Oral microbiome profiles relate periodontal disease and brain health - the PAROMIND Study

Oral microbiome profiles relate periodontal disease and brain health - the PAROMIND Study | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Oral microbiome profiles relate periodontal disease and brain health - the PAROMIND Study Marvin Petersen, Carolin Walther, Katrin Borof, Guido Heydecke, and 13 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6580781/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract The relationship between oral microbiome composition and brain health in the general population remains poorly understood. In this study, we inferred a microbiome similarity network based on 16S rRNA sequencing data of crevicular fluid collected from 1,026 participants in the Hamburg City Health Study, which revealed a continuous disease gradient mirroring the microbial pathogenicity spectrum of periodontitis. Leveraging this network, we systematically examined associations between periodontal microbiome profiles and 37 brain health-related phenotypes, including cognitive function, brain structure, mental health, inflammatory biomarkers, diet, vascular risk factors, and demographics. Higher abundance of periodontitis-related microbial taxa was linked to poorer cognitive performance, elevated leukocyte counts and lower MIND diet adherence after covariate adjustment, but no significant associations were found for the remaining brain health phenotypes. Notably, we identified both previously known as well as novel microbial associations with brain health phenotypes. These findings advance the understanding of the oral microbiome-brain axis, highlighting potential pathways connecting periodontal health and brain function with potential implications for future causal and interventional studies. Biological sciences/Microbiology/Bacteria Health sciences/Neurology/Neurological disorders Oral microbiome Periodontitis Brain health Cognition Brain structure Neuroimaging Mental health Inflammation Nutrition Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction The human oral microbiome harbors a diverse community of microorganisms that influences the health and well-being of their hosts. 1 In recent years, periodontitis, which is the inflammatory disruption in the host–microbial homeostasis of the periodontal pocket, has gained increasing attention as a key factor impacting brain health. Studies indicate that periodontitis and the linked bacterial communities are associated with the incidence of cognitive decline and Alzheimer’s dementia. 2 – 7 Given its estimated prevalence between 10% and 50% in elderly people, this renders periodontitis a relevant public health concern. 8 As therapeutic interventions can alter the progression of periodontitis, comprehending its impact on the brain is vital for effective prevention and management of cognitive sequelae. Among bacterial species of the subgingival biofilm, several are known to be associated with periodontitis including: Porphyromonas gingivalis , Tannerella forsythia and Treponema denticola . 9 , 10 Mechanistic models have been proposed to explain the connection between these bacterial communities and cognitive decline. The connection is considered to arise from bacterial species promoting systemic inflammation, neurodegenerative processes, and blood-brain barrier disruption. 11 – 14 Despite existing research efforts, our understanding of the association between oral microbiome composition and brain health remains limited. While recent studies have predominantly focused on clinical Alzheimer’s disease and mild cognitive impairment cohorts, 7 , 15 – 19 understanding these connections in the general population requires research within broader community-based settings, which remains relatively limited. 20 Additionally, many investigations are constrained by small sample sizes and a lack of multi-modal imaging and biomarker data potentially leading to inconsistent findings. 7 These issues are further complicated by the inherent complexity of microbiome data. Tapping into these research needs, we aim to advance the understanding of the oral microbiome-brain axis by applying advanced analysis techniques capable of integrating the large-scale, multi-domain data crucial for achieving robust insights. Specifically, our primary hypothesis was that subgingival microbial profiles indicative of periodontitis would be associated with variations in brain health-related host phenotypes within a population-based cohort. To investigate this hypothesis, we analyzed population-based data of n = 1,026 individuals from the Hamburg City Health Study. 21 Our approach focuses on a topology-based analysis that integrates abundance data of the subgingival microbiome derived from 16S rRNA sequencing of subgingival samples with in-depth clinical and lifestyle data including oral health assessments, cognitive test scores, mental health scores, neuroimaging, circulating inflammatory markers, dietary patterns, and vascular risk measures. Results Our methodology, which integrates microbiome, clinical and lifestyle data into a unified analysis framework, is illustrated in Fig. 1 . Quality assessment of subgingival microbiome abundance and brain MR imaging data resulted in a final analysis sample of 1,026 individuals ([mean ± SD] age 63.72 ± 8.2 years, 42.7% female; for details see Table 1 ). Table 1 Sample characteristics Metric Value a Age, years 63.72 ± 8.17 (1,026) [range: 46–78] Female, n (% female) 438 (42.7) Education, ISCED 2.42 ± 0.58 (999) Oral health measures Clinical attachment loss 2.60 ± 0.83 (842) Plaque index 19.22 ± 26.30 (830) Bleeding on probing index 15.38 ± 19.91 (1,011) DMFT index 19.27 ± 4.90 (842) Cognitive scores Animal Naming Test 24.66 ± 6.67 (977) Mini Mental State Exam 27.85 ± 1.68 (973) Multiple Choice Vocabulary Intelligence Test 31.42 ± 3.43 (838) Trail Making Test A, sec 39.82 ± 14.07 (925) Trail Making Test B, sec 89.44 ± 37.66 (917) Word List Recall 7.70 ± 1.88 (949) Mental health scores PHQ-9 3.21 ± 3.61 (1,025) PHQ-15 4.27 ± 3.90 (1,025) Geriatric Depression Scale 1.83 ± 2.20 (914) GAD-7 2.41 ± 2.90 (1,025) Global neuroimaging features Free-water (gray matter) 0.50 ± 0.05 (909) Free-water (white matter) 0.31 ± 0.03 (909) Tissue fractional anisotropy (gray matter) 0.19 ± 0.01 (909) Tissue fractional anisotropy (white matter) 0.41 ± 0.01 (909) Mean cortical thickness, mm 2.63 ± 0.11 (986) Mean subcortical volume, ml 3188.31 ± 313.66 (986) Inflammatory markers hsCRP 0.25 ± 0.45 (970) Leukocytes 6.13 ± 1.75 (1,006) Diet MEDAS score 4.46 ± 1.89 (970) DASH score 4.46 ± 1.06 (970) MIND score 6.30 ± 1.78 (970) Vascular risk measurements Systolic blood pressure 142.12 ± 20.06 (997) Diastolic blood pressure 83.46 ± 10.55 (997) Body mass index 26.60 ± 4.32 (995) Smoking, % 173 (16.9) Triglycerides 117.67 ± 75.68 (1,007) Cholesterol 208.28 ± 40.83 (1,008) LDL 121.56 ± 36.53 (997) HDL 63.51 ± 19.00 (1,008) HbA1c 5.63 ± 0.54 (1,007) Abbreviations: DMFT index = decayed/missing/filled teeth index, HDL = high density lipoprotein, hsCRP = high sensitivity c-reactive peptide, LDL = low density lipoprotein, ISCED = International Standard Classification of Education, mm = millimeters, sec = seconds a Presented as median [IQR] (N) A similarity network of subgingival bacterial communities We applied an unsupervised, topology-based technique to abundance data of 85 different bacterial genera to infer a low-dimensional network representation of oral microbiome similarity. The resulting microbiome similarity network consisted of 577 nodes – representing participant groups with highly similar subgingival microbiome compositions – and 10,230 edges – connecting nodes that share at least one participant. The network represents a map of inter-individual variability capturing transitions in microbial composition across the cohort. For a visualization of the network see Fig. 2 a. In an enrichment analysis, we examined whether the microbiome similarity network captures inter-individual differences in participant traits. Therefore, we computed enrichment scores, which are node-level indices that indicate whether participant traits are significantly higher or lower than expected by chance in specific regions, i.e., participant clusters, of the network. Enrichment scores were obtained for genus-level abundance data as well as all non-microbiome phenotypes. Furthermore, we calculated the enrichment ratio (= number of significantly enriched nodes / total number of nodes) based on the enrichment scores for each phenotype quantifying the overall amount of enrichment on the microbiome similarity network. This measure indicates how strongly the network-based organization of participants reflects inter-individual variance in a respective phenotype. Refer to Fig. 2 b and 2 c for an exemplary display of Porphyromonas abundance and the corresponding enrichment map on the microbiome similarity network. Enrichment analysis of genus-level abundance Microbiome abundance showed significant enrichment across all detected genera ranging from 42.5–87.7%. Figure 3 a displays the enrichment ratios alongside mean relative abundance and phylogenetic associations. We performed a dominance analysis by identifying for each network node the bacterial genus with the highest enrichment scores (Fig. 3 b). Of the 85 investigated genera, 15 showed the highest enrichment score for at least 5 nodes. Distribution of dominant genera in the network representation of oral microbiome similarity followed a horizontal pathogenicity gradient: Bacterial genera with strongest enrichments at the left end of the microbiome similarity network were periodontitis-associated taxa including Fusobacterium (n nodes = 208), Campylobacter (65), Treponema (13), Dialister (8), Saccharibacteria (TM7) [G-5] (6) and Porphyromonas (5). In the center of the network, Aggregatibacter (18), Gemella (8), Capnocytophaga (6) and Leptotrichia (5) exhibited highest enrichment scores. At the right end, genera with strongest enrichments were of low periodontal pathogenicity or related to other dental diseases including Streptococcus (126), Veillonella (42), Neisseria (26), Rothia (17) and Haemophilus (7). Enrichment analysis of non-microbiome phenotypes All non-microbiome phenotypes showed significant enrichment on the microbiome similarity network including measures of oral health status, cognition, mental health, brain structure, circulating inflammatory markers, diet, vascular risk factors and demographics (Fig. 4 ). The top 10 ranking non-microbiome phenotypes were leukocytes (78.0%), plaque index (77.3%), verbal fluency (75.4%), general cognitive ability (74.7%), smoking behavior (74.5%), Mini Mental State Exam (73.8%), bleeding on probing (BOP) index (72.6%), clinical attachment loss (69.5%), Memory (69.2%) and age (68.6%) (Fig. 4 a). Complementary microbiome covariate identification using linear approaches ( envfit , adonis , and capscale ) confirmed these results. Among the phenotypes with the top 10 highest enrichment ratios, all showed a significant linear association (p FDR < 0.05) with subgingival microbiome profiles across all three complementary linear approaches ( supplementary figure S1 ) . Based on the enrichment maps, we assessed inter-participant differences of phenotypes. Phenotype enrichment maps are presented in Fig. 4 b. Most participant traits varied along the microbiome similarity network in a linear left-right trajectory aligning with the pathogenicity gradient identified via dominance analysis. Put differently, there was a gradient of enrichment from left to right that reflected the transition of participant characteristics. Participants at the left end of this gradient exhibited higher severity of clinical periodontitis, elevated circulating inflammatory markers, older age, a higher percentage of smokers (indicated by significant positive enrichment), as well as lower cognitive function, lower brain structural integrity, a less healthy diet, a lower percentage of females, and lower education levels (indicated by significant negative enrichment). Conversely, the right end of the gradient featured participants with opposite traits: lower severity of clinical periodontitis, lower circulating inflammatory markers, younger age, fewer smokers, as well as higher cognitive performance, higher brain structural integrity, a healthier diet, a higher percentage of females, and higher education levels. Mental health scores and vascular risk factors beyond smoking displayed a non-linear enrichment pattern not aligned with the pathogenicity gradient. Co-enrichment analysis Principal component analysis of enrichment scores revealed that enrichment patterns of phenotypes differed along two dominant axes of inter-phenotype variation explaining 55.69% (principal component 1, PC1) and 17.49% (principal component 2, PC2) of variance, respectively. Phenotypes with similar enrichment patterns co-localized in the principal component space formed by these axes (Fig. 5 , supplementary figure S2 ). Phenotypes with enrichment patterns indicating increasing values from right end of the network to left were localized on the left extreme of the principal component space including periodontitis-associated bacterial genera, clinical oral health measurements, circulating inflammatory markers, smoking behavior and white matter free-water measured by diffusion-weighted MRI. Phenotypes increasing left to right were located on the right including health-associated bacterial genera, cognitive performance measurements, diet scores, cortical thickness and subcortical volume as measured by MRI. Phenotypes showing a non-linear enrichment pattern or transition pattern in the left-right right-left orientation were localized in the middle including mental health scores and vascular risk measurements apart from smoking. For a heatmap depicting the Spearman correlation of enrichment scores for genus-level abundance and non-microbiome phenotypes see Fig. 6 . For the same co-enrichment heatmap indicating genus-genus correlation as well as correlation between non-microbiome phenotypes see supplementary figures S3 and S4 . Microbiome similarity groups differ in periodontal disease, inflammatory markers, cognition and diet Applying k-means clustering on the microbiome similarity network topology resulted in an unsupervised separation of the sample into two participant groups along the pathogenicity gradient. The following statistics to characterize the identified participant groups were adjusted for potential confounders including age, sex, education and vascular risk factors. The procedure is illustrated in Fig. 7 a. Participants assigned to nodes in both groups were excluded from the analysis (n excluded =137). We compared genus-level abundance to identify bacterial genera that showed significant differences between groups (Fig. 7 b and 7 c). Of the 85 tested genera, 8 showed no significant group differences ( supplementary figure S5 ). Top 15 bacterial genera which were significantly higher in the “A” group were Treponema (β std = -1.26), Fusobacterium (β std = -1.26), Fretibacterium (β std = -1.21), Tannerella (β std = -1.09), Dialister (β std = -1.01), Mogibacterium (β std = -0.98), Porphyromonas (β std = -0.96), Peptostreptococcaceae [G-4] (β std = -0.93), Peptostreptococcaceae [G-2] (β std = -0.91), Parvimonas (β std = -0.87), Veillonellaceae [G-1] (β std = -0.85), Peptostreptoccocaceae [G-9] (β std = -0.85), Prevotella (β std = -0.85), Catonella (β std = -0.83), and Desulfobulbus (β std = -0.83) (all p FDR < 0.001). Top 15 bacterial genera which were significantly higher in the “B” group were Streptococcus (β std = 1.37), Haemophilus (β std = 1.16), Rothia (β std = 1.02), Granulicatella (β std = 0.98), Gemella (β std = 0.93), Riemerella (β std = 0.86), Actinomyces (β std = 0.82), Veillonella (β std = 0.81), Neisseria (β std = 0.75), Kingella (β std = 0.75), Capnocytophaga (β std = 0.75), Lautropia (β std = 0.75), Corynebacterium (β std = 0.69), Arachnia (β std = 0.59), Cardiobacterium (β std = 0.56) (all p FDR < 0.001). In addition, the “A“ group showed significantly higher clinical severity of periodontitis, higher circulating inflammatory markers, lower cognitive performance and lower brain structural integrity compared to the “B” group (Fig. 7 d): a significantly higher clinical attachment loss (β std = -0.56, p FDR < 0.001), DMFT index (β std = -0.15, p FDR = 0.047), plaque index (β std = -0.32, p FDR < 0.001), bleeding on probing index (β std = -0.53, p FDR < 0.001), leukocytes (β std = -0.21, p FDR = 0.014) as well as a significantly lower general cognitive ability (β std = 0.18, p FDR = 0.026), Mini Mental State Exam score (β std = 0.22, p FDR = 0.014) as well as MIND score (β std = 0.20, p FDR = 0.026). The remaining cognitive scores, diet scores and hsCRP showed no significant differences ( supplementary figure S6 ). Moreover, brain structural indices and mental health scores also showed no significant differences. Sensitivity analysis To assess the robustness of our result we performed two-fold sensitivity analyses. First, we conducted repeated enrichment analyses on randomly selected subsamples, reducing the sample size from 100–10% in 1% decrements, with 100 random samples at each step ( supplementary figure S7 ). Throughout this process, the Spearman correlation of enrichment ratios (covering both non-microbiome and microbiome phenotypes) demonstrated high robustness (Spearman ρ > 0.8) to different sample compositions until the sample size was reduced to approximately 80%. Additionally, the Adjusted Rand Index (ARI) remained high (ARI > 0.8), indicating strong consistency in microbiome similarity network-based group assignments from k-means clustering, until the sample size was reduced to approximately 70%. Second, to evaluate the potential influence of pipeline design choices on our findings, we reanalyzed the data using 17 different pipeline configurations, varying the components and parameters of the topological data analysis ( supplementary table S8 ). The results demonstrated high robustness, with Spearman correlations between enrichment ratios from the original and alternative configurations averaging 0.82 ± 0.03. Furthermore, the ARI for group assignments derived from k-means clustering via the original pipeline, compared to those from different configurations, remained consistently high at 0.84 ± 0.06. Discussion The oral cavity hosts the second most diverse microbiome in the human body, following the gut. 22 Interactions within these oral microbial communities, along with host factors, affect both oral and systemic health. 