Discovery of Streptococcus pneumoniae Serotype 6 Variants with Glycosyltransferases Synthesizing Two Differing Repeating Units*

Background: Two atypical serogroup 6 strains have been discovered. Results: They express capsular polysaccharide with two different repeating units and share A150T mutation in WciNα. Conclusion: This mutation shifts WciNα from a galactosyltransferase to a bi-specific glycosyltransferase, creating two new hybrid serotypes: 6F and 6G. Significance: Pneumococci can change capsule serotypes by point mutations and may alter their interactions with the immunity of the host. Streptococcus pneumoniae is a persistent, opportunistic commensal of the human nasopharynx and is the leading cause of community-acquired pneumonia. It expresses an anti-phagocytic capsular polysaccharide (PS). Genetic variation of the capsular PS synthesis (cps) locus is the molecular basis for structural and antigenic heterogeneity of capsule types (serotypes). Serogroup 6 has four known members (6A–6D) with distinct serologic properties, homologous cps loci, and structurally similar PSs. cps of serotypes 6A/6B have wciNα, encoding α-1,3-galactosyltransferase, whereas serotypes 6C/6D have wciNβ encoding α-1,3-glucosyltransferase. Two atypical serogroup 6 isolates (named 6X11 and 6X12) have been discovered recently in Germany. Flow cytometric studies using monoclonal antibodies show that 6X11 has serologic properties of 6B/6D, whereas 6X12 has 6A/6C. NMR studies of their capsular PSs revealed that 6X11 and 6X12 have two different repeating units with a distribution of ∼40:60 6B:6D and 75:25 6A:6C PS, respectively. Sequencing of the wciNα gene in 6X12 and 6X11 revealed single and double nucleotide substitutions, respectively, resulting in the amino acid changes A150T and D38N. Substitution of alanine with threonine at position 150 in a 6A strain was associated with hybrid serologic and chemical profiles like 6X12. The hybrid serotypes represented by 6X12 and 6X11 strains are now named serotypes 6F and 6G. Single amino acid changes in cps genes encoding glycosyltransferases can alter substrate specificities, permit biosynthesis of heterogeneous capsule repeating units, and result in new hybrid capsule types that may differ in their interaction with the immune system of the host.

Streptococcus pneumoniae (pneumococcus) is an opportunistic commensal of the human nasopharynx and is the leading cause of community-acquired pneumonia, otitis media, bacterial sepsis, and meningitis (1). Pneumococcal pathogenesis is primarily associated with the polysaccharide (PS) 2 capsule, which shields pneumococci from phagocytosis and greatly enhances virulence (2,3). More than 90 different capsule types (serotypes) have been determined, each of which has a unique capsular PS synthesis (cps) locus, chemical structure, and serologic property (4 -6).
Some serotypes are serologically cross-reactive and can be grouped into one serogroup. For instance, serogroup 6 traditionally has had two cross-reactive serotypes, 6A and 6B, that have identical PSs except for the linkage between rhamnose and ribitol caused by genetic differences in the rhamnosyltransferase gene wciP ( Fig. 1) (7). Recently, two new members were added to serogroup 6 when serotypes 6C and 6D were discovered among the isolates that were respectively typed as 6A and 6B by Quellung reaction (8 -10). 6C and 6D PSs differ from 6A and 6B PSs, respectively, by having glucose (GlcЈ) in place of galactose (Gal) (Fig. 1). Reflecting this structural difference, the capsule gene (cps) loci of serotypes 6A and 6B have wciN␣ encoding a galactosyltransferase, whereas serotypes 6C and 6D have wciN␤, which is distinct from wciN␣ and encodes a glucosyltransferase (11).
