Streptococcus pneumoniae Serotype 11D Has a Bispecific Glycosyltransferase and Expresses Two Different Capsular Polysaccharide Repeating Units*

Background: Streptococcus pneumoniae serotype 11D capsular polysaccharide (CPS) structure is unknown. Results: Serotype 11D PS contains two different repeating units; one has αGlcNAc, and the other contains αGlc. Conclusion: The 11D CPS is due to the bispecific glycosyltransferase WcrL. Based on codon 112, WcrL can transfer αGlc, αGlcNAc, or both. Significance: Minimal genetic changes can make bacteria produce different polysaccharides. Streptococcus pneumoniae (pneumococcus) expresses a capsular polysaccharide (CPS) that protects against host immunity and is synthesized by enzymes in the capsular polysaccharide synthesis (cps) locus. Serogroup 11 has six members (11A to -E) and the CPS structure of all members has been solved, except for serotype 11D. The cps loci of 11A and 11D differ by one codon (N112S) in wcrL, which putatively encodes a glycosyltransferase that adds the fourth sugar of the CPS repeating unit (RU). Gas chromatography and nuclear magnetic resonance analysis revealed that 11A and 11D PSs contain identical CPS RUs that contain αGlc as the fourth sugar. However, ∼25% of 11D CPS RUs contain instead αGlcNAc as the fourth sugar, suggesting that 11D wcrL encodes a bispecific glycosyltransferase. To test the hypothesis that codon 112 of WcrL determines enzyme specificity, and therefore the fourth sugar in the RU, we generated three isogenic pneumococcal strains with 11A cps loci containing wcrL encoding Ser-112 (MBO128) or Ala-112 (MBO130). MBO128 was serologically and biochemically identical to serotype 11D. MBO130 has a unique serologic profile; has as much αGlcNAc as 11F, 11B, and 11C CPS do; and may represent a new serotype. These findings demonstrate how pneumococci alter their CPS structure and their immunologic properties with a minimal genetic change.

The structural diversity of surface-associated carbohydrates, such as lipopolysaccharide or polysaccharide (PS) 2 capsule, plays important roles in the survival of microbes in their host. For instance, the capsular PS (CPS) of the important human pathogen Streptococcus pneumoniae (pneumococcus) helps the bacterium avoid the innate immune system by preventing interaction of host phagocytes with bacterial surface antigens or with complement components deposited on the pneumococcal cell wall (1). However, adaptive immunity in response to vaccination or natural exposure to the pneumococcus can produce anti-capsule antibodies capable of opsonizing encapsulated pneumococci and mediating phagocytosis. Probably to evade the selective pressure of capsule-specific immunity, over 90 biochemically and antigenically distinct pneumococcal capsule serotypes have evolved (2). Thus, serotype diversity is central to the continued survival of S. pneumoniae.
Pneumococcal serotype diversity is dictated by the capsule synthesis (cps) locus, which contains the serotype-specific genes that encode the CPS biosynthesis machinery (reviewed in Ref. 3). For all but two serotypes, CPS is synthesized in a Wzx/ Wzy-dependent pathway that presumably produces only one type of CPS repeating unit (RU). The synthetic cycle begins with the transfer of sugar 1-phosphate to the lipid carrier undecaprenyl-phosphate mediated by an integral membrane glycosyltransferase (4). Cytosolic glycosyltransferases then sequentially add glycosidic residues from NDP-sugar donors to the nascent oligosaccharide. Finally, mature oligosaccharide RUs are exported to the bacterial surface by a Wzx flippase and are polymerized into surface-associated PS by a Wzy polymerase, thereby releasing undecaprenyl-phosphate for recycling. Because of the need to recycle undecaprenyl-phosphate for use in other essential cellular functions (e.g. peptidoglycan synthesis, etc.), cps gene mutations that disrupt the completion of the synthesis cycle are theoretically lethal to pneumococci. Thus, the synthetic cycle may be largely inflexible to changes that affect later steps of the pathway. This stringency limits capsule type diversity to the finite number of cps loci encoding successful biosynthetic machinery.
