Dysbiotic shift in the oral microbiota of patients with Alzheimer's disease compared to their healthy life partners-a combinatorial approach and a paired study design

Abstract

Background: The oral microbiota has been associated with Alzheimer's disease (AD). However, earlier studies provided conflicting results using varying sampling methods, sequencing techniques, and statistics, as well as independent subjects.

Methods: To robustly identify disease-associated microbial features, we recruited patients and their healthy life partners from the same households sharing a more similar microbiota compared to independent individuals increasing statistical power via paired design and combined three different sequencing methods - including metagenomics-and several bioinformatic pipelines. We recruited 26 AD-patients and their life partners. Salivary and supragingival samples were collected and a clinical examination of the mouth was performed.

Results: Both groups showed comparable oral health. By focusing primarily on recurrently identified species across the different datasets we were able to identify a Core dysbiosis. This Core dysbiosis surprisingly spares the most central of oral diseases pathogens, namely Porphyromonas gingivalis. However, it includes numerous other species commonly associated with oral pathologies such as Prevotella nigrescens, Streptococcus anginosus, Dialister invisus, Anaeroglobus geminatus, Olsenella uli and Mogibacterium timidum. In contrast, more host-compatible species such as Prevotella melaninogenica or Streptococcus parasanguinis are identified in controls.

Conclusions: This is the first study using a combined sequencing approach and a paired study design to identify robust features of the oral microbiota of AD-patients. Although promising, the results should nevertheless be interpreted with caution, as the cross-sectional study design limits the possibilities of interpretation, and larger, longitudinal data are necessary for causal conclusions. However, this combined approach on multiple processing levels to identify intra-partnership differences still offers the possibility to better identify disease-associated microbial features potentially involved in AD-pathogenesis.

Trial registration: This study was prospectively registered at the German Clinical Trials Register (DRKS00023456) at the 30th of November 2020.

Keywords: Alzheimer; Dementia; Metagenomic; Next generation sequencing; Oral microbiota; Periodontal disease; Spouses.

Fig. 1

Flowchart illustrating the overall workflow of the study divided by patient recruitment and sampling, laboratory work and preprocessing and the final bioinformatic processing of the data. If not shown otherwise all samples were included in the according analysis step. Abbreviations: AD, Alzheimer's disease; ALDEx2, ANOVA-Like Differential Expression 2; ANCOM-BC2, Analysis of Compositions of Microbiomes with Bias Correction 2; Bracken, Bayesian Reestimation of Abundance with KrakEN; CSF, Cerebro-spinal-fluid; DNA, Deoxyribonucleic acid; FDI, Fédération Dentaire Internationale; gDNA, Genomic DNA; GLMMMiRKAT, Generalized Linear Mixed Model Microbiome Regression-based Kernel Association Test; LEfSe, Linear Discriminant Analysis Effect Size; Limma, Linear Models for Microarray; MMSE, Mini Mental Status Examination; NIA-AA, National Institute of Aging and Alzheimer’s Association; PCoA, Principal Coordinate Analysis; PERMANOVA, Permutational Multivariate Analysis Of Variance; SBI, Sulcus bleeding index; SRS, Scaling with Ranked Subsampling; Supra, Supragingival; V3/V4, Variable region 3/4

Fig. 2

Mean relative abundances of the salivary samples for the genus- (a) and phylum-level (b). The columns represent the 16S full-length (FL), 16S short-read (SR) and metagenomic (WGS) sequencing results divided by patients and controls. Taxa with a mean abundance below 1% are summed as Others_1%. Abbreviations: FL, full-length; SR, short-read; WGS, whole-genome-sequencing

Fig. 3

Bray–Curtis based PCoAs of metagenomic (a), 16S full-length (b), and 16S short-read (c) sequencing of salivary samples. Patients are shown in blue, and controls in yellow. The samples from the control group show a tighter grouping across all three methods while the samples from the patient group are more widely dispersed. (d) P values for PERMANOVA, Betadispersion, as well as GLMMMiRKAT (Bray–Curtis and Omnibus) testing are presented in the table. Statistically significant differences (p < 0.05) are indicated by *. Abbreviations: PCoA, Principal Coordinate Analysis; PERMANOVA, Permutational Multivariate Analysis Of Variance; GLMMMiRKAT, Generalized Linear Model Microbiome Regression-based Kernel Association Test

