Evolutionary inactivation of a sialidase in group B Streptococcus

Group B Streptococcus (GBS) is a leading cause of bacterial sepsis and meningitis in newborns. GBS possesses a protein with homology to the pneumococcal virulence factor, NanA, which has neuraminidase (sialidase) activity and promotes blood-brain barrier penetration. However, phylogenetic sequence and enzymatic analyses indicate the GBS NanA ortholog has lost sialidase function – and for this distinction we designate the gene and encoded protein nonA/NonA. Here we analyze NonA function in GBS pathogenesis, and through heterologous expression of active pneumococcal NanA in GBS, potential costs of maintaining sialidase function. GBS wild-type and ΔnonA strains lack sialidase activity, but forced expression of pneumococcal NanA in GBS induced degradation of the terminal sialic acid on its exopolysaccharide capsule. Deletion of nonA did not change GBS-whole blood survival or brain microvascular cell invasion. However, forced expression of pneumococcal NanA in GBS removed terminal sialic acid residues from the bacterial capsule, restricting bacterial proliferation in human blood and in vivo upon mouse infection. GBS expressing pneumococcal NanA had increased invasion of human brain microvascular endothelial cells. Thus, we hypothesize that nonA lost enzyme activity allowing the preservation of an effective survival factor, the sialylated exopolysaccharide capsule.