Serogroup 11 is among the most extensively characterized pneumococcal serogroups. The six antigenically distinct serotypes in serogroup 11 (i.e. serotypes 11A-11F) have highly homologous cps loci (5,6). To investigate the molecular sources of antigenic diversity in this serogroup, we previously examined the structures of serotype 11A, 11B, 11C, 11E, and 11F CPSs (7). Their CPS structures share a similar tetrasaccharide RU but differ in their acetyl and polyalcohol content. These structural modifications can be correlated to their antigenic properties in conventional serotyping assays. For instance, serotypes 11A and 11E do not react with serotyping factor serum 11b, whereas serotypes 11B, 11C, and 11F do (7). Because ␣-N-acetylglucosamine (␣GlcNAc), as the fourth residue of the tetrasaccharide RU, is the only carbohydrate feature shared by serotype 11B, 11C, and 11F CPS but is not present in 11A and 11E CPS, factor serum 11b reactivity is strongly linked to the presence of ␣GlcNAc (7).
The structure of serotype 11D CPS has not been reported. Because of its reactivity with factor serum 11b, we hypothesized that serotype 11D CPS also contains GlcNAc. However, the published serotype 11D cps locus (5) differs from previously published 11A cps sequences (2,8,9) by only one base pair in the cps gene wcrL. Whether change of this codon (named codon 112, for its position in the gene) could mediate antigenic differences between the serotypes is unclear. Here, we demonstrate that serotype 11D CPS surprisingly contains two distinct RUs. Furthermore, we show that alteration of codon 112 in wcrL dictates expression of serotype 11A, serotype 11D, or a novel capsule serotype, 11X3.
Antibodies and Flow Cytometric Serotyping Analysis (FCSA)-Mouse hybridomas expressing monoclonal antibodies (mAbs) labeled Hyp11AM1 and Hyp11AM7 were derived from mice immunized with serotype 11A CPS, as described (11). Rabbit polyclonal factor sera 11b, 11c, and 11g were purchased from SSI. Flow cytometry was used to detect capsule epitope expression on bacteria using polyclonal antisera or mAb, as described (7). Briefly, frozen bacterial stocks were thawed and washed, and bacterial density was adjusted to A 630 ϭ 0.2 before diluting 100-fold. Test strains were then incubated with factor sera at a 1:1000 dilution or hybridoma supernatants containing mAb at a 1:100 dilution for 30 min at 4°C. After washing, bound antibodies were detected with goat anti-rabbit Ig fluorescein isothiocyanate-conjugated antibodies or with rabbit anti-mouse IgM phycoerythrin-Cy7-conjugated antibodies (Southern Biotechnology Associates, Inc., Birmingham, AL). Stained bacteria were analyzed using a FACSCalibur (BD Biosciences), and data analysis was done with FCS Express version 3.0 (De Novo Software, Los Angeles, CA). Bacteria stained with the secondary antibody alone were used as negative controls.
Polysaccharide Purification-Capsular PS was purified from pneumococcal strains MNZ272 (denoted 11A PS), SSISP 11F/2 (denoted 11F PS), and SSISP 11D/1 (denoted 11D PS) and recombinant strains MBO129, MBO128, and MBO130 (see below), as described (7). Strains were grown overnight at 37°C in a chemically defined medium (12) from JRH Biosciences (Lenexa, KS) supplemented with choline chloride (1 g/liter), sodium bicarbonate (2.5 g/liter), and cysteine-HCl (0.73 g/liter). Deoxycholic acid (1 g/liter) was then added to the culture, and the pH was adjusted to 7.0. The culture was incubated at 37°C for 20 min to induce lysis. The pH of the lysate was adjusted to 6.0, and the lysate was centrifuged at 18,000 ϫ g for 30 min to remove cell debris and precipitate the deoxycholic acid. The supernatant was collected and incubated in 30%, 50%, and then 75% ethanol, each step at 4°C for 2 days. Between steps, lysates were centrifuged to remove precipitate. After the final incubation at 75% ethanol, the supernatant was decanted, and CPS precipitates were dissolved in 0.2 M NaCl and then desalted by dialysis against water. The solution containing the CPS was loaded onto a column (45 ml of DEAE-Sepharose, GE Healthcare), and the CPS was eluted with a linear NaCl gradient from 0 to 1 M. Fractions containing CPS detected by a multibead inhibition assay (13) were pooled, desalted, lyophilized, and redissolved in 10 mM Tris-HCl (pH 7.4) buffer containing 100 mM NaCl, to a concentration of ϳ20 mg/ml. The sample was separated by a size exclusion chromatography column (120 ml of Sephacryl S-300 HR, Amersham Biosciences). High molecular weight fractions containing CPS were pooled, desalted, lyophilized, and stored at Ϫ20°C until analyzed.