23 In this study, we present a comprehensive population-level analysis examining the association between the subgingival microbiome and multi-domain brain health-related phenotypes ranging from cognitive functions, mental health, brain structure, inflammatory blood biomarkers, to dietary behavior and vascular risk factors. Using 16S rRNA sequencing and topological data analysis, we inferred a microbiome similarity network that revealed a continuous pathogenicity gradient, mapping individuals based on periodontal microbiome profiles. Leveraging this network, we identified associations of periodontitis-related microbial communities with multiple brain health phenotypes: Individuals with higher abundance of periodontitis-related taxa exhibited significantly lower cognitive performance, lower MIND diet scores as well as increased leukocyte counts adjusting for demographics and cardiovascular risk. These findings were robust, as confirmed by sensitivity analyses involving diverse random subsamples and varying analytical configurations. Our findings highlight several key insights into the oral microbiome-brain axis, potential pathophysiological pathways and clinical implications. Topological data analysis reveals a latent axis of periodontitis-related microbial composition We performed a dominance analysis to determine which bacterial genera are particularly abundant within specific participant clusters of the microbiome similarity network. Based on this analysis, we show that the microbiome similarity network captured interindividual variations and a gradient of microbial compositions from periodontitis-associated taxa to health-associated taxa (Fig. 3 b): Periodontitis-associated genera such as Porphyromonas , Fusobacterium , Treponema, Saccharibacteria (TM7) and Campylobacter were enriched at the left part of the network, consistent with their established roles in biofilm formation, immune modulation, and periodontal tissue destruction. 9 , 24 – 27 Also positioned at this pathogenic end were bacteria of the genus Dialister , which have only recently been associated with periodontitis and could be a key periodontal pathogen, warranting further research into its mechanisms. 28 In contrast, genera such as Streptococcus , Haemophilus , Rothia, Veillonella and Neisseria were enriched at the right part, reflecting their roles in maintaining an oral health equilibrium, their overall low periodontal pathogenicity or their association with oral conditions different from periodontitis, such as dental caries. 29 – 31 Enrichment ratios for bacterial genera were overall high, affirming that the observed interindividual differences were relevantly captured by the microbiome similarity network. These findings indicate that periodontitis emerged as the most parsimonious explanation for the topology of the microbiome similarity network, with participants harboring higher abundances of periodontitis-related taxa clustering at one end, and those with higher abundances of other genera clustering at the other. Put differently, the identified network revealed a latent axis and continuum of periodontitis (referred to as pathogenicity gradient) mirroring the pathogenicity spectrum of periodontitis and suggesting that the disease constitutes a key driver of the observed interindividual variance of microbiome compositions. This evidence demonstrates that a topological data analysis-based approach can uncover subtle yet biologically meaningful patterns of disease severity in large-scale, highly complex datasets. Microbial compositions are linked to non-microbiome phenotypes The microbiome similarity network offers a means to integrate complex, high-dimensional microbiome data with comprehensive health information by mapping non-microbial phenotypes on the network and statistically testing their association with the network structure. We observed that the investigated non-microbiome phenotypes differentially enriched on the network, implying varying degrees of association with microbial configuration. These enrichment patterns ranged from linear enrichments correlated or anti-correlated with the pathogenicity gradient (e.g., leukocytes) – where variance in these variables coincided with changes in the abundance of periodontitis-related taxa – to non-linear patterns indicating weaker or more complex associations with the pathogenicity gradient (e.g., PHQ-9) (Fig. 4 b). Co-enrichment analyses further detailed these observations, showing that microbial and non-microbial phenotypes with visually overlapping enrichment patterns occupied similar positions in principal component space and exhibited strong correlations in their enrichment scores (Figs. 5 and 6 ; supplementary figures S2 -S4 ). Among the phenotypes most closely aligned with the pathogenicity gradient were the plaque index, clinical attachment loss, and bleeding on probing index, reinforcing periodontitis as the primary pathology captured by the network. 32 , 33 These findings highlight not only the complex covariance structure of subgingival microbiome composition but also the potential of genus-level abundance data as a promising biomarker source, reflecting a broad range of clinical and lifestyle factors. Periodontal dysbiosis and brain health The mechanisms behind the relationship between periodontitis and brain health are multifaceted and remain to be fully elucidated. Our analysis sought to address these complexities by integrating data on multiple brain health phenotypes into a single analysis framework. Notably, multiple markers of cognitive performance (verbal fluency, general cognitive ability, Mini Mental State Exam, Memory) and systemic inflammation (leukocyte counts, hsCRP) were among non-oral phenotypes with strongest enrichment ratios (Fig. 4 a) and varied in tandem with the pathogenicity gradient from one network end to the other (Fig. 4 b). Specifically, participants with higher abundance of periodontitis-associated taxa and poorer oral health exhibited significantly lower cognitive performance and elevated systemic inflammation after adjusting for covariates (Fig. 7 c and 7 d). Importantly, these findings build on prior reports linking shifts in oral microbiota to cognitive changes among dementia patients and demonstrate that such associations also manifest in a healthy population of non-demented, middle-aged individuals. 18 , 34 Furthermore, our findings align with earlier analyses indicating relationships between clinical markers of periodontitis and cognitive health. 35 – 38 Our analysis highlights associations between brain health and the abundance of various bacterial genera (Fig. 7 b and 7 c). These findings are consistent with prior research implicating genera such as Porphyromonas , Treponema , Fusobacterium , and Prevotella in pathways potentially relevant to Alzheimer’s disease. Prominently, experimental evidence from animal models demonstrated that Porphyromonas gingivalis contributes to Alzheimer’s pathology by contributing to the formation of amyloid-β, neurofibrillary tangles and neuroinflammation. 11 , 13 Furthermore, postmortem studies report a higher abundance of Porphyromonas gingivalis in the brain samples of AD patients compared to non-demented controls, with intracerebral presence of the bacterium being related to a six-fold increased risk of AD. 11 , 39 Treponema denticola has been demonstrated to induce tau hyperphosphorylation and neuroinflammatory processes in mice. 40 The serum levels of antibodies against Prevotella intermedia and Fusobacterium nucleatum are significantly increased in Alzheimer's disease patients. 15 Additionally, elevated Prevotella intermedia abundance is associated with APOE4 -carrier status, and Fusobacterium species have been linked to cerebrovascular lesions indicative of small vessel pathology. 17 , 41 Our findings align with these results, indicating that the prevalence of these pathogenic bacteria in the subgingival biofilm corresponds to reduced brain health in cognitively normal individuals. At the same time, we found genera not yet reported in relation to brain health – Fretibacterium , Tannerella , Dialister , Mogibacterium , Peptostreptococcaceae and Parvimonas – warranting further experimental research to determine whether they independently drive cognitive deficits or merely coincide with other pathogenic species. Our findings emphasize the role of periodontitis-related taxa in systemic inflammation and possible connections to neurodegenerative pathways. We observed that individuals positioned at the pathogenic end of the oral microbiome similarity network exhibited significantly higher leukocyte levels after covariate adjustment (Fig. 7 d), indicating presence of systemic inflammation. This aligns with prior studies linking clinically diagnosed periodontitis to elevated systemic inflammatory markers. 42 Pathomechanistically, such responses can stem from dense leukocytic infiltration during gingival inflammation, the stimulation of bone marrow for sustained inflammatory cell production, or the systemic response to dissemination of pathogens from ulcerated gingival tissues. 42 Importantly, we present novel evidence linking periodontitis-associated microbial compositions to systemic inflammation. Notably, this systemic inflammation was accompanied by reduced cognitive performance, supporting the hypothesis that chronic oral infections contribute to both systemic inflammation and cognitive disease. Although our findings cannot prove causality due to the cross-sectional nature of the study design, they align with the inflammatory hypothesis of Alzheimer’s disease, suggesting that immune responses triggered by oral pathogens may disrupt the blood-brain barrier and activate microglia, thereby driving neurodegeneration. 43 Previous analyses also highlight an association between periodontal health and brain structure. 20 , 44 , 45 However, the global brain structural measures we assessed did not show significant group differences after adjustment for covariates. We speculate that this could be because any potential structural effects related to the oral microbiota in this cohort are subtle or regionally specific, rather than global, and thus not captured by the overall brain metrics used. Future analysis should therefore focus on assessing localized brain structural changes, potentially using voxel-based or region-of-interest methods, to determine if subtle or regionally specific associations exist with oral microbial profiles. In contrast to cognitive measures, mental health scores displayed weaker associations with the pathogenicity gradient, suggesting more nuanced or less direct links with periodontitis-related microbiome composition. These associations did not persist after adjusting for demographic and cardiovascular risk factors. We interpret these results as pointing to a more specific link of periodontitis to cognitive rather than mental health. Nevertheless, prior studies have reported connections between anxiety, depression, and clinical periodontitis or oral microbiome composition. 46 Longitudinal research will be essential to elucidate how these factors collectively influence oral and brain health. Influence of diet and vascular risk factors on oral and brain health We investigated the role of nutritional behavior as a key factor influencing oral microbiome composition. 47 Enrichment analysis revealed that adherence to cognitively beneficial diets, such as the MEDAS, MIND, and DASH diets, 48 – 50 aligned with the pathogenicity gradient: participants with greater abundance of periodontitis-related taxa tended to adhere less to these diets. Notably, after adjusting for covariates, individuals with a higher abundance of periodontitis-related genera exhibited significantly lower adherence to the MIND diet. These findings corroborate previous results from the HCHS, which demonstrated that higher adherence to DASH, MEDAS and an anti-inflammatory dietary was associated with lower odds of periodontal disease. 51 Our findings suggest that dietary patterns known to promote cognitive health may also shape the oral microbiome toward a less pathogenic composition. Given that the MIND diet specifically emphasizes foods with anti-inflammatory and neuroprotective properties, 52 it is plausible that such nutritional patterns could influence both oral and brain health through microbial community shifts and immune modulation. A healthier diet may foster a healthier oral microbiome based on its emphasis on fiber over simple sugars selectively promoting beneficial microbes while limiting resources for harmful ones. 53 Furthermore, it could reduce gingival inflammation through anti-inflammatory compounds, making the environment less hospitable to pathogens potentially disrupting brain health. 54 Longitudinal and interventional studies are needed to disentangle these pathways and clarify causal relationships between diet, oral microbial ecology, and brain health. Among demographic and vascular risk factors, smoking behavior was the only factor clearly associated with the pathogenicity gradient. Smoking has been documented to promote the colonization of periodontal pathogens, facilitate biofilm formation, and compromise the immune response, thereby exacerbating periodontal disease and systemic inflammation. 55 Our findings highlight the role of smoking on promoting dysbiosis and inflammation and positions it as a direct contributor to the pathogenicity gradient and brain health phenotypes. Clinical implications Our findings not only shed light on the intricate systemic pathophysiology of periodontal dysbiosis and brain health, but they also hint at potential avenues of clinical utilization. Our study reveals an association between subgingival microbiome signatures and variance in cognitive performance and brain structure that could enhance early screening measures for risk of cognitive decline and improve targeted recruitment of individuals at critical early stages of cognitive impairment and dementia. Particularly, identifying microbial community changes before emergence of cognitive symptoms could enhance diagnostics by providing microbiome-based markers as a convenient complement to existing dementia biomarkers. Regarding future therapeutic interventions, microbiome signatures of early periodontal dysbiosis may guide the development of oral microbiome-directed therapies to slow or prevent dementia progression. 56 However, the definitive role of microbiome biomarkers in cognitive disorders is yet to be determined and large-scale longitudinal as well as interventional studies are required for moving in this direction. Strengths and limitations Strengths of this work lie in its considerable sample size; comprehensive taxonomic profiling of the oral microbiome; in-depth phenotyping of clinical and lifestyle data; as well as a novel, robust data analysis pipeline that unifies multiple complex data domains into a single analysis framework. However, our study also exhibits limitations. Due to the cross-sectional nature of the performed microbiome-phenotype association analyses, our findings cannot firmly establish causal links and should be regarded as hypothesis-generating. Specifically, the design cannot preclude the possibility of reverse causation; for example, individuals with poorer cognitive function might face challenges in maintaining optimal oral hygiene, which could subsequently influence their periodontal health status and subgingival microbiome composition. However, it is important to note that our associational findings are substantiated by experimental research in animal models which provides evidence that periodontal pathogens can indeed trigger neuroinflammation and contribute to neurodegenerative pathology, lending weight to a potential contributing causal role. 11 , 13 Moreover, even after adjusting for confounders in our statistical models, we cannot fully exclude their potential influence on the relationship between subgingival microbiome composition and non-microbiome phenotypes. Longitudinal and experimental studies are needed that expand on our findings to further discern the effects of the oral microbiome on brain health. Additionally, data on participant relatedness or household sharing were unavailable; these factors are known to influence microbiome similarity and could therefore potentially act as unmeasured confounders in our analyses. Finally, our reliance on 16S rRNA gene sequencing limits the current analysis primarily to taxonomic composition, precluding direct assessment of microbial functional potential or metabolic activity. While computational tools can predict function from 16S data, integrating such inferred analyses was beyond the scope of this initial broad, multi-domain study focused on taxonomic associations. Our planned next steps include exploring computationally inferred functions (e.g., using PICRUSt2 57 ) on the existing dataset. However, for direct functional characterization, investigation of specific metabolic pathways, resolution of strain-specific taxonomic variations (including virulence factors), and identification of potential non-bacterial microbiome members, future studies utilizing high-resolution methods such as shotgun metagenomics will be essential to gain deeper mechanistic insights into the oral microbiome-brain axis. Conclusion Drawing on a comprehensive investigation integrating multi-domain data with cutting-edge analysis techniques we characterized latent associations of periodontitis-related subgingival microbial composition with cognitive function, mental health, brain structural integrity, systemic inflammation, diet and vascular risk in predominantly healthy individuals. Notably, periodontitis-related oral dysbiosis was associated with lower cognitive performance, lower cortical and subcortical volume, and higher leukocytes. As this research field progresses, oral microbiome profiling could contribute to improved dementia risk stratification and guide preventive interventions. Materials and methods Study design Our methodology is illustrated in Fig. 1 . This analysis integrates different data domains related to oral and brain health into a unified analysis framework: subgingival microbiome composition, oral health status, cognitive function, mental health status, brain structure, circulating inflammatory markers, dietary patterns, vascular risk and demographics. In brief, we employed topological data analysis, specifically the Mapper algorithm, to create a low-dimensional network representation of subgingival microbiome abundance data. 58 This network positions individuals based on similarities in their subgingival microbiome composition. Subsequently, we conducted an enrichment analysis using Spatial Analysis of Functional Enrichment (SAFE). 