gene, an ortholog of pneumococcal nanA (SAK_RS09520 or SAK_1891), annotated in the published genome of GBS strain A909. Amino acid sequence alignment analysis showed that SAK_RS09520 contains a sialidase domain but lacks the conserved lectin like-domain LamG and cell wall anchoring motifs present in pneumococcal NanA (Fig. 1A). GBS NonA shared 58% amino acid sequence identity with pneumococcal NanA across the sialidase domain, with lesser degrees of sequence identity with pneumococcal NanB and NanC (27-28% amino acid sequence identities) (Supplementary Table 1). Next, tBLASTn analysis revealed that a subset of species in the genus Streptococcus contains nanA orthologs and a phylogenetic analysis was performed using orthologous bacterial nanA sequences. Both Bayesian-and maximum likelihood phylogenetic analyses of these orthologs revealed similar patterns of genetic classification with high posterior probabilities or bootstrap values (Fig. 1B,  Supplementary Fig. 1 and Supplementary Table 2). The sialidase genes of Gram-positive and rod-shaped bacteria, Erysipelothrix rhusiopathiae, Clostridium perfringens, and Virgibacillus sp., were used to root, since with the exception of homologous genes from other streptococci, these genes exhibited the highest similarity with pneumococcal nanA. E. rhusiopathiae and C. perfringens are known to produce active sialidases [18][19][20] . Both trees indicate that the nanA ortholog genes of S. mitis and S. pseudopneumoniae diverged from each other, having shared a common ancestor. Of note, nanA genes of S. pneumoniae strains CGSP14 and NT_110_58 were distinct from those of other pneumococcal strains. The phylogenetic analysis revealed that the GBS nonA represents a single lineage in a cluster otherwise composed of S. iniae, which is a pathogen of fish and occasional nosocomial infections in humans 21 . NonA of both S. iniae and GBS lack an LPXTG motif, which is conserved in the NanA proteins of other streptococcus species. All streptococcal sialidases except GBS NonA possess the LamG domain ( Fig. 1A and Supplementary Fig. 2). Further analysis of the genome database indicates that five GBS strains (GX026, SA20-06, 2-22, 138spar, and 138P) carry a nonA gene containing a nonsense mutation (Supplementary Table 3). In addition, we measured bacterial sialidase activities using streptococcal type strains and clinical isolates ( Supplementary Fig. 3). Type strains of S. oralis, S. intermedius, and S. pseudopneumoniae showed positive sialidase activities. In contrast, the sialidase activity of S. mitis, GBS, and S. iniae strains was always below the detection limit. Previously, Killian et al. reported that 100% of 17 S. pneumoniae and 3 S. pseudopneumoniae strains and 69% of 54 S. mitis strains showed positive sialidase activity 22 . Some S. mitis strains appear to have reduced the genome sizes and may have lost virulence-associated factors including NanA in a reductive evolutionary process 23,24 . Thus, it is likely that S. mitis strains exhibit a diversity of sialidase activity. Furthermore, the result of an ancestral reconstruction technique suggests the possibility that sialidase activity was lost in the nonA lineage rather than gained in the nanA lineage (Fig. 2). Together these results suggest that streptococcal nanA orthologs diverged into two major groups, one consisting of S. mitis, S. intermedius and S. pneumoniae, and the other consisting of GBS and the S. iniae group. In the S. iniae/GBS group NonA appears to have lost its functional role.
To examine the relationship of pneumococcal nanA, nanB, nanC, and GBS nonA, a phylogenetic analysis was performed using the genes. Bayesian-and maximum likelihood phylogenetic analyses of the genes revealed similar patterns of genetic classification with high posterior probabilities or bootstrap values ( Fig. 3 and Supplementary Fig. 4). The pneumococcal nanB/nanC were well separated from pneumococcal nanA or GBS nonA. We performed an additional evolutionary analysis on nanA, nonA, nanB, nanC, bgaA, and strH genes. BgaA and StrH, another pneumococcal exoglycosidases, remove galactose that is β 1-4 linked to N-acetylglucosamine, and N-acetylglucosamine that is β 1 linked to mannose, respectively 13 . Selection analysis through non-synonymous/synonymous ratio calculations by Fixed Effects Likelihood (FEL) and Fast, Unconstrained Bayesian AppRoximation (FUBAR) analyses suggested similar results. There were more codons evolving under negative selection in the nanA genes of S. pneumoniae strains (  and Supplementary Figs 5-10). In contrast, fewer codons evolving under negative selection were detected in the nonA genes of GBS as well as the nanB and nanC genes. Similar results were obtained with the bgaA and strH genes, indicating that nanA is under strong selective pressure. On the other hand, there were no or very few codons that appear to be evolving under positive selection in these genes. We conducted a likelihood ratio test to investigate whether pneumococcal nanA and GBS nonA genes have the same distribution of substitution rates across sites ( Table 2). The distributions of substitution rates indicate no significant differences in between pneumococcal nanA and GBS nonA genes. However, there was a significant difference in selective regimes (dN/dS    and proportions), especially in the proportions of codons under selection. These results suggest that a functional change of NanA would be deleterious in S. pneumoniae. In fact, pneumococcal NanA is a multifunctional protein, that promotes bloodstream survival 17 and penetration of host endothelial cell barriers system 14 . In contrast to pneumococcal NanA, the GBS NonA does not appear to be under strong selective pressure, which supports our hypothesis that NonA no longer functions in GBS.
Forced expression of NanA in GBS degrades terminal sialic acids of its capsule. To investigate the role of NonA in bacterial pathogenesis, we constructed an isogenic GBS Δ nonA mutant strain and then complemented the Δ nonA strain with the functional pneumococcal NanA as described in the Methods section. The expression of the nonA gene in a GBS wild-type (WT) strain was higher than that of the well-characterized cylE gene encoding the GBS β -hemolysin/cytolysin ( Supplementary Fig. 11). Sialidase activities of GBS WT, Δ nonA, and Δ nonA[pNanA] strains were determined using a fluorometric sialidase assay (Fig. 4A). Neither the WT nor Δ nonA GBS strains showed sialidase activity associated the bacterial cells or culture supernatants, but sialidase activity could be detected with heterologous expression of the pneumococcal enzyme. We next investigated whether heterologous expression of a functional sialidase (NanA) would degrade the terminal sialic acid moiety on the GBS capsular polysaccharide repeating unit by flow cytometry with FITC-labeled Erythrina cristagalli agglutinin (ECA; Fig. 4B). ECA binds to terminal (unsialylated) galactose and the ECA binding level inversely reflects the level of sialylation on the GBS capsule. GBS WT and Δ nonA mutant strains showed similar histogram patterns and did not interact with the FITC-labeled ECA. On the other hand, the complemented GBS Δ nonA[pNanA] strain showed substantially higher fluorescence intensity when incubated with FITC-labeled ECA as compared to the strains incubated without FITC-labeled ECA. These results indicated that the GBS WT strain possessed no sialidase activity and the forced expression of the active sialidase in GBS could have the effect of degrading its own terminal sialic acid, a known immune evasion virulence factor of the pathogen with anti-complement, anti-phagocytic, and immunosuppressive properties 4-8 .

