Module 4.7.1.7

Undecaprenyl Phosphate Synthesis

THIERRY TOUZÉ* AND DOMINIQUE MENGIN-LECREULX

[SECTION EDITOR: LYNN SILVER]

Posted February 12, 2008

Université Paris-Sud, CNRS, UMR 8619, Laboratoire Enveloppes Bactériennes et Antibiotiques, Institut de Biochimie et Biophysique Moléculaire et Cellulaire, Bâtiment 430, F-91405 Orsay Cedex, France

*Corresponding author. Mailing address: Laboratoire Enveloppes Bactériennes et Antibiotiques, IBBMC, UMR 8619 CNRS, Université Paris-Sud, Bât. 430, 91405 Orsay Cedex, France. Phone: 33 1 69 15 61 34, Fax: 33 1 69 85 37 15, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it .

Undecaprenyl phosphate (C55-P), also referred to as bactoprenol, is a 55-carbon long-chain polymer of isoprene units belonging to the family of linear isoprenoids (C55 to C100). This lipid is used for the biosynthesis of the bacterial cell wall peptidoglycan (92, 93) and, for this reason, is essential. C55-P is also required for the synthesis of many other cell wall polymers, such as the lipopolysaccharides (68), the teichoic acids, the membrane-derived oligosaccharides (44), and the capsular polysaccharides and the enterobacterial common antigen (70). C55-P is utilized as a chemical carrier that facilitates the transport of the hydrophilic oligosaccharide precursors across the cytoplasmic membrane toward the periplasm, where the polymerization of the newly formed glycan chain takes place. Accordingly, the oligosaccharide is transferred from a nucleotide-sugar intermediate to the lipid carrier through the action of a glycosyltransferase; thereafter, the complex is translocated from the cytoplasmic to the outer surface of the inner membrane prior to the polymerization of the glycoconjugate. Two enzymatic steps are implicated in the de novo synthesis of C55-P; it originates from the dephosphorylation of its precursor, undecaprenyl pyrophosphate (C55-PP), which is synthesized by eight consecutive condensations of isopentenyl pyrophosphate (C5-PP) with farnesyl pyrophosphate (C15-PP). C55-PP is also released after each polymerization round when the oligosaccharide unit is transferred to the growing glycan chain in the periplasm. It is then recycled, after being transformed into the active monophosphate form and translocated from the outer surface to the inner side of the cytoplasmic membrane. C55-P can alternatively be generated by the phosphorylation of undecaprenol, which is produced by the dephosphorylation of C55-P. Undecaprenol has been detected in a few gram-positive bacteria and was therefore assumed to constitute a reserve pool for the subsequent utilization of C55-P. However, the existence of undecaprenol in gram-negative bacteria has never been documented, raising the question of the existence of such a reserve pool in enteric bacteria. In the present review, we focus on the current status of our knowledge of the enzymatic biosynthesis of C55-P in Escherichia coli; the genetics, the biochemistry, the enzyme structures, and the catalytic mechanisms of the enzymes are discussed in detail. Finally, we point out the fact that while the cellular content of C55-P is critical for bacterial cell survival, the pool level of this lipid has never been studied in detail and the mechanisms that control its synthesis, as well as that of its precursor, remain unknown.

BIOSYNTHESIS OF C55-PP

Isoprenoids, including linear isoprenoids, carotenoids, retinoids, prenylated proteins, ubiquinones, cholesterol, and rubbers, are natural compounds all synthesized by sequential head-to-tail 1'-4 condensations of the five-carbon building unit, the homoallylic substrate C₅-PP, with allylic pyrophosphates with various numbers of carbon atoms (14) (Fig. 1). These reactions are catalyzed by polyprenyl diphosphate synthases, also referred to as prenyltransferases, which can be divided into two families according to the stereochemical outcome, cis (Z-prenyltransferases) or trans (E-prenyltransferases), of the double bond in the newly added isoprene unit (97). Each prenyltransferase is selective for the allylic substrate it recognizes and is also specific with respect to the end product chain length (15, 97). The reactions catalyzed by prenyltransferases start by the formation of an allylic cation after the elimination of a pyrophosphate ion, followed by the electrophilic attack of the C-1 carbonium of the allylic substrate on the C-4 atom of C₅-PP, with the stereospecific removal of a proton at the C-2 position of C₅-PP to form a new C―C bond and a new double bond (57) (Fig. 1). The 55-carbon C₅₅-PP is synthesized by the consecutive condensations of eight C₅-PP units with the 15-carbon-chain trans,trans-C₁₅-PP, a reaction catalyzed by the cis-prenyltransferase undecaprenyl pyrophosphate synthase (UppS), yielding di-trans,octa-cis–C₅₅-PP (1, 27, 60) (Fig. 2).

Fig. 1Reaction catalyzed by prenyltransferases. Here, the newly formed double bond is in a cis (Z) configuration. "A" designates the prenyltransferase residue that plays the role of the C₅-PP C-2 proton acceptor before this proton is transferred to the leaving pyrophosphate group. IPP, isopentenyl pyrophosphate; R, allylic substrate carbon chain.

Fig. 2Synthesis of C₅₅-P. The synthesis of C₅₅-P involves two enzymatic steps: first, the undecaprenyl pyrophosphate synthase (UppS) catalyzes the sequential condensation of trans,trans-C₁₅-PP with eight molecules of C₅-PP, yielding C₅₅-PP; thereafter, C₅₅-PP is dephosphorylated. OP, phosphate group; OPP, pyrophosphate group.

In contrast to the cis-prenyltransferase UppS, the farnesyl pyrophosphate synthase (FPS) is the prototype of the trans-prenyltransferases. It catalyzes the sequential addition of dimethylallyl pyrophosphate and geranyl pyrophosphate to C₅-PP, yielding C₁₅-PP, which is the branching point in the synthesis of many isoprenoid compounds, as it serves as an allylic primer substrate for additional C₅-PP condensations via cis- or trans-prenyltransferases, including UppS (49). Structure determination and site-directed mutagenesis studies revealed the precise mechanism of prenyl chain elongation catalyzed by FPS a decade before any UppS-encoding gene was identified. Like all trans-prenyltransferases, FPS is characterized by the so-called isoprenoid fold that consists of several α-helices joined by connected loops surrounding a large cavity forming the active site. The wall of this cavity is covered with highly conserved residues, among which the two aspartate-rich DDXXD motifs are found. These two regions are critical since the substrate’s negatively charged pyrophosphate moieties bind to the aspartate carboxylate side chains via chelation with magnesium, each motif being responsible for the binding of either a homoallylic (C₅-PP) or an allylic substrate (49, 57, 84).