59 This analysis statistically assesses regions of the network derived via Mapper to identify where specific phenotypes are significantly higher or lower than expected by chance. This approach allowed us to examine regional differences in phenotypes, revealing whether the network representation captures variance in specific participant traits related to oral and brain health. Finally, we performed a post-hoc group comparison between participants of two distinct participant clusters within the microbiome network, enabling us to assess the link of subgingival microbiome composition and clinical and lifestyle phenotypes while adjusting for relevant covariates. Study population PAROMIND is a cross-sectional study nested within the Hamburg City Health Study (HCHS). The HCHS is a prospective, single-center, population-based cohort study investigating adults aged 45–75 to enhance the detection of major chronic disease risks through extensive clinical and imaging phenotyping. 21 Participants were included in PAROMIND for periodontal examination and subgingival sampling if they reported no antibiotic use within the preceding three months, had no requirement for endocarditis prophylaxis, and possessed more than two remaining teeth. The study protocol includes assessments of oral health (including collection of gingival crevicular fluids for microbiome analyses), cognition, mental health, diet, vascular risk, brain MRI, and blood sampling. For each participant, the clinical assessments and biological sampling were generally completed during a single comprehensive baseline visit; the separate brain MRI appointment followed shortly thereafter, ensuring reasonable contemporaneity between these measures for cross-sectional analysis. Ethics statement PAROMIND and the HCHS were approved by the local ethics committee of the Landesärztekammer Hamburg (State of Hamburg Chamber of Medical Practitioners, PV5131). The conduct of PAROMIND is governed by ethical guidelines of Good Clinical Practice (GCP), Good Epidemiological Practice (GEP) and the Declaration of Helsinki. 60 Written informed consent was obtained from all participants investigated in this work. Molecular analysis and phenotyping of the oral microbiome Oral microbiome phenotyping followed a standardized procedure targeting the subgingival environment within periodontal pockets. This specific environment was chosen because it contains the subgingival biofilms most directly implicated in periodontitis pathogenesis, distinguishing its microbial community from those in other oral sites like saliva or supragingival plaque. During the dental examination, gingival crevicular fluid samples were collected from periodontal pockets with sterile paper points. Two samples per participant were obtained: (1) from a single deep periodontal pocket and (2) one pooled sample (4 paper points) from the deepest periodontal pockets per quadrant. Each paper point remained in situ for 15 seconds. Subsequently, samples were placed in a sterile 2ml-Eppendorf tube and stored at -80°C in the HCHS biobank. For the following molecular analysis, only pooled samples were further processed. Bacterial composition was determined via 16S rRNA Illumina sequencing. DNA Isolation was performed using the DNA extraction kit from Innuprep Analytics (Analytik Jena AG, Überlingen, Germany). For the initial lysis with lysozyme and mutanolysin (3 mg lysozyme, 100 U mutanolysin, in 200 µl Tris EDTA buffer), all samples were incubated at 37°C for 10 min and further isolation was performed according to the manual with elution volumes of 100 µl. For 16S rDNA sequencing of isolated DNA, variable regions V3 and V4 of the 16S rRNA gene were amplified using the primer pair 341F (5‘-CCTACGGGAGGCAGCAG-3‘) and 806R (5‘-GGACTACHVGGGTWTCTAAT-3’) 27F-338R in a dual barcoding approach. 61 , 62 Agarose gel electrophoresis was used to verify resulting PCR products before normalization using the SequalPrep Normalization Plate Kit (Thermo Fischer Scientific, Waltham, MA, USA), pooling and sequencing on the Illumina MiSeq with v3 2x300bp chemistry (Illumina Inc., San Diego, CA, USA). Demultiplexing after sequencing was based on 0 mismatches in the barcode sequences. Data processing was performed using the DADA2 workflow for big datasets (v. 1.10.42, https://benjjneb.github.io/dada2/bigdata.html ), resulting in abundance tables of amplicon sequence variants (ASVs). For this, all sequencing runs were handled separately and finally collected in a single abundance table per dataset, which underwent chimera filtering. ASVs underwent taxonomic annotation using the Bayesian classifier provided in DADA2 and using the expanded Human Oral Microbiome Project (eHOMD) version V15.23. Samples with less than 10,000 sequences were not considered for further analysis. Downstream analyses were performed at the genus-level because this taxonomic rank provided a suitable balance, enhancing statistical power across the large cohort, improving interpretability and comparability with existing literature, and reducing data dimensionality while still providing sufficient resolution for the study's objectives. Genera that were observed with a frequency of less than 0.1% of all genera detected in a sample were discarded. Oral health assessment A certified study nurse assessed probing depth and gingival recession at six sites of the tooth (mesio-buccal, buccal, disto-buccal, disto-palatinal, palatinal and mesio-palatinal). The clinical attachment loss (CAL) was calculated (CAL = probing depth + gingival recession). 63 The BOP index was determined by probing 2 sites per tooth (vestibular, oral) and expressed as a percentage of bleeding sites. The number of bleeding sites on probing is a reliable and consistent measure to assess the degree of gingival inflammation. Measurements were completed with a standard periodontal probe (PCP 15, Hu-Friedy, Chicago, IL, USA). Additionally, the DMFT index was calculated. The oral health assessment followed a standardized protocol for the reporting of the prevalence and severity of periodontal diseases. 64 Cognitive and mental health assessments Cognitive testing was conducted using the extended version of the Consortium to Establish a Registry for Alzheimer’s Disease Neuropsychological Assessment Battery (CERAD-NP/Plus). 65 A trained study nurse administered all tests. For this analysis, we considered cognitive scores measuring executive function (Trail Making Test B), information processing speed (Trail Making Test A), memory (Word List Recall Test), reasoning (Multiple Choice Vocabulary Intelligence Test B), verbal fluency (Animal Naming Test), and the Mini Mental State Exam. To ensure higher scores indicated better cognitive performance across all tests, we inverted the results of Trail Making Test A and B. Subsequently, we performed a principal component analysis (PCA) on all individual test scores. Following previous procedures, the first principal component, which accounted for the greatest variance (40.5%), was defined as a measure of general cognitive ability (g) (for details see supplementary figure S9 ). 66 According to the principal component loadings, higher values of this measure corresponded to lower cognitive performance; thus, it was also inverted. Furthermore, participants underwent mental health assessments via established questionnaires for depression (PHQ-9, Geriatric Depression Scale), somatic symptom severity (PHQ-15), and anxiety (GAD-7). 67 – 69 Neuroimaging of brain macro- and microstructure Neuroimaging markers representing different aspects of macro- and microstructural brain integrity were computed based on T1-weighted and diffusion-weighted MRI following previous procedures. 70 After cortical surface reconstruction and subcortical segmentation based on T1-weighted images with FreeSurfer (v. 6.0.1), cortical thickness and subcortical volume were estimated representing morphometric measures of neurodegenerative processes. 71 , 72 The mean cortical thickness and mean subcortical volume were z-scored. Subsequently, the two resulting z scores were averaged to obtain a single summary measure reflecting cortical thickness and subcortical volume. Following preprocessing of the diffusion-weighted images, free-water imaging was employed to compute free-water which quantifies the amount of extracellular water as well as tissue fractional anisotropy reflecting neurite architecture and integrity. 73 , 74 The free-water and tissue fractional anisotropy values were then averaged across cortical and subcortical gray matter voxels, as well as white matter voxels, to obtain global measures of gray and white matter, respectively. For a detailed account on the acquisition protocol, quality assessment, preprocessing and computation procedures on the different imaging measures see supplementary text S10 . Diet The dietary behavior of all participants was assessed using a validated food frequency questionnaire with 102 items, developed for the European Prospective Investigation into Cancer and Nutrition Study (EPIC). 75 The adherence to different dietary patterns was measured based on the food frequency questionnaire scores, including the Mediterranean diet (MEDAS diet), the Dietary Approaches to Stop Hypertension (DASH diet), and the Mediterranean-DASH Intervention for Neurodegenerative Delay (MIND diet). 48 – 50 Adherence to the Mediterranean diet was determined using the German version of the Mediterranean Diet Adherence Screener, which assigns a score of 0 or 1 to 14 specific food items, producing a total adherence score ranging from 0 (no adherence) to 14 (maximum adherence). 48 The DASH diet was evaluated using a previously established scoring method that assigns a score of 0, 0.5, or 1 to each of 10 items, resulting in an adherence score between 0 (no adherence) and 10 (maximum adherence). 50 Finally, adherence to the MIND diet was calculated according to standard procedures: scores of 0, 0.5, or 1 to are assigned 10 healthy and 5 unhealthy food items, culminating in a total adherence score ranging from 0 (no adherence) to 15 (maximum adherence). 49 Topological data analysis We implemented a topological data analysis pipeline that integrates two key components: (1) the Mapper algorithm, which performs an unsupervised reconstruction of a topological network based on genus-level abundance data, capturing microbiome composition similarity, and (2) SAFE, which conducts statistical tests to examine the relationship between the network's structure and different phenotypes. 58 , 59 This method effectively combines dimensionality reduction with topological insights, offering a powerful tool for understanding the intrinsic geometry of high-dimensional microbiome data and its relationships with other data domains. The analysis was performed in python v3.8.1 based on the packages NetworkX v2.2 ( https://github.com/networkx/networkx ), safepy ( https://github.com/baryshnikova-lab/safepy ), scikit-learn v1.5.1 ( https://github.com/scikit-learn/scikit-learn ) and tmap v1.2 ( https://github.com/GPZ-Bioinfo/tmap ) as well as R v4.4.0 based on the package vegan v2.6-6.1 . 58 , 59 Data visualization was based on plotly v5.22 ( https://github.com/plotly/plotly.py ) and iTOL v6 ( https://itol.embl.de/ ). HTML versions of many presented plots can be found on OSF allowing interactive data exploration ( https://osf.io/vqj8m/ ). Mapper: Reconstruction of the microbiome similarity network Genus-level subgingival microbiome abundance data served as input to the Mapper algorithm, a topological data analysis technique that simplifies complex high-dimensional data by constructing a topological network capturing essential relationships and patterns in the data. This network preserves the data's underlying topological and geometric structure by positioning participants with similar subgingival microbiome profiles nearby. Conceptually, this representation is analogous to a topographical map that reveals the essential features of a landscape. Importantly, the network can represent non-linear associations that conventional linear techniques might miss. Mapper has previously been used to analyze the dynamic organization of brain function, 76 , 77 the shape of genetic data in breast-cancer patients, 78 biomolecular folding pathways, 79 brain structure in patients with fragile X syndrome, 80 and neuronal data from the visual cortex. 81 The applied Mapper pipeline comprises multiple analysis steps to reconstruct the topological network (for an illustration see Fig. 1 b): filtering, covering, clustering and network reconstruction. 82 First, as the filtering step, we conducted a principal coordinate analysis of the Aitchison distance of genus-level abundance (n participants x n genera ). 83 By that, we obtained two components – also called lenses – capturing the major axes of variation in the microbial community composition across the participants. Based on these axes, overlapping covers were defined (overlap = 1.5, resolution = 30) to segment the data into overlapping bins, each representing a local region of inter-individual variation. Unsupervised clustering of datapoints within each bin was performed using Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN, epsilon threshold = 0.95). 84 By this, nodes are obtained that represent participant groups with similar configuration of the subgingival microbiome. Participants can belong to multiple nodes with the number varying per participant. Lastly, the network reconstruction was accomplished by connecting clusters sharing common participants. Of note, not all datapoints are retained by Mapper resulting in the omission of some participants (n not retained = 109). SAFE: Enrichment analysis SAFE is an annotation technique for biological networks that enables the computation and statistical assessment of local enrichment for specific phenotypes. 59 In our work, we performed SAFE on the microbiome similarity network derived using the Mapper algorithm to identify regions within the network that are significantly enriched for specific participant traits. Specifically, we investigated the enrichment of genus-level microbiome abundance, oral health measures (clinical attachment loss, plaque index, bleeding on probing index, DMFT index), cognitive scores (general cognitive ability, Animal Naming Test, Mini Mental State Exam, Multiple Choice Vocabulary Intelligence Test B, Trail Making Tests A and B, Word List Recall), mental health scores (PHQ-9, PHQ-15, Geriatric Depression Scale, GAD-7), imaging measures (cortical thickness and subcortical volume, gray matter free-water, white matter free-water, gray matter tissue fractional anisotropy, white matter tissue fractional anisotropy), circulating inflammatory markers (hsCRP, leukocytes), diet scores (MEDAS score, DASH score, MIND score), vascular risk factors (systolic and diastolic blood pressure, body mass index, smoking behavior, blood triglycerides, cholesterol, low density lipoprotein, high density lipoprotein, HbA1c), and demographics (age, sex, education). Following previous analyses leveraging SAFE, 59 , 85 the microbiome similarity network was spring-embedded, i.e., nodes in the network were positioned so that those with connections are placed closer together, while repulsive forces push non-connected nodes apart, resulting in a visually balanced and interpretable representation of the network's structure. 86 Next, we computed node attributes for all phenotypes by averaging the values of each variable across all participants within a node. Subsequently, enrichment scores were calculated for each node following a four-step process (Fig. 1 c). First, we defined the local neighborhood by identifying all nodes within a maximum distance threshold of 0.75 from the central node. The distance was measured using the map-weighted shortest path length (MSPL). 59 Second, we calculated a neighborhood score by summing the attribute values of neighboring nodes. Third, we computed a p-value by comparing the empirical neighborhood score against a distribution derived from 5000 permutations. Permutations were performed by randomly reassigning attributes to nodes while preserving the network topology. 59 The resulting p-value was corrected for multiple comparisons across all phenotypes. Finally, we assigned an enrichment score to the neighborhood center by applying a -log10 transformation to the corrected p-value. Given the 5000 permutations, the maximal enrichment score is -log10(1/5000) = 3.70, with -log10(0.05) = 1.30 indicating significance. Positive enrichment scores indicate that observed values are higher than the permuted distribution, negative enrichment scores indicate that they are lower than the permuted distribution. This procedure is repeated for each node of the network, resulting in an enrichment map indicating where attributes are higher or lower than expected by chance. To understand the primary taxonomic patterns shaping the network topology, we performed a dominance analysis alongside examining individual phenotype enrichments. This involved labeling each network node by the single bacterial genus with the highest positive enrichment score. Visualizing this node-level dominance helps to reveal major taxonomic transitions across the network. While highlighting the most strongly enriched genus in each region provides valuable pointers to key drivers, this is a simplification; each node still represents a complex microbial community, not just the dominant genus identified. To measure how strongly the microbiome similarity network reflects a specific phenotype we computed the enrichment ratio as the number of significantly enriched nodes (corrected p < 0.05) divided by the total number of nodes. A higher enrichment ratio indicates that more nodes are significantly enriched, suggesting that the network’s topology captures a greater extent of a phenotype’s variance. As a complement, we used three different statistical methods from the vegan package to test the linear association between non-microbiome phenotypes and microbiome configuration: 1) envfit (n permutations = 10,000) testing the linear association between each phenotype and the PCoA ordination (all components) of the Aitchison distance matrix; 2) adonis (n permutations = 5,000) testing the proportion of variance in the Aitchison distance matrix explained by each phenotype individually, and capscale (n permutations = 5,000) testing the extent to which the ordination based on the Aitchison distance matrix is constrained by each phenotype individually. All p-values of the complementary approaches were false discovery rate-corrected. The complementary linear analyses were only performed for non-microbiome phenotypes and not for abundance of individual genera. To determine whether specific phenotypes co-enrich, i.e., exhibit similar enrichment patterns, we performed an ordination of the enrichment scores using principal component analysis retaining the first two principal components and assessed the pairwise Spearman correlation between the enrichment scores. Group analysis To further examine the relationship between the microbiome similarity network and clinical as well as lifestyle phenotypes, we performed a post-hoc group analysis which allowed us to adjust for relevant covariates. Therefore, nodes of the microbiome similarity network were clustered in two non-overlapping groups using k-Means clustering of the node positions. We chose k-Means as it is widely used and arguably represents the simplest unsupervised clustering technique. 