Although there are many pneumococcal serotypes, only a few serotypes are primarily associated with invasive disease, and current pneumococcal vaccines target those serotypes (12). For instance, serogroup 6 includes commonly pathogenic serotypes and is targeted in all pneumococcal vaccines (13). Widespread pneumococcal vaccination exerts a selection pressure against the serotypes in the vaccine and promotes the emergence of novel, nonvaccine serotypes (14,15). Thus, serotype surveys of pneumococcal isolates are performed in many countries. In Germany, ϳ20,000 invasive pneumococcal disease isolates were collected from 1992 to 2012, and ϳ7% belong to serogroup 6 (16). However, two isolates (6X11 and 6X12) could not be assigned to one of the four known serotypes. Our studies are aimed at characterizing the serologic, genetic, and chemical basis of 6X11 and 6X12 capsules.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Other Reagents-Reference strains expressing serotypes 6A, 6B, 6C, or 6D or no capsule were TIGR6A, TIGR6B, TIGR6C, TIGR6D, and TIGRJS, respectively, which were produced in the genomic background of TIGR4 by previously described genetic manipulations (8,17,18). MNZ21 is a clinical isolate expressing serotype 6D (10). The 6X11 and 6X12 strains are clinical adult isolates from Germany named PS6657 and PS16864, respectively. The strains were rederived from a single colony on blood agar plates to avoid a potential mixture of two different serotypes. All pneumococci were grown at 37°C in 5% CO 2 in Todd Hewitt broth (BD Biosciences, San Jose, CA) containing 0.5% yeast extract (THY), harvested at A 600 of 0.4 -0.6, and aliquoted in THY with 15% glycerol. The aliquots were kept at Ϫ80°C until needed.
Flow Cytometry-The flow cytometric serotyping assay was performed as described (10,21). Frozen bacteria aliquots were thawed, washed, normalized to A 630 of 0.02 in FACS buffer (PBS containing 4% fetal bovine serum (Thermo Scientific Hyclone, Logan, UT)), added to a V-bottomed ELISA plate (Nunc, Roskilde, Denmark), and incubated with a culture supernatant of a hybridoma at 1:40 final dilution. After incubation for 30 min at 4°C without shaking, the plates were washed with FACS buffer and incubated again for 30 min at 4°C with phycoerythrin-conjugated goat anti-mouse IgM antibody diluted 1:1,000 or phycoerythrin-Cy7-conjugated goat antimouse IgG antibody, diluted 1:3,000. After washing, bacteria were resuspended in FACS buffer and examined with a flow cytometer (FACSCalibur; Becton Dickinson, Mountain View, CA). The data were analyzed with FCS Express V 3.0.
Capsular Polysaccharide Purification-Capsular PSs were purified from pneumococcal strains MNZ21, 6X11, and 6X12. In addition, four mutants (MBO172, MBO182, MBO184, and MBO177) were produced as described below. The strains were grown in 2 liters of chemically defined medium (JRH Biosciences, Lenexa, KS) supplemented with choline chloride (1 g/ liter), sodium bicarbonate (2.5 g/liter), and cysteine-HCl (0.73 g/liter) and lysed with 0.1% deoxycholic acid. After removing cell debris by centrifugation, PS was precipitated stepwise in 30, 50, and 70% ethanol and was recovered by dissolution in 200 mM NaCl. After desalting by dialysis against MilliQ water, the PS was loaded onto a 60-ml column of DEAE-Sepharose (GE Healthcare) and eluted with a linear gradient of NaCl from 0 to 1 M. The resulting fractions were tested for PS using a multibead inhibition assay (22). The PS-containing fractions were pooled, desalted by dialysis, lyophilized, dissolved in 1-2 ml of water, and loaded onto a gel filtration column (120 ml of Sephacryl S-300 HR; GE Healthcare). Using a buffer containing 10 mM Tris-HCl and 100 mM NaCl, the PS was eluted from the column, and all fractions were tested for PS with the multiplexed inhibition type immunoassay (20). The fractions containing the first PS peak were pooled, lyophilized, and used for NMR studies.
