Chapter 70

Membrane-Derived Oligosaccharides (Periplasmic Beta-d-Glucans) of Escherichia coli

EUGENE P. KENNEDY

DISTRIBUTION OF PERIPLASMIC GLUCANS IN GRAM-NEGATIVE BACTERIA

In the first detailed investigation of the biosynthesis of membrane phospholipids in Escherichia coli, Kanfer and Kennedy (33) measured the turnover of the principal membrane phospholipids under conditions of steady-state, logarithmic growth. Pulse-chase experiments with ³²P revealed that radioactivity was continuously lost from the hydrophilic head group of phosphatidylglycerol, suggesting some dynamic function of this phospholipid. In a further investigation of this phenomenon, van Golde et al. (64) labeled cells of E. coli K-12 with [2-³H]glycerol and discovered that the phospho-sn-1-glycerol head group of phosphatidylglycerol, labeled either with ³²P or [2-³H]glycerol, was continuously transferred as a unit to novel water-soluble oligosaccharides which, from their relation to the metabolism of membrane phospholipids, were called membrane-derived oligosaccharides (MDO).

Recent discoveries in rather diverse fields of microbiology have led to the realization that the MDO of E. coli are typical representatives of a class of cell constituent, the periplasmic β -d-glucans, widely distributed in gram-negative bacteria (Table 1). Glucans of this kind have not been found in gram-positive bacteria, which lack a well-defined periplasmic compartment. The periplasmic glucans have important but still poorly understood functions in osmotic adaptation and in cell signaling that make them the objects of increasing interest and research.

Table 1Periplasmic ?-D-glucans of gram-negative bacteria.

As is the case with other bacterial cell surface carbohydrates, such as lipopolysaccharides and cell capsules, the periplasmic β-glucans vary considerably in details of chemical structure from species to species. The following general features, however, indicate their close structural, metabolic, and functional relationship throughout the range of gram-negative bacteria in which they have been reported thus far.

(i) These glucans are localized in the periplasm. This has been established for the β-d-glucans of E. coli (57), Rhizobium trifolii (1), Agrobacterium tumefaciens (46), and Bradyrhizobium japonicum (44). High levels of cyclic β-glucan are sometimes also excreted into the growth medium. Excretion involves the function of the ndvA gene of Rhizobium (60) or the homologous chvA gene (11) of Agrobacterium species, but little is known about the mechanism by which the glucans enter the periplasm or are excreted into the medium. There is some evidence that cyclic glucans appear first in the periplasm and are later released into the growth medium (21). (ii) Periplasmic glucans contain d-glucose as the sole sugar, linked by β-d-glucosidic bonds. The sole exception appears to be a cyclic β-1,2-glucan from a strain of Xanthomonas reported to contain also a single alpha-linked glucose residue (2). (iii) Periplasmic glucans are substituted to varying degrees with phospho-sn-1-glycerol, phosphoethanolamine, and phosphocholine residues derived metabolically from the head groups of membrane phospholipids. In addition, these glucans are often also substituted with succinic (64) and methylmalonic (28) acids in O-ester linkage.

FUNCTIONS OF PERIPLASMIC GLUCANS

There is a broad similarity in the functions so far assigned to the periplasmic glucans in the varied organisms in which they are found. These functions are of two principal types: osmotic adaptation and cell signaling.

Periplasmic Glucans and Osmotic Adaptation

To live and grow, cells require soluble substances in the cytoplasm with a minimum total concentration of about 300 mosM (17, 27). Because plasma membranes are freely permeable to water, all cells face a fundamental problem of osmoregulation from challenges in the environment of two distinct types.

In medium containing solutes at higher concentrations than about 300 mosM, cells will undergo shrinkage (plasmolysis) because of movement of water out of the cell, down its gradient of activity. Bacterial and plant cells respond to this type of osmotic stress by increasing their intracellular content of potassium (17) or glucose and its derivatives (52), and also accumulate high levels of a variety of other substances that have been termed "compatible solutes." The higher level of osmolarity of the cytosol caused by the accumulation of these solutes balances that of the medium, preventing decrease of cell volume. Adaptation of this type to hyperosmolarity is particularly important for plants growing in surroundings of high salinity or during periods of drought, and for bacteria to withstand desiccation.

