OUP user menu

Versatility of choline metabolism and choline-binding proteins in Streptococcus pneumoniae and commensal streptococci

Regine Hakenbeck, Abderrahim Madhour, Dalia Denapaite, Reinhold Brückner
DOI: http://dx.doi.org/10.1111/j.1574-6976.2009.00172.x 572-586 First published online: 1 May 2009


The pneumococcal choline-containing teichoic acids are targeted by choline-binding proteins (CBPs), major surface components implicated in the interaction with host cells and bacterial cell physiology. CBPs also occur in closely related commensal species, Streptococcus oralis and Streptococcus mitis, and many strains of these species contain choline in their cell wall. Physiologically relevant CBPs including cell wall lytic enzymes are highly conserved between Streptococcus pneumoniae and S. mitis. In contrast, the virulence-associated CBPs, CbpA, PspA and PcpA, are S. pneumoniae specific and are thus relevant for the characteristic properties of this species.

  • Streptococcus pneumoniae
  • Streptococcus oralis
  • Streptococcus mitis
  • choline-binding proteins
  • teichoic acid


The phosphorylcholine moieties of the pneumococcal teichoic acids (TAs), described 40 years ago as a component of the wall teichoic acid (WTA) (Brundish & Baddiley, 1968), are important structures for host cell interactions of the pathogen. They are targeted by the C-reactive protein (CRP), an acute-phase serum protein that shows a rapid increase in response to infection or injury (Volanakisk & Kaplan, 1971). The reaction of CRP with phosphorylcholine can lead to complement activation and protection against pneumococcal infection in vivo (Volanakisk & Kaplan, 1971). The choline residues also bind to certain myeloma proteins (Leon & Young, 1971), and the PAF receptor (Cundell et al., 1995). Furthermore, the choline-containing TA represent the receptor for bacteriophages (for a review, see (López & García, 2004). The pneumococcal lipoteichoic acid (LTA) is involved in initiating imflammatory cascades by activating the alternative complement pathway (Hummell et al., 1985), and stimulates the host immune response via the Toll-like receptor 2 signal pathway (Draing et al., 2008; Schröder et al., 2008). Interestingly, other respiratory pathogens able to colonize mucosal surfaces are also decorated with phosphorylcholine on their cell surface, including Neisseria, Haemophilus influenzae and Pseudomonas aeruginosa (Kolberg et al., 1997; Weiser et al., 1998a, b).

The choline-containing TAs represent the ligand for choline-binding proteins (CBPs), which have been described as typical for Streptococcus pneumoniae. They contain a choline-binding domain consisting of characteristic repeat motifs. These cell wall-associated proteins play an important role in cell wall physiology, and for the interaction with host cells as well. Some CBPs have also been identified in closely related, commensal streptococci such as Streptococcus mitis and Streptococcus oralis. This review summarizes our knowledge of TAs and CBPs in S. pneumoniae and closely related commensal species.

TA and LTA structure

Choline has long been known to be an essential component in the growth media of the pneumococcus (Rane & Subbarow, 1940; Badger et al., 1944). Tomasz (1967) discovered that it was a component of the cell wall, and it was identified as part of the C-substance or TA soon after Brundish & Baddiley (1968). TAs of S. pneumoniae are unusual complex molecules and both LTA and WTA share the same repeating unit (Fischer et al., 1993). The pentamer repeating unit consists of ribitol phosphate, two molecules of N-acetylgalactosamine (GalNAc), the rare sugar 2-acetamido-4-amino-2,4,6-trideoxy-d-galactose (AATGal) and glucose (Jennings et al., 1980; Behr et al., 1992). Phosphocholine residues are attached to GalNAc. Mass spectrum data indicated that about half of the typical LTA molecules of strain R6, an unencapsulated derivative of the type 2 D39 strain, have repeating units with one phosphocholine group, and the other half with two (Behr et al., 1992; Seo et al., 2008). In contrast, only 15% of the repeats of strain Rx1 (which is also derived from D39) carry two phosphocholine groups (Fischer, 2000). In general, the number of phosphocholine residues appears to be variable (Karlsson et al., 1999). In addition, glucose was replaced by galactose in the TA of a serotype 5 strain (Vialle et al., 2005). Variations in the TA content are associated with the phase variation in S. pneumoniae, where transparent variants contain approximately twofold more TA compared with opaque variants, which have a higher content of capsular polysaccharide (Weiser et al., 1994).

Recent analyses (Draing et al., 2008) revealed that the ribitol carries substitutions at the hydroxyl groups that were not detected in previous structural determinations (Fischer et al., 1993). In the TIGR4 strain, 40% of the hydroxyl groups carried an N-acetyl-d-galactosamine, 50% of d-alanine and 10% remained unsubstituted (Draing et al., 2008). In R6, however, d-alanine was not detected. This strain contains a mutation in dltA that is part of the dltABCD operon encoding proteins responsible for d-alanylation (Kovács et al., 2006). The DltA gene encodes a predicted d-alanine-d-alanyl carrier protein ligase. The parental serotype 2 D39 strain, and another D39 derivative RX, contain a functional dltA. Comparison between the R6 mutant and dltA wild-type strains suggested that d-alanylation of the TAs protects the pathogen against the action of cationic antimicrobial peptides, and that it is relevant for the induction of a variety of cytokines (Kovács et al., 2006; Draing et al., 2008). In addition to the differences concerning the presence of d-Ala, the LTA from the R6 strain differs from R36A, its progenitor, and several other genetically distinct strains, by MS analysis, but the molecular basis is not yet known (Seo et al., 2008).

