Identification of a Pneumococcal Glycosidase That Modifies O-Linked Glycans

Abstract

Colonization of the airway by Streptococcus pneumoniae is typically asymptomatic; however, progression of bacteria beyond the oronasopharynx can cause diseases including otitis media and pneumonia. The mechanisms by which S. pneumoniae establishes and maintains colonization remain poorly understood. Both N-linked and O-linked glycans are abundant in the airway. Our previous research demonstrated that S. pneumoniae can sequentially deglycosylate N-linked glycans and suggested that this modification of sugar structures may aid in colonization. There is published evidence that S. pneumoniae expresses a secreted O-glycosidase that cleaves galactose β1-3 N-acetylgalactosamine (Galβ1-3GalNAc) from core-1 O-linked glycans; however, the biological function of this enzyme has not previously been determined. We established that the activity is not secreted but is instead surface associated in a sortase-dependent manner. Genome analysis revealed an open reading frame predicted to encode a sortase-dependent surface protein with sequence similarity to the O-glycosidase of Bifidobacterium longum. Deletion of this pneumococcal open reading frame confirmed that this gene encodes an O-glycosidase. Experiments using a model glycoconjugate demonstrated that this O-glycosidase, together with the neuraminidase NanA, is required for S. pneumoniae to cleave sialylated core-1 O-linked glycans. The ability of the O-glycosidase mutant to cleave this glycan structure was restored by both genetic complementation and the addition of O-glycosidase. The mutant showed a reduction in adherence to human airway epithelial cells and a significantly decreased ability to colonize the upper respiratory tract, suggesting that cleavage of core-1 O-linked glycans enhances the ability of S. pneumoniae to colonize the human airway.

Streptococcus pneumoniae (pneumococcus) is the leading cause of community-acquired pneumonia and is also a major cause of otitis media, bacteremia, and meningitis (16, 28, 33). While S. pneumoniae is an important human pathogen, the bacterium more frequently colonizes the oronasopharynx asymptomatically (14, 42). Although colonization is often cleared by the host, it is a necessary precursor to disease and is therefore essential to pneumococcal pathogenesis. Despite the importance of this process, little is known about the mechanisms by which the bacterium establishes and maintains colonization.

The epithelial cell surface, host defense molecules, and the mucous layer are decorated with diverse sugar structures, including both N- and O-linked glycans. S. pneumoniae is adept at manipulating sugars, carrying at least six surface-associated glycosidases that likely modify human glycoconjugates (3, 8, 9, 32, 41, 46). We previously demonstrated that S. pneumoniae glycosidases mediate sequential deglycosylation of N-linked glycans (24). This deglycosylation has been proposed to contribute to colonization by providing sugar residues for bacterial growth, revealing receptors for adherence, and modifying the function of host defense molecules (6, 24). The ability of S. pneumoniae to modify O-linked glycans and the contribution of this deglycosylation to pneumococcal pathogenesis have not yet been examined. In addition to contributing to pneumococcal pathogenesis by the mechanisms proposed for N-linked glycans, the deglycosylation of O-linked glycans could also contribute to pathogenesis by enabling progression through the mucous layer.

S. pneumoniae expresses an O-glycosidase, designated Eng (endo-α-N-acetylgalactosaminidase), that acts to specifically cleave core-1 O-linked glycans (4, 5, 25, 41). The core-1 structure consists of a galactose β1-3 N-acetylgalactosamine (Galβ1-3GalNAc) disaccharide linked to a serine or threonine residue (Fig. 1). This core structure can be modified further to create more complex glycan structures. The addition of terminal sialic acid, as shown in Fig. 1, is a common modification. This structure is present in the human airway, for example, decorating the hinge region of immunoglobulin A (IgA) (30). In conjunction with other glycosidases, including the neuraminidase NanA, O-glycosidase could enable S. pneumoniae to deglycosylate O-linked glycans. Despite the demonstration that S. pneumoniae expresses secreted O-glycosidase activity, any contribution of this glycosidase to pneumococcal pathogenesis has not been investigated.

FIG. 1.

