Dephosphorylation of the Escherichia coli Transcriptional Antiterminator BglG by the Sugar Sensor BglF Is the Reversal of Its Phosphorylation

Abstract

The Escherichia coli BglF protein catalyzes transport and phosphorylation of β-glucosides. In addition, BglF is a membrane sensor which reversibly phosphorylates the transcriptional regulator BglG, depending on β-glucoside availability. Therefore, BglF has three enzymatic activities: β-glucoside phosphotransferase, BglG phosphorylase, and phospho-BglG (BglG-P) dephosphorylase. Cys-24 of BglF is the active site which delivers the phosphoryl group either to the sugar or to BglG. To characterize the dephosphorylase activity, we asked whether BglG-P can give the phosphoryl group back to Cys-24 of BglF. Here we provide evidence which is consistent with the interpretation that Cys-24–P is an intermediate in the BglG-P dephosphorylation reaction. Hence, the dephosphorylation reaction catalyzed by BglF proceeds via reversal of the phosphorylation reaction.

The bgl operon in Escherichia coli, induced by an environmental stimulus (β-glucosides), is regulated by a membrane-bound sensor, BglF, and a cytoplasmic regulator, BglG (4). BglG is an RNA-binding protein which, in the presence of the inducing sugar, antiterminates transcription of the bgl operon (12, 17). BglF, also designated EII^bgl, is an enzyme II (EII) of the phosphoenolpyruvate (PEP)-dependent carbohydrate phosphotransferase system (PTS) which catalyzes concomitant transport and phosphorylation of β-glucosides (10). In addition to its role in sugar transport, BglF modulates BglG activity by phosphorylating and dephosphorylating it according to β-glucoside availability (1, 2, 22), consequently controlling its dimeric state (3). Thus, in the absence of β-glucosides, BglF phosphorylates BglG; phospho-BglG (BglG-P) is an inactive monomer. The presence of β-glucosides stimulates BglF to dephosphorylate BglG; dephosphorylated BglG is an active dimer which leads to expression of the bgl operon.

Like many other EIIs of the PTS, BglF consists of three well-defined domains: IIA^bgl possesses the first phosphorylation site (site 1), His-547, which is phosphorylated by HPr-P; IIB^bgl possesses the second phosphorylation site (site 2), Cys-24, which accepts the phosphoryl group from IIA^bgl-P; IIC^bgl, the membrane-spanning domain, presumably forms the sugar translocation channel and at least part of the sugar-binding site (8, 19, 23). Yet, BglF is the first EII shown to reversibly phosphorylate a transcription regulator in addition to phosphorylating its sugar substrate. It has been shown that site 2 of BglF, Cys-24, is the phosphate donor to both the sugar and the BglG protein (8), but the mechanism by which BglF dephosphorylates BglG remains unclear. The finding that a mutant BglF protein in which Cys-24 was replaced by a serine (C24S) lost the ability to dephosphorylate BglG-P in vitro in the presence of the β-glucoside salicin (data not shown) indicates that Cys-24 plays an important role in BglG dephosphorylation. A likely possibility is that Cys-24 accepts the phosphoryl group back from BglG-P, implying that the reaction of BglG dephosphorylation is the reversal of its phosphorylation. The reversibility of the phosphorylation reactions between the different components of the PTS favors this possibility. However, it is not clear whether BglG can be defined as a PTS member. In the present study, we have investigated the mechanism of BglG dephosphorylation by BglF. Our results show that the phosphoryl group of BglG-P can indeed be transferred back to Cys-24 of BglF.

Role of Cys-24 in BglF dephosphorylase activity: experimental plan.

