Streptococcus pneumoniae strains were cultured in Todd-Hewitt broth (BD Biosciences, San Jose, CA, USA) supplemented with 0.2% yeast extract THY medium (BD Biosciences) at 37°C. For selection and maintenance of mutants, spectinomycin (Fujifilm Wako Pure Chemical Corporation, Osaka, Japan) was added to the medium at 120 μg/mL. The Escherichia coli strain XL10-Gold (Agilent, Santa Clara, CA, USA) was used as a host for derivatives of plasmid pQE-30. All E. coli strains were cultured in Luria-Bertani broth supplemented with 100 μg/mL carbenicillin (Nacalai Tesque, Kyoto, Japan) at 37°C with agitation.

S. pneumoniae TIGR4 isogenic pfbA mutant strain was generated as previously described with minor modifications (Bricker and Camilli, 1999; Yamaguchi et al., 2008; Mori et al., 2012). Briefly, the upstream region of pfbA, an aad9 cassette, the downstream region of pfbA, and pGEM-T Easy vector (Promega, Madison, WI, USA) were amplified by PrimeSTAR® MAX DNA Polymerase (TaKaRa Bio, Shiga, Japan) using the specific primers listed in Supplementary Table 1. The DNA fragments were assembled using a GeneArt® Seamless Cloning and Assembly Kit (Thermo Fisher Scientific, Waltham, MA, USA). The constructed plasmid was then transformed into E. coli XL-10 Gold, and the inserted DNA region was amplified by PCR. The products were used to construct mutant strains by double-crossover recombination with the synthesized competence-stimulating peptide-2. Mutation was confirmed by PCR amplification of genomic DNA isolated from the mutant strain.

Human promyelocytic leukemia cells (HL-60, RCB0041) were purchased from RIKEN Cell Bank (Ibaraki, Japan). HL-60 cells were maintained in RPMI 1640 medium (Thermo Fisher Scientific) supplemented with 10% FBS, and were incubated at 37°C in 5% CO₂. HL-60 cells were differentiated into neutrophil-like cells for 5 days in culture media containing 1.2% DMSO (Collins et al., 1979; Wen et al., 2016). Cell differentiation was confirmed by nitro blue tetrazolium reduction assay (Collins et al., 1979).

Human TLR2/NF-κB/SEAP stably transfected HEK293 cells and human TLR4/MD-2/CD14/NF-κB/SEAP stably transfected HEK293 cells (Novus Biologicals, Centennial, CO, USA, currently sold by InvivoGen, San Diego, CA, USA) were maintained in DMEM with 4.5 g/L glucose, 10% FBS, 4 mM l-glutamine, 1 mM sodium pyruvate, 100 units/mL penicillin, 100 μg/mL streptomycin, 10 μg/mL blasticidin, and 500 μg/mL G418 and DMEM with 4.5 g/L glucose, 10% FBS, 4 mM l-glutamine, 1 mM sodium pyruvate, 100 units/mL penicillin, 100 μg/mL streptomycin, 10 μg/mL blasticidin, 2 μg/mL puromycin, 200 μg/mL zeocin, and 500 μg/mL G418, respectively. A secreted alkaline phosphatase reporter assay was performed according to the manufacturer's instructions (Novus Biologicals).

Phylogenetic analysis was performed as described previously (Yamaguchi et al., 2016, 2017, 2019), with minor modifications. Briefly, homologs and orthologs of the pfbA gene were searched using tBLASTn and DELTA-BLAST (Gertz et al., 2006; Boratyn et al., 2012). Sequences from tBLASTn and DELTA-BLAST results with e-values <1 × 10⁻⁸⁰ and >40% query coverage were selected for phylogenetic tree analysis. Sequences from incomplete coding sequences were excluded from the DELTA-BLAST results. The sequences were aligned using Phylogears2 (Venditti et al., 2006; Tanabe, 2008) and MAFFT v.7.221 using an L-INS-i strategy (Katoh and Standley, 2013), and ambiguously aligned regions were removed using Jalview or trimAl (Talavera and Castresana, 2007; Capella-Gutiérrez et al., 2009; Waterhouse et al., 2009). The best-fitting codon evolutionary models for phylogenetic analyses were determined using Kakusan4 (Tanabe, 2011). Bayesian Markov chain Monte Carlo analyses were performed with MrBayes v.3.2.6 (Ronquist et al., 2012), with sampling until the standard deviation of split frequencies was 0.01 or 8 × 10⁶ generations. To validate phylogenetic inferences, maximum likelihood phylogenetic analyses were performed with RAxML v.8.1.20 (Stamatakis, 2014). Phylogenetic trees were generated using FigTree v.1.4.4 (Rambaut, 2014) based on the calculated data. The pfbA genes of Gram-positive Granulicatella strains and Streptococcus merionis were used to root as outgroups.