NMR Spectroscopy-Briefly, 1-2 mg of lyophilized PS was dissolved in 0.6 ml of D 2 O for one-or two-dimensional NMR studies. The one-dimensional 1 H NMR data of purified PSs were collected on a Bruker DRX (600 MHz) spectrometer equipped with a cryoprobe at 25°C located in University of Alabama at Birmingham Comprehensive Cancer Center. The data were analyzed with the ACD/NMR Processor Academic Edition (Advanced Chemistry Development, Inc., Toronto, Canada). Assignments of 1 H and 13 C signals for 6X11, 6X12, 6A, 6B, and 6D PS were achieved by two-dimensional nuclear Overhauser spectroscopy ( 1 H-1 H NOESY), correlation spectroscopy ( 1 H-1 H COSY), total correlation spectroscopy ( 1 H-1 H TOCSY), 1 H-13 C heteronuclear multiple quantum coherence (HMQC), and 1 H-13 C heteronuclear multiple bond correlation (HMBC) data collected at 45°C on a Bruker Avance II (700 MHz 1 H) spectrometer equipped with a cryoprobe. The data were processed with NMRPIPE (23) and analyzed with NMR-VIEW (24). 1 H and 13 C chemical shifts for 6A, 6B, and 6C capsular PSs have been reported elsewhere (25,26).
Site-directed Mutagenesis-Mutant pneumococcal strains were made by transforming TIGR6A with appropriate genetic constructs as described (11). An unencapsulated variant, named MBO163, was made by allelic exchange of the wchA-wciN-wciO gene region with a Janus cassette (Cassette 1) (see Fig. 6) as described (17,27,28). Additional genetic constructs with desired mutations at wciN␣ were generated by overlap extension PCR. Cassette 2 codes for WciN␣ with A150T mutation, cassette 3 codes for D38N and A150T mutations, cassette 4 D38N, and cassette 5 codes for A150S mutation. Replacement of the Janus cassette in MBO163 with cassettes 2, 3, 4, and 5 resulted in four recombinant strains, which were named MBO172, MBO182, MBO184, and MBO177, respectively (see Fig. 6). All mutations have been confirmed by DNA sequencing (Heflin Sequencing Core, University of Alabama at Birmingham). DNA sequences of TIGR6A, MBO163, MBO172, Sandwich ELISA-Flat-bottom 96-well ELISA plates (Corning Costar Corp., Acton, MA) were coated for 5 h with an IgG mAb Hyp6CG6 (2 g/ml) in PBS. All incubations were done at 37°C in a humid incubator, unless otherwise stated. After washing with PBS containing 0.05% Tween 20, the plates were blocked with 5% skim milk (BD Difco, Sparks, MD) in PBS for 2 h. After washing, previously diluted purified PSs were added to the wells at concentrations ranging from 3000 to 0.03 ng/ml and incubated for 1 h. The 6A and 6C PSs were from Statens Serum Institute, whereas the 6B PS was from ATCC. The 6X12 and 6D PSs were purified from strains 6X12 and TIGR6D, respectively. After washing, an IgM mAb Hyp6DM5 (specific for 6C and 6D) or Hyp6AM3 (specific for 6A) at 1:50 dilution was added to the wells. After incubation for 1 h, the plates were washed, and alkaline phosphatase-conjugated goat anti-mouse IgM antibodies (Sigma) were added to the wells at a 1:10,000 dilution and incubated for 1 h. The amount of enzyme immobilized to wells was determined by adding paranitrophenyl phosphate substrate (Sigma) (1 mg/ml) in diethanolamine buffer and incubated at room temperature for 2 h. The optical density at 405 nm was read with a microplate reader (BioTek Instruments Inc, Winooski, VT).
Structural Characterization of 6X11 and 6X12 PS Capsules-To explain the unusual serologic behavior of the two German strains, we characterized the molecular structure of their capsular PS using NMR spectroscopy. The one-and two-dimensional NMR data revealed that the 6X12 PS contains two distinct forms of capsule repeating units (RUs) and that the 1 H and 13 C chemical shifts of these two forms are essentially identical to those of 6A and 6C PSs. Because the anomeric signals in the 1 H NMR spectra correspond to residues in PS RUs, we first compared the anomeric signals of 6X12 PS with those of 6A and 6C PSs. Three signals have been observed in the anomeric region of 6A (5.60, 5.10, and 5.02 ppm), corresponding to the anomeric protons of ␣Gal, ␣Glc, and Rha (26). Similarly, the one-dimensional 1 H spectrum of 6C has three signals (5.57, 5.10, and 5.02 ppm), which respectively correspond to the anomeric protons of ␣GlcЈ, ␣Glc, and Rha. In contrast, 6X12 PS had the three anomeric signals of 6A PS, as well as a fourth signal at 5.57 ppm, which corresponds to ␣GlcЈ of 6C PS (Fig.  3A). Thus, 6X12 PS appears to contain RUs of both 6A and 6C PSs, even though it is purified from a single bacterial colony.