87 k-Means requires to predefine the number of clusters to assign datapoints to. Given that enrichment analysis indicated that most participant traits varied along the microbiome similarity network in a linear left-right trajectory, i.e., participants on the left end differed from those on the right end, the number clusters was predefined as n clusters = 2 post-hoc. After the clustering, participants were categorized based on the resulting groupings. Importantly, individuals present in both groups – due to being assigned to nodes in both groups – were excluded from the analysis (n = 137). Given that not all datapoints are retained by the Mapper algorithm during the network reconstruction step, participants that were not represented in the microbiome similarity network were not considered for this analysis. The groups were statistically compared for the phenotypes using multiple linear regression and age, sex and education and vascular risk factors (systolic and diastolic blood pressure, body mass index, smoking behavior, triglycerides, cholesterol, LDL, HDL, HbA1c) were included as covariates: $$\:Phenotype\:\sim\:Group+Age+Sex+Education+Vascular\:risk\:factors$$ Covariates were selected a priori based on literature demonstrating their potential influence on both oral health/microbiome and brain health indices. 45 , 88 – 90 Sensitivity analyses To verify the robustness of our results against variations in the sample as well as the analysis pipeline, we conducted a comprehensive sensitivity analysis. This involved repeating the entire analysis across random subsamples of different sizes. Specifically, we varied the sample sizes from 100–10% of the total dataset, in 1% decrements. For each decrement, we randomly sampled subsets 100 times, resulting in a total of 9,000 iterations (90 different sample sizes × 100 random samples per steps). For each iteration, we assessed the robustness of our findings by comparing the results from the subsamples to the original results. The stability of the enrichment analysis results was evaluated by calculating the Spearman correlation of enrichment ratios for the non-microbiome phenotypes and genus-level abundance. In addition to this, we measured the agreement of the participant-group assignments resulting from k-Means clustering used for group analysis employing the Adjusted Rand Index (ARI) which ranges from 0 (no agreement) to 1 (full agreement). The parameters and components of our analysis pipeline were selected based on established practices, including strategies to optimize sample coverage as described in the tmap documentation, or utilized default software settings. Recognizing that pipeline parameters can influence resulting network characteristics (e.g., node count, connectivity), we conducted a sensitivity analysis using alternative configurations to verify that our results were not biased by these initial design choices. We systematically explored variations in parameter values and pipeline components, altering the Mapper cover overlap from the original 1.5 to 1, 1.2, 1.4, 1.6, 1.8 and 2; the Mapper cover resolution from 30 to 20, 25, 35, 40 and 45; and the Mapper epsilon threshold from 0.95 to 0.99, and 0.90. Additionally, we adjusted the SAFE distance threshold from 0.75 to 0.5 and 0.99, and the SAFE neighborhood radius from 0.1 to 0.05 and 0.15. We adjusted one parameter from the original pipeline per iteration, resulting in a total of 15 iterations. The robustness of our results was evaluated by comparing the findings from the original configuration to those from alternative setups. Matching the approach for the previous sensitivity analysis, we calculated the Spearman correlation of enrichment ratios for clinical and genus abundance phenotypes and the ARI for group assignments from k-Means clustering to assess stability. Declarations Data availability HCHS data can be obtained by qualified researchers on reasonable request to the study’s steering committee. The analysis code for this work is publicly available on GitHub (https://github.com/csi-hamburg/oral_microbiome_brain_health). Interactive versions of the plots can be found on OSF (https://osf.io/vqj8m/). Acknowledgments The authors would like to acknowledge all participants, all cooperation partners, patrons and the Deanery from the University Medical Center Hamburg-Eppendorf for supporting the HCHS. Special thanks are due to the staff at the Population Health Research Department for conducting the study. The publication of its results has been approved by the Steering Board of the HCHS. We also thank the staff of the microbiome and sequencing laboratories of the Institute of Clinical Molecular Biology Kiel for their excellent support. Competing interests JG has received speaker fees from Lundbeck, Janssen-Cilag, Lilly, Otsuka and Boehringer outside the submitted work. JF reported receiving personal fees from Acandis, Cerenovus, Microvention, Medtronic, Phenox, and Penumbra; receiving grants from Stryker and Route 92; being managing director of eppdata; and owning shares in Tegus and Vastrax; all outside the submitted work. RT is a co-inventor of an international patent on the use of a computing device to estimate the probability of myocardial infarction (PCT/EP2021/073193, International Publication Number WO2022043229A1). RT is shareholder of the company ART-EMIS GmbH Hamburg. GT has received fees as consultant or lecturer from Acandis, Alexion, Amarin, Bayer, Boehringer Ingelheim, BristolMyersSquibb/Pfizer, Daichi Sankyo, Portola, and Stryker outside the submitted work. The remaining authors declare no conflicts of interest. Author Contributions Each author has made a significant contribution to the manuscript and all authors read and approved its final version. We describe contributions to the paper using the CRediT contributor role taxonomy. M.P.: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Visualization, Writing—original draft, Writing—review & editing; C.W.: Conceptualization, Data curation, Investigation, Methodology, Resources, Writing—original draft, Writing—review & editing; K.B.: Data curation, Methodology, Investigation, Resources, Software, Writing—review & editing; G.H.: Resources, Writing—review & editing; T.B.: Resources, Writing—review & editing; M.A.: Resources, Writing—review & editing; C.M.: Resources, Writing—review & editing; F.L.N.: Data curation, Resources, Writing—review & editing; B.Z.: Resources, Writing—review & editing; J.F.: Resources, Writing—review & editing; J.G.: Resources, Writing—review & editing; S.K.: Resources, Writing—review & editing; R.T.: Resources, Writing—review & editing; C.B.: Methodology, Writing—review & editing; G.T.: Supervision, Funding, Writing—review & editing; B.C.: Conceptualization, Funding, Project administration, Resources, Supervision, Writing—original draft, Writing—review & editing; G.A.: Conceptualization, Funding, Project administration, Resources, Supervision, Writing—original draft, Writing—review & editing. Funding This work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation): PAROMIND – 514762487 – GZ AA 93/9-1 and GZ CH 2631/4-1 (B.C., G.A.); Sonderforschungsbereich 936 – 178316478 – C2 (G.T., B.C.); and Schwerpunktprogramm 2041 – 454012190 (G.T.). Microbiome sequencing received infrastructure support from the DFG Research Unit 5042 „miTarget" and the DFG Excellence Cluster 2167 "Precision Medicine in Chronic Inflammation" (PMI). References Baker JL, Welch M, Kauffman JL, McLean KM, J. S., He X (2024) The oral microbiome: diversity, biogeography and human health. Nat Rev Microbiol 22:89–104 Chen C-K, Wu Y-T, Chang Y-C (2017) Association between chronic periodontitis and the risk of Alzheimer’s disease: a retrospective, population-based, matched-cohort study. Alz Res Therapy 9:56 Nilsson H, Sanmartin Berglund J, Renvert S (2018) Longitudinal evaluation of periodontitis and development of cognitive decline among older adults. J Clin Periodontol 45:1142–1149 Iwasaki M et al (2019) Periodontitis, periodontal inflammation, and mild cognitive impairment: A 5-year cohort study. J Periodontal Res 54:233–240 Choi S et al (2019) Association of Chronic Periodontitis on Alzheimer’s Disease or Vascular Dementia. J Am Geriatr Soc 67:1234–1239 Kamer AR, Craig RG, Niederman R, Fortea J, de Leon (2020) M. J. Periodontal disease as a possible cause for Alzheimer’s disease. Periodontol 2000 83:242–271 Dioguardi M et al (2020) The Role of Periodontitis and Periodontal Bacteria in the Onset and Progression of Alzheimer’s Disease: A Systematic Review. J Clin Med 9:495 Demmer RT, Papapanou PN (2010) Epidemiologic patterns of chronic and aggressive periodontitis. Periodontol 2000 53:28–44 Socransky Ss, Haffajee Ad, Cugini Ma, Smith C, Kent Jr. (1998) R. L. Microbial complexes in subgingival plaque. J Clin Periodontol 25:134–144 Belstrøm D et al (2017) Microbial profile comparisons of saliva, pooled and site-specific subgingival samples in periodontitis patients. PLoS ONE 12:e0182992 Dominy SS et al (2019) Porphyromonas gingivalis in Alzheimer’s disease brains: Evidence for disease causation and treatment with small-molecule inhibitors. Sci Adv 5:eaau3333 Lei S et al (2023) Porphyromonas gingivalis bacteremia increases the permeability of the blood-brain barrier via the Mfsd2a/Caveolin-1 mediated transcytosis pathway. Int J Oral Sci 15:1–12 Hajishengallis G, Chavakis T (2021) Local and systemic mechanisms linking periodontal disease and inflammatory comorbidities. Nat Rev Immunol 21:426–440 Wang RP-H, Ho Y-S, Leung WK, Goto T, Chang R (2019) C.-C. Systemic inflammation linking chronic periodontitis to cognitive decline. Brain Behav Immun 81:63–73 Sparks Stein P et al (2012) Serum antibodies to periodontal pathogens are a risk factor for Alzheimer’s disease. Alzheimers Dement 8:196–203 Laugisch O et al (2018) Periodontal Pathogens and Associated Intrathecal Antibodies in Early Stages of Alzheimer’s Disease. J Alzheimers Dis 66:105–114 L’Heureux JE et al (2025) Oral microbiome and nitric oxide biomarkers in older people with mild cognitive impairment and APOE4 genotype. PNAS Nexus 4:pgae543 Troci A et al (2024) Disease- and stage-specific alterations of the oral and fecal microbiota in Alzheimer’s disease. PNAS Nexus 3:pgad427 Noble JM et al (2014) Serum IgG Antibody Levels to Periodontal Microbiota Are Associated with Incident Alzheimer Disease. PLoS ONE 9:e114959 Rubinstein T et al (2024) Periodontitis and brain magnetic resonance imaging markers of Alzheimer’s disease and cognitive aging. Alzheimer’s Dement 20:2191–2208 Jagodzinski A, Koch-gromus U, Adam G, Anders S, Augustin M (2019) Rationale and Design of the Hamburg City Health Study. Eur J Epidemiol. 10.1007/s10654-019-00577-4 Escapa IF et al (2018) New Insights into Human Nostril Microbiome from the Expanded Human Oral Microbiome Database (eHOMD): a Resource for the Microbiome of the Human Aerodigestive Tract. mSystems 3, e00187-18 Curtis MA, Diaz PI, Van Dyke T (2020) E. The role of the microbiota in periodontal disease. Periodontol 2000 83, 14–25 Sedghi LM, Bacino M, Kapila YL (2021) Periodontal Disease: The Good, The Bad, and The Unknown. Front Cell Infect Microbiol 11 Horiuchi A, Kokubu E, Warita T, Ishihara K (2020) Synergistic biofilm formation by Parvimonas micra and Fusobacterium nucleatum . Anaerobe 62:102100 Macuch PJ, Tanner ACR (2000) Campylobacter Species in Health, Gingivitis, and Periodontitis. J Dent Res 79:785–792 Hajishengallis G, Darveau RP, Curtis MA (2012) The Keystone Pathogen Hypothesis. Nat Rev Microbiol 10:717–725 Antezack A, Etchecopar-Etchart D, La Scola B, Monnet-Corti V (2023) New putative periodontopathogens and periodontal health-associated species: A systematic review and meta-analysis. J Periodontal Res 58:893–906 Costalonga M, Herzberg MC (2014) The oral microbiome and the immunobiology of periodontal disease and caries. Immunol Lett 162:22–38 Li X, Liu Y, Yang X, Li C, Song Z (2022) The Oral Microbiota: Community Composition, Influencing Factors, Pathogenesis, and Interventions. Front Microbiol 13 Zhou P, Manoil D, Belibasakis GN, Kotsakis GA (2021) Veillonellae: Beyond Bridging Species in Oral Biofilm Ecology. Front Oral Health 2:774115 Hajishengallis G (2015) Periodontitis: from microbial immune subversion to systemic inflammation. Nat Rev Immunol 15:30–44 Hajishengallis G (2014) Immunomicrobial pathogenesis of periodontitis: keystones, pathobionts, and host response. Trends Immunol 35:3–11 Fogelholm N et al (2023) Subgingival microbiome at different levels of cognition. J Oral Microbiol 15:2178765 Said-Sadier N et al (2023) Association between Periodontal Disease and Cognitive Impairment in Adults. Int J Environ Res Public Health 20:4707 Cs D et al (2018) Tooth loss is associated with accelerated cognitive decline and volumetric brain differences: a population-based study. Neurobiol Aging 67 Zhang R-Q et al (2023) Poor Oral Health and Risk of Incident Dementia: A Prospective Cohort Study of 425,183 Participants. J Alzheimers Dis 93:977–990 Matsuyama Y et al (2022) Differences in Brain Volume by Tooth Loss and Cognitive Function in Older Japanese Adults. Am J Geriatr Psychiatry 30:1271–1279 Sixin L, Stuart G, D., Rui Z (2023) Association Between Oral Bacteria and Alzheimer’s Disease: A Systematic Review and Meta-Analysis. J Alzheimer’s disease: JAD 91 Tang Z et al (2022) Treponema denticola Induces Alzheimer-Like Tau Hyperphosphorylation by Activating Hippocampal Neuroinflammation in Mice. J Dent Res 101:992–1001 Kato Y et al (2024) Fusobacterium in oral bacterial flora relates with asymptomatic brain lesions. Heliyon 10 Botelho J et al (2021) Periodontitis and circulating blood cell profiles: a systematic review and meta-analysis. Exp Hematol 93:1–13 Kamer AR et al (2008) Inflammation and Alzheimer’s disease: Possible role of periodontal diseases. Alzheimer’s Dement 4:242–250 Adam HS et al (2022) The prospective association between periodontal disease and brain imaging outcomes: The Atherosclerosis Risk in Communities study. J Clin Periodontol 49:322–334 Mayer C et al (2023) Association between periodontal disease and microstructural brain alterations in the Hamburg City Health Study. Journal of Clinical Periodontology n/a Malan-Müller S et al (2024) Probing the oral-brain connection: oral microbiome patterns in a large community cohort with anxiety, depression, and trauma symptoms, and periodontal outcomes. Transl Psychiatry 14:1–16 Santonocito S et al (2022) A Cross-Talk between Diet and the Oral Microbiome: Balance of Nutrition on Inflammation and Immune System’s Response during Periodontitis. Nutrients 14:2426 Hebestreit K et al (2017) Validation of the German version of the Mediterranean Diet Adherence Screener (MEDAS) questionnaire. BMC Cancer 17:341 Morris MC et al (2015) MIND diet slows cognitive decline with aging. Alzheimer’s Dement 11:1015–1022 Folsom AR, Parker ED, Harnack LJ (2007) Degree of Concordance With DASH Diet Guidelines and Incidence of Hypertension and Fatal Cardiovascular Disease*. Am J Hypertens 20:225–232 Altun E et al (2021) Association between Dietary Pattern and Periodontitis—A Cross-Sectional Study. Nutrients 13:4167 Seo Y et al (2024) Effect of MIND diet on cognitive function in elderly: a narrative review with emphasis on bioactive food ingredients. Food Sci Biotechnol 33:297–306 Santonocito S et al (2022) A Cross-Talk between Diet and the Oral Microbiome: Balance of Nutrition on Inflammation and Immune System’s Response during Periodontitis. Nutrients 14:2426 Woelber JP et al (2019) The influence of an anti-inflammatory diet on gingivitis. A randomized controlled trial. J Clin Periodontol 46:481–490 Jiang Y, Zhou X, Cheng L, Li M (2020) The Impact of Smoking on Subgingival Microflora: From Periodontal Health to Disease. Front Microbiol 11 Schwahn C et al (2021) Effect of periodontal treatment on preclinical Alzheimer’s disease—Results of a trial emulation approach. Alzheimer’s Dement alz 12378. 