A second, and less widely studied, type of osmotic challenge occurs when cells with their minimum cytoplasmic osmolarity of about 300 mosM find themselves in very dilute medium. Under these conditions, water must flow into the cytoplasm, until at equilibrium its flow is resisted by a hydrostatic pressure of about 6.4 atm (ca. 648.9 kPa). It is well established that the peptidoglycan layer of the bacterial cell envelope plays an essential role in maintaining the structural integrity of the cell against such pressures. Lesions in the peptidoglycan caused by treatment of the cells with lysozyme in the presence of EDTA, or by treatment of growing cells with penicillin, cause the swelling and lysis of cells in dilute medium.

Many species of gram-negative bacteria respond to hypoosmotic challenge by synthesizing large amounts of periplasmic glucans. This type of osmotic adaptation was first observed in E. coli (34) with maximum production of cell-associated glucan occurring during growth in medium of low osmolarity. This is also the case for species of Agrobacterium (46), Rhizobium (69), and Bradyrhizobium (43).

The accumulation of high levels of glucans may greatly modify the physical properties of the periplasm. Although the amounts of periplasmic glucan, as well as the pattern of substitution with phosphoglycerol and succinyl residues, may vary considerably depending upon the exact conditions of growth and the specific strain under study, the value of 3.5% of the dry weight, reported for E. coli DF214 (53), may be taken as representative. For an MDO molecule of M _r of about 2,300, this represents about 15 nmol of MDO per mg (dry weight) of cells. Cayley et al. (13) found that the volume of the periplasm for cells of E. coli K-12 growing in medium of low osmolarity was about 0.31 μl/mg (dry weight). Assuming that this value is approximately correct for strain DF214, the concentration of MDO in the periplasm is calculated to be about 50 mM. If the MDO molecules bear three net negative charges because of their content of phosphoglycerol and succinyl substituents, neutralized by freely dissociable monovalent cations, the contribution of MDO to the osmotic strength of the periplasm is about 200 mosM. Although the uncertainties in this calculation are apparent, this appears to be a reasonable approximation.

In their pioneering study of the periplasmic compartment in E. coli and Salmonella typhimurium, Stock et al. (61) concluded that the osmolarity of the periplasmic space is about the same as that of the cytoplasm. It follows that when cells are grown in medium of low osmolarity, the osmolarity of the periplasm must be higher than that of the medium. The presence of MDO makes a significant contribution to the maintenance of the osmolarity of the periplasm under these conditions. Further, because MDO molecules bear a net negative charge and cannot diffuse through the outer membrane, they strongly influence the Donnan potential across the outer membrane. Sen et al. (58) measured the partition of labeled choline into the periplasm of cells of E. coli HN455 (blocked in the active transport of choline), enabling them to determine the Donnan potential across the outer membrane. They reported that the Donnan potential could largely be accounted for by the calculated content of MDO in the periplasm.

Mutations that block the synthesis of periplasmic glucans impair the growth of cells of Rhizobium meliloti (16) and A. tumefaciens (12) in medium of low osmolarity, but not in medium of high osmolarity, indicating that the presence of periplasmic glucans is an important part of the adaptation of these strains to hypoosmotic medium. The growth of hrpM mutants of Pseudomonas syringae, thought to be defective in the production of MDO-like β-glucans, is similarly decreased in dilute medium, but normal growth may be restored by increasing the level of certain sources of nitrogen, such as ammonium chloride or urea, to the level of 10 mM (47). The further addition of 10 μM glutamine, which somewhat reduces the growth of wild-type P. syringae on ammonia or urea under these conditions, completely blocks the growth of the hrpM mutant. These results suggest a role for periplasmic glucans in the glutamine-regulated uptake of certain nitrogen sources during growth on hypoosmotic medium. In contrast, however, there was no impairment of the growth of mdoA mutants of E. coli K-12 on medium of low osmolarity (34).

Mutations in E. coli that block the synthesis of periplasmic glucans, although not appreciably impairing the rate of growth under conditions so far studied, do have pleiotropic effects on cell motility, the regulation of capsule production, and the pattern of outer membrane proteins (19). Similar effects are seen in species of Rhizobium (22) and Agrobacterium (12) and in P. syringae (47). It may be noted that these pleiotropic effects are all on systems that are regulated by the osmolarity of the growth medium. There may be a hierarchy in hypoosmotic adaptation in which the presence of periplasmic glucans serves as a signal to other osmoregulated systems, reporting the osmolarity of the medium.