Schematic representation of the core-1 O-linked glycan structure commonly found on human glycoconjugates. The sugar residues are labeled beneath their corresponding symbols. Lines represent linkages between the sugar residues, and numbers indicate the specific linkages. Arrows above the schematic indicate potential cleavage sites of the glycosidases.

In this study, we identify the gene encoding the pneumococcal O-glycosidase. The product of this gene is surface associated in a sortase-dependent manner, and our evidence suggests that this gene encodes the only detectable O-glycosidase expressed by S. pneumoniae. We also establish that the O-glycosidase cleaves the core-1 structure from glycoconjugates following cleavage of terminal sialic acid by the neuraminidase NanA. Furthermore, we identify a role for this enzyme in colonization of the upper respiratory tract and adherence to human epithelial cells, which suggests that the O-glycosidase may contribute to pneumococcal pathogenesis.

MATERIALS AND METHODS

Bacterial strains, culture media, and chemicals.

Parental and genetically modified strains of S. pneumoniae and the five pneumococcal clinical isolates utilized in this study are described in Table 1. The five clinical isolates represent five different multilocus sequence types and five distinct serotypes. Broth cultures were routinely grown at 37°C in Todd-Hewitt broth (Becton Dickinson) supplemented with 0.2% (wt/vol) yeast extract (Becton Dickinson). C medium with 5% yeast extract, pH 8, was used for transformations (27). S. pneumoniae was also grown at 37°C and 5% CO₂ overnight on tryptic soy (Becton Dickinson) plates with 1.5% agar that were spread with 5,000 units of catalase (Worthington Biochemical Corporation) prior to plating of bacteria. Mutants were selected on tryptic soy agar plates that contained either streptomycin (200 μg/ml) or kanamycin (500 μg/ml), as appropriate.

TABLE 1.

Strains of Streptococcus pneumoniae used in this study

Strain	Serotype	Characteristics/genotypea	Source or reference
TIGR4	4	Clinical isolate	40
TIGR4 Smr	4	Lys56→Thr in RpsL [rpsL(K56T)], conferring Smr	2
TIGR4 ΔsrtA	4	Cmr	J. N. Weiser
TIGR4 Δeng	4	Δeng rpsL(K56T) (Smr)	This study
1121 Smr	23F	Lys56→Thr in RpsL [rpsL(K56T)], conferring Smr	This study
1121 Δeng	23F	Δeng rpsL(K56T) (Smr)	This study
1121 Δeng ΔnanA	23F	ΔnanA (Cmr) Δeng rpsL(K56T) (Smr)	This study
C06_29	15B/C	Clinical isolate	6
C06_31	23F	Clinical isolate	6
C06_31 Smr	23F	Lys56→Thr in RpsL [rpsL(K56T)], conferring Smr	This study
C06_31 Δeng	23F	Δeng rpsL(K56T) (Smr)	This study
C06_39	35F	Clinical isolate	6
C06_57	6A/B	Clinical isolate	6
C06_57 Smr	6A/B	Lys56→Thr in RpsL [rpsL(K56T)], conferring Smr	This study
C06_57 Δeng	6A/B	Δeng rpsL(K56T) (Smr)	This study
C06_58	19A	Clinical isolate	6

^aSm^r, resistance to streptomycin; Cm^r, resistance to chloramphenicol.

Unless otherwise specified, all chemicals, substrates, and enzymes were purchased from Sigma Chemical Co.

Mutation of glycosidases.

