Identification of a Novel Keto Sugar Component in Streptococcus pneumoniae Serotype 12F Capsular Polysaccharide and Impact on Vaccine Immunogenicity

Streptococcus pneumoniae is a Gram-positive encapsulated bacterium that is a common colonizer of the human respiratory tract as well as a cause of both noninvasive and invasive pneumococcal disease (IPD). Children <5 y of age (1, 2) and the elderly (3) are the most vulnerable populations affected by S. pneumoniae. In the 1990s, before the introduction of the first pneumococcal capsular polysaccharide conjugate vaccine (PCV7) into the childhood immunization program, the annual rates of S. pneumoniae–caused IPD among children <5 y of age were estimated to be almost 100 cases per 100,000 (4) in the United States. The bacterium also caused almost 5,000,000 cases of acute otitis media per year among children <5 y of age (5, 6).

Although >90 serotypes of S. pneumoniae have been described to date, most pneumococcal disease in humans is commonly caused by <30 serotypes (6, 7). The bacterial capsule defines pneumococcal serotypes, and the serotype diversity is caused by variation in the chemical structure (e.g., oligosaccharide units or attached side groups) of capsular polysaccharides (7).

The capsular polysaccharide is also an important pneumococcal virulence factor and is associated with helping the bacteria survive in the bloodstream. Immunization with vaccines targeting the capsular polysaccharides has been effective at preventing disease (7, 8) by eliciting protective Abs with opsonophagocytic activity (OPA) that facilitate complement-mediated uptake and killing of the pneumococcus by human phagocytic effector cells such as neutrophils (9–11).

PCV20 (12) contains the 13 polysaccharide conjugates included in PCV13 as well as polysaccharide conjugates for seven additional serotypes (8, 10A, 11A, 12F, 15B, 22F, and 33F) (13). Serotype 12F was selected for inclusion as one of the seven additional serotypes in PCV20 based on its relative prevalence as a cause of IPD in both adult and pediatric populations and its generalized geographic distribution. In addition, serotype 12F is associated with outbreaks of IPD (14–16).

Although polysaccharide structures have been previously established for all known S. pneumoniae serotypes (7), the polysaccharide structure of serotype 12F was reanalyzed using state-of-the-art high-resolution and -sensitivity nuclear magnetic resonance (NMR) and high-resolution mass spectrometry (MS) approaches, and a novel structural feature in the 12F polysaccharide was identified. This involved partial replacement of a repeat unit (RU) N-acetyl-galactosamine sugar residue with an atypical sugar residue, 2-acetamido-2,6-dideoxy-xylo-hexos-4-ulose (Sug). Sug incorporation in the 12F RU structure was confirmed to be ubiquitous and variable in incorporation level across a series of clinical S. pneumoniae 12F isolates. Using the opsonophagocytic killing assay, an accepted indication of functional immune responses against pneumococci in adults, the impact of 12F Sug incorporation on elicitation of immune responses was evaluated.

The S. pneumoniae serotype 12F strain was obtained from the American Type Culture Collection (ATCC 6312). The S. pneumoniae serotype 12F strain PFESP00919 used in producing the specific polysaccharide in PCV20 was obtained from the laboratory of Gerald Schiffman. Various serotype 12F clinical isolates from the Pfizer-sponsored TEST and ATLAS surveillance programs were curated by International Health Management Associates.

Purified and lyophilized S. pneumoniae serotype 12F polysaccharide was purchased from the American Type Culture Collection (ATCC 196-X), which was deposited by Merck and Company and is representative of the type used as a component of the unconjugated polysaccharide vaccine PPSV23. Purified and lyophilized S. pneumoniae cell wall polysaccharide (C-Poly) was purchased from Cedarlane (supplied by Statens Serum Institut, København, Denmark).

Culture stocks of S. pneumoniae strains were prepared by growing to late exponential phase in a soy hydrolysate-glucose–based medium and subsequently cryopreserved at −70°C after adding glycerol to a 20% final concentration. Production of 12F polysaccharide at scales of 0.25–2000 l was enabled by fermentation of S. pneumoniae strains in a soy hydrolysate-glucose–based medium.

Starter seed cultures were cultured from frozen cell suspensions at designated fermentation temperature without shaking to late logarithmic phase and then used to inoculate bioreactors, which were operated at designated fermentation parameters. Fermentations were grown to early stationary phase and were terminated by adding N-lauryl sulfate to a final concentration of 0.1%.

The N-lauryl sulfate lysate was clarified by flocculation followed by acidification. The precipitate that formed was removed by centrifugation and depth filtration. The purified polysaccharide was obtained through various filtration steps.