Monosaccharide Analysis-A 40-g sample of lyophilized CPS was subjected to methanolysis in 3 N methanolic HCl at 80°C for 16 h (19). Following evaporation of the methanolic HCl under vacuum, the residue was washed and dried several times with methanol. Re-N-acetylation of amino sugars in the samples was achieved by dissolving in 200 l of methanol and adding 20 l of pyridine and 20 l of acetic anhydride. After 30 min at room temperature, the samples were dried under vacuum, dissolved in 100 l of methanol, and transferred to glass conical inserts in sample vials. After evaporation to dryness, samples were trimethylsilylated by dissolving in 100 l of Tri-Sil (Pierce) under argon for 20 min at 80°C.
The reaction products were analyzed on a gas chromatograph/mass spectrometer (Varian 4000, Agilent Technologies, Santa Clara, CA) fitted with a 30-m (0.25-mm inner diameter) VF-5ms capillary column (19). Column temperature was maintained at 160°C for 3 min and then increased to 260°C at 3°C/ min and finally held at 260°C for 2 min. The effluent was analyzed by mass spectrometry (MS) using the electron impact ionization mode. Peptidoglycan (PGN) contamination was monitored by identification and quantification of N-acetylmuramic acid (MurNAc).
The one-dimensional 1 H NMR data for purified CPSs (2 mg/ml in D 2 O) were acquired on a Bruker DRZ (600-MHz) spectrometer equipped with a cryoprobe at 25°C. The data were analyzed with ACD/NMR Processor Academic Edition (Advanced Chemistry Development, Inc., Toronto, Canada).
Creation of Isogenic Strains Differing in Codon 112 of WcrL-Three isogenic recombinant strains were created to examine the effect of different amino acids at WcrL codon 112 on CPS structure. MBO129 was generated by transforming TIGR-JS, which contains a Janus cassette in place of a cps locus, with lysate from strain MNZ272 and by streptomycin (300 g/ml) selection as described (10). Two additional strains (MBO128 and MBO130) putatively expressing WcrL variants were derived using site-directed mutagenesis of MBO129 ( Fig. 4 and Table 2). The entire cps loci of MBO128-130 were sequenced (GenBank TM accession no. JX102570 -JX102572) and were found to have wcrL genes encoding Ser-112, Asn-112, and Ala-112 for MBO128, MBO129, and MBO130, respectively.
Serotype 11D CPS Contains GlcNAc-Because serotype 11D reacts with factor serum 11b, we hypothesized serotype 11D capsule contains N-acetylglucosamine (GlcNAc). To examine this possibility, we used gas chromatography and mass spectroscopy (GC/MS) to determine the carbohydrate content of CPS purified from strains SSISP 11D/1, SSISP 11F/2, and MNZ272 (herein referred to as 11D CPS, 11F CPS, and 11A CPS, respectively) and serotype 11A CPS purchased from ATCC. Because all characterized serogroup 11 CPSs contain 2 mol of galactose (Gal) per RU (7), the GC/MS spectra of CPS samples in this study were normalized according to Gal peaks (marked with solid circles in Fig. 2A). 11F CPS contains Gal, glucose (Glc), and GlcNAc at a molar ratio of 2:1:1 per RU (7) and was used as a positive control. As expected, the GC/MS spectrum of 11F CPS showed peaks characteristic of Gal (7.75-10 min), Glc (between 10 and 11 min) and GlcNAc (between 15.5 and 16.5 min) ( Fig. 2A). The GlcNAc peak was smaller than the Glc peaks because GlcNAc is detected less efficiently than Glc and Gal by the employed GC/MS method. 3 In contrast to 11F CPS, the GC/MS spectra of the two preparations of serotype 11A CPS, which is known to have Gal and Glc at a molar ratio of 2:2 (7), showed only Gal and Glc signals and no GlcNAc-associated signal between 15.5 and 16.5 min (Fig. 2, B and C). Reflecting the increased amount of Glc per RU, Glc peaks of both 11A CPS spectra were taller than those of 11F CPS.
