Expression of the Streptococcus pneumoniae Type 3 Synthase in Escherichia coli ASSEMBLY OF TYPE 3 POLYSACCHARIDE ON A LIPID PRIMER*

Synthesis of the type 3 capsular polysaccharide of Streptococcus pneumoniae is catalyzed by the mem-brane-localized type 3 synthase, which utilizes UDP-Glc and UDP-GlcUA to form high molecular mass [3- (cid:1) - D -GlcUA-(1 3 4)- (cid:1) - D -Glc-(1 3 ] n . Expression of the synthase in Escherichia coli resulted in synthesis of a 40-kDa protein that was reactive with antibody directed against the C terminus of the synthase and was the same size as the native enzyme. Membranes isolated from E. coli contained active synthase, as demonstrated by the ability to incorporate Glc and GlcUA into a high molecular mass polymer that could be degraded by type 3 polysaccha-ride-specific depolymerase. As in S. pneumoniae , the membrane-bound synthase from E. coli catalyzed a rapid release of enzyme-bound polysaccharide when incubated with either UDP-Glc or UDP-GlcUA alone. The recombinant enzyme expressed in E. coli was capable of releasing all of the polysaccharide from the enzyme, although the chains remained associated with the membrane. The recombinant enzyme was also able to reini-tiate polysaccharide synthesis following polymer release by utilizing a lipid primer present in the membranes. At low concentrations of UDP-Glc and UDP-GlcUA (1 (cid:2) M in the presence of Mg 2 (cid:3) Analytical Methods— Chromatography m M ammonium m M m M polysaccharide-specific membranes m

Type 3 polysaccharide, which is composed of the repeating subunit [3)-␤-D-GlcUA-(13 4)-␤-D-Glc- (13], is one of 90 differ-ent capsule types identified in the human pathogen Streptococcus pneumoniae (1)(2)(3). Synthesis of type 3 polysaccharide requires only a single glycosyltransferase, which utilizes UDP-Glc and UDP-GlcUA to form both glycosidic linkages of the repeating disaccharide (4 -6). The type 3 synthase shares significant homology with a number of processive ␤-glycosyltransferases, including the hyaluronan synthases from prokaryotes and eukaryotes, the cellulose synthases from plants and bacteria, the chitin synthases from yeast, and the Nod factor synthases from Rhizobium (5, 7) It has recently been classified in family 2 of the glycosyltransferases (8). Elongation of type 3 chains in pneumococcal membranes has been demonstrated to occur at the nonreducing end of the polysaccharide (9), consistent with a previously suggested mechanism of growth for members of the processive ␤-glycosyltransferase family that involves the dual addition of sugars to the growing polymer (10,11). An improved understanding of the three-dimensional structure of inverting glycosyltransferases has generated a more recent proposal that most family transferases utilize conserved structural domains and four conserved aspartates to form a single center where both the binding of the UDP-sugars and the glycosyl transfer reaction would take place (12). It was further speculated that a single active-site transferase could catalyze the formation of an alternating polysaccharide if the addition of each sugar served to fine-tune the affinity of the acceptor site for the succeeding nucleotide-sugar.
In a recent study of the polysaccharide release reaction mediated by the type 3 synthase in isolated S. pneumoniae membranes, the presence of either UDP-Glc or UDP-GlcUA was shown to dramatically affect the binding affinity of the enzyme for the conjugate UDP-sugar (13). In addition, polymer synthesis is significantly impaired following UDP-sugar-mediated release of the polysaccharide from the synthase in S. pneumoniae membranes, suggesting that some additional factor may be necessary for reinitiation of polymer synthesis (13). To further explore the chain initiation process as well as possible interactions between the nucleotide-sugar-and carbohydrate-binding sites, we have characterized the biosynthetic and polysaccharide release reactions of the recombinant type 3 synthase contained in Escherichia coli membranes.
