Capsular Polysaccharide Expression in Commensal Streptococcus Species: Genetic and Antigenic Similarities to Streptococcus pneumoniae

ABSTRACT Expression of a capsular polysaccharide is considered a hallmark of most invasive species of bacteria, including Streptococcus pneumoniae, in which the capsule is among the principal virulence factors and is the basis for successful vaccines. Consequently, it was previously assumed that capsule production distinguishes S. pneumoniae from closely related commensals of the mitis group streptococci. Based on antigenic and genetic analyses of 187 mitis group streptococci, including 90 recognized serotypes of S. pneumoniae, we demonstrated capsule production by the Wzy/Wzx pathway in 74% of 66 S. mitis strains and in virtually all tested strains of S. oralis (subspecies oralis, dentisani, and tigurinus) and S. infantis. Additional analyses of genomes of S. cristatus, S. parasanguinis, S. australis, S. sanguinis, S. gordonii, S. anginosus, S. intermedius, and S. constellatus revealed complete capsular biosynthesis (cps) loci in all strains tested. Truncated cps loci were detected in three strains of S. pseudopneumoniae, in 26% of S. mitis strains, and in a single S. oralis strain. The level of sequence identities of cps locus genes confirmed that the structural polymorphism of capsular polysaccharides in S. pneumoniae evolved by import of cps fragments from commensal Streptococcus species, resulting in a mosaic of genes of different origins. The demonstrated antigenic identity of at least eight of the numerous capsular polysaccharide structures expressed by commensal streptococci with recognized serotypes of S. pneumoniae raises concerns about potential misidentifications in addition to important questions concerning the consequences for vaccination and host-parasite relationships both for the commensals and for the pathogen.

proteins, an initial sugar transferase, several glycosyltransferases, a polymerase (Wzy), and a flippase (Wzx), and in some cases phosphotransferases, acetyltransferases, and pyruvyltransferases. As a signature of horizontal transfer of cps genes between strains, all pneumococcal cps loci harbor several transposase genes (10). Remarkably, a total of 1,973 genes with predicted function were identified in the cps loci of the first 90 serotypes (10,11). A similar Wzy/Wzx-dependent pathway is widely used in pathogenic bacteria for synthesis of cell wall polysaccharides, including lipopolysaccharides, capsular polysaccharides, extracellular polysaccharides, and glycosylation of certain surface glycoproteins (12).
It has been generally assumed that commensal bacteria that colonize mucosal membranes do not express capsular polysaccharides. However, other mitis group streptococci, including Streptococcus oralis, Streptococcus sanguinis, Streptococcus gordonii, and Streptococcus mitis, may produce several extracellular polysaccharides, including simple glucans that serve as nutritional storage and matrix in biofilms and more structurally complex polysaccharides (13)(14)(15)(16)(17). Some of the latter were shown to function as ligands in pilus-mediated interspecies interactions during oral biofilm formation and have been referred to as coaggregation receptor polysaccharides (CRPs) (17,18). At least some of the CRPs are synthesized by an operon of genes similar to that of pneumococcal cps loci (19)(20)(21)(22)(23). An operon involved in the biosynthesis of capsular or coaggregation receptor polysaccharides was demonstrated in the type strain of S. mitis (24) but is absent in the first reported complete S. mitis genome (25). In our subsequent study, 12 out of 15 S. mitis genomes included a complete cps locus, one of which was identical to that of S. pneumoniae serotype 19C (26). Systematic studies of other mitis group species have not been performed (10,18,27).
The extensive structural diversity of pneumococcus capsular polysaccharides and their genetic basis have been an enigma in view of the otherwise genetically conserved pneumococcal genome. Capsular switching by genetic transformation is a common phenomenon in the population of pneumococci (28)(29)(30), and mutations may also lead to a change of serotype (31,32). However, such genetic events cannot explain the complexity of the gene structure of cps loci. In a recent study, we demonstrated evidence in support of the conclusion that this diversity evolved by pneumococcal import of genes relevant to polysaccharide biosynthesis from a range of commensal streptococci (26). Different polysaccharides have been described for selected strains of S. oralis and S. mitis (16,17), but their structural variation within the species has not been examined systematically (33). Interestingly, serological cross-reactions between pneumococci and other streptococci of uncertain identity were observed in the past (34)(35)(36)(37). To improve the knowledge of cell surface polysaccharides and their genetic basis, we performed comparative immunochemical and genetic analyses of capsular polysaccharides of selected species of streptococci identified according to current taxonomic principles. The aim was to examine and compare the expression of capsular polysaccharides in S. pneumoniae and the commensal species S. mitis, S. oralis (including the subspecies tigurinus and dentisani), and S. infantis as well as other mitis and anginosus group streptococci to understand their functional significance and the potential impact of immunogenic antigens shared by pathogenic and commensal bacteria.

Bacterial suspensions.
After stabilization with formaldehyde, the cells of strains of commensal species used for immunization formed aggregates, in contrast to suspensions of encapsulated pneumococci. The subsequent treatment of the bacterial suspensions with proteinase K dissolved these aggregates, indicating that proteins protruding through the capsular polysaccharide caused the autoaggregation.
Characterization of nonpneumococcal mitis group antisera. All rabbits immunized with 12 strains of S. mitis and S. oralis selected for having a cps locus responded well with high levels of antibodies (titers of 32 or higher in the 3rd bleedings) that induced precipitation in immunodiffusion assays of only one or two antigens in the crude (i.e., untreated) extracts prepared from the homologous strains (exemplified in Fig. 1). Treatment of the crude antigen extracts with either sodium metaperiodate or proteinase K before testing (example shown in Fig. 1B) demonstrated that the outer line was formed by a polysaccharide antigen (protease resistant and sensitive to periodate treatment), whereas the inner line was formed by an unidentified protein antigen (protease sensitive and resistant to periodate). Based on this observation, all bacterial extracts were treated with proteinase K before use as antigens for serotyping by immunodiffusion. In this way, specificity for polysaccharide antigens was ensured and only one precipitation line was formed in each positive reaction (see examples in Fig. 1). None of the antisera reacted with group O antigen, i.e., the common cell wall polysaccharide antigen shared by S. mitis and S. pneumoniae (16), which is also known as C-polysaccharide. Therefore, the sera were considered specific for capsular polysaccharides when protease-treated extracts were used in the tests.
Detection of cps locus regulatory genes by PCR. Initially, 66 S. mitis strains were subjected to PCR analysis for detection of the initial regulatory gene wzg characteristic of the S. pneumoniae cps locus. Four strains of S. pneumoniae served as positive controls. Like the four S. pneumoniae controls, 40 of 66 S. mitis strains gave a strong reaction with both primer sets, and 9 strains gave a strong reaction with one primer set and a weaker or, in one case, negative reaction with the other primer set (selected data are shown in Table S1 in the supplemental material). Among the 66 S. mitis strains, 17 yielded no amplicon with either of the two primer sets. According to this finding combined with the more detailed genetic analyses (see below), 74% of these randomly selected S. mitis strains possessed a putatively functional cps locus. This may be an underestimate of the proportion of S. mitis strains that have a complete cps locus as genome sequencing of one of the PCRnegative strains revealed a complete cps locus.
Serotyping of streptococcal strains. Precipitations appeared in the gels when the prepared antisera were tested against polysaccharides extracted from the homologous streptococcal strain (shaded in Table S2 in the supplemental material). As reactions with the common antigen could be ruled out, this demonstrates that the 12 streptococcal strains selected for immunization all possessed a cell-wall-associated polysaccharide distinct from the common group O antigen (examples shown in Fig. 1 and 2).