To determine whether BglG gives the phosphoryl group back to Cys-24 of BglF, we planned the experiment schematically described in Fig. 1. We took advantage of the reversibility of the phosphorylation reactions between the different PTS components (19 and references therein) and the cross-phosphorylation between BglF and IIA^glc/IICB^glc (the glucose EII complex, which consists of a soluble and a membrane-bound protein, respectively) (8, 23, 27). We added [³²P]BglG, as the only phosphoryl source, to a mixture containing BglF, IIA^glc, IICB^glc, and methyl α-glucoside, a carbohydrate that is specifically phosphorylated by IICB^glc (27). If BglG-P can transfer its phosphoryl group back to Cys-24 in BglF (Fig. 1, reaction 1), it can be subsequently delivered either to site 1 of BglF (His-547) or to site 1 of the glucose permease (His-90 in IIA^glc) (Fig. 1, reactions 2a and 2b, respectively). The delivery to site 1 of BglF (reaction 2a) can be prevented by using a BglF variant in which His-547 is mutated. Such a mutant (H547R) was previously shown to catalyze BglG dephosphorylation in vivo in the presence of β-glucosides (8), indicating that His-547 is not essential for the dephosphorylase activity. From site 1 of IIA^glc, the phosphoryl group can be transferred to site 2 of the glucose permease (Cys-421 in IICB^glc) and then to methyl α-glucoside (Fig. 1, reactions 3 and 4, respectively). If wild-type BglF is used, the phosphoryl group can flow from its site 1 directly to IICB^glc. In any case, if reaction 1 can occur, the equilibrium is expected to be shifted toward sugar phosphorylation (due to phosphoryl drainage by the methyl α-glucoside), and thus a decrease in the level of [³²P]BglG should be observed. Following the same working hypothesis, the BglF mutant protein C24S, in which Cys-24 is replaced by a serine, is not expected to affect the level of the radioactively labeled BglG-P upon incubation with IIA^glc, IICB^glc, and methyl α-glucoside.

FIG. 1.

Experimental design for studying the role of Cys-24 in BglF dephosphorylase activity. Shown is a schematic representation of phosphoryl flow from BglG-P to methyl α-glucoside. G, BglG; α-MG, methyl α-glucoside; P, phosphoryl group.

Analysis of the dephosphorylase activity of BglF.

To carry out the experimental plan described above, we prepared radioactively labeled BglG-P and MBP–BglG-P (BglG fused to maltose-binding protein and purified on an amylose column as described in reference 8). [³²P]BglG was prepared by adding a BglG-containing extract to phosphorylation system A (8), which contained [³²P]PEP, a cytoplasmic fraction of Salmonella typhimurium strain LJ144 that expresses increased amounts of enzyme I, HPr, and IIA^glc (21), and membranes prepared from E. coli K38 cells overproducing BglF. [³²P]MBP-BglG was prepared by adding purified MBP-BglG to phosphorylation system B, which contained [³²P]PEP, purified enzyme I and HPr, and membranes prepared from E. coli LM1 cells overproducing BglF. The LM1 strain contains mutations in the crr and nagE genes that result in lack of IIA^glc and II^nag activities (16). [³²P]BglG and [³²P]MBP-BglG were separated from [³²P]BglF and [³²P]PEP as previously described (1). For the experiments described below, it is most relevant that the [³²P]BglG preparation contained IIA^glc (which copurified together with [³²P]BglG), whereas the [³²P]MBP-BglG preparation lacked IIA^glc.

The ability of wild-type BglF and that of BglF mutated in either one of its phosphorylation sites (C24S or H547R; reference 8) to dephosphorylate BglG-P in the presence of the glucose permease and methyl α-glucoside were tested. In the first set of experiments, [³²P]BglG was incubated with membranes of E. coli LM1 cells overproducing either wild-type BglF or one of its mutants, together with methyl α-glucoside. These mixtures contained both IIA^glc (see above) and IICB^glc (present in the membrane fraction of the E. coli LM1 strain). In the presence of wild-type BglF, dephosphorylation of [³²P]BglG occurred in a time-dependent manner (Fig. 2A, lanes 1 to 3). Because methyl α-glucoside can only be phosphorylated by the glucose permease, which cannot dephosphorylate [³²P]BglG (see below), this result suggested that BglF-P was an intermediate in this experiment. The C24S mutant did not dephosphorylate [³²P]BglG under these conditions (Fig. 2A, lanes 4 to 6), indicating that Cys-24 plays an important role in BglG-P dephosphorylation. However, the H547R mutant dephosphorylated [³²P]BglG in the presence of methyl α-glucoside (Fig. 2A, lanes 7 to 9). Since Cys-24 is the only phosphorylation site in the H547R mutant, the sugar-induced dephosphorylation of BglG-P in the latter case can only be explained if the phosphoryl flow is as suggested in Fig. 1, i.e., from BglG-P to Cys-24 of BglF, subsequently to IIA^glc, then to IICB^glc, and finally to the sugar. Therefore, this result strongly suggests that the Cys-24 residue can accept the phosphoryl group from BglG-P. The slightly more efficient dephosphorylation of [³²P]BglG by wild-type BglF compared to the H547R mutant can be explained by the fact that with wild-type BglF, the first step in the phosphoryl flow involves an intramolecular transfer (from Cys-24 of BglF to His-547 of BglF and then to IICB^glc) versus an intermolecular transfer, as in the case of the H547R mutant (from Cys-24 of BglF to His-90 of IIA^glc and then to IICB^glc). No dephosphorylation of [³²P]BglG was observed when membranes of LM1 cells that do not express the bglF gene were included in the incubation (Fig. 2A, lanes 10 to 12). This result emphasizes that [³²P]BglG cannot be dephosphorylated by IIA^glc, IICB^glc, and methyl α-glucoside in the absence of BglF.