To unambiguously determine the structure of 6X12 PS, we collected a set of two-dimensional NMR data (NOESY, COSY, TOCSY, 1 H-13 C HMQC, and 1 H-13 C HMBC) for 6X12 and 6A PSs. Signal assignments for 6A and 6C PS are described elsewhere and were used as a guide to assign 1 H and 13 C signals of 6X12 (26). As shown in Fig. 4A Table 1. Based on the signal assignment of the second monosaccharide in 6A and 6C RUs, ␣Gal, and ␣GlcЈ, respectively, we determined that 6X12 PS is a mixture of two RUs: ϳ75% 6A and 25% 6C. We conclude that 6X12 strain produces a novel "hybrid" capsular PS (Fig. 4C).
A similar strategy has been utilized to characterize the molecular structure and sugar composition of 6X11 PS. Complete assignment of 1 H and 13 C signals for 6B and 6D PS has been achieved using homonuclear and heteronuclear two-dimensional NMR data as described above. We describe in detail the assignment strategy of the 6D PS. Three 1 H signals of anomeric proton have been observed at 5.56, 5.14, and 5.10 ppm. The anomeric proton at 5.56 ppm, which is connected to a carbon signal at 99.35 ppm in the two-dimensional HMQC spectrum, is correlated to proton signals at 3.98, 3.84, 3.53, and 4.05 ppm in the two-dimensional TOCSY spectrum. Strong NOE cross-peaks have been observed between anomeric signal at 5.56 ppm and two signals at 3.84 and 3.98 ppm. Medium or weak NOEs have also been observed between the anomeric sig-nal at 5.56 ppm and signals at 3.53 and 3.81 ppm. 1 46 ppm is connected to a proton signal at 3.94 ppm. These correlations and others identified from the COSY data allowed for the unambiguous assignment of the proton signals at 5.56, 3.98, 3.84, 3.53, 4.05, and 3.81 to the H1, H2, H3, H4, H5, and H6 protons of the ␣GlcЈ moiety, respectively.
The signal of anomeric proton at 5.11 ppm, correlated to a carbon signal at 97.07 ppm in the HMQC spectrum, has crosspeaks to proton signals at 3.67, 3.94, 3.70, and 3.97 ppm in the TOCSY spectrum. The signal at 3.94 ppm, which is connected to a carbon signal at 81.46 ppm in the HMQC, is assigned to the H3 proton of the ␣Glc because it is linked to the anomeric proton of ␣GlcЈ in the HMBC spectrum. The pattern of crosspeak correlations in the COSY, NOESY, and HMBC is very similar to that observed for the ␣GlcЈ moiety, indicating that the proton signals at 3.67, 3.94, 3.7, 3.97, and 3.78 ppm belong to the H2, H3, H4, H5, and H6, respectively, of the ␣Glc moiety.
The third anomeric proton at 5.14 ppm belonging to the Rha moiety has cross-peaks to proton signals at 4.26, 3.87, 3.58, 3.79, and 1.3 ppm in the TOCSY spectrum. NOE cross-peaks have also been observed between the signal at 5.14 ppm and signals at 4.26, 3.87, 3.58, 3.79, and 1.3 ppm. Numerous NOE, COSY, and TOCSY cross-peak correlations between these five proton signals have also been identified. In the HMBC, the anomeric proton at 5.14 ppm has long range cross-peaks to carbon signals at 68.63, 71.16, 76.85, and 78.66 ppm. These observations allowed for the assignment of the Rha moiety as shown in Table 1. The carbon signal at 78.7 is connected to a proton signal at 4.10 ppm in the HMQC spectrum, which in turn has cross-peaks to proton signals at 3.63, 3.79, 3.83, 3.78, 4.09, and 4.23 ppm in the TOCSY spectrum. In the HMQC spectrum, proton signals at 3.63 and 3.79 ppm have cross-peaks to the same carbon signal at 64.46 ppm, and signals at 4.09 and 4.23 ppm have cross-peaks to the same carbon signal at 66.30 ppm. These observations and others from the NOESY, COSY, and HMBC data led to the assignment of signals of the ribitol moiety as shown in Table 1.