Growth Conditions and Membrane Preparations-S. pneumoniae membranes from type 3 strain WU2 (16) and its non-encapsulated derivative JD908 (17) were prepared as previously described (9). Membranes were stored at Ϫ80°C in a solution containing 100 mM Hepes (pH 8.0), 10% glycerol, and 10 mM sodium thioglycolate. E. coli strains JD422 and JD424 have been described (5). JD422 is E. coli TG-1 containing the expression vector pKK223-3. JD424 is TG-1 containing cps3S cloned into pKK223-3. The clone contains a 2.1-kb Sau3AI fragment from S. pneumoniae WU2 that contains the 3Ј-end of cps3D and the complete cps3S. E. coli strains were grown in L-broth (1% Tryptone, 0.5% yeast extract, 0.5% NaCl, and 0.1% glucose) containing 100 g/ml ampicillin at 37°C with constant shaking to a cell density of 1.2 ϫ 10 9 cfu/ml. 1 Expression of the synthase (cps3S) was induced with 1 mM isopropyl-␤-D-thiogalactopyranoside, and the culture was grown for an additional 2 h at 37°C. The cells were harvested by centrifugation at 4°C for 10 min at 10,000 ϫ g, washed twice with phosphate-buffered saline (137 mM NaCl, 2.7 mM KCl, 5.4 mM Na 2 HPO 4 ⅐7H 2 O, and 1.8 mM KH 2 PO 4 (pH 7.4)) containing 10% glycerol, and frozen at Ϫ80°C. Cell pellets were thawed and suspended to 1% of the original culture volume in 20% sucrose, 30 mM Tris-HCl (pH 8.2), 10 mM MgCl 2 , 1 mM dithiothreitol, 1 mM EDTA (pH 8.0), 0.5 g/ml leupeptin, and 0.7 g/ml pepstatin. Lysozyme was added to 400 g/ml, and the suspension was incubated at 4°C for 40 min with constant mixing. The suspension was centrifuged at 4°C for 10 min at 12,000 ϫ g, and the pellet was suspended in sterile deionized water at 1% of the original culture volume. Phenylmethanesulfonyl fluoride, MgCl 2 , DNase, and RNase were added to final concentrations of 1 mM, 60 mM, 1 g/ml, and 1 g/ml, respectively, and the suspension was sonicated three times for 30 s using a microtip. The suspension was incubated at 4°C for 20 min with constant mixing. Membranes were harvested by centrifugation at 100,000 ϫ g for 1 h. The membranes were washed once with 100 mM Hepes (pH 7.5) containing 10% glycerol and 10 mM EDTA and twice with 100 mM Hepes (pH 7.5) containing 10% glycerol. The final pellet was suspended to 0.5% of the original culture volume in 100 mM Hepes (pH 7.5) containing 10% glycerol and stored at Ϫ80°C.
Assay of Synthase Activity-Type 3 synthase activity was determined using 0.02 Ci of either UDP-[ 14 C]Glc or UDP-[ 14 C]GlcUA in 100-l reactions containing 100 mM imidazole (pH 7.0), 10 mM MnCl 2 , UDP-Glc, UDP-GlcUA, and membranes isolated as described above. The reactions were incubated at 35°C for 10 min and terminated by the addition of 10 l of 12.5 M acetic acid. The reaction components were separated by ascending paper chromatography on Whatman No. 3MM paper with a solvent containing 95% ethanol and 1 M ammonium acetate (pH 5.5) (65:35, v/v) for 16 -18 h and quantified by liquid scintillation counting.
Preparation of the 14 C-Labeled Polysaccharide-Synthase Complex-Type 3 polysaccharide was synthesized in a 400-l reaction mixture consisting of 100 mM imidazole (pH 7.0), 10 mM MnCl 2 , 20 M UDP-Glc, 20 M UDP-[ 14 C]GlcUA (135 Ci/mmol), and E. coli membranes (7 mg of protein). The reaction was incubated for 30 s at 35°C and stopped by placing on ice. The volume of the reaction was brought to 2 ml with 100 mM Hepes (pH 7.5) containing 10% glycerol, and the membranes were washed five times as previously described (13). A high concentration of protein was necessary for this experiment because of the short reaction time.
Preparation of Glycolipid Products-Glycolipids were synthesized in reaction mixtures containing 0.3 Ci of labeled nucleotide-sugar, other nucleotide-sugars as indicated, 100 mM Hepes (pH 7.5), 10 mM MgCl 2 , and E. coli membranes (2 mg of protein) in a total volume of 0.6 ml. Following a 12-min incubation at 35°C, the reaction was terminated by the addition of 2 ml of an ice-cold solution containing 100 mM Hepes (pH 7.5) and 10% glycerol, and the membranes were sedimented by centrifugation at 100,000 ϫ g for 30 min. The membranes were washed two more times with 2.5 ml of wash solution as described above. The membranes were suspended in 0.2 ml of 0.1 M NaCl and heated at 100°C for 5 min. The labeled glycolipids were extracted with 0.25 ml of a solution containing 0.1 M NaCl, 0.3% Nonidet P-40, and 2 mM EDTA (pH 7.5). The insoluble material was sedimented by centrifugation at 13,000 ϫ g for 5 min; the supernatants were saved; and the pellets were extracted a second time. The supernatants containing the glycolipid fraction were combined for further analyses.