FIG. 2.

BglG-P delivers the phosphoryl group to methyl α-glucoside via Cys-24 of BglF. Phosphorylated BglG proteins were prepared as described in the text. (A) [³²P]BglG was incubated with membranes of LM1 cells that overproduce the various BglF derivatives (wild-type BglF, C24S, and H547R) in the presence of 0.2% methyl α-glucoside at 30°C for the times indicated. (B) [³²P]MBP-BglG was incubated with membranes of LM1 cells that overproduce the various BglF derivatives in the presence of 0.2% methyl α-glucoside and IIA^glc (obtained from J. Reizer) at 40 μg/ml at 30°C for the times indicated. (C) Same as panel B but without IIA^glc. (D) [³²P]BglG was incubated with membranes of ZSC112ΔG cells (with the pstG gene deleted, thus not expressing IICB^glc) that overproduce wild-type BglF in the presence of 0.2% methyl α-glucoside at 30°C for the times indicated. (E) [³²P]BglG was incubated with membranes of LM1 cells that overproduce wild-type BglF in the absence of methyl α-glucoside at 30°C for the times indicated. H547R and C24S, mutations in the first and second phosphorylation sites of BglF (site 1 and site 2), respectively. No BglF, membranes of cells which do not express BglF, but are otherwise identical to the other membrane preparations used in each experiment, were included in the different phosphorylation systems. Samples were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis followed by autoradiography. Arrowheads indicate the positions of BglG and MBP-BglG. The faster-migrating component in panels A, D, and E is a degradation product of BglG (1).

In the second set of experiments, the abilities of the different BglF derivatives to dephosphorylate MBP–BglG-P were tested. In addition to membranes of E. coli LM1 enriched for the different BglF variants and methyl α-glucoside, purified IIA^glc was added in this case, because the [³²P]MBP-BglG preparation lacked IIA^glc (see above). The results, presented in Fig. 2B, are essentially the same as the results obtained with [³²P]BglG; i.e., wild-type BglF and BglF H547R, but not BglF C24S, dephosphorylated MBP–BglG-P.

To further confirm that the pathways of the phosphoryl flow during BglG dephosphorylation are as predicted by us (Fig. 1), we tested whether IIA^glc (which can be replaced by the IIA^bgl domain), IICB^glc, and methyl α-glucoside are essential for the dephosphorylation of BglG. We first tested the abilities of the different BglF derivatives to dephosphorylate [³²P]MBP-BglG in the presence of methyl α-glucoside and IICB^glc but in the absence of IIA^glc. The results, presented in Fig. 2C, indicate that whereas wild-type BglF (which has an unaltered IIA^bgl domain) can dephosphorylate MBP–BglG-P in the absence of IIA^glc, the H547R mutant fails to do so. This result suggests that the pathway by which the H547R protein catalyzes transfer of the phosphoryl group from BglG-P to methyl α-glucoside involves phosphoryl delivery from Cys-24 to IIA^glc. Next, we tested whether methyl α-glucoside can trigger BglF to dephosphorylate BglG-P in the absence of IICB^glc. To this end, we used membranes of E. coli ZSC112ΔG cells overexpressing the bglF gene. In this strain, the ptsG gene is deleted; thus, it lacks IICB^glc (7). The results, presented in Fig. 2D, demonstrate that IICB^glc is indispensable for this reaction. Last, we tested whether methyl α-glucoside is essential for the dephosphorylation of BglG-P by BglF and the glucose permease complex. As shown in Fig. 2E, dephosphorylation of BglG-P is not observed without methyl α-glucoside.

