Historically, some of the earliest work on Escherichia coli genetics and physiology in the Jacob and Monod groups concentrated on the maltose system, but it was rapidly overshadowed by the spectacular success in understanding the lactose operon. Fortunately, interest in the maltose system persisted as a result of its being an example of a “positively regulated” group of genes and the mysterious connection between maltose metabolism and the susceptibility of E. coli to phage λ infection.

The maltose system is responsible for the uptake and efficient catabolism of α(1→4)-linked glucose polymers (maltodextrins) up to 7 to 8 glucose units. This system has turned out to be a far richer source of interesting molecules and regulatory phenomena than anyone might have anticipated 30 years ago when the malA and malB regions were mapped on the E. coli chromosome (239). For example, in the course of studying the receptor activity of E. coli responsible for λ attachment, Randall and Schwartz discovered the λ receptor (LamB) (207) and, in collaboration with Hofnung, Szmelcman, and one of us (W.B.), showed that this outer membrane protein constituted the channel for the passage of sugars, especially maltodextrins, across the outer membrane (263, 265). This information focused attention on the possibility of porins as specific channels and culminated in the solution of the three-dimensional structure of the maltoporin channel associated with malto-oligosaccharide ligands (231). Together with a large body of information about the functional sites within LamB, the structure represents one of the most complete descriptions of a membrane transport protein.

From the earliest time that the maltose system was studied, transcriptional control was a major focus because of the interest in a positively controlled system. The MalT protein, which is the activator at all mal promoters (112, 206, 214), and the sites at which it binds (MalT boxes) have provided an important alternative view to the lac operon about how transcription is activated. Although the precise mechanism of how MalT interacts with and stimulates RNA polymerase is still unknown, there is considerable information about the structural requirements for MalT binding to DNA as well as its ability to form a nucleoprotein complex with the catabolite activator protein (CAP).

The periplasmic binding protein-dependent ATP binding cassette (ABC) transporter of the maltose system has been studied to understand the mechanism of transport and the mechanism of protein localization to different cellular compartments (e.g., inner membrane, outer membrane, and periplasm) (21, 156, 173, 250), as well as the determinants of membrane protein topology. The periplasmic maltose binding protein (MBP) has been characterized structurally (257), and a variety of functional sites within its structure for interacting with the chemosensory apparatus and the ABC transporter have been identified. The actual membrane transporter (MalFGK₂), made up of two integral membrane proteins (MalF and MalG) and two copies of the ATP-hydrolyzing subunit (MalK), has been characterized biochemically, and there is some information about functional sites that are important for substrate recognition, ATP binding, and subunit contacts and interactions with MBP. The ability to manipulate the transporter genetically offers a rare opportunity to dissect the mechanism of transport by a combined approach of biochemistry and genetics. The direct participation of the transporter in transcriptional regulation indicates that gene regulation in bacteria may depend on the rates of substrate entry in addition to the physical presence of the substrate itself.

Reviews on the maltose system of E. coli have appeared (22, 241). In this review, we attempt to summarize what is currently known about this system. We have focused on three areas: (i) transport of maltodextrins into the cell via an outer membrane porin and a periplasmic binding protein-dependent ABC transporter, (ii) metabolism of the internalized sugars, and (iii) transcriptional regulation of the mal genes. We attempt to clarify what we believe are significant functional relationships between transport activity and transcriptional control. Also, we discuss the interconnections between the maltose-degrading enzymes and those of glucose catabolism and gluconeogenesis from the perspective of how endogenous inducers of the mal system are produced. Because the initial breakdown of maltodextrins results in both glucose and glucose-1-phosphate, these enzyme activities must be coordinated with those for glycogen synthesis and breakdown. Finally, other less obvious connections with phosphate metabolism and trehalose metabolism are also discussed. We will not consider aspects of secretion (169), folding (272), or assembly (151, 223) of the components of the maltose system.

Table 1 summarizes all known mal genes in E. coli. Their definition is based on the function of MalT. Thus, all genes regulated by MalT belong to the maltose regulon, while the expression of malT itself is independent of MalT. malP and malQ encode essential enzymes for maltose and maltodextrin metabolism, whereas malS and malZ encode nonessential maltodextrin-metabolizing enzymes. malEFG and malK lamB encode the binding protein-dependent ABC transporter. Some mal genes are organized in clusters. The malA region at 76.5 min contains two divergently orientated operons, with malT transcribed clockwise and malPQ transcribed counterclockwise (65, 126, 200). Likewise, all maltose transport genes are clustered at 91.4 min in the malB region with two divergently organized operons: malEFG in the counterclockwise orientation and malK lamB malM in the clockwise orientation (203, 254).

TABLE 1 mal genes and their products, genes controlling mal gene expression, and genes related to maltodextrinmetabolism

Other gram-negative enteric bacteria have additional genes, not present in E. coli, that are under the control of MalT. For instance, Klebsiella pneumoniae harbors a battery of malT-controlled pul genes that are necessary for the biosynthesis and the secretion of pullulanase (71), an extracellular enzyme that degrades pullulan to maltotriose (14, 285). Pullulan, first isolated from Pullularia pullulans (13), consists of α(1→6)-glucosidically linked maltotriose units. The MalT-dependent pulA gene, encoding the lipoprotein pullulanase (163), is positioned divergently from a series of genes, also dependent on MalT, that are necessary for the export of pullulanase through the outer membrane (193, 195). Likewise, Klebsiella oxytoca harbors a series of cym genes encoding proteins for the uptake and metabolism of cyclic dextrins (89). The expression of these genes may also directly or indirectly depend on malT. The expression of the maltose regulon in Vibrio cholerae affects the virulence of this organism, and mutations in malQ and malF render it less virulent (149). The products of all the MalT-dependent E. coli genes have been identified, but the function of one, the periplasmic MalM protein (104, 222), is still unclear. The observation that MalM is also present in Salmonella typhimurium may suggest that this periplasmic protein has an important function (234).

Table 1 contains many more genes that do not belong to the maltose regulon but whose encoded proteins affect the metabolism of maltodextrins or the regulation of mal gene expression at different levels. Dextrin-metabolizing or synthesizing enzymes affect the level of internal maltotriose, the inducer of the system which is recognized by MalT; cyclic AMP (cAMP)/CAP, as well as the product of the mlc gene, controls the expression of malT itself; and the levels of MalK, the ATP-hydrolyzing subunit of the transport system, as well as of MalY (a βC-S lyase) and of Aes (an esterase with homology to lipases) affect the expression of the maltose transport genes. There are some more genes which, when mutated, have a subtle effect on mal gene expression. Among them is asuE, encoding a tRNA-modifying enzyme, and treR, encoding a repressor for the trehalose-utilyzing system in E. coli. These aspects will be discussed in detail below.