Saponification, Acid Hydrolysis, and Enzymatic Digestions of Glycolipid Products-Saponification was carried out at 35°C for 20 min in 0.4 ml of 80% methanol containing 0.1 N NaOH. The mixture was neutralized with acetic acid, and the methanol was removed with a stream of nitrogen. Acid hydrolysis was carried out in 1.0 N HCl at 100°C for the indicated times. The hydrolysates were neutralized with NaOH, and samples were analyzed by Sephacryl S-300 and paper chromatography. Phospholipase digestions were conducted in reaction mixtures of the appropriate buffer, pH, and temperature as recommended by Sigma, with the addition of Nonidet P-40 and 10 mM CaCl 2 , which have been shown to enhance these activities (18,19). A minimum of 10 units of phospholipase was added, except for 1 unit of C. perfringens phospholipase C. The reactions were terminated by heating to 100°C for 4 min; the mixtures were clarified by centrifugation at 13,000 ϫ g for 5 min; and the products were analyzed by paper, thinlayer, and Sephacryl S-300 chromatography as indicated. Digestions with almond ␤-glucosidase (5 units for 3 h) and E. coli ␤-glucuronidase (276 units for 6 h) were conducted as described previously (9). Prior to incubation with ␤-glucuronidase, the GlcUA-containing lipid was digested with phospholipase D, and the liberated component was purified by chromatography in butanol/acetic acid/water (44:16:40).

Type 3 Synthase Expression in E. coli-
We previously reported that overexpression of the type 3 synthase in E. coli results in death of the cells, with no detectable accumulation of protein, as determined by Coomassie Blue and silver staining (5). Arrecubieta et al. (6) observed similar toxic effects, but were able to observe low levels of recombinant protein. Using a rabbit polyclonal antiserum specific for the C-terminal 14 residues of the synthase, we have confirmed expression of the protein in our original strains and shown that it is of the same apparent molecular size as that expressed in S. pneumoniae (Fig. 1). The apparent molecular size (40 kDa) was smaller than the predicted size (48 kDa), consistent with observations for the streptococcal hyaluronan synthase (20). It was in contrast, however, to the apparent molecular size of 49 kDa reported by Arrecubieta et al. (6) for the type 3 synthase. In our system, cps3S expression was leaky, resulting in a constant low level production of the synthase that was increased only modestly by induction with isopropyl-␤-D-thiogalactopyranoside ( Fig. 1). The presence of the protein was detectable only with the synthase-specific antiserum or using the assays described below.
Type 3 synthase activity was determined by the incorporation of 14 C from either UDP-[ 14 C]Glc or UDP-[ 14 C]GlcUA into a high molecular mass product. Membranes from the recombinant strain (JD424) actively synthesized polymer when incubated with high concentrations (100 M) of the UDP-sugars, whereas only minimal incorporation (Ͻ3% of JD424 levels) occurred with membranes from the control strain containing the vector only (JD422). The activity of the recombinant enzyme was dependent on the presence of both UDP-Glc and UDP-GlcUA, was linear for 30 min, and was proportional to protein concentration. Activity was observed in the presence of both Mn 2ϩ and Mg 2ϩ and was optimal at pH 7.0 -8.0, which compares with an optimal pH range of 7.0 -8.5 in S. pneumoniae (9,13). The apparent K m values for UDP-Glc and UDP-GlcUA were 26 and 20 M in the presence of Mn 2ϩ and 76 and 58 M in the presence of Mg 2ϩ , respectively. Using S. pneumoniae membranes, the respective values were 12, 8.5, 65, and 31 M (9). The high molecular mass product obtained with the E. coli membranes was similar in size to that observed with the S. pneumoniae membranes and was confirmed to be type 3 polysaccharide by digestion with a type 3-specific depolymerase from Bacillus circulans. The E. coli product, like the S. pneumoniae product (9), was degraded by the depolymerase to a tetrasaccharide (data not shown).