The mechanism by which periplasmic glucans may exert such regulatory roles is not known. The altered pattern of production of porins OmpF and OmpC in mutants blocked in MDO production, first described by Fiedler and Rotering (19), is found only when the mutant cells are growing in medium of low ionic strength as well as low osmolarity (20). The normal pattern of porins was restored by the addition of 15 mM KC1 to the low-osmolarity medium, but not by the addition of 30 mM sucrose (20). The osmoregulation of porins OmpF and OmpC involves the function of EnvZ (59), a cytoplasmic membrane protein with a periplasmic domain. Presumably the function of EnvZ in the regulation of OmpF and OmpC requires a minimum ionic strength in the periplasm, which in medium of low ionic strength is furnished by the presence of MDO.

Periplasmic Glucans and the Recognition of Specific Eukaryotic Hosts

The periplasmic cyclic glucans also play an important but poorly understood role in the recognition of specific plant hosts by species of Rhizobium (22), leading to symbiotic nitrogen fixation, and by Agrobacterium (10), leading to the development of plant tumors. There is strong evidence also that linear periplasmic glucans closely similar to the MDOs of E. coli are essential virulence factors for the infection of plants by P. syringae (40, 41, 48). In soil bacteria, periplasmic glucans thus play a rather general role in the recognition and infection of specific eukaryotic hosts, whether leading to symbiosis or to disease. It would obviously be of great interest to determine whether periplasmic glucans are also virulence factors for the infection of animal hosts by gram-negative bacteria such as pathogenic strains of Escherichia, Salmonella, Klebsiella, and Pseudomonas that are known to produce periplasmic glucans.

A more comprehensive account of the cyclic periplasmic glucans of the Rhizobiaceae may be found in the recent review by Breedveld and Miller (7). The following discussion will focus on the MDO (periplasmic β-d-glucans) of E. coli.

MDO OF E. COLI

Structure

The MDO of E. coli K-12 are a heterogeneous family of closely related oligosaccharides containing glucose as the sole sugar and substituted with phospho-sn-1-glycerol, phosphoethanolamine, and O-succinyl ester residues (35, 64). The discovery of phosphoethanolamine residues in MDO was surprising in view of the metabolic stability of phosphatidylethanolamine in the earlier experiments of Kanfer and Kennedy (33). It is possible that the strain of E. coli B studied in those early experiments (33) fortuitously lacked the enzyme needed for the transfer of phosphoethanolamine residues to MDO.

In addition to heterogeneity of substitution, species of MDO also vary in the number of glucose residues per mole. After removal of the succinyl residues and dephosphorylation with HF under conditions of minimum cleavage of glycosidic bonds, MDOs were converted to fluorescent derivatives (15). High-pressure chromatography led to the separation of distinct species (Fig. 1). MDO appear to contain 6 to 12 glucose residues per mole, with the principal species containing 8 to 9 glucose units.

Fig. 1Heterogeneity of size of MDO fraction B, derived from frozen cells of E. coli B (Grain Processing Corp.). After treatment with 50% HF to remove phosphodiester substitutents, and after mild alkaline hydrolysis to remove succinyl residues, the MDO was converted to fluorescent derivatives of 2-aminopyridine (15) and analyzed by high-pressure liquid chromatography on a column of amino Spherisorb.

Source: The degree of polymerization (dp) was determined by mass spectrometry in the laboratory of Vernon Reinhold.

The glucose units of MDO are joined by β-1,2 and β-1,6 linkages (55). The structure is highly branched. The backbone probably consists of β-1,2-linked glucose units (65) to which the branches are attached by β-1,6 linkages as in the tentative formula of Fig. 2. The glucose unit at the reducing terminus is known to be linked through its 2-position to the remainder of the molecule (55).

Fig. 2Tentative structure of MDO fraction A-2. The precise linkage of the glucose residues has not been determined and is arbitrarily presented.