Human blood was obtained via venipuncture from healthy donors after obtaining informed consent. The protocol was approved by the institutional review boards of Osaka University Graduate School of Dentistry (H26-E43). Human neutrophils and monocytes were prepared using Polymorphprep (Alere Technologies AS, Oslo, Norway), according to the manufacturer's instructions. Human blood was carefully layered on the Polymorphprep solution in centrifugation tubes, which were then centrifuged at 450 × g for 30 min in a swing-out rotor at 20°C. Monocyte and neutrophil fractions were transferred into tubes containing ACK buffer (0.15 M NH₄Cl, 0.01 M KHCO₃, 0.1 mM EDTA), then centrifuged, washed in phosphate-buffered saline (PBS), and resuspended in RPMI 1640 medium.

The pneumococcal cells grown to the mid-log phase were resuspended in PBS. TIGR4 strains (3–11 × 10³ CFUs/well) with or without rPfbA (0, 10, or 100 nM) were combined with human neutrophils or neutrophil like-differentiated HL-60 cells (2 × 10⁵ cells/well), and R6 strains (1.4–2.0 × 10² CFUs/well) were combined with human neutrophils (1 × 10⁵ cells/well). The mixture was incubated at 37°C in 5% CO₂ for 1, 2, and 3 h. Viable cell counts were determined by plating diluted samples onto TS blood agar. The growth index was calculated as the number of CFUs at the specified time point/number of CFUs in the initial inoculum. Bacterial phagocytosis was blocked by addition of cytochalasin D (20 μM), and pneumococcal killing was blocked by protease inhibitor cocktail set V (Merck, Darmstat, Germany; 500 μM AEBSF, 150 nM Aprotinin, 1 μM E-64, and 1 μM leupeptin hemisulfate, EDTA-free) at 1 h before incubation. To determine whether TLR2 and TLR4 signaling affect pneumococcal survival, 100 μM TIRAP (TLR2 and TLR4) inhibitor peptide or control peptide (Novus Biologicals) were added to neutrophils at 1 h before incubation.

For time-lapse observations, isolated neutrophils were resuspended in RPMI 1640 at 1 × 10⁶ cells/mL. Next, 10 μL of S. pneumoniae R6 wild type or ΔpfbA strains (1 × 10⁶ CFUs) was added to 2 mL of the cells, and the mixture was incubated and observed at 37°C. Time-lapse images were captured using an Axio Observer Z1 microscope system (Carl Zeiss, Oberkochen, Germany).

Recombinant PfbA (rPfbA) or BSA was coated onto 0.5 μm-diameter fluorescent beads (FluoroSphere, Thermo Fisher Scientific), according to the manufacturer's instructions. rPfbA was purified as previously described (Yamaguchi et al., 2008). Isolated neutrophils or monocytes were then resuspended in RPMI 1640 at 1.0 × 10⁷ cells/mL, after which 900 μL of RPMI 1640 containing 1 μL of rPfbA-, BSA-, or non-coated fluorescent beads was added to 100 μL of cells, and then the mixtures were rotated at 37°C for 1 h. The cells were washed twice and fixed with 2% glutaraldehyde-RPMI 1640 at 37°C for 1 h, then washed again three times and analyzed with a CyFlow flow cytometer (Sysmex, Hyogo, Japan) using FlowJo software ver. 8.3.2 (BD Biosciences, Franklin Lakes, NJ, USA).

HEK cells expressing TLR2 or TLR4 were stimulated with S. pneumoniae and/or rPfbA for 16 h, according to the manufacturer's instructions (Novus Biologicals). To avoid the effect of bacterial replication on this assay, S. pneumoniae were pasteurized by incubation at 56°C for 30 min. To perform the assay under the same condition, rPfbA was also incubated at 56°C for 30 min. Lipopolysaccharides from E. coli O111:B4 (Sigma-Aldrich Japan Inc., Tokyo, Japan) for the TLR-4 cell line and Pam3CSK4 and Zymozan (Novus Biologicals) for the TLR-2 cell line were used as positive controls under the same conditions. Secreted alkaline phosphatase (SEAP) was analyzed using the SEAPorter Assay (Novus Biologicals) according to the manufacturer's instructions. Quantitative data (ng/mL) were obtained using a standard curve for the SEAP protein.