An unmarked mutant in SP0368, the last open reading frame in the predicted transcriptional unit, was generated using the Janus cassette selection system (39). Construction of a mutant using this method requires two rounds of transformation. The first round introduces a Janus cassette, which contains kanamycin resistance and streptomycin sensitivity (rpsL⁺) genes, into the genome of a streptomycin-resistant (Sm^r) S. pneumoniae strain in place of the gene of interest. DNA fragments flanking the region to be deleted were amplified (primers O.1 and O.2 and primers O.4 and O.5) and sequentially joined to the Janus cassette PCR product (primers J.1 and J.2), using a variation of splicing by overlap extension-PCR (6) first described by Horton et al. (22). Genomic DNA was prepared as previously described (43). In order to minimize PCR-generated errors, all PCRs were conducted using a high-fidelity proofreading polymerase (Pfx-50; Invitrogen). The Janus construct was transformed into S. pneumoniae, and the transformants were selected on kanamycin and confirmed by PCR (with primers O.7 and O.8). The second round of transformation replaced the Janus cassette with an engineered segment of DNA consisting of the two DNA fragments flanking SP0368 spliced together via splicing by overlap extension. Fragments flanking SP0368 were amplified using primers O.1 and O.3 (upstream fragment) and primers O.5 and O.6 (downstream fragment). Construction of an unmarked SP0368 mutant (1121 Δeng) was confirmed by PCR with primers flanking the construct (primers O.7 and O.8) and by sequencing (Table 2). The complemented strain (1121 Δeng/eng⁺) was generated by transforming 1121 Δeng with the Janus construct and then with 1121 Sm^r DNA. Complementation was confirmed by PCR with primers flanking the construct (primers O.7 and O.8) and by activity assays. To further investigate the pneumococcal O-glycosidase, unmarked mutations were also generated in TIGR4 and two recent clinical isolates (C06_31 and C06_57).

TABLE 2.

Primers used in this study

Group	Primer	Primer sequence (5′→3′)a	Nucleotide location (GenBank accession no.)
nanA	N.1	TATCGAGTAGGGTAGTTCTT	149-169 (X72967)
	N.2	ACGGGGCAGGTTAGTGACAT	1953-1972 (V01277)
	N.3	TAGTTCAACAAAGGAAAATTGGATAA	1101-1126 (V01227)
	N.4	AGCACGAACTGGAATCTTACCT	1008-1037 (U43526)
Janus	J.1	CCGTTTGATTTTTAATGGATAATG	7-30 (AY334019)
	J.2	GGGCCCCTTTCCTTATGCTT	247511-247527 (AE005672)
eng	O.1	CTGCTGGCGAAGACTGGGAT	340567-340586 (AE005672)
	O.2	CATTATCCATTAAAAATCAAACGGCATTTTTGAGCAAGCGTTTACA1	341344-341364 (AE005672)
	O.3	CTCACCTGTTGCTGGCACATTTTTGAGCAAGCGTTTACA2	341344-341364 (AE005672)
	O.4	AAGCATAAGGAAAGGGGCCCTGCCAGCAACAGGTGAGAG3	346586-346604 (AE005672)
	O.5	GTGCCATGAAAGAAACGACAG	347431-347451 (AE005672)
	O.6	TGCCAGCAACAGGTGAGAG	346586-346604 (AE005672)
	O.7	GCGTCCAATCTATGAAAC	340450-340467 (AE005672)
	O.8	CAGCCGTGGAAACTCTAAACA	347691-347718 (AE005672)
	O.9	AGCGCCAGCCTTGTTTGT	343742-343759 (AE005672)
	O.10	AACGCTTCAGAAACTTATCC	343468-343487 (AE005672)

^aUnderlining indicates the reverse complement sequences of primers J.1 (1), O.6 (2), and J.2 (3).

The gene encoding the pneumococcal neuraminidase NanA is predicted to be in a single transcriptional unit (40). Therefore, an insertion-deletion mutant was constructed in strain 1121 Δeng as previously described and confirmed by PCR and sequencing, using primers A.1 and A.2 and primers A.3 and A.4 (24). Growth rates and maximum optical densities at 600 nm were not significantly different for mutant strains compared to parental strains (data not shown). Since opacity can also affect the adherence of S. pneumoniae, all mutants were confirmed to be of the same opacity as their parental strains (13, 19, 23).

O-glycosidase activity.

O-glycosidase activity was measured using the colorimetric substrate Galβ1-3GalNacα1pNP (Calbiochem). Bacterial strains were grown to the appropriate optical density at 600 nm. Aliquots of the culture were lysed with toluene to prevent growth during the assay and were used for activity assays. The reactions were started with the addition of 40 μl of 50 mM sodium acetate buffer (pH 5.0) containing 250 μM Galβ1-3GalNacα1pNP to 40 μl of bacterial sample. After incubation at 37°C for 1 h, the reactions were stopped with the addition of 60 μl of 1 M sodium carbonate. The reactions were centrifuged to remove any bacterial cells, and the amount of p-nitrophenol released was determined by measuring the absorbance at 400 nm. All assays were performed in triplicate on three independent occasions. Data for an appropriate medium-only control were subtracted from all data.