The GC/MS spectrum of 11D CPS (Fig. 2D) displayed Gal and Glc peaks but also showed a small but distinct signal (marked with an arrow) that had both the retention time and the molecular weight of GlcNAc. The Glc/Gal integral ratio of 11D CPS was lower than the Glc/Gal ratio observed in 11A CPS, suggesting that 11D CPS RUs contain less Glc than Gal. Furthermore, the GlcNAc-associated signal was too small to represent 1 mol of GlcNAc per RU when compared with the spectra for 11F CPS.
PGN contains GlcNAc and can contaminate CPS preparations. Because PGN has GlcNAc and MurNAc in equimolar amounts, if the GlcNAc signal in 11D CPS originated from PGN, the 11D CPS spectrum would also show signals originating from MurNAc, which elutes between 20 and 21 min. 3 However, the 11D spectrum showed no peaks with elution time between 20 and 21 min. Furthermore, no GlcNAc-or MurNAcassociated signals were appreciable in the GC/MS spectrum of 11A CPS (Fig. 2C), which was purified using the same techniques as 11D CPS (Fig. 2D). Thus, the GlcNAc-associated signal seen in the 11D spectrum is unlikely to be from contaminant PGN. We concluded that 11D CPS is composed of GlcNAc, Gal, and Glc.

Serotype 11D CPS Is a Heteropolymer Composed of Two Different Subunits That Contain either ␣-Linked Glc or GlcNAc-
To further elucidate the structure of the serotype 11D capsule, we produced de-O-acetylated (dO) 11D CPS by mild hydrolysis, examined it with two-dimensional NMR ( 1 H-13 C HSQC), and compared its NMR spectrum with that of dO 11A CPS that was previously published (Fig. 3A) (14). Although the spectra of both dO CPS samples displayed a signal at 4.98/100.7 ppm, assigned to the H1 of ␣Glc, the 11D dO CPS spectra demonstrated an additional weak anomeric signal at 4.96/99.1 ppm, which was identified to be H1 of ␣GlcNAc in our previous studies of 11F CPS (7). This conclusion was further supported by the presence of another weak signal at 4.02/54.47 ppm, which arises from an H2 containing geminal N-acetylation (Fig. 3A) and is distinct from the 4.32/50.0 ppm and 4.10/51.3 ppm signals assigned to ␣-N-acetylgalactosamine (␣GalNAc) H2 and ␤GalNAc H2, respectively, in pneumococcal teichoic acid (cwPS) (15). A diffusion-ordered spectroscopy experiment (16) revealed no evidence of free ␣GlcNAc in the CPS sample solution. Thus, ␣GlcNAc is integral to the glycosidic backbone of 11D PS.
In addition to identifying the H1 and H2 signals of GlcNAc, standard homonuclear and heteronuclear experiments were performed to assign all resonance peaks in dO 11D CPS to Gro and 4 carbohydrate residues: ␣Glc, ␣Gal, ␤Glc, and ␤Gal (Table  1). These studies also assigned a small but distinct peak at 4.55/ 103.28 to variant H1 of ␤Glc.
