Exploring the nasopharyngeal microbiota composition in infants with whooping cough: A test-negative case-control study

Abstract

Purpose: The purpose of this study was to characterize the nasopharyngeal microbiota of infants with possible and confirmed pertussis compared to healthy controls.

Methods: This prospective study included all infants <1 year with microbiologically confirmed diagnosis of pertussis attended at a University Hospital over a 12-month period. For each confirmed case, up to 2 consecutive patients within the same age range and meeting the clinical case definition of pertussis but testing PCR-negative were included as possible cases. A third group of asymptomatic infants (healthy controls) were also included. Nasopharyngeal microbiota was characterized by sequencing the V3-V4 region of the 16S rRNA gene. Common respiratory DNA/RNA viral co-infection was tested by multiplex PCR.

Results: Twelve confirmed cases, 21 possible cases and 9 healthy controls were included. Confirmed whooping cough was primarily driven by detection of Bordetella with no other major changes on nasopharyngeal microbiota. Possible cases had limited abundance or absence of Bordetella and a distinctive microbiota with lower bacterial richness and diversity and higher rates of viral co-infection than both confirmed cases and healthy controls. Bordetella reads determined by 16S rRNA gene sequencing were found in all 12 confirmed cases (100%), 3 out of the 21 possible cases (14.3%) but in any healthy control.

Conclusion: This study supports the usefulness of 16S rRNA gene sequencing for improved sensitivity on pertussis diagnosis compared to real-time PCR and to understand other microbial changes occurring in the nasopharynx in children <1 year old with suspected whooping cough compared to healthy controls.

Fig 1. Alpha diversity metrics at the bacterial genus level between confirmed cases, possible cases and healthy controls.

A) Boxplots showing the richness (Observed and Chao1 metric) and diversity (Shannon and Inverse Simpson indices) between the suspected whooping cough group (including confirmed and possible cases) and the healthy controls. B) Boxplots showing the richness and diversity between the confirmed cases, possible cases and the healthy controls. For both, A) and B), significance threshold is set as * p-value < 0.1, ** p-value < 0.05, *** p -value < 0.01.

Fig 2. Nasopharyngeal microbiota bacterial genus composition profiling and differences between whooping cough children and healthy controls.

A) Table showing the top ten abundant bacterial genera based on their mean relative abundance per group (mean ± SD). Color-code is kept through the groups to highlight the most abundant genera. On the right, barplots showing the mean contribution of the ten top bacterial genera per group in relation to the overall gut microbiota composition. B) Boxplots showing the relative abundance of those bacterial genera with a differential abundance between suspected whooping cough and healthy groups based on a Wilcoxon Rank Sum test (p-value significant threshold set to 0.05). Differences between groups are shown as * p-value < 0.1, ** p-value < 0.05, *** p -value < 0.01. C) Boxplots showing the relative abundance of those bacterial genera with an overall differential abundance between confirmed cases, possible cases and healthy groups based on a Kruskal-Wallis test (p-value significant threshold set to 0.05). Further unadjusted pairwise-statistical differences between groups are shown as * p-value < 0.1, ** p-value < 0.05, *** p -value < 0.01.

Fig 3. Increased Bordetella detection sensitivity by 16S rRNA gene sequencing.

A) Barplot showing Bordetella detection by 16S rRNA gene sequencing. Height of bars represent the total number of unique 16S rRNA gene sequences (Amplicon Sequencing Variants, ASVs) matching Bordetella genus per sample. Red/magenta flags at the bottom of the bars represent possible and confirmed cases, respectively. Color-gradient in blue represents total relative abundance of Bordetella genus reads to overall bacterial composition per sample. Black dots show the Bordetella bacterial load detected by real-time PCR (copies/ml). B) Barplots showing the compositon of the overall Bordetella genus abundance at the single ASVs level. The x-axis represent each of the ASVs matching Bordetella genus (n = 55). The y-axis represent their contribution to overall Bordetella genus relative abundance (100%). Each ASV is colored differentently. More detailed color-coding for those ASVs with a major contribution is shown as the figure legend at the bottom.

Fig 4. Bordetella genus is the best, although not unique, bacterial target for whooping cough diagnosis.

A) The random forest model resulting AUC-ROC curve showing its performance to discern between confirmed and possible cases of whooping-cough based on the nasopharyngeal microbiota composition. B) Barplot showing the single variable analysis of the random forest model. The length of bars along the x-axis represents the Mean Decrease Accuracy (MDA) value for each bacterial taxon as a metric for its importance in the model. The MDA significance threshold is set to 1 and variables with an absolute MDA score >2.5 are highlighted. Bacterial taxa are ordered from more to less discriminative and split by the group they correlate to (red for possible cases, magenta por confirmed cases).