As expected, serotype 6B PS had three 1 H resonances at 5.59, 5.16, and 5.14 ppm, assigned to ␣Gal, Rha, and ␣Glc, respectively (25). Serotype 6D PS also had three anomeric 1 H signals at 5.56, 5.14, and 5.11 ppm, which respectively correspond to ␣GlcЈ, Rha, and ␣Glc (Fig. 3B). Interestingly, 6X11 PS had four anomeric signals: three identical to those of 6D PS and a small fourth signal at 5.59 ppm, assigned to ␣Gal of 6B PS (Fig. 3B). Thus, 6X11 PS purified from a single bacterial colony appears to contain RUs composed of 6D and 6B PS (Fig. 3B). As shown in Fig. 4B, the 1 H- 13 Table 1). These signals are assigned to ␣GlcЈ in 6D as described above. Collectively, the NMR data demonstrate that 6X11 PS contains two different RUs (ϳ40% 6B and 60% 6D), confirming that 6X11 is a new "hybrid" capsule type. The structural model of 6X11 is summarized in Fig. 4C.
cps Loci of 6X11 and 6X12 Are Nearly Identical to Those of 6A-To determine the genetic basis for the two anomalous German strains, we determined their capsule gene loci (cps) sequences from dexB to aliA ( Fig. 5; GenBank TM accession numbers KC832410 and KC832411). Compared with a serotype 6A cps sequence (GenBank TM accession number CR931638), 6X11 and 6X12 sequences were 99.9 and 98.9% identical, respectively. The sequence differences were limited to ϳ10 -100 individual nucleotides that were randomly distributed (Fig. 5). wciP allelism can distinguish serotypes 6A/6C from 6B/6D; the former group has wciP␣, whereas the latter has wciP␤ (7). When the wciP alleles of the two German strains were examined, the 6X12 sequence was identical to a typical 6A wciP␣ (GenBank TM accession number CR931638), but the 6X11 sequence was identical to the wciP␤ of a typical serotype 6B isolate (Gen-Bank TM accession number JF911503) (30). Thus, this finding explains the association of 6X12 with serotypes 6A/6C and 6X11 with 6B/6D.
Next we examined the sequences of wciN, which genetically distinguishes serotypes 6A/6B from serotypes 6C/6D. Serotypes 6A/6B have WciN␣, which adds UDP-Gal, whereas serotypes 6C/6D have WciN␤ that is completely different from WciN␣ and adds UDP-Glc (9 -11). When the two German strains were examined, they had wciN␣ but not wciN␤. Careful comparison of the 6X12 wciN␣ sequence to a canonical 6A wciN␣ sequence (CR931638) revealed a single nucleotide substitution (G to A) at position 488 of the wciN␣ coding strand, changing alanine 150 to threonine 150 (Fig. 5). 6X11 wciN␣ had two mutations comprised of a substitution (G to A) at position 113 changing aspartic acid 38 to asparagine 38, as well as the aforementioned G448A point mutation resulting in A150T of its corresponding gene product. The two mutations were highly unusual because they were absent among all wciN␣ sequences of strains expressing serotype 6A or 6B in the literature. One amino acid change can convert a mono-specific glycosyltransferase to a bi-specific transferase (31). Thus, the two mutations A150T and D38N may broaden the specificity of WciN␣ from UDP-Gal only to UDP-Gal and UDP-Glc and be responsible for the observed serologic and biochemical changes.