Identification of the UppS Gene

The first UppS-encoding gene was identified in Micrococcus luteus in 1998 by Shimizu et al. in Japan, providing the first identification of a cis-prenyltransferase enzyme gene (79). To achieve this goal, a genomic DNA library of M. luteus B-P 26 was generated in E. coli, and colonies harboring overexpressed prenyltransferase activity were screened in the presence of C₁₅-PP and radiolabeled C₅-PP by the colony autoradiography method developed by Raetz (65). Shortly after the isolation of the M. luteus UppS gene, many cis-prenyltransferases were also identified via comparisons of primary amino acid sequences from eubacteria, archaea, higher plants, and animals (17, 23, 37, 78). UppS enzymes from E. coli and various gram-negative and gram-positive bacteria were also identified (3, 43). Subsequently, UppS enzymes from E. coli and Streptococcus pneumoniae were shown to be essential for cell survival. The primary structures of the cis-prenyltransferases are very distinct from those of the trans-prenyltransferases; for instance, the cis-prenyltransferases lack the aspartate-rich motifs typically found in all trans-prenyltransferases, suggesting that the two types of enzymes do not share a common catalytic mechanism. Despite the lack of any aspartate-rich motif, divalent cations were also found to be required for the activity of UppS. An extensive site-directed mutagenesis analysis of all acidic residues in UppS was then performed, revealing that one aspartate residue, D26 in E. coli UppS, is absolutely required for C₅-PP binding as well as catalysis (61).

Structure of the UppS Enzyme

The three-dimensional crystal structures of the UppS apoenzymes of M. luteus and E. coli were both reported in 2001, revealing closely similar structures (25, 48) that are very different from the trans-prenyltransferase isoprenoid fold. UppS forms a homodimer of 29-kDa subunits that are tightly associated through a large interface among the edges of a central β-sheet (β6) and a pair of α-helices (α5, α6), with the main contact established by the coiled-coil structure formed between the long α5 helices ( Fig. 3). Each monomer contains a catalytic domain that consists of six parallel β-strands (β1 to β6), forming a central β-sheet core surrounded by five of seven α-helices (α1, α2, α3, α4, and α7) (Fig. 3). Many of the residues conserved among UppS enzymes are located within two α-helices (α2 and α3) and four β-strands (β1, β2, β3, and β4), all together forming a large cleft on the molecular surface of the enzyme. The interior of this cleft is mostly hydrophobic, whereas the entrance presents a cluster of positively charged residues (R39, R194, and R200 in E. coli UppS) in the vicinity of the D26 acidic residue previously shown to be critical for C₅-PP binding. Additionally, a motif referred to as a structural P loop that is commonly found in phosphate recognition enzymes is also located at the entrance of the cleft (47). The latter motif consists of four residues located at the N terminus of an α-helix (α1 in E. coli UppS), with a glycine (G27) at the N terminus followed by three residues conserved among cis-prenyltransferase enzymes (N28, R29, and R30). In light of these structures, it was hypothesized that the interior of the cleft, 30 Å deep, could accommodate the hydrophobic carbon tail of the allylic substrate C₁₅-PP and also that of the elongating prenyl chain. The negatively charged pyrophosphate group of C₁₅-PP and that of the C₅-PP units were assumed to interact with the positively charged cluster, the P-loop structure, and/or the D26 amino acid via the required metal ion. Moreover, both UppS apoenzymes contain a region connecting the α3 helix to the β2 strand that could not be resolved in the crystal structures due to its invisible electron density (Fig. 3). This loop was therefore hypothesized to be a flexible domain (E. coli residues 72 to 83). As deduced from its position, the loop, which contains many residues conserved among cis-prenyltransferases, is located near the entrance of the cleft and was thus hypothesized to participate in substrate binding and the catalytic process. A high degree of flexibility around the cleft must be required for substrate binding, catalysis, and/or product release. Subsequently, site-directed mutagenesis of conserved residues within this loop in E. coli and M. luteus UppS enzymes confirmed that this region plays an essential role in C₅-PP binding, as well as catalysis (see below) (26, 46, 48).

Fig. 3Overall structure of undecaprenyl pyrophosphate synthase (UppS) from E. coli. The figure was prepared with the atomic coordinates (1JP3) deposited in the Protein Data Bank by Ko et al. in 2001 (48). One dimer is shown by red α-helices and green β-strands, and the other monomer is shown by purple α-helices and blue β-strands. The region from residue 72 to 83 was not visible in the electron density map, suggestive of a flexible loop; a dotted line was added in one monomer to resolve the localization of the corresponding region. The hydrophobic cleft at the molecular surface of the molecule is indicated by arrows. This figure was prepared using PyMol.

Interestingly, in the E. coli structure, the two subunits are slightly different, suggestive of two distinct conformations relevant to the catalytic mechanism (48). Indeed, the α3 helix is kinked at different locations and to different degrees in the two subunits, leading to a closed form with a tighter space inside the cleft for better binding of the aliphatic chain and an open form for the release of the product or the binding of a newly allylic substrate.

UppS Catalytic Site

More recently, the crystal structures of the UppS enzyme from E. coli with bound allylic and homoallylic substrates, as well as a magnesium ion, were reported by Chang et al. and Guo et al., providing better insights into the molecular mechanism of C₅₅-PP synthesis (12, 13, 33). In order to generate a stable ternary complex without the formation of a product, the authors used an almost inactive C₁₅-PP analog, C₁₅-FsPP, in which the oxygen atom between the farnesyl and pyrophosphate groups was replaced by a sulfur atom. A comparison of the crystal structures of the apoenzyme and the substrate-bound UppS enzyme confirmed the dynamic conformational change between the closed and the open forms and the role of the flexible loop in the catalytic process that became apparent once the enzyme was in complexes with its substrates.

In the UppS–C₁₅-PP binary complex, the substrate hydrocarbon chain is bound by several side chains of hydrophobic residues protruding from helices α2 and α3 and forming the catalytic cleft. Among those residues, L85, L88, and F89 from helix α3 are reoriented through C₁₅-PP binding, leading to the closed conformation that likely ensures the correct positioning of the allylic substrate for the subsequent nucleophilic attack by C₅-PP (Fig. 4) (11). The fact that the side chains of these three hydrophobic residues move in order to come closer to the bound C₁₅-PP suggests that these side chains participate directly in the catalysis through their interaction with the allylic substrate. The pyrophosphate group of C₁₅-PP is linked to the structural P-loop motif and to the positively charged cluster by the backbone N atoms of G29 and R30 as well as the side chains of N28, R30, and R39 through hydrogen bonding and electrostatic interactions. No magnesium was found to be associated with the negatively charged pyrophosphate group of C₁₅-PP. Importantly, C₁₅-PP binding triggers the change from the open to the closed conformation, in which the flexible loop becomes more ordered. Upon C₅-PP addition to the binary UppS–C₁₅-PP complex, the pyrophosphate group of C₅-PP binds to the positively charged cluster residues R194 and R200 and to the amino acids N74 and R77, located at the entrance of the cleft and in the flexible loop, respectively (33).