Native 12F polysaccharide was hydrolyzed with acetic acidic at 70°C to lower polysaccharide molecular mass. After hydrolysis, polysaccharide was purified by diafiltration. Hydrolyzed polysaccharide was oxidized using 2,2,6,6-tetramethylpiperidinyloxy/N-chlorosuccinimide (TEMPO/NCS) to oxidize primary alcohols (17) and purified by diafiltration. Oxidized polysaccharide was conjugated with the cross-reacting material 197 (CRM₁₉₇) carrier protein using aqueous reductive amination chemistry with sodium cyanoborohydride, and unreacted aldehydes were capped with sodium borohydride. After capping, conjugate was purified by diafiltration.

For NMR structure elucidation, 12F polysaccharide in water was treated with tip sonication to reduce the average polymer length using a Bransonic 450 sonicator at ∼10–25% amplitude using a 0.5-inch titanium tip for up to 90 min over an ice bath. Samples were filtered with a 0.22-µm filter, and in some cases were size separated serially using 50- and 10-kDa molecular mass cutoff (MMCO) spin columns. Sonicated samples with or without size separation were dialyzed using 3-kDa MMCO dialysis cassettes against water, frozen, lyophilized, and redissolved in D₂O up to ∼135 mg/ml with ∼0.55 mM trimethylsilyl propanoic acid-d4 (TSP-d4) at pH ∼6.0–7.0. PFESP00919 12F polysaccharide was reduced by dissolving in water (2.5 mg/ml) at pH 7.0, adding ∼11.5 mEq NaBH₄ from solution, and mixing at 150 rpm overnight at 23°C, followed by dialysis against water using 7-kDa MMCO dialysis cassette, then sample freezing, lyophilization, and redissolution in D₂O. NMR data were collected at 75°C at 500 MHz using a Bruker 5-mm DCH cryoprobe and a Bruker BioSpin Avance III console, and at 600 MHz using a Bruker 5-mm TCI cryoprobe and a Bruker BioSpin Avance Neo console. Standard Bruker pulse sequences were used to collect one-dimensional (1D) ¹H and ¹³C, and two-dimensional (2D) correlation spectroscopy (COSY), nuclear Overhauser effect spectroscopy (NOESY), heteronuclear single-quantum correlation spectroscopy (HSQC), HSQC–total correlation spectroscopy (TOCSY), and heteronuclear multiple-bond correlation spectroscopy (HMBC) spectra. The data processing was conducted using M-Nova v12.0 and NMRViewJ (18). The chemical shift reference was TSP-d4 at 0 and −1.8 ppm ¹H and ¹³C, respectively.

1D ¹H NMR relative quantification of the mole percent of Sug in the 12F polysaccharide RU was carried out by dissolving lyophilized PFESP00919 12F polysaccharide to ∼5 mg/ml using DMSO-d6, 5% (v/v) D₂O, and 50 mM maleic acid with stirring overnight in a glass vial at 25°C, followed by transfer to a 5-mm NMR tube and 1D ¹H analysis at 75°C. Spectra were processed with Bruker TopSpin v3.5 using 0.5-Hz EM broadening, automatic phasing with manual adjustment as needed, fifth-order polynomial baseline correction, and chemical shift referencing to residual protonated DMSO at 2.5 ppm. The 12F RU mole percent of Sug was quantified as the percent relative area of the Sug H6 signal relative to the pyranose residue 2-acetamido-2,6-dideoxy-l-galactose (FucpNAc) H6 signal.

2D HSQC NMR relative quantification of the mole percent of Sug in 12F polysaccharide RU from different 12F strains was carried out by dissolving purified and lyophilized polysaccharide to ∼20 mg/ml in D₂O, followed by 10 min of tip sonication at 20% amplitude using a Bransonic 450 sonicator, then filtration using a 0.22-μm filter. 2D ¹H-¹³C HSQC was carried out at 75°C using the “hsqcetgp” Bruker pulse sequence. Typical pulse parameters include the following: NS = 32 for an ∼3-h experiment, f1 = 256, ¹³C SW = 20 ppm, ¹³C O1 = 15 ppm, cnst2 optimized at 130 Hz for backbone methyl ¹J_CH coupling, and D1 = 1.5 s. Data were processed and visualized using NMRViewJ (18). The level of Sug incorporation in the 12F RU was measured as the percent of the ratio of Sug methyl cross-peak intensity over the FucpNAc methyl cross-peak intensity.