Analyses were performed to explore whether the detected polysaccharides were unique or shared by unrelated streptococcal strains. Polysaccharide extracts prepared from 84 nonpneumococcal mitis group streptococci (including the strains used for immunization) and from pneumococcal strains of 90 different serotypes were examined in a checkerboard system: i.e., all extracts were examined in the 12 nonpneumococcal sera and in 14 pneumococcal diagnostic pool sera (pools A to I and P to T). Altogether, more than 4,500 tests were performed by double immunodiffusion (see representative examples in Fig. 1 and 2). Thirty-five of the 84 nonpneumococcal strains exhibited a positive reaction in one or more of the antisera (Table S2). Some of the strains showed identity to or at least cross-reaction with a known pneumococcal serotype. When possible, the serotypes were established by confirmatory double immunodiffusion tests either by comparison with antigens prepared from known pneumococcal serotypes or by the use of pneumococcal group or type sera (examples of reactions are shown in Fig. 2). Based on the results, strains were assigned to recognized S. pneumoniae serotypes or serogroups, to provisional new serogroups designated smI to smIV (Table S1 and Table S2), or to unique structures demonstrated in a single strain only. The characteristics of each of these serogroups are described below.
Antigenic identity to S. pneumoniae serotypes. (i) Serotype 19C. Among the S. mitis strains, only the homologous strain showed a positive reaction with the SK564 antiserum (Table S2). In addition, capsular polysaccharide prepared from the pneumococcal serotype 19C and 19B strains gave distinct positive reactions. Confirmatory tests revealed serological identity between the SK564 polysaccharide and pneumococcal serotype 19C (Table S2).
(iii) Serotype 40. Only S. mitis strain SK611 and pneumococcal types 40 and 7C reacted with the SK611 antiserum (Table S2). A confirmatory test revealed serological identity with the pneumococcal serotype 40 polysaccharide and confirmed the previously demonstrated partial identity to type 7C (Fig. 2D).
(iv) Serotype 2. S. oralis subsp. dentisani strain SK95 did not react with any of the prepared antisera but reacted in pneumococcal diagnostic pool antisera A and T. Additional tests demonstrated that the polysaccharide of S. oralis subsp. dentisani strain SK95 was serologically identical with the pneumococcal serotype 2 polysaccharide (Fig. 2E).
(v) Serotype 36. Immunodiffusion tests of polysaccharides extracted from S. infantis strains SK140 and SK1076 showed a close serological relationship to pneumococcal serotype 36 (reaction for SK140 shown in Fig. 2F). The antigenic identity was not definitively confirmed as antisera were not available for the two S. infantis strains (see results of the genetic analysis below).
(vi) Serotype 21. Two S. mitis strains, SK1123 and SK1124, cross-reacted with each other and with pneumococcal serotype 21 polysaccharide when analyzed with pneumococcal serum pool E (reaction not shown) (Table S2). Capsular polysaccharides distinct from S. pneumoniae serotypes. (i) S. mitis serogroup smI. Immunodiffusion analyses revealed cross-reactions between the three S. mitis strains SK137, SK597, and SK608 (Table S2), but "spurs" at the ends of some of the precipitation lines imply minor structural differences in the three polysaccharides (Fig. 1D). Interpretation of the precipitation lines seen in Fig. 1D suggests that strains SK137 and SK597 express identical polysaccharides, while the polysaccharide of SK608 apparently lacks an epitope relative to the two other strains. Four additional S. mitis strains, SK135, SK138, SK602, and SK677, not used for immunization, reacted with the same three antisera (Table S2). Thus, these seven mitis strains possess similar polysaccharides, although the comparison demonstrated minor differences (Fig. 1D). Based on these results and according to the tradition for pneumococcal serology (i.e., the Kauffmann-Lund Homologous antiserum (as) of SK611 (second bleeding, diluted as indicated) was added to the surrounding wells. The inner sharp line represents an unidentified protein antigen (arrow), while the outer diffuse line represents the capsular polysaccharide. The highest serum dilution that precipitates the polysaccharide was 1:8 (i.e., the titer of this antiserum is 8). (B) Precipitation lines were identified as follows. SK142 antiserum was added to the center well, and lysates made from homologous cells of S. mitis strain SK142 (antigen) were added to the surrounding wells as indicated. The crude lysate (untreated, mutanolysin-lysozyme extract; upper wells) contained two different antigens (arrows A) precipitated by the antiserum. The two antigens were distinct and did not cross-react (arrow D). Acetate buffer (pH 5.0) had no influence on the reactions (control well, upper left). Proteinase K treatment (lower right) digested the band closest to the center well, while the outer band (arrow B) was resistant to the proteinase. In contrast, sodium metaperiodate (lower left) decomposed the outer band, while the inner band was resistant to this treatment (arrow C). Thus, a polysaccharide antigen formed the outer band, while an unidentified protein antigen formed the inner band. (C and D) Example demonstrating cross-reaction or identity between capsular polysaccharide antigens prepared from different S. mitis strains. The center wells contained antiserum, and the surrounding wells contained crude mutanolysin-lysozyme extract (antigen) as indicated. Antigens from all three strains precipitated by antiserum SK608 showed identity. Antigens from SK597 and SK137 precipitated by antiserum SK597 showed identity, while this antiserum revealed nonidentity (arrows) between polysaccharide antigens prepared from the two strains SK597 and SK608. Wells marked with "X" contained buffer only (negative control). Bars, 5 mm. nomenclature [38]), we assigned the seven mentioned S. mitis strains to a "serogroup," i.e., strains displaying extensive serological cross-reactivity due to common antigenic determinants but allowing for minor structural differences. The serogroup I strains did not cross-react with any pneumococcal capsular polysaccharide.
(iii) S. mitis serogroup smIII. In addition to the homologous strain, three S. mitis strains (SK334, SK596, and SK1073) reacted with the antiserum to the SK271 polysaccharide (Table S2). The strains of this group showed one continuous precipitation line when tested against the SK271 antiserum. None of the four strains showed antigenic similarities to pneumococcal capsular polysaccharides.
(iv) S. mitis serogroup smIV. Strain SK637 assigned to serogroup IV was unique among the strains, and no cross-reaction was observed with any pneumococcal serotype (Table S2).
Reactions with anti-S. oralis SK23/ATCC 35037 T serum. In addition to the homologous strain, four strains belonging to three different Streptococcus species (i.e., S. mitis [SK578 and SK646], S. oralis subsp. tigurinus [SK313], and S. infantis [SK959]), reacted in the antiserum raised against S. oralis SK23 with precipitates suggesting identity or close similarity. None of them reacted with pneumococcus typing antisera (Table S2). Since none of the strains had mutually related cps loci (see the results of genetic analysis), the observed reaction may have been caused by antigens unrelated to the capsular polysaccharide.
Serologically unclassified strains. Of a total of 84 nonpneumococcal mitis group streptococci serologically examined in this study, 49 strains reacted neither in any of the pneumococcus typing antisera nor in the 12 antisera raised against selected S. mitis capsular polysaccharides (Table S2). Among 49 S. mitis strains that showed evidence of a cps locus by PCR, 21 (43%) strains did not react in any of the pneumococcus typing antisera or antisera raised against selected S. mitis capsular polysaccharides (Table S2).
Genetic analyses of cps loci. The genomes of 22 S. mitis, 3 S. pseudopneumoniae, 10 S. oralis subsp. oralis, 5 S. oralis subsp. tigurinus, 5 S. oralis subsp. dentisani (previously "S. mitis biovar 2"), and 6 S. infantis strains, as well as the nonclassified strain ATCC 6249 (incorrectly labeled as S. mitis), were examined for the presence and structure of cps loci (Table S2). The search initially focused on the sequence between the genes dexB and aliA/sarA, which flank the cps locus in all S. pneumoniae serotypes and in When tested against the homologous antiserum, the polysaccharide prepared from strain SK611 showed identity to pneumococcal serotype 40 polysaccharide (arrow A) and partial identity to pneumococcal serotype 7C polysaccharide (arrow B). Partial identity between the two pneumococcal serotypes 7C and 40 was also demonstrated (arrow C). (E) Identity between polysaccharide antigens prepared from S. mitis SK95 and S. pneumoniae type 2 (Sp 2) as shown by reactions with pneumococcal type 2 antiserum applied to the center well. (F) S. infantis strain SK140 was identified as type 36 by comparison with polysaccharides prepared from known pneumococcal serotypes by reaction with pneumococcal pool D antiserum. Polysaccharide prepared from SK140 shows identity to pneumococcal type 36 capsular polysaccharide antigen (arrow A) but is dissimilar from the three types 16A, 16F, and 37 (arrows B, D, and E, respectively). As expected, partial identity was observed between the two pneumococcal serotypes 16A and 16F (arrow C), while the two pneumococcal serotypes 36 and 37 were distinct from each other (arrow F). Bars, 5 mm. Skov Sørensen et al. previously examined strains of S. mitis (10,26). Full cps loci spanning from 16,938 to 26,507 bp in length (sequence between end of dexB and start of aliA) and including the four regulatory genes wzg, wzh, wzd, and wze, glycosyltransferases, polymerase, and flippase were demonstrated in 16 of the 22 genomes of S. mitis strains. In the genome sequences of SK255 and SK569, the genes of the cps locus were present on two or three different contigs. The gap between these contigs in each of these strains was closed by Sanger sequencing of PCR amplicons of the gap regions.