Fig. 4C₁₅-PP binding in the active site of UppS. The ribbon diagram of the protein in two different orientations is shown close up, centered at the substrate binding site. The figure was prepared with the atomic coordinates 1X06 deposited in the Protein Data Bank by Guo et al. in 2005 (33). It corresponds to the structure of E. coli UppS in a complex with FsPP, C₅-PP, and Mg²⁺. In this structure, densities for C₅-PP were low, which dissuaded the authors from modeling the corresponding molecule; therefore, only the FsPP substrate is shown. For clarity, FsPP is represented as either line (up) or sphere (down) models. The side chains of residues interacting with FsPP are shown as line models; C, O N, P, and S atoms are in green, red, blue, orange, and yellow, respectively. The flexible loop that becomes visible upon allylic substrate binding is colored in red.

UppS Catalytic Mechanism and Role of the Metal Ion

The structures of the haloenzyme revealed the central role played by the acidic residue D26 and the metal ion in the catalytic mechanism. When UppS was in a complex with C₁₅-FsPP alone, no magnesium associated with the substrate pyrophosphate group was found (12). This observation was consistent with the fact that the metal ion was previously shown by fluorescence quenching measurements to be dispensable for C₁₅-PP binding to UppS but that C₅-PP binding as well as the ensuing condensation reaction necessitate the magnesium (12, 61). Surprisingly, in the UppS–C₁₅-FsPP–C₅-PP ternary complex, the magnesium is not bound to C₅-PP as expected but is instead coordinated by the pyrophosphate group of the allylic substrate, C₁₅-FsPP, and the carboxylate residue D26 (Fig. 4). To further highlight the function of the D26 residue in catalysis, Guo et al. have also determined the crystal structures of a D26A E. coli UppS mutant form whose activity is severely altered. When the two allylic and homoallylic binding sites in the mutant form were occupied, the metal ion was not fixed to the allylic substrate but remained associated with C₅-PP, in contrast to the wild-type UppS structure. In the light of these structures and fluorescence quenching measurements, it is now suggested that UppS first binds the allylic substrate, C₁₅-PP, without the need for magnesium and that this binary complex then binds the homoallylic substrate, C₅-PP, whose pyrophosphate group interacts with the acidic D26 residue via a magnesium bridge (16, 33). Thereafter, the D26 residue may control the migration of the magnesium from C₅-PP to C₁₅-PP, where the magnesium may facilitate the ionization of the pyrophosphate group of C₁₅-PP, leading to the formation of the allylic cation. To ensure the ensuing reaction, the flexible loop possesses residues (N74 and S71) that, with respect to their positions, are good candidates for proton acceptors for the subtraction of the proton from C₅-PP C-2, promoting the electrophilic attack and the formation of the new double bond (33, 82) (Fig. 1). Site-directed mutagenesis of these loop residues leads to a significant decrease of UppS catalytic activity that is consistent with the potential role of these residues (48, 60). The complete reaction results in the net transfer of the C₅-PP C-2 proton to the C₁₅-PP pyrophosphate group that is subsequently released (Fig. 1). The protonation may not be direct and may involve some other residues as proton carriers (33). In the current model, it is assumed that the flexible loop plays an essential role by pulling the α3 helix toward the binding site for the conversion of the subunit from an open form (apoenzyme and product bound) into the active closed conformation (substrate bound), as required for catalysis (11). The conversion from the open to the closed conformation must be triggered by interactions with the substrate molecule, and conversely, the product release must drive the relaxation of the binding site, which becomes available for another cycle of reaction. Thus, this catalytic loop may be central in controlling the mechanism of substrate binding and ultimate product release, which occurs once the product has reached its definite length (33).

Mechanism of Chain Length Control

It is intriguing to notice that, despite the fact that all cis-prenyltransferases are homologous and catalyze the same condensation reaction, with the same building unit and often the same allylic substrate, their products vary greatly in size. Moreover, each individual enzyme catalyzes the formation of an ultimate product of invariable size. This property implies the existence of a tight chain length control mechanism. With respect to the lengths of their end products, cis-prenyltransferases have been classified into three families, the short-, medium-, and long-chain cis-prenyltransferases, UppS belonging to the medium-chain family (82). From the E. coli apoenzyme UppS structure, it was first hypothesized that the ultimate chain length of the 55-carbon C₅₅-PP was determined by the presence of several large amino acids covering the bottom of the hydrophobic cleft that accommodates the elongating allylic substrate. Site-directed mutagenesis of these residues revealed the critical role played by the leucine 137 residue, as its replacement with the smaller residue alanine resulted in the synthesis of a longer-chain product (48). Moreover, the comparison of the UppS primary sequence with the sequence of one cis-prenyltransferase homolog from Saccharomyces cerevisiae, which synthesizes longer-chain products, revealed that the latter enzyme contains an alanine at the corresponding position (7). Ko et al. have proposed that the UppS leucine residue may function as the floor of the catalytic pocket to block the further elongation of the growing isoprenoid chain. A comparable chain length control mechanism in the case of trans-prenyltransferases had previously been highlighted, even though no structural relationship was visible (83, 84, 96, 97).

Inversely, the comparison of the UppS enzymes with the FPS enzyme from Mycobacterium tuberculosis, which is the only short-chain cis-prenyltransferase identified to date, has pointed out the role of the small alanine residue at position 69 in the E. coli UppS enzyme (48, 78). In E. coli UppS, this alanine is located in close proximity to the bound chain of the elongating allylic substrate positioned about halfway in the course of C₁₅-PP-to-C₅₅-PP elongation. In the M. tuberculosis enzyme, the corresponding amino acid is replaced by a bulky leucine residue, which was thus hypothesized to interfere with prenyl chain elongation via steric hindrance, yielding a limited number of condensation cycles. Ko et al. have also shown that the E. coli UppS A69L mutant form catalyzes the formation of a C₃₀-prenyl product, suggesting that the nature of the residue at this precise position is a determinant of the length of the end product (48).