Liquid chromatography (LC)–MS and LC–tandem MS (MS/MS) data were collected in a positive ion mode on a Thermo Scientific Orbitrap Fusion Lumos Tribrid mass spectrometer equipped with an Agilent 1260 HPLC system. Samples were injected and separated on a Waters hydrophilic interaction chromatography (HILIC) BEH spherical hybrid column (1.7 µm, 2.1 × 150 mm; part 186004742) held at 60°C. Mobile phase A (MPA) is water with 0.1% formic acid, and mobile phase B is acetonitrile with 0.1% formic acid. The elution gradient was delivered at 200 μl/min; 10–95% MPA in 35 min, returned to 10% MPA in 1 min, and equilibrated for 15 min. The radio frequency entrance lens voltage in the ion source was increased to 60 V (typically) to allow efficient desolvation and transmission of the high mass ions (of possibly multiply charged intact polysaccharide ions) for subsequent in-source fragmentation. The source fragmentation (in-source CID) voltage was then increased to 80 eV to induce cleavage of the glycosidic bonds. This combination of settings appeared to be critical for comprehensive fragmentation of the polysaccharide RU. LC-MS and LC-MS/MS data from the reduction of 12F polysaccharide with NaBD₄ were collected in a positive ion mode on a Thermo Scientific Orbitrap Q Exactive mass spectrometer equipped with an Agilent 1260 HPLC system. Samples were injected and separated on a HILIC BEH spherical hybrid column (1.7 µm, 2.1 × 150 mm; part number 18600474) held at 60°C. MPA is water with 0.1% trifluoroacetic acid, and mobile phase B is acetonitrile with 0.1% trifluoroacetic acid. The elution gradient was delivered at 200 μl/min; 30–70% MPA in 35 min, returned to 30% MPA in 1 min, and equilibrated for 14 min. The radio frequency entrance lens voltages were increased to 60 V and in-source fragmentation was set to 80 eV to induce in-source fragmentation of polysaccharide. The LC-MS/MS data were acquired using higher energy C trap dissociation (HCD) with stepped collision energies of 20, 30, and 40.

Pneumococcal OPA assays were developed for S. pneumoniae serotype 12F isolates similar to those that have been described previously (19–22). The interpolated OPA Ab titer is expressed as the reciprocal of the serum dilution resulting in a 50% reduction in the number of bacterial colonies when compared with the control wells that did not contain serum. Sera from subjects immunized with a vaccine generated from conjugated 12F polysaccharide (PCV20 [n = 84]), a vaccine generated from nonconjugated 12F polysaccharide (PPSV23 [n = 28]), or a vaccine that does not contain a 12F polysaccharide conjugate (PCV13 [n = 84]) were assessed for OPA against S. pneumoniae serotype 12F isolates.

S. pneumoniae bacteria were streaked from frozen cell banks to TSA II blood agar plates and grown overnight at 37°C in 5% CO₂. Colony morphology on the following day was characteristic of S. pneumoniae, and visual inspection did not reveal any evidence of contaminating microorganisms. A 10-μl inoculation loop was used to harvest bacterial growth to seed 300-μl THY media for expansion in liquid for 4–5 h at 37°C in 5% CO₂. A 200-μl aliquot of the cell suspension was removed to a 96-well plate and heated 5 min at 95°C. After cooling to 4°C, genomic DNA was extracted using magnetic bead technology according to the manufacturer’s instructions with minor modifications (Agencourt GenFind v2 system; Beckman Coulter, Brea, CA).

Whole-genome sequencing (WGS) and data analysis were performed as described (23). Briefly, WGS was performed using a MiSeq desktop sequencer (Illumina, San Diego, CA) with 2 × 300 bp paired-end sequencing chemistry. Preparation of the DNA library for bacterial WGS was done according to the Nextera XT DNA sample kit preparation protocol (Illumina, San Diego, CA). The DNA libraries generated from the bacterial strains were pooled and loaded onto the MiSeq instrument as instructed in the manufacturer’s protocol. Primary DNA sequence reads were run through the “Merge Overlapping Pairs” program followed by assembly into contigs using the “De Novo Assembly” program in the Qiagen CLC Genomics Workbench (Qiagen, Redwood City, CA) using default parameters. The assembled sequence contigs were entered into BIGSdb (24). Analyses included determination of nucleotide sequence of aroE, ddl, gdh, gki, recP, spi, and xpt from the WGS of each strain; these sequences were used to assign a sequence type at PubMLST (http://pubmlst.org/). Detailed comparison with the capsular polysaccharide biosynthetic operon (cps) of the ATCC 6312 reference isolate (GenBank accession number CR931660 [https://www.ncbi.nlm.nih.gov/nuccore/CR931660]) was also performed. Primary nucleotide sequence data for the S. pneumoniae isolates from this study have been deposited to the National Center for Biotechnology Information Sequence Read Archive (BioProject accession number PRJNA927336).