NonA does not contribute to GBS invasion into hBMECs.
To examine the role of GBS NonA compared to the previously established role of pneumococcal NanA in the invasion of blood-brain barrier endothelium, we performed adherence/invasion assay using human brain microvascular endothelial cells (hBMECs) (Fig. 5). To quantify bacterial invasion, hBMECs were incubated with GBS strains for 1 hour (h) and further incubated for 1 h in medium containing antibiotics. WT GBS and the Δ nonA mutant did not differ in their adherence or invasion phenotypes to hBMEC (Fig. 5). The association of the Δ nonA[pNanA] strain was decreased as compared to that of other strains; however, invasion of the Δ nonA[pNanA] strains into human brain microvascular endothelial cells were significantly higher than that of GBS WT and Δ nonA strains. These results indicated that pneumococcal NanA, but not the endogenous GBS NanA homologue, can contributes to bacterial invasion of brain endothelial cells.

Expression of sialidase in GBS inhibits its survival.
We compared bacterial survival rate ex vivo in human blood to compare the function of GBS NonA and active pneumococcal NanA in the GBS background (Fig. 6). The series of GBS strains were mixed with freshly collected human blood, and the mixture incubated for 3 h. Viable cell counts were determined hourly by plating diluted samples onto THY agar. No significant change in survival was noted comparing the isogenic Δ nonA mutant to the parent GBS WT strain. However, when NanA was introduced in this background, survival was approximately 72-79%, 39-44%, and 23% of WT levels after 1, 2, and 3 h, respectively. These results indicated that forced expression of NanA inhibits the survival of GBS in human blood, likely through the degradation of its terminal sialic acid on the capsule.
Finally, to investigate the potential role of NonA vs. an active sialidase in GBS pathogenesis, we infected GBS strains in mice intravenously, and compared bacterial CFU in blood and brains from mice 20 h after infection. Recovered CFUs of WT and Δ nonA strains in mouse blood were almost same. However, the Δ nonA[pNanA] expressing sialidase activity had on 14% the level of WT survival in the blood, consistent with the findings in the human blood survival assay.

Discussion
S. pneumoniae contains three sialidases, NanA, NanB, and NanC. NanB works as a virulence factor in pneumococcal infection and NanC catalyzes intermediate metabolic compounds which acts as sialidase inhibitor 25,26 . The nanC gene was reported to significantly associate with clinical isolates from invasive diseases 12,27 . Our evolutionary analysis indicated that ~15% of the codons in the pneumococcal nanA gene evolved under negative selection, while few codons of nanB and nanC evolved in positive or negative selection. The results suggest that selection pressures exist such that the key enzymatic functions of NanA may not change. In contrast to the   conservations on nanA in S. pneumoniae, the GBS NanA homologue (NonA) appears to have lost its sialidase activity. Sialyltransferases are highly conserved in GBS strains of each serotype when compared to other glycosyltransferase genes in a same operon, and the difference in genetic diversity support a hypothesis that sialic acid is critical for GBS survival in the human host 28 . Our results showed that restoration of an active sialidase function inhibited GBS survival in human blood ex vivo and mouse blood in vivo. Therefore, sialidase activity would be deleterious to the fitness of GBS, and GBS nonA appears to be a non-functional gene. We recently reported a similar relationship between bacterial capsule and glycosidase in another pathogenic streptococci, group A Streptococcus (GAS, Streptococcus pyogenes) 29 . Almost all serotypes of GAS express a hyaluronan exopolysaccharide capsule and contain an inactivated version of the hyaluronidase (HylA) with a single nucleotide mutation resulting in Asp to Val substitution at amino acid position 199 30 . However, serotype M4 strains express an active HylA, while lacking hyaluronan capsule biosynthesis operon. The operon was predicted to represent a more recent evolutionary acquisition in most serotypes. Although hyaluronan capsule is a major GAS virulence factor, heterologous expression studies to generate partial encapsulation of M4 wild-type strain and full encapsulation of an isogenic mutant Δ hylA strain did not increase virulence. In this human bacterial pathogen, the conflicts between polysaccharide capsule and glycosidase would exert conflicting selective pressures, and resulted in mutual exclusivity. In the present work, we find a similar mutual exclusivity between sialidase activity and the GBS polysaccharide capsule.
It is widely thought that pathogenic microbes may explain some human polymorphisms 31,32 . Sialylated pathogens can dampen the immune response through interaction with Siglecs, and this molecular mimicry is considered to be one of the primary forces in the rapid evolution of human Siglecs 4,33-35 . For example, Siglec-13 and -17 may have been genetically eliminated during hominid evolution, because of interactions with pathogenic bacteria, including GBS, that cause invasive infections 33 . In addition, Siglec-14 and -5 expressed on neutrophils and monocytes appear to have evolved to provide a balanced response to pathogens and infants with Siglec-14 deficiency were the most prone to GBS immune subversion 36 . Thus, there exists a multifaceted interaction between pathogen and human evolution at the molecular level. The synergy of evolutionary bioinformatics and functional analysis may help to investigate the interplay between pathogen and host within an evolutionary framework and to identify new genetically stable therapeutic targets within pathogens and/or their human hosts.