Mutations in malT [malT(Con)], resulting in the constitutive expression of all mal genes, have been isolated (49, 67). These mutations center around amino acids 243 and 358 of the polypeptide chain. Whereas in an in vitro assay the wild-type MalT protein is inactive in the absence of maltotriose, the MalT(Con) proteins are active in the absence of maltotriose but still bind maltotriose with an increased affinity. All mutant proteins can still be stimulated further in their transcriptional activity by maltotriose (49). The last C-terminal 95 amino acids of MalT constitute the DNA-binding domain. In this area, MalT exhibits homology to a number of prokaryotic transcriptional activators (280) of the UhpA-LuxR family of regulatory proteins (260). Truncated forms of MalT lacking its C-terminal DNA binding domain are negatively dominant over the function of the wild-type protein, indicating that the protein interacts with itself when binding to DNA (40).

The expression of malT is not autoregulated by MalT but is subject to catabolite repression and therefore requires the presence of the cAMP/CAP complex (35, 36, 65). According to the current view, the mechanism of catabolite repression is based mainly on the regulation of intracellular cAMP levels. This, in turn, is a function of adenylate cyclase, controlled by the enzyme IIA^Glc of the phosphoenolpyruvate (PEP)-dependent sugar phosphotransferase system (PTS) in its phosphorylated (activating) and dephosphorylated (inactivating) states (191). Indeed, mutants lacking adenylate cyclase or CAP are unable to grow on maltose. It is rather informative to monitor mal gene expression in a wild-type strain during growth in rich media such as Luria-Bertani broth or tryptone broth, in the absence of external inducer. After inoculation (1:20 dilution) from an overnight culture, expression of a malK-lacZ fusion or maltose transport activity is high. It dramatically decreases (indicating a lack of new synthesis) as growth commences and reaches its lowest level during mid-log-phase growth. Since this phenomenon coincides with a cessation of cell division, it was once erroneously concluded that the synthesis of maltose binding protein (74) as well as galactose binding protein (246) might be connected to cell division. However, this phenomenon most probably reflects the concentration of internal cAMP (15), pointing once more to the importance of cAMP in catabolite repression.

Mutations in the control region of the malT gene (Fig. 1) have shown that its expression in the wild type is limited at both the transcriptional and translational levels. Mutations with increased malT expression have been isolated (malTp1, malTp7) (Fig. 1) (34). The use of deletions introduced upstream of the malT promoter revealed that the 120 bp upstream of the transcriptional start site, encompassing the binding sites for the polymerase and the cAMP/CAP complex, are sufficient for malT expression. However, deletion of DNA further upstream increased the expression of malT, indicating that a binding site for a protein reducing malT expression had been removed (200, 205). In addition, a Tn10 insertion (malT-P418::Tn10) (Fig. 1) that increases malT expression has been isolated. The position of this insertion is still upstream of the DNA region which, when deleted, increases expression (Fig. 1). Whether or not these sites are involved in the dominant negative effect of certain mutations in envZ (105, 286, 289), resulting in an overphosphorylated OmpR protein (2, 224, 291), is still unclear. Interestingly, the DNA upstream of the malT promoter whose deletion results in an increase in malT expression contains sequences that are similar to sequences identified as binding sites for phosphorylated OmpR (133, 155) (Fig. 1). Although ompR null mutants do not exhibit an increase in malT expression, mutations in envZ leading to uncontrolled phosphorylation of OmpR reduce the transcription of malT (33).

FIG. 1 Regulatory region upstream of malT. The unboxed shaded area represents the cAMP/CAP binding site for the malT promoter. The boxed shaded area represents the binding site of Mlc, the regulator of malT expression, as determined by footprint analysis. The (more ...)

Expression of malT is controlled by a repressor called Mlc. This was detected by isolating a chromosomal insertion that increased malT expression. The insertion was found to be in mlc, a known gene whose product, when overproduced, impairs the utilization of glucose (131). The increase in malT expression could still be observed in mutants carrying the deletions as well as the insertion upstream of the malT promoter, excluding these regions as binding sites for Mlc. Mlc shows homology to NagC, a gene regulator functioning as a repressor for the divergent nag operons (encoding proteins for the uptake and degradation of N-acetyl-d-glucosamine) and as an activator as well as a repressor for the glmUS operon encoding enzymes for the biosynthesis of N-acetyl-d-glucosamine (188). By footprint analysis, it was shown that Mlc binds to malT DNA at a position 1 to 23 bp from the start of the malT transcript (Fig. 1). In mutants lacking Mlc, the expression of malT is increased by a factor of 2 to 3 when grown in glycerol. On the other hand, overexpression of Mlc from a multicopy plasmid strongly reduces malT expression (68). The identity of the effector for Mlc is still unclear. Internal free glucose or glucose derived from the metabolism of disaccharides such as trehalose (141) or even maltose (28) slightly induces malT expression in a Mlc-dependent fashion. However, free glucose, even 100 mM, does not prevent DNA binding of Mlc during footprint or band shift assays. Therefore, it is likely that a metabolic product of glucose could be the controlling compound for Mlc. Mlc not only regulates malT expression but also represses the man operon (259) when overproduced (189). Mlc may thus represent a new global regulator for sugar-metabolizing systems.

The second structural feature found to be essential for MalT-dependent mal gene expression was two additional MalT boxes in a direct repeat upstream of the −38 region (206, 279, 281). The analysis of these MalT boxes also led to an extension in the consensus sequence, which is now defined as 5′-GGGGA(T/G)GAGG-3′ (281). In contrast to the orientation of the MalT box, which is proximal to the transcription initiation site and always oriented 5′ to 3′ in the direction of transcription, the orientation of the two repeats can be in either direction. This motif of MalT-dependent promoter structure with three MalT boxes has also been found by sequence analysis upstream of malS (235) and malZ (267), two MalT-dependent genes encoding maltodextrin-hydrolyzing enzymes in E. coli.

Figure 2 shows a schematic view of the arrangement of the MalT and cAMP/CAP-binding sites of the different E. coli mal gene promoters.

FIG. 2 Scheme for the promoter structures of the different mal operons. Arrows indicate transcriptional start points. Solid large arrow-shaped bullets represent MalT binding sites (MalT boxes), and open rectangles represent cAMP/CAP binding sites. The open arrow-shaped (more ...)

By using the decanucleotide 5′-GGGGAGGAGG-3′, the MalT-binding consensus box, and nucleotide sequences present in the polylinker of a vector plasmid as connecting sequences, Danot and Raibaud constructed several functional semisynthetic promoters and tested them for MalT-dependent gene expression (45). These studies confirmed the importance of the structural motif formed by two MalT-binding sites in a direct repeat. This motif is involved in promoter activation either alone or in conjunction with a third MalT-binding site proximal to the transcription start site. In this configuration, the promoters are active irrespective of the orientation of the repeat. Provided that the alignment along the axis of the helix is retained, the distance of the repeat to the proximal MalT-binding site can be varied to some extent. This analysis defines the inducible MalT-dependent promoter that does not require the cAMP/CAP complex (45). The role of these sites in the contact and activation of RNA polymerase by MalT has been assessed by mutating each site and measuring the contribution of the remaining sites (46).