Moreover, a revised model with respect to the repeating unit and for the lipid anchor has been proposed (Seo et al., 2008). Here, the terminus of the repeating unit and thus of the intact LTA molecule is [GalNAc(α1→3)GalNAc(β→1)], which is consistent with Forssman antigeniticity of the pneumococcal LTA (Fischer et al., 1993). Thus, the polysaccharide repeating unit begins with GalNac and ends with AATGal, with ribitol being in the middle of the repeat (Seo et al., 2008). Its lipid anchor is Glc(α1→3) diacyl-glycerol, which is a common lipid in pneumococcal membranes. Synthesis of the lipid anchor requires the activity of the SP1076 α-monoglucosyl diacylglycerol synthase (Berg et al., 2001).

Choline metabolism in S. pneumoniae

There are several operons implicated in choline metabolism in S. pneumoniae and two of them are clustered at the lic locus. The lic locus consists of the lic1 and lic2 operons, which are transcribed in opposite directions. The five genes of lic1 include tarI (spr1148), tarJ (spr1149), licA, licB and licC. While LicABC are required for choline uptake and activation, TarIJ are needed for the formation of activated ribitol. Choline is transported by the LicB transporter (Kharat & Tomasz, 2006), phosphorylated by the LicA kinase (Whiting & Gillespie, 1996) and subsequently converted to cytidine 5′-phosphate (CDP)-choline by the cytidylyl transferase LicC (Campbell & Kent, 2001; Rock et al., 2001; Kwak et al., 2002). TarJ is an NADPH-dependent alcohol dehydrogenase producing ribitol 5-phosphate from ribulose 5-phosphate. TarI is a cytidylyl transferase synthesizing CDP-ribitol from ribitol 5-phosphate and cytidine 5′-triphosphate. Disruption of tarIJ as well as licABC was not possible, suggesting that the gene products are essential for cell growth (Kharat & Tomasz, 2006; Baur et al., 2009).

The lic2 operon consists of three genes: tacF (spr1150), licD1 and licD2. The LicD1 and LicD2 proteins are most likely responsible for attachment of phosphorylcholine residues to the TA precursors. A licD2 mutant contained only about 50% of choline in their cell walls and was deficient in a number of surface proteins (Zhang et al., 1999). A licD1 inactivation could not be attained in this study. The recently proposed flippase TacF is required for the transport of the TA subunits across the cytoplasmic membrane (Damjanovic et al., 2007).

The lic1 operon is transcribed from two promoters, one of which is subject to CiaR-mediated activation, the response regulator of the two-component system CiaRH (Halfmann et al., 2007). The lic2 operon is transcribed from one promoter located in front of tacF, which is independent of CiaR. Using a β-galactosidase reporter gene placed behind licC or licD2, the last genes of the lic1 and lic2 operons, it was determined that the licD2 gene is strongly upregulated during choline starvation, whereas licC is not (Desai et al., 2003). The lic2 promoter, however, located in front of tacF and also responsible for licD2 transcription, did not respond to choline starvation (unpublished data); strain differences may account for this apparent discrepancy. The actual function of two putative choline transport genes (SP1860/SP1861 and spr1677/spr1678, respectively) remains to be clarified (Hoskins et al., 2001; Tettelin et al., 2001).

Choline-independent S. pneumoniae

Choline in the growth medium can be replaced by different aminoalcohols, for example ethanolamine. However, cells grown under these conditions show numerous abnormalities, such as growth in long chains, lack of autolysis and resistance to bacteriophage Dp-1, and have a transformation minus phenotype (Tomasz et al., 1968; López et al., 1982; Ware et al., 2005). It has been suggested that in the wild type in the absence of choline or another aminoalcohol, aminoalcohol-free TA linked to the polyprenol phosphate carrier accumulates, thus rendering the carrier unavailable to peptidoglycan synthesis (Fischer et al., 2000). Chain formation in the absence of choline or in the presence of ethanolamine has been linked to CBPs implicated in cell separation such as LytB, which might no longer be associated with the choline-free cell wall, and thus cannot function properly. This effect can also be achieved by growing cells in the presence of 2% choline (Briese & Hakenbeck, 1983, 1985; Giudicelli & Tomasz, 1984). This effect is probably due to the release or inhibition of LytB, a CBP with endo-β-N-acetylglucosaminiase activity involved in cell separation (De Las Rivas et al., 2002).

Similar phenotypes have been observed using choline dendrimers shown to be potent inhibitors of LytA, LytB, LytC and Pce, and to bind to CBPs with high affinity and specificity (Hernández-Rocamora et al., 2009). In contrast, the addition to the growth medium of choline analogues (esters of bicyclic amines) induced a different phenotype such as growth inhibition, morphological alteration and reduced viability (Maestro et al., 2007). These compounds also strongly interact with the choline-binding module of LytA and inhibit the in vitro activity of LytA, LytC and Pce. It is possible that the effects on bacterial growth of the choline analogues are based on the interaction with targets other than CBPs, thereby interfering for example with the biosynthesis of the choline-containing TAs. Nonetheless, such compounds represent an interesting basis for the development of new antimicrobial agents.