Previous published structure analysis of the S. pneumoniae 12F capsular polysaccharide employed methylation, specific degradations, and limited low-resolution and low-sensitivity 1D-NMR spectroscopy. The macromolecule was described as hexasaccharide RUs containing pyranose residues d-glucose (Glcp), d-galactose (Galp), 2-acetamido-2-deoxy-d-galactose (GalpNAc), FucpNAc, and 2-acetamido-2-deoxy-d-mannuronic acid (ManpNAcA) in the proportions 2:1:1:1:1 (Fig. 1A) (25). During the development of PCV20, the strain PFESP00919 12F capsular polysaccharide structure was analyzed using contemporary multidimensional NMR (2D-NMR) spectroscopy instrumentation with high sensitivity and resolution, coupled with polysaccharide molecule size reduction to optimize line shape and nuclear spin relaxation to achieve complete and unambiguous assignment of the 12F RU. In addition, modern high-resolution MS was applied to elucidate 12F carbohydrate structure within the RU and between the RUs as explained below. Taken together, 2D-NMR and MS data provide complementary and conclusive evidence for a heterogeneous 12F RU structure, revealing that ∼75% of the PFESP00919 12F polysaccharide RUs are characterized by the sugars and linkages described by Leontein et al. (25) as illustrated in Fig. 1A, along with ∼25% of RUs characterized by replacement of backbone GalpNAc with an atypical keto sugar Sug (refer to Fig. 1B). To our knowledge, this novel finding both confirms and updates the previous structure (25) as well as highlights the central role that high-sensitivity, high-resolution techniques such as NMR and MS play in the identification and characterization of polysaccharide structural heterogeneity. Several recent examples from three different S. pneumoniae serogroups: 7 (26), 11 (27), and 6 (28), further demonstrate that state-of-the-art characterization techniques should be used to confirm polysaccharide structure and potential heterogeneity.

The 12F polysaccharide is heterologous (Fig. 1B), being comprised of a primary spin system (∼75% of RUs) specified by backbone incorporation of GalpNAc, and a secondary spin system whereby the keto sugar Sug replaces GalpNAc (∼25% of RUs). The 1D ¹H proton spectrum of size-reduced PFESP00919 12F polysaccharide is shown in Fig. 2. Complete ¹H and ¹³C chemical shift assignment is listed in Table I for the primary spin system. Table II lists the novel and resolved chemical shifts for the secondary spin system due to Sug replacement of GalpNAc. The absolute configuration of FucpNAc, Galp, GalpNAc, ManpNAcA, and Glcp residues was previously determined (25), and the complete ¹H/¹³C NMR chemical shift data in Tables I and II confirm that all residues are in pyranose forms based on the absence of low-field nonanomeric ring carbon shifts that are not substituted. NMR analyses of the PFESP00919 12F polysaccharide primary spin system, including anomeric ¹H and ¹³C resonance assignment and relevant coupling constant measurement (¹J_C1,H1 and ³J_H1,H2), corroborate anomeric configuration assignment (29) in the primary spin system of 12F previously determined using a degradation and derivatization approach (25). Homonuclear and heteronuclear correlations used to confirm assignment of primary and secondary 12F RU spin systems are provided in Supplemental Tables I and II. To our knowledge, this is the first NMR-based structure of S. pneumoniae serotype 12F capsular polysaccharide.

To our knowledge, a novel high-resolution/accurate mass LC-MS method was applied to characterize the 12F polysaccharide that confirms the RU, branching structure, and component sugars. The 12F polysaccharide eluted as a single peak via HILIC. In-source fragmentation was used to produce a surrogate pseudomolecular ion at m/z 1094.3890 representative of the intact 12F polysaccharide RU (Fig. 3), as well as the corresponding fragment ions that help elucidate the branching pattern and sequence of the sugars based on the unique compositional information from accurate masses. The fragment ions primarily consisted of single charged ions in the arrangements of monosaccharides, disaccharides, trisaccharides, tetrasaccharides, and pentasaccharides. The RU sugars were identified using the letter designation indicated in Fig. 1B: residues A, B, and E form the backbone chain, whereas residues C, D, and F represent the branched residues. There appears to be preferential fragmentation between A and E (e.g., /ABE/ABE/ABE/), thus forming the ions associated with the RU backbone. There were some low-abundant ions, EA (m/z 391) and EFA (m/z 553), which are formed from cleavage between B and E, as well as the low-abundant fragment ions ABEA, ABEFA, and ABCEFA, which arise from cleavage between A and B. Cross-ring fragmentation via carbon–carbon bond cleavages was not observed. Furthermore, both galactose and glucose are hexose residues (162 Da) with the same elemental compositions, and therefore they cannot be differentiated based on accurate mass alone. All experimental accurate mass determinations matched theoretical values to 1 ppm or less. Major fragment ions (i.e., disaccharides, trisaccharides, tetrasaccharides, and pentasaccharides) were subjected to subsequent MS/MS with precursor ion selection and HCD to yield smaller, more specific multiresidue fragment ions and individual monosaccharide ions to confirm the respective fragment ion sugar assignments in Fig. 3 in terms of compositions and branching patterns (data not shown). The HCD mass spectrum of m/z 1094 via LC-MS/MS (data not shown) essentially provided the same fragment ion mass-intensity data as shown by LC-MS in Fig. 3.