The complete cps loci of the 16 S. mitis strains included up to 26 genes, excluding dexB and aliA (Fig. 3, 4, and 5). In the remaining five S. mitis strains (NCTC 10712, SK321, SK642, SK1080, and B6), the locus between dexB and aliA consisted of 5,055 to 7,513 bp encoding one or two oligopeptide ABC transporters, AliC and AliD (periplasmic oligopeptide-binding protein OppA), the UDP-galactopyranose mutase Glf (in all but SK642), and the exopolysaccharide biosynthesis transcriptional activator EpsA/ Wzg (in NCTC 10712 and SK321) (see Fig. S1 in the supplemental material), suggesting degradation of an originally complete cps locus.
An apparent discrepancy was noted for S. mitis SK321 between the positive PCR-based demonstration of the regulatory gene wzg and the absence of a complete cps locus flanked by dexB and aliA. A search for cps genes in other parts of the genome identified a nearly complete cps-like locus in another part of the genome flanked by genes encoding recombination helicase AddA (SMSK321_0547) and a hypothetical protein (SMSK321_0548) upstream and a conserved hypothetical protein (SMSK321_0567) and RNase HII (SMSK321_0568) downstream of the cps locus.
An analysis extended to all other strains revealed a similar cpslike locus at the same genome site in SK137 (SK137_1072 to SK137_1090) in addition to the complete cps locus between dexB and aliA. The cps-like loci in the two strains were organized like classical cps loci, except that in SK321 only a fragment of the wzg gene was present, and in SK137 two of the four regulatory genes, wzg and wzh, were missing. Theoretically, the missing wzg gene in the nonclassical locus of SK321 (cps2) may be functionally complemented by the wzg gene in the truncated cps locus located between dexB and aliA. Comparison of the classical and nonclassical cps loci in SK137 showed no significant homology even between genes that were annotated to carry out similar functions, such as the regulatory genes and the polymerase and flippase genes (see Fig. S2 in the supplemental material). However, the nonclassical cps-like loci in SK137 and SK321 were highly similar, except for a duplication of a glycosyl transferase gene (SMSK321_559 and SMSK321_562) in the SK321 cps2 locus, and both showed partial similarity to the S. pneumoniae serotype 36 cps locus (Fig. S2). None of them included aliB-like genes.
Among the 10 S. oralis subsp. oralis genomes analyzed, nine included a full cps locus ranging in size between 17,845 and 24,479 bp. In the type strain ATCC 35037/SK23 and in strains ATCC 10557/SK10 and SK144, an acetyltransferase gene was found upstream of the cps locus instead of dexB. In the four strains Uo5, SK143, SK610, and C104, which showed identical cps loci (see below), the aliA gene was not present immediately downstream of the cps locus. The remaining S. oralis strain, SK141, had a 5,888-bp sequence between dexB and aliA with a structure similar to that of the mentioned S. mitis strains with incomplete cps loci (Fig. S1).
The five S. oralis subsp. tigurinus strains, including strain J22, which previously was described as a strain of S. sanguis and S. oralis, respectively (13,21), all possessed a full cps locus flanked upstream by dexB and downstream by aliA (Fig. 4B).
In comparison with S. pneumoniae, a number of differences in the overall structure of the cps operons were observed ( Fig. 3 to 5). As described previously, all pneumococcal cps operons include one to several transposase genes and several RUP (repeat units in pneumococci) elements (10,26), which is not the case in any of the other mitis group streptococci examined in this study, except for the two truncated S. pseudopneumoniae cps loci (Fig. S1). In addition, immediately downstream from dexB, all nonpneumococcus strains had one or two oligopeptide ABC transporter genes ("aliB-like"), whereas only fragments were present in some pneumococcal cps loci.
The structures of cps loci of representative strains were further examined and compared mutually and with cps loci of recognized pneumococcal serotypes. The results will be discussed in accordance with the immunochemical results.
Strains of commensal species with cps loci identical to recognized pneumococcal serotypes. Comparisons of complete sequences and gene contents of cps loci of commensal streptococcus strains with those of recognized S. pneumoniae serotypes revealed many examples of identity or close similarity (Fig. 3).
cps sequences were available for 3 (SK575, SK579, and SK616) of the 10 S. mitis strains assigned to S. pneumoniae serotype 45 based on serological identity or similarity. In agreement with results of the antigenic analyses, identical cps locus structures were found, except for a short fragment of a putative acetyltransferase gene (SPC45_0022) and a putative IS1381 transposase (SPC45_ 0023) in the S. pneumoniae serotype 45 strain Eddy 72 and aliC and aliD genes in the S. mitis strains (Fig. 3A). Strain SK575 had a gene encoding UDP-galactopyranose mutase Glf (SK575_26) at the end of the locus just upstream of the aliA gene. Orthologs of glf, but fragmented, were present in the two other S. mitis strains, SK579 and SK616 (Fig. 3A). It is not clear if the fragmentation of the reading frames in the two strains is authentic or due to sequencing errors. The genetic analysis offers no explanation for the signs of an extra epitope identified in SK616 relative to S. pneumoniae serotype 45 and strains SK575 and SK579.
In accordance with the antigenic analysis, the cps locus structure of S. mitis strain SK564 was identical to that of S. pneumoniae serotype 19C as previously described (26) (Fig. 3B). No other nonpneumococcus strain in the collection showed similarity to this structure. The structure of the cps locus of S. mitis SK569 was identical to that of SK564 apart from a truncated UDPgalactopyranose mutase gene (not shown), apparently resulting in loss of the antigenic relationship.
Two strains of S. mitis, SK578 and SK1126, and four strains of S. infantis, ATCC 70779 T , SK140, SK970, and SK1076, showed cps loci closely related to S. pneumoniae serotype 36 (sequence iden-tities of 80 to 88%), but with minor differences that may not influence the expressed polysaccharide ( Fig. 3C and 5). While there was extensive sequence homology between regulatory genes, the glycosyltransferases, etc., the flippase and polymerase genes were very distant from those of serotype 36 and were arranged in opposite order in the two S. mitis strains (Fig. 3C). Rather, these two genes showed 82% identity to the orthologous genes in S. pneumoniae serotype 14 strain Gro Norge. The cps loci of S. infantis strains SK140, SK970 (not shown), and SK1076 were unique among the five strains in including a gene (wcwK/wefC) annotated as coding for a capsular polysaccharide phosphotransferase, which had been described as a stealth protein (39) (Fig. 3C).  The cps locus of S. mitis SK629 was functionally identical to that of the S. pneumoniae serotype 33D cps locus but with distinct evolutionary histories for three genes. The gene encoding the initial sugar transferase wcjH was an ortholog (91% identity) of the gene in S. pneumoniae serotypes 39, 43, 47F, and 35F. Interestingly, a 415-nt fragment of the 791-nt glycosyltransferase gene wciB following the initial sugar transferase in serotype 39 was also present in SK629. Likewise, the acetyltransferase gene in SK629 showed no homology to the gene in S. pneumoniae serotype 33D but was an ortholog of the wcyO gene in serotype 39 (66% identity). Finally, the polymerase gene wzy showed no significant homology to any pneumococcal cps polymerase gene (Fig. 3D).