We performed miRNA array analysis using neutrophil like-differentiated HL-60 cells incubated with S. pneumoniae strains and/or 100 nM rPfbA for 1 h. We compared rPfbA-treated and non-treated cells, wild type and ΔpfbA-infected cells, and ΔpfbA with and without rPfbA-infected cells. In each cell sample, six replicates were pooled and total RNA including microRNA was isolated from the pooled cells by miRNeasy Mini Kit (Qiagen, Hilden, Germany). Approximately 1,000 ng RNA was used for microarray analysis using Affymetrix GeneChip miRNA 4.0 arrays (Affymetrix, Santa Clara, CA, USA) through Filgen Inc. (Nagoya, Japan). Briefly, the quality of total RNA was assessed using a Bioanalyzer 2100 (Agilent). Hybridization was performed using a FlashTag Biotin HSR RNA Labeling Kit, GeneChip Hybridization Oven 645, and GeneChip Fluidics Station 450. The arrays were scanned by Affymetrix GeneChip Scanner 3000 7G. The GeneChip miRNA 4.0 arrays contain 30,424 total mature miRNA probe sets including 2,578 mature human miRNAs, 2,025 pre-miRNA human probes, and 1,196 Human snoRNA and scaRNA probe sets.

Mouse infection assays were performed as previously described (Okerblom et al., 2017; Yamaguchi et al., 2017, 2019; Hirose et al., 2018). For the lung infection model, CD-1 mice (Slc:ICR, 8 weeks, female) were infected intratracheally with 4.3–6.7 × 10⁶ CFUs of S. pneumoniae. For intratracheal infection, the vocal cords were visualized using an operating otoscope (Welch Allyn, NY, USA), and 40 μL of bacteria was placed onto the trachea using a plastic gel loading pipette tip. Mouse survival was monitored twice daily for 14 days. At 24 h after intratracheal infection, bronchoalveolar lavage fluid (BALF) was collected following perfusion with PBS.

For the sepsis model, CD-1 mice (Slc:ICR, 8 weeks, female) were infected intravenously with 3.3–6.5 × 10⁵ CFUs of S. pneumoniae via the tail vein. Mouse survival was monitored twice daily for 14 days. At 24 and 48 h after infection, blood aliquots were collected from mice following induction of general euthanasia. Brain, lung, and liver samples were collected following perfusion with PBS. Brain and lung whole tissues as well as the anterior segment of the liver were resected. Bacterial counts in the blood as well as organ homogenates were determined by separately plating serial dilutions, with organ counts corrected for differences in organ weight. Detection limits were 50 CFUs/organ and 50 CFUs/mL in blood.

The concentrations of TNF-α in BALF and plasma were determined using a Duoset® ELISA Kit (R&D Systems, Minneapolis, MN, USA). Mice plasma was obtained by centrifuging the heparinized blood. All mouse experiments were conducted in accordance with animal protocols approved by the Animal Care and Use Committees at Osaka University Graduate School of Dentistry (28-002-0).

Statistical analysis of in vitro and in vivo experiments was performed using a non-parametric analysis, Mann-Whitney U-test, or Kruskal-Wallis test with Dunn's multiple comparisons test. Mouse survival curves were compared using a log-rank test. p < 0.05 was considered to indicate a significant difference. The tests were carried out with Graph Pad Prism version 6.0h or 8.1.2 (GraphPad Software, Inc., San Diego, CA, USA).