Fig. 4

Bacterial species identified as statistically significant by LEfSe in the metagenomic (a), 16S full-length (b), and 16S short-read (c) sequencing analysis of salivary samples. Taxa that have achieved an LDA score of at least 2 or −2 in the linear discriminant analysis are shown. Species with a higher abundance in the patient group are shown in green and for the control group in red. Abbreviations: LEfSe, Linear Discriminant Analysis Effect Size; LDA, Linear Discriminant Analysis

Fig. 5

Bacterial species identified as statistically significant by ANCOM-BC2 and without association to number of teeth or SBI in the metagenomic (a), 16S full-length (b) and 16S short-read (c) sequencing analysis of salivary samples. Bars with whiskers represent log-fold-change with standard error. Taxa with a higher abundance in the patient group are shown in blue and for the control group in yellow. Abbreviations: ANCOM-BC2, Analysis of Compositions of Microbiomes with Bias Correction 2; SBI, Sulcus bleeding Index

Fig. 6

Mean relative abundances of the supragingival samples for the genus- (a) and phylum-level (b). The columns represent the 16S full-length (FL) and 16S short-read (SR) sequencing results divided by patients and controls. Taxa with a mean abundance below 1% are summed as Others_1%. Abbreviations: FL, full-length; SR, short-read

Fig. 7

Bray–Curtis based PCoAs of 16S full-length (a), and 16S short-read (b) sequencing of supragingival samples. Patients are shown in blue, and controls in yellow. For the 16S full-length sequencing the samples from the control group show a slightly tighter grouping while the samples from the patient group are more widely dispersed. For the 16S short-read sequencing the two groups show a comparable distribution but a clearer separation of the two groups. (c) P values for PERMANOVA, Betadispersion, as well as GLMMMiRKAT testing (Bray–Curtis and Omnibus) are presented in the table. Statistically significant differences (p < 0.05) are indicated by *. Abbreviations: PCoA, Principal Coordinate Analysis; PERMANOVA, Permutational Multivariate Analysis Of Variance; GLMMMiRKAT, Generalized Linear Model Microbiome Regression-based Kernel Association Test

Fig. 8

Bacterial species identified as statistically significant by LEfSe in the 16S full-length (a), and 16S short-read (b) sequencing analysis of supragingival samples. Taxa that have achieved an LDA score of at least 2 or −2 in the linear discriminant analysis are shown. Species with a higher abundance in the patient group are shown in green and for the control group in red. Abbreviations: LEfSe, Linear Discriminant Analysis Effect Size; LDA, Linear Discriminant Analysis

Fig. 9

Bacterial species identified as statistically significant by ANCOM-BC2 and without association to number of teeth or SBI in the 16S full-length (a) and 16S short-read (b) sequencing analysis of supragingival samples. Bars with whiskers represent log-fold-change with standard error. Taxa with a higher abundance in the patient group are shown in blue and for the control group in yellow. Abbreviations: ANCOM-BC2, Analysis of Compositions of Microbiomes with Bias Correction 2; SBI, Sulcus bleeding Index

Fig. 10

Venn-diagrams showing the overlap in taxon identification for the salivary samples between the results of the metagenomic (WGS), 16S full-length (FL) and 16S short-read (SR) sequencing. Presented are the results for different taxonomic levels and filtering approaches: A and b show the species-level with the overlap of the unfiltered dataset (a) and the dataset filtered by a minimum relative abundance of 0.1% (b). C and d show the equivalent results for the genus-level. Abbreviations: FL, full-length; SR, short-read; WGS, whole-genome-sequencing

Fig. 11

Venn-diagrams showing the overlap in taxon identification for the supragingival samples between the results of the 16S full-length (FL) and 16S short-read (SR) sequencing. Presented are the results for different taxonomic levels and filtering approaches: A and b show the species-level with the overlap of the unfiltered dataset (a) and the dataset filtered by a minimum relative abundance of 0.1% (b). C and d show the equivalent results for the genus-level. Abbreviations: FL, full-length; SR, short-read

Fig. 12

Heatmaps visualizing the relative abundances of the species (a) and genera (b) of the Core dysbiosis in the salivary samples divided by patients and controls, as well as sequencing methods. Taxa with higher abundances in patients, respectively controls are marked with a P and C. Abbreviations: FL, full-length; SR, short-read; WGS, whole-genome-sequencing

Fig. 13

Heatmaps visualizing the relative abundances of the species (a) and genera (b) of the Core dysbiosis in the supragingival samples divided by patients and controls, as well as sequencing methods. Taxa with higher abundances in patients, respectively controls are marked with a P and C. Abbreviations: FL, full-length; SR, short-read