Of the five known MalT-dependent operons in E. coli, two (malPQ and malZ) are independent of cAMP/CAP and two, malK lamB malM and malEFG, depend on it. The upstream region of the last mal gene, malS, does contain a sequence resembling a cAMP/CAP binding site, but whether it is functional is unclear. malK lamB malM and malEFG are oriented divergently from each other. Their transcription start sites are 271 bp apart. Located between these start sites is a 210-bp regulatory region which comprises two series of MalT boxes (three MalT boxes, including the repeat, in front of malK, and two MalT boxes, the repeat, in front of malE) separated by three cAMP/CAP binding sites. The entire control region is required for the full expression of both operons. Sequential removal of the two distal MalT boxes reduces but does not abolish the expression of malE, whereas the removal of any MalT box abolishes the expression of the malK operon (206). This suggested that multiple copies of MalT and cAMP/CAP form a unique nucleoprotein structure at this regulatory region that is required for maximal activity of both promoters. The observation that the function of these sites is sensitive to the phase of the DNA helix has led to a model in which the DNA is wrapped around the complex composed of MalT and cAMP/CAP (199, 206). The state of supercoiling also appears to be important for the effectiveness by which the two operons are transcribed. Not only is relaxed DNA less efficient in transcription initiation, but also the sites occupied by the activator complex are shifted in supercoiled relative to relaxed DNA (216).

By a series of elegant experiments, it was found that binding of cAMP/CAP in the intergenic regulatory region between malE and malK results in a repositioning of MalT binding. In the absence of cAMP/CAP, MalT binds with high affinity to the three MalT boxes upstream of the malK transcriptional start site (identified previously). In the presence of cAMP/CAP, MalT binding is shifted by three nucleotides toward the Pribnow box of the malK promoter. The cAMP/CAP effect requires the malKp-distal MalT binding sites (218). The repositioning is caused by DNA bending, which can be mimicked by replacing cAMP/CAP with the integration host factor (217). The role of the CAP and MalT binding sites in transcriptional regulation of the two divergently oriented promoters continues to be the subject of extensive studies (213).

Recently, it has been observed that Lrp, the leucine-responsive protein (32), also affects the transcription of malT as well as of some malT-dependent genes (268). By using operon fusions to malT, malE, and malK in the genetic background of strain MC4100, this Lrp dependence could not be observed in the laboratory of W.B. It has also been reported that the expression of the malK lamB malM and malEFG operons, but not of the malT gene, is controlled by the pH of the medium, being elevated at high pH (121). The mediator of this pH-dependent control is unknown.

FIG. 3 Crystal structures of the open and closed forms of MBP. (A) Structural representation of MBP in the open (top) and closed (bottom) forms. The yellow structure represents maltose bound within the binding site. Courtesy of Sherry Mowbray (247); reprinted (more ...)

MBP has been subjected to a detailed mutational analysis, and regions affecting transport (78, 264) or chemotaxis (77) have been defined. Also, intensive studies on the renaturation process of denaturated MBP (37, 96) and on the biosynthesis and secretion of the protein (37, 73, 90, 261, 272) have been published. A recent report even indicates that periplasmic binding proteins, MBP among them, exhibit properties of molecular chaperonins in the periplasm (212).

The interaction of MBP with MalF and MalG has been studied by the genetic approach of mutant and suppressor analysis (274). This study, in combination with the knowledge of the crystal structure of MBP, has led to the conclusion that one lobe of MBP interacts with MalF and the other lobe interacts with MalG (128). The starting point for this genetic analysis was the isolation of mutants (actually, two point mutations were always required in either malF or malG) that transported maltose in the absence of MBP. Surprisingly, some of these MBP-independent mutations became maltose negative in the presence of wild-type MBP, a situation that allowed isolation of mutants with suppressor mutations in malE (274). Most of these suppressor mutations in MBP were dominant negative in combination with a wild-type MalFGK₂ complex, indicating a higher affinity towards the membrane components when in the non-ATP hydrolysis-triggering mode (128, 248). Similarly, after the introduction of two cysteines (G69C and S337C) by site-directed mutagenesis into each domain of MBP, the formation of an interdomain disulfide cross-link that holds the protein in a closed conformation could be observed. This mutant MBP confers a dominant negative phenotype for growth on maltose, for maltose transport, and for maltose chemotaxis (303). The observation that wild-type MBP interferes with the transport activity of the MalFGK₂ complex of the MBP-independent mutant permitted further studies on the interaction of MBP with the MalFGK₂ complex. Transport studies with vesicles revealed that the wild-type binding protein inhibited transport only at high concentrations while it actually stimulated transport at low concentration (59). This analysis, together with the properties of dominant negative mutations in MBP, has led to the proposal that the MalFGK₂ complex must be able to attain at least two different conformations in its interaction with MBP, only one of which is able to trigger ATP hydrolysis by the MalK subunit (21, 248).

Even though the major recognition entity of the maltose transport system is the MBP, it is clear that not all substrates that are bound by the binding protein are transported. This is particularly obvious with cyclodextrins and the p-nitrophenyl derivatives of maltooligosaccharides that are bound very well by MBP but are not transported by the intact system (87, 150, 269). This is due to the modes of substrate recognition by MBP. Substrates that can be attached to MBP at the reducing end are transported; those which are bound within the dextrinyl chain (such as cyclodextrins) are not (107, 108, 173).

MBP also functions as the substrate recognition site for maltose chemotaxis (114). The sites in MBP that are essential for the interaction with the maltose transport machinery and the chemotaxis machinery are distinct but partially overlapping (77, 97, 98, 302, 303). The α helix 7 of MBP appears to be exclusively involved in the interaction with the membrane components of the transport system. Mutations in this α helix belonging to the C-lobe of MBP affect transport without affecting the binding of substrate or the interaction with the chemotactic machinery (264).

FIG. 4 Two-dimensional structure of MalF and MalG. A topological model of MalF and MalG based on the analysis of malF::phoA and malG::phoA fusions is shown. MBP-independent mutants always require two mutations, one in the p (proximal) region and one in the d (more ...)

Studies with mutations in malF and their intragenic suppressors represent the first attempts to determine the positions of the different MSS relative to each other. Thus, mutations in MSS 7 were suppressed by mutations in MSS 6 or 8, perhaps an indication that these helices are next neighbors. In addition, mutations in MSS 6 leading to altered substrate specificity occur only on one side of the helical wheel, pointing to a participation of this surface in the substrate-specific transport channel (80).

The MalF subunit of the maltose transport system of E. coli and other gram-negative enteric bacteria (44) represents something of an exception among the membrane-bound components of binding protein-dependent ABC transporters, since these MalF proteins contain a large periplasmic loop between MSS 3 and 4. This loop may not mediate the interaction with the binding protein, since all pairs of mutations in malF leading to an MBP-independent phenotype (and which can be suppressed by mutations in MBP for the dominant negative phenotype caused by wild-type MBP) are positioned outside the loop. It may be involved in the docking of MBP to the membrane components. Consistent with this view is the observation that in none of the MBP-independent mutants is docking of MBP prevented. Also, no periplasmic loop is found in the MalF homologs of binding protein-dependent maltose transport systems in gram-positive bacteria (196, 225, 276) or even in the Archaea (299), whose members contain lipid-anchored high-affinity binding proteins for maltose on the outside of the cytoplasmic membrane. In the latter case, docking is no problem because close vicinity to the membrane components is based on the lipid anchorage.