Polysaccharide Association with the Enzyme-Membrane Complex-To assess whether polysaccharide remained associated with the E. coli synthase-membrane complex during the course of the biosynthetic reaction, samples were taken from reaction mixtures and sedimented by centrifugation to separate released polysaccharide from membrane-associated polysaccharide. Virtually all of the polysaccharide product remained bound to the membranes even though most of the UDP-Glc was depleted by 30 min (data not shown). These results suggested that the polysaccharide was not released from the enzyme-membrane complex, which was in sharp contrast to the results observed with membranes from S. pneumoniae (9,13). When polysaccharide release was assayed directly by incubating E. coli membranes in the presence of a single UDP-sugar, no significant release of polysaccharide from the membranes was observed (data not shown). Furthermore, there was no significant inhibition of polymer synthesis using E. coli membranes that had been pretreated with either UDP-Glc or UDP-GlcUA, which again was in contrast to the results observed in S. pneumoniae (13).
To explore the possibility that polymer was being released from the enzyme but not from the membrane, we examined polysaccharide synthesis in a dual isotope experiment. First, [ 14 C]GlcUA-labeled polysaccharide-synthase-membrane complex was prepared (as described under "Experimental Procedures") in a brief (30 s) reaction ( Fig. 2A). Then, the labeled membranes were incubated with either UDP-Glc (Fig. 2B) or UDP-GlcUA (Fig. 2C) prior to a second round of polysaccharide synthesis in a reaction mixture containing UDP-[ 3 H]Glc and UDP-GlcUA. In each instance, the formation of a high molecular mass 3 H-labeled product was observed, but none of the 14 C-labeled material was elongated. In contrast, when the membranes were preincubated in the absence of either UDPprecursor, approximately half of the 14 C-labeled product was extended into a higher molecular mass form that coeluted with the 3 H-labeled polysaccharide (Fig. 2D). Both the 14 C-and 3 H-labeled high molecular mass products were degraded by the type 3-specific depolymerase, indicating that both labels were incorporated into type 3 polysaccharide. These data indicate that incubation with a single substrate will actuate release of all the 14 C-labeled polysaccharide from the enzyme and that the synthase is capable of reinitiating synthesis. These results contrast with those obtained in S. pneumoniae, where ϳ50% of the chains were released, and reinitiation did not occur (13).
Further experiments were carried out at low protein/substrate ratios to minimize changes in the substrate concentration during the course of the reaction. In a reaction containing 5 M UDP-precursors and MgCl 2 , two products were present following a 30-min incubation (Fig. 3). Over 80% of the larger product was chased into a higher molecular mass polysaccharide by continuing the incubation with 200 M concentrations of the UDP-sugars. These data indicated that Ͻ20% of the polysaccharide chains had been released from the synthase during the initial 30-min incubation. In contrast, a low molecular mass product corresponding in size to detergent micelles was not chased into a larger product. The low molecular mass product, which was the primary product at low UDP-sugar precursor concentrations (2 M or less), was further characterized.
Identification of the Low Molecular Mass Product as a Glycolipid-At low UDP-sugar concentrations, membranes from the synthase-containing strain (JD424) and from the vector control strain (JD422) incorporated radioactivity into products with very different properties. Membranes from JD422 incorporated Glc into a product that remained at the origin and another that migrated ϳ75% of the distance of the Glc standard (Fig. 4A). When analyzed by Sephacryl S-500, the origin product appeared to be a high molecular mass polymer, and the second product corresponded in size to a small oligosaccharide (data not shown). The JD422 membranes were unable to utilize UDP-GlcUA as a substrate, and the presence of UDP-GlcUA did not stimulate the formation of any additional Glc-labeled products (Fig. 4A). When incubated with low concentrations of UDP-sugars, membranes from the synthase-containing JD424 strain were greatly reduced in the above activities and instead incorporated isotope into products with different chromatographic properties. The GlcUA-labeled products (synthesized in reactions containing only UDP-GlcUA) migrated more rapidly than a monosaccharide when analyzed by paper chromatography (Fig. 4B). When incubated with UDP-Glc alone, JD424 membranes incorporated low levels of Glc into compounds with a chromatographic pattern similar to that of the GlcUA-labeled products. The presence of unlabeled UDP-GlcUA in the reaction mixture stimulated the incorporation of Glc 10-fold, suggesting copolymerization of these two sugars. Following saponification with 0.1 N NaOH for 15 min at 37°C (Fig. 5A) or acid hydrolysis for 7 min in 1 N HCl at 100°C (data not shown), both the GlcUA-and Glc-labeled products (synthesized in the presence of UDP-GlcUA) migrated more slowly than a monosaccharide when separated by paper chromatography and yielded a pattern that was suggestive of a series of oligosaccharides. Ninety percent of the products were liberated by acid hydrolysis for 7 min in 1 N HCl at 100°C, and 100% were liberated in 10 min. Fifty percent were liberated in 30 min in 0.1 N HCl at 100°C.
