. E. coli strains

S1 mapping.
Total RNA was prepared from strains JM-G1, JM-G2, JM-G35 and JM-G55 growing on minimal MOPS medium with glycerol or GlcN as carbon source. The probe used was the PCR fragment, Glc5-Glc2 (Fig. 2), labelled at Glc2 to monitor expression from the chromosomal copy of ptsG. A second probe covering the ptsG-lacZ junction (Glc1-Lac22) was used to monitor mRNA from the ptsG-lacZ fusion. Hybridization and S1 analysis were carried out as described previously (Plumbridge, 1998b ).

(15K): Fig. 2. The ptsG gene and location of the umgC mutation. The relative locations of the IIC and IIB domains are indicated. The two changes found in the umgC strain (G176D and E472K) are shown and the site of phosphorylation, Cys421, within IIB is indicated. ptsG is expressed from two promoters, p1 and p2, and controlled by Mlc binding to two operators located upstream as shown. The locations of the oligonucleotides Glc2, Glc5 and Glc9 are shown with an asterisk at their 5 end. Only relevant restriction sites are indicated. The EcoRI site shown with a dotted line is derived from pBR322 during the cloning.

DNA sequencing and cloning of the umgC mutation.
PCR fragments covering the ptsG gene (Glc5 to Glc9, Fig. 2) were amplified from chromosomal DNA of the wild-type and the umgC mutated allele (IBPC703), and sequenced using the Thermosequenase Radiolabelled Terminator Cycle Sequencing Kit (Amersham Pharmacia) and a series of internal primers. Plasmid pTZ(PtsG) has been described previously (Plumbridge, 1999 ). pTZ(UmgC) was made in an analogous way starting from IBPC703. Both plasmids contain DNA corresponding to the MluI to Glc9 oligonucleotide (see Fig. 2) so that the ptsG promoter and Mlc operator region is absent and ptsG is expressed from the lac promoter. To separate the G176D mutation from the E472K mutation on pTZ(UmgC), the N-terminal and C-terminal halves of the gene were recloned using the SacII restriction site located near aa 285. The presence of the mutations on the plasmids was verified by sequencing. The relative effects of umgC and mlc mutations on growth on GlcN and Man
The umgC mutation of JM2053, conferring growth on GlcN in a manXYZ background, was initially transduced into a strain derived from IBPC5321 (IBPC703). IBPC5321 was subsequently shown to carry a spontaneous mlc (mlc-1) mutation (Plumbridge & Vimr, 1999 ). To verify that the mlc-1 mutation in IBPC703 was not necessary for the GlcN⁺ phenotype, the umgC mutation was transduced out of IBPC703 using zcf229::Tn10 into JM-G11, a manXYZ::Tn9 mlc⁺ strain derived from JM101, to give JM-G35. Both JM-G11 and JM-G35 carry a ptsG-lacZ fusion on a λ phage as reporter for ptsG expression. About 5% of the Tc^R bacteria became GlcN⁺ Man⁺ while retaining the manXYZ::Tn9 mutation. This shows that the umgC mutation is sufficient to confer growth on GlcN to a manXYZ strain in the absence of an mlc mutation. As will be shown, the umgC mutation is an allele of ptsG but for simplicity it will be called umgC to distinguish it from ptsG⁺ and ptsG^- alleles.

The capacities of the manXYZ- or umgC-encoded transporters to allow growth on Man and GlcN were investigated by measuring the growth rates of strains using either of these systems in the presence or absence of the mlc mutation (Table 2). Mlc controls the expression of both manXYZ and ptsG. An mlc mutation increases their expression 4- and 20-fold respectively, in low catabolite repression media like glycerol (Plumbridge, 1998a , b , 1999 ). The effect of the mlc mutation on manXYZ expression is detected by an increase in the growth rate on both GlcN and Man. The introduction of the mlc mutation into a wild-type strain decreases the doubling time (DT) on GlcN from about 150 to 100 min (30 °C) (Table 2, line 2). A similar decrease in DT is seen with Man. This decrease is almost exclusively due to the effect of the mlc mutation on manXYZ expression since the PtsG-encoded transporter is effectively incapable of allowing growth on GlcN and allows only slow growth on Man. This is shown by the fact that introduction of a manXYZ mutation into a wild-type strain (ptsG⁺mlc⁺) eliminates measurable growth on GlcN and increases the DT on Man from about 140 to 300 min (Table 2, line 3). Subsequent introduction of an mlc mutation into the manXYZ strain has only a very small effect on growth on Man or GlcN. The growth rate on Man is slightly improved and growth on GlcN can now be measured, but with a DT of about 600 min (Table 2, line 4). This shows that GlcN can be transported, but very inefficiently, by the PtsG transporter when its expression is increased.