To investigate how ␣GlcNAc is integrated into the 11D CPS backbone, we obtained the NOESY spectrum of dO 11D CPS and found correlation between two weak signals at 4.98 ppm (anomeric proton of GlcNAc) and the signal at 4.07 ppm. The resonance peak at 4.07/79.12 ppm that is present in the dO 11D PS spectrum but not the dO 11A PS spectrum (Fig. 3A) was previously identified as ␣Gal H4 of 11F PS (7). Also the peak   ␣Glc H1 and ␣Gal H4, respectively, of the ␣Glc(1-4)␣Gal linkage characteristic of 11A CPS (14). These observations led to the model of dO 11D CPS with two different RUs (Fig. 3B); the major RU contains ␣Glc as the fourth glycosyl residue, whereas the minor RU contains ␣GlcNAc. This model is also consistent with the presence of the H1 of variant ␤Glc (that theoretically would be 1-6-linked to the ␣GlcNAc in the adjacent RU) in the dO 11D PS spectrum (Table 1). This structural model of 11D CPS predicts the presence of three distinct phosphate bonds (i.e. ␣Glc-4P, ␣GlcNAc-4P, and Gro-3P). To confirm the model, we examined their presence in native 11D CPS with 1 H-31 P HMBC. The 1 H-31 P HMBC is ideally suited for identifying long range correlations in PSs (17). As expected, the 11D dO CPS spectrum (Fig. 3C) contained a 1 H/ 31 P signal at 4.12/Ϫ0.15 ppm and a diastereotopic signal at 3.98,4.045/Ϫ0.13 ppm, assigned to ␣Glc H4 and Gro H3, respectively (14). The spectrum contains a signal at 4.18 ppm, which is assigned to ␣Glc H6. However, overlays of the 1 H-13 C HSQC spectra of dO11A CPS and 11D CPS reveal no serotype differences in the ␣Gal H6 and ␤Gal H5 proton signals, and these residues of 11D CPS have no geminal or vicinal phosphate substitutions.
To investigate if our findings are not limited to our own bacterial strains, we obtained and de-O-acetylated 11A, 11D, and 11F CPS from a commercial source and examined the anomeric region of their 1 H NMR spectra. As shown in Fig. 3D, the 11A spectrum contains anomeric peaks corresponding to ␣Glc, and the 11F spectrum contains an ␣GlcNAc peak. In contrast, the 11D CPS spectrum shows both ␣Glc and ␣GlcNAc peaks. Integration of 1 H NMR peaks arising from ␣Glc and ␣GlcNAc calculated that ϳ75% of the 11D RUs contained ␣Glc and ϳ25% contained ␣GlcNAc, consistent with observations made in the GC/MS spectra of 11D CPS. We concluded that the 11D CPS glycosidic backbone is a heteropolymer, in contrast to the homopolymer backbones of 11A and 11F PSs.
Polymorphism at Codon 112 (c112) of wcrL Is Sufficient for Serotype Switching-Along with previous research (7), our structural findings support that the identity of the fourth glycosyl residue in the CPS RU correlates with factor serum 11b reactivity. To examine the molecular basis of serum 11b reactivity, we performed a comparative analysis of serogroup 11 wcrL, which putatively encodes the glycosyltransferase that mediates the addition of the fourth glycosyl residue (18). Alignment of deduced WcrL amino acid sequences revealed a correlation between c112 and the identity of the fourth glycosyl residue (Fig. 5A). Namely, wcrL alleles from serotypes that contain ␣Glc as the fourth glycosyl residue (11A and 11E) encode Asn-112, wcrL allele from serotypes that contain ␣GlcNAc (11B, 11C, and 11F) encode Ala-112, and the wcrL gene from serotype 11D, which contains both ␣Glc and ␣GlcNAc, encodes Ser-112.
To confirm that c112 of wcrL mediates the identity of the fourth glycosyl residue, we performed site-directed mutagenesis to create three strains that are isogenic except for c112 of WcrL ( Fig. 4 and Table 2). Strains MBO128, MBO129, and MBO130 contain wcrL alleles that encode Ser-112, Asn-112, and Ala-112, respectively. According to FCSA, Hyp11AM1 mAb comparably bound all three strains, revealing that all isogenic strains express similar levels of capsule (Fig. 5B). Despite this, factor serum 11b antibodies comparably bound strains MBO128 and MBO130 but did not bind strain MBO129 (Fig.  5B). In contrast, another mAb specific for serogroup 11 epitopes, Hyp11AM7, bound MBO128 and MBO129 but not MBO130. We concluded that the three strains are serologically distinct and that the polymorphism at c112 is sufficient to mediate serotype switching among three different serotypes.
