Prevotella are major contributors of sialidases in the human vaginal microbiome

Abstract

Elevated bacterial sialidase activity in the female genital tract is strongly associated with poor health outcomes including preterm birth and bacterial vaginosis (BV). These negative effects may arise from sialidase-mediated degradation of the protective mucus layer in the cervicovaginal environment. Prior biochemical studies of vaginal bacterial sialidases have focused solely on the BV-associated organism Gardnerella vaginalis. Despite their implications for sexual and reproductive health, sialidases from other vaginal bacteria have not been characterized. Here, we show that vaginal Prevotella species produce sialidases that possess variable activity toward mucin substrates. The sequences of sialidase genes and their presence are largely conserved across clades of Prevotella from different geographies, hinting at their importance globally. Finally, we find that Prevotella sialidase genes and transcripts, including those encoding mucin-degrading sialidases from Prevotella timonensis, are highly prevalent and abundant in human vaginal genomes and transcriptomes. Together, our results identify Prevotella as a critical source of sialidases in the vaginal microbiome, improving our understanding of this detrimental bacterial activity.

Keywords: Prevotella; bacterial vaginosis; mucin; sialidase; vaginal microbiome.

Fig. 1.

Vaginal Prevotella species encode diverse sialidases. (A) Vaginal bacteria sialidase enzymes are hypothesized to remove sialic acids (such as Neu5Ac) from the mucin glycans that comprise the protective mucus layer covering vaginal epithelial cells. (B) Prevotella timonensis CRIS 5C-B1, Prevotella bivia DNF00188, Prevotella denticola DNF00960, and Prevotella amnii CRIS21A-A encode proteins with predicted sialidase domains. Previously characterized Gardnerella sialidases’ protein domains are also displayed for comparison. Individual colors represent different domains and features: signal peptide (red), sialidase domain IPR011040 (green), and carbohydrate binding domain 93 (blue). The amino acid length is displayed for each bar. (C) Prevotella isolates display sialidase activity that correlates with the presence of candidate sialidase genes, indicated by +/–. Bacterial isolates were cultured in PYGT media containing 1 % glucose and 10 % horse serum. Sialidase activity was measured in whole culture samples. Positive control represents the full hydrolysis of 4-methylumbelliferyl N-acetyl-α-D-neuraminic acid (4-MU-Neu5Ac) by Arthrobacter ureafaciens sialidase (AUS). Gardnerella vaginalis ATCC14018 is indicated as a (–) because it encodes NanH1 but not NanH2 or NanH3. G. vaginalis JCP8066 encodes NanH1 and NanH3. Data represent the average ± SD of >5 biological replicates. Significance was assessed using one-way ANOVA followed by the multiple comparisons test, ****P < 0.0001, *P < 0.05. Significance values represent comparison to the media blank.

Fig. 2.

Prevotella sialidases are active at varying pH levels and are inhibited by small molecules. (A) Purified sialidase enzymes hydrolyze 4-MU-Neu5Ac, indicating they possess sialidase activity. For each time point, 2.5 nM enzyme was incubated with 200 µM 4-MU-Neu5Ac in sodium acetate buffer, pH 5.5 at 37 °C. Data represent the average + SEM of three independent experiments. (B) Sialidase activity at varying pH. 200 µM 4-MU-Neu5Ac was prepared in 0.1 M citric acid/0.2 M phosphate buffer at pH values ranging from 3 to 8. Data represent the average ± SEM of three independent experiments. Inhibitory activity of (C) Neu5Ac2en and (D) Zanamivir toward Prevotella and Gardnerella sialidases (IC₅₀ values can be found in SI Appendix, Table S8). Purified sialidases were preincubated with inhibitor for 15 min before adding 4-MU-Neu5Ac to determine activity. Data represent the average ± SEM of three independent experiments.

Fig. 3.

Prevotella sialidases have varied substrate specificities. (A) Quantification of sialidase activity toward different substrates: 4-MU-Neu5Ac, 3′SL and 6′SL, human serum IgG, and human serum IgA. Purified sialidase enzyme (20 to 100 nM) was incubated with each substrate in 20 mM sodium acetate buffer pH 5.5 for 2 h at 37 °C. AUS was used a positive control to release the total sialic acid from each substrate (31). Sialidase activity toward (B) BSM and (C) purified human salivary MUC5B. (D) Structures of Neu5Ac and N-glycolylneuraminic acid (Neu5Gc). Data represent the average ± SEM of three independent experiments. N.D. represents values were not detected below the limit of detection. (E) Sialidase activity toward Neu5Gc from BSM. Data represent the average ± SEM of three independent experiments. Significance was assessed using one-way ANOVA followed by the multiple comparisons test, ****P < 0.0001, *P < 0.05. Significance values represent comparison to the no enzyme control.

Fig. 4.

Sialidases are widely distributed across Prevotella and Gardnerella isolates obtained from the United States and South African vaginal samples. Phylogenetic trees of vaginal isolates from the Vaginal Microbiome Research Consortium (N. American) and FRESH (S. African) studies. (A) Sialidase genes present in individual Prevotella isolate genomes (gene presence is shown as a filled circle). (B) Genome neighborhoods of Prevotella and Gardnerella sialidases. (C) Sialidase genes present in individual Gardnerella isolate genomes. The phylogeny is based on 49 concatenated ribosomal proteins and serves as a proxy for the core genome. The scale bar indicates nucleotides substitutions per site. Each genome was searched using the sialidase protein alignment with HMMER (version 3.1b2). See SI Appendix, Figs. S13 and S15 for the unrooted Prevotella and Gardnerella phylogenetic trees. * = 56% amino acid ID to PtNanH1, ** = 61% AA ID to PtNanH1, *** = 78% amino acid ID to PdNanH, **** = 70% amino acid ID to PbNanH. The orange filled circle indicates a sialidase that is unique from the Prevotella sialidases characterized here.

Fig. 5.

Prevotella sialidase genes and transcripts are prevalent across vaginal community state types. (A) Total sialidase abundance and prevalence in paired MG and (B) MT samples (n = 176). Abundance was determined by Diamond blastX and displayed as reads per kilobase million (RPKM). The cutoff for sialidase prevalence is sialidase reads > 0. Significance was assessed by a one-way ANOVA followed by a Brown–Forsythe and Welch test, **P < 0.0001, **P < 0.01. (C and D) The abundance and prevalence of specific sialidase genes in all available samples (MG; n = 193, MT; n = 188). The figure displays sialidase genes with >15 % prevalence in MG samples. Abundance values were calculated by adding RPKM to a pseudocount (1E8). The diamond indicates the average sialidase expression across all samples ± SD. See complete list of representative sialidase genes in SI Appendix, Fig. S17. (E) Contribution of sialidase abundance within individual vaginal samples across MG and MT. Paired MG and MT were used to investigate the relative contribution of several vaginal sialidases. Stacked bar graphs are aligned to pair the corresponding MG and MT sample.