10.1002/alz.12378 Douglas GM et al (2020) PICRUSt2 for prediction of metagenome functions. Nat Biotechnol 38:685–688 Liao T, Wei Y, Luo M, Zhao G-P, Zhou H (2019) tmap: an integrative framework based on topological data analysis for population-scale microbiome stratification and association studies. Genome Biol 20:293 Baryshnikova A (2016) Systematic Functional Annotation and Visualization of Biological Networks. Cell Syst 2:412–421 Petersen M et al (2020) Network Localisation of White Matter Damage in Cerebral Small Vessel Disease. Sci Rep 10:9210 Muyzer G, de Waal EC, Uitterlinden AG (1993) Profiling of complex microbial populations by denaturing gradient gel electrophoresis analysis of polymerase chain reaction-amplified genes coding for 16S rRNA. Appl Environ Microbiol 59:695–700 Caporaso JG et al (2012) Ultra-high-throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. ISME J 6:1621–1624 Eke PI, Page RC, Wei L, Thornton-Evans G, Genco RJ (2012) Update of the case definitions for population-based surveillance of periodontitis. J Periodontol 83:1449–1454 Holtfreter B et al (2015) Standards for reporting chronic periodontitis prevalence and severity in epidemiologic studies: Proposed standards from the Joint EU/USA Periodontal Epidemiology Working Group. J Clin Periodontol 42:407–412 Moms JC et al (1989) The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD). Part I. Clinical and neuropsychological assesment of Alzheimer’s disease. Neurology 39:1159–1159 Fawns-Ritchie C, Deary IJ (2020) Reliability and validity of the UK Biobank cognitive tests. PLoS ONE 15:e0231627 Kroenke K, Spitzer RL, Williams JBW (2001) The PHQ-9: Validity of a brief depression severity measure. J Gen Intern Med 16:606–613 Kroenke K, Spitzer RL, Williams JBW (2002) The PHQ-15: Validity of a New Measure for Evaluating the Severity of Somatic Symptoms. Psychosom Med 64:258–266 Spitzer RL, Kroenke K, Williams JBW, Löwe B (2006) A Brief Measure for Assessing Generalized Anxiety Disorder: The GAD-7. Arch Intern Med 166:1092 Petersen M et al (2023) Brain imaging and neuropsychological assessment of individuals recovered from a mild to moderate SARS-CoV-2 infection. Proceedings of the National Academy of Sciences 120, e2217232120 Fischl B, Dale AM (2000) Measuring the thickness of the human cerebral cortex from magnetic resonance images. Proceedings of the National Academy of Sciences 97, 11050–11055 Fischl B et al (2002) Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron 33:341–355 Reveley C et al (2022) Diffusion MRI anisotropy in the cerebral cortex is determined by unmyelinated tissue features. Nat Commun 13:6702 Pasternak O, Sochen N, Gur Y, Intrator N, Assaf Y (2009) Free water elimination and mapping from diffusion MRI. Magn Reson Med 62:717–730 Boeing H, Wahrendorf J, Becker N (1999) EPIC-Germany–A source for studies into diet and risk of chronic diseases. European Investigation into Cancer and Nutrition. Ann Nutr Metab 43:195–204 Saggar M et al (2018) Towards a new approach to reveal dynamical organization of the brain using topological data analysis. Nat Commun 9:1399 Saggar M, Shine JM, Liégeois R, Dosenbach NUF, Fair D (2022) Precision dynamical mapping using topological data analysis reveals a hub-like transition state at rest. Nat Commun 13:4791 Nicolau M, Levine AJ, Carlsson G (2011) Topology based data analysis identifies a subgroup of breast cancers with a unique mutational profile and excellent survival. Proc Natl Acad Sci U S A 108:7265–7270 Yao Y et al (2009) Topological methods for exploring low-density states in biomolecular folding pathways. J Chem Phys 130:144115 Romano D et al (2014) Topological methods reveal high and low functioning neuro-phenotypes within fragile X syndrome. Hum Brain Mapp 35:4904–4915 Singh G et al (2008) Topological analysis of population activity in visual cortex. J Vis 8:11 Singh G, Memoli F, Carlsson G (2007) Topological Methods for the Analysis of High Dimensional Data Sets and 3D Object Recognition. The Eurographics Association Gloor GB, Macklaim JM, Pawlowsky-Glahn V, Egozcue JJ (2017) Microbiome Datasets Are Compositional: And This Is Not Optional. Front Microbiol 8 McInnes L, Healy J, Astels S (2017) hdbscan: Hierarchical density based clustering. J Open Source Softw 2:205 Costanzo M et al (2016) A global genetic interaction network maps a wiring diagram of cellular function. Science 353:aaf1420 Fruchterman TMJ, Reingold EM (1991) Graph drawing by force-directed placement. Software: Pract Experience 21:1129–1164 Ikotun AM, Ezugwu AE, Abualigah L, Abuhaija B, Heming J (2023) K-means clustering algorithms: A comprehensive review, variants analysis, and advances in the era of big data. Inf Sci 622:178–210 Alfaro-Almagro F et al (2021) Confound modelling in UK Biobank brain imaging. NeuroImage 224:117002 Petersen M et al (2024) A latent clinical-anatomical dimension relating metabolic syndrome to brain structure and cognition. eLife 12, RP93246 Sun Q, Li M (2025) Association between periodontitis and cognitive impairment in older adults: A cross-sectional study of the National Health and Nutrition Examination Survey. Clin Epidemiol Global Health 102020. 10.1016/j.cegh.2025.102020 How KY, Song KP, Chan KG (2016) Porphyromonas gingivalis: An Overview of Periodontopathic Pathogen below the Gum Line. Front Microbiol 7:53 Letunic I, Bork P (2024) Interactive Tree of Life (iTOL) v6: recent updates to the phylogenetic tree display and annotation tool. Nucleic Acids Res 52:W78–W82 Additional Declarations Yes there is potential Competing Interest. JG has received speaker fees from Lundbeck, Janssen-Cilag, Lilly, Otsuka and Boehringer outside the submitted work. JF reported receiving personal fees from Acandis, Cerenovus, Microvention, Medtronic, Phenox, and Penumbra; receiving grants from Stryker and Route 92; being managing director of eppdata; and owning shares in Tegus and Vastrax; all outside the submitted work. RT is a co-inventor of an international patent on the use of a computing device to estimate the probability of myocardial infarction (PCT/EP2021/073193, International Publication Number WO2022043229A1). RT is shareholder of the company ART-EMIS GmbH Hamburg. GT has received fees as consultant or lecturer from Acandis, Alexion, Amarin, Bayer, Boehringer Ingelheim, BristolMyersSquibb/Pfizer, Daichi Sankyo, Portola, and Stryker outside the submitted work. The remaining authors declare no conflicts of interest. Supplementary Files NMEDAN139424epc.pdf Editorial Policy Checklist NMEDAN139424rs.pdf Reporting Summary Supplement.pdf Supplementary materials Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6580781","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":455220809,"identity":"e7636952-070d-4fd7-8e59-2082fd73cd10","order_by":0,"name":"Marvin Petersen","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA3UlEQVRIiWNgGAWjYNCDDwwJQDKBKLUGYJJxBslamHmI0aLbfvYBc0HFHznz/sPHHtu2pckxsCcfwKvF7Ey6AfOMMwbGMjfS0o1z23KMGXie4bfG7EAaAzNvm0HiDAkeM+nctorEBokcA/xazj8DavlnUD+D//w3acu2ivoGifwP+LXcANnSYJAgwZDDJs3YlpPAIJGDVwdQyzOGwzOOGRvOkEgzk+w5l2bYxvOMkMPSGB8X1MjJS/AffibxoyxZnp89+QF+a4DgMAqPjaB6IGAmRtEoGAWjYBSMYAAAyUg+5WHAfb4AAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0001-6426-7167","institution":"University Medical Center Hamburg-Eppendorf","correspondingAuthor":true,"prefix":"","firstName":"Marvin","middleName":"","lastName":"Petersen","suffix":""},{"id":455220810,"identity":"3a5dc3bd-7164-4dff-872a-486a654935aa","order_by":1,"name":"Carolin Walther","email":"","orcid":"","institution":"University Medical Center Hamburg-Eppendorf","correspondingAuthor":false,"prefix":"","firstName":"Carolin","middleName":"","lastName":"Walther","suffix":""},{"id":455220811,"identity":"0df2e243-dd56-4df6-8ba0-10f00014ffa9","order_by":2,"name":"Katrin Borof","email":"","orcid":"","institution":"University Medical Center Hamburg-Eppendorf","correspondingAuthor":false,"prefix":"","firstName":"Katrin","middleName":"","lastName":"Borof","suffix":""},{"id":455220812,"identity":"bcd5175b-750c-4dd7-981b-67a9dfcaad52","order_by":3,"name":"Guido Heydecke","email":"","orcid":"","institution":"University Medical Center Hamburg-Eppendorf","correspondingAuthor":false,"prefix":"","firstName":"Guido","middleName":"","lastName":"Heydecke","suffix":""},{"id":455220813,"identity":"73b05c65-7f7e-4156-bd6f-c1037c3f546b","order_by":4,"name":"Thomas Beikler","email":"","orcid":"","institution":"University Medical Center Hamburg-Eppendorf","correspondingAuthor":false,"prefix":"","firstName":"Thomas","middleName":"","lastName":"Beikler","suffix":""},{"id":455220814,"identity":"1c1d6cc5-db26-46e9-bd8c-0152a39ff072","order_by":5,"name":"Malik Alawi","email":"","orcid":"https://orcid.org/0000-0002-5993-7709","institution":"University Medical Center Hamburg-Eppendorf","correspondingAuthor":false,"prefix":"","firstName":"Malik","middleName":"","lastName":"Alawi","suffix":""},{"id":455220815,"identity":"87ff8a7d-9abe-4ed7-9557-11f6b58d4f18","order_by":6,"name":"Christian Mueller","email":"","orcid":"","institution":"University Heart Center Hamburg","correspondingAuthor":false,"prefix":"","firstName":"Christian","middleName":"","lastName":"Mueller","suffix":""},{"id":455220816,"identity":"ebce6e9e-e77b-484d-9623-fae3536785cb","order_by":7,"name":"Felix Naegele","email":"","orcid":"","institution":"University Medical Center Hamburg-Eppendorf","correspondingAuthor":false,"prefix":"","firstName":"Felix","middleName":"","lastName":"Naegele","suffix":""},{"id":455220817,"identity":"225dfb98-1bb1-4ceb-9d33-2bca3c3d2ef8","order_by":8,"name":"Birgit-Christiane Zyriax","email":"","orcid":"https://orcid.org/0000-0002-5377-5956","institution":"University Medical Center Hamburg-Eppendorf","correspondingAuthor":false,"prefix":"","firstName":"Birgit-Christiane","middleName":"","lastName":"Zyriax","suffix":""},{"id":455220818,"identity":"c1c0a3f2-d0e7-4069-8109-deac399815af","order_by":9,"name":"Jens Fiehler","email":"","orcid":"","institution":"UKE","correspondingAuthor":false,"prefix":"","firstName":"Jens","middleName":"","lastName":"Fiehler","suffix":""},{"id":455220819,"identity":"637bc598-4a08-466f-b494-5a04c5de18d9","order_by":10,"name":"Jürgen Gallinat","email":"","orcid":"","institution":"Department of Psychiatry and Psychotherapy University Medical Center Hamburg-Eppendorf","correspondingAuthor":false,"prefix":"","firstName":"Jürgen","middleName":"","lastName":"Gallinat","suffix":""},{"id":455220820,"identity":"9c7c117e-bbf5-4039-9f28-a5235670010c","order_by":11,"name":"Simone Kuehn","email":"","orcid":"","institution":"University Medical Center Hamburg-Eppendorf","correspondingAuthor":false,"prefix":"","firstName":"Simone","middleName":"","lastName":"Kuehn","suffix":""},{"id":455220821,"identity":"2580ceb5-08dd-4643-a2ec-3f71ef799dcd","order_by":12,"name":"Raphael Twerenbold","email":"","orcid":"https://orcid.org/0000-0003-3814-6542","institution":"University Medical Center Hamburg-Eppendorf","correspondingAuthor":false,"prefix":"","firstName":"Raphael","middleName":"","lastName":"Twerenbold","suffix":""},{"id":455220822,"identity":"ed485256-9ce2-4f31-813a-30fc39784a54","order_by":13,"name":"Corinna Bang","email":"","orcid":"","institution":"Institute of Clinical Molecular Biology, Christian-Albrechts-University Kiel, University Hospital Schleswig Holstein Campus Kiel","correspondingAuthor":false,"prefix":"","firstName":"Corinna","middleName":"","lastName":"Bang","suffix":""},{"id":455220823,"identity":"50df4f24-0ecc-4755-9254-6f4c01f5d695","order_by":14,"name":"Goetz Thomalla","email":"","orcid":"","institution":"University Medical Center Hamburg-Eppendorf","correspondingAuthor":false,"prefix":"","firstName":"Goetz","middleName":"","lastName":"Thomalla","suffix":""},{"id":455220824,"identity":"fbd9a523-9b52-45fa-b732-86cfd16873cb","order_by":15,"name":"Bastian Cheng","email":"","orcid":"https://orcid.org/0000-0003-2434-1822","institution":"University Medical Center Hamburg-Eppendorf","correspondingAuthor":false,"prefix":"","firstName":"Bastian","middleName":"","lastName":"Cheng","suffix":""},{"id":455220825,"identity":"14873c4d-4a80-478b-8eea-853907622489","order_by":16,"name":"Ghazal Aarabi","email":"","orcid":"","institution":"University Medical Center Hamburg-Eppendorf","correspondingAuthor":false,"prefix":"","firstName":"Ghazal","middleName":"","lastName":"Aarabi","suffix":""}],"badges":[],"createdAt":"2025-05-02 20:35:46","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6580781/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6580781/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":82701868,"identity":"73f0b39c-e3eb-4823-8242-e04f65177392","added_by":"auto","created_at":"2025-05-14 09:38:51","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1694669,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMethodology. \u003c/strong\u003ea) Population-based data from the Hamburg City Health Study were used including subgingival microbiome abundance, assessments of oral health status, cognitive function and mental health, multimodal brain MRI, circulating inflammatory markers, diet questionnaires, and vascular risk measures. b) We applied the Mapper algorithm to transform the high-dimensional genus-level subgingival microbiome profiles into an interpretable low-dimensional network representation. Mapper consists of multiple steps: (1) Inputting participant-level genus abundance; (2) Projecting data points to a low-dimensional space using a filter function for dimensionality reduction; (3) Dividing the low-dimensional space into overlapping covers, each containing a subset of data points; (4) Clustering data points within each cover based on their distances in the original high-dimensional space; (5) Constructing a network from the clustering results, where each node represents a cluster of participants, and links between nodes indicate shared participants between clusters. Modified from Liao et al.\u003csup\u003e58\u003c/sup\u003e c) Spatial Analysis of Functional Enrichment (SAFE) was used to annotate the network derived via Mapper, identifying regions significantly enriched with specific attributes (e.g., systolic blood pressure). SAFE involves the following steps: (1) Computing node attributes by averaging variables of interest across participants within each node; (2) Defining the local neighborhood by identifying all nodes within a maximum distance threshold from the center node, with distance measured by map-weighted shortest path length (MSPL); (3) Calculating a neighborhood score by summing the attribute values of neighboring nodes; (4) Computing a p-value by comparing the empirical neighborhood score against a randomly permuted distribution, achieved through random node-to-attribute reassignments while preserving network topology; (5) Assigning an enrichment score to the neighborhood center using -log10 transformation of the multiple testing-corrected p-value. This procedure is repeated for each node, resulting in an enrichment map indicating where attributes are higher or lower than expected by chance. Modified from Baryshnikova et al.\u003csup\u003e59\u003c/sup\u003e \u003cem\u003eAbbreviations\u003c/em\u003e: BOP index = bleeding on probing index; clin. = clinical; cort. = cortical; DMFT index = decayed/missing/filled teeth index; hsCRP = high-sensitivity c-reactive protein; subcort. = subcortical.\u003c/p\u003e","description":"","filename":"fig1.png","url":"https://assets-eu.researchsquare.com/files/rs-6580781/v1/c7bd9159ccb2a1f06e13ebe0.png"},{"id":82702470,"identity":"3222d6d1-bdbb-44a9-b95d-382745eb433e","added_by":"auto","created_at":"2025-05-14 09:46:51","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":6689238,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMicrobiome similarity network based on genus-level abundance.\u003c/strong\u003e a) Microbiome similarity network obtained by applying the Mapper algorithm to the abundances of oral microbiome genera consisting of 577 nodes and 10230 edges. Nodes represent groups of participants with similar microbiome profiles. Edges connect nodes that share common participants. Within the network, proximity signifies microbiome profile similarity among participants. b) Exemplary network distribution of z-scored \u003cem\u003ePorphyromonas\u003c/em\u003e abundance. \u003cem\u003ePorphyromonas\u003c/em\u003e was selected given its high relevance in periodontal disease.\u003csup\u003e91\u003c/sup\u003e c) Enrichment map of \u003cem\u003ePorphyromonas\u003c/em\u003e abundance. Nodes with positive enrichment scores are colored blue indicating a higher \u003cem\u003ePorphyromonas\u003c/em\u003e abundance of the node and its neighborhood than expected by random permutation. Nodes with negative enrichment scores are colored red indicating lower \u003cem\u003ePorphyromonas\u003c/em\u003e abundance than expected by permutation. Non-significant nodes are shown in a lighter shade. The enrichment ratio represents a measure of how strongly a phenotype is enriched on a network. It is computed by dividing the count of significantly enriched nodes by all nodes.\u003c/p\u003e","description":"","filename":"fig2.png","url":"https://assets-eu.researchsquare.com/files/rs-6580781/v1/9baa4c09084a4a47407545bc.png"},{"id":82701833,"identity":"9360082e-a4cd-49ec-8ab0-19c3a37cfb62","added_by":"auto","created_at":"2025-05-14 09:38:50","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":3609540,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eEnrichment analysis of microbiome phenotypes. \u003c/strong\u003ea) Circular phylogenetic tree at the level of oral microbiome genera. The inner band shows phyla with corresponding coloring of the tree, the mid band displays the enrichment ratio and the outer band shows overall relative abundance. The phylogenetic tree was visualized using iTOL (v6).\u003csup\u003e92\u003c/sup\u003e b) Dominance analysis: Nodes of the microbiome similarity network are colored to represent the bacterial genus with the highest enrichment score. Genera with dominance in five or more nodes are shown. \u003cem\u003eAbbreviations\u003c/em\u003e: GM = gray matter, WM = white matter.\u003c/p\u003e","description":"","filename":"fig3.png","url":"https://assets-eu.researchsquare.com/files/rs-6580781/v1/55da52fab582bc35e4cf93bb.png"},{"id":82701883,"identity":"ae6e3b4b-1e89-4088-87b6-3dbbe66753da","added_by":"auto","created_at":"2025-05-14 09:38:52","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":10928781,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eEnrichment analysis of non-microbiome phenotypes.\u003c/strong\u003e a) The bar plot displays the enrichment ratio (= significantly enriched network nodes / all network nodes) of non-microbiome phenotypes. The enrichment ratio reflects how strongly a phenotype’s variance is captured by the topology of the microbiome network obtained via Mapper. The bars of the plot are colored by variable category. b) Enrichment maps: Enrichment scores (-log\u003csub\u003e10\u003c/sub\u003e-transformed p-values) are mapped on the microbiome similarity network. Nodes with positive enrichment scores, indicating a higher neighborhood score than expected by chance, are colored blue, while nodes with negative enrichment scores, indicating a lower neighborhood score than expected by chance, are colored red. Non-significant nodes are shown in a lighter shade. Non‑microbiome phenotypes are first ordered by domain (denoted by colored dots in the lower‑left) and then by enrichment ratio. Enrichment maps for demographics, vascular risk factors and genus-level abundances are not displayed and can be found in the online supplement (\u003ca href=\"https://osf.io/vqj8m/\"\u003ehttps://osf.io/vqj8m/\u003c/a\u003e)\u003cem\u003e. Abbreviations\u003c/em\u003e: BOP index = bleeding on probing index; DMFT index = decayed/missing/filled teeth index; GM = gray matter, hsCRP = high sensitivity c-reactive peptide; LDL = low density lipoprotein, T. fractional anisotropy = tissue fractional anisotropy; WM = white matter.