The fragment ion at m/z 1076.3785 appears to be associated with the RU in which GalpNAc was replaced with Sug residue, but we acknowledge that the accurate mass difference between 1094 (with GalpNAc) and 1076 (with Sug) also represents a nominal water loss. The ion, m/z 1076, was further fragmented by LC-MS/MS using HCD, and produced different intensities of the fragment ions at m/z 186 and 168 that are associated with Sug residue and loss of water from GalpNAc, respectively (data not shown). Separation of the GalpNAc- and Sug-containing polysaccharides by HILIC and other modes of chromatography was not possible, indicating that the Sug residue is systematically incorporated into the 12F polysaccharide structure. Additional doubly charged species representing 2 RUs were detected, including (2RU)²⁺ at m/z 1094.89 and (2RU-hexose)²⁺ at m/z 1013.36, both with relative abundances of 8%. Both of these species contain a −18 Da fragment ion consistent with the incorporation of one Sug residue in a 2 RU species. However, no respective –36 Da fragment ions were detected, which also indicates a lack of evidence for the presence of two distinct polysaccharide chains where one chain would contain backbone Sug in all RUs and the second chain would contain backbone GalpNAc in all RUs.

To our knowledge, the new MS methodology provided separate synergistic orthogonal 12F structural elucidation and corroborates the sequence and organization of the sugars in the RU determined by NMR. This in-source fragmentation LC-MS and LC-MS/MS methodology has been applied to other polysaccharides from different bacterial strains with similar results and has also been useful to elucidate the structure of activated polysaccharides that have been purposely modified for the conjugation to protein carriers such as CRM₁₉₇ (data not shown).

Based on the newly understood 12F structural heterogeneity in PFESP00919, we examined the distribution of the Sug moiety in contemporary S. pneumoniae 12F isolates and explored the potential relevance of Sug incorporation on 12F strain susceptibility to killing in an OPA assay. To assess whether Sug was incorporated at a consistent percentage in the PFESP00919 production strain during fermentation and whether it is also broadly represented in circulating 12F pneumococcal strains, multiple fermentation batches of the PFESP00919 12F strain (Fig. 4) and a range of clinical 12F isolates were analyzed by NMR (Table III). Sug substitution in 12F polysaccharide from the PFESP00919 12F strain was evaluated in 18 different fermentation batches. As shown in Fig. 4, the percentage of Sug incorporation was consistent across the scale of fermentation batches, ranging from 0.25 to 2000 l, and consistent between all research, pilot, and clinical batches tested. In cases where triplicate analysis was performed, the relative SD was typically <0.5%. This analysis indicates that the ∼25% Sug incorporation observed in the 12F polysaccharide synthesized by the vaccine PFESP00919 strain used in PCV20 is a stable feature of this strain. Sug substitution was subsequently assessed in the ATCC 196-X polysaccharide, which is a component of the unconjugated polysaccharide vaccine PPSV23. We find that Sug is present in ATCC 196-X 12F polysaccharide but at a lower level of substitution of ∼0.1%, as determined by 2D HSQC NMR analysis of the unique Sug position 6 hydrate cross-peak. Furthermore, LC-MS/MS analysis confirms that the m/z 1076 species is present in the ATCC 196-X polysaccharide, but at much lower levels than PFESP00919 12F polysaccharide, confirming that Sug is present in ATCC 196-X polysaccharide and indicating that this particular fragment ion modulates with the incorporation level of Sug. These data show that Sug substitution is present to varying degrees in 12F strains used to generate the 12F conjugate vaccine PCV20 as well as the unconjugated polysaccharide vaccine PPSV23. To determine whether 12F Sug substitution is universal in circulating clinical isolates collected globally between 2004 and 2015, purified polysaccharide was analyzed from twenty-two 12F clinical isolates (confirmed by the Quellung reaction and sequencing of the cps operon) (30, 31) by 2D-NMR spectroscopy. All 17 of the 12F clinical isolates showed evidence of Sug substitution ranging from 1.9 to 10.2% of the RUs (Table III). These data show that Sug substitution is ubiquitous but at variable levels across circulating 12F clinical isolates. Additionally, 12F polysaccharide from repeat fermentation of strains PFESP00928 and PFESP00922 was analyzed, and the replicate analysis revealed an equivalent percentage of Sug incorporation, which indicates that there is little variation in the degree of incorporation for any given strain as observed with the PFESP00919 strain (Fig. 4).