When separated by gel filtration on Sephacryl S-300, both the GlcUA-and Glc-labeled products eluted at a position sim- ilar to that of Nonidet P-40 micelles (Fig. 5B). After saponification (Fig. 5B) or acid hydrolysis (data not shown), they eluted at the same position as small oligosaccharides. These properties are indicative of glycolipids, which might be expected to be soluble in organic solvents. Table I shows the distribution of the products when partitioned with chloroform, methanol, and water, similar to the procedure of Bligh and Dyer (21). At a neutral pH, 77% of the GlcUA-labeled products and 89% of the Glc-labeled products were found in the aqueous phase; but in an acidic solvent, 76 and 72% of these respective products partitioned at the interphase or in the organic phase. All of these properties are consistent with those expected for anionic glycolipids containing oligosaccharides of moderate length.
The glycolipid products were unaffected by heating with 50% phenol at 70°C for 2.5 h, which clearly distinguished them from undecaprenyl diphosphate and monophosphate sugars, which, under this condition, have respective half-lives of 4 -10 and 60 min (22). The products were completely hydrolyzed by phospholipase D from S. chromofuscus as assessed by paper chromatography using a solvent of butanol/acetic acid/water (44:16:40). Most of the digested Glc-labeled product remained at the origin, suggesting a composition of somewhat larger oligomers than the GlcUA-labeled product (Fig. 6). The glycolipid products were unaffected by digestion with peanut phospholipase D or with phospholipase C from either B. cereus or C. perfringens. Phospholipase A 2 from bee venom partially hydrolyzed the products, but only when added at 1000-fold over the level of S. chromofuscus phospholipase D, suggesting that a contaminating phospholipase may have been responsible for the partial liberation of the saccharide moiety.
The GlcUA-labeled product (labeled as described above in the absence of UDP-Glc) was resolved by thin-layer chromatography into approximately five bands, none of which were sensitive to digestion with the type 3-specific depolymerase. The Glc-labeled product was resolved by thin-layer chromatography into approximately eight bands and a component that remained at the origin. The origin material and the slowest migrating band were completely degraded by digestion with  After incubation for 30 min at 35°C, the reaction was terminated by placing on ice, and the membranes were washed three times. One-half of the mixture was solubilized and analyzed on Sephacryl S-500 (q). The remaining membranes were incubated for an additional 20 min as described above, except that the radioisotope was omitted, and 200 M UDP-sugars was added. The products (E) were analyzed as described above. The elution positions of Glc and Nonidet P-40 micelle standards are indicated. the depolymerase (data not shown). In a separate experiment, depolymerase digestion of the phospholipase-treated Glc-labeled product released 22 and 9% of the isotope into components that migrated like a tetrasaccharide and a disaccharide, respectively (Fig. 6).
␤-Glucuronidase digestion of the phospholipase D-treated GlcUA-labeled product released all of the GlcUA, confirming the presence of a ␤-glycosidic linkage and also indicating that growth was occurring by sugar addition at the nonreducing end. Approximately 15-25% of the Glc was released by ␤-glycosidase digestion of the Glc-labeled product, suggesting that a number of internal glucose residues are present in glycolipid products formed in the presence of both UDP-sugars.
Formation of Polymer on a Lipid Primer-Because there were no oligomeric products smaller in size than that corresponding to detergent micelles, the data suggested that polysaccharide synthesis was initiated on a micellar component, which, at low UDP-sugar concentrations, was released from the synthase as a glycolipid before the chain could be extended. As shown in Fig. 7, glycolipid formation occurred much more rapidly than polymer synthesis during the first 30 s in a reaction containing 3 M concentrations of both UDP-sugars. After a pronounced lag, polymer synthesis rapidly increased, whereas following an initial burst, glycolipid formation continued to decrease as the reaction progressed. In a number of similar experiments, it was shown that the lag in polymer synthesis was more marked at lower temperatures and at lower UDPprecursor concentrations and that, conversely, the lag was shorter at higher temperatures and higher UDP-sugar concentrations (data not shown). The presence of UDP-Glc stimulated the incorporation of GlcUA into glycolipid during the first minute of the reaction (compare with UDP-GlcUA alone); however, this stimulation diminished at longer reaction times, presumably as more synthase was engaged in polymer synthesis. The 3 M concentrations of the UDP-sugars present in the reaction are significantly below their K m values and would be insufficient to engage all of the synthase in polymer synthesis.