The alignment of multiple cis-prenyltransferases yielding products of different sizes revealed that the eukaryotic enzymes that catalyze the formation of long-chain isoprenoids contain three to seven extra amino acid residues located within theα3 helix (23, 76, 77). To test the potential role of these residues, Kharel et al. have constructed M. luteus UppS mutant forms containing insertions corresponding to the sequences present in the eukaryotic homologs (45). The resulting modified enzymes catalyze the formation of longer prenyl chains than the wild-type enzyme, indicating the determinant role of this region in chain length regulation. This finding was confirmed by the fact that the deletion of the corresponding sequences in the long-chain isoprenoid prenyltransferases results in the synthesis of shorter-prenyl-chain products. It is noteworthy that the nature of these additional residues as well as their respective positions within the α3 helix are decisive, as the addition of alanine residues in the M. luteus UppS sequence does not lead to the formation of longer products, showing that a simple extension of the hydrophobic tunnel is clearly not sufficient. Kharel et al. have noticed that these extra residues are localized at the hinge oftheα3 helix and concluded that these residues may control the bending direction of the growing prenyl chain so that it is accommodated by the cleft to fit a precise size (45).

C55-PP DEPHOSPHORYLATION

In addition to being released after its enzymatic biosynthesis, C₅₅-PP is released during the synthesis of the different cell wall polymers when the saccharide is finally transferred to the extending glycan strand. It was established previously that about 60% of C₅₅-PP is recycled in this way in peptidoglycan synthesis (80). C₅₅-PP must then be translocated from the outer surface of the inner membrane to the cytoplasmic side to be reused, since the transfer of the glycosyl moiety to the lipid carrier is known to occur on the cytoplasmic side of the membrane. The mechanism that allows this translocation event still remains to be discovered.

The following step in C₅₅-P synthesis is then the dephosphorylation of C₅₅-PP (Fig. 2). This step is essential before the lipid carrier becomes available since the oligosaccharide transfer between the UDP-sugar and the lipid requires the monophosphate form of the lipid. The dephosphorylation of C₅₅-PP occurs in the course of the de novo synthesis of C₅₅-P and also at the end of each cycle of the polymerization reaction, when the lipid is released in the pyrophosphate form. The dephosphorylation step was previously hypothesized to be the target of colicin M, a toxin synthesized and released into the growth medium by some strains of E. coli to kill other susceptible E. coli strains by causing the inhibition of peptidoglycan synthesis (35, 36). However, we recently demonstrated that colicin M exerts its bacteriolytic effect via the degradation of lipid I and lipid II peptidoglycan precursors, releasing the lipid carrier in the form of the alcohol undecaprenol (20). In 1972, Goldman and Strominger reported for the first time that the dephosphorylation of C₅₅-PP in M. luteus is catalyzed by a membrane-bound phosphatase (31). However, it is only recently that the responsible enzymes have been identified (21, 22). Thus, the second and last step of the C₅₅-P synthesis pathway has been characterized to only a limited extent up to now.

BacA Proteins Form the First Family of Bacterial C55-PP Phosphatases

Since the 1970s, the enzymatic dephosphorylation of C₅₅-PP had been known to be the target of bacitracin, an antibiotic composed of a mixture of small peptides and produced by some strains of Bacillus licheniformis and Bacillus subtilis (80, 88). The main component of bacitracin is the macrocyclic dodecapeptide bacitracin A, which interacts tightly with C₅₅-PP in the presence of a bound divalent cation, thereby inhibiting the formation of C₅₅-P through the sequestration of its precursor. In 1993, Cain et al. reported the identification of a chromosomal gene from E. coli whose amplification in a high-copy-number plasmid confers resistance to bacitracin; this gene was named bacA (8). The authors hypothesized that the BacA protein might catalyze the phosphorylation of undecaprenol, the potential reserve pool of the lipid carrier (see below). BacA overproduction would, according to this hypothesis, result in additional supplies of C₅₅-P, bypassing the C₅₅-PP dephosphorylation step and thereby overcoming the inhibitory effect of bacitracin on peptidoglycan synthesis (8).

More recently, the role of E. coli bacA in C₅₅-P synthesis was investigated in more detail by our group (21). The highly hydrophobic 30-kDa BacA protein was overproduced and purified from membrane extracts. It was clearly demonstrated that this integral membrane protein catalyzes the dephosphorylation of C₅₅-PP, not the phosphorylation of undecaprenol as previously postulated. The resistance to bacitracin observed in bacA-overexpressing cells could then be explained by the fact that high-level expression of the enzyme should dramatically deplete the pool of C₅₅-PP, the target of bacitracin, thereby increasing the level of resistance toward the antibiotic; in other words, the overproduced enzyme would compete with bacitracin for C₅₅-PP binding.

Homologs of E. coli bacA in many bacteria have since been identified; these correspond to a large family of bacterial proteins, referred to as the PF02673 family in the PFAM database. The bacA genes of Staphylococcus aureus and S. pneumoniae were shown to be nonessential, but their disruption results in bacitracin hypersensitivity as well as decreased virulence in mouse models of infection (9). Likewise, the inactivation of the E. coli bacA gene is not lethal (21). The nonessential character of bacA was intriguing, considering the absolute requirement for C₅₅-P for cell viability, as exemplified by the bacteriolytic effect of the sequestration of C₅₅-PP by bacitracin and the fact that the UppS-encoding gene is essential (see above). In fact, El Ghachi et al. reported that the C₅₅-PP phosphatase activity present in a bacA null mutant is not totally abolished but is reduced by 75% compared to that in the wild-type strain (21). This finding suggested the existence of another protein(s) able to dephosphorylate C₅₅-PP and provide a sufficient supply of C₅₅-P in the bacA mutant to sustain its viability and growth. However, no bacA homolog in the genome of E. coli has been found.

Identification of a Novel Family of C55-PP Phosphatases

B. licheniformis, a bacitracin producer, possesses an ABC transporter that pumps out bacitracin as a mean of self-protection. The efflux pump is encoded by the bcrABC operon. It was initially assumed, based on the fact that the three genes form a single operon, that the efflux pump was a tripartite complex including all three proteins, BcrA, BcrB, and BcrC (62). B. subtilis 168, which does not produce bacitracin, expresses homologous proteins BceA, BceB, and BceC conferring cell resistance to the antibiotic (5, 52, 58, 59). However, in contrast to the homolog in B. licheniformis, the bceC gene, also referred to as ywoA, is not organized in an operon with the other two genes, and several lines of evidence have suggested that BceAB and BceC contribute independently to bacitracin resistance (5). Moreover, Harel et al. reported in 1999 that E. coli encodes a homolog of B. licheniformis BcrC of unknown function, named YbjG, whose disruption causes increased bacitracin sensitivity and whose overexpression, conversely, confers antibiotic resistance (34). It was also reported that all BcrC homologs, based on analyses of the primary sequences, belong to a superfamily of phosphatases listed in the PFAM database: the PAP2 superfamily (PF01569) (5).