Four choline-independent mutants in S. pneumoniae have been reported. In the absence of choline they grow in long chains and do not autolyze. The mutant JY2190 has been derived from Rx1 after several rounds of growth in a choline-free medium, but with decreasing concentrations of ethanolamine (Yother et al., 1998). A similar mutant was constructed in the R6 background named R6Chi (Damjanovic et al., 2007). A third mutant, R6Cho, was selected by transformation of R6 with DNA from an S. oralis strain known to contain choline in its cell wall, but that can grow in a choline-free medium although much slower (Severin et al., 1997). Also with DNA from S. mitis SK598 DNA, a strain that grew in long chains with and without choline, choline-independent S. pneumoniae transformants were obtained (González et al., 2008).

The choline-independent phenotype of R6Chi is due to a single tacF point mutation G700→T resulting in the amino acid change V234→F (Damjanovic et al., 2007). The authors proposed that the mutant TacF can transport TA subunits with or without choline, whereas the parental TacF is strictly specific for choline-containing subunits, in agreement with the hypothesis formulated before by Fischer (2000). Other mutations in tacF have been identified based on comparative analysis of tacF genes in S. pneumoniae, S. mitis and Streptococcus pseudopneumoniae and subsequent transformation using PCR fragments covering different regions of tacF (González et al., 2008). García and colleagues showed that a minimum of two mutations in TacF were required to confer improved fitness to choline-independent S. pneumoniae mutants when grown in a medium lacking aminoalcohols. In support of this hypothesis, the JY2190 strain was identified as a double mutant in tacF (González et al., 2008).

In the mutant R6Cho, a different mutational pathway became apparent (Kharat et al., 2008). In this transformant, a deletion occurred flanking another lic homologue annotated as licD1 in R6 (Hoskins et al., 2001) and as licD3 in the TIGR4 strain (Tettelin et al., 2001). This region was replaced by a large 20-kb fragment of S. oralis ATCC35037 DNA carrying both tacF and licD genes. In this background deletion of tacFR6 was possible. It has been suggested that the S. oralis tacF product is able to export TAs with or without choline phosphoryl groups, because the parent S. oralis strain grows in a choline-containing medium as well as in a choline-free medium (Horne & Tomasz, 1993). The function of other genes in this locus remains to be clarified.

Choline in commensal streptococci

There are many reports describing choline in the cell wall of commensal streptococci (Kilpper-Bälz et al., 1985; Ronda et al., 1991; Gillespie et al., 1993; Horne & Tomasz, 1993; Kolberg et al., 1997; Gmur et al., 1999; Bergström et al., 2000; García et al., 2004). The presence of choline in the cell wall of commensal bacteria was also suggested by a PCR screen for the lic1 and lic2 loci in a collection of commensal streptococci (unpublished data), confirming the presence of the genes in many (but not necessarily all) S. mitis strains. Recently, two monoclonal antibodies were used to detect epitopes characteristic of the backbone and the phosphocholine residues of the TA (also named C-polysaccharide) in a large sample of S. pneumoniae, S. pseudopneumoniae, S. mitis, S. oralis and Streptococcus infantis (Kilian et al., 2008). Most of the S. mitis strains (52 out of 57) reacted with both antibodies, whereas two strains lacked the phosphocholine residue, but contained ethanolamine instead as described for one strain previously (Bergström et al., 2003), and three showed evidence of phosphocholine, but lacked the characteristic epitope of the TA backbone. In contrast, all 28 strains of the Oralis cluster contained phosphocholine, but lacked the backbone characteristic of the pneumococcal TA (Kilian et al., 2008).

When 2% choline is added to the growth medium of some S. mitis and S. oralis strains, the cells grow in long chains similar to that observed with S. pneumoniae. We have tested the phenotype during growth with 2% choline and in the absence of choline, using a variety of S. oralis and S. mitis isolates. Remarkably, chain formation appears to be independent of the requirement for choline, and the observed phenotypes were variable within the S. mitis and within the S. oralis strains. For example, some S. mitis strains that grow in chains in 2% choline medium are strictly choline dependent, whereas others are not and also show chain formation in the absence of choline; some S. oralis strains grow with and without choline equally well, but only some show chain formation in the absence of choline (see Fig. 1 for examples). Growth in the absence of choline results in unusual septum formation in S. oralis ATCC35037 (Horne & Tomasz, 1993), suggesting that the choline-containing cell wall is involved in this process probably indirectly and perhaps via some CBP. These results confirm the genetic variability within the S. mitis and S. oralis group as revealed by comprehensive multilocus sequence typing (Chi et al., 2007) and other comparative sequence analysis (Kilian et al., 2008). Nevertheless, it is clear that choline-containing cell wall polymers as well as CBPs are present at least in many strains of these species.

Figure 1

Growth of Streptococcus oralis strains with and without choline. (a–d) Streptococcus oralis RSA60; (a′–d′) S. oralis RSA38. Both strains have been isolated in South Africa (Arthur et al., 1995). (a, a′) Closed circles: growth in a chemically defined medium containing 5 μg mL−1 choline (CDM) (van de Rijn & Kessler, 1980); open circles: CDM without choline. Bacteria were examined by phase-contrast microscopy during the exponential growth phase. (b, b′) CDM, 5 μg mL−1 choline; (c, c′) CDM plus 2% choline; (d, d′) CDM minus choline.

CBPs in S. pneumoniae

The family of CBPs represents modular proteins that are located on the pneumococcal cell surface. Whereas many CBPs appear to be conserved in S. pneumoniae strains, the number of CBPs depends on the strain analyzed and some proteins are highly variable. Their choline-binding domain (CBD) can be arranged in mostly 20-amino acid (aa)-long choline-binding repeating units with characteristic conserved amino acids (Pfam accession PF01473), with the exception of the three lytic enzymes LytA, LytB and LytC. An alignment of CBPs that contain highly similar repeats and also a striking similarity in their C-terminal tail is shown in Fig. 2. The CBD is mainly located at the C-terminus, with the two exceptions LytB and LytC, and is sometimes preceded by a proline-rich linker sequence as in PspA and PspC (García et al., 1994; Yother & White, 1994). Four to five of the repeat units suffice to interact with the phosphorylcholine moiety of TA (Yother & White, 1994).