The formation of polymer on a lipid primer was clearly established in scaled-up reactions containing 3 M concentrations of both UDP-precursors. Only glycolipid was synthesized during the first 5 s of the reaction. It eluted at the position of a micelle on Sephacryl S-300 and, following saponification, appeared to have the size of a small oligosaccharide (Fig. 8A). By FIG. 5. Paper chromatography and gel filtration following saponification of JD424 glycolipid products. The GlcUA-labeled product (q) and the Glc-labeled product synthesized in the presence of UDP-GlcUA (E) were prepared as described in the legend to Fig. 4 and extracted in detergent solution as described under "Experimental Procedures." A, samples (20,000 cpm) were saponified in 0.1 N NaOH for 15 min at 37°C, neutralized, and then analyzed by paper chromatography as described in the legend to Fig. 4. B, samples (20,000 cpm) were analyzed by Sephacryl S-300 chromatography before (OO) and after (---) saponification. Column fractions (0.5 ml) were assayed for radioactivity by scintillation counting.

TABLE I
Partitioning of the Glc-and GlcUA-labeled products between aqueous and organic phases at neutral and acidic pH Membranes containing labeled glycolipids (5000 cpm) were heated at 100°C for 5 min and then partitioned in a mixture consisting of 0.4 ml of chloroform, 0.4 ml of methanol, and 0.4 ml of either 50 mM Hepes (pH 7.5) or 10% acetic acid. The mixture was vortexed vigorously, and the upper aqueous and lower organic phases were removed and placed in scintillation vials. The interface material was suspended in 0.2 ml of methanol and placed in counting vials. The organic liquids were removed under a stream of warm air, and the radioactivity was counted. 20 s, the intact product still consisted of a single micellar fraction; however, following saponification, a small quantity of intermediate-size polysaccharides was present in addition to the small oligosaccharides (Fig. 8B). By 60 s, some of the product had increased sufficiently in size so that it eluted before the glycolipid micellar component (Fig. 8C). This fraction was clearly reduced in size by saponification, although it was still slightly larger than the glycolipid micellar product. By 120 s, the majority of the product was much larger than the glycolipid micellar component, and saponification resulted in only a slight reduction in size of the polymeric fraction (Fig.  8D). At longer reaction times and at higher UDP-sugar concentrations, the reduction in size following saponification became imperceptible. These data strongly indicate that type 3 polysaccharide synthesis is initiated on a lipid primer.
Effect of UDP-sugar Concentration on Polysaccharide Size-The effect of UDP-sugar concentration on the size of polymer formed by synthase-containing E. coli membranes in the presence of 10 mM MgCl 2 is shown in Fig. 9A. Polymer synthesis was carried out at low levels of protein to substrate to minimize changes in the substrate concentration during the course of the reaction. At 1 M concentrations of both UDP-precursors, only the low molecular mass product was observed. As the concentrations of UDP-Glc and UDP-GlcUA were increased from 1 to 20 M, the formation of glycolipid progressively decreased, with a concomitant increase in the formation of polysaccharide of steadily increasing molecular size. In a similar experiment carried out in the presence of 10 mM MnCl 2 , formation of polysaccharide chains of a size comparable to those synthesized in MgCl 2 occurred at 4-fold lower UDP-sugar concentrations (Fig. 9B). The transition from formation of primarily glycolipid to polymer occurred at UDP-sugar concentrations of 2 M in MgCl 2 and 0.5 M in MnCl 2 (Fig. 10). The 4-fold difference corresponds very closely with the relative K m values of the type 3 synthase for the UDP-sugars in Mg 2ϩ and Mn 2ϩ .

DISCUSSION
S. pneumoniae membranes were previously shown to assemble type 3 polysaccharide by the addition of Glc and GlcUA to the nonreducing termini of pre-existing polysaccharide primers (9). When the concentration of either UDP-Glc or UDP-GlcUA drops below a critical concentration, which correlates with the apparent K m values for the UDP-precursors in the biosynthetic reaction, the growing polysaccharide chain is released from the synthase by an abortive translocation process (13). Following release of the polymer from S. pneumoniae membranes, the type 3 synthase is greatly reduced in its ability to form additional polymer, suggesting that some unknown factor is involved in the initiation of polysaccharide synthesis. This investigation has shown that E. coli membranes contain a lipid that is capable of serving as a primer for type 3 synthase. The lipid nature of the primer was indicated by its 1) ready saponification, 2) co-migration with detergent micelles, 3) partitioning between aqueous and organic mixtures, 4) hydrolysis by phospholipase D, and 5) chromatographic mobility. The results are more consistent with a phosphoglycerol lipid than a polyprenol; however, the resistance of the primer to hydrolysis by phospholipase C cannot be explained.