The ketone/hydrate of the Sug residue is sensitive to reduction using NaBH₄, which is a standard, final treatment applied in reductive amination after conjugation of polysaccharides to carrier proteins. The NaBH₄ specifically reduces the (activated) aldehyde groups at the nonconjugated sites throughout the polysaccharide. Treatment of 12F polysaccharide with NaBH₄ is expected to specifically reduce the Sug position 4 ketone to an alcohol with a mixed population of novel FucpNAc and pentose residue N-acetyl-d-quinovosamine (QuipNAc) epimers (32), characterized by position 4 hydroxyl at axial and equatorial orientations, respectively (refer to Fig. 5B). As expected, NaBH₄-treated PFESP00919 12F polysaccharide is characterized by specific changes in NMR spectra of the Sug residue including deshielding of position 6 methyl carbon (>4 ppm ¹³C shift increase), loss of position 4 carbon sp² ketone population, as well as >18 ppm shift decrease of the position 4 carbon sp³ population due to reduction of the hydrate to two novel populations of FucpNAc and QuipNAc spin systems (Fig. 5A) at ∼1:1 ratio and ∼12 mol% each. With NaBH₄ reduction, the 12F primary spin system is unchanged, additional epimeric heterogeneity is introduced, and the Sug epitope is altered.

LC-MS and LC-MS/MS analysis of NaBD₄-reduced PFESP00919 12F polysaccharide confirms specific Sug reduction due to observed 3.0219-Da positive mass shift of fragment ions that contain a single Sug residue (Supplemental Fig. 1). This unique mass shift from deuterium is unnatural and provides unique fragment ions that do not contain natural isotopes of ¹²C, ¹³C, ¹H, ¹⁶O, and ¹⁴N. In Supplemental Fig. 1B, the fragment ions from reduced 12F RU comprise one to five sugar residues, which permits mapping of the deuterium incorporation in 12F RU residues. In this study, the unique mass shift of fragment ions containing the reduced and deuterated Sug residue (A^D; 188.0907 Da) confirms Sug replacement of backbone GalpNAc (203.0794 Da) in the 12F RU, especially because the branching pattern is maintained. The fragment ion of a single reduced Sug residue containing deuterium is observed at m/z 189.0981 (Supplemental Fig. 1B), whereas the fragment ion at m/z 204.0869 representing the single GalpNAc residue is missing (Supplemental Fig. 1C). There appears to be preferential fragmentation between A and E in the production of the precursor ion (m/z 1079) as mentioned previously, as abundance of the EA fragment ion is low. The same steps of NaBD₄ reduction and LC-MS analysis were applied to the low Sug-containing ATCC 196-X 12F polysaccharide, and an analogous pattern of reduced Sug-containing ions was generated albeit with lower levels of Sug residue but with the same pattern of incorporation (data not shown). The relative abundance of m/z 1079 is 33%, which is close to 25% of total ion current and consistent with NMR quantitation for PFESP00919 12F polysaccharide. NMR and MS data confirm that purified 12F polysaccharide, treated with NaBH₄ (NaBD₄) reduction used during the conjugation process of PFESP00919 12F polysaccharide, results in specific Sug reduction in purified 12F polysaccharides from both PFESP00919 and ATCC 196-X strains.

To confirm that PFESP00919 12F polysaccharide Sug residue is reduced by NaBH₄ treatment during conjugate formation, eight manufacturing scale lots of 12F conjugate were analyzed by 1D ¹H and 2D HSQC NMR to probe for populations of resolved Sug, novel FucpNAc and QuipNAc methyl resonances, and cross-peaks. Although no 1D ¹H signals are observed for Sug methyl resonance in all 12F conjugate production lots prepared with PFESP00919 polysaccharide, interference from carrier protein signals raises the limit of detection and quantification of methyl resonance in Sug, novel FucpNAc, and QuipNAc resonances by the 1D ¹H method. By 2D ¹H-¹³C-HSQC, methyl resonance in Sug, novel FucpNAc and QuipNAc ¹H/¹³C cross-peaks are resolved, and for all conjugate lots analyzed, both reduced novel FucpNAc and QuipNAc cross-peaks are observed at an equal intensity, and there is no detectable cross-peak for nonreduced Sug (Fig. 5A). The 2D-NMR data indicate that the Sug residue epitope of 12F polysaccharide in production conjugate lots is altered to populations of novel FucpNAc and QuipNAc by conjugation process NaBH₄ treatment.