Deglycosylation of O-linked glycans.

To assess the ability of pneumococci to cleave O-linked glycans from glycoconjugates, lectin blots were performed with samples incubated with different bacterial strains and glycosidases. Bacterial cultures were grown to stationary phase, 20 μl of culture was added to 0.2 μg of fetuin, and the reaction mixtures were incubated at 37°C overnight. Controls consisting of medium alone incubated overnight with the protein were also performed. To terminate the reactions, gel loading buffer was added and samples were placed at −20°C. Where specified, the following enzymes were added: 0.0008 unit of purified Clostridium perfringens neuraminidase (resuspended in 10 mM potassium phosphate buffer, pH 7) and 0.00016 unit of S. pneumoniae O-glycosidase (in 50 mM sodium phosphate, pH 7.5) (QA Bio). These quantities of glycosidases were selected by in vitro assays, which demonstrated that they possessed approximately the same level of activity as that in 20 μl of stationary-phase culture.

To investigate deglycosylation following glycosidase treatment, sodium dodecyl sulfate-polyacrylamide gel electrophoresis (12.5%) was performed, and samples were transferred to Immobilon-P membranes (Millipore) and detected using the lectins Maackia amurensis agglutinin (MAA), peanut agglutinin (PNA), and Datura stramonium agglutinin (DSA) from a DIG glycan differentiation kit (Roche) according to the manufacturer's instructions. MAA recognizes terminal sialic acid α2-3 linked to galactose, PNA recognizes terminal galactose β1-3 linked to N-acetylgalactosamine, and DSA recognizes GlcNAc present in both N- and O-linked glycans. The contrast on digital images was manipulated using Adobe Photoshop Elements 2.0.

Adherence of S. pneumoniae to human epithelial cells.

The contribution of SP0368 to pneumococcal adherence was tested essentially as previously described (19, 24). Adherence of the parental strain and eng mutants to Detroit 562 cells (ATTC CCL-138), a human pharyngeal epithelial carcinoma cell line, was assessed. To more accurately mimic the interaction of bacteria with the epithelial cell layer, the protocol was altered so that bacterial cells were not centrifuged onto the epithelial cells. Adherence after 60 min of incubation was expressed as a percentage of that of the parental strain, 1121 Sm^r, under the same experimental conditions. Where indicated, 0.0000125 or 0.000625 U of recombinant O-glycosidase per 1.9 cm² of well surface area was added with the bacterial inoculum. Three independent experiments performed in triplicate were used for statistical analysis.

Mouse model of pneumococcal colonization.

Nasopharyngeal colonization was performed essentially as described previously (31). For strains 1121 Sm^r, 1121 Δeng, and 1121 ΔnanA Δeng, 20 6- to 8-week-old C57BL/6 mice (Jackson Laboratory) were inoculated intranasally with 2 × 10⁷ mid-log-phase organisms. The density of colonization was assessed by upper respiratory tract lavage and quantitative culture for 10 mice at 36 h and the remaining mice at day 5. The animal data are presented as means (CFU ml⁻¹) ± standard errors of the means.

Statistical analysis.

O-glycosidase activities of mutants and results of adherence assays were compared by two-tailed Student's t tests. Analysis of variance was used to assess differences in activity between recent clinical isolates. Differences in murine colonization were assessed by one-tailed Student's t tests.

RESULTS

Analysis of the pneumococcal genome identified a putative O-glycosidase.

S. pneumoniae is known to express O-glycosidase activity (4, 15, 41), as it was previously detected in culture supernatants (15); however, the majority of other pneumococcal glycosidases are predicted to be surface associated in a sortase-dependent manner (40). To narrow the number of O-glycosidase candidate genes, we investigated whether the S. pneumoniae protein is surface associated and if sortase A (SrtA) is required for surface localization. The O-glycosidase activities associated with the whole culture, cells, and supernatant of TIGR4 and TIGR4 ΔsrtA were determined. While the majority of the O-glycosidase activity was cell associated in the parental strain, mutation of srtA resulted in the majority of the activity being secreted (Fig. 2). These data demonstrate that the O-glycosidase, like the majority of other pneumococcal glycosidases, is surface associated in a sortase-dependent manner.