Both type 3 polymer and glycolipid were synthesized by the synthase-containing E. coli strain JD424, and neither was synthesized by the vector control strain JD422. Formation of the glycolipid preceded the onset of polymer synthesis, and glycolipid synthesis diminished as the level of synthase engaged in polymer synthesis increased. A fraction of the glycolipid was cleaved by the type 3 polysaccharide-specific depolymerase from B. circulans, suggesting that the same repeating disaccharide sequence was present in both polymer and glycolipid. Finally, we could find no evidence of any other primer that was smaller than that corresponding to the detergent micelle elution volume when analyzed by gel filtration. All of these findings indicate that the type 3 synthase initiates polysaccharide synthesis on a lipid primer. We have recently identified a primer in S. pneumoniae with properties identical to those of the E. coli primer. 2 The formation of novel glycolipids due to the heterologous expression of a glycosyltransferase is not unexpected in view of several recent demonstrations that the expression of other glycosyltransferases in E. coli can give rise to novel glycolipid and phosphoglycolipid products (23,24). One of the puzzling FIG. 6. Effects of phospholipase D and depolymerase on the Glc-labeled glycolipid. Glc-labeled lipid was synthesized as described in the legend to Fig. 4 and solubilized, and one-third of the mixture was analyzed by paper chromatography (q). The remaining products were digested for 1 h at 30°C with 5 units of phospholipase D from S. chromofuscus in 100 l of 25 mM Tris (pH 8) and 10 mM CaCl 2 . The mixture was heated at 100°C for 4 min; precipitated protein was removed by centrifugation; and one-half of the digest (E) was analyzed by paper chromatography. The remaining products were digested for 1 h at 35°C with 4 g of depolymerase in 50 l of 75 mM Mes (pH 6). The depolymerase digest was analyzed by paper chromatography as above (OE). Disaccharide (Di) and tetrasaccharide (Tetra) standards were obtained from depolymerase digests of S. pneumoniae type 3 polysaccharide. features, as noted in those studies, is the ability of the glucosyltransferases to glycosylate hydrophobic acceptors embedded in the surface layer of the membrane and also hydrophilic accep-tors. Of particular interest is the glucosylation of phosphatidylglycerol in recombinant E. coli expressing the Staphylococcus aureus diacylglycerol glucosyltransferase, although no such phosphoglycolipids have been found in S. aureus (24).
We do not yet know the extent of the modification of the endogenous lipid composition in the membranes of recombinant E. coli expressing the type 3 synthase; however, the composition of carbohydrate products synthesized from UDP-Glc was significantly altered. The membranes from wild-type E. coli synthesized an oligomeric product with properties similar to those of membrane-derived oligosaccharides, which are thought to utilize phosphatidylglycerol by an undefined pathway (25,26). Membranes from E. coli containing the recombinant type 3 synthase synthesized almost none of this oligosaccharide product and also much less of a high molecular mass glucan, suggesting that these activities had been greatly reduced as a consequence of endogenous type 3 synthase activity.  Fig. 7. The washed membranes were solubilized, and the products were analyzed by Sephacryl S-300 chromatography before (q) and after (E) saponification. The elution positions of Nonidet P-40 micelles (arrow 1) and Glc (arrow 2) are indicated.  9 containing MgCl 2 (squares) and MnCl 2 (circles) were analyzed by paper chromatography with a solvent of butanol/acetic acid/water (44: 16:40). The chromatograms were cut into 2-cm strips, and the radioactivity present at the origin (open symbols) and in a component migrating more rapidly than standard Glc (closed symbols) was determined.
Possibly, the inability to express high levels of the synthase in E. coli is a result of these alterations in lipid composition.