Source:

The multiple substitution of species of MDO with phospho-sn-1-glycerol and O-succinyl ester residues gives these molecules varying amounts of net negative charge, which is the basis for their separation on DEAE-cellulose at pH 7.4 into fractions designated MDO A, B, and C (64). Each of these fractions can be further subfractionated on Dowex-1 acetate at pH 3.7, giving rise to species of MDO designated A-1, A-2, etc. Subfractionation during this second step presumably reflects varying substitution with O-succinyl ester residues, of which the unesterified carboxyl groups are completely ionized during chromatography on DEAE-cellulose at pH 7.4 but not at pH 3.7 during chromatography on Dowex.

MDO fraction A has been subfractionated and analyzed by Kennedy et al. (35) and Schneider and Kennedy (54). Fraction A-1, after separation on Dowex-1 resin at pH 3.7, was found to contain an average of one phosphoglycerol and two succinyl residues per mole with a total estimated net negative charge of 3 at neutral pH. About half of the molecules in this subfraction also contain phosphoethanolamine residues in phosphodiester linkage, which do not contribute to the net negative charge at neutral pH. Fraction A-2 contains three phospho-sn-1-glycerol residues per mole, with no detectable succinate or phosphoethanolamine.

A tentative structure consistent with what is known about MDO A-2 is shown in Fig. 2. Some of the details, such as the exact pattern of branching, are arbitrarily presented. The molecular weight of the tripotassium salt of MDO A-2 (Fig. 2) is calculated to be 2,305. MDOs of fractions B and C have a higher number of anionic substituents and presumably somewhat higher molecular weights.

Biosynthesis

Synthesis of Polyglucose Chains.

Schulman and Kennedy (56) reported genetic and biochemical evidence that UDP-glucose is an essential intermediate in the biosynthesis of MDO. Weissborn and Kennedy (65) discovered a novel membrane-bound glucosyltransferase that catalyzes the transfer of glucose units from UDP-glucose to a suitable "primer" such as octyl-β-d-glucoside or β-d-glucosyl-(1,2)-d-glucopyranoside (sophorose), with the formation of β-1,2-linked polyglucose chains. Although the overall reaction appears to be quite simple, it is in fact surprisingly complex.

Therisod et al. (63) found that acyl carrier protein (ACP) is an essential component of the transglucosylase system. This result was unexpected, because the function of ACP in the transglucosylase bears no apparent relation to its well-known functions in lipid biosynthesis. This disparity was further emphasized by the finding that the conversion of ACP to apo-ACP by enzymic removal of its covalently bound phosphopantetheine prosthetic group does not alter its transglucosylase activity (62). All previously known functions of ACP involve its phosphopantetheine prosthetic group.

Although the ACP of E. coli is a small protein of only 77 amino acid residues, it appears to contain at least two distinct functional domains, one involving phosphopantetheine linked to serine 36, needed for its multiple functions in lipid synthesis, and some other domain(s) needed for transglucosylase activity. The further study of the function of chemically and enzymically modified forms of ACP has revealed some of the structural features of the ACP molecule that are important in the glucosyltransferase reaction (C. Karkaria and E. P. Kennedy, unpublished data), as follows. (i) The carboxyl terminus of ACP may be extensively modified without loss of glucosyltransferase activity. (ii) The first six amino acids of the amino terminus appear to contain a region essential for glucosyltransferase function. (iii) A synthetic peptide representing residues 26 to 50 of ACP does not activate the glucosyltransferase itself, but blocks the activation of the system by holo-ACP. Peptide 26–50 therefore appears to contain a domain important for binding of ACP to the membrane component of the system. (iv) ACP in which the free amino residues are acetylated by treatment with O-acetyl-N-hydroxysulfosuccinimide retains activity. The activity of the acetylated ACP becomes resistant to treatment with trypsin, which rapidly inactivates unmodified ACP.