According to the structure of its cps locus, S. mitis strain SK667 likely belongs to the S. pneumoniae group 18 serovars. As shown in Fig. 3E, it is closely similar to the cps loci of serotypes 18F and 18A. Relative to SK667, the serotype 18F locus includes two genes annotated as acetyltransferase genes (SPC18F_0011 and SPC18F_ 0016). While the former is shared with SK667 (95% identity), the latter is present in SK667 as an intact open reading frame but with nt 175 to 651 missing relative to the intact 1,002-nt gene in serotype 18F. In contrast to all the serogroup 18 cps loci, the cps locus of SK667 includes three open reading frames (SK667_1776, SK667_1775, and SK667_1772) between rmlB and rmlD annotated as representing hypothetical proteins. The first of the three is annotated as encoding a conserved protein detected in several strains identified as S. pneumoniae from Thailand (for example, WP_050292519.1) (40). The second and largest open reading frame encodes a nuclease-related domain protein with homology to proteins in a strain of Streptococcus parasanguinis and in Streptococcus salivarius K12. The function of these proteins, if any, in polysaccharide biosynthesis is unknown. In spite of the structural similarity of the SK667 cps genes to those of pneumococcus serogroup 18, the SK667 polysaccharide extract did not react with pool Q sera, which react with serotypes 18F, 18A, 18B, and 18C. Thus, an antigenic relationship cannot be confirmed with the available antisera.
The three S. oralis subsp. dentisani strains SK95, 7747, and F0407 showed cps loci identical to that of S. pneumoniae serotype 2, apart from a putative cross-wall-targeting SCP domain protein gene present in SK95 and F0407 but absent in strain 7747 and the pneumococcal locus and apart from the two aliB-like genes in the S. oralis subsp. dentisani strains (Fig. 3F). This close genetic simi- Capsule Expression in Commensal Streptococci larity in cps locus structure is in accordance with complete identity of the polysaccharides of SK95 and S. pneumoniae serotype 2 when analyzed with the serotype 2 antiserum. Although the identity was not definitively proven due to the lack of an antiserum against the SK95 polysaccharide, it is likely that the two polysaccharides are identical. The identity of the cps locus of the three geographically independent S. oralis subsp. dentisani strains SK95, F0407, and 7747 out of five analyzed suggest that this is a common serotype in this taxon.
Another S. oralis subsp. dentisani strain, F0392, was unique among the commensal species but showed 93% nucleotide identity and the same gene content and organization as the cps locus of S. pneumoniae serotype 5 (Fig. 3G). As only the genome sequence was available, the identity could not be verified by serological analysis.
The cps locus of S. oralis subsp. oralis SK304 was identical to that of S. pneumoniae serotype 16A apart from the inverted rmlD gene as in all S. oralis cps loci with the rhamnose pathway genes (Fig. 3H). The identical arrangement was previously described for S. oralis subsp. tigurinus strain J22 by Yoshida et al. (21). The cps structure demonstrated in S. oralis SK304 (Fig. 4B) has not been previously detected.
The serologically detected identity between S. mitis strain SK611 and S. pneumoniae serotype 40 could not be validated by genetic analysis due to the lack of an available sequence of SK611.
The type strain of S. oralis subsp. tigurinus, Az_3a, was unique in the collection of commensal streptococci in possessing a cps locus virtually identical to that of S. pneumoniae serotype 33A, except that the terminal genes annotated as coding for acetyltransferases in both loci are highly dissimilar, although they may have identical functions. Rather, the acetyltransferase gene in S. oralis subsp. tigurinus strain Az_3a T is an orthologue (88% identity) of the terminal acetyltransferase gene wzyO (SPC21_0022) in S. pneumoniae serotype 21 strain 546/62. The cps locus of Az_3A T is unique among the commensal strains by lacking intact aliC or aliD genes downstream of dexB and by including remnants of an insertion sequence. Immediately upstream of the aliA gene, the Az_3a T cps locus contains a 401-nt open reading frame with homology to the terminal part of the 1,293-nt IS1167 transposase gene of S. pneumoniae serotype 33A (Fig. 3I).
A summary of the genetic and antigenic identities observed between strains of commensal species and recognized serotypes of S. pneumoniae is shown in Table S1.
Strains of S. mitis with limited homology to S. pneumoniae serotypes. Among the remaining S. mitis strains for which cps locus sequences were available, four groups corresponding to the serologically defined groups smI to smIV were detected (Fig. 4A). Each of these groups of loci showed genes with high sequence identity to recognized S. pneumoniae serotypes, while other genes lacked significant homology. Interestingly, the cps loci of three of the groups, smI, smII, and smIII, included a gene encoding phosphotransferase DUF 3184 family protein previously demonstrated to have "stealth protein activity" (39).
Group smI. In agreement with the immunochemical analysis, three representatives of group I (SK137, SK597, SK608) showed identical cps structures, with the exception that the cps locus of SK597 encoded both versions of the oligopeptide ABC transporters AliC and AliD, while the two other strains possessed the aliD gene only (see below). The serological analysis of SK608 suggested a missing epitope in the polysaccharide relative to that observed for SK137 and SK697. However, the gene content of the cps operons of the three strains does not provide an explanation for this possible difference in epitope structure. The glycosyltransferase genes are highly similar in the three strains. Apart from the wciB gene downstream of the initial transferase, all glycosyltransferase genes lack significant homologies to pneumococcal cps loci. Noticeably, the operon also encodes an LPTXG cell wall anchor protein (Fig. 4A).
Group smII. In agreement with the serological analysis, the cps locus of the type strain of S. mitis NCTC 12261/SK142 was unique in the strain collection (Fig. 4A). Although the overall structure showed no significant homology to any of the pneumococcal cps loci, several genes had orthologs both in cps loci of S. pneumoniae and several other species. In addition to genes encoding the four regulatory proteins and the initial sugar transferase WchA gene (SM12261_0992), the rhamnosyltransferase WchF gene (SM12261_0993), a glycosyltransferase gene (SM12261_0994), the flippase Wzx gene (SM12261_0999), and the four rhamnose pathway genes rmlA to -D are shared with many serotypes of S. pneumoniae. Other genes were rare or absent among pneumococci, such as genes encoding the putative acetyltransferase (SM12261_0995) found only in serotypes 7F, 7A, 22F, and 22A, the two putative glycosyltransferase genes SK12261_0996 and SK12261_0997 found in none of the pneumococcal serotypes, the polymerase Wzy gene (SM12261_ 0998) in serotypes 13, 35F, 35B, 36, and 47F and all members of serogroup 18, the putative glycosyltransferase gene SM12261_1001 in serotypes 20 and 21, and the putative galacto-furanosyltransferase gene (SM12261_1002) in serotypes 10F, 10C, 29, 35B, 36, and 43.
Group smIII. Comparison of the cps loci of strains SK271 and SK1073 among the four S. mitis strains assigned to this serogroup showed an identical structure, in agreement with the observed serological identity. The structure of the cps locus is shown in Fig. 4A. Apart from the four rhamnose pathway genes rmlA to -D, no overall identity or close similarity to any of the S. pneumoniae cps loci was observed. The conserved hypothetical protein encoded by gene SK271_1556 is 63% identical to the WcwD protein encoded by the cps locus of S. pneumoniae serotype 7F and has orthologs in many other Streptococcus species. Genes that were not represented by orthologs in any of the cps loci of recognized serotypes of S. pneumoniae were encoding an acetyltransferase ("LbH_MAT_like," SK271_1557) previously demonstrated in socalled atypical pneumococci (40) and one glycosyltransferase (SK271_1555) with only 33% amino acid sequence identity to glycosyltransferases in S. pneumoniae.
Group smIV. The single S. mitis strain, SK637, assigned to this group had a cps locus spanning 17,161 nt (Fig. 4A). Although the overall structure was distinct, all genes had orthologs in S. pneumoniae cps loci. A span of three glycosyltransferase genes, SK637_1569 to -67 had orthologs in the cps locus of S. pneumoniae serotype 39.