FIG. 2.

O-glycosidase activity is cell associated in a sortase-dependent manner. O-glycosidase activity was compared in TIGR4 and a sortase mutant (TIGR4 ΔsrtA). Strains were grown to stationary phase and, when noted, centrifuged to separate cells from supernatants prior to toluene treatment. Reaction mixtures were incubated for 1 h with the colorimetric substrate Galβ1-3GalNacα1pNP. Reactions were stopped with 1 M sodium carbonate, and activity was measured by recording the optical density at 400 nm. Values are the means of three independent experiments ± standard deviations. *, significant difference between TIGR4 total and supernatant levels (P < 0.0001); **, significant difference between TIGR4 ΔsrtA total and cell levels (P < 0.0002).

There are 19 predicted sortase-dependent cell-associated proteins in the TIGR4 genome (40). The basic local alignment search tool (BLAST) (http://www.ncbi.nlm.nih.gov/blast/Blast.cgi) demonstrated that one of the open reading frames predicted to encode a sortase-dependent cell-associated protein, SP0368, shares a high level of sequence similarity to engBF of Bifidobacterium longum (38% sequence identity and 54% sequence similarity over 1,749 amino acids), which encodes a protein with core-1-specific O-glycosidase activity (17). SP0368 is predicted to encode a 137.8-kDa protein, which in addition to including a predicted secretion signal (amino acids 1 to 38) and an LPXTG motif is predicted to contain a family 32 carbohydrate binding module (amino acids 1497 to 1601) (29). This locus is distal to but predicted to be in the same transcriptional unit as that for the polypeptide transporter AliA (Fig. 3A) (40).

FIG. 3.

SP0368 encodes the pneumococcal O-glycosidase (Eng). (A) Schematic representation of the genetic region surrounding the putative O-glycosidase gene, SP0368. Open reading frames from the TIGR4 sequence are represented by block arrows. Arrows above the schematic indicate predicted transcription start sites. (B) O-glycosidase activity in 1121 Δeng is significantly reduced compared to that in 1121 Sm^r. Strains were grown to stationary phase, lysed with toluene, and incubated for 1 h with the colorimetric substrate Galβ1-3GalNacα1pNP. The reaction was stopped with 1 M sodium carbonate, and activity was measured by recording the optical density at 400 nm. Values are the means of three independent experiments ± standard deviations. *, significant difference between 1121 Sm^r and 1121 Δeng (P < 0.002).

SP0368 encodes an S. pneumoniae O-glycosidase.

To determine if SP0368 encodes an O-glycosidase, a nonpolar unmarked deletion was constructed in the 1121 background. Successful construction of the mutant was confirmed by both PCR and sequencing. Strain 1121 Δeng was shown to lack O-glycosidase activity by use of the colorimetric substrate Galβ1-3GalNacα1pNP (Fig. 3B). SP0368 was previously proposed to be essential for in vitro growth of TIGR4 (34). To investigate this possibility, we constructed an SP0368 mutant in TIGR4, and the resulting strain had no detectable O-glycosidase activity (data not shown). Together, these data suggest that SP0368 is not essential for in vitro growth and that this gene encodes the only pneumococcal O-glycosidase detectable under these assay conditions.

All 29 pneumococcal genome sequences available contain a homologue of SP0368. The predicted amino acid sequence of SP0368 contains low levels of sequence diversity of up to 3.4%. These sequence changes are not evenly distributed throughout the predicted sequence, suggesting that the locus may have evolved by recombination. In addition, one strain contains a 15-bp duplication at nucleotide 367, which results in a five-amino-acid insertion. Although there were statistically significant differences in O-glycosidase activity among different recent clinical isolates, the five isolates tested all contained SP0368 homologues and expressed O-glycosidase activity (Fig. 4 and data not shown). Deletion of the SP0368 homologue in two of these recent clinical isolates (C06_31 and C06_57) resulted in the absence of detectable O-glycosidase activity (data not shown). These data further support the hypothesis that all pneumococcal strains encode an O-glycosidase and that SP0368 encodes the only pneumococcal O-glycosidase activity.