It has been suggested the synthesis and modification of MDO glucan chains takes place while the nascent glucans are bound to a carrier, presumably a phospho- or diphosphopolyprenol (66). The lack of effect of colicin M on the biosynthesis of MDO in cells of E. coli led Harkness et al. (26) to suggest that the biosynthesis of MDO may not involve bactoprenol phosphate. Colicin M is thought to cause the lysis of cells by interfering with the regeneration of bactoprenol phosphate in the cycle of reactions in which polyprenol-linked intermediates function in the synthesis of peptidoglycan (25). More direct evidence supporting the role of polyprenol phosphate in MDO biosynthesis, however, was reported by Weissborn et al. (66). The ACP-requiring glucosyltransferase reaction in cell-free membrane preparations was found to be strongly inhibited by both bacitracin and amphomycin, antibiotics that function by forming specific complexes with polyprenol phosphates. Furthermore, the activity of the glucosyltransferase was greatly stimulated by the addition of polyprenyl phosphates (P) such as decaprenyl-P and synthetic dihydroheptaprenyl-P but not by farnesyl-P (66).

The cell-free membrane-bound glucosyltransferase preparations also catalyzed a rapid formation of decaprenyl-P-glucose, presumed to be decaprenol alpha-d-glucopyranosyl phosphate on the basis of its alkali stability (66). Yamazaki et al. (68) have shown that the alpha anomer is much more stable to alkali than the beta. Pulse-chase experiments revealed that the glucose units transferred from UDP-glucose to glucan chains attached to the model acceptor octyl-β-d-glucoside do not pass through the pool of decaprenol α-d-glucopyranosyl phosphate. No accumulation of other prenol derivatives labeled from UDP-[³H]glucose could be detected, suggesting that the presumed lipid intermediates do not dissociate from the enzyme.

The products of the glucosyltransferase system in vitro are linear oligomers containing only β-1,2-linked glucose units, suggesting that the β-1,2 backbone of an MDO molecule is synthesized first and the branches are introduced subsequently. Branching may possibly involve UDP-glucose as a donor in a reaction catalyzed by an enzyme specific for the formation of β-1,6 links, or may alternatively involve a rearrangement of glucose units originally linked to the polymer through β-1,2 linkages. The formation of branch points in glycogen is well known to involve the latter type of process.

The glucose polymers produced in vivo by the glucosyltransferase system are of two discrete classes of size (65). The system thus does not synthesize polymers by addition of one glucose unit at a time to a pool of free intermediates, since this would lead to products with a continuous range of size, but must involve some processive mechanism. The principal products in vitro are approximately the same size as MDO molecules, but larger products are also found. Some chain-terminating event that limits the size of MDO molecules in vivo functions only imperfectly in the in vitro enzyme system.

The biosynthesis of the β-1,2-glucan chains of MDO is very different from that of the cyclic β-1,2-glucans of the family Rhizobiaceae. Zorreguieta and Ugalde (70) reported that a membrane-bound enzyme from species of Rhizobium and Agrobacterium catalyzed the synthesis of cyclic glucan from UDP-glucose. The growing glucan chains were found to be covalently linked to a membrane protein, estimated from its migration during electrophoresis in buffers containing sodium dodecyl sulfate to be 235 kDa in size. This protein is the product of the ndvB or chvB gene in Rhizobium or Agrobacterium, respectively, and is now known from the gene sequence to be 319 kDa in size (29). No such protein has been detected in E. coli. In further contrast, the synthesis of cyclic glucan appears not to require ACP or polyprenol phosphate, both of which are involved in MDO synthesis.

Presumably the last step in the synthesis of cyclic glucan is the cyclization reaction itself, in which the glucose unit at the reducing end of the nascent glucan is transferred from its attachment to the 319-kDa protein to the 2-position of the glucose unit at the nonreducing terminus. Obviously, such a reaction is not involved in the biosynthesis of open-chain glucans such as the MDO of E. coli, and this may be the reason why the E. coli glucosyltransferase is quite different from that of the members of the Rhizobiaceae.

In the Rhizobiaceae, cyclic glucan appears to be released into the cytosol, from which it is secreted in a process which is little understood, but which requires the function of the ndvA or chvA gene (60). The structure of the NdvA/ChvA protein, deduced from the gene sequence, resembles that of HlyB, a protein that functions in the ATP-requiring export of hemolysin A from E. coli. In contrast, it is thought that in the biosynthesis of MDOs, synthesis of nascent glucan chains takes place while these are linked to a polyprenol carrier that is needed for transit of the glucan chains from the cytosolic to the periplasmic face of the inner membrane. On the periplasmic face, the glucans are substituted with sn-1-glycerophosphate (31) or other residues that may be signals for release of the glucan from the carrier into the periplasm.