\u003c/p\u003e","description":"","filename":"fig4.png","url":"https://assets-eu.researchsquare.com/files/rs-6580781/v1/c45d29fde326742c92a40dfa.png"},{"id":82701838,"identity":"da23d123-2b8d-4711-af63-6af9905140dd","added_by":"auto","created_at":"2025-05-14 09:38:50","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":476867,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePrincipal component analysis of enrichment scores.\u003c/strong\u003e The plot illustrates the phenotypes’ location in principal component space, where proximity suggests similarity in enrichment patterns. Points are color-coded by phenotype category. Dominant genera and selected non-microbiome measures from each phenotype group were highlighted with annotations. For a fully annotated scatter plot, refer to the \u003cem\u003esupplementary materials\u003c/em\u003e.\u003c/p\u003e","description":"","filename":"fig5.png","url":"https://assets-eu.researchsquare.com/files/rs-6580781/v1/944e7df1d6d0bde755ba7421.png"},{"id":82701845,"identity":"42f85650-5bf3-474c-bb38-601fe05fe48e","added_by":"auto","created_at":"2025-05-14 09:38:50","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":1076682,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCo-enrichment heatmap\u003c/strong\u003e. The heatmap presents the Spearman correlation of enrichment scores for genus-level abundance and non-microbiome phenotypes. The order is determined by hierarchical clustering of the enrichment scores. \u003cem\u003eAbbreviations\u003c/em\u003e: BOP index = bleeding on probing index, DMFT index = decayed/missing/filled teeth index, GM = gray matter, LDL = low density lipoprotein, T. fractional anisotropy = tissue fractional anisotropy, WM = white matter.\u003c/p\u003e","description":"","filename":"fig6.png","url":"https://assets-eu.researchsquare.com/files/rs-6580781/v1/2b545b7729c1d0bfbb3d9d11.png"},{"id":82703751,"identity":"14aff09e-0816-4fa0-987c-4a33646a4b74","added_by":"auto","created_at":"2025-05-14 09:54:51","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":2556514,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGroup analysis.\u003c/strong\u003e a) Approach: The nodes of the microbiome similarity network were divided into two distinct groups (“A” and “B”) using k-means clustering. Participants were categorized based on these groupings, with individuals present in both groups being excluded from the analysis (n = 137). The two participant groups were then statistically compared using linear regression analysis, adjusting for age, sex, education and vascular risk factors. b) Relative abundance of bacterial genera within groups. c) Top 15 positive and negative differences in microbiome genera: Negative regression coefficients (higher in the “A” group) are shown in red, while positive coefficients (higher in the “B” group) are shown in blue. Significance is indicated by asterisks. For a bar plot displaying coefficients of all investigated genera see \u003cem\u003esupplementary materials\u003c/em\u003e. d) Shown are box plots for significant associations. Box plot colors correspond to the groups. Statistical significance level is indicated by asterisks. For box plots of the remaining non-microbiome phenotypes see \u003cem\u003esupplementary materials\u003c/em\u003e. \u003cem\u003eAbbreviations\u003c/em\u003e: β\u003csub\u003estd\u003c/sub\u003e = standardized beta coefficient for the group variable. BOP index = bleeding on probing index, DMFT index = decayed/missing/filled teeth index, GM = gray matter; \u003cem\u003ePeptostreptococ.\u003c/em\u003e = \u003cem\u003ePeptostreptococcaceae\u003c/em\u003e; WM = white matter.\u003c/p\u003e","description":"","filename":"fig7.png","url":"https://assets-eu.researchsquare.com/files/rs-6580781/v1/27db5ab6fe3b384d68f50052.png"},{"id":93116049,"identity":"c4a476dc-9a95-467c-ab8a-be853cd8aaa3","added_by":"auto","created_at":"2025-10-09 08:49:18","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":22466201,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6580781/v1/e3bd9754-05cb-4048-9f9d-fde84f515081.pdf"},{"id":82701832,"identity":"864c3eab-1310-412e-a2e3-423410d3c6fe","added_by":"auto","created_at":"2025-05-14 09:38:50","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1805605,"visible":true,"origin":"","legend":"Editorial Policy Checklist","description":"","filename":"NMEDAN139424epc.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6580781/v1/d6a1d1eea00b5a03b6535b39.pdf"},{"id":82701834,"identity":"a583168f-58ab-49d1-bd9c-05e6bac9b775","added_by":"auto","created_at":"2025-05-14 09:38:50","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":3402896,"visible":true,"origin":"","legend":"Reporting Summary","description":"","filename":"NMEDAN139424rs.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6580781/v1/2b0bd56962b1644117025ff5.pdf"},{"id":82701835,"identity":"c3c5c5f0-1fd0-442b-a8cf-0e7609a8e9ec","added_by":"auto","created_at":"2025-05-14 09:38:50","extension":"pdf","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":1964754,"visible":true,"origin":"","legend":"Supplementary materials","description":"","filename":"Supplement.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6580781/v1/2729df902cb5d2649f6c5548.pdf"}],"financialInterests":"\u003cb\u003eYes\u003c/b\u003e there is potential Competing Interest.\nJG has received speaker fees from Lundbeck, Janssen-Cilag, Lilly, Otsuka and Boehringer outside the submitted work. JF reported receiving personal fees from Acandis, Cerenovus, Microvention, Medtronic, Phenox, and Penumbra; receiving grants from Stryker and Route 92; being managing director of eppdata; and owning shares in Tegus and Vastrax; all outside the submitted work. RT is a co-inventor of an international patent on the use of a computing device to estimate the probability of myocardial infarction (PCT/EP2021/073193, International Publication Number WO2022043229A1). RT is shareholder of the company ART-EMIS GmbH Hamburg. GT has received fees as consultant or lecturer from Acandis, Alexion, Amarin, Bayer, Boehringer Ingelheim, BristolMyersSquibb/Pfizer, Daichi Sankyo, Portola, and Stryker outside the submitted work. The remaining authors declare no conflicts of interest.","formattedTitle":"Oral microbiome profiles relate periodontal disease and brain health - the PAROMIND Study","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe human oral microbiome harbors a diverse community of microorganisms that influences the health and well-being of their hosts.\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e In recent years, periodontitis, which is the inflammatory disruption in the host\u0026ndash;microbial homeostasis of the periodontal pocket, has gained increasing attention as a key factor impacting brain health. Studies indicate that periodontitis and the linked bacterial communities are associated with the incidence of cognitive decline and Alzheimer\u0026rsquo;s dementia.\u003csup\u003e\u003cspan additionalcitationids=\"CR3 CR4 CR5 CR6\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e Given its estimated prevalence between 10% and 50% in elderly people, this renders periodontitis a relevant public health concern.\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e As therapeutic interventions can alter the progression of periodontitis, comprehending its impact on the brain is vital for effective prevention and management of cognitive sequelae.\u003c/p\u003e \u003cp\u003eAmong bacterial species of the subgingival biofilm, several are known to be associated with periodontitis including: \u003cem\u003ePorphyromonas gingivalis\u003c/em\u003e, \u003cem\u003eTannerella forsythia\u003c/em\u003e and \u003cem\u003eTreponema denticola\u003c/em\u003e.\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e Mechanistic models have been proposed to explain the connection between these bacterial communities and cognitive decline. The connection is considered to arise from bacterial species promoting systemic inflammation, neurodegenerative processes, and blood-brain barrier disruption.\u003csup\u003e\u003cspan additionalcitationids=\"CR12 CR13\" citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eDespite existing research efforts, our understanding of the association between oral microbiome composition and brain health remains limited. While recent studies have predominantly focused on clinical Alzheimer\u0026rsquo;s disease and mild cognitive impairment cohorts,\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e,\u003cspan additionalcitationids=\"CR16 CR17 CR18\" citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u003c/sup\u003e understanding these connections in the general population requires research within broader community-based settings, which remains relatively limited.\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e Additionally, many investigations are constrained by small sample sizes and a lack of multi-modal imaging and biomarker data potentially leading to inconsistent findings.\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e These issues are further complicated by the inherent complexity of microbiome data.\u003c/p\u003e \u003cp\u003eTapping into these research needs, we aim to advance the understanding of the oral microbiome-brain axis by applying advanced analysis techniques capable of integrating the large-scale, multi-domain data crucial for achieving robust insights. Specifically, our primary hypothesis was that subgingival microbial profiles indicative of periodontitis would be associated with variations in brain health-related host phenotypes within a population-based cohort. To investigate this hypothesis, we analyzed population-based data of n\u0026thinsp;=\u0026thinsp;1,026 individuals from the Hamburg City Health Study.\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e Our approach focuses on a topology-based analysis that integrates abundance data of the subgingival microbiome derived from 16S rRNA sequencing of subgingival samples with in-depth clinical and lifestyle data including oral health assessments, cognitive test scores, mental health scores, neuroimaging, circulating inflammatory markers, dietary patterns, and vascular risk measures.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eOur methodology, which integrates microbiome, clinical and lifestyle data into a unified analysis framework, is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Quality assessment of subgingival microbiome abundance and brain MR imaging data resulted in a final analysis sample of 1,026 individuals ([mean\u0026thinsp;\u0026plusmn;\u0026thinsp;SD] age 63.72\u0026thinsp;\u0026plusmn;\u0026thinsp;8.2 years, 42.7% female; for details see Table \u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSample characteristics\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMetric\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eValue\u003csup\u003e\u003cem\u003ea\u003c/em\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge, years\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e63.72\u0026thinsp;\u0026plusmn;\u0026thinsp;8.17 (1,026) [range: 46\u0026ndash;78]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFemale, n (% female)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e438 (42.7)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEducation, ISCED\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.42\u0026thinsp;\u0026plusmn;\u0026thinsp;0.58 (999)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eOral health measures\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eClinical attachment loss\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.60\u0026thinsp;\u0026plusmn;\u0026thinsp;0.83 (842)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePlaque index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e19.22\u0026thinsp;\u0026plusmn;\u0026thinsp;26.30 (830)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBleeding on probing index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e15.38\u0026thinsp;\u0026plusmn;\u0026thinsp;19.91 (1,011)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDMFT index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e19.27\u0026thinsp;\u0026plusmn;\u0026thinsp;4.90 (842)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eCognitive scores\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAnimal Naming Test\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e24.66\u0026thinsp;\u0026plusmn;\u0026thinsp;6.67 (977)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMini Mental State Exam\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e27.85\u0026thinsp;\u0026plusmn;\u0026thinsp;1.68 (973)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMultiple Choice Vocabulary Intelligence Test\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e31.42\u0026thinsp;\u0026plusmn;\u0026thinsp;3.43 (838)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTrail Making Test A, sec\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e39.82\u0026thinsp;\u0026plusmn;\u0026thinsp;14.07 (925)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTrail Making Test B, sec\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e89.44\u0026thinsp;\u0026plusmn;\u0026thinsp;37.66 (917)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWord List Recall\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7.70\u0026thinsp;\u0026plusmn;\u0026thinsp;1.88 (949)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eMental health scores\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePHQ-9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.21\u0026thinsp;\u0026plusmn;\u0026thinsp;3.61 (1,025)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePHQ-15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4.27\u0026thinsp;\u0026plusmn;\u0026thinsp;3.90 (1,025)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGeriatric Depression Scale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.83\u0026thinsp;\u0026plusmn;\u0026thinsp;2.20 (914)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGAD-7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.41\u0026thinsp;\u0026plusmn;\u0026thinsp;2.90 (1,025)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eGlobal neuroimaging features\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFree-water (gray matter)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.50\u0026thinsp;\u0026plusmn;\u0026thinsp;0.05 (909)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFree-water (white matter)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.31\u0026thinsp;\u0026plusmn;\u0026thinsp;0.03 (909)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTissue fractional anisotropy (gray matter)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.19\u0026thinsp;\u0026plusmn;\u0026thinsp;0.01 (909)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTissue fractional anisotropy (white matter)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.41\u0026thinsp;\u0026plusmn;\u0026thinsp;0.01 (909)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMean cortical thickness, mm\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2.63\u0026thinsp;\u0026plusmn;\u0026thinsp;0.11 (986)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMean subcortical volume, ml\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3188.31\u0026thinsp;\u0026plusmn;\u0026thinsp;313.66 (986)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eInflammatory markers\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ehsCRP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.25\u0026thinsp;\u0026plusmn;\u0026thinsp;0.45 (970)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLeukocytes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6.13\u0026thinsp;\u0026plusmn;\u0026thinsp;1.75 (1,006)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eDiet\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMEDAS score\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4.46\u0026thinsp;\u0026plusmn;\u0026thinsp;1.89 (970)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDASH score\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4.46\u0026thinsp;\u0026plusmn;\u0026thinsp;1.06 (970)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMIND score\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6.30\u0026thinsp;\u0026plusmn;\u0026thinsp;1.78 (970)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003eVascular risk measurements\u003c/span\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSystolic blood pressure\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e142.12\u0026thinsp;\u0026plusmn;\u0026thinsp;20.06 (997)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiastolic blood pressure\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e83.46\u0026thinsp;\u0026plusmn;\u0026thinsp;10.55 (997)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBody mass index\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e26.60\u0026thinsp;\u0026plusmn;\u0026thinsp;4.32 (995)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSmoking, %\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e173 (16.9)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTriglycerides\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e117.67\u0026thinsp;\u0026plusmn;\u0026thinsp;75.68 (1,007)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCholesterol\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e208.28\u0026thinsp;\u0026plusmn;\u0026thinsp;40.83 (1,008)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLDL\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e121.56\u0026thinsp;\u0026plusmn;\u0026thinsp;36.53 (997)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHDL\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e63.51\u0026thinsp;\u0026plusmn;\u0026thinsp;19.00 (1,008)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHbA1c\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e5.63\u0026thinsp;\u0026plusmn;\u0026thinsp;0.54 (1,007)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c2\" namest=\"c1\"\u003e \u003cp\u003eAbbreviations: DMFT index\u0026thinsp;=\u0026thinsp;decayed/missing/filled teeth index, HDL\u0026thinsp;=\u0026thinsp;high density lipoprotein, hsCRP\u0026thinsp;=\u0026thinsp;high sensitivity c-reactive peptide, LDL\u0026thinsp;=\u0026thinsp;low density lipoprotein, ISCED\u0026thinsp;=\u0026thinsp;International Standard Classification of Education, mm\u0026thinsp;=\u0026thinsp;millimeters, sec\u0026thinsp;=\u0026thinsp;seconds\u003c/p\u003e \u003cp\u003e\u003csup\u003ea\u003c/sup\u003ePresented as median [IQR] (N)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eA similarity network of subgingival bacterial communities\u003c/h2\u003e \u003cp\u003eWe applied an unsupervised, topology-based technique to abundance data of 85 different bacterial genera to infer a low-dimensional network representation of oral microbiome similarity. The resulting microbiome similarity network consisted of 577 nodes \u0026ndash; representing participant groups with highly similar subgingival microbiome compositions \u0026ndash; and 10,230 edges \u0026ndash; connecting nodes that share at least one participant. The network represents a map of inter-individual variability capturing transitions in microbial composition across the cohort. For a visualization of the network see Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eIn an enrichment analysis, we examined whether the microbiome similarity network captures inter-individual differences in participant traits. Therefore, we computed enrichment scores, which are node-level indices that indicate whether participant traits are significantly higher or lower than expected by chance in specific regions, i.e., participant clusters, of the network. Enrichment scores were obtained for genus-level abundance data as well as all non-microbiome phenotypes. Furthermore, we calculated the enrichment ratio (=\u0026thinsp;number of significantly enriched nodes / total number of nodes) based on the enrichment scores for each phenotype quantifying the overall amount of enrichment on the microbiome similarity network. This measure indicates how strongly the network-based organization of participants reflects inter-individual variance in a respective phenotype. Refer to Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb and \u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec for an exemplary display of \u003cem\u003ePorphyromonas\u003c/em\u003e abundance and the corresponding enrichment map on the microbiome similarity network.