Smoking, pregnancy and the subgingival microbiome

Abstract

The periodontal microbiome is known to be altered during pregnancy as well as by smoking. However, despite the fact that 2.1 million women in the United States smoke during their pregnancy, the potentially synergistic effects of smoking and pregnancy on the subgingival microbiome have never been studied. Subgingival plaque was collected from 44 systemically and periodontally healthy non-pregnant nonsmokers (control), non-pregnant smokers, pregnant nonsmokers and pregnant smokers and sequenced using 16S-pyrotag sequencing. 331601 classifiable sequences were compared against HOMD. Community ordination methods and co-occurrence networks were used along with non-parametric tests to identify differences between groups. Linear Discriminant Analysis revealed significant clustering based on pregnancy and smoking status. Alpha diversity was similar between groups, however, pregnant women (smokers and nonsmokers) demonstrated higher levels of gram-positive and gram-negative facultatives, and lower levels of gram-negative anaerobes when compared to smokers. Each environmental perturbation induced distinctive co-occurrence patterns between species, with unique network anchors in each group. Our study thus suggests that the impact of each environmental perturbation on the periodontal microbiome is unique, and that when they are superimposed, the sum is greater than its parts. The persistence of these effects following cessation of the environmental disruption warrants further investigation.

Figure 1. Dissimilarity in microbial community configuration between the groups.

Linear discriminant analysis of relative abundances of species-level operational taxonomic units (s-OTUs) is shown. The microbial profiles of subjects clustered by pregnancy and smoking status, creating four statistically significant clusters (p < 0.05, MANOVA/Wilks).

Figure 2. Alpha diversity and equitability in the four groups.

Kernel plot of density curves for Shannon Diversity Index in non-pregnant, non-smoking controls, pregnant women, smokers and pregnant smokers are shown in 2A, while the same plots of Shannon Equitability Index are shown in Fig. 2B. The index was not significantly different between groups (p > 0.05, Wilcoxon signed rank test).

Figure 3. Gram staining characteristics and oxygen requirements of species.

Significant differences (p < 0.05, Wilcoxon signed rank test) were observed between non-pregnant, non-smoking controls, smokers and pregnant women when the s-OTUs were stratified based on these criteria. Significance among pairwise comparison is denoted by same alphabets in red on top of bars. There were no significant differences between pregnant smokers and nonsmokers.

Figure 4. Distribution of significant species by groups.

Relative abundances of s-OTUs that were significantly different between groups (p < 0.05, Wilcoxon signed rank test) are shown in Fig. 4A and those that were abundant in each group are shown in Fig. 4B.

Figure 5. Co-occurrence networks in each group.

Non-pregnant, non-smoking control group are shown in Fig. 5A, smokers in 5B, pregnant women in 5C and pregnant smokers in 5D. Each network graph contains nodes (circles sized by relative abundance per group) and edges (lines). Nodes represent species-level OTU’s and edges represent Spearman’s ρ. Edges are colored green for positive correlation and red for negative correlation. Genera contributing to network robustness (anchoring OTUs) were computed as a function of genus degree. Only significant correlations (p < 0.05, t-test) with a coefficient of at least 0.75 are shown. Network anchors are highlighted in brown font and capital lettering.