Taken together, these observations suggested that BcrC is not part of the efflux pump, as initially considered, but rather is a protein contributing to bacitracin resistance by itself. BcrC was then hypothesized to catalyze the dephosphorylation of C₅₅-PP, thereby competing with bacitracin for C₅₅-PP binding, similar to BacA. Different studies were recently undertaken to elucidate the precise role of bcrC and its homologs. The membrane-bound BcrC enzyme from B. subtilis was purified, and it was subsequently demonstrated that BcrC indeed exhibits C₅₅-PP phosphatase activity comparable to that previously described for E. coli BacA (4).

E. coli encodes two other BcrC homologs in addition to YbjG (Fig. 5): the phosphatidyl glycerolphosphate phosphatase PgpB and a protein of unknown function, YeiU. All three homologs are predicted to be integral membrane proteins (22). The overproduction of the three proteins causes a significant increase of cellular C₅₅-PP phosphatase activity in membrane extracts, as measured in vitro. Their overproduction also causes a significant rise in the level of bacitracin resistance in vivo. Taken together, the data strongly suggest that, similar to BacA, the three enzymes catalyze the C₅₅-PP dephosphorylation reaction (22). Like the BacA gene, their genes may be independently disrupted without any significant effect on growth under laboratory conditions. The construction of different combinations of mutations by allelic replacement has been undertaken; mutation combinations involving the inactivation of the three genes bacA, ybjG, and pgpB could not be obtained, suggesting a lethal effect (22). The lethality of these triple-mutant combinations was further demonstrated by using a conditional triple mutant [bacA(Ts) ΔybjG ΔpgpB] in which bacA expression was impaired at 42°C and the other two genes were deleted. The triple mutant grew normally at the permissive temperature, but after a shift to the restrictive temperature, the cells rapidly lysed. This result showed that PgpB and YbjG, together with BacA, contribute to the total C₅₅-PP phosphatase activity required for E. coli cell viability. Interestingly, the presence of only one of the three chromosomal genes is sufficient for normal cell growth under laboratory conditions, raising the question of the relevance of multiple C₅₅-PP phosphatases in E. coli. Surprisingly, the overproduction of YeiU, which also leads to a significant increase in C₅₅-PP phosphatase activity as measured in vitro, is not able to restore the viability of the triple mutant, suggesting that YeiU may have another role besides dephosphorylating C₅₅-PP during carrier lipid synthesis. Indeed, we recently demonstrated that YeiU protein establishes a link between C₅₅-PP dephosphorylation and lipid A modification (89). Lipid A is an acylated disaccharide of glucosamine; it serves as the hydrophobic anchor of lipopolysaccharide and is required to maintain the integrity of the outer membrane (67). In E. coli, the majority of lipid A contains monophosphate moieties at the 1 and 4' positions, but approximately one-third of the lipid A molecules contain a diphosphate unit at the 1 position (termed lipid A 1-diphosphate) (66). We previously demonstrated that the deletion of yeiU results in the disappearance of lipid A 1-diphosphate species in the membrane and that purified YeiU exhibits phosphotransferase activity, utilizing specifically C₅₅-PP as a phosphate donor substrate to catalyze the formation of lipid A 1-diphosphate (Fig. 6) (89). Furthermore, we showed that the catalytic site of YeiU is oriented toward the periplasmic space by taking advantage of a thermosensitive strain in which MsbA is no longer able to export lipid A molecules across the inner membrane at the restrictive temperatures (Fig. 6) (10, 19). In the latter thermosensitive mutant, the phosphorylation of lipid A by YeiU was shown to occur only at the permissive temperatures. According to its involvement in the modification of the lipid A structure, YeiU was renamed LpxT (89).

Fig. 5Amino acid sequence alignment of PAP2 enzymes. Only the regions which encompass the three conserved motifs, designated C1, C2, and C3, are presented here. The corresponding sequences of five integral membrane proteins, the E. coli (EC) PAP2 C₅₅-PP phosphatases (PgpB, YbjG, and YeiU), B. subtilis YwoA (BS-YwoA), and B. licheniformis BcrC (BL-BcrC), along with the soluble globular protein E. blattae nonspecific acid phosphatase (EB-NSAP) are aligned. The first line shows the PAP2 consensus identified by Carman and Stuckey in 1997 (81). The consensus residues, as well as additional conserved residues, are indicated in red; the asterisks and the plus signs point out identical and similar residues, respectively. The line above the E. blattae NSAP sequence indicates the location of α-helices, designated H, based on the three-dimensional structure of the protein (41). The black lines below the sequences indicate the locations of the predicted transmembrane segments (T3 to T6) in E. coli PAP2 C₅₅-PP phosphatases. Unaligned residues are indicated by lowercase letters.

Fig. 6Model for C₅₅-P carrier lipid biogenesis pathway. C₅₅-PP, the precursor of the lipid carrier, is de novo synthesized in the cytoplasm, from which it is partitioned into the inner side of the cytoplasmic membrane, but C₅₅-PP is also released at the outer side of the membrane after the transfer of the glycan moieties to the acceptor polymers. In the proposed model, BacA ensures the dephosphorylation of the de novo-synthesized C₅₅-PP on the cytosolic side of the membrane whereas the PAP2 phosphatase enzymes, YeiU, YbjG, and probably PgpB, are involved exclusively in the recycling of C₅₅-PP on the outer side of the membrane. YeiU was shown to establish a link between C₅₅-P biogenesis and lipid A structure modification by transferring the distal phosphate group from C₅₅-PP to lipid A, generating free lipid carrier and lipid A 1-diphosphate (lipid A 1-diP) species. Because of its involvement in the modification of the lipid A structure, YeiU was renamed LpxT. As indicated by question marks, it is yet unknown how C₅₅-P and the C₅₅-PP-sugar complex are translocated from one leaflet of the membrane to the other.

In summary, two families of bacterial C₅₅-PP phosphatases coexist in E. coli: the BacA family and a novel family belonging to the phosphatase PAP2 superfamily. Although BacA and the PAP2 C₅₅-PP phosphatases show similar catalytic activities, their primary sequences are totally different and they belong to completely different PFAM families. These data raised the question of whether all these enzymes, in particular, the PAP2 family members, are involved exclusively in C₅₅-P synthesis. A recent study of the LpxT protein proved the contrary (89). Additionally, the PgpB enzyme has already been described as having broad substrate specificity in vitro, as it dephosphorylates phosphatidic acid and lysophosphatidic acid, as well as diacylglycerolpyrophosphate and phosphatidyl glycerolphosphate (18, 28). Whether this broad substrate specificity is shared by other PAP2 C₅₅-PP phosphatases and BacA remains to be established. Thus, the full characterization of the different proteins is now required to get the mechanistic details and to understand the respective roles of these proteins in the C₅₅-P synthesis pathway. The analysis of the expression of these different enzymes should also be undertaken for better insight into C₅₅-P metabolism regulation.