Table 2. Effect of the umgC mutation on growth rates on GlcN and Man, and on expression of ptsG-lacZ

The effect of the introduction of the mlc mutation into a manXYZ strain can be contrasted with the introduction of the umgC mutation. The latter allows growth rates comparable to the wild-type manXYZ-encoded transporter. The DT of the manXYZ umgC strain on GlcN is about 150 min compared to 600 min for the manXYZ mlc strain, using the wild-type PtsG transporter (Table 2, line 5). Further introduction of an mlc mutation into these manXYZ umgC strains had no appreciable effect on the growth rates. The manXYZ umgC and the manXYZ umgC mlc strains exhibit similar growth rates on all sugars tested. The effect of the umgC and mlc mutations was also tested during growth on glycerol and Glc. The umgC mutation produced a slight decrease in the growth rate on glycerol and possibly on Glc, but the changes were relatively small (Table 2).

These experiments strongly suggest that constitutive expression of wild-type ptsG, as produced by the mlc mutation, is not sufficient to allow PtsG to transport GlcN efficiently and so imply that the umgC mutation is acting in a different, specific way to permit GlcN and Man uptake and that the umgC mutation is not just enhancing ptsG expression as expected for a constitutive mutation.

The umgC mutation enhances ptsG expression
The effect of the umgC mutation on growth rates compared to that of the transcription factor mlc strongly suggests that the umgC mutation is altering PtsG sugar specificity and thus allowing it to transport GlcN. However, the umgC mutation does also increase ptsG expression as shown by an effect in trans of the umgC mutation on the ptsG-lacZ fusion. Growth of a wild-type strain on either GlcN or Man increases ptsG expression three- to fourfold but not so much as growth on Glc (eightfold; Table 2, line 1). These and other PTS sugars are capable of inducing ptsG expression presumably by partially relieving Mlc repression, although the molecular mechanism of the derepression has not been fully elucidated (Plumbridge, 1999 ). The presence of the mlc mutation derepresses expression in all four media relative to mlc⁺, but the level of ptsG-lacZ expression observed is different in each medium (Table 2, line 2). The P1 promoter of the ptsG gene is dependent upon cAMP/CAP and the level of expression in the different sugars is related to the level of catabolite repression generated by growth on the different sugars (Epstein et al., 1975 ; Hogema et al., 1998 ; Plumbridge, 1999 ).

The effect of the umgC mutation on the level of expression of the ptsG-lacZ fusion was compared to that of the mlc mutation (Table 2, lines 2 and 5). Two effects are discernible. Even in a non-PTS medium like glycerol, the umgC mutation causes about a fourfold increase in ptsG-lacZ expression. Introduction of an mlc mutation to the umgC strain increased expression in glycerol to the same level as in the mlc strain. In GlcN or Man media the umgC mutation also increases ptsG-lacZ expression about five- to sixfold (i.e. at least 20-fold compared to the basal level of the wild-type strain in glycerol), producing levels of expression even higher than that produced by the mlc mutation in the same medium. Subsequent introduction of the mlc mutation into the umgC strain had no further effect on expression during growth of the PTS sugars (Table 2, line 6). This can be correlated with the similar growth rates of strains carrying the umgC and umgC mlc mutations. Although the presence of the umgC mutation affects ptsG expression in the absence of any PTS substrate, expression is considerably increased in the presence of the UmgC substrates, Man or GlcN. It can also be noted that even in Glc medium the umgC mutation appears to increase ptsG expression somewhat, producing about a 30% increase in ptsG-lacZ expression.

The strong effect of the umgC mutation during growth on GlcN and Man in the manXYZ background on expression of the ptsG-lacZ reporter is reduced by the presence of a manXYZ⁺ allele. The presence of a second transporter for Man and GlcN reduces the level of expression of the ptsG-lacZ fusion, presumably by reducing the flux through the UmgC transporter. On the other hand the level of expression of ptsG-lacZ during growth on glycerol is not affected by the presence of the manXYZ ⁺ allele.