N112A Mutation Results in Exclusive Addition of ␣GlcNAc to the Serogroup 11 Glycosidic Backbone-Factor serum 11b reactivity provides only indirect evidence for the presence of ␣GlcNAc in the RU. To obtain direct evidence, we performed GC/MS analysis of CPS purified from strains MBO128, MBO129, and MBO130 (Fig. 6, A-D). c112 of wcrL encodes Ala in both MBO130 and SSI 11F. The GC/MS spectra spanning 7.5-22.5 min of elution time were comparable for 11F CPS and MBO130 CPS (Fig. 6, A and B). Both spectra contained strong, GlcNAc-associated signals eluting between 15.5 and 16.5 min. In contrast, the spectrum for MBO129 CPS contained prominent Gal and Glc signals with no signals between 15.5 and 16.5 min (Fig. 6C). This spectrum was indistinguishable from the 11A CPS spectrum (Fig. 2). The spectrum for MBO128 (Fig.  6D), whose wcrL c112 encodes Ser as in SSISP 11D/1, was similar to that of 11D CPS (Fig. 2), including a small GlcNAc-associated signal (identified with an arrow). Taken together, these findings indicate that MBO130 CPS contains only ␣GlcNAc.

DISCUSSION
We have elucidated the structure of 11D CPS and shown it to be a novel CPS containing two different RUs; about 25% of its RUs have ␣GlcNAc as the fourth residue, whereas about 75% have ␣Glc (Fig. 4C). The presence of ␣GlcNAc in the 11D CPS can readily explain the reactivity of serotype 11D with factor serum 11b, which has been found to react with other serogroup 11 serotypes containing ␣GlcNAc (7). However, factor serum 11b reacts equally well with 11D and 11F, although ␣GlcNAc RUs account for only about 25% of the total RUs of serotype 11D CPS (Fig. 4C). This similar reactivity may occur because small amounts of ␣GlcNAc may significantly alter conformation of 11D CPS. For instance, a minor modification of type III CPS of Group B streptococcus has been shown to change its antigenicity significantly (20). Alternatively, ␣GlcNAc RUs could be more common at the terminus of a CPS chain, and factor serum 11b may preferentially react with terminal ␣GlcNAc. 11D CPS is predicted to be formed by linking the first (reducing end) residue of the growing chain to the fourth (nonreducing end) residue of a single RU by the polymerase (Wzy) (3). If Wzy has a higher affinity for RUs with ␣Glc rather than ␣GlcNAc in the 4th position, the 11D CPS chain may terminate more often with ␣GlcNAc RU. We also show by site-directed mutation that a non-synonymous change of one single nucleotide (c112 of WcrL) of 18 kb in the cps is sufficient to convert serotype 11A to serotype 11D. c112 of wcrL encodes asparagine in 11A but serine for 11D. This finding resolves a major conundrum, that serotypes 11A and 11D have almost identical cps loci despite their serologic differences (8), and should facilitate distinguishing serotypes 11A and 11D molecularly. We further demonstrate the importance of c112 by showing that an N112A mutation creates a viable pneumococcal strain producing a novel CPS; it differs from 11D CPS because it does not have ␣Glc and should differ from 11F because its pendant is expected to be phosphoglycerol, not phosphoribitol. In addition, the N112A mutant has distinct serological properties, and therefore the N112A mutant represents a new serotype. We propose giving the new serotype a provisional name, 11X3, because it has not yet been found in nature. Nevertheless, this site-directed mutation experiment clearly demonstrates that c112 of wcrL is the basis for serotype 11D and the identity of the fourth glycosidic residue.