To determine whether the Sug content of capsular polysaccharide impacts the ability of vaccine-induced Abs to bind and kill S. pneumoniae serotype 12F isolates, immunogenicity of the 12F-containing vaccines PCV20 and PPSV23 was assessed against a representative panel of S. pneumoniae 12F isolates with a range of Sug substitution levels using the OPA assay. The S. pneumoniae OPA assay assesses the ability of functional Ab to bind to pneumococcal bacteria in the presence of a functional complement source (baby rabbit complement), thereby facilitating bacterial engulfment and killing by a phagocytic human cell line (differentiated HL-60 cells). OPA assays were developed for six 12F isolates, and titers were generated for a panel of sera from subjects immunized with 12F-containing vaccines PCV20 or PPSV23, or the 12F-deficient vaccine PCV13 as a negative control.

As shown in Fig. 6, both PCV20 and PPSV23 immune sera were able to elicit bacterial killing responses to isolates with Sug substitution levels between ∼1.9 and ∼25.2%, with no statistically significant differences between titers across isolates, whereas sera from subjects immunized with PCV13, which does not contain a 12F capsular polysaccharide conjugate, showed no or low titers against all of the 12F isolates tested. The OPA data indicate that both PCV20 and PPSV23 are able to elicit a killing response against isolates with a range of Sug substitution. Because there is basal-to-no Sug epitope present in the PCV20 and PPSV23 vaccines, the Abs generated against them target the primary RU lacking Sug shown in Fig. 1A. These in vitro data indicate that the Sug epitope is not required to generate an effective immunogenic response against 12F, and that both PCV20 and PPSV23 vaccines generate killing response to circulating 12F isolates with a range of Sug incorporation.

The present study reports structural characterization of 12F pneumococcal serotype polysaccharide using high-resolution and sensitivity-enhanced multidimensional NMR and MS, revealing the presence of Sug substitution in the backbone, a feature not observed in previous studies. A survey of the serotype 12F PFESP00919 production strain, publicly deposited strains, clinical isolates, and purchased polysaccharide demonstrated that Sug substitution is ubiquitously present at varying levels, ranging from the lower level of ∼0.1% in the commercially available polysaccharide purchased from the American Type Culture Collection (ATCC 196-X) and used to formulate PPSV23, to the higher level of ∼25% in the PFESP00919 strain used for PCV20. The difference in the level of Sug incorporation in the 12F polysaccharide component used in the PPSV23 and PCV20 vaccines is therefore specific to the S. pneumoniae isolates grown and is a product of fermentation. Because analytical methods for characterization of polysaccharide structure have improved in the past 40 y, it is likely that Sug incorporation in serotype 12F polysaccharides was always present but until now was undetected. The findings will allow for updating the reference serotype 12F polysaccharide structure appropriately.

Identification of partial Sug substitution in the 12F polysaccharide was unexpected, but not without precedent, as similar substitutions have been found in recent years in other S. pneumoniae serotypes through the application of advanced multidimensional NMR methodology. Pneumococcal polysaccharide capsules are synthesized by a series of enzymes encoded by the cps locus. The synthetic cycle begins with the transfer of sugar 1-phosphate to the lipid carrier undecaprenyl phosphate mediated by an integral membrane glycosyltransferase. Glycosidic residues are then sequentially added to the developing oligosaccharide by cytosolic glycosyltransferases. Once mature, oligosaccharide RUs are exported to the bacterial surface by a flippase and are polymerized into surface-associated polysaccharide by a polymerase (7). It is now understood that some of the enzymes encoded in the pneumococcal cps locus are bispecific glycosyltransferases, which can catalyze the addition of two different sugars to the polysaccharide chain. This has been shown to be the case for S. pneumoniae serogroup 11 (27), as well as serogroups 6 and 7 (26, 28, 33), demonstrating the precedent for detection of partially substituted sugars at levels of ∼20% in pneumococcal polysaccharide capsule structure. Earlier polysaccharide structural characterization methods were likely not sensitive enough to detect polysaccharide RU heterogeneity (i.e., single sugar partial substitution) at the level of ∼20%, and it can be reasonably expected that future technology and/or methodological advances will allow identification of even more dilute partial substitution in polysaccharides. Whereas heterogeneity has been previously observed in specific polysaccharide serotypes, to our knowledge, this is the first study detailing variability in heterogeneity among strains of a specific serotype, and further illustrates the boundary of current bacteria polysaccharide structure understanding. Expanded analysis of bacterial polysaccharide dilute heterogeneity may reveal pathways for prokaryotes to evolve novel serotypes and thereby evade host recognition.