The umgC mutation enhances ptsG mRNA levels
The increase in ptsG-lacZ expression is paralleled by an increase in chromosomal ptsG mRNA levels as measured by an S1 protection experiment using the Glc5-Glc2 probe (Fig. 2). The umgC mutation produces a small increase during growth on glycerol (Fig. 3, lane 2) but a considerable increase during growth on GlcN (Fig. 3, lane 6). The mlc mutation in glycerol produces a comparable enhancement in ptsG mRNA (Fig. 3, lanes 3 and 4,) but the mlc mutation in GlcN (Fig. 3, lane 7) seems to produce a somewhat lower level of ptsG mRNA than the umgC mutation, with or without the mlc mutation (Fig. 3, lanes 6 and 8). This correlates with a lower value for the ptsG-lacZ fusion in the mlc strain grown on GlcN than for the umgC strain in the same medium (Table 2). The position of the two transcription start sites for ptsG, the major P1 transcript and the minor P2 transcript, are not changed by the umgC mutation and both mRNAs are similarly affected, which is the expected observation if umgC is altering Mlc regulation, since Mlc controls both promoters (Plumbridge, 1998b ). The S1 analysis was repeated using a probe specific for the ptsG-lacZ fusion. The pattern of expression was the same as for the chromosomal copy of ptsG (data not shown).

(52K): Fig. 3. S1 analysis of the effect of umgC mutation on chromosomal ptsG mRNA. The probe used was the Glc5Glc2 fragment labelled at Glc2. It was hybridized overnight in 40 mM PIPES (pH 6·9), 400 mM NaCl, 80% formamide (52 °C) with 25 µg total RNA extracted from JM-G1 (lanes 1 and 5), JM-G35 (umgC; lanes 2 and 6), JM-G2 (mlc; lanes 3 and 7) and JM-G55 (umgC mlc; lanes 4 and 8) grown in minimal glycerol medium (lanes 14) or minimal GlcN medium (lanes 58). C indicates the strain carried the umgC mutation. Lane 9 shows the control with 25 µg tRNA. After digestion with S1 nuclease (250 units ml-1 in 30 mM sodium acetate, pH 4·5, 250 mM NaCl, 1 mM ZnCl2, 5% glycerol) for 40 min at 37 °C, S1-resistant hybrids were analysed on a 6% denaturing acrylamide gel. The major and minor ptsG transcripts, P1 and P2, are indicated.

The umgC mutation enhances ptsH and manX expression
Both the ptsHIcrr and manXYZ operons are controlled by Mlc but the factors of induction, three- to fourfold, are lower than that for ptsG (Plumbridge, 1998a , b , 1999 ; Table 3). The umgC mutation produces a small increase in expression of both ptsH and manX during growth on glycerol and larger increases during growth on GlcN and Man, comparable to the presence of the mlc mutation (Table 3). The observation that umgC has an equivalent effect on ptsG, ptsH and manX is in agreement with the idea that the umgC mutation is affecting their expression via Mlc regulation. Expression of the nagE-encoded GlcNAc transporter is not controlled by Mlc and is not significantly affected by the umgC mutation (data not shown).

Table 3. Effect of the umgC mutation on ptsH-lacZ and manX-lacZ expression

The umgC allele has two mutations, one in the IIC domain and one in the IIB domain of EIICB^Glc
Sequencing the ptsG chromosomal locus defined by oligonucleotides Glc5 to Glc9, which include 200 bp upstream of the start codon, of the umgC strain detected two point mutations in the ptsG ORF: G176D in what is predicted to be a periplasmic loop of the IIC domain (Buhr & Erni, 1993 ), and E472K near the C-terminal end of the protein (total 477 aa) within the soluble IIB domain (Fig. 2). The ptsG locus (from MluI site to oligo Glc9) carrying the umgC mutation was cloned into a plasmid vector under control of the lac promoter. Plasmids carrying each mutation individually were also constructed. The ability of the plasmids carrying the single and double mutations to allow growth on GlcN and Glc was measured. Both plasmids carrying the original double umgC mutation and the G176D single mutation in the IIC domain allowed growth on GlcN but not the plasmid carrying the E472K IIB mutation (Table 4). In fact, the plasmid carrying the E472K mutation appears to have a negative effect on growth on GlcN and Man. The G176D and double UmgC plasmids also produced an increased expression of ptsG-lacZ during growth on glycerol, showing that it is the IIC mutation which is responsible for both the change in sugar specificity and the increase in ptsG expression.