Cellular Localization of C55-PP Dephosphorylation

The different C₅₅-PP phosphatases are all membrane-bound proteins (see above); they are predicted to contain several transmembrane segments, six in PAP2 C₅₅-PP phosphatases and eight in BacA (unpublished data). This prediction raises the question of the localization of their respective catalytic sites with respect to the cytoplasmic membrane. C₅₅-P is required on the cytoplasmic side of the inner membrane, where the linkage to the glycan unit occurs. Its precursor, C₅₅-PP, is de novo synthesized in the cytoplasm, from which it is partitioned into the inner side of the membrane, but C₅₅-PP is also released on the outer side of the membrane after each late polymerization reaction.

As already mentioned, the YeiU-LpxT reaction was shown to take place on the periplasmic side of the membrane (89). This finding was supported by the results of recent experiments with PhoA or green fluorescent protein fusion proteins that demonstrated that the PAP2 phosphatase signature residues of LpxT and YbjG are oriented toward the periplasm (85). Taking into account that PgpB belongs to the same family as LpxT and YbjG, it is not unreasonable to hypothesize that the PgpB active site may also face the periplasm (Fig. 6). In contrast, the membrane topology of BacA, which lacks any features of a typical phosphatase, is yet unknown. According to these data, it is tempting to speculate that the PAP2 phosphatase enzymes may be involved exclusively in the recycling of C₅₅-PP that is released after the transfer of the glycan moieties but that BacA may ensure the dephosphorylation of the de novo-synthesized C₅₅-PP on the cytosolic side of the membrane (Fig. 6) (85). However, the observation that only one copy of either BacA, YbjG, or PgpB is required for cell viability (22) suggests that any one of the latter three phosphatases may act on the de novo-synthesized C₅₅-PP, but the involvement of another, unidentified C₅₅-PP phosphatase specifically carrying out this reaction on the inner side of the membrane can still be imagined. The first hypothesis implies that the de novo-synthesized C₅₅-PP is translocated across the membrane to reach the site of dephosphorylation and, conversely, that C₅₅-P is efficiently transported back to the inner side of the membrane, where the transfer of the glycosyl moieties onto the lipid carrier occurs. This idea is supported by the fact that a high rate of transbilayer movement of the lipid carrier has been observed in "Micrococcus lysodeikticus" (Micrococcus luteus) (53, 54), and finally, it leaves open the relevant question of the BacA membrane topology.

The mechanism by which C₅₅-P and C₅₅-PP are translocated from one side of the membrane to the other has not yet been identified, and whether it is the same process that mediates the transfer of the free lipid carrier or the different C₅₅-PP–sugar complexes is also not yet known. All these questions will have to be addressed in the coming years.

PAP2 Enzymes

E. coli YbjG, LpxT, and PgpB and their homologs form a new class of proteins within the PAP2 phosphatase superfamily that was identified by Stukey and Carman in 1997 (81, 86). The members of this superfamily, which are largely widespread among all kingdoms, are characterized by a conserved signature: K(X₆)RP-(X_12-54)-PSGH-(X_31-54)-SR(X₅)H(X₃)D (Fig. 5). Among the various members of this superfamily, the homology is generally limited strictly to the signature sequence, in which three distinct motifs are visible, designated C1, C2, and C3, respectively. Each motif consists of 4 to 11 residues, including 3 to 4 conserved amino acids, and the order of the three motifs in the primary structure is always the same. The PAP2 superfamily comprises two different types of enzymes, the histidine phosphatases and the vanadium haloperoxidases, including type 2 phosphatidic acid phosphatases, or lipid phosphate phosphatases, the mammalian glucose-6-phosphate phosphatases, bacterial acid phosphatases, vanadium chloroperoxidases, vanadium bromoperoxidases, and several mostly uncharacterized enzymes (56). Interestingly, this superfamily offers an example in which soluble globular enzymes and integral membrane proteins share the same signature sequence. This trait implies that the architectures of the active sites should be similarly organized in very different types of proteins that have emerged most likely through divergent evolution. The crystal structures of two soluble PAP2 proteins, the chloroperoxidase from the fungus Curvularia inaequalis (55) and the nonspecific acid phosphatase (NSAP) from Escherichia blattae (41), in complexes or not with vanadate and molybdate, respectively, have been determined. The ligand-bound structures are likely to reflect the transition state intermediate-bound enzymes, as the ligands used are both analogous to the phosphate group. These structures have revealed that the catalytic site, as well as the overall structure, is conserved in the two enzymes, even though the similarity is restricted to the three conserved motifs (38, 51). Noteworthily, the overall structures of these enzymes revealed a very flat, mostly α-helical organization. The three catalytic motifs (C1, C2, and C3) are in close proximity to one another, and several conserved residues are involved in the substrate binding and/or catalysis in a concerted manner (Fig. 7). Structure-function analyses of different eukaryotic members have confirmed the role of some of the conserved residues in the catalytic process (see below) (29, 30, 75).

Fig. 7Transition state structure of the PAP2 E. blattae NSAP polypeptide with molybdate. (A) The ribbon diagram of the central region encompassing residues 91 to 210 is shown; the entire molecule comprises 249 amino acids. The three catalytic domains, C1, C2, and C3, are localized at the end of four long α-helices, and they are colored in red, grey, and pink, respectively. The catalytic triad residues and the molybdate ion are shown in line models; Mo, C, N, and O atoms are in blue, green, dark blue, and red, respectively. (B) Detailed interactions of molybdate with the catalytic triad and three other consensus residues. The dotted yellow lines indicate polar contacts. N-ter, N terminus; C-ter, C terminus.