Genetic analysis of cps loci in S. oralis subspecies oralis, dentisani, and tigurinus and in S. infantis. The cps locus of the S. oralis subsp. dentisani strain 7746 T was not identical to any S. pneumoniae cps locus but showed partial similarity in gene content to that of S. pneumoniae serotype 33F. In addition to the four regulatory genes, orthologs of serotype 33F cps genes included those encoding the initial sugar transferase (WchA), the flippase (Wzx), a putative acetyltransferase (WciG), and the UDP-galactopyranose mutase Glf of serotype 33F. The final pseudogene of another putative acetyltransferase (WcjE), present in serotype 33F, is not present in the 7746 T cps operon (Fig. 4B). The glycosyl transferase gene wciN and the LicD protein phosphotransferase gene wcrO were shared with the cps locus of serotype 33C.
Surface polysaccharides encoded by the cps locus of S. oralis have been studied both genetically and structurally by Cisar and coworkers (13,17,20,21), who have been using the term coaggregation receptor polysaccharides (CRPs) according to their demonstrated function and specificity in interspecies coaggregation processes during biofilm formation on tooth surfaces. According to the designations used by Cisar and coworkers (41), the type strain of S. oralis ATCC 35037 and S. oralis strain ATCC 10557 both had a cps locus corresponding to the type 3G coaggregation receptor polysaccharide. S. oralis subsp. tigurinus strain SK313 had a cps locus identical to that of type 2G represented by S. oralis subsp. tigurinus strain J22 (previously named S. sanguis and S. oralis, respectively) (Fig. 4B). The cps locus type 4Gn represented by S. oralis strain C104 was found also in strains SK143, Uo5, and SK610. Type 1Gn found in S. oralis strain 34 was unique among the strains examined in this study. None of these cps locus types showed homology in gene structure to that of recognized S. pneumoniae serotypes. An exception was S. oralis SK144 (structural type 5Gn). The difference between the cps locus of this strain and that of SK143 (4Gn) was the acetyltransferase gene missing in SK144 (Fig. 4B). An ortholog of this acetyltransferase gene is found in S. pneumoniae serotypes 10C and 10F, with which both SK143 and SK144 share a significant part of the cps locus genes (Fig. 4B). One additional type, not previously reported, was demonstrated by our genetic analysis. As described above, S. oralis strain SK304 had a cps locus identical to that of S. pneumoniae serotype 16A (Fig. 3H and 4B).
The cps loci of two S. oralis subsp. tigurinus strains, SK255 and SK1074, each were unique in the collection. Besides by its gene content, the SK255 cps locus included an integrase core protein gene between the final acetyltransferase and the flanking aliA (Fig. 4B). The other unique S. oralis subsp. tigurinus strain, SK1074 (Fig. 4B), showed from 76 to 95% nucleotide sequence identity with the genes encoding the four regulatory proteins, the initial sugar transferase, the putative rhamnosyl transferase WchF, and the four rhamnose pathway proteins RmlA to -D in the cps loci of S. pneumoniae serotypes 2 and 7F. Likewise, the flippase gene wzx and the glycosyltransferase gene immediately upstream shared 66 to 72% nucleotide sequence identity with genes in the S. pneumoniae serotype 47A cps locus. All remaining seven genes in the central part of the cps locus of SK1074 lacked homologs among available sequences from Streptococcaceae, although several were annotated as encoding glycosyltransferases and an acetyltransferase. Surprisingly, no gene showed homology to any available sequence of a polymerase. Finally, like two other strains of S. oralis subsp. tigurinus (Az_3a T and SK255), the cps of SK1074 included a fragment of a transposase gene.
The cps locus of S. infantis SPAR10, flanked by dexB and ftsA, was closely similar to that of S. oralis strains Uo5, C104, SK143, and SK610, except for two additional glycosyl transferase genes (wcaA and epsK) in SPAR10 and an acetyltransferase (wciG) in the S. oralis strains (Fig. 4B and 5). In S. infantis SK1302, the locus showed partial identity in structure and sequence to that of S. pneumoniae serotype 15F. The exceptions are the genes downstream of the flippase gene wzx (Fig. 5). As strain SK1302 was lost, the identity of the capsular structure could not be definitively proven by serological analysis.
Genes unique to cps loci of commensal streptococci. The cps loci of many commensal streptococci include one or two genes encoding periplasmic oligopeptide-binding protein, so-called "AliB-like" or "AmiA" proteins. According to Park et al. (42), the genes may be termed aliC and aliD. A phylogenetic analysis of the genes extracted from all cps loci examined in this study plus reference sequences from the report of Park et al. (42) allowed us to assign names to the individual genes. According to the tree shown in Fig. S3a in the supplemental material, two major clades, each containing one of the two reference sequences aliD and aliC, were observed. Within each clade, separate clusters reflecting the overall phylogeny of the individual species are seen. These clusters, therefore, constitute allelic versions of the same gene (i.e., aliD and aliC, respectively). The gene aliD was present in all complete cps loci of commensal streptococci and in strains of the three S. pneumoniae serotypes 25A, 25F, and 38. As demonstrated by Bentley et al. (10), the cps loci of these three S. pneumoniae serotypes include an almost complete sequence (1,917 and 1,959 nt) of aliD but with three premature stop codons created by two minor sequence deletions. The phylogenetic analysis presented in Fig. S3a shows that the additional aliB-like genes present in strains of S. infantis and S. oralis subspecies oralis, dentisani, and tigurinus are aliC. This gene is absent in S. mitis strains, with the exception of SK597 and SK629 (Fig. 3 to 5). The genes in Fig. 3 to 5 (Fig. S1 and S2) are named according to this phylogenetic analysis. In all S. pneumoniae cps loci other than serotypes 25A, 25F, and 38, a pseudogene consisting of the first 153 to 174 nt of the 1,959 nt in S. mitis aliD was present. As previously demonstrated by Hathaway et al. (43), the truncated cps region of nonencapsulated pneumococci contains one or two aliB-like genes. The clustering of these genes in the tree (Fig. S3a) shows that they are orthologs of aliC and aliD organized as in strains of commensal streptococci. Table S5 in the supplemental material provides a summary of cps-locus encoded proteins in commensal streptococci that do not have significant matches among cps locus-encoded S. pneumoniae proteins (above 50% amino acid sequence identity over Ͼ30% of the length).
Phylogenetic analysis of selected cps locus genes. A phylogenetic analysis of wzy gene sequences from all S. pneumoniae serotypes and commensal streptococci with the available information on the linkage specificity of the encoded polymerase is presented in Fig. S3b. Combined with the significant sequence diversity among S. pneumoniae serotypes, the identities of many pneumococcal genes with wzy genes of several commensal streptococci are in agreement with our observation that the diversity of S. pneumoniae cps loci and capsular serotypes emerged by acquisition of genes from other species (26).
Other proteins encoded by cps locus genes. The cps loci of three S. mitis strains belonging to serogroup I, SK137, SK597, and SK608, included a gene encoding a putative cell-wall-anchored protein with an LPXTG motif at the N terminus (Fig. 4A). The encoded 985-aa, 979-aa, and 999-aa proteins showed 84 to 92% mutual amino acid identity and belong to the G5 superfamily. Bentley et al. (10) identified a putative surface-anchored protein gene at the end of the cps locus of S. pneumoniae serotype 14. However, the three S. mitis proteins showed no homology to the S. pneumoniae protein or to any other protein in the NCBI data-base, and their function in the context of capsular polysaccharide synthesis, if any, remains obscure.
The cps locus of two of the five strains of S. oralis subsp. dentisani, SK95 and F0407, included a gene encoding a protein with a putative choline-binding, cross-wall-targeting lipoprotein signal (SCP domain extracellular protein) between the two periplasmic oligopeptide-binding protein genes aliC and aliD (Fig. 3F). BLASTP screening of the NCBI nonredundant protein database shows that homologs are present in many commensal streptococci and in an unpublished S. pneumoniae strain, 2080767 II, isolated from blood (SAMEA2382970).