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eEnrichment analysis of genus-level abundance\u003c/h3\u003e\n\u003cp\u003eMicrobiome abundance showed significant enrichment across all detected genera ranging from 42.5\u0026ndash;87.7%. Figure\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea displays the enrichment ratios alongside mean relative abundance and phylogenetic associations. We performed a dominance analysis by identifying for each network node the bacterial genus with the highest enrichment scores (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb). Of the 85 investigated genera, 15 showed the highest enrichment score for at least 5 nodes. Distribution of dominant genera in the network representation of oral microbiome similarity followed a horizontal pathogenicity gradient: Bacterial genera with strongest enrichments at the \u003cem\u003eleft\u003c/em\u003e end of the microbiome similarity network were periodontitis-associated taxa including \u003cem\u003eFusobacterium\u003c/em\u003e (n\u003csub\u003enodes\u003c/sub\u003e = 208), \u003cem\u003eCampylobacter\u003c/em\u003e (65), Treponema (13), \u003cem\u003eDialister\u003c/em\u003e (8), \u003cem\u003eSaccharibacteria\u003c/em\u003e (TM7) [G-5] (6) and \u003cem\u003ePorphyromonas\u003c/em\u003e (5). In the center of the network, \u003cem\u003eAggregatibacter\u003c/em\u003e (18), \u003cem\u003eGemella\u003c/em\u003e (8), \u003cem\u003eCapnocytophaga\u003c/em\u003e (6) and \u003cem\u003eLeptotrichia\u003c/em\u003e (5) exhibited highest enrichment scores. At the right end, genera with strongest enrichments were of low periodontal pathogenicity or related to other dental diseases including \u003cem\u003eStreptococcus\u003c/em\u003e (126), \u003cem\u003eVeillonella\u003c/em\u003e (42), \u003cem\u003eNeisseria\u003c/em\u003e (26), \u003cem\u003eRothia\u003c/em\u003e (17) and \u003cem\u003eHaemophilus\u003c/em\u003e (7).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eEnrichment analysis of non-microbiome phenotypes\u003c/h3\u003e\n\u003cp\u003eAll non-microbiome phenotypes showed significant enrichment on the microbiome similarity network including measures of oral health status, cognition, mental health, brain structure, circulating inflammatory markers, diet, vascular risk factors and demographics (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). The top 10 ranking non-microbiome phenotypes were leukocytes (78.0%), plaque index (77.3%), verbal fluency (75.4%), general cognitive ability (74.7%), smoking behavior (74.5%), Mini Mental State Exam (73.8%), bleeding on probing (BOP) index (72.6%), clinical attachment loss (69.5%), Memory (69.2%) and age (68.6%) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea). Complementary microbiome covariate identification using linear approaches (\u003cem\u003eenvfit\u003c/em\u003e, \u003cem\u003eadonis\u003c/em\u003e, and \u003cem\u003ecapscale\u003c/em\u003e) confirmed these results. Among the phenotypes with the top 10 highest enrichment ratios, all showed a significant linear association (p\u003csub\u003eFDR\u003c/sub\u003e \u0026lt; 0.05) with subgingival microbiome profiles across all three complementary linear approaches (\u003cem\u003esupplementary figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e)\u003c/em\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eBased on the enrichment maps, we assessed inter-participant differences of phenotypes. Phenotype enrichment maps are presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb. Most participant traits varied along the microbiome similarity network in a linear left-right trajectory aligning with the pathogenicity gradient identified via dominance analysis.\u003c/p\u003e \u003cp\u003ePut differently, there was a gradient of enrichment from left to right that reflected the transition of participant characteristics. Participants at the left end of this gradient exhibited higher severity of clinical periodontitis, elevated circulating inflammatory markers, older age, a higher percentage of smokers (indicated by significant positive enrichment), as well as lower cognitive function, lower brain structural integrity, a less healthy diet, a lower percentage of females, and lower education levels (indicated by significant negative enrichment). Conversely, the right end of the gradient featured participants with opposite traits: lower severity of clinical periodontitis, lower circulating inflammatory markers, younger age, fewer smokers, as well as higher cognitive performance, higher brain structural integrity, a healthier diet, a higher percentage of females, and higher education levels. Mental health scores and vascular risk factors beyond smoking displayed a non-linear enrichment pattern not aligned with the pathogenicity gradient.\u003c/p\u003e\n\u003ch3\u003eCo-enrichment analysis\u003c/h3\u003e\n\u003cp\u003ePrincipal component analysis of enrichment scores revealed that enrichment patterns of phenotypes differed along two dominant axes of inter-phenotype variation explaining 55.69% (principal component 1, PC1) and 17.49% (principal component 2, PC2) of variance, respectively. Phenotypes with similar enrichment patterns co-localized in the principal component space formed by these axes (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e, \u003cem\u003esupplementary figure \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e\u003c/em\u003e). Phenotypes with enrichment patterns indicating increasing values from right end of the network to left were localized on the left extreme of the principal component space including periodontitis-associated bacterial genera, clinical oral health measurements, circulating inflammatory markers, smoking behavior and white matter free-water measured by diffusion-weighted MRI. Phenotypes increasing left to right were located on the right including health-associated bacterial genera, cognitive performance measurements, diet scores, cortical thickness and subcortical volume as measured by MRI. Phenotypes showing a non-linear enrichment pattern or transition pattern in the left-right right-left orientation were localized in the middle including mental health scores and vascular risk measurements apart from smoking. For a heatmap depicting the Spearman correlation of enrichment scores for genus-level abundance and non-microbiome phenotypes see Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e. For the same co-enrichment heatmap indicating genus-genus correlation as well as correlation between non-microbiome phenotypes see \u003cem\u003esupplementary figures \u003cspan refid=\"MOESM3\" class=\"InternalRef\"\u003eS3\u003c/span\u003e and S4\u003c/em\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eMicrobiome similarity groups differ in periodontal disease, inflammatory markers, cognition and diet\u003c/h3\u003e\n\u003cp\u003eApplying k-means clustering on the microbiome similarity network topology resulted in an unsupervised separation of the sample into two participant groups along the pathogenicity gradient. The following statistics to characterize the identified participant groups were adjusted for potential confounders including age, sex, education and vascular risk factors. The procedure is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ea. Participants assigned to nodes in both groups were excluded from the analysis (n\u003csub\u003eexcluded\u003c/sub\u003e=137).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eWe compared genus-level abundance to identify bacterial genera that showed significant differences between groups (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eb and \u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ec). Of the 85 tested genera, 8 showed no significant group differences (\u003cem\u003esupplementary figure S5\u003c/em\u003e). Top 15 bacterial genera which were significantly higher in the \u0026ldquo;A\u0026rdquo; group were \u003cem\u003eTreponema\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e = -1.26), \u003cem\u003eFusobacterium\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e = -1.26), \u003cem\u003eFretibacterium\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e = -1.21), \u003cem\u003eTannerella\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e = -1.09), \u003cem\u003eDialister\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e = -1.01), \u003cem\u003eMogibacterium\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e = -0.98), \u003cem\u003ePorphyromonas\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e = -0.96), \u003cem\u003ePeptostreptococcaceae\u003c/em\u003e [G-4] (β\u003csub\u003estd\u003c/sub\u003e = -0.93), \u003cem\u003ePeptostreptococcaceae\u003c/em\u003e [G-2] (β\u003csub\u003estd\u003c/sub\u003e = -0.91), \u003cem\u003eParvimonas\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e = -0.87), \u003cem\u003eVeillonellaceae\u003c/em\u003e [G-1] (β\u003csub\u003estd\u003c/sub\u003e = -0.85), \u003cem\u003ePeptostreptoccocaceae\u003c/em\u003e [G-9] (β\u003csub\u003estd\u003c/sub\u003e = -0.85), \u003cem\u003ePrevotella\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e = -0.85), \u003cem\u003eCatonella\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e = -0.83), and \u003cem\u003eDesulfobulbus\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e = -0.83) (all p\u003csub\u003eFDR\u003c/sub\u003e \u0026lt; 0.001). Top 15 bacterial genera which were significantly higher in the \u0026ldquo;B\u0026rdquo; group were \u003cem\u003eStreptococcus\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;1.37), \u003cem\u003eHaemophilus\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;1.16), \u003cem\u003eRothia\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;1.02), \u003cem\u003eGranulicatella\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.98), \u003cem\u003eGemella\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.93), \u003cem\u003eRiemerella\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.86), \u003cem\u003eActinomyces\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.82), \u003cem\u003eVeillonella\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.81), \u003cem\u003eNeisseria\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.75), \u003cem\u003eKingella\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.75), \u003cem\u003eCapnocytophaga\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.75), \u003cem\u003eLautropia\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.75), \u003cem\u003eCorynebacterium\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.69), Arachnia (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.59), \u003cem\u003eCardiobacterium\u003c/em\u003e (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.56) (all p\u003csub\u003eFDR\u003c/sub\u003e \u0026lt; 0.001).\u003c/p\u003e \u003cp\u003eIn addition, the \u0026ldquo;A\u0026ldquo; group showed significantly higher clinical severity of periodontitis, higher circulating inflammatory markers, lower cognitive performance and lower brain structural integrity compared to the \u0026ldquo;B\u0026rdquo; group (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ed): a significantly higher clinical attachment loss (β\u003csub\u003estd\u003c/sub\u003e = -0.56, p\u003csub\u003eFDR\u003c/sub\u003e \u0026lt; 0.001), DMFT index (β\u003csub\u003estd\u003c/sub\u003e = -0.15, p\u003csub\u003eFDR\u003c/sub\u003e = 0.047), plaque index (β\u003csub\u003estd\u003c/sub\u003e = -0.32, p\u003csub\u003eFDR\u003c/sub\u003e \u0026lt; 0.001), bleeding on probing index (β\u003csub\u003estd\u003c/sub\u003e = -0.53, p\u003csub\u003eFDR\u003c/sub\u003e \u0026lt; 0.001), leukocytes (β\u003csub\u003estd\u003c/sub\u003e = -0.21, p\u003csub\u003eFDR\u003c/sub\u003e = 0.014) as well as a significantly lower general cognitive ability (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.18, p\u003csub\u003eFDR\u003c/sub\u003e = 0.026), Mini Mental State Exam score (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.22, p\u003csub\u003eFDR\u003c/sub\u003e = 0.014) as well as MIND score (β\u003csub\u003estd\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;0.20, p\u003csub\u003eFDR\u003c/sub\u003e = 0.026). The remaining cognitive scores, diet scores and hsCRP showed no significant differences (\u003cem\u003esupplementary figure S6\u003c/em\u003e). Moreover, brain structural indices and mental health scores also showed no significant differences.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eSensitivity analysis\u003c/h2\u003e \u003cp\u003eTo assess the robustness of our result we performed two-fold sensitivity analyses.\u003c/p\u003e \u003cp\u003eFirst, we conducted repeated enrichment analyses on randomly selected subsamples, reducing the sample size from 100\u0026ndash;10% in 1% decrements, with 100 random samples at each step (\u003cem\u003esupplementary figure S7\u003c/em\u003e). Throughout this process, the Spearman correlation of enrichment ratios (covering both non-microbiome and microbiome phenotypes) demonstrated high robustness (Spearman ρ\u0026thinsp;\u0026gt;\u0026thinsp;0.8) to different sample compositions until the sample size was reduced to approximately 80%. Additionally, the Adjusted Rand Index (ARI) remained high (ARI\u0026thinsp;\u0026gt;\u0026thinsp;0.8), indicating strong consistency in microbiome similarity network-based group assignments from k-means clustering, until the sample size was reduced to approximately 70%.\u003c/p\u003e \u003cp\u003eSecond, to evaluate the potential influence of pipeline design choices on our findings, we reanalyzed the data using 17 different pipeline configurations, varying the components and parameters of the topological data analysis (\u003cem\u003esupplementary table S8\u003c/em\u003e). The results demonstrated high robustness, with Spearman correlations between enrichment ratios from the original and alternative configurations averaging 0.82\u0026thinsp;\u0026plusmn;\u0026thinsp;0.03. Furthermore, the ARI for group assignments derived from k-means clustering via the original pipeline, compared to those from different configurations, remained consistently high at 0.84\u0026thinsp;\u0026plusmn;\u0026thinsp;0.06.\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe oral cavity hosts the second most diverse microbiome in the human body, following the gut.\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e Interactions within these oral microbial communities, along with host factors, affect both oral and systemic health.\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e In this study, we present a comprehensive population-level analysis examining the association between the subgingival microbiome and multi-domain brain health-related phenotypes ranging from cognitive functions, mental health, brain structure, inflammatory blood biomarkers, to dietary behavior and vascular risk factors. Using 16S rRNA sequencing and topological data analysis, we inferred a microbiome similarity network that revealed a continuous pathogenicity gradient, mapping individuals based on periodontal microbiome profiles. Leveraging this network, we identified associations of periodontitis-related microbial communities with multiple brain health phenotypes: Individuals with higher abundance of periodontitis-related taxa exhibited significantly lower cognitive performance, lower MIND diet scores as well as increased leukocyte counts adjusting for demographics and cardiovascular risk. These findings were robust, as confirmed by sensitivity analyses involving diverse random subsamples and varying analytical configurations. Our findings highlight several key insights into the oral microbiome-brain axis, potential pathophysiological pathways and clinical implications.\u003c/p\u003e\n\u003ch3\u003eTopological data analysis reveals a latent axis of periodontitis-related microbial composition\u003c/h3\u003e\n\u003cp\u003eWe performed a dominance analysis to determine which bacterial genera are particularly abundant within specific participant clusters of the microbiome similarity network. Based on this analysis, we show that the microbiome similarity network captured interindividual variations and a gradient of microbial compositions from periodontitis-associated taxa to health-associated taxa (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb): Periodontitis-associated genera such as \u003cem\u003ePorphyromonas\u003c/em\u003e, \u003cem\u003eFusobacterium\u003c/em\u003e, \u003cem\u003eTreponema, Saccharibacteria (TM7)\u003c/em\u003e and \u003cem\u003eCampylobacter\u003c/em\u003e were enriched at the left part of the network, consistent with their established roles in biofilm formation, immune modulation, and periodontal tissue destruction.