Transfer of Phosphoglycerol Residues to MDO.

An enzyme designated phosphoglyceroltransferase I catalyzes the transfer of phosphoglycerol residues from phosphatidylglycerol to MDO or to certain synthetic β-glucoside acceptors (31). The products were shown to be sn-1,2-diacylglycerol and β-glucoside-6-phosphoglycerol. The enzyme is localized in the inner or cytoplasmic membrane with its catalytic face on the external or periplasmic surface.

A second enzyme, phosphoglyceroltransferase II, localized in the periplasm, catalyzes the interchange of phosphoglycerol residues among species of MDO and certain β-glucoside model substrates (23). Although the physiological function of phosphoglyceroltransferase II appears to be the transfer of phosphoglycerol residues rather than their hydrolysis, at low concentrations of acceptor the enzyme acts as a cyclase/hydrolase with the liberation of cyclic sn-1(3),2-glycerophosphate (23). These findings suggest a mechanism in which Enz-P-glycerol is an intermediate. The phosphoglycerol moiety is transferred to MDO when this substrate is available. In the absence of this substrate, slow cyclization and release of cyclic glycerophosphate from the enzyme takes place.

Phosphoglyceroltransferase II does not recognize phosphatidylglycerol as donor of sn-1-glycerophosphate residues, as shown by the fact that mutants in the mdoB locus, defective in phosphoglyceroltransferase I, produce MDO containing no phosphoglycerol residues although they contain wild-type levels of phosphoglyceroltransferase II (30).

It is thought that when MDO chains linked to their lipid carrier appear on the external face of the inner membrane, phosphoglyceroltransferase I catalyzes the transfer to MDO of phosphoglycerol residues from phosphatidylglycerol. Because unsubstituted MDO is not detected in significant amounts in pulse-chase experiments, substitution appears to be a signal for release into the periplasm. Phosphoglyceroltransferase II, however, may catalyze the transfer of phosphoglycerol from the nascent MDO to periplasmic MDO before the nascent chain can be released from the lipid carrier. This reaction regenerates unsubstituted MDO on the lipid carrier, which may act as acceptor for another phosphoglycerol residue from phosphatidylglycerol. According to this view, phosphoglyceroltransferase II functions, not in the primary transfer of phosphoglycerol from phosphatidylglycerol, but in secondary transfer reactions that lead to multiple substitutions of MDO with phosphoglycerol.

Transfer of Phosphoethanolamine Residues.

Miller and Kennedy (45) provided direct evidence for the transfer of phosphoethanolamine residues from phosphatidylethanolamine to MDO. Vesicles containing phosphatidyl[³H]ethanolamine were taken up by unlabeled cells of E. coli as described by Jones and Osborn (32), and the transfer of labeled phosphoethanolamine to MDO was observed during subsequent growth of the cells on unlabeled medium. The transfer of phosphoethanolamine to MDO was osmotically regulated in a fashion closely similar to that of the overall synthesis of MDO (45).

MDO Biosynthesis and the Diacylglycerol Cycle.

The continuous transfer of phosphoglycerol and phosphoethanolamine residues from membrane phospholipids to MDO leads to the generation of diacylglycerol, which is phosphorylated by ATP to form phosphatidic acid in a reaction catalyzed by diacylglycerol kinase, an enzyme first described in E. coli by Pieringer and Kunnes (49). This kinase does not play a primary role in the de novo synthesis of phospholipids, which instead utilizes phosphatidic acid generated by the acylation of sn-3-glycerophosphate (14).

Mutants in the dgk gene are defective in diacylglycerol kinase. Such mutants grow readily in medium of high osmolarity but not in low osmolarity (51). This suggests that the biosynthesis of MDO, favored by low osmolarity, is the principal (although not the sole) source of diacylglycerol in E. coli. This conclusion is supported by the finding that further mutations that block the synthesis of MDO increase the rate of growth of dgk mutants in medium of low osmolarity (67).

Genetics of the Biosynthesis of MDO.