Conclusions.

Dephosphorylation of BglG, catalyzed by BglF, could, in principle, proceed via several alternative mechanisms. The fact that substitution of Cys-24 by a serine abolished the ability of BglF to dephosphorylate salicin suggested that Cys-24 not only plays a role in delivering the phosphoryl group to BglG but also accepts it back upon addition of sugar. This result does not support the possibility that BglF acts as BglG cophosphatase that stimulates BglG autodephosphorylation. This is because a cophosphatase is not expected to contribute important residues to the active site, and therefore, amino acid substitutions in the active site should not affect it, as has been observed for NtrB (15). Nevertheless, the inability of the C24S mutant to dephosphorylate BglG-P in the presence of salicin could also be explained by a substantial overlap between the phosphorylase and dephosphorylase sites or by a model which is similar to the one suggested for the dephosphorylation of OmpR-P by its kinase EnvZ; i.e., water replaces the phosphorylated side chain of a residue at the active site, thus leading to hydrolysis (13). Hence, the question of whether Cys-24-P is an intermediate in BglG-P dephosphorylation became crucial and only a positive answer could validate the reverse phosphotransfer model.

The results presented in this paper are consistent with the interpretation that Cys-24–P is formed as an intermediate during dephosphorylation of BglG by BglF. Hence, BglG dephosphorylation, catalyzed by the transmembrane sensor BglF, seems to proceed via reversal of the phosphorylation reaction. Since Cys-24 is also the phosphate donor to the sugar, we suggest that all BglF enzymatic activities are associated with the same active site. This proposed pathway has several attractive implications concerning the mechanism of bgl regulation. First, BglG phosphorylation and dephosphorylation will not occur simultaneously, which seems to be the simplest solution to guarantee that they are mutually exclusive. Second, the sugar itself, by draining the phosphoryl groups from Cys-24, will divert the phosphoryl flow away from BglG-P, thus directly initiating the signal transduction that leads to production of the proteins required for its utilization. This mechanism is reminiscent of the inducer exclusion mechanism by which PTS sugars prevent transport and metabolism of non-PTS sugars (19). The PTS sugars, by draining the phosphoryl groups from their respective EIIs, shift the equilibrium of the phosphoryl flow among PTS components and eventually lead to dephosphorylation of IIA^glc; dephosphorylated IIA^glc binds to non-PTS permeases and prevents their action. The similarity between the two mechanisms does not seem to be fortuitous. Rather, it highlights the affiliation of BglG with the PTS family of proteins. Third, because the same site on BglF, Cys-24, phosphorylates BglG and dephosphorylates BglG-P, it is obvious that the phosphorylation state of this site determines whether BglF functions as a phosphorylase or a dephosphorylase. The ability of β-glucosides to dephosphorylate Cys-24 is the simplest solution for accomplishing the desired modulation between the activities of BglF. Comparison with the mechanism of inducer exclusion suggests also that carbohydrates other than β-glucosides could, in principle, result in the induction of the bgl operon, since any PTS carbohydrate could dephosphorylate HPr, and consequently BglF and BglG, via its specific EII. In practice, this depends on whether the rate of indirect dephosphorylation via HPr is fast enough to achieve net dephosphorylation of BglF-P, and consequently of BglG-P.

Unlike sensors of two-component systems which catalyze transfer of phosphoryl groups from phosphorylated histidines to conserved aspartates on respective regulators (reviewed in references 18, 20, and 24), BglF transfers a phosphoryl group from a cysteine to a histidine residue on the BglG regulator (5, 9). Reversible phosphotransfer from histidine to cysteine is known to occur between site 1 and site 2 of several EIIs (19). Homologues of BglG were found in various organisms, but only in the case of the B. subtilis SacY protein was phosphorylation shown to involve a putative EII, SacX (14). Like BglG, SacY is dephosphorylated in vivo in the presence of the inducing sugar, sucrose in this case (14), but the dephosphorylation reaction has not been characterized yet. BglG and some of its homologues have been shown to be phosphorylated by HPr (6, 11, 25, 26), which is phosphorylated on a histidine residue, but their dephosphorylation by HPr has not been reported.