Table 4. Effect of the umgC mutations on plasmids on growth rates and expression of the ptsG-lacZ fusion

The umgC mutation affects PtsG sugar transport specificity
The PtsG protein consists of two domains: an N-terminal membrane-bound EIIC domain predicted to include eight membrane-spanning helices, which is responsible for sugar transport across the cytoplasmic membrane, and a soluble EIIB domain located on the cytoplasmic side of the membrane (Buhr & Erni, 1993 ). The EIIB domain contains the Cys421 residue which is phosphorylated by EIIA^Glc and then transfers the phosphate to the transported sugar concomitantly with its passage through the cytoplasmic membrane. Several mutations have been described which affect the sugar specificity of EII^Glc in E. coli. G320V, located in the predicted transmembrane helix VIII (Buhr & Erni, 1993 ), allows PtsG to take up mannitol (Begley et al., 1996 ), five mutations (F37Y, G176D, G281D, I283T and L289Q) have been characterized which allow PtsG to transport ribose (Oh et al., 1999 ) and one mutation, V12F, has been described which allows the transport of fructose (Kornberg et al., 2000 ). Ribose is not a PTS sugar and both ribose and fructose were taken up as free sugars by facilitated diffusion rather than by phosphorylation-accompanied transport. Interestingly, the V12F mutation, located in what is predicted to be an N-terminal α-helix on the cytoplasmic side of the membrane, was previously isolated in a selection for strains which grew better in a Glc- and oxygen-limited environment (Manché et al., 1999 ). It shows enhanced Glc transport characteristics as well as the ability to transport fructose (Kornberg et al., 2000 ; Manché et al., 1999 ). Two of the ribose-positive mutations were found in regions predicted to be in periplasmic loops of the membrane-spanning region (F37Y and G176D), whilst the three others were clustered in a sequence predicted to form transmembrane helix VII (G281D, I283T and L289Q). In addition, mutations which impair transport but do not affect phosphorylation (Buhr et al., 1992 ) and mutations which partially uncouple transport from phosphorylation (Ruijter et al., 1992 ) are found throughout the IIC domain. The wide distribution of these mutations affecting sugar specificity or phosphorylation implies a flexible structure where changes at many locations, both within the membrane and in the connecting loops, affect transport characteristics.

Interestingly, the IIC mutation found in the umgC strain studied here is the same as one of the ribose-positive mutations, G176D (Oh et al., 1999 ). Moreover, recent experiments, which repeated the selection for a GlcN⁺ phenotype, yielded several different mutations, including changes in both V12 and G176 (Notley-McRobb & Ferenci, 2000 ; Zeppenfeld et al., 2000 ). As in the case of V12F, the umgC mutation also allows uptake of fructose (data not shown), implying that V12F and G176D allow a general reduction in sugar specificity and transportphosphorylation coupling. It should be noted that the umgC mutation does not eliminate all sugar specificity. It does not, for example, allow the uptake of GlcNAc; introduction of the umgC mutation to a manXYZ nagE (GlcNAc^-) strain does not enable the strain to grow on GlcNAc.

The umgC mutation affects Mlc regulation
Two distinct, but interrelated, phenotypes can be assigned to the umgC mutation: an enhanced transport capacity for GlcN or Man and an increase in the expression of ptsG and other genes regulated by Mlc. Several pieces of evidence support the hypothesis that the umgC mutation affects regulation by Mlc: the umgC mutation produces parallel effects in cis on ptsG mRNA levels and in trans on ptsG-lacZ expression; three different Mlc-controlled operons, ptsG, manXYZ and ptsHI, are similarly affected; and the mlc mutation is epistatic to umgC mutations. During growth on GlcN and Man, when the umgC mutation has its maximum effect, there is no additional increase in expression when an mlc mutation is introduced, consistent with the idea that the mlc and umgC mutations are affecting the same regulatory pathway. However, it is clear that the increase in ptsG expression produced by the umgC mutation is not the only reason that growth on GlcN and Man improves since an mlc mutation, which produces increased ptsG expression, does not allow good growth on GlcN or Man. Notice that there is relatively little effect of the umgC mutation during growth on Glc. This is presumably because the level of catabolite repression generated by Glc dominates the expression level of ptsG and other Mlc-controlled genes, all of which are strongly cAMP/CAP-dependent.