Annotation of genes of the S. mitis SK137 cps locus. We previously determined the structure of the SK137 capsular polysaccharide (16) (see Fig. S4 in the supplemental material). This allows us to propose the function of some of the proteins encoded by the genes in the cps operon of this strain and thus annotate most of the genes in the capsular biosynthetic locus (see Table S3 in the supplemental material). The gene downstream of dexB encodes an AliD periplasmic regulatory protein (SMSK13_0341). This gene is similar to aliA immediately downstream of the cps locus. It has been suggested that this group of proteins are involved in substrate recognition (44,45) and may not participate directly in the polysaccharide synthesis. Seven other genes of the SK137 cps locus are common to the Wzy-dependent capsular polysaccharide biosynthesis pathway. They encode enzymes/proteins involved in the process of regulation and cell wall linkage (genes 0342, transcriptional regulator; 0343, tyrosine-protein phosphatase; 0344, chain length determinant protein; and 0345, tyrosine-protein kinase), oligosaccharide chain elongation (0348 plus 0349, Wzy repeat unit polymerase), and transfer of repeat units across the cell wall (0355, Wzx flippase) (11). Based on alignment of related protein sequences and a search among published polysaccharide synthesis pathways, the functions of some of the remaining enzymes encoded by the SK137 cps locus genes are suggested (Table S2). One gene (0356) encodes a mutase (Glf) that catalyzes the transformation of galactopyranose to galatofuranose, a monosaccharide appearing twice in the polysaccharide structure (residues I and VII [ Fig. S4]) of SK137. The suggested specificities of the six transferases are listed in Table S3 and see Fig. S4. The process is started by an initial transferase (Fig. S4, bond 1, gene 0346) that links glucosyl-1-phosphate from UDP-glucose to a lipid carrier (11,46). The second monosaccharide next to the glucose moiety is Galf. The linkage (bond 2a; see Fig. S5 in the supplemental material) is established by an enzyme (0347) similar to the product of the transferase wciB gene (76 to 84% identity) present in various S. pneumoniae serotypes, which like SK137, have the D-Galf-(1¡3)-␤-D-Glcp unit (11,27,47) (Table S2). The third and fourth sugars are Galp moieties attached by ␤(1-6) glycosidic linkages (bonds 2b and 2c, Fig. S5). The transferases (genes 0350 and 0351, core-2/I-branching enzyme [pfam02485]) catalyzing these two bonds are somewhat related (40 to 45% identity) to the products of the wcrM and wcrG genes in S. pneumoniae serotypes 29 and 35B and in serotypes 10A and 39, respectively (11,27,48) (Table S3). The five mentioned pneumococcal polysaccharides contain a glycosylic linkage shared with SK137. However, the repeat units differ as one of the two Gal molecules in the disaccharides from pneumococci is acetylated. The fifth sugar, Glcp-1-P (V, step 2 days), is attached by an ␣-glucose 1-phosphotransferase (0354), similar to the product of the wcrK gene (45% identity) present in S. pneumoniae serotype 7B (11,27). The next bond (2e, Fig. S5) is an ␣(1-6) glycosylic linkage that may be formed by the action of an ␣-glycosyl transferase. This step is, however, uncertain because the putative gene (0352) is not closely related to any other gene encoding an enzyme with a known function. The last sugar, Galf (VII; Fig. S5), is transferred by a galactofuranosyl transferase (0353) related to the product of the wcrH gene (38% identity), which forms the same, although inverted, linkages (i.e., ␤ instead of ␣) in pneumococcal serotype 10F polysaccharide (11), and wefE (38% identity) in S. oralis (49). The polymerase (0348 plus 0349, one gap) connects the repeat units by catalyzing the formation of the D-Glcp-(1¡6)-␤-D-Galf linkages in the final SK137 capsular polysaccharide.
cps loci in other commensal Streptococcus species. Using a BLASTp search of selected representatives of genomes of other species of streptococci commensal to the upper respiratory tract and oral cavity, we identified complete cps loci in all examined strains of Streptococcus anginosus, Streptococcus intermedius, Streptococcus constellatus, Streptococcus cristatus, Streptococcus parasanguinis, Streptococcus australis, and Streptococcus gordonii (see Table S4 in the supplemental material), with the reservation that some were distributed on more than one contig. All cps loci in these species were located in the genomes immediately downstream of a gene encoding an anaerobic ribonucleoside-triphosphate reductaseactivating protein. All contained the four regulatory genes wzg, wzh, wzd, and wze, except for S. australis ATCC 700641, from which wze was missing. In several of the anginosus group streptococci, transposase genes or gene fragments were present as in cps loci of S. pneumoniae. In none of these strains did the cps locus include aliB-like genes.

DISCUSSION
Expression of a capsular polysaccharide is considered a hallmark of most invasive species of bacteria. In invasive strains of S. pneumoniae, the capsule is among the principal virulence factors, as demonstrated by results of in vitro experiments, experimental infections, and the success of the current conjugate vaccines based on selected serotypes of capsular polysaccharides. Consequently, it was previously assumed that capsule production distinguishes S. pneumoniae from closely related commensals of the mitis group streptococci. The findings of this study effectively disprove this assumption. Our genetic analyses demonstrated complete cps loci in 74% of 66 random S. mitis strains, in all but one of 20 S. oralis strains, including the subspecies oralis, tigurinus, and dentisani, and in all six S. infantis strains. Searches of complete genome sequences in GenBank further revealed complete cps loci in all examined strains of the mitis group species Streptococcus cristatus, Streptococcus parasanguinis, Streptococcus australis, and Streptococcus gordonii and in the more distantly related anginosus group species Streptococcus anginosus, Streptococcus intermedius, and Streptococcus constellatus (Table S4). The antigenic analyses confirm that the capsular polysaccharides are expressed. The only exception appears to be S. pseudopneumoniae, which had a significantly truncated cps locus similar to that of occasional S. mitis and S. oralis strains (Fig. S1). The high prevalence of intact cps loci in S. mitis is at odds with the observation recently reported by Yang et al. (23) that none of 12 S. mitis strains examined by them contained a cps/rps operon.
The location of the cps locus in the genomes reflects, to a large degree, the extensive synteny of genomes of mitis group Streptococcus species. Like in S. pneumoniae, the cps locus was flanked by dexB and aliA in all strains of S. mitis, S. oralis subsp. tigurinus, and S. oralis subsp. dentisani. In S. infantis strains, the flanking gene downstream of the cps locus was not aliA but ftsA. Surprisingly, S. oralis subsp. oralis strains showed different patterns. While strains SK304, C104, and 34 were identical to S. pneumoniae, S. mitis, S. oralis subsp. tigurinus, and S. oralis subsp. dentisani, other strains of S. oralis subsp. oralis lacked either the downstream aliA or upstream dexB gene ( Fig. 4B; see Table S1 in the supplemental material). Furthermore, in view of the close genetic relationship of S. pseudopneumoniae to S. pneumoniae and S. mitis, it is surprising that aliA is not found downstream of its truncated cps locus, in contrast to that of nonencapsulated strains of S. pneumoniae and S. mitis (Fig. S1).
The cps loci of pneumococci are among the genome areas most frequently affected by recombination events (50). Horizontal transfer of cps genes between strains is facilitated by the several transposase genes and RUP elements present in the cps loci of all pneumococcal serotypes (10). As part of the many genetic traits that contribute to the genomic stability of S. mitis, contrasting with the genomic plasticity of S. pneumoniae, we previously demonstrated that transposases and RUP elements are lacking in cps loci of S. mitis (26). This is confirmed by this study for S. mitis and furthermore demonstrated for other commensal species, with the exception of cps loci of strains of S. oralis subsp. tigurinus, S. pseudopneumoniae, and species of the more distant anginosus group, which included transposase genes ( Fig. 3 and 4; Fig. S1).