Even though it represents the major substrate recognition site, MBP cannot be the only substrate binding site of the transport system. MBP-independent mutants with mutations in MalF or MalG that retain specificity for maltose have been isolated, indicating the presence of a latent substrate binding site in the MalFGK₂ membrane complex (273). The V_max of these mutants for maltose transport can sometimes be similar to the wild-type value, while the apparent K_m of uptake is always increased by a factor of about 10³. Transport in these mutants is still active, i.e., against the concentration gradient, and dependent on the MalK-mediated hydrolysis of ATP. Characteristically, these MBP-independent mutants with mutations in MalF always carry two mutations, one in MSS 5 near the periplasmic surface, in the p (proximal) region, and the other one deep within the cytoplasmic membrane, in either MSS 6, 7, or 8, in the d (distal) region (41). It is interesting that substrate specificity mutations have been isolated in MSS 6 and that nearest-neighbor analysis predicts that MSS 6, 7, and 8 are close to each other (80) and at the same time form the d region. The attractive concept that mutations in the p region affect binding-protein recognition and mutations in the d region affect the coupling to the ATPase activity mediated by the associated MalK subunit apparently does not hold true, however. Both mutations are required for the uncoupled MalK-ATPase phenotype.

The screening of mutations in malF and malG for the ability to transport other sugars revealed that one mutation, malF515, which changes Leu334 to Trp (L334W) on the periplasmic side of MSS 5, permitted the transport of lactose in addition to maltose. MalK as well as MalG is required for lactose transport, as is MBP. The requirement for MBP is particularly intriguing because MBP clearly does not bind lactose. Therefore, these results are consistent with the idea that the unliganded form of MBP interacts with the membrane complex (160). The position for such a “specificity” mutation is surprising since it is on the periplasmic side of MalF, which is not expected to form the substrate recognition site. This may indicate that substrate specificity for lactose is already present in MalF or MalG but needs the proper opening of the gate for the substrate to enter. In a different construct, large portions of the MalF protein have been deleted (beginning from the C terminus up to MSS 1). These deletion mutants are able to transport lactose in a MalG-, MalK-, and MBP-dependent manner without an additional mutation (159). This may indicate that the substrate recognition site of the membrane complex for lactose resides in MalG.

Little is known about the mechanism of energy coupling of ATP hydrolysis to the accumulation of maltose. Since neither the substrate nor any of the proteins involved have been seen to become phosphorylated during transport, it has become commonplace to interpret energy coupling as the transfer of the energy gained through ATP hydrolysis to protein conformational energy in MalF and MalG, followed by its subsequent binding protein-triggered release, resulting in the unidirectional translocation of the substrate through the membrane. Mutants with mutations in malF or malG that no longer require the triggering by substrate-loaded MBP have been isolated. It is noteworthy that the MalFGK₂ complex of these mutants no longer requires the presence of substrate-loaded MBP to stimulate the ATPase activity of MalK (57). ATPase activity has become uncoupled in these mutants, similar to the uncoupled ATPase activity of the purified wild-type MalK subunit (166). This has led to the notion that the transport system is a signal transduction pathway that begins in the periplasm with the recognition of maltose by the binding protein and ends with the control of MalK-ATPase activity in the cytoplasm (41, 57). The helical domain of the ATPase in MalK is likely to be involved in the signal transduction between MalF and MalG (167).

No quantitative studies on the synthesis of MalK in comparison to its partners in the membrane, MalF and MalG, are available. Even though the stoichiometric composition of the biochemically active complex has been determined as MalFGK₂ (55), it is unclear whether MalK is synthesized in excess or, depending on the regulatory conditions, in varying amounts. The latter possibility is not unlikely since malF and malG are located in a different operon from malK and since substantial amounts of MalK have been found, in contrast to MalF and MalG (7, 251). As discussed below, MalK plays an important role in the MalT-dependent control of mal gene expression. It is only natural to see MalK not only as an energy module to drive transport but also as a control unit of transport activity. Obviously, the availability of ATP should influence transport activity. However, since the K_m of MalK for ATP is in the micromolar range (58) and the physiological concentration of ATP is in the millimolar range, the cells would presumably have to become quite exhausted of ATP before they stopped transporting maltose.

We would like to postulate an additional level of control for the function of MalK in transport, which works by regulating the affinity of MalK for its partners, MalF and MalG. We observed that transport of glycerol-3-phosphate by the binding protein-dependent Ugp transport system is inhibited by internal P_i (29). The Ugp system exhibits a surprising degree of sequence similarity to the maltose system in all subunits in spite of the differences in substrate specificity and regulation (178). In an attempt to find whether UgpC, the MalK analog of the Ugp system, was the target of P_i inhibition, we exchanged MalK with UgpC (115) and tested a possible inhibition of maltose transport by P_i. Although P_i did not inhibit the UgpC-mediated uptake of maltose, it also failed to inhibit the uptake of glycerol-3-phosphate via the Ugp system when UgpC was overproduced. Apparently, the reduced affinity of UgpC for its cognate membrane partners in the presence of P_i can be compensated for by increasing the UgpC concentration. It suggests the possibility that the degree of association of the ATP-hydrolyzing subunit with the membrane components may control transport activity. An equivalent finding in the histidine transport system in Salmonella typhimurium has been interpreted in the same way. High concentrations of internal histidine were able to inhibit transport of histidine in trans (152). These observations may indicate that the transport activity of ABC transporters is controlled from the inside via the ATP-hydrolyzing subunit by binding the accumulated substrate.

The MalK subunit is usually pictured as being peripherally membrane associated through binding to the intrinsic membrane proteins MalF and MalG. This membrane association of MalK may in fact be more intimate (296). Studies with the functionally related binding protein-dependent histidine transport system of Salmonella typhimurium have shown that HisP, the MalK analog of this system, might actually be accessible from the periplasm (6), even though the interaction of the cognate periplasmic binding protein, HisJ, takes place with the intrinsic membrane proteins of the system and not with HisP (192). Similar findings of accessibility of MalK on the periplasmic side of the membrane have been reported for MalK in S. typhimurium (236). Models of the transport mechanism mediated by binding protein-dependent ABC systems picture the input of energy by ATP hydrolysis as conservation of a strained protein structural conformation in MalF and MalG that is released by triggering with substrate-loaded binding protein (21, 59, 250). The analogy of the repetitive movement of SecA through the membrane that has been postulated as crucial for Sec-dependent protein secretion (140, 293) to a possible repetitive movement of a subdomain of MalK through the membrane (236) is less convincing. The transported molecule would be far smaller than the moving machinery, and the moving part of MalK should then exhibit substrate specificity for the transported molecule, which has never been observed. In contrast, UgpC, the ATP-hydrolyzing subunit of the glycerol phosphate transport system, can complement malK mutants to transport maltose (115), indicating that the ATPase subunit does not carry substrate specificity for the transported molecule.