We searched pfbA-homologs using tBLASTn and DELTA-BLAST and performed phylogenetic analysis (Figure 1 and Supplementary Figure 1). The tBLASTn search of the NCBI Nucleotide collection database showed that pfbA-homologs were present only in S. pneumoniae, S. pseudopneumoniae, S. gordonii, and Streptococcus merionis. To expand the targeted information, we performed DELTA-BLAST using the NCBI Non-redundant proteins sequences database, and utilized the obtained sequences for phylogenetic analysis after excluding incomplete coding sequences. The pfbA gene homologs were identified in mitis group Streptococcus (S. pneumoniae, S. pseudopneumoniae, S. mitis, S. oralis, S. infantis, S. gordonii, and S. anginosus), Sphingomonas paucimobilis, Haemophilus haemolyticus, Granulicatella species, and S. merionis. In these bacteria, S. paucimobilis and H. haemolyticus would have obtained the pfbA gene by occasional horizontal gene transfer from S. pneumoniae, given that the gene was only detected in one strain in each species and showed 100% identity with the pneumococcal variant. Genus Granulicatella bacteria were previously identified as a nutritionally variant Streptococcus (Christensen and Facklam, 2001), and Granulicatella species and S. merionis each contain a gene according to the query, with coverage and identity >50%. S. merionis strain NCTC13788 (also known as WUE3771, DSM 19192, and CCUG 54871), isolated from the oropharynges of Mongolian jirds (Meriones unguiculatus), contained 16S rRNA that belongs in a cluster distinct from the mitis group (Tappe et al., 2009). The pfbA genes in S. pneumoniae and S. pseudopneumoniae are highly conserved and formed an independent cluster from a cluster of mostly other mitis-group strains, whereas this gene showed genetic diversity in other mitis-group bacteria. Interestingly, the pfbA gene of some S. mitis strains belongs to a cluster between pneumococcal clusters. These results indicated that during the evolutionary process, some S. mitis strains lost and re-gained the gene via horizontal gene transfer from S. pneumoniae.

To investigate whether PfbA contributes to evasion of neutrophil killing, we determined pneumococcal survival rates after incubation with human neutrophils. After 3 h incubation, the TIGR4 ΔpfbA strain showed a significantly decreased bacterial survival rate. In addition, to clarify whether the observed effects were attributable to PfbA, we also performed the assay using rPfbA. In the presence of 100 nM rPfbA, TIGR4 ΔpfbA strain demonstrated a recovered survival rate nearly equal to that of the wild-type strain (Figure 2A). In pneumococcal survival assays with neutrophil-like differentiated HL-60 cells, TIGR4 strains showed similar results (Figure 2B). We also performed the assay using the non-encapsulated strain R6 and human neutrophils. The R6 ΔpfbA strain showed significantly decreased survival rates as compared to the wild-type strain after incubation for 1, 2, and 3 h (Figure 2C). As the R6 strain showed this phenotype at earlier time points than the TIGR4 strain, we performed pneumococcal survival assays using R6 strains with inhibitors (Figure 2D). Neutrophil phagocytic killing of S. pneumoniae requires the serine proteases (Standish and Weiser, 2009). Thus, in the present study, we used a protein inhibitor cocktail as a positive control of a neutrophil killing inhibitor. While the R6 ΔpfbA strain showed significantly decreased survival rates at 1 h after incubation with human fresh neutrophils in the absence of inhibitors, treatment with an actin polymerization inhibitor, cytochalasin D, reduced the differences among the wild-type and ΔpfbA strains as well as the protein inhibitor cocktail.

We confirmed the anti-phagocytic activity of PfbA using flow cytometry and PfbA-coated fluorescent beads (Figure 3A). The fluorescence intensity of neutrophils and monocytes incubated with PfbA-coated beads was substantially lower as compared with cells incubated with non- or BSA-coated beads. These results indicated that neutrophils and monocytes phagocytosed the non- and BSA-coated fluorescent beads, whereas the PfbA-coated fluorescent beads escaped phagocytosis by neutrophils and monocytes.

We performed real-time observations for time-lapse analysis of the interaction between S. pneumoniae and neutrophils (Figure 3B). S. pneumoniae strain R6 wild-type and ΔpfbA strains were separately incubated with fresh human neutrophils in RPMI 1640 medium. After coming into contact with neutrophils, the ΔpfbA strain was phagocytosed within 1 min, whereas the wild-type strain was not phagocytosed after more than 5 min. Time-lapse analysis also showed the ΔpfbA strain engulfed by neutrophil phagosomes.

In summary, these results showed that pfbA deficiency in both S. pneumoniae TIGR4 and R6 strains decreased the survival rate after incubation with human neutrophils and differentiated HL-60 cells. Furthermore, rPfbA-addition recovered the survival rate of the ΔpfbA strain, with this recovery also observed following treatment with cytochalasin D and a protein-inhibitor cocktail. Additionally, PfbA-coated beads escaped phagocytosis, and time-lapse analysis showed that the ΔpfbA strain was more easily phagocytosed than the WT strain. These findings suggested that PfbA can directly inhibit phagocytosis.