Figure 2

Alignment of the C-terminus including choline-binding repeats of Streptococcus pneumoniae TIGR4 CBPs. ‘Repeats’ are marked by brackets above; the C-terminal conserved end is boxed. *Mark positions implicated in choline binding of CbpF (Molina et al., 2009). PspC* refers to the truncated fragment of a pseudogene.

The CBD of LytA has been crystallized and the structure was determined at a resolution of 2.4–2.8 Å. The LytA-CBD adopts a solenoid fold that is maintained by the binding of choline residues (Fernández-Tornero et al., 2001, 2002). Also, the crystal structure of a recombinant Pce lacking 85 residues at the C-terminus has been resolved in complex with the reaction product phosphocholine and choline analogues. The choline-binding module is very elongated (100 Å) and organized in a superhelical fold that includes the first 20 aa of the C-terminus (Hermoso et al., 2005).

Similar cell wall-binding modules have been described in glucan-binding proteins (glucansucrases and glucosyltransferases) of Leuconostoc mesenteroides and of oral streptococci including Streptococcus mutans, in the surface antigen SpaA of Erysipelothrix rhusiopathiae (Makino et al., 2000), and as cell wall lytic enzymes and toxins of Clostridium spp. as well (Wren et al., 1991; Croux et al., 1993; Shah et al., 2004). Because of the high degree of similarity to CBDs, these proteins give rise to erroneous annotations as has been pointed out before (Shah et al., 2004). The repeating units are responsible for carbohydrate binding (Shah et al., 2004). Interestingly, the cell wall-binding domain of the Clostridium difficile toxin A shows a weak binding to choline, and it dimerizes in the presence of choline (Demarest et al., 2005). However, the crystal structure of a C-terminal fragment of ToxA is arranged in a β-solenoid fold containing five repeats (Ho et al., 2005), all features described for the cell wall anchor domain of LytA (Fernández-Tornero et al., 2001).

The non-CBP module located in most cases at the N-terminus contains the functional domain specific for each CBP. With the exception of LytA, it is preceded by a signal peptide (for review, see Bergmann & Hammerschmidt, 2006). Individual pneumococcal strains contain a variable number (>10) of CBPs. Some CBPs carry different names depending on the authors, and also the annotation differs between the first two genomes of the unencapsulated laboratory strain S. pneumoniae R6 and the type 4 TIGR4 strain (Table 1). The CBPs of strain D39, the ancestor of R6, do not differ from the R6 genes, except that in some cases the start codons have been redefined (Lanie et al., 2007). A summary of CBPs found in another four S. pneumoniae genomes available in the data bank of NCBI (http://www.ncbi.nlm.nih.gov/genomes/lproks.cgi) and the Sanger center (http://www.sanger.ac.uk/Projects/S_pneumoniae/) compared with those of TIGR4 and R6 is summarized in Table 1: a Hungarian isolate Hu-6 of type 19A, a serotype 19F strain G54, the type 14 strain CGSP14 and a representative of the multiresistant and high-level penicillin-resistant clone Spain23F−1 (Aanensen et al., 2007). Differences will be discussed below in the context of CBPs and related cell wall-associated proteins in commensal streptococci.

View this table:
Table 1

Choline-binding proteins in Streptococcus pneumoniae and Streptococcus mitis

Current nameS. pneumoniaeS. mitis
TIGR geneName (aa)R6 geneName (aa)G54Hu19A_6CGSP14Sanger 23F−1B6NCTC12261
CbpISP0069CbpI (211)+++
PspASP0117PspA (744)spr0121PspA (619)++Variant2 frames
CbpCSP0377CbpC (340)spr0337CbpF (338)++
CbpJSP0378CbpJ (332)++
CbpGSP0390CbpG (285)spr0349CbpF trunc. (218)++++
spr0350CbpF trunc. (42)
CbpFSP0391CbpF (340)spr0351PcpC (294)(+)+2 frames+++
SP0667cons. hypo. (332)spr0583hypo (329)+++2 genes++
PceSP0930CbpE (627)spr0831CbpE (627)++++++
LytBSP0965LytB (658)spr0867LytB (721)++++++
SP1417PspC, degen. (129)spr1274SpsA trunc. (129)++++
LytCSP1573LytC (490)spr1431LytC (501)++++++
LytASP1937LytA (318)spr1754LytA (318)++++ΦB6
PcpASP2136PcpA (621)spr1945PcpA (690)++++
PspCSP2190CbpA (693)spr1995PcpC (701)+++Longer
CbpDSP2201CbpD (448)spr2006CbpD (448)++++++
  • * pm, point mutation; two genes.

  • Frameshift annotated, but confirmed as functional gene product (Moscoso et al., 2005a).

Several CBPs have been implicated in virulence. Among these, PcpA, PspA and PspC (also named CbpA, SpsA, PbcA and Hic), in addition to the four cell wall hydrolytic enzymes LytA, LytB, LytC and Pce (LytD and CbpE), have been extensively characterized. Recent review articles describe the role of pneumococcal CBPs in cell wall metabolism and virulence in detail (Bergmann & Hammerschmidt, 2006; Hammerschmidt et al., 2007; Vollmer et al., 2007).