A number of studies have analyzed the function of LamB (maltoporin) in the diffusion of maltodextrins (16, 88, 92, 136, 154, 171). All these studies demonstrate the presence of a maltodextrin binding site that is essential for the facilitated diffusion process of maltodextrins. A major contribution to understanding the molecular basis of transport through porin channels has come from determination of the three-dimensional structure of LamB, which was done in the presence of a series of maltooligosaccharides. These studies showed that each subunit of the trimeric protein contains a wide channel formed by an 18-stranded (230), antiparallel β-barrel. Three inwardly folded loops contribute to a constriction about halfway through the channel (231) (Fig. 5). The crystal structures of maltoporin complexed with maltose, maltotriose, or maltohexaose reveal an extended binding site within the channel. The maltooligosaccharides are in apolar van der Waals contact with the “greasy slide,” a hydrophobic path that is composed of aromatic residues and is located at the channel lining (79, 287). Interestingly, the LamB protein contains two of four sequences (235) that are conserved in amylases and form the maltodextrin binding sites there (282). One of them, FYQRHD, at positions 106 to 111 of LamB, is actually part of the sugar binding site as determined by X-ray crystallography (79). A similar structure for the maltoporin from S. typhimurium, including the prominent greasy slide for maltodextrins, has been reported (162).

FIG. 5 Crystal structure of LamB. A schematic drawing of the λ-receptor (maltoporin) monomer is shown. The cell exterior is at the top, and the periplasmic space at the bottom. The area of the subunit in trimer contacts is facing the viewer. The 18 antiparallel (more ...)

Calculations show that the diffusion of maltose through LamB and the following uptake by the ABC transporter are closely matched at maltose concentrations in the micromolar range. The V_max for the disaccharide maltose in a fully induced strain has been measured to be 20 nmol per min per 10⁹ cells (265). The diffusion of maltose through the outer membrane, mediated by about 10,000 trimeric LamB porin molecules per cell, is apparently not yet limiting at micromolar maltose concentrations, the K_m of the transport system (27). We calculated the diffusion of maltose for different external concentrations through these 30,000 pores with an approximate diameter of 1 nm, using a diffusion constant of 10⁻⁵ cm²/sec and an outer membrane thickness of 6 nm. We found that a rate of diffusion of 20 nmol per min per 10⁹ cells, the V_max of the fully induced ABC transporter, is reached at 0.1 μM. Considering that not all maltose molecules that reach the external face of the lambda receptor will actually enter the pore, this concentration is most probably a lower limit, and in reality the concentration might be closer to 1 μM. Therefore, one can conclude that the overall uptake of maltose at its K_m is close to the maximal rate of diffusion through the outer membrane. The validity of this conclusion has been experimentally tested with mutants containing reduced amounts of lambda receptor (27). In other words, a transport system that requires a V_max of about 20 nmol per min per 10⁹ cells and is equipped with at least equal amounts of a specific diffusion pore to those in the fully induced lambda receptor cannot exhibit an apparent K_m considerably lower than 1 μM, in spite of a periplasmic binding protein with a considerably lower K_D of binding. In contrast, transport systems with an inherently low V_max, such as binding protein-dependent amino acid transport systems, may very well approach in their K_m of transport the K_D of the corresponding binding protein.

Incoming maltose and maltodextrins of up to seven glucose moieties are metabolized to glucose and glucose-1-phosphate by the combined action of three cytoplasmic enzymes, amylomaltase (MalQ), maltodextrin phosphorylase (MalP), and maltodextrin glucosidase (MalZ) (Fig. 6). Since maltodextrins larger than six glucose moieties are not very well transported by the ABC transporter, they are reduced in size by a periplasmic amylase, the MalS protein.

FIG. 6 Maltose degradation by the maltose enzymes. The enzymes amylomaltase (malQ), maltodextrin phosphorylase (malP), and maltodextrin glucosidase (malZ) are indicated by their genes. After transport of maltose by the binding protein-dependent ABC transporter, (more ...)

It is important to stress that maltose, strictly speaking, is not a substrate of amylomaltase but only an acceptor in the transfer reaction catalyzed by the enzyme. Thus, it follows that for maltose degradation, the cell has to be able to internally produce small amounts of maltodextrins as primers with the minimum size of maltotriose. Amylomaltase is essential for maltose degradation, and malQ mutants are unable to grow on maltose. Not only are malQ mutants Mal⁻, but also their growth is inhibited by maltose (127). Under these conditions, they accumulate large amounts of free maltose inside the cell (265). In the presence of an alternative carbon source, mutations arise that are found preferentially in malT or in malK. Why no mutations in other genes arise whose products are essential in transport remains a mystery (241). Even in the absence of external maltose, malQ mutants contain significant concentrations of free glucose, maltose, maltotriose, and larger maltodextrins, provided that the strain contains the glycogen-synthesizing enzymes (69, 81). Therefore, within the cell, glycogen must continually be degraded to maltodextrins, which cannot be repolymerized (under production of glucose) in the absence of amylomaltase.

The enzymatic reaction of amylomaltase can be exploited for the convenient synthesis of uniformly labeled maltodextrins (maltotriose and larger) of high specific activity starting from uniformly labeled maltose. When amylomaltase is incubated, even after extensive dialysis with radiochemically pure [¹⁴C]maltose as the only substrate, [¹⁴C]glucose and ¹⁴C-maltodextrins are formed immediately. They can be separated and purified by chromatographic procedures and exhibit the same specific radioactivity (with respect to the glucosyl moiety) as the initial maltose. The primers used for this reaction are present in minute amounts, bound to the enzyme (180).

Obviously, it is important that maltodextrin phosphorylase does not attack maltotetraose and maltotriose, since dextrins of a minimum size are required for full activity of amylomaltase. Notably, glycogen phosphorylase, the glgP-encoded enzyme, supposedly involved in the degradation of glycogen and linear dextrins (300), cannot replace maltodextrin phosphorylase in the utilization of maltose and maltodextrins as a carbon source.

The reaction catalyzed by maltodextrin phosphorylase is, of course, reversible. The rate in the phosphorolysis direction will be stimulated by increasing concentrations of cytoplasmic P_i. In turn, the concentration of internal P_i will vary depending on the availability of external P_i and phosphorus-containing organic compounds, as well as the state of induction of the phoB-dependent pho regulon (298). There are other connections between the mal and the pho regulon. Aside from the sequence similarity between the proteins of the maltose- and phoB-dependent Ugp transport systems for glycerol-3-phosphate (178) and the functional exchangeability between their ATP binding subunits, MalK and UgpC (115), there is a common response (i.e., repression) of the two systems to dominant mutations in envZ, resulting in overphosphorylation of OmpR, the response regulator of the two-component regulatory EnvZ-OmpR system involved in osmoregulation (33).