Some lectins of pathogens work as ligand for TLR2 and TLR4 (Ricci-Azevedo et al., 2017). We previously reported that PfbA can interact with glycolipid and glycoprotein fractions of red blood cells, several monosaccharides, d-sucrose, and d-raffinose (Beulin et al., 2017; Radhakrishnan et al., 2018). Hence, to determine whether PfbA works as a TLR ligand, we performed a SEAP assay using HEK-293 cells stably transfected with either TLR2 or TLR4, NF-κB, and SEAP (Figure 4A). Pam3CSK4 and Zymozan were used as positive controls for the TLR2 ligand, while LPS was used for TLR4. The SEAP assay indicated that pasteurized S. pneumoniae TIGR4 wild-type cells activated NF-κB via TLR2, whereas ΔpfbA cells did not stimulate cells expressing either TLR2 or TLR4. Pasteurized rPfbA also activated NF-κB dose-dependently through TLR2, but not TLR4. In addition, in the presence of pasteurized rPfbA, ΔpfbA cells activated the cells expressing TLR2. To confirm the effect of pasteurization, we performed this assay using intact and pasteurized bacteria or recombinant proteins (Supplementary Figure 2). Use of the same number of live bacteria generated less amounts of SEAP as compared to that of a smaller number of bacteria, possibly due to cell death caused by the introduction of abundant live bacteria. To minimize this effect, we performed the assay using a smaller number of bacteria and observed that both pasteurized proteins and bacteria exhibited significantly decreased activity relative to intact variants. These results indicated that pasteurization at least partially denatured PfbA and decreased its recognition by TLR2, and that PfbA is responsible for pneumococcal NF-κB activation through TLR2 recognition of the structure.

Next, to determine whether TLR signaling suppresses survival of pneumococci incubated with neutrophils, we performed a neutrophil survival assay using a TIRAP inhibitor peptide (Figure 4B). Data are presented as the ratio calculated by dividing CFUs in the presence of inhibitor peptide by CFUs in the presence of control peptide. TIRAP is an adaptor protein involved in MyD88-dependent TLR2 and TLR4 signaling pathways. Since the TIRAP inhibitor peptide blocks the interaction between TIRAP and TLRs, the peptide works as a TLR2 and TLR4 inhibitor. The inhibitor peptide treatment increased survival rates of the ΔpfbA strain, but did not affect wild-type survival rates. These results indicated that PfbA contributes to the evasion of neutrophil phagocytosis, and TIRAP inhibitor treatment did not change survival rates of pneumococci incubated with neutrophils. On the other hand, the S. pneumoniae ΔpfbA strain was more easily killed by neutrophils as compared to the wild-type strain, and this phenotype was abolished by the TIRAP inhibitor.

To investigate the role of PfbA in pneumococcal pathogenesis, we infected mice with S. pneumoniae strains intratracheally and compared bacterial CFUs and TNF-α levels in BALF from mice at 24 h after infection. There were no differences observed in survival time between mice infected with wild type and ΔpfbA strains (Figure 5A). However, recovered CFUs of wild-type bacteria were significantly greater than those of ΔpfbA strains in mouse BALF. In addition, the level of TNF-α in BALF was almost the same in wild type and ΔpfbA infection (Figure 5B).

We also investigated the role of PfbA in mice following intravenous infection as a model of sepsis. In the infection model, the ΔpfbA strain showed significantly higher levels of virulence as compared to the wild-type strain (Figure 6A). Furthermore, we compared the TNF-α levels in plasma and examined the bacterial burden in blood, brain, lung, and liver samples obtained at 24 and 48 h after intravenous infection (Figures 6B,C and Supplementary Figure 3). At 24 h after infection, TNF-α ELISA findings showed a significantly greater level in the plasma of pfbA mutant strain-infected mice as compared to the wild-type strain-infected mice. The numbers of CFUs of both the wild-type and pfbA mutant strains in the blood and brain samples were comparable. On the other hand, in the lung and liver samples, the pfbA mutant strain-infected mice showed slightly but significantly reduced numbers of CFUs as compared with the wild-type strain-infected mice. At 48 h after infection, there were no significant differences in TNF-α level and bacterial burden in each organ between the wild-type- and pfbA mutant strain-infected mice (Supplementary Figure 3). Bacteria were not detected in the blood of two of the wild-type strain-infected mice and five of the pfbA mutant strain-infected mice. Meanwhile, three of the wild-type strain-infected mice yielded more than 10⁶ CFUs/mL, while seven of the pfbA mutant strain-infected mice did (Supplementary Table 2). The pfbA mutant strain infection caused a polarized bacterial burden in the hosts at 48 h after infection as compared with this not being observed following wild-type strain infection. PfbA deficiency did not increase bacterial burden in mouse organs, but increases pneumococcal pathogenicity in intravenous infection.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.