In contrast to S. pneumoniae, release of the polysaccharide from type 3 synthase in E. coli was not accompanied by release from the membranes. Presumably, this was due to the retention of a lipid anchor at the reducing end of the polymer. E. coli strains synthesize a variety of cell-surface polysaccharides, including group 2 and 3 capsular K antigens, the enterobacterial common antigen, and colanic acid, which are attached to the outer membrane through a diacylglycerol phosphate moiety (27)(28)(29). Group 2 and 3 K antigens and colanic acid are exported across the inner membrane by an ABC transporter; and, at least in the case of colanic acid, the phosphatidic acid has been postulated to be the common recognition for the transporter (30). Little is known about the attachment of diacylglycerol phosphate to these polysaccharides during their biosynthesis (31).
At low concentrations of UDP-Glc and UDP-GlcUA, the type 3 synthase functioned largely in a non-processive manner, transferring only one to several sugar residues to the lipid primer. As the UDP-sugar concentrations were increased above 0.2 M in the presence of MnCl 2 and above 1 M in the presence of MgCl 2 , the polysaccharide chains were dramatically lengthened. Possibly, the growing nascent chain must attain a minimum length before it will become permanently engaged with the carbohydrate-binding site, thus allowing extensive processive elongation. We previously observed that heparin synthase exhibits a Ͼ100-fold decrease in K m for the carbohydrate substrate as the size of the oligosaccharide increases from a tetrasaccharide to an octasaccharide (32), suggesting that the affinity of the transferase for the growing polysaccharide increases markedly in the early stages of polymerization. A similar occurrence here would, in part, explain the difficulty in trapping oligosaccharides of intermediate size.
Type 3 synthase in S. pneumoniae is a bifunctional glycosyltransferase capable of adding Glc and GlcUA to the nonreducing terminus of the polysaccharide chain (9), which presumably remains bound to a carbohydrate-binding site specific for one to several repeating disaccharide sequences at the nonreducing end of the type 3 chain. The apparent K m values for UDP-Glc and UDP-GlcUA in the reaction catalyzed by the synthase in E. coli membranes were 26 and 20 M in the presence of Mn 2ϩ and 76 and 58 M in the presence of Mg 2ϩ , respectively. The correlation between the differences in these values in the presence of Mn 2ϩ and Mg 2ϩ and the approximate 4-fold difference in UDPsugar concentrations at which a glycolipid-to-polymer transition occurred in reactions in the presence of these two metal ions demonstrates the importance of the concentration of the UDP-precursors in modulating this reaction. The dramatic increase in polysaccharide chain length corresponding to a slight increase in UDP-sugar concentration is indicative of a highly cooperative mechanism, which allows the synthase to engage in non-processive addition of sugars at low UDP-precursor concentrations and processive polymer formation at higher substrate levels.
When S. pneumoniae membranes are incubated in the complete absence of the conjugate UDP-sugar, either UDP-Glc or UDP-GlcUA will actuate the release of the polysaccharide from the synthase with respective apparent K m values of 880 and 0.004 M (13). The presence of UDP-Glc decreases the affinity of the S. pneumoniae synthase for UDP-GlcUA by 3 orders of magnitude, and the presence of the latter increases the binding affinity of the former by 2 orders of magnitude. The release by either UDP-sugar is inhibited by the presence of the conjugate UDP-sugar, and all the results are consistent with the hypothesis that both polymerization and release are catalyzed by interaction of the UDP-sugars with the same set of binding sites on the synthase. These data are consistent with a possible allosteric interaction between two nucleotide-binding sites. However, in view of the recent proposal that the processive glycosyltransferase family 2 synthases contains only a single nucleotide-binding site (12), a variety of mechanisms need to be considered for the type 3 synthase that would allow for both a processive and non-processive reaction mode.
In E. coli, 100% of the polymer was released from the membranes, whereas only 50% release occurred with S. pneumoniae membranes (13). These results suggest that other factors may affect release in S. pneumoniae or that there are differences in the enzyme expressed in the two systems (modification in S. pneumoniae, for example). Reductions in surface capsule levels are important for colonization of the nasopharynx, whereas enhanced levels are necessary in systemic infections (33). In the type 3 strain WU2, ϳ60% of the polysaccharide is released from the cell during laboratory culture in enriched medium (34), but different environmental conditions may alter this ratio. 3 Neither the mechanisms that regulate expression of the type 3 capsule in S. pneumoniae nor the mechanisms that control the percentage of polymer that remains cell-associated are known. Thus, an understanding of the processive/non-processive synthetic mechanism of the type 3 synthase and the modulation of polysaccharide release by UDP-sugar concentrations may provide insights into how S. pneumoniae regulates the amount of surface-localized polysaccharide.