Comparison of Soluble and Membrane PAP2 Enzymes

The three motifs forming the active site in the soluble globular enzymes are located at the ends of four long antiparallel α-helices that are part of a four-helix bundle (Fig. 7). Similarly, in the membrane-spanning LpxT, YbjG, and PgpB proteins, the signature residues are located at the edges of four predicted transmembrane α-helices (Fig. 5). Thus, we can speculate that the overall scaffold surrounding the active site may be conserved in both soluble and membrane proteins. Noteworthily, some of the signature residues in the membrane enzymes, in particular, the residues from the C2 and C3 motifs, are predicted to be localized at the membrane interface or may even be deeply embedded in the membrane (Fig. 5). This possibility implies that the binding pocket should be formed in close proximity to the phospholipids’ hydrophobic core, possibly providing the phosphatase specificity toward lipid phosphate substrates. Thus, in regard to the localization of the catalytic site with respect to the membrane, C₅₅-PP may interact with the enzyme after lateral diffusion into the membrane, so that the pyrophosphate group should dock directly into the active site. In such a way, the substrate aliphatic chain would then be positioned near the transmembrane α-helices, with which it could thus interact in a more or less specific manner. After phosphate hydrolysis, C₅₅-P should diffuse away into the membrane plane, freeing the active site for another cycle of dephosphorylation.

Catalytic Mechanism of PAP2 Enzymes

PAP2 enzymes are histidine phosphatases, and their catalysis takes place by a two-step mechanism (56, 95). First, there is a nucleophilic attack on the phosphoryl group by a histidine, resulting in the formation of a covalent bond between a nitrogen atom of the imidazole ring of histidine and the phosphate group. In PAP2 enzymes, this catalytic histidine was deduced from the haloenzyme structures to belong to motif C3, since this amino acid is covalently bound to the ligand (Fig. 7). In order to form this phosphoenzyme catalytic intermediate, a charge-relay system involving the aspartate residue from motif C3 is established (Fig. 7). The phosphate transfer must then be terminated by the protonation of the substrate-leaving group by another domain of the enzyme, the best candidate being the histidine residue from motif C2 (Fig. 7). This scenario has been confirmed by mutagenesis analyses of the mammalian glucose-6-phosphatase enzyme (29). In the second and last catalytic step, the phosphoenzyme intermediate must be hydrolyzed, releasing the enzyme and inorganic phosphate. By acting this time as a base, the histidine residue from motif C2 may facilitate this nucleophilic attack of the phosphate group by water or another acceptor molecule.

Apart from the catalytic triad, the other signature residues are also assumed to participate in the catalytic process (41). Based on findings from crystal structure examination, they are likely to be important in both attracting the negatively charged phosphate group and stabilizing it in the transition state structure. These residues comprise a cluster of positively charged amino acids surrounding the catalytic C3 histidine residue, the lysine and arginine residues from C1, and the arginine residue from C3 (Fig. 5). The serine and glycine residues from the C2 motif are also important in contributing to the holding of the phosphate group in the crystal structure. The other conserved residues, in particular, the two proline residues from C1 and C2, are assumed to be structurally important rather than participating in the enzymatic process per se.

Comparison of E.coli PAP2 C55-PP Phosphatases

As previously mentioned, the YbjG and LpxT enzymes have been demonstrated to have their catalytic sites located on the periplasmic side of the plasma membrane (85), and the active site of their homolog PgpB is likely to be oriented likewise. The latter three proteins are predicted to contain six transmembranes α-helices (unpublished data) (Fig. 8) and to have both their amino- and carboxyl-terminal ends oriented toward the cytoplasm, i.e., opposite the localization of the catalytic site that is carried by loops 3 and 5. On the noncatalytic side, the putative extramembranous regions are very limited in size, whereas on the catalytic side, the loops 1 and 3 have variable lengths (see below). Among the PAP2 signature residues, the catalytic triad mediating the nucleophilic attack on the lipid phosphate ester bond, as deduced from the crystal structures, namely, C2 histidine and C3 histidine and aspartate, is completely conserved in the three C₅₅-PP phosphatases, which suggests that the catalytic mechanism is perfectly conserved. Interestingly, all the other signature residues are also conserved in PgpB. In contrast, YbjG and LpxT possess several variations (Fig. 5). In YbjG, the C1 lysine is replaced by a glycine and the C2 glycine is replaced by an aspartate. Finally, LpxT contains many variations within the consensus sequence: in C1, the lysine is replaced by a leucine and the arginine is replaced by a serine; in C2, the serine is replaced by a glycine and the glycine is replaced by an aspartate; and in C3, the serine is replaced by a proline. The fact that LpxT presents several "mutations" within its catalytic pocket may diminish its pyrophosphatase activity compared to those of its homologs, thereby explaining why it is not able to complement the triple mutant while it displays C₅₅-PP phosphatase activity in vitro. Nevertheless, these variations must also be functionally relevant in providing substrate specificity as well as function specialization, thereby explaining the occurrence of several PAP2 C₅₅-PP phosphatase homologs in E. coli.

Fig. 8Topological model of E. coli PAP2 C₅₅-PP phosphatases. N-ter, N terminus; C-ter, C terminus; T1 to T6, transmembrane segments 1 to 6; L1 to L5, loops.

We notice that apart from the signature residues, additional amino acids are conserved among the PAP2 C₅₅-PP phosphatase enzymes. When comparing the primary structures of E. coli homologs, along with those of B. subtilis and B. licheniformis BcrC proteins, we found additional conserved hydrophobic residues in the C2 and C3 motifs; these residues include a phenylalanine in the C2 motif, as well as a glycine and a tryptophan in the C3 motif (Fig. 5).

Among the significant differences among E. coli PAP2 C₅₅-PP phosphatases, we notice that the putative loop 3, which bears the C1 and C2 sequences, is larger in PgpB (66 amino acids) than in LpxT and YbjG (33 and 22 amino acids, respectively). Additionally, LpxT possesses a putative larger loop 1, 36 amino acids in length, whereas this corresponding loop is composed of fewer than 20 residues in the other two homologs. These loops may be involved in the regulation of catalysis, considering that they are localized on the same side of the catalytic site with respect to the membrane. It has been reported that eukaryotic soluble PAP2 proteins possess a pattern within the corresponding region that was assumed to specify a distinct function or to be involved in regulation (40, 64). These observations provide a framework for site-directed mutagenesis and kinetic studies to reveal the mechanistic details and specificities of the different E. coli homologs.

Phosphotransferase Activities

The recent discovery of the phosphotransferase activity of LpxT raises the question of whether the other different PAP2 C₅₅-PP phosphatases may also catalyze the transfer of the distal phosphate group of C₅₅-PP to different acceptor molecules. We previously showed that LpxT was the only one catalyzing the transfer of the phosphate group from C₅₅-PP to lipid A, but the other two PAP2 enzymes may well use another periplasmic molecule as an acceptor, conferring some kind of specialization on these different homologs. The mammalian sphingomyelin synthases, which form a specific class among the PAP2 superfamily (64), transfer the phosphocholine from phosphatidylcholine onto ceramide. These membrane proteins contain an additional pattern that is predicted to be present in the first loop and has been postulated to participate in concert with the conventional C1, C2, and C3 motifs in the phosphotransfer (64).