A promising therapeutic application using CBPs has been demonstrated using purified LytA against β-lactam-resistant S. pneumoniae in a peritonitis-sepsis mouse model (Rodríguez-Cerrato et al., 2007). The enzyme reduced peritoneal bacterial counts more than five log10, and was more effective than the phage-encoded lysozyme Cpl-1 (Rodríguez-Cerrato et al., 2007), which had also been used in a variety of other mouse models (Loeffler et al., 2003; Entenza et al., 2005; Grandgirard et al., 2008; Witzenrath et al., 2009).

Cell wall hydrolytic CBPs

The major autolysin LytA, an N-acetylmuramoyl-l-alanine amidase responsible for stationary-phase autolysis typical of pneumococci, is highly conserved among pneumococci, exhibiting only slight sequence diversity at the DNA level (Gillespie et al., 1997; Llull et al., 2006). It has been reported to be introduced by recombination with phages that frequently encode a lytA homologue (Whatmore & Dowson, 1999). Because it lacks a signal peptide, it is still unknown as to how it reaches the extracytoplasmic murein substrate (Díaz et al., 1989).

LytA is synthesized in a low-activity E-form (Tomasz & Westphal, 1971), which is converted to the active C-form upon interaction with choline residues of the cell wall TA at a low temperature (Höltje & Tomasz, 1976). Conversion is also observed with LTA in non-micellar form, and by preincubation with 2% choline in vitro, followed by dilution below 0.04% choline in the assay, whereas low choline concentrations (0.1%) in the assay prevent conversion (Briese & Hakenbeck, 1984, 1985). The E-form of the enzyme behaves as a monomer (Höltje & Tomasz, 1976), and in the presence of choline LytA forms dimers through the interaction of the CBD. Key amino acids implicated in dimer formations have been investigated (Usobiaga et al., 1996). The C-form enzyme also readily digests choline-free murein, indicating that the requirement for choline for LytA activity is only given in case of the WTA–murein complex (Garcia-Bustos & Tomasz, 1987; Severin et al., 1997). The E-form enzyme is only found in cytoplasmic fractions and not in the cell wall (Briese & Hakenbeck, 1983, 1985; Giudicelli & Tomasz, 1984), and one might assume that LytA is eventually released from the cell only through the lytic activities of other enzymes.

The physiological role of LytA is unclear. It has been suggested that it mediates the release of pneumolysin, one of the major virulence factors of pneumococci, into the extracellular environment (Lock et al., 1992). This might be coupled to the state of genetic competence: LytA activity in competent cells causes lysis of noncompetent cells (allolysis) and thereby releases pneumolysin (Guiral et al., 2005). However, it is not known whether this mechanism is important during infection. A lytA-independent, unknown mechanism for pneumolysin release has been reported in certain strains of S. pneumoniae (Balachandran et al., 2001). Pneumolysin sequesters to the cell wall fraction in a lytA-independent fashion (Price & Camilli, 2009). The mechanism by which Ply is released from the cell wall in the host is yet to be determined.

LytA mutants containing type 3 and type 6 capsules are virulent to mice similar to the parental strain (Tomasz et al., 2008), whereas a lytA mutant of the type 2 strain D39 displayed a significantly reduced virulence in a mouse intraperitoneal model (Berry & Paton, 2000). A crucial role of autolysin in the pathogenesis of pneumococcal meningitis has been documented (Hirst et al., 2008).

The LytB N-acetylglucosaminidase (García, 1999a) is involved in cell separation: lytB mutants grow in long chains of nonseparated cells (De Las Rivas et al., 2002). A purified GFP-LytB fusion protein localizes at the cell poles of wild-type and lytB mutant cells (De Las Rivas et al., 2002). How the protein is targeted to these sites is not known.

LytC has a C-terminal catalytic domain that has sequence similarity to N-acetylmuramidases (lysozyme), and an N-terminal CBD (García et al., 1999b). Its optimal temperature is 30 °C, and a double mutant in lytA and lytC (but not in lytC alone) shows reduced lysis at this temperature. Although in the initial genome sequence of TIGR4 lytC has been reported to contain an authentic frameshift mutation, further analysis revealed a functional gene product (Moscoso et al., 2005a). A three-dimensional model of LytC has been reported based on homology modelling of each module (Monterroso et al., 2005).

Both LytC and LytB mutants have a reduced ability to colonize the nasopharynx of rats (Gosink et al., 2000).

The phosphorylcholine esterase Pce (CbpE; LytD) has been found in cell extracts as an activity that removes only about 20% of the cell wall-associated phosphorylcholine residues in vitro (Höltje & Tomasz, 1974). Mutants in pce show no obvious phenotype in the laboratory, but have an altered colony morphology, and a type 3 strain lacking pce was more virulent in the intraperitoneal mouse model (De Las Rivas et al., 2001; Vollmer & Tomasz, 2005). Its structure has been resolved (Hermoso et al., 2005), and in an independent effort, the structure of the catalytic region was determined (Garau et al., 2005).

CbpD plays a role during competence-induced cell lysis (Steinmoen et al., 2003; Guiral et al., 2005; Kausmally et al., 2005). This has been supported by the fact that the amount of DNA released into the medium during competence induction is strongly reduced in cbpD mutants (Kausmally et al., 2005). Presumably, the two bacteriocins CibAB that are induced during competence act as a trigger for the murein lytic enzymes LytA, LytC and CbpD (Guiral et al., 2005). The N-terminal domain of CbpD belongs to the AcmB/LytN subfamily of CHAP enzymes, and therefore it has been postulated that it functions as a cell wall hydrolytic amidase (Kausmally et al., 2005).

