Effect of HIV/HAART and Other Clinical Variables on the Oral Mycobiome Using Multivariate Analyses

Abstract

The oral microbiome is considered an important factor in health and disease. We recently reported significant effects of HIV and several other clinical variables on the oral bacterial communities in a large cohort of HIV-positive and -negative individuals. The purpose of the present study was to similarly analyze the oral mycobiome in the same cohort. To identify fungi, the internal transcribed spacer 2 (ITS2) of the fungal rRNA genes was sequenced using oral rinse samples from 149 HIV-positive and 88 HIV-negative subjects that had previously undergone bacterial amplicon sequencing. Quantitative PCR was performed for total fungal content and total bacterial content. Interestingly, samples often showed predominance of a single fungal species with four major clusters predominated by Candida albicans, Candida dubliniensis, Malassezia restricta, or Saccharomyces cerevisiae Quantitative PCR analysis showed the Candida-dominated sample clusters had significantly higher total fungal abundance than the Malassezia or Saccharomyces species. Of the 25 clinical variables evaluated for potential influences on the oral mycobiome, significant effects were associated with caries status, geographical site of sampling, sex, HIV under highly active antiretroviral therapy (HAART), and missing teeth, in rank order of statistical significance. Investigating specific interactions between fungi and bacteria in the samples often showed Candida species positively correlated with Firmicutes or Actinobacteria and negatively correlated with Fusobacteria, Proteobacteria, and Bacteroidetes Our data suggest that the oral mycobiome, while diverse, is often dominated by a limited number of species per individual; is affected by several clinical variables, including HIV positivity and HAART; and shows genera-specific associations with bacterial groups.IMPORTANCE The oral microbiome is likely a key element of homeostasis in the oral cavity. With >600 bacterial species and >160 fungal species comprising the oral microbiome, influences on its composition can have an impact on both local and systemic health. We recently reported significant effects of HIV and several other clinical variables on the oral bacterial community in a large cohort of HIV-positive and -negative subjects. We describe here a comprehensive analysis of the oral mycobiome in the same cohort. Similar to the bacterial community, HIV under highly active antiretroviral therapy (HAART) had a significant impact on the mycobiome composition, but with less impact compared to other clinical variables. Additionally, unlike the oral bacterial microbiome, the oral mycobiome is often dominated by a single species with 4 major clusters of fungal communities. Together, these results suggest the oral mycobiome has distinct properties compared with the oral bacterial community, although both are equally impacted by HIV.

Keywords: Candida; Malassezia; clinical variables; fungal microbiome; human immunodeficiency virus; mycobiome; oral microbiome.

FIG 1

Clustering of oral mycobiome samples. (A) A dendrogram showing hierarchical clustering of fungal communities by Bray-Curtis dissimilarity using the unweighted pair group method with arithmetic mean. The tree is ordered with tighter clusters to the left. Clusters derived by cutting the tree into 8 clusters are shown by the colors on the bar below. (B) Stacked bar chart of relative abundances of species in the samples in the same order as in A, with the most abundant 13 species colored as shown. (C) Total fungal concentration for the samples, colored by HIV status.

FIG 2

Nonmetric multidimensional scaling (NMDS) of fungal communities based on Bray-Curtis dissimilarities stratified by community clusters determined as in Fig. 1. Clusters are identified with the predominant fungi. Ellipses are 95% confidence areas of the group centroids. Clusters 2 and 8 do not have enough members to calculate an ellipse.

FIG 3

Effect of significant clinical variables on the fungal communities. (A to E) The identical NMDS ordination as in Fig. 2 (refer to Fig. 2 for the coordinates of the fungal community clusters on the graph). Ellipses indicate 95% confidence intervals of group centroids. The blue arrows indicate the direction of increase of ordinal variables determined by the envfit function. Directions of effects of categorical variables can be determined by the ellipse positions. (A) Points colored by HIV status. (B) Points colored by sex. (C) Points colored by the clinic location where they were collected. (D) Points sized by caries index, and arrow indicating the direction of increase. (E) Points sized by the number of missing teeth excluding third molars, and arrow indicating the direction of increase. (F) ANOVA table of a distance-based redundancy analysis of the marginal effects of the variables. Model variables were selected by an initial screening by PERMANOVA followed by a stepwise process based on adjusted R² values.

FIG 4

Differences between the four major fungal community clusters. (A) Total fungal content by qPCR versus fungal community cluster. A Kruskal-Wallis test of overall differences was highly significant (P = 2.4 × 10⁻¹²), and bars indicate significant pairwise Mann-Whitney Wilcoxon tests (***, P < 0.001). (B) A contingency table of HIV status versus fungal community cluster. P = 0.027 by Fisher’s exact test. (C) An NMDS of bacterial communities colored by the predominant fungal cluster with ellipses indicating 95% confidence intervals of the group centroid.

FIG 5

Correlation analyses of the most abundant fungal and bacterial species. Fungal and bacterial species that had a mean relative abundance of greater than 1% were selected, and the Spearman correlation over samples was calculated and is shown as on the scale. In the top panel, the correlation was determined based on relative abundance, while in the bottom panel it was based on qPCR-adjusted absolute abundance for both the fungus and bacteria. Interactions where the correlation was not statistically significantly different from zero are marked with Xs. The bacteria are grouped by phylum as indicated.