To determine the effect of mutations on the molecular structures of PS, we obtained 1 H NMR spectra of all four isogenic mutants. The 1 H NMR spectra of 6X12 and MBO172 were identical in all regions, providing evidence that A150T mutation alone was responsible for the altered capsule type seen for 6X12. In contrast, the spectrum of MBO184, which has a D38N mutation, was similar to that of 6A PS, whereas the spectrum of MBO182 with both D38N and A150T mutations showed a more prominent GlcЈ peak compared with Gal peak. Thus, D38N mutation alone does not alter WciN␣ specificity but  enhances its preference for UDP-Glc introduced by A150T mutation. When the 1 H NMR spectra of MBO177 was examined, it had a bigger GlcЈ peak than Gal peak, suggesting that it produces more 6C RUs units than 6A RUs (Fig. 7B). Thus, the A150S mutation is more effective than A150T mutation in altering WciN␣ substrate specificity. Taken together, our data indicate that A150 is the key residue responsible for WciN␣ specificity observed in the two German strains.
6X12 PS Is a Hybrid of 6A and 6C PS RUs-To show that 6X12 PS is a hybrid composed of 6A and 6C RUs mixed together in a single polymer (and not a mixture of two different polymers: one composed of only 6A RUs and the other composed of only 6C RUs), we used two sandwich ELISAs. One ELISA is specific for 6C RU, and the other is specific for a polymer containing both 6A and 6C RUs (Fig. 8). As expected, the ELISA for 6C RU detected 6C PS, a mixture of 6A and 6C PS,  6A, 6B, and 6X12). For immunological comparison, all the strains were stained with a 6A (Hyp6AG4) or 6C (Hyp6DM5)-specific mAbs, the amounts of mAb bound to bacteria were determined with a flow cytometer, and the amounts (mean fluorescence intensity, MFI) were plotted in both axes. The amount of Hyp6AG4 bound to MBO182 was artificially reduced by 20% to provide better visual separation between MBO182 and MBO177. For chemical comparison, the capsular PSs were purified from the mutants and were analyzed by NMR to obtain chemical shifts in the anomeric region. and 6X12 PS (Fig. 8A). A mixture of 6A and 6C PSs at a 3:1 ratio was used to mimic the binary capsule model based on the predicted percentage of each RU from the NMR data in Fig. 3. 6D PS gave a weak signal at a very high concentration (Fig. 8A) because of the cross-reactivity of the detection antibody (Hyp6DM5) with 6D PS. Interestingly, only 6X12 PS produced a positive signal in the ELISA designed to detect PS chains containing both 6A and 6C RUs (Fig. 8B). Thus, our findings demonstrate that 6X12 PS is a hybrid PS with two different RUs mixed together in a single polymer.

DISCUSSION
Although the two German strains clearly belong to serogroup 6, our studies demonstrate that they have serologic properties, biochemical features, and genetic markers that are stable, unique, and distinct from the other four members in serogroup 6. It is well known that some established serotypes differ by one to three nucleotides in their cps. For instance, serotypes 9V and 9A differ by one single nucleotide (4), serotypes 15B/15C and 18B/18C differ by two nucleotides (4,33), and serotypes 6A/6B differ by three nucleotides (7,34). Thus, although the German strains genetically differ from serotypes 6A and 6B by one or two nucleotides, the genetic changes of the German strains alter the enzyme function, and therefore we propose the two German strains be recognized as representing two new serotypes, 6F and 6G. Serotype 6F is represented by 6X12 and has serologic properties of both serotypes 6A and 6C. Serotype 6G has properties of 6B and 6D and is represented by the strain 6X11.