It is known that S. pneumoniae can undergo serotype conversions when a missense mutation inactivates enzymatic activity of the involved gene (2,21). We show here that a simple missense mutation that preserves the enzymatic activity can still cause serotype conversions. There are additional examples of simple missense mutations causing serotype conversions among other bacteria. Neisseria meningitidis serogroups Y and W-135, respectively, produce capsular PSs containing either Glc or Gal (22). Site-directed mutational analysis showed that amino acid residue 310 of SiaD determines both its substrate specificity and its serogroup association (22). Campylobacter jejuni produces lipo-oligosaccharide that mimics host gangliosides using a sialyltransferase Cst-II (23). Depending on the amino acid at position 51, Cst-II can produce lipo-oligosaccha-   (Table 2). Dotted lines represent sites of construct integration and recombination. Another interesting feature is that 11D WcrL is a bispecific transferase capable of catalyzing the transfer of two different substrates. WcrL may be useful for studying the structural basis for donor bispecificity. For instance, heparan sulfate synthesis in humans involves five EXT genes, which may encode bispecific transferases, including EXTL2, that can transfer either GlcNAc or GalNAc (26 -28). But molecular mechanisms for the bispecificity are not yet clear (27). A well studied example of a bispecific transferase is an animal ␤1,4-galactosyltransferase mutant, which can use both Gal and GalNAc as donors. However, this enzyme is not found in nature (29). Bacterial PSs provide additional examples. Serotype 3 pneumococci produce capsule by alternatively linking two different monosaccharides into a polymer with a synthase (3). But its donor specificity is determined by target substrate (3). Meningococcal strains expressing both serotypes Y and W135 were identified, and molecular analysis suggested that SiaD with serine at 310 may use both uridine diphosphate (UDP)-Glc and UDP-Gal as donor (30). However, the role of residue 310 could not be confirmed by a site-directed mutagenesis study (22). 11D wcrL not only putatively encodes a natural bispecific transferase, but also it can be compared with wcrL alleles of 11A and 11X3. Thus, 11D WcrL would be useful for studying the molecular basis of bispecific transferases.
The three different donor specificities of WcrL of 11A, 11D, and 11X3 may be explained at the atomic level with a steric hindrance model involving the 112th amino acid in WcrL (Fig.  7). The bulky side group of the Asn-112 in 11A WcrL may interact with the N-acetyl (NAc) of a UDP-␣GlcNAc molecule  and inhibit its function as a N-acetylglucosyltransferase but permit its function as a glucosyltransferase. In contrast, a small and hydrophobic Ala-112 in the 11X3 WcrL binding pocket would preferentially accommodate UDP-␣GlcNAc to UDP-␣Glc. The Ser-112 found in 11D WcrL has a side group of intermediate size and may offer a more relaxed donor site specificity than Asn-112 but still restrict the ability to use UDP-␣GlcNAc as a donor substrate. Because both substrates are abundant in pneumococci, the observed ratio may reflect that 11D WcrL has a higher affinity for UDP-Glc than UDP-GlcNAc as its donor substrate. Interestingly, a similar model has been described for Gal/GalNAc specificity of ␤1,4-galactosyltransferases that distinguish vertebrates from invertebrates (31). Briefly, in these galactosyltransferases, one specific amino acid residue (residue 289) largely determines the enzyme specificity, and the amino acid has bulky side groups for Gal-transferases found among vertebrates, whereas the amino acid has small side groups for GalNAc-transferases of invertebrates (31)(32)(33).
Because CPS synthesis involves various transferases and other enzymes that must work together (34), protective CPSs have generally been considered to be not so malleable. However, to evade host immunity, many ways to produce different structures with different serologic properties have evolved in bacteria. The pneumococcus, which is a very successful commensal accidental pathogen, can express diverse capsule types by slightly altering transferases or inactivating acetyltransferases. It is clear that pneumococci may produce more capsule types than we have identified so far. For instance, we show here that serotype 11A can have two variants by altering codon 112 of WcrL. All three serotypes should be able to generate three additional variants by inactivating an acetyltransferase wcjE in its cps, as was shown for the 11A/11E pair (2). Thus, we need to view the bacterial PS as more malleable than we have previously recognized. Also bacteria would explore all of the variations in transferases to survive the host immune system, and therefore they should provide many examples of functional variations in transferases. Thus, studies of bacterial PS, such as pneumococ-cal capsule, should help us investigate structure-function relationships of glycosyltransferases.