FIG. 4.

Levels of O-glycosidase activity differ between strains. Five recent clinical isolates of different genetic backgrounds and serotypes were used to investigate differences in enzyme activity between strains. 1121 Sm^r and 1121 Δeng were used as controls. Strains were grown to an optical density at 600 nm of 0.6, lysed with toluene, and incubated for 1 h with the colorimetric substrate Galβ1-3GalNacα1pNP. The reaction was stopped with 1 M sodium carbonate, and activity was measured by determining the optical density at 400 nm. Values are the means of three independent experiments ± standard deviations. Analysis of variance demonstrated that the O-glycosidase activities between the recent clinical isolates were significantly different (P < 0.001).

NanA and Eng sequentially deglycosylate O-linked glycans.

Our previously published work demonstrated that S. pneumoniae sequentially deglycosylates N-linked glycans on human proteins (24). We used the model glycoconjugate fetuin to determine if S. pneumoniae can sequentially deglycosylate O-linked glycans, which are also common in the airway. Fetuin is a well-characterized serum protein that has an average of three N-linked glycans and three O-linked glycan chains per molecule (35-38). The majority of the O-linked glycan chains on fetuin are the sialylated core-1 structure (Fig. 1) (44), which is also present on many proteins in the human airway, including IgA.

The extent of deglycosylation was determined using the lectins MAA and PNA, which detect terminal sialic acid α2-3 linked to galactose and terminal Galβ1-3GalNAc, respectively (Fig. 5). Incubation of fetuin with the parental strain, 1121 Sm^r, resulted in removal of all detectable O-linked glycans, as determined by the absence of binding of both MAA and PNA. As predicted, terminal sialic acid was detected following incubation of fetuin with medium alone. The presence of the protein in all lanes was confirmed by a third lectin, DSA, which detected the GlcNAc present in the N-linked glycan structures of fetuin (10; data not shown).

FIG. 5.

Neuraminidase and O-glycosidase sequentially deglycosylate fetuin. Medium alone (M) or stationary-phase cultures of 1121 Sm^r, 1121 Δeng, 1121 Δeng/eng⁺ (eng⁺), and 1121 ΔnanA Δeng were incubated overnight with 0.2 μg of fetuin at 37°C. Where indicated, 0.0008 unit of purified Clostridium perfringens neuraminidase (+N) or 0.00016 unit of recombinant S. pneumoniae O-glycosidase (+O) was added. Following incubation, samples were resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (12.5%), transferred to a membrane, and detected by the lectins MAA (A) and PNA (B), which detect α2-3-linked sialic acid and Galβ1-3GalNAc, respectively.

We predicted that deglycosylation of O-linked glycans is dependent on the activities of the neuraminidase NanA and the O-glycosidase encoded by SP0368. The Eng mutant cleaved only the terminal sialic acid to expose Galβ1-3GalNAc, as detected by the binding of PNA. Cleavage of the disaccharide could be restored both by addition of pneumococcal O-glycosidase and by genetically complementing the mutant (1121 Δeng/eng⁺). The neuraminidase NanA has previously been shown to contribute to cleavage of terminal sialic acid from N-linked glycans. To confirm a role for NanA in deglycosylation of O-linked glycans, a double mutant, 1121 Δeng ΔnanA, was shown to be unable to modify O-linked glycans, as determined by the detection of terminal sialic acid. The addition of purified neuraminidase resulted in the exposure of Galβ1-3GalNAc, while the addition of both neuraminidase and O-glycosidase resulted in cleavage of the entire glycan structure. These data demonstrate that S. pneumoniae can sequentially deglycosylate O-linked glycans and that NanA and Eng are essential for this activity in this model system.

Eng contributes to pneumococcal adherence and colonization.