Although bispecific glycosyltransferases are known to recognize alternative substrate sugar intermediates, in most cited instances, the actual structure of the polysaccharides produced cannot unequivocally be explained to be a mixture of two discrete homopolymers or a single heteropolymer. The latter is conceptually more likely in that there is no mechanism for a single, soluble glycosyltransferase to distinguish between two separate growing polysaccharide chains. This work on 12F clearly shows that Sug and N-acetyl-galactosamine are coincorporated into a single heteropolymer that would be expected to randomly alternate the subunit constituent sugars, similar to a report of heteropolymer formation for S. pneumoniae serogroup 6 (28).

The incorporation of Sug into the 12F heteropolymer may also be governed by a bifunctional glycosyl transferase that confers a strain-specific level of Sug incorporation. The variable differences in Sug levels cannot be accounted for by the genetics of the cps operon. With the exception of strain ATCC 6312, the deduced amino acid sequence of protein coding regions across the cps operon is identical for all strains listed in Table III. Differences relative to strain ATCC 6312 (GenBank CR931660) include a T59A amino acid substitution in the gene coding for FnlA, a protein responsible for conversion of Sug to N-acetyl-fucosamine, and a K134M substitution in the gene coding for WcxE, annotated as an unassigned glycosyltransferase. Accordingly, a likely explanation for the observed strain-specific differences in Sug incorporation comes from differences in primary metabolic flux that would influence intracellular concentrations of constituent sugars.

The observation that S. pneumoniae serotype 12F polysaccharide contains varying levels of previously undescribed Sug raised the question of whether vaccines targeting the 12F serotype elicit an immunogenic response that varies based on 12F Sug incorporation level. Commercially available pneumococcal vaccines target the polysaccharide capsule, which is a known virulence factor. Importantly, both the PPSV23 polysaccharide vaccine and the PCV20 conjugate vaccine present nonsignificant Sug epitope due to a very low level in PPSV23, and due to Sug reduction to an equal mix of novel FucpNAc and QuipNAc during the conjugation process in PCV20. Protection conferred by pneumococcal vaccines is believed to be due to vaccine-elicited serotype-specific functional Ab responses that mediate OPA, whereby the serotype-specific Ab binds to the pneumococcal bacteria and induces uptake and killing by phagocytes in the presence of complement. This functional activity is measured in the OPA assay. Sera generated from both PCV20 and PPSV23 vaccines elicited killing responses against 12F isolates with a range of Sug incorporation, indicating that the level of Sug incorporation in the capsular polysaccharide of a given isolate does not impact susceptibility to killing. These findings are substantiated by the basal OPA titers against the same range of 12F isolates using sera generated from the PCV13 vaccine that contains no 12F polysaccharide conjugate.

Taken together, the data presented show that upon re-evaluation with modern structural characterization technology and newer methods, the S. pneumoniae serotype 12F capsular polysaccharide contains a previously unrecognized Sug incorporation. This incorporation was found to be present in all isolates examined, indicating that it was likely always present but not detected by earlier characterization methods. The level of Sug incorporation varied among strains from <2% to up to ∼25%, with most clinical isolates incorporating a low–medium level of Sug. The level of Sug incorporation was found to be consistent across multiple fermentation runs, indicating that the level of Sug incorporation in 12F polysaccharide is reproducible. Vaccines presenting 12F polysaccharide with basal-to-no Sug epitope were able to elicit functional, opsonophagocytic killing of all 12F clinical isolates tested, regardless of level of Sug incorporation. This suggests that level of Sug incorporation in the 12F polysaccharide used to produce a vaccine, such as PCV20, does not impact the immunogenicity of the vaccine, and that PCV20 is likely to protect against circulating S. pneumoniae clinical isolates.

The authors are or were employees of Pfizer Inc. and may own Pfizer stock.

We thank the following current and former (§) Pfizer Inc. colleagues: Christopher Kohanski^§ for 12F polysaccharide quantitative NMR analysis; Elena G. Novikova, Elizabeth Baranyi, Michele Weaver, Wei Chen, Assem Nemeri,^§ Vijaya Raghuraman,^§ and Marta Suazo^§ for 12F polysaccharide purification; Urvi Rajyaguru and Lubomira Andrew for microbiology and determination of bacterial genome sequence; Li Hao for analysis of bacterial genome sequence; Nishith V. Merchant for help in writing the purification section of Materials and Methods; and Paul A. Liberator, Kathrin U. Jansen^§, and Suddham Singh for critical review and comments. We also thank Nataliya Kushnir (Pfizer Inc.) for scientific writing assistance.