We previously demonstrated that the structural polymorphism of capsular polysaccharides in S. pneumoniae evolved by import of relevant genes from a range of commensal Streptococcus species (26). Therefore, it was not surprising to find strains of commensal streptococci with cps loci identical or nearly identical in gene structure to those of recognized pneumococcus serotypes (Table S1; Fig. 3 to 5). A total of 26% of the detected S. mitis capsules were structurally identical to pneumococcal serotypes. However, the patterns of nucleotide sequence identities over the range of the cps locus clearly demonstrate that the cps gene import by S. pneumoniae does not occur in toto but as blocs of genes resulting in a mosaic of genes of different origins. This is most clearly demonstrated in the range of identities between genes of S. pneumoniae serotype 5 (strain Ambrose) and S. oralis subsp. dentisani strain F0392 (Fig. 3G), between S. pneumoniae serotype 33D (strain CSF 79) and S. mitis SK629 (Fig. 3D), and between S. pneumoniae serotype 16A (strain R105) and S. oralis SK304 (Fig. 3H). It is conceivable that the pneumococcal import of cps locus genes is a still ongoing process that will result in novel capsular polysaccharide structures in S. pneumoniae, some of which may be identical or similar to structures that presently are unique to commensal species. Like those of S. pneumoniae, the capsular polysaccharides of commensal species, in particular S. mitis, showed a significant degree of structural diversity as indicated by the antigenic and genetic evidence. In addition to eight structures identical to recognized pneumococcal serotypes and four structures unique to S. mitis strains, 43% of the examined strains with a putative complete cps locus did not react in any of the available antisera. Apart from the potentially different mosaics of genes that may lead to different structures of polysaccharides, several genes in the cps loci of commensal species annotated as glycosyl transferases lacked homologs in the current S. pneumoniae serotypes (Table S5). Although the exact transferase activities of these enzymes are yet unknown, it is possible that they can expand the structural diversity of capsular polysaccharides if imported by pneumococci.
The demonstrated antigenic identity of at least eight capsular polysaccharides from commensal streptococci with recognized serotypes of S. pneumoniae (serotypes 2, 5, 16A, 18F, 19C, 33A, 33D, 36, and 45) raises important questions concerning the consequences for host-parasite relationships and the potential impact on pneumococcal infections. Does colonization with such commensal strains influence the prevalence of cross-reacting pneumococcal serotypes, induce immunity, or increase infection susceptibility to them? As the necessary comprehensive epidemiological data are not available, the questions can be approached only from a theoretical point of view. One of the capsules detected in the examined collection of commensal streptococci (i.e., S. oralis subsp. dentisani strain F0392) was identical to S. pneumoniae serotype 5, which is among the frequent causes of pneumococcal infection and is included in the current 13-valent conjugate vaccine (51). Therefore, inadvertent elimination of members of the commensal microbiota of the upper respiratory tract by the pneumococcus vaccination is of potential concern. Conversely, there is increasing evidence that commensal bacterial species induce immunological tolerance at the mucosal level but not in the systemic compartment of the immune system, thus facilitating their harmonious coexistence with the host as long as they remain in their natural habitat (52). Therefore, the demonstrated cross-reacting commensal streptococci are neither likely to induce protection nor provide enhanced susceptibility to pneumococcal infection by the mechanisms hypothetically related to the production of an IgA1 protease (53,54).
In pneumococcal infections, the capsular polysaccharide confers a strong antiphagocytic activity on the bacteria, at least partly by reducing the complement deposition on the bacterial surface (55). In vitro studies show that serotypes that are resistant to neutrophil-mediated killing tend to be more heavily encapsulated (56). While the pneumococcal capsules are estimated to be approximately 200 to 400 nm thick (57), information on the size of capsules of commensal Streptococcus species is largely lacking. Yurchak and Austrian (37) reported that the capsular reaction test ("capsular quellung") is not optimal for detection of surface polysaccharides in nonpneumococal streptococci due to the relatively small amounts of capsular polysaccharide, but the identity of the strains is not clear. It is generally assumed that capsular polysaccharides in S. pneumoniae cover other antigens located on the surface of the bacterial cell wall, as we showed for the pneumococcal group O antigen (C-polysaccharide) (57). Our observation that the formalin-treated cells of commensal streptococci aggregated by a proteinase K-sensitive mechanism suggests that the capsular polysaccharide does not cover surface-exposed proteins as in pneumococci. An additional important difference may be that capsule expression in S. pneumoniae is under regulation by a mechanism mediated by a type I restriction modification system (SpnD39III) (58), which is lacking in commensal Streptococcus species (Fig. S5).
In addition to the capsule, commensal streptococci may produce two other extracellular polysaccharides. One is the cell wall polysaccharide analogous to the so-called C-polysaccharide or Lancefield group O antigen in pneumococci (16). Species such as S. oralis subsp. oralis, S. sanguinis, and S. gordonii, in addition, produce an extracellular glucan synthesized by a surfaceassociated glycosyltransferase (59) by a mechanism similar to that of the serotype 3 capsular polysaccharide in pneumococci. In contrast to capsules (60), these extracellular polysaccharides are not covalently linked to the call wall. In oral streptococci, the extracellular glucan is known to play a role as intercellular matrix in the biofilms formed by these bacteria on, for example, tooth surfaces (61). The molecular interplay between these polysaccharides and the capsular polysaccharide, which is actively exported to the surface, is yet unknown.
Pioneering work by Cisar and his coworkers demonstrated that surface polysaccharides synthesized by the Wzy/Wzx pathway in strains of S. oralis, S. gordonii, and S. sanguinis mediate coaggregation between members of the biofilm formed on tooth surfaces (17,23). These polysaccharides have been referred to as coaggregation receptor polysaccharides (CRPs), but are the equivalent of capsular polysaccharides of pneumococci and the commensal species examined in this study. Although the coaggregation mechanism has been mapped only in strains of S. oralis, S. gordonii, and S. sanguinis, it is likely that the polysaccharides demonstrated in this study for many additional species, including S. mitis, have similar functions. In addition, capsulation may protect bacteria from attack by bacteriophages as demonstrated for pneumococci (62). However, it is still unknown to what extent expression of a capsular polysaccharide may contribute to the survival of commensal streptococci that gain access to the bloodstream and thus may play a role in the pathogenesis of subacute bacterial endocarditis.
Interestingly, the cps loci of all examined strains of S. mitis, S. oralis subspecies oralis, tigurinus, and dentisani, and S. infantis include one or two genes encoding an AliB-like protein. These proteins belong to a family of paralogous membrane-bound lipoproteins, AmiA, AliA, and AliB, that participate in oligopeptide transport in S. pneumoniae. The gene encoding AliA is found in all pneumococci and strains of S. mitis and S. oralis subspecies dentisani and tigurinus, as well as in some strains of S. oralis immediately downstream of the cps locus but with no known function in capsular polysaccharide biosynthesis. One or two alleles of the aliB-like genes, aliC and aliD, were found in all complete as well as truncated cps loci of S. mitis, S. oralis subsp. dentisani, S. oralis subsp. tigurinus, and S. infantis immediately downstream of dexB ( Fig. 3 to 5; Fig. S1) but not in S. anginosus, S. intermedius, S. constellatus, S. cristatus, S. parasanguinis, S. australis, S. sanguinis, and S. gordonii. Orthologous genes at the start of the cps locus were previously demonstrated in nonencapsulated strains of S. pneumoniae (40,42,44). Remarkably, the cps locus of none of the encapsulated pneumococci includes functional aliB-like genes although there are traces of their prior existence in the form of small fragments in most serotypes and almost full-length pseudogenes in strains of serotypes 25A, 25F, and 38 (10,43) and in S. pseudopneumoniae (Fig. S1). The pressure that eliminated the genes in encapsulated pneumococci but not in their noninvasive counterparts (i.e., nonencapsulated pneumococci and the most closely related species S. mitis, S. oralis, and S. infantis) is unknown, but might be related to their parallel adaptation to either pathogenic or mutualistic lifestyles (26). Claverys et al. (45) and Hathaway et al. (44) reported that AliA and AliB-like proteins are involved in sensing environmental conditions by their ability to detect and respond to foreign bacterial peptide fragments in their environment. In this context, their proximity to capsular biosynthesis genes and their potential regulatory effects in commensal streptococci are of obvious interest.