In order to investigate the in vivo contribution of the O-glycosidase, 1121 Δeng, 1121 Δeng ΔnanA, and 1121 Sm^r were compared in a mouse model of nasopharyngeal colonization. We hypothesized that the O-glycosidase mutant would be reduced significantly in adherence. Significant reductions in the level of upper respiratory tract colonization (P ≤ 0.05) between 1121 Δeng (2.10 × 10³ ± 0.90 × 10³) and 1121 Sm^r (2.52 × 10⁴ ± 1.34 × 10⁴) and between 1121 Δeng ΔnanA (2.34 × 10³ ± 1.3 × 10³) and 1121 Sm^r (2.52 × 10⁴ ± 1.34 × 10⁴) were observed at 36 h postinoculation. There was no significant difference in the level of upper respiratory tract colonization between 1121 Δeng and 1121 Δeng ΔnanA. By 5 days postinoculation, there was no significant difference in the ability of the three strains to colonize the upper respiratory tract of mice (1121 Δeng level, 3.30 × 10³ ± 1.04 × 10³; 1121 Δeng ΔnanA level, 6.60 × 10³ ± 2.62 × 10³; and 1121 Sm^r level, 1.00 × 10⁴ ± 0.21 × 10⁴). These data suggest that the O-glycosidase may contribute to establishment of colonization.

The ability of S. pneumoniae to modify O-linked glycans may contribute to pneumococcal pathogenesis in multiple ways. Adherence to the epithelial cell surface is likely essential for efficient pneumococcal colonization. A significant reduction in adherence of 1121 Δeng to upper airway epithelial cells (D562) was observed compared to that of the parental strain (Fig. 6). Genetic complementation of the mutation (1121 Δeng/eng⁺) restored parental levels of adherence. A neuraminidase mutant of 1121 did not show any reduction in adherence to D562 cells, suggesting that cleavage of sialic acid is not required for the role of Eng in pneumococcal adherence (data not shown). The adherence of the O-glycosidase mutant was not complemented by the addition of recombinant enzyme, despite the use of enzyme in excess of that produced by the number of bacteria included in the assay (data not shown).

FIG. 6.

Relative adherence of an eng mutant to human epithelial cells is reduced. 1121 Sm^r, 1121 Δeng, and 1121 Δeng/eng⁺ cells were grown to an optical density at 600 nm of 0.6, and adherence to D562 cells was determined. Mutant strain adherence is expressed as a percentage of parental strain adherence over a 60-min incubation period under the same conditions. Values are the means of three independent experiments ± standard deviations. *, significant difference between 1121 Δeng and 1121 Sm^r (P < 0.02).

DISCUSSION

The data presented here demonstrate that SP0368 encodes an O-glycosidase and support the hypothesis that this enzyme is a member of the recently identified glycoside hydrolase family 101 (17). Recently, Caines et al. published the structure of the protein encoded by SP0368 and independently demonstrated that this recombinant protein possesses O-glycosidase activity (7). In this report, we explore the biological relevance of this enzymatic activity. Our in vitro deglycosylation studies demonstrate that this enzyme, in conjunction with NanA, deglycosylates a model glycoconjugate. In addition, we identified the first biological function of this protein, demonstrating that it contributes to pneumococcal colonization of the upper respiratory tract and to adherence to human epithelial cells.

Pneumococcal O-glycosidase activity was previously proposed to be secreted (15). Our studies demonstrate that Eng is surface associated in a sortase-dependent manner and suggest that the activity detected in culture supernatants was likely due to enzyme released from the surface during cell wall turnover. This mechanism of surface attachment is utilized by the majority of pneumococcal glycosidases, suggesting that the surface localization of these enzymes may be important for their contribution to pathogenesis.

A mutagenesis screen of pneumococcal open reading frames in TIGR4 suggested that SP0368 was essential for growth in vitro (34). Song et al. (34) proposed that genes which could not be mutated in three attempts were essential; however, this designation was not confirmed. We successfully deleted this locus in TIGR4 and three other genetic backgrounds. The construction of a mutant allowed us to extend previous observations by demonstrating that SP0368 encoded the only detectable O-glycosidase expressed by several pneumococcal strains. The construction of an O-glycosidase mutant also allows us to investigate the contribution of this enzyme to pathogenesis.