Other pneumococcal CBPs

PcpA contains several leucine-rich repeats in its N-terminal domain (Sánchez-Beato et al., 1998). It has been identified as a virulence factor in a large-scale sequence-tagged mutagenesis (STM) study in murine models of sepsis and pneumonia using the type 4 TIGR4 strain (Hava & Camilli, 2005), together with other CBPs including PspC (CbpA), CbpC and PspA. PcpA is under the control of the manganese-dependent regulator PsaR (Johnston et al., 2006), and it has recently been shown that it elicits protection against lung infection and sepsis in murine models (Glover et al., 2008).

The PspA protein is a highly variable protein as shown by immunological studies (Crain et al., 1990) and by sequencing a collection of genetically diverse clinical isolates (Hollingshead et al., 2000), revealing a highly complex mosaic structure of the gene. The N-terminal part of PspA is separated from the C-terminal CBD by a proline-rich domain. It has been predicted to be arranged in long antiparallel coiled-coils (for a review, see Jedrzejas et al., 2001). PspA is highly immunogenic, and immunization of mice identified protective epitopes in the N-terminal domain (Talkington et al., 1991). Its impact on virulence became clear after documenting its ability to interfere with complement activation, and to bind lactoferrin (Hammerschmidt et al., 1999; Tu et al., 1999). The recent structure of the complex of the N-terminal lactoferrin-binding domain of PspA and the N-lobe of lactoferrin reveals how PspA protects pneumococci from exposure to the lactoferrin domain (Senkovich et al., 2007). There is recent evidence that PspA gene expression is driven by the VicRK (RR02/HK02) regulatory system (Ng et al., 2004).

PspC (CbpA, SpsA, PbcA, Hic) is the major pneumococcal adhesin. This multifunctional protein contributes to binding of pneumococcal cells to epithelial cells (Rosenow et al., 1997) and binds to factor H, thus avoiding complement activation (Duthy et al., 2002; Jarva et al., 2002). It binds to the secretory component associated with the secretory immunoglobulin A or the polymeric immunoglobulin receptor (Hammerschmidt et al., 2000). Its impact on virulence has been shown in several mouse infection models (Zhang et al., 2000; Duthy et al., 2002; Iannelli et al., 2002).

Comparison of pspC of 43 S. pneumoniae strains showed a highly variable sequence due to a mosaic structure of pspC, similar to pspA (Iannelli et al., 2002). The pspC locus is highly polymorphic, and the authors describe eleven groups of PspC proteins that were highly variable with respect to the length of the proline-rich region, but in all cases the DNA encoding for the signal peptides was conserved (Iannelli et al., 2002). Some variants described as Hic (Janulczyk et al., 2000), which are located at the same position in the genome, contain an LPXTG motif typical for cell wall anchor proteins instead of the CBD (Iannelli et al., 2002). Its impact on virulence has been shown in several mouse infection models (Zhang et al., 2000; Duthy et al., 2002; Iannelli et al., 2002). Most interestingly, a subgroup of strains contains two tandem copies of pspC. The two-component system RR06/HK06 controls pspC expression (Standish et al., 2005).

Another two CBPs have been implicated in virulence. CbpC has been identified in an STM study as a virulence factor in mouse models (Hava & Camilli, 2005), but its function is unknown. Mutants in CbpG have a decreased ability to adhere to human nasopharyngeal epithelia cells, and this property has been linked to a reduced colonization of the rat nasopharynx. Moreover, loss of function of CbpG also showed reduced virulence in a mouse sepsis model (Gosink et al., 2000). It has been suggested that it cleaves host extracellular matrix and thus participates in the bacterial adhesion process (Mann et al., 2006).

CbpF has been crystallized recently (Molina et al., 2007), and its structure in complex with choline has now been reported (Molina et al., 2009). CbpF is a member of a new subfamily of the CBPs that are entirely composed of a combination of so-called canonical choline-binding repeats at the C-terminus involved in choline-binding and nonconserved choline-binding motifs of the N-terminal module that are highly modified. Most remarkably, CbpF inhibited LytC activity (but not LytA or Pce) in vivo and in vitro, and the pneumococcal phage-associated lysozymes Cpl-1 and Cpl-7 as well, and this effect was not due to competition for choline residues (Molina et al., 2009). Thus, CbpF might be crucial in regulating the LytC activity.

Besides sequence information, little is known about CbpI and CbpJ. Preliminary crystallization data of CbpI (Paterson et al., 2006) are available.


One highly conserved gene encoded by SP0667/spr0583 in the TIGR4/R6 is present in all pneumococcal strains, with one exception. In the Sanger 23F strain, a protein encoded by SPN23F06030 contains 94 identical amino acids compared with the C-terminal part of SP0667, preceded by a hypothetical protein SPN23F06010 (containing an internal duplication of SPN23F06030) that is completely absent in the other genomes analyzed here. Although SP0667 and homologues contain a domain highly similar to the CBD in the CBPs (COG5263; glucan-binding domain; YG repeat), it has not been annotated as CBP in the first S. pneumoniae genomes (Hoskins et al., 2001; Tettelin et al., 2001), but is referred to as a hypothetical protein in R6, as a ‘pneumococcal surface protein, putative’ in TIGR4, CGSP14 and the Sanger strain, as an ‘excalibur containing-containing protein’ in Hu19_6 and only in G54 as a CBP. It has been noted in the R6 genome that it is ‘similar to 1,4-β-N-acetylmuramidase’, and in fact blast searches reveal a much higher score for the repeats found in LytC compared with other CBPs.