Strain T10GP, a derivative of E. coli K-12 constructed by Pluschke et al. (50), was found in the laboratory of C. R. Raetz to be blocked in the synthesis of MDO. The mutation (mdoA1) in this strain causes defects in the membrane-bound, ACP-requiring glucosyltransferase described above. It has been mapped to the 23-min region of the E. coli chromosome (6). Lacroix et al. (36) cloned a 5.5-kb piece of DNA that complements both the mdoA1 mutation and a second mutation, mdo-200::Tn10, of identical phenotype. Further work showed that the mdoA region consists of an operon with two genes, mdoG and mdoH (37). The mdoA1 and mdo-200::Tn10 alleles belong to the mdoH complementation group.

The mdoGH operon has been sequenced (40) and has been found to be homologous to the hrpM locus of P. syringae, which functions in the virulence of this organism against plant hosts (48). Introduction of a plasmid bearing ORF2 of the hrpM locus from P. syringae into an mdoH::Tn10 mutant of E. coli restored the synthesis of MDO (41). These experiments offer strong evidence that the mechanism of biosynthesis of MDO is very similar in E. coli and P. syringae and further indicate that periplasmic glucans play an important role in the infection of plant hosts by P. syringae.

Mutations in either mdoG or mdoH block the synthesis of MDO, but only mutations in mdoH abolish the activity of the membrane-bound glucosyltransferase (37, 67). The mdoG gene encodes a 56-kDa periplasmic protein (40) of unidentified function. Possible roles for this protein include the introduction of β-1,6 branches into the MDO structure, or the release of the MDO chains from a polyprenol carrier. The mdoG gene corresponds to ORF1 of the hrpM locus of P. syringae (40).

The mdoH gene encodes a membrane protein of 97 kDa (40). It corresponds to ORF2 of the hrpM locus. Because defects in mdoH lead to loss of the membrane-bound glucosyltransferase, it is presumed that mdoH is the structural gene for the glucosyltransferase, or for a part of it, but direct evidence on this point is lacking.

Mutations in a gene designated mdoB lead to loss of activity of phosphoglyceroltransferase I and to the production in vivo of MDO lacking phosphoglycerol substituents (18, 30). The mdoB gene is localized near 99 min on the E. coli chromosome (30, 39).

Work on the genetics of MDO biosynthesis has been hampered by the lack of a facile procedure for the selection of mdo mutants. Such a procedure has now been developed (67), based on the fact that dgk strains of E. coli, defective in the enzyme diacylglycerol kinase, grow slowly when diacylglycerol accumulates and cannot be metabolized. Further mutations that block the synthesis of MDO permit dgk cells to grow more rapidly, offering an effective positive selection for the MDO^– phenotype. Of 28 such mutants studied in detail, 4 were found to map near min 27 and were identified as galU mutants blocked in the production of UDP-glucose. The remaining 24 were mutations in the mdoA locus, of which 23 were in the mdoH and one in the mdoG gene (67).

Osmotic Regulation of the Biosynthesis of MDO.

The biosynthesis of MDO in E. coli is strikingly regulated by the osmolarity of the medium in which the cells are growing. The level of MDO in cells grown in a medium of low osmolarity was about 16 times higher than when grown in the same medium with added 0.4 M NaCl (34). As indicated above, such osmotic regulation appears to be a general feature of the biosynthesis of periplasmic glucans.

Relatively little is known about the mechanism by which the synthesis of periplasmic glucans is regulated by the osmolarity of the growth medium. Two general types of mechanism have been considered (53), as follows. (i) The expression of genes coding for biosynthetic enzymes may be increased by low and reduced by high osmolarity. (ii) The activity of rate-limiting biosynthetic enzymes may be regulated by the ionic strength of the cytosol, which in turn responds to the osmolarity of the medium. Evidence has been reported in support of each of these alternative hypotheses.

Lacroix et al. (37) prepared multicopy plasmids bearing fusions of the lacZ gene to the beginning of the mdoH gene and found that the expression of the lacZ gene in one such construction was under osmotic control, with a sevenfold higher expression by cells growing at low osmolarity as compared with cells grown in the same medium with added 0.3 M NaCl. The levels of mdoGH-specific mRNA were also found to about fivefold higher in cells grown in medium of low osmolarity.