In conclusion, capsular polysaccharides synthesized by the Wzy/Wzx pathway are generally expressed by commensal streptococci associated with humans. The level of sequence identities of cps locus genes confirms that the structural polymorphism of capsular polysaccharides in S. pneumoniae evolved by import of cps fragments from commensal Streptococcus species, resulting in a mosaic of genes of different origins. Like in S. pneumoniae, a significant structural diversity of capsular polysaccharides was demonstrated in commensal species, in particular in S. mitis. The demonstrated antigenic identity of many capsular polysaccharides expressed by commensal streptococci and S. pneumoniae raises important questions concerning the consequences for hostparasite relationships both for the commensals and for the pathogen S. pneumoniae.
Genetic analyses. The structures of the capsular polysaccharide biosynthesis locus, cps, of 90 capsular serotypes of S. pneumoniae reported by Bentley et al. (10) and 52 commensal strains extracted from available complete or draft genome sequences were examined and compared in a Sybil database constructed as described previously (64) and established as part of this study (accessible at http://sybil-clovr.igs.umaryland.edu/sybil/ Kilian_CPS_loci). The latter included S. mitis (n ϭ 22), S. pseudopneumoniae (n ϭ 3), S. oralis subsp. oralis (n ϭ 10), S. oralis subsp. tigurinus (n ϭ 5), S. oralis subsp. dentisani (n ϭ 5), S. infantis (n ϭ 6), and Streptococcus sp. strain ATCC 6249 (n ϭ 1). A complete list of these strains and accession numbers for the sequences is shown in Table S1. Nucleotide and protein sequence BLAST analyses were performed at the NCBI database. Illustrations generated in Sybil were manually edited in Adobe Illustrator.
Cluster analysis of selected cps genes were carried out in MEGA version 6.06 (65) using the Minimum Evolution algorithm and bootstrap analysis with 500 replicates. PCR detection of cps locus genes. The presence of the regulatory gene wzg of the cps operon was examined in 66 S. mitis strains by PCR using two sets of primers: wzg-1-for (AATGCRRCITCIAAYTAYTCARTATTC) combined with wzg-1-rev (CCRTARGTRTCAATICCRCTIAYATA) and wzg-2-for (AGTGTIAYRGSICCRACWGRIACIRATAAKGA) combined with wzg-2-rev (TCIATCAWYTTCAARAAIGARGTRAARTTCAAICG), where "I" stands for deoxyinosine. The amplicons of 401 and 575 bp, respectively, generated by the wzg-1 and wzg-2 primer sets were detected by agarose gel electrophoresis. For the PCR, we used PuReTaq Readyto-Go PCR beads (GE Healthcare, United Kingdom) in a 25-l reaction mixture containing 1 ng genomic DNA and 50 pmol of each primer. A thermocycling program of 96°C for 1 min, 30 cycles of 96°C for 30 s, 55°C for 30 s, and 72°C for 1 min followed by an extension at 72°C for 5 min was used.
Antisera. Antisera were raised against the following 12 streptococcus strains: the type strains of S. mitis (SK142 ϭ NCTC 12261) and S. oralis (SK23 ϭ ATCC 35037) and 10 additional S. mitis strains, SK137, SK271, SK564, SK569, SK575, SK597, SK608, SK611, SK637, and SK1124, selected based on positive PCRs for cps locus genes suggesting the potential for surface polysaccharide expression. Briefly, bacterial cells stabilized in 1% formaldehyde were collected by centrifugation (3,000 ϫ g, 30 min), washed in phosphate-buffered saline (PBS), and treated with 10 g proteinase K per ml concentrated cell suspension for 1 h at room temperature (66). Bacterial aggregations were hereby dissolved, or extra proteinase K was added. After the treatment, the enzyme and peptides of digested proteins were removed by washing the cells twice in PBS. White New Zealand female rabbits (2 kg) were immunized by intravenous injections of 1 ml the stabilized proteinase K-treated whole-cell vaccine as described previously (67) under an official permit and in agreement with the national guidelines for animal research. Titers of the prepared antisera were individually examined by double immunodiffusion (Fig. 1). Sera from weekly bleedings of two rabbits immunized with the same antigen and with an agglutination titer equal to 16 or higher were pooled. Diagnostic pneumococcal antisera (pools, group, and type sera) were obtained from Statens Serum Institut, Copenhagen (68).
Preparation of streptococcal polysaccharide extracts for immunoprecipitation. Bacterial cells were harvested from 40-ml overnight broth cultures by centrifugation (3,000 ϫ g, 30 min) and lysed and treated as follows. (i) Nonpneumococcal cells were suspended in 1-ml lysis buffer (0.1 M NaCl, 0.05 M HEPES, 1 mM CaCl 2 , 1 mM MgCl 2 [pH 7.5]) containing 100 U of mutanolysin (Sigma) and 1 mg lysozyme (Sigma) and incubated at 37°C for several hours until more than 95% of the cells were digested as evaluated by Gram staining. (ii) Pneumococcal cells were suspended in 1 ml 0.1% sodium deoxycholate in PBS. This lysis buffer activates the autolysin and induces complete lysis of the pneumococcal cells. Cell debris was removed from the bacterial extracts by centrifugation (10,000 ϫ g, 30 min), and proteins in the supernatants were digested by adding proteinase K (10 g/ml) for 2 h at 50°C. The protease activity was finally blocked by adding 15 l stock solution of the protease inhibitor phenylmethylsulfonyl fluoride (PMSF [with 17.4 mg/ml isopropanol]) per ml of extract to a final concentration of 1.5 mM. A 0.01% solution (1 mg per 10 ml of saline) of purified pneumococcal C-polysaccharide (Statens Serum Institut) was used as a control.
Examination of crude bacterial antigen extracts by immunoprecipitation. Crude extracts of the bacterial strains used for immunization of rabbits were prepared for evaluation of the specificity of the raised sera. Mutanolysin-lysozyme extracts were prepared from live bacterial cells as described above, except that proteinase K was omitted initially. The crude extracts contained mixtures of polysaccharide and protein antigens and were treated as follows. (i) Each extract was kept without further treatment. (ii) Extract was mixed with proteinase K stock solution (100 g/ml, 2 h, 50°C). (iii) Extract was mixed (10:1) with freshly prepared 0.1 M sodium metaperiodate solution in 0.5 M acetate buffer (pH 5.0). After incubation of the sample on ice for 1 h, the reaction was stopped by adding 25 l of polyethylene glycol solution (30% PEG 8000 in 0.3 M NaHCO 3 -NaOH [pH 8.0]). (iv) Extract was mixed (10:1) with 0.5 M acetate buffer (pH 5.0) without periodate (negative control). The four different specimens (i to iv) made from each extract were then compared by double immunodiffusion against the homologous antiserum as shown in Fig. 1. Double immunodiffusion in agarose was used for the examination of reaction between streptococcal polysaccharide extracts (antigens) and the different rabbit antisera. The immunoprecipitation was carried out in 2.0-mm layers of 1% agarose (Litex HAS; Lonza) in HEPES-saline buffer (0.1 M NaCl, 0.05 M HEPES [pH 7.5]) cast on transparent polyester films (Gelbond; Lonza). In most experiments, 15 l undiluted antiserum was applied to a center well (4 mm), and 15-l extract samples were applied to the six surrounding wells (4 mm) placed at a distance, edge to edge, of 5 mm. For details, see Fig. 1 and 2. Plates were kept for 2 days at 5°C and then drained, washed, and stained for 10 min (0.5% Coomassie brilliant blue R-250 in ethanol-water-glacial acetic acid [45/45/10 vol/vol/vol]) and rinsed several times afterward in the same solvent until the background was clear. Finally the rinsed plates were dried, and photos were taken.