In addition, the utilization of a fermentable carbon source such as glucose, trehalose, or maltose at P_i concentrations below 1 mM requires the derepression of the pho regulon. phoB mutants do not turn deep red on MacConkey indicator plates (0.3 mM P_i) unless 5 mM P_i is added. Also, whereas strains that turn red on these plates derepress the pho regulon, nonfermenters that appear pale remain repressed for the pho regulon. Apparently, the utilization of a fermentable carbon source requires higher cellular P_i concentrations than does that of a noncarbohydrate carbon source present in MacConkey plates (111).

malQ mutants cannot grow on maltose or maltotriose. This is somewhat surprising since MalZ should release glucose (which could be used as a carbon source) from maltotriose. Most probably, the simultaneous accumulation of maltose resulting from the MalZ-dependent hydrolysis of maltotriose in the malQ mutant is toxic. By selecting malQ mutants to grow on maltose, a mutation in malZ (malZ292) was obtained. Surprisingly, the purified mutant MalZ enzyme is not able to hydrolyze purified maltose and shows the same maltodextrin-hydrolyzing activity as the wild-type enzyme. However, by using [¹⁴C]glucose in the presence of unlabeled maltotriose or maltotetraose, the mutant enzyme, unlike the wild-type enzyme, is able to transfer maltodextrins onto the labeled glucose. The mutant MalZ enzyme has acquired a quality which is similar to that of amylomaltase. However, amylomaltase is strictly a transferase, which is not true for the mutant MalZ enzyme. The latter also remains a hydrolase, allowing the net hydrolysis of maltose to glucose in the presence of maltodextrins (185). As in the case of amylomaltase, the utilization of maltose by the mutant MalZ enzyme in vivo requires the endogenous formation of maltodextrins that can be used as primers. The first step in the catalysis performed by amylomaltase and the MalZ enzyme is identical, namely, release of glucose. However, whereas in the case of amylomaltase the maltodextrinyl residue can be transferred only onto the C-4 carbon hydroxyl group of the nonreducing glucose residue of the acceptor molecule, in the case of MalZ it can also be transferred to water.

The malZ gene has been found and characterized as a typical mal gene that is dependent on MalT (235, 267). Recently, we found that malZ expression can also occur in the absence of MalT, in particular under conditions of high medium osmolarity. Using the method of primer extension with reverse transcriptase, we found that besides the MalT-dependent transcript described first (235), malZ is also transcribed into a second mRNA, originating upstream of the start site of the main promoter (145). The significance of this phenomenon is unclear.

pgm mutants lacking phosphoglucomutase activity are able to grow on maltose, even though seemingly only one half of the maltose molecule can be used as a carbon and energy source. These strains exhibit a maltose blue phenotype (1, 220) caused by the massive production of dextrins due to the accumulation of glucose-1-phosphate and the reversal of the maltodextrin phosphorylase reaction. The problem created by the accumulation of maltodextrins is eased by the action of maltodextrin glucosidase, the MalZ enzyme, which produces glucose from the maltodextrins, followed by glucokinase-dependent phosphorylation to glucose-6-phosphate. Indeed, pgm malZ double mutants are unable to grow on maltose. Similarly, pgm mutants can grow on galactose (1) but only when all the enzymes of the maltose system, including MalZ, are present (69). pgm mutants, even null mutations created by insertion elements (153), still show residual Pgm-like activity when the enzyme activity is assayed by coupling the formation of glucose-6-phosphate from glucose-1-phosphate to the NADP⁺-dependent oxidation by glucose-6-phosphate dehydrogenase. This activity is caused by a sugar-phosphate transferase which is able not only to hydrolyze sugar phosphates but also to transfer the phosphate moiety to another sugar molecule (258). Thus, in the presence of free glucose, glucose-1-phosphate can be transformed to glucose-6-phosphate, mimicking phosphoglucomutase activity.

Figure 6 summarizes the pathway by which, according to our view, maltose is degraded into glucose and α-glucose-1-phosphate by the maltose-specific enzymes.

To explain how MalK acts as a repressor, it has been argued that the protein could be an enzyme that degrades the inducer (31). This possibility was attractive, since malT(Con) mutants, which require less or no inducer (49), are resistant to the repression mediated by overproduced MalK. However, the synthesis of maltotriose, the inducer of the maltose system (202), is not disturbed by overproduction of MalK and does not lead to induction under these conditions (69). Thus, MalK does not seem to act on the level of inducer degradation. A second possibility is that MalK, as long as it is not engaged in transport, interacts with MalT and inactivates it. MalT has been proposed to exist in an equilibrium of two forms, only one of which is acting as a transcriptional activator. Thus, when stabilized by inducer, MalT would be able to act as a transcriptional activator (66), whereas when stabilized by MalK, it would be inactive. Hence, MalK and the inducer would compete for MalT, shifting the equilibrium either with maltotriose to the inducing conformation or with MalK to the inactive form, causing repression. The first evidence for such a scheme comes from the laboratory of Shuman (183). A biotinylation site was introduced at the C terminus of MalT, leading to a biotinylated MalT (MalT-BCCP). Avidin-coated beads were then used to attach MalT-BCCP on their surface. These beads were able to bind MalK, demonstrating an interaction between MalK and MalT. Also, the phenotype of strains harboring MalT-BCCP is in line with the consequences of a MalK-MalT interaction. The presence of MalT-BCCP is negatively dominant over wild-type MalT. Overproduction of MalK abolished the dominant negative phenotype, indicating the titration of MalT-BCCP by MalK (183).

As discussed above, MalK forms dimers in vivo, as indicated by fusions to a truncated λ repressor (which is unable to dimerize). MalK mutants that are unable to repress mal gene expression no longer show a tendency to form dimers in vivo (68). This may indicate that the dimeric form of MalK is the one that interacts with MalT.

Figure 7 shows a model of how MalK could mediate repression. MalK is pictured as a protein that is able to shift between a state of association with MalFG and a state of association with MalT. In the presence of substrate to be transported, MalK would shift to the transport conformation allowing MalT to function as mal gene activator. In the absence of substrate to be transported, MalK would be free to interact with MalT and cause repression. The model is reminiscent of findings with the proline utilization system, where PutA, the proline dehydrogenase, functions as a membrane-bound enzyme in the presence of proline and as a cytoplasmic repressor in its absence (168).

FIG. 7 MalK cycle of transport and repression. A proposed model for the function of MalK in transport and regulation is shown. When maltose (solid circle) is present in the medium, it is recognized by MBP (MalE) and then transported through the membrane via (more ...)