In these experiments, however, the actual levels of glucosyltransferase, encoded or regulated by the chromosomal mdoH gene, were not measured. When such measurements were made by Rumley et al. (53), no significant difference could be found in the glucosyltransferase activity of membranes from cells grown in medium of high or low osmolarity. This result is consistent with the findings of Zorreguieta et al. (69), who reported that levels of the membrane-bound enzyme that catalyze the synthesis of periplasmic cyclic β-d-glucan in A. tumefaciens are not reduced when cells are grown in medium of high osmolarity.

The osmotic regulation of a second enzyme of MDO biosynthesis, phosphoglyceroltransferase I, which catalyzes the transfer of phosphoglycerol head groups from phosphatidylglycerol to nascent MDO molecules (Fig. 1), was examined in an earlier study (5). Its activity was also shown to be independent of the osmolarity of the medium in which the cells are grown. These results suggest that most or all of the enzymes of MDO biosynthesis are expressed constitutively and not induced by growth in medium of low osmolarity.

If osmotic regulation of levels of the enzyme needed for MDO synthesis takes place primarily at the genetic level, when the medium is changed from high to low osmolarity there should be a considerable lag to allow for the synthesis of new MDO biosynthetic enzymes before MDO synthesis reaches its new, higher rate. When this was tested (53), no lag could be detected, consistent with the conclusion that enzymes of MDO biosynthesis are present but inactive during growth of cells in medium of high osmolarity.

Bohin and Kennedy (5) measured the production of MDO chains in cells making the transition from low to high osmolarity. Although the conditions of the experiments did not permit sensitive kinetic analysis of early time points in the process, it appeared that the rate of synthesis of MDO was down-regulated without detectable time lag as the osmolarity of the medium was increased. A mechanism for the shutoff of MDO synthesis based on simple dilution of the cellular level of MDO biosynthetic enzymes could be excluded. The results are consistent with a mechanism involving inhibition of the biosynthetic enzymes in cells growing in medium of high osmolarity.

Increases in the osmolarity of the growth medium lead to increases in the ionic strength of the cytosol (27, 38). Inhibition of the membrane-bound MDO glucosyltransferase by high ionic strength might be a mechanism by which the activity of the transferase is automatically reduced when cells are growing in medium of high osmolarity. Such inhibition of the cell-free glucosyltransferase by high levels of salt added to the assay system was in fact observed (53). This result with the E. coli enzyme again parallels findings with the enzyme from A. tumefaciens catalyzing the synthesis of cyclic glucan, the activity of which is also inhibited in vitro by added salt (69).

At present, there is no explanation for the discrepancy between the experiments of Lacroix et al. (37), which point to osmotic regulation of MDO synthesis at the level of gene expression, and those of Rumley et al. (53), which indicate that these genes are expressed constitutively. One possibility is that osmotic regulation of the mdoGH operon, borne on a plasmid as in the experiments of Lacroix et al. (37), may be different from the regulation of the chromosomal genes.

Feedback Regulation of MDO Biosynthesis.

Feedback inhibition also appears to be an important mode of the regulation of MDO biosynthesis. Strain DF214 pgi zwf cannot synthesize glucose and so cannot synthesize MDOs unless glucose is added to the medium. Rumley et al. (53) studied the kinetics of the accumulation of MDO by cells of this strain. When labeled glucose was added to the medium, synthesis of labeled MDO began immediately at a high rate. The rate of synthesis fell as MDO accumulated in the periplasm, leveling off at a rate of about one-seventh the initial rate as the content of MDO reached a plateau level of about 3.5% of the dry weight.

Breedveld et al. (8) found that the synthesis of cyclic glucans catalyzed by cell-free membranes of Rhizobium leguminosarum was strikingly inhibited by the addition to the assay system of cyclic glucan at concentrations comparable to those calculated to accumulate in the periplasmic space in vivo. These workers suggested that feedback inhibition may therefore be an important mechanism for the regulation of the biosynthesis of cyclic glucans in this organism, and further suggested that loss of cyclic glucans from the periplasm when the outer membrane becomes leaky under certain conditions of growth allows their continuing synthesis, leading to the observed accumulation of very high amounts of glucan in the medium.

ACKNOWLEDGMENTS

Work in this laboratory on the biosynthesis and function of periplasmic glucans has been supported by Public Health Service grants GM19822 and GM22057 from the National Institute of General Medical Sciences.