The major difference in the repeat structure of SP0667 is that a perfect 40mer repeat can be aligned, whereas a 20mer is not possible in contrast to other CBPs for example PspA (Fig. 3; see also Fig. 2). An alignment between both types of repeats can only be achieved by introducing several gaps, but amino acids implicated in choline binding of CbpF are also conserved in the SP0667 repeat (Fig. 3). Because these residues involved in choline binding are asymmetrically distributed within these two repeats, an arrangement in 40meric units might generally be more suitable in presenting the CBD. Although the ligand for the S. pneumoniae SP0667 protein is still unknown, it is likely that it is a CBP, but that might interact with the pneumococcal cell wall in a somewhat different way. The N-terminal part has no homologue in other bacteria.

Figure 3

Apparent choline-binding domain in SP0667. (a) The C-terminal repeating units of SP0067 and those of PcpA are shown as alignment. The first amino acid of the part shown is indicated. (b) The alignment shown at the bottom includes the CbpF repeat, and amino acids involved in choline binding are marked (*).

CBPs in commensal species and evolution of CBPs

Because choline has been recognized in the cell wall of S. oralis and S. mitis, it is not too surprising that CBPs have also been detected in these organisms. Choline esterase activity has been detected in S. oralis (Ronda et al., 1991), and LytB variants were recognized in the Mitis group (Moscoso et al., 2005b). We have searched the unfinished genome of S. mitis NCTC12261 (http://cmr.jcvi.org/tigr-scripts/CMR/shared/Genomes.cgi), and the recently finished genome sequence of S. mitis B6 (to be published elsewhere), an unusually high-level penicillin and multiple antibiotic-resistant strain isolated in Germany (Hakenbeck et al., 1998; König et al., 1998), for the presence of CBPs, and their relationship with the S. pneumoniae proteins (Table 1).

The S. mitis B6 genome contains a LytA gene as part of its prophage ΦB6. The association of LytA homologues with S. pneumoniae bacteriophages has long been recognized (López & García, 2004), and ΦB6 provides a new example of this phenomenon in the Mitis group, together with the recently described LytA from the S. mitis phage ΦHER (Romero et al., 2005). Both phage enzymes contain a Thr317 instead of Val317 as is found in the S. pneumoniae LytA, which is located in one of the major hydrophobic zones implicated in dimer formation (Romero et al., 2005). Indeed, whereas the wild-type LytAR6 formed dimers in the absence of choline, a mutant V317T remained mainly monomeric under these conditions, although dimerization occurred at increasing choline concentrations (Romero et al., 2005).

Remarkably, all other five cell wall lytic enzymes are present in both S. mitis B6 and NCTC12261: LytB, LytC, Pce, LytE and CbpD. All these proteins contain a CBD. This implies that a choline-containing TA must be present in these strains. Moreover, the high identity of the nonrepeat domain of the homologues on the amino acid level between the two species indicates that the functions of these proteins are also well conserved in commensal streptococci, i.e. have not evolved in the context of pathogenesis of S. pneumoniae. The S. mitis B6 LytB is smaller and contains some variation in the nonrepeat region compared with the pneumococcal protein. We have confirmed the role of the LytB gene product in cell separation in S. mitis NCTC10712, another S. mitis type strain (Kilian et al., 1989), which is almost identical to the B6 LytB (GenBank accession number FJ824668). This strain develops genetic competence under laboratory conditions in contrast to many other S. mitis strains, and therefore we chose this strain to construct a lytB mutant by replacing the entire gene with the spectinomycin resistance gene aad9 from Enterococcus faecalis (LeBlanc et al., 1991). Streptococcus mitis NCTC10712 is strictly choline dependent (as is the B6 strain) and grows in long chains of unseparated cells in the presence of 2% choline, and this phenotype is indeed expressed in the lytB mutant under normal growth conditions (Fig. 4). In this context, it is interesting that sequence analysis of LytB in a collection of S. mitis strains revealed a high variation in the numbers as well as in the amino acid sequences of the choline-binding modules (Moscoso et al., 2005b), perhaps indicating adjustment to biochemically different TA repeat units. It is likely that the other lytic enzymes also function in a manner similar to the pneumococcal proteins, but experimental proof is required.

Figure 4

Growth of a lytB mutant Streptococcus mitis NCTC10712. (a) Cells were grown in CDM (closed circles), and growth was followed by nephelometry (N). Open circles: CDM without choline. Bacteria were examined by phase-contrast microscopy during the exponential growth phase. (b) Streptococcus mitis NCTC10712; (c) S. mitis NCTC10712 in CDM with 2% choline; and (d) lytB mutant of S. mitis NCTC10712 in CDM.

Homologues of cbpI and cbpF (S. pneumoniae TIGR4 nomenclature) were also found in S. mitis B6 and NCTC12261. Both regions are particularly interesting because they reveal processes related to the evolution of CBPs.

The CbpI gene appears to be an example of CBP decay in S. pneumoniae. The cbpI gene SP0069 in S. pneumoniae TIGR4 is preceded by two genes (SP0067 and 0068). In other S. pneumoniae strains, all three genes and the downstream region are either completely missing, or one gene includes SP0067/0068 as in the case of S. pneumoniae G54 (Table 1). In contrast, both S. mitis strains contain one gene covering the region of the SP0067–SP0069 fragments (Fig. 5). This suggests that this gene has evolved before the evolution of S. pneumoniae in the commensal species, and that it is decaying in the pathogen. However, the apparent intragenic mutations will have to be verified by direct sequence analysis.