Methods
Phylogenetic and evolutionary analysis. Phylogenetic and evolutionary analyses were performed as previously described with minor modifications 37 . Homologues of nanA were searched for using tBLASTn of NCBI BLAST 38 . Sequences from complete genomes with e-values < 2 × 10 −85 and > 40% query coverage were selected for phylogenetic tree analysis. The sequences were aligned using MAFFT ver. 7.221 with FFT-NS-i strategy 39 and edited using by Jalview 40 . Regions coding sialidase domain were used for further phylogenetic analysis. Edited sequences were aligned again using MAFFT with L-INS-i strategy. The best-fitting codon evolutionary models for maximum likelihood and Bayesian phylogenetic trees were determined by Kakusan4 41 . Maximum likelihood phylogenetic trees with bootstrap values were generated by RAxML ver. 8.1.20 42  were performed using two-rate Fixed Effects Likelihood (FEL) and Fast Unconstrained Bayesian AppRoximation (FUBAR) in the HyPhy software package 44,45 . Comparing codon selection between nanA and nonA genes was performed using LR tests in the HyPhy 44 .
Ancestral states for bacterial sialidases were reconstructed in Mesquite version 3.04 46 using a parsimony model with characters treated as unordered. The reconstruction was performed on the phylogenetic tree generated by MrBayes. States of active or inactive sialidase were assigned "0" or "1" for each taxon. Unavailable data were coded as missing.  . Blood and brain were collected at 20 h after infection. All mice were perfused with PBS prior to brain isolation. S.E. were represented by vertical lines. The difference between groups was analyzed using an Mann-Whitney U-test. 49455) 47 was kindly provided by Dr. T. Hoshino, Nagasaki University, Japan. The Escherichia coli strain TOP10 (Invitrogen) was used as a host for derivatives of plasmids pSET4s, pDCerm, and pDESTerm. All E. coli strains were cultured in Luria-Bertani (LB) broth at 37 °C with agitation. For selection and maintenance of mutants, antibiotics were added to the media at the following concentrations: ampicillin (Wako), 100 μ g/ml for E. coli; kanamycin (Sigma-Aldrich), 50 μ g/ml for E. coli; chloramphenicol (Sigma-Aldrich), 10 μ g/ml for E. coli; spectinomycin (Wako), 100 μ g/ml for E. coli and 150 μ g/ml for GBS; and erythromycin (Sigma-Aldrich), 400 μ g/ml for E. coli and 5 μ g/ml for GBS. Human brain endothelial cell line (hBMEC) was maintained in RPMI 1640 medium (Invitrogen) supplemented with 10% FBS, 10% NuSerum (BD), and 1% MEM nonessential amino acids, and were incubated at 37 °C in 5% CO 2 .

Bacterial strains and cell lines. Streptococcal strains listed in Supplementary
Construction of mutant strain. The construction of in-frame deletion mutants was conducted using a temperature-sensitive shuttle vector, pSET4s, as reported previously 48,49 . During the course of construction, a merodiploid strain was created after the first allelic replacement and then resolved to possess either mutant or wild type alleles after the second allelic replacement. To minimize the effect of secondary mutations and epigenetic changes that may have arisen during mutagenesis, a clone possessing the wild type allele was used as a wild-type strain. Both the wild-type and an in-frame deletion mutant strain arose from the same merodiploid ancestor. The correct in-frame deletion of genes was confirmed by site-specific PCR using purified chromosomal DNA. To create NonA and NanA-swapped GBS strain, Δ nonA[pNanA], pNanA plasmid was introduced respectively into GBSΔ nonA strain by electroporation 14 . pNanA was constructed by ligating nanA gene from S. pneumoniae strain D39 into pDESTerm plasmid 50 .