The German strains provide important insights into the molecular basis of WciN␣ specificity. WciN␣ belongs to Pfam01501, which includes many glycosyltransferases used by viruses, bacteria, and eukaryotes and has the DXD motif well known for binding divalent cations (35,36). A mutation of residue 38 of WciN␣ is present in only one of the two German strains, and its neighboring residues are not conserved among Pfam01501 members. Indeed, our data revealed that mutation of residue 38 does not alter ligand specificity, although it may augment the impact of mutation in residue 150. In contrast, the mutation at residue 150 is present in both German strains and is located within a highly conserved region. For instance, residues 148 -152 of WciN␣ are conserved in human glycogenin-1, except for residue 150 (FNAGV versus FNSGV) (32). Because residue 150 is variable, residues 149 -151 are herein named as the NXG motif to simplify its description. Crystallographic studies found the NXG motif to form a part of the ligand-binding pocket: NXG of Neisseria meningitidis LgtC surrounds the "C1" of the donor ligand (37), and the NXG of human glycogenin-1 interacts with the hydroxyl groups of "C2" and "C3" of Glc (32). Molecular modeling of WciN␣ using PHYRE2 (38) also predicted the NXG motif to form a ligand-binding pocket. Pfam01501 members with NAG are often galactosyltransferases like WciN␣, whereas members with NSG, like glycogenin-1, are often glucosyltransferases. Our studies clearly show that substitution of alanine in NAG to threonine or serine alters the ligand specificity of WciN␣, making it capable of transferring both galactose as well as glucose. Taken together, we conclude that NXG is critical to WciN␣ ligand specificity and probably to the specificity of all Pfam01501 members.
More interesting is that both A150T and A150S mutations turn WciN␣ into bi-specific transferases, and the mutants produce novel PSs with two different RUs. Such a novel PS raises interesting points, such as the fact that one RU may favor termination of PS chains or require a higher substrate level than the other. Because of the interesting biochemical properties, characterization of bi-specific transferases is of high importance. To date, only two eukaryotic transferases have been reported, of which one is natural (39) and the other is artificially created (31). In contrast, studies of bacterial transferases suggest several examples of bi-specific transferases. Pneumococci (34) and meningococci (40) with mixed capsule types have been described. Although they may produce PSs with mixed RUs, their genetic and chemical bases have been incompletely characterized. Better described is LOS, which is produced by Campylobacter jejuni with a Cst-II variant (T51N) (41). The mutation was shown to be responsible for producing two RUs with different glycosidic linkages. Using WciN␣ variants, we have provided a clear example that can transfer two different ligands. mAbs that are specific for the different RUs are available, genetic manipulations are easily performed with pneumococci, and a simple in vitro substrate for WciN␣ was described recently (42). WciN␣ is therefore useful for studying the molecular basis of bi-specific transferases. The significance of PS with multiple RUs in host-pathogen interactions is unclear at the moment. As the most exposed structure for bacteria, capsular PS is critical to host-pathogen interaction. Also a minor structural change can dramatically alter its interaction with the adaptive or innate immunity of the host. For instance, pneumococcal serotypes 19A and 19F, which differ by one linkage in their RUs, are starkly different in their cross-reactivity with vaccine-induced antibodies (43) and also binding of factor H (44). Also, C. jejuni strains with the Cst-II variant elicit a unique autoimmune disease (41). Thus, investigation of serotypes 6F and 6G will provide new insights in understanding survival advantages by comparing their complement binding and reaction with antibodies with other serogroup 6 members.
Because capsular PS is important in host-pathogen interaction, capsule evolution has been extensively studied. Most studies found that pneumococci regularly switch capsule types by acquiring new DNA from other bacteria through genetic recombination (45,46). However, we show two single base mutations that are synergistic in capsule type alterations. Perhaps, there may be a third mutation that may complete serotype change from 6A to 6C. Thus, if point mutations give survival benefits to pneumococci, the presence of such mutational stepping stones would open an evolutionary pathway for pneumococci to alter their capsule structure without a source of foreign DNA. This serotype shift is probably useful in invading deeper tissues from the nasopharynx or in rapid responses to vaccination. Interestingly, evidence for such a capsule type shift by mutation has been recently described for Streptococcus iniae infecting vaccinated fish in fish farms (47).
Because protective immunity is associated with the surface PS, pathogens are commonly divided into discrete serotypes based on their surface PS structure. With increased knowledge of the genetic basis for PS synthesis, genetic markers are often used alone to determine serotypes (48 -50). However, genetic differences between two serotypes may be only one or two nucleotides. Furthermore, the presence of PSs with mixed RUs blurs serologic boundaries and the definitions of serotypes. Thus, one should recognize that one typing method may be inadequate to identify new serotypes. Multiple analytical approaches will be required for accurate identification of serotypes useful for correct prediction of protective immunity.