Figure 5

CbpI region in Streptococcus pneumoniae and Streptococcus mitis. The arrows indicate ORFs in the genomes of S. mitis B6 (black arrow), S. pneumoniae TIGR4 (three frames) and S. pneumoniae G54 (two frames). The triangle indicates a three-codon insert in the S. pneumoniae genes. The flanking genomic regions are identical in all three cases.

The situation concerning the two tandem pairs CbpC/J (SP0377/0378) and CbpG/CbpF (SP0390/0391) is more complex. There is only one CBP at the position of SP0377 in both S. mitis strains that also shows homology to this protein, although it is more similar to CbpF. On the other hand, the region SP0389–0399 including CbpG/F is missing (Fig. 6). In other S. pneumoniae strains, only one CBP with higher similarity to either CbpC or CbpJ is present at the location of the TIGR4 CbpC/J genes. Thus, the tandem arrangement cbpC/J appears to be specific to the TIGR4 strain, suggesting that gene duplication and diversification might have occurred in this particular case. CbpC, J and F contain closely related N- and C-terminal domains, whereas CbpG shares an identical 36 aa C-terminus only, suggesting that gene transfer events and recombination at these loci also played a role.

Figure 6

CbpC/J and CbpG/F region in Streptococcus pneumoniae TIGR4, R6, CGSP14 and Streptococcus mitis B6. The genomic regions including the two pairs of tandem CbpC/J and CbpG/F (TIGR4 nomenclature) are shown. Gray areas indicate sequence similarity. CbpC is shown as a black arrow; similar shading of the genes (arrows) indicates homologues.

In contrast, in the other S. pneumoniae strains the CbpG/CbpF region is well conserved, except that either cbpG (R6) or cbpF (CGSP14) appears to be fragmented (Table 1 and Fig. 6).

Gene duplication has been recognized as an important mechanism related to gene transfer to evolve new functions (Otto & Yong, 2002; Hooper & Berg, 2003). The complex pattern of homologous regions concerning these four CBPs points to a common evolutionary origin of these paralogues, involving duplication of one ancestor CBP, followed by duplication of the tandem genes and diversification that might be triggered by recombination events in S. pneumoniae.

None of the three S. pneumoniae CBPs PcpA, PspA or PspC are present in the two S. mitis genomes i.e. CBPs that have been shown to play a major role in the virulence of S. pneumoniae and that are found in the genomes analyzed here. This strongly suggests that one key difference between the pathogenic and the commensal species includes evolutionary events concerning these CBPs. There are several indications that these genes have been introduced into S. pneumoniae, and that further diversification has occurred in this species as is apparent by the variability of pspC (hic) and pspA. For example, the PcpA gene is located adjacent to an IS1318 element including a degenerate orfA gene. However in S. mitis, B6 pcpA is absent, but IS1318 appears to be intact. In several S. pneumoniae strains, IS elements were found flanking the PspC gene (SP2190) (Iannelli et al., 2002). PspC and the adjacent genes SP2191–Sp2193, which include a cognate two-component regulatory system in S. pneumoniae, replace a distinct region that is perfectly conserved in both S. mitis strains, suggesting acquisition of the pspC region in S. pneumoniae via gene transfer. The TIGR4 PspA gene and flanking regions (SP0115–0117) replace other genes in S. mitis. In the R6 strain, pspA (spr0121) and another 18 genes (spr0103–0121) are not present in both S. mitis strains, whereas spr0102–0108 are absent in TIGR4 but present in S. mitis, suggesting that this region might represent a hot spot for recombination events.

The SP0067/spr0583 gene is highly conserved in the two S. mitis strains (Table 1), and probably also in other S. mitis as suggested by hybridization experiments using specific oligonucleotides (unpublished data).

Concluding remarks

Phosphorylcholine is being discussed as a potential candidate for a plurispecific vaccine. However, one should be aware of the fact that such vaccines also target commensal organisms. It is clear by now that S. mitis as well as S. oralis possess a choline-containing cell wall polysaccharide whose structure remains to be clarified, but most likely is very similar to that of S. pneumoniae. Whereas the cell wall lytic CBPs are highly conserved among S. pneumoniae and S. mitis, other CBPs including major virulence factors S. pneumoniae are species specific and show a considerable variation between pneumococcal strains. CBPs are surface-exposed components, and are thus targeted by human defense mechanisms, enforcing the selection of sequence variants in order to evade the immunological responses.

CBPs appear to be a playground for fast-evolving surface components. They often occur in a tandem arrangement that can be considered as the result of gene duplication events, and truncated versions are frequently encountered in the genomes. Both features are indicative of evolutionary forces acting on these surface proteins, and gene transfer events are also likely (Otto & Yong, 2002; Hooper & Berg, 2003). We propose that the three major virulence components PspA, CbpA and PcpA play a key role in the origin of the species S. pneumoniae. The fast increase of genomic information will help to reveal in more detail the evolutionary mechanisms of these interesting cell wall-associated proteins.


The help of Michael Nuhn in bioinformatic analysis is greatly appreciated. This work has been supported by the EU (Intafar, LSHM-CT-2004-512138; EuroPathoGenomics LSHB-CT-2005-512061) and the Stiftung Rheinland Pfalz für Innovation (project 838).


  • Editor: Josep Casadesus


View Abstract