UNDECAPRENOL PHOSPHORYLATION

During the 1970s, several studies reported that the gram-positive bacteria S. aureus and Lactococcus plantarum contain a membrane-associated ATP-dependent kinase that catalyzes the conversion of free undecaprenol into C₅₅-P (39, 42, 72, 73, 74). The existence of such an activity suggested that the C₅₅-isoprenoid alcohol is present in these bacteria, which was proved to be the case for S. aureus (39) and Lactobacillus plantarum (32, 87), as well as Enterococcus faecalis (90) and Listeria monocytogenes (94). In these bacteria, undecaprenol should originate from the dephosphorylation of C₅₅-P and it was assumed to form a reserve pool, which may thus participate in the regulation of the pool level of the active form of the lipid carrier. The formation of this pool implies the existence of an undecaprenol phosphatase enzyme that hitherto has never been described. Despite the fact that undecaprenol kinase activities were found in some of these bacteria, the identities of the proteins were not determined.

This issue remained outstanding until recently, when Lis and Kuramitsu showed that diacylglycerol kinase (DGK) from Streptococcus mutans was able to catalyze the phosphorylation of undecaprenol in vitro (50). This alternative function of DGK was further supported by the demonstration that a dgk mutant of S. mutans exhibits increased susceptibility toward bacitracin, suggesting that, in the mutant, the pool of undecaprenol can no longer be utilized to bypass C₅₅-PP dephosphorylation to yield C₅₅-P.

The existence of a pool of free undecaprenol in other bacteria, in particular, enteric bacteria, as well as the presence of C₅₅-P phosphatase or undecaprenol kinase activities, has never been reported, raising the question of the existence of such a pool in these bacteria. Moreover, the DGK from E. coli was proved to be inactive toward undecaprenol (6, 50). Recently, we have been able to detect undecaprenol in colicin M-treated E. coli cells, whereas we did not observe this compound in nontreated cells (20). Colicin M cleaves the lipid I and lipid II peptidoglycan precursors between the undecaprenyl and the 1-pyrophospho-MurNAc moieties, thereby causing the accumulation of undecaprenol. All together, these data support the hypothesis that a reserve pool of undecaprenol may not exist in E. coli cells. This important issue in C₅₅-P synthesis and recycling should be clarified in the near future.

CONCLUSION AND FUTURE RESEARCH ISSUES

The 55-carbon C₅₅-P is a key lipid in metabolism in E. coli, as well as in most bacteria. It constitutes the branching point for the synthesis of very disparate cell wall components, in particular, the peptidoglycan; the disruption of peptidoglycan synthesis causes rapid cell lysis. C₅₅-P facilitates the portage of glycosyl units across the cytoplasmic membrane; however, the mechanism by which the free lipid carriers or those in complexes are translocated remains a central question. It was recently demonstrated that the translocation of a sugar–C₅₅-P complex, peptidoglycan precursor lipid II, across the membrane does not occur spontaneously, implying the existence of a translocation machinery likely composed of one or more integral membrane proteins (91). Rick et al. and Rush et al. proposed that a flippase may mediate this translocation event similarly to the well-characterized MsbA protein, which translocates the phospholipids and lipid A molecules from one side of the inner membrane to the other (19, 69, 71, 100). Alternatively, Zhou and Troy have proposed a mechanism involving the glycosyltransferases themselves, which upon binding with the lipid carrier would form a membrane channel through which the lipid-sugar complex could pass (98, 99).

The metabolism of C₅₅-P has been overlooked for many years, up to the identification about 10 years ago of the enzyme catalyzing the essential first step of C₅₅-PP synthesis, the undecaprenyl pyrophosphate synthase (UppS). Thereafter, apo- and halo-UppS structures were resolved, and together with the findings of intensive mutagenesis and biochemical studies, these structures have provided a high level of understanding of C₅₅-PP synthesis. In contrast, the second step of C₅₅-P metabolism remains a large research issue for the coming years. We have highlighted two families of membrane proteins that participate in concert in this step; in E. coli, C₅₅-PP dephosphorylation can proceed through the action of the BacA protein and three PAP2 C₅₅-PP phosphatases, YbjG, PgpB, and LpxT. Interestingly, none of these enzymes are required for growth under laboratory conditions, raising the question of the relevance of the various C₅₅-PP phosphatases, in particular, the PAP2 homologs in E. coli. To date, we have no structural or mechanistic details relevant to BacA protein, and mechanistic hypotheses regarding the PAP2 phosphatases have been formulated only on the basis of the comparison of these phosphatases with soluble orthologs whose structures have been resolved. Therefore, the two types of enzymes should be the object of future intensive investigations to get further insights into their structures and catalytic mechanisms and to answer the central question of their respective roles in C₅₅-P synthesis. The main difficulty in analyzing the C₅₅-PP phosphatase enzymes resides in the fact that they are all integral membrane proteins. Biochemical and structural studies of such proteins have proved to be very difficult, even though great progress has been made in this field in recent years. Moreover, this fact raises the question of their topology and the localization of their catalytic sites with respect to the cytoplasmic membrane. We have also pointed out that C₅₅-P is actively recycled following its release after the transfer of the carbohydrate moiety to the growing glycan strand. The recycling implies that C₅₅-P moves back to the inner side of the cytoplasmic membrane by a mechanism that is yet to be identified.

The pool level of C₅₅-P is critical for bacterial cell survival, as exemplified by the different mechanisms of bacitracin resistance. Indeed, the resistance can be achieved by either the overexpression of a C₅₅-PP phosphatase competing with the antibiotic for C₅₅-PP binding (8, 34) or the inhibition of the synthesis of a nonessential cell wall component, such as the exopolysaccharides and the membrane-derived oligosaccharides, thereby providing a supplementary supply of free C₅₅-P (24, 63). Thus, tight regulation is likely to take place, which controls the pool level of C₅₅-P and its distribution between the different pathways to ensure cell integrity and harmonious cell growth. Interrelation between the biosynthesis of peptidoglycan and that of some other cell wall components has been highlighted (2), suggesting that the pool level of C₅₅-P is the limiting factor in the synthesis of the various cell wall components, but hitherto little has been known about the mechanisms that control C₅₅-P synthesis, and the measurement of the pool level of C₅₅-P and that of its precursor has never been the subject of great attention. Thus, the mechanisms that participate in the regulation of C₅₅-P synthesis will also be of a great interest in the coming years.

ACKNOWLEDGMENTS

We thank Didier Blanot for critical reading of the manuscript.

This work was supported by the European Community (FP6 projects COBRA).