Two observations support such a MalK cycle. First, there is a particular mutation in malK (malK941) (147) which leads to a superrepressor MalK protein. This mutation is near the ATP binding site. Labelling of the mutant protein with radioactive 8-azido-ATP is still possible; therefore, binding of ATP, but no ATP hydrolysis, probably also takes place. This would mean that the absence or presence of ATP hydrolysis determines the quality of MalK as a repressor and that ATP bound to MalK is required for repression. Second, mutations in malF and malG that result in high constitutive levels of MalK ATPase activity (41, 57) show a high level of mal gene expression in the absence of inducer, presumably because the ATP-hydrolyzing form of MalK is mimicking transport activity and is unable to repress MalT activity (183).

The MalK protein is unusually long in comparison to other ATP-hydrolyzing subunits of binding protein-dependent ABC transport systems; the difference is the presence of a C-terminal extension. On the other hand, UgpC, the corresponding subunit of the Ugp system, exhibits the same C-terminal extension as MalK and shows sequence similarity to MalK except at its C-terminal extension. MalK and UgpC can be exchanged between the two systems and can complement the transport defect of the heterologous system, but only in the absence of the cognate subunit (115), indicating a lower affinity in forming the heterologous complex. A similar exchange of MalK from S. typhimurium by LacK, the ATP-hydrolyzing subunit of a binding protein-dependent ABC transporter for lactose in Agrobacterium radiobacter, has been reported (296). Similarly, CymD, the ATP-hydrolyzing subunit of the cyclodextrin transport system of K. oxytoca, can substitute for MalK in E. coli (179).

Considering the additional segment at the C terminus of MalK, where all the regulation-negative mutations have been found (147), it was of interest to find whether UgpC, which carries the same C-terminal extension as MalK, would also be active as a repressor. While the overproduction of UgpC had no repressing effect on mal gene expression, it did reduce the expression of the remaining ugp genes (271). Thus, repression exerted by an ATP-hydrolyzing subunit of a binding protein-dependent transport system is not a peculiarity of MalK but may represent a novel form of regulation occurring in this type of transport system.

The MalK subunit is also the target of PTS-mediated inducer exclusion that is brought about by the interaction of nonphosphorylated EIIA^Glc with MalK (60, 147, 275). This interaction leads to a reduction in V_max without changing the K_m of the transport system (60). This is reminiscent of the effect of the UgpC-mediated inhibition of the Ugp transport system by internal P_i (29, 298), which we explain by the dissociation of UgpC from its protein partners in the membrane. Thus, it appears that the state of MalK (membrane bound or cytosolic) is important for its function in transport and regulation.

MBP-dependent ABC transport systems have also been found in gram-positive bacteria (102, 196), in thermophilic bacteria (120, 225), and in a hyperthermophilic archaeon (130, 299). The canonical composition of these transport systems is the same as in gram-negative bacteria, even though the respective binding proteins are anchored in the membrane by a diacylated glycerol linked via a thioether to the N-terminal cysteine of the binding protein. Surprisingly, in many of these systems that have been analyzed by sequencing, the genes encoding the MalEFG analogs are clustered in the same orientation as in the E. coli maltose system whereas the gene encoding the corresponding ATP-hydrolyzing enzyme is localized elsewhere on the chromosome or has not been found yet (225). In Streptomyces, the same ATP-hydrolyzing subunit appears to serve at least two different transport systems (233). These observations, together with the finding that ATP-hydrolyzing subunits can (under certain conditions) be functionally exchanged among heterologous systems, indicate that this component may have been the first one to appear during evolution, with additional proteins determining altered substrate specificity arriving later in evolution. Also, it would not be surprising if, in the future, several “ancient” systems that share their ATP-hydrolyzing subunits were found.

The βC-S lyase activity of MalY is not the cause of the repression of the mal genes. A mutation in the pyridoxal phosphate binding site of MalY was constructed that resulted in the loss of βC-S lyase activity, even though the protein was still active as a mal gene repressor. In addition, overproduction of MetC had no effect on the regulation of the mal genes (301). As in the case of MalK, mutants with mutations in MalY that still exhibit nearly unchanged enzymatic activity but have lost their ability to repress the maltose system can be isolated. The clustering of these mutations favors a mechanism of protein-protein interaction with MalT as causing repression. There are several indications to suggest that MalY acts by the same mechanism as MalK: malT(Con) mutants are largely insensitive to the effect of either MalK or MalY. Also, when maltose enters the cell via a constitutive transport system for trehalose (see below), an inducer that renders MalT insensitive to overproduction of MalK as well as MalY can be formed. On the other hand, the inhibiting effect of MalY can still be observed in mlc mutants that have elevated levels of MalT. Therefore, the effect of MalY and MalK on the activity of MalT depends on the amount of MalT and on the presence of inducer. A high inducer concentration (or the malT(Con) mutation) is more effective than an increase in the concentration of MalT.

The relatively high “uninduced level” of mal expression in E. coli when the cells are grown without maltodextrins in the medium (i.e., with glycerol) points to the endogenous synthesis of the inducer. Several lines of evidence support this conclusion: first, malK-lacZ fusions (lacking MalK function) display a much higher β-galactosidase activity than does the corresponding malK⁺ merodiploid strain. This shows not only that MalK acts as a repressor but also that MalT is activated by internal sources. Second, as outlined above, maltose metabolism initiated by amylomaltase needs a maltodextrin primer with the minimal chain length of maltotriose. Thus, in the absence of endogenous maltodextrins, E. coli could not grow on maltose. Third, mutants with mutations in enzyme II^Glc (ptsG), enzyme II^Man (ptsM), or glucokinase (glk) can be induced by glucose entering the cell in its unphosphorylated form via the galactose permease. This indicates that the inducer is formed from glucose. The induction by glucose does not occur in a mutant also lacking phosphoglucomutase (pgm). Such a strain exhibits a very low level of mal gene expression (in comparison to its pgm⁺ parent) as measured by transport activity (69). Since the induction-preventing mutation is in pgm, it is clear that the synthesis of inducer can arise from gluconeogenesis. The only inducer established so far to be able to activate MalT in vitro is maltotriose (202). Therefore, the cell must have the ability to synthesize maltotriose or another sugar with inducer capabilities. There are at least two ways in which maltotriose can be synthesized endogenously.

It appears that even in the absence of trehalose metabolism or when the cells are growing on glycerol and no glucose is taken up from the medium, the mal genes are endogenously induced by the formation of small amounts of maltotriose (or another unidentified inducer). In wild-type strains, the basal expression of the mal genes is low, curbed by the action of the MalK protein, but it becomes significant in malK-lacZ fusion strains lacking MalK function. Since this constitutivity is also seen in strains defective in glycogen synthesis, it is obvious that the endogenous inducer can also be derived from gluconeogenesis without bringing in glucose from the medium. These observations have led us to formulate a scheme for the synthesis of the internal inducer that is based on the function of an as yet unidentified maltose/maltotriose phosphorylase. This enzyme should reversibly form maltose (and maltotriose) from glucose (or maltose) and glucose-1-phosphate (22). The formation of these maltodextrins and therefore the postulated enzyme(s) for their synthesis are independent of any maltose enzyme, since maltotriose is created in malT mutants as well (69). The gluconeogenetic origin of glucose-1-phosphate seems clear; less clear is the origin of free internal glucose. There are several possibilities. The first is the hydrolytic function of a phosphoglucotransferase, an enzyme that was purified some time ago but whose gene has not yet been identified. This enzyme transfers the phosphate moiety from glucose-phosphate not only to another sugar but also to water, thus forming free glucose (258). The other possibility is the release of free glucose from UDP-glucose by analogy to the release of galactose from UDP-galactose (297). Indeed, galU mutants lacking UDP-glucose pyrophosphorylase are somewhat impaired in the constitutivity of a malK-lacZ fusion.