The G176D mutation both causes the change in sugar specificity and the increase in ptsG expression
The analysis of the two mutations found in the umgC strain, G176D and E472K, individually on plasmids show that it is the G176D mutation which is responsible for both phenotypes, i.e. growth on GlcN and increased ptsG expression. This same mutation has already been shown to permit the uptake of ribose. Since several other mutations affecting PtsG sugar specificity map to the IIC domain, it is not surprising that the G176D mutation is responsible for the change in sugar specificity. It is perhaps less expected that the regulatory phenotype, increased expression of Mlc-controlled genes, is also associated with this same mutation. Previous results (Plumbridge, 1999 ) implicated the soluble IIB domain and the level of PtsG phosphorylation in the regulation of Mlc-controlled genes. Only in the presence of a ptsG⁺ allele were Mlc-regulated genes derepressed by growth on PTS sugars or by mutations which interrupted the flow of phosphates from PEP to the sugar.

Growth on Glc reasonably leads to induction by relieving Mlc repression but unlike other classical repressors no low-molecular-mass metabolite has been identified which is capable of displacing Mlc from its DNA targets. One hypothesis to explain the role of PtsG in Mlc induction is that transport of Glc permits an interaction between Mlc and membrane-bound PtsG, thus sequestering Mlc away from its DNA targets (Fig. 1; Plumbridge, 1999 ). Recent experiments have indeed shown that such an interaction between membrane-bound PtsG and Mlc exists (S.-J. Lee, W. Boos and J. Plumbridge, unpublished). The predicted location of G176D in a periplasmic loop strongly suggests that there is no direct interaction between G176D and Mlc which is presumably confined to the cytoplasm or the cytoplasmic side of the inner membrane. Instead it would imply that the G176D mutation has a global effect on IICB^Glc structure which is monitored by Mlc in the soluble part of the cell. Since the G176D mutation affects regulation in the absence of PTS sugar transport, the mutation might be mimicking an active transport conformation which is able to signal to the IIB domain. There is evidence for such a conformational coupling between the IIC and IIB domains of the mannitol-specific PTS permease (Meijberg et al., 1998 ). The V12F mutation in the N-terminal cytoplasmic helix of the IIC domain, which shows similar characteristics to G176D (increased ptsG expression and the ability to transport fructose and GlcN), could, however, be in direct contact with Mlc.

The E472K mutation near the C terminus of the IIB domain appears to have no direct effect on ptsG expression levels. As the G176D mutation in the umgC strain is sufficient for growth on GlcN, one can wonder what is the origin of the E472K mutation. The presence of a plasmid carrying E472K appears to have a detrimental effect during growth on GlcN or Man compared to the wild-type ptsG allele (Table 4), suggesting that E472K might decrease PtsG activity. If the G176D mutation produces a very strong increase in PtsG transport activity, as has been shown for V12F (Kornberg et al., 2000 ; Manché et al., 1999 ), a second mutation reducing this hyperactivity might have arisen. This hypothesis is supported by the observation that when another example of strain JM2053, which had been stored independently for many years, was analysed, it was also found to carry two mutations: G176D and a UAG stop codon at position E472 (L. Notley-McRobb & T. Ferenci, personal communication) rather than the AAG lysine found in JM2053 obtained from the CGSC and analysed in this study.

Biochemical evidence for the model in Fig. 1 comes from the demonstration that Mlc binds to PtsG-containing membranes (Lee et al., 2000 ; Tanaka et al., 2000 ). I am very grateful to H. Kornberg, T. Ferenci and K. Jahreis for open discussions and communicating results prior to publication, to A. Kolb, J. Deutscher and W. Boos for many useful discussions, to H. Kornberg and H. Putzer for critical comments on an earlier version of the manuscript and to B. Erni, H. Kornberg, J.-H. Alix and the CGSC for the gift of bacterial strains.