Sialidase activity assay. Sialidase activities of bacterial cells and supernatants were determined by
Neuraminidase assay kit (abcam). Streptococcal strains were grown to the mid-log phase (OD 600 = 0.4-0.5) and centrifuged. To prepare bacterial cell fraction, the bacterial pellet was washed by PBS and resuspended in PBS. The supernatant was used as a supernatant fraction. The samples were incubated for 2 hours at 37 °C and fluorescence intensity was measured at Ex/Em = 350 nm/460 nm.

Real-time reverse transcription-PCR (RT-PCR) assay.
Total RNA of GBS strains grown to the exponential phase (OD 600 = 0.5) was isolated with RNeasy mini kit and RNase-Free DNase Set (Qiagen). Then, cDNA was synthesized with Transcriptor First Strand cDNA Synthesis Kit (Roche). Real-time RT-PCR analysis was conducted using StepOnePlus Real-Time PCR system (Thermo Fisher Scientific) and KAPA SYBR Fast qPCR Kit (KAPA Biosystems). Data for gyrA were used as internal control. Primers are listed in Supplementary Table 6. ECA-binding assay. ECA-binding assay was performed as previously described 51 . GBS strains were grown to the mid-log phase and resuspended in PBS to adjust OD 600 to 0.1. The bacteria were incubated on ice with FITC-conjugated Erythrina cristagalli agglutinin (ECA; Vector Laboratories, CA) at 10 μ g/mL for 30 min. And then, bacterial cells were washed and resuspended in PBS. The ECA-binding activities on the surface of live bacterial cells were analyzed with a CyFlow SL flow cytometer. hBMEC association and invasion assay. The bacterial association to and invasion of hBMEC were quantified with minor modifications as described previously [52][53][54] . GBS strains were grown to mid-log phase (OD 600 = 0.5) and resuspended in PBS (OD 600 = 0.1). hBMECs were seeded at 2 × 10 5 cells per well in RPMI1640 supplemented with 10% FBS in 24-well plates 1 d prior to bacterial infection. In each well, ~2.0 × 10 6 CFU of bacteria was added to infect with ~2.0 × 10 5 hBMECs at an multiplicity of infection (MOI) of 10 in a final volume of 500 μ l, and the plate was centrifuged at 1600 rpm for 5 min to initiate their contact. To determine bacterial adhesion, the infected cells were incubated for 1 h, washed three times with PBS, and harvested with a trypsin and 0.025% Triton X-100 solution. The number of bacterial association was quantified by serial dilution plating. To examine bacterial invasion, hBMECs were washed following 1 h-incubation, and 500 μ L of media containing 100 μ g/mL of gentamicin was added and cells were incubated for an additional 1 h. The cells were washed and lysed, and the number of bacterial invasion was quantified. The bacterial association or invasion rate was calculated by dividing the number of bacterial association/invasion by the number of original inoculums. The invasion rate of bacterial association was also calculated by dividing the number of bacterial invasion by the number of bacterial association. Blood bactericidal assay. A blood bactericidal assay was performed as previously described 52,55,56 . Blood was obtained via venopuncture from healthy donors. It was performed under written informed consent according to a protocol approved by the institutional review boards of Osaka University Graduate School of Dentistry. The GBS cells grown to the mid-log phase were washed and resuspended in PBS, and OD 600 was adjusted to 0.1. Bacterial cells (10 μ l) were combined with fresh human blood (190 μ l), and then the mixture was incubated at 37 °C in 5% CO 2 for 1, 2, and 3 hours. Viable cell counts were determined by plating diluted samples onto THY agar. Growth index was calculated as the number of CFU at the specified time point/number of CFU in the initial inoculum.

Mice infection assay.
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 (24-025-2). CD-1 (ICR: IGS) mice (6 weeks, female; Oriental) were infected with 3.5 × 10 7 CFU of GBS via the tail vein. After 20 h post-infection, blood aliquots were collected from mice just after general euthanasia. The samples of brain/ meninges were collected following perfusion with PBS. Bacterial counts in blood and brain homogenates were determined by plating serial dilutions. Bacterial counts in brain were corrected for differences in each brain weight.
Scientific RepoRts | 6:28852 | DOI: 10.1038/srep28852 Statistical analysis. Statistical analysis of in vitro and in vivo experiments was performed using a nonparametric analysis, Mann-Whitney U test. The tests were carried out with Graph Pad prism version 6.0 e (GraphPad Software, Inc.). In evolutionary analysis, P < 0.1 was regarded as a significant difference as well as HyPhy default setting.