If the model of inducer synthesis is correct, pgm mutants should be strongly reduced in their basal expression of the maltose genes and uninducible by trehalose. This is exactly what is observed when measuring maltose transport of pgm mutants grown in glycerol or, even more dramatically, grown on trehalose. However, malK-lacZ (Mal⁻) fusions are only weakly (two- to threefold) reduced in their constitutive expression when tested in pgm mutants. Similarly, the induction of a malK-lacZ fusion (be it Mal⁺ or Mal⁻) by trehalose is only weakly reduced in pgm mutants. At present, there is no reasonable explanation for the difference in these two tests for mal gene expression.

This is in sharp contrast to the situation for cells growing logarithmically in batch cultures at high glucose concentrations. Under these conditions, transport of glucose via the PTS exerts strong catabolite repression, blocking the transcription of malT and the malT-dependent genes of the malK lamB malM and malE malF malG operons. In addition, transport of maltose is diminished by inducer exclusion. These effects are caused by the interaction of unphosphorylated enzyme IIA^Glc of the PTS with MalK (60, 147, 275).

Since only the uninduced level of mal gene expression is affected by osmolarity, one may argue that it is the synthesis of endogenous inducer that is inhibited under conditions of high osmolarity. Most recently, we found that the presence of glucokinase is necessary for the effective repression of uninduced or malK-dependent constitutive mal gene expression. Since the expression of glk at high osmolarity remains unchanged (as measured by a translational glk-lacZ fusion), the most likely explanation is a stimulation of glucokinase activity at high osmolarity. In strains carrying ptsM and ptsG mutations in addition to the glk mutation, no osmoregulation of mal gene expression is observed. This demonstrates that osmoregulation (i.e., repression) of mal gene expression is caused by the removal of internal free glucose.

Maltose and trehalose are disaccharides composed of two α-glycosidically linked glucose residues, but their structures are quite different. Whereas maltose contains an α(1→4)-glycosidic linkage and exhibits a reducing end glucose, trehalose contains two α(1→4)-glycosidic linkages and no reducing sugar residue (Fig. 8). Due to the particular symmetrical and backfolded structure of this nonreducing sugar, trehalose, in contrast to maltose, has properties that make it an excellent agent to preserve protein and membrane structures (42). It has been observed that trehalose is synthesized as a protecting agent or osmolyte in many cell types including E. coli under conditions of dehydration as well as high osmolarity (137). Thus, it is not surprising that E. coli has developed two entirely different systems for the metabolism of maltose and trehalose, both of which are excellent carbon sources for this organism. Unlike maltose, trehalose uses a PTS-dependent enzyme II^Tre (encoded by treB) for its uptake as trehalose-6-phosphate (142) followed by its internal hydrolysis by trehalose-6-phosphate hydrolase (encoded by treC) to glucose and glucose-6-phosphate (219). treB and treC form an operon that is controlled by the adjacent treR gene, encoding a repressor (129). The internally acting inducer of the system is trehalose-6-phosphate, which inactivates TreR. There is one complicating fact: in addition to its degradative abilities, E. coli is able to synthesize internal trehalose at high osmolarity. The two necessary enzymes are trehalose-6-phosphate synthase (encoded by otsA), catalyzing the transfer of glucose from UDP-glucose to glucose-6-phosphate, thus forming trehalose-6-phosphate, and trehalose-6-phosphate phosphatase (encoded by otsB), catalyzing the hydrolysis of trehalose-6-phosphate to free trehalose (137). otsB and otsA form an operon that is induced at high osmolarity in an RpoS-dependent manner (119). The expression of ostB otsA and of treB treC is mutually exclusive in that high osmolarity prevents the induction of the treB treC operon. Despite these entirely different pathways of maltose and trehalose uptake and metabolism, there are curious links between the two systems.

FIG. 8 Structures of maltose and trehalose. (A) Maltose (α-d-glucopyranosyl-1→4-glucopyranose) in its predominant α enantiomeric form. (B) Trehalose, shown as the naturally occurring α-d-glucopyranosyl-1→1-α-d (more ...)

One may ask what purpose these different regulatory circuits have that govern the expression of the maltose system. There are two main branches of regulation: one is stimulatory, represented by endogenous induction and cAMP/CAP dependency, and the other is repressive, represented by MalK and other proteins that inactivate the central gene activator. From the blueprint of the high-affinity and binding protein-dependent transport system that is also connected to the cell’s chemotactic machinery, it appears that the maltose system is geared for the scavenging of low levels of maltose and maltodextrins. Even before the arrival of the true substrate, the system is becoming prepared by endogenous induction, particularly when fasting on glucose or other carbohydrate carbon sources. However, only in the presence of the true substrate is full-fledged induction developed. It makes perfect sense that the function of holding back full induction should be connected to a component of the transport system itself. In this way, the system responds not only to the presence of substrate by induction but also to loss of substrate by repression. This could in principle also be achieved by a gene regulator that is an activator in the presence of the inducer and a repressor in the absence of the inducer (the ara system is the classical example [232]), but apparently, a finer tuning results from the regulation seen in the maltose system. That the transport step itself is connected to regulation is not unprecedented. We find this in the regulation of the pho regulon by the Pst transport system (288); in the Bgl system, where an antiterminator is controlled by its cognate phosphotransferase system (3); and in the regulation of the Fec system by the transport of ferric citrate through its cognate outer membrane receptor FecA (139).

Aside from these two main streams (endogenous induction and MalK-dependent repression), there appear to be several additional regulatory interconnections such as the existence of a repressor for the activator or the inhibition of the activator by seemingly unrelated enzymes (MalY or Aes). The characteristic feature of these effectors is that they have hardly any recognizable effect under normal physiological conditions but may become important under extreme growth conditions. It is possible that the maltose system is nothing special in this respect, and several more “classical” systems that show this kind of network regulation may be discovered.

We are indebted to Armin Geyer (Department of Chemistry, University of Konstanz), who produced Fig. 8. We thank Sherry Mowbray, Tilman Schirmer, and Florante Quiocho for their help in providing Fig. 3 and 5. We thank Erika Oberer-Bley for her help in preparing the manuscript.

Work done in the laboratory of W. Boos was supported by grants from the Deutsche Forschungsgemeinschaft (SFB156 and Schwerpunkt Netzwerkregulation in Bakterien). Work in the Shuman laboratory was supported by NIH grant GM51118. The collaboration of the two laboratories was supported by the Max-Planck-Forschungspreis 1991.