Published data have demonstrated that the protein encoded by SP0368 is expressed in vivo, as convalescent-phase sera contain antibodies that react with the protein and immunization with the protein provides significant protection against intravenous challenge in mouse models of infection (18). All available genome sequences contain this gene, and all strains tested expressed detectable O-glycosidase activity, suggesting that this enzyme activity is common to all S. pneumoniae strains. Giefing et al. reported that SP0368 was absent in approximately 8% of isolates screened (18). This disparity may be explained by the sequence diversity within the gene, resulting in a failure to detect the gene by PCR.

Our findings demonstrate that Eng and NanA are required to cleave sialylated core-1 O-linked glycans from glycoconjugates. This expands our previous finding that S. pneumoniae can sequentially deglycosylate N-linked glycans and demonstrates that pneumococcal glycosidases can also sequentially deglycosylate O-linked glycans (24). Recently published data support a role for NanA in modification of mucin, which is heavily decorated with O-linked glycans (45). It was proposed that cleavage of sialic acid from mucin supports the growth of S. pneumoniae, although since mucin contains both N- and O-linked glycans it was unclear which structures were being cleaved by NanA. In this study, we demonstrate that NanA cleaves terminal sialic acid from O-linked glycans.

We recently reported that under some conditions, a second pneumococcal neuraminidase, NanB, contributes to cleavage of terminal sialic acid from N-linked glycans (6). We cannot rule out a role for NanB in modification of O-linked glycans, although there was no detectable cleavage of sialic acid by the NanA mutant in this study. In addition, recently published data suggest that the main function of NanB may be as an intramolecular trans-sialidase (20). This study utilized a simple and well-characterized O-linked glycan. It is possible that other pneumococcal glycosidases could contribute to the deglycosylation of more complex structures.

Our demonstration that O-glycosidase mutants were reduced in the ability to colonize the upper respiratory tract of mice suggests that the degradation of O-linked glycans may contribute to pneumococcal pathogenesis. A significant reduction in colonization was observed at 36 h postinoculation but not at 5 days postinoculation. These data imply that the enzyme contributes to establishment of colonization, although further studies would be required to investigate this possibility.

Pneumococcal modification of N-linked glycans is proposed to alter the clearance function of host defense molecules, expose binding sites on the epithelial surface, and contribute to bacterial growth (6, 24). Modification of host O-linked glycans may also contribute to pneumococcal pathogenesis by multiple mechanisms. Published data imply that S. pneumoniae binds to host cell glycoconjugates, including GalNAcβ1-4Gal, found in gangliosides GM1 and GM2 (26), the glycolipid globoside (GalNAcβ1-3Galα1-4Galβ1-4GlcCer) (12), and the naturally occurring milk oligosaccharide lacto-N-neotetraose (1). If S. pneumoniae utilizes glycans as receptors on epithelial cells, it is possible that glycosidases act to expose these receptors. This possibility is strengthened by the previous finding that the pneumococcal glycosidases NanA (a neuraminidase) and BgaA (a β-galactosidase) play a role in adherence of some strains (24).

The significant reduction of the O-glycosidase mutant in adherence to human epithelial cells suggests that the degradation of O-linked glycans may contribute to pneumococcal adherence. The residual adherence of the Eng mutant is presumably a result of other adherence mechanisms, perhaps including those identified previously (11, 47). Given the critical nature of adherence to colonization, there are likely several mechanisms, each of which will have to be elucidated individually (21). The mechanism by which the O-glycosidase contributes to adherence is unclear. The absence of a role for neuraminidase in adherence of 1121 suggests that sequential deglycosylation of O-linked glycans does not contribute to adherence of this strain. 1121 Δeng could not be complemented by the addition of purified enzyme. This result may be due to the level of O-glycosidase activity under the conditions utilized or our inability to add sufficient enzyme. Alternatively, these data may suggest that the role of O-glycosidase in adherence is independent of its enzymatic activity.

In summary, we have identified the gene encoding the pneumococcal O-glycosidase. This enzyme, in conjunction with the neuraminidase NanA, sequentially deglycosylates O-linked glycan structures found in humans. These findings highlight the observation that modification of both N- and O-linked glycans likely contributes to pneumococcal colonization. Furthermore, we identified the first putative function for this enzyme in pneumococci. Given the wide distribution of O-linked glycans, further investigation will likely reveal multiple functions for bacterial modification of these structures in pneumococcal pathogenesis.