\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan additionalcitationids=\"CR25 CR26\" citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e Also positioned at this pathogenic end were bacteria of the genus \u003cem\u003eDialister\u003c/em\u003e, which have only recently been associated with periodontitis and could be a key periodontal pathogen, warranting further research into its mechanisms.\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e In contrast, genera such as \u003cem\u003eStreptococcus\u003c/em\u003e, \u003cem\u003eHaemophilus\u003c/em\u003e, \u003cem\u003eRothia, Veillonella and Neisseria\u003c/em\u003e were enriched at the right part, reflecting their roles in maintaining an oral health equilibrium, their overall low periodontal pathogenicity or their association with oral conditions different from periodontitis, such as dental caries.\u003csup\u003e\u003cspan additionalcitationids=\"CR30\" citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e Enrichment ratios for bacterial genera were overall high, affirming that the observed interindividual differences were relevantly captured by the microbiome similarity network.\u003c/p\u003e \u003cp\u003eThese findings indicate that periodontitis emerged as the most parsimonious explanation for the topology of the microbiome similarity network, with participants harboring higher abundances of periodontitis-related taxa clustering at one end, and those with higher abundances of other genera clustering at the other. Put differently, the identified network revealed a latent axis and continuum of periodontitis (referred to as pathogenicity gradient) mirroring the pathogenicity spectrum of periodontitis and suggesting that the disease constitutes a key driver of the observed interindividual variance of microbiome compositions. This evidence demonstrates that a topological data analysis-based approach can uncover subtle yet biologically meaningful patterns of disease severity in large-scale, highly complex datasets.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eMicrobial compositions are linked to non-microbiome phenotypes\u003c/h2\u003e \u003cp\u003eThe microbiome similarity network offers a means to integrate complex, high-dimensional microbiome data with comprehensive health information by mapping non-microbial phenotypes on the network and statistically testing their association with the network structure. We observed that the investigated non-microbiome phenotypes differentially enriched on the network, implying varying degrees of association with microbial configuration. These enrichment patterns ranged from linear enrichments correlated or anti-correlated with the pathogenicity gradient (e.g., leukocytes) \u0026ndash; where variance in these variables coincided with changes in the abundance of periodontitis-related taxa \u0026ndash; to non-linear patterns indicating weaker or more complex associations with the pathogenicity gradient (e.g., PHQ-9) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb). Co-enrichment analyses further detailed these observations, showing that microbial and non-microbial phenotypes with visually overlapping enrichment patterns occupied similar positions in principal component space and exhibited strong correlations in their enrichment scores (Figs.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e and \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e; \u003cem\u003esupplementary figures \u003cspan refid=\"MOESM2\" class=\"InternalRef\"\u003eS2\u003c/span\u003e-S4\u003c/em\u003e). Among the phenotypes most closely aligned with the pathogenicity gradient were the plaque index, clinical attachment loss, and bleeding on probing index, reinforcing periodontitis as the primary pathology captured by the network.\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e,\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e These findings highlight not only the complex covariance structure of subgingival microbiome composition but also the potential of genus-level abundance data as a promising biomarker source, reflecting a broad range of clinical and lifestyle factors.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003ePeriodontal dysbiosis and brain health\u003c/h2\u003e \u003cp\u003eThe mechanisms behind the relationship between periodontitis and brain health are multifaceted and remain to be fully elucidated. Our analysis sought to address these complexities by integrating data on multiple brain health phenotypes into a single analysis framework.\u003c/p\u003e \u003cp\u003eNotably, multiple markers of cognitive performance (verbal fluency, general cognitive ability, Mini Mental State Exam, Memory) and systemic inflammation (leukocyte counts, hsCRP) were among non-oral phenotypes with strongest enrichment ratios (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea) and varied in tandem with the pathogenicity gradient from one network end to the other (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb). Specifically, participants with higher abundance of periodontitis-associated taxa and poorer oral health exhibited significantly lower cognitive performance and elevated systemic inflammation after adjusting for covariates (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ec and \u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ed). Importantly, these findings build on prior reports linking shifts in oral microbiota to cognitive changes among dementia patients and demonstrate that such associations also manifest in a healthy population of non-demented, middle-aged individuals.\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e,\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e Furthermore, our findings align with earlier analyses indicating relationships between clinical markers of periodontitis and cognitive health.\u003csup\u003e\u003cspan additionalcitationids=\"CR36 CR37\" citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eOur analysis highlights associations between brain health and the abundance of various bacterial genera (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eb and \u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ec). These findings are consistent with prior research implicating genera such as \u003cem\u003ePorphyromonas\u003c/em\u003e, \u003cem\u003eTreponema\u003c/em\u003e, \u003cem\u003eFusobacterium\u003c/em\u003e, and \u003cem\u003ePrevotella\u003c/em\u003e in pathways potentially relevant to Alzheimer\u0026rsquo;s disease. Prominently, experimental evidence from animal models demonstrated that \u003cem\u003ePorphyromonas gingivalis\u003c/em\u003e contributes to Alzheimer\u0026rsquo;s pathology by contributing to the formation of amyloid-β, neurofibrillary tangles and neuroinflammation.\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e Furthermore, postmortem studies report a higher abundance of \u003cem\u003ePorphyromonas gingivalis\u003c/em\u003e in the brain samples of AD patients compared to non-demented controls, with intracerebral presence of the bacterium being related to a six-fold increased risk of AD.\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e \u003cem\u003eTreponema denticola\u003c/em\u003e has been demonstrated to induce tau hyperphosphorylation and neuroinflammatory processes in mice.\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e The serum levels of antibodies against \u003cem\u003ePrevotella intermedia\u003c/em\u003e and \u003cem\u003eFusobacterium nucleatum\u003c/em\u003e are significantly increased in Alzheimer's disease patients.\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e Additionally, elevated \u003cem\u003ePrevotella intermedia\u003c/em\u003e abundance is associated with \u003cem\u003eAPOE4\u003c/em\u003e-carrier status, and \u003cem\u003eFusobacterium\u003c/em\u003e species have been linked to cerebrovascular lesions indicative of small vessel pathology.\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e,\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e Our findings align with these results, indicating that the prevalence of these pathogenic bacteria in the subgingival biofilm corresponds to reduced brain health in cognitively normal individuals. At the same time, we found genera not yet reported in relation to brain health \u0026ndash; \u003cem\u003eFretibacterium\u003c/em\u003e, \u003cem\u003eTannerella\u003c/em\u003e, \u003cem\u003eDialister\u003c/em\u003e, \u003cem\u003eMogibacterium\u003c/em\u003e, \u003cem\u003ePeptostreptococcaceae\u003c/em\u003e and \u003cem\u003eParvimonas\u003c/em\u003e \u0026ndash; warranting further experimental research to determine whether they independently drive cognitive deficits or merely coincide with other pathogenic species.\u003c/p\u003e \u003cp\u003eOur findings emphasize the role of periodontitis-related taxa in systemic inflammation and possible connections to neurodegenerative pathways. We observed that individuals positioned at the pathogenic end of the oral microbiome similarity network exhibited significantly higher leukocyte levels after covariate adjustment (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ed), indicating presence of systemic inflammation. This aligns with prior studies linking clinically diagnosed periodontitis to elevated systemic inflammatory markers.\u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e Pathomechanistically, such responses can stem from dense leukocytic infiltration during gingival inflammation, the stimulation of bone marrow for sustained inflammatory cell production, or the systemic response to dissemination of pathogens from ulcerated gingival tissues.\u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e Importantly, we present novel evidence linking periodontitis-associated microbial compositions to systemic inflammation. Notably, this systemic inflammation was accompanied by reduced cognitive performance, supporting the hypothesis that chronic oral infections contribute to both systemic inflammation and cognitive disease. Although our findings cannot prove causality due to the cross-sectional nature of the study design, they align with the inflammatory hypothesis of Alzheimer\u0026rsquo;s disease, suggesting that immune responses triggered by oral pathogens may disrupt the blood-brain barrier and activate microglia, thereby driving neurodegeneration.\u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003ePrevious analyses also highlight an association between periodontal health and brain structure.\u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e,\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e,\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e However, the global brain structural measures we assessed did not show significant group differences after adjustment for covariates. We speculate that this could be because any potential structural effects related to the oral microbiota in this cohort are subtle or regionally specific, rather than global, and thus not captured by the overall brain metrics used. Future analysis should therefore focus on assessing localized brain structural changes, potentially using voxel-based or region-of-interest methods, to determine if subtle or regionally specific associations exist with oral microbial profiles.\u003c/p\u003e \u003cp\u003eIn contrast to cognitive measures, mental health scores displayed weaker associations with the pathogenicity gradient, suggesting more nuanced or less direct links with periodontitis-related microbiome composition. These associations did not persist after adjusting for demographic and cardiovascular risk factors. We interpret these results as pointing to a more specific link of periodontitis to cognitive rather than mental health. Nevertheless, prior studies have reported connections between anxiety, depression, and clinical periodontitis or oral microbiome composition.\u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e Longitudinal research will be essential to elucidate how these factors collectively influence oral and brain health.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eInfluence of diet and vascular risk factors on oral and brain health\u003c/h2\u003e \u003cp\u003eWe investigated the role of nutritional behavior as a key factor influencing oral microbiome composition.\u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e Enrichment analysis revealed that adherence to cognitively beneficial diets, such as the MEDAS, MIND, and DASH diets,\u003csup\u003e\u003cspan additionalcitationids=\"CR49\" citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e aligned with the pathogenicity gradient: participants with greater abundance of periodontitis-related taxa tended to adhere less to these diets. Notably, after adjusting for covariates, individuals with a higher abundance of periodontitis-related genera exhibited significantly lower adherence to the MIND diet. These findings corroborate previous results from the HCHS, which demonstrated that higher adherence to DASH, MEDAS and an anti-inflammatory dietary was associated with lower odds of periodontal disease.\u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e Our findings suggest that dietary patterns known to promote cognitive health may also shape the oral microbiome toward a less pathogenic composition. Given that the MIND diet specifically emphasizes foods with anti-inflammatory and neuroprotective properties,\u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e it is plausible that such nutritional patterns could influence both oral and brain health through microbial community shifts and immune modulation. A healthier diet may foster a healthier oral microbiome based on its emphasis on fiber over simple sugars selectively promoting beneficial microbes while limiting resources for harmful ones.\u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e Furthermore, it could reduce gingival inflammation through anti-inflammatory compounds, making the environment less hospitable to pathogens potentially disrupting brain health.\u003csup\u003e\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e\u003c/sup\u003e Longitudinal and interventional studies are needed to disentangle these pathways and clarify causal relationships between diet, oral microbial ecology, and brain health.\u003c/p\u003e \u003cp\u003eAmong demographic and vascular risk factors, smoking behavior was the only factor clearly associated with the pathogenicity gradient. Smoking has been documented to promote the colonization of periodontal pathogens, facilitate biofilm formation, and compromise the immune response, thereby exacerbating periodontal disease and systemic inflammation.\u003csup\u003e\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u003c/sup\u003e Our findings highlight the role of smoking on promoting dysbiosis and inflammation and positions it as a direct contributor to the pathogenicity gradient and brain health phenotypes.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eClinical implications\u003c/h2\u003e \u003cp\u003eOur findings not only shed light on the intricate systemic pathophysiology of periodontal dysbiosis and brain health, but they also hint at potential avenues of clinical utilization. Our study reveals an association between subgingival microbiome signatures and variance in cognitive performance and brain structure that could enhance early screening measures for risk of cognitive decline and improve targeted recruitment of individuals at critical early stages of cognitive impairment and dementia. Particularly, identifying microbial community changes before emergence of cognitive symptoms could enhance diagnostics by providing microbiome-based markers as a convenient complement to existing dementia biomarkers. Regarding future therapeutic interventions, microbiome signatures of early periodontal dysbiosis may guide the development of oral microbiome-directed therapies to slow or prevent dementia progression.\u003csup\u003e\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e\u003c/sup\u003e However, the definitive role of microbiome biomarkers in cognitive disorders is yet to be determined and large-scale longitudinal as well as interventional studies are required for moving in this direction.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eStrengths and limitations\u003c/h2\u003e \u003cp\u003e Strengths of this work lie in its considerable sample size; comprehensive taxonomic profiling of the oral microbiome; in-depth phenotyping of clinical and lifestyle data; as well as a novel, robust data analysis pipeline that unifies multiple complex data domains into a single analysis framework. However, our study also exhibits limitations. Due to the cross-sectional nature of the performed microbiome-phenotype association analyses, our findings cannot firmly establish causal links and should be regarded as hypothesis-generating. Specifically, the design cannot preclude the possibility of reverse causation; for example, individuals with poorer cognitive function might face challenges in maintaining optimal oral hygiene, which could subsequently influence their periodontal health status and subgingival microbiome composition. However, it is important to note that our associational findings are substantiated by experimental research in animal models which provides evidence that periodontal pathogens can indeed trigger neuroinflammation and contribute to neurodegenerative pathology, lending weight to a potential contributing causal role.\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e Moreover, even after adjusting for confounders in our statistical models, we cannot fully exclude their potential influence on the relationship between subgingival microbiome composition and non-microbiome phenotypes. Longitudinal and experimental studies are needed that expand on our findings to further discern the effects of the oral microbiome on brain health. Additionally, data on participant relatedness or household sharing were unavailable; these factors are known to influence microbiome similarity and could therefore potentially act as unmeasured confounders in our analyses. Finally, our reliance on 16S rRNA gene sequencing limits the current analysis primarily to taxonomic composition, precluding direct assessment of microbial functional potential or metabolic activity. While computational tools can predict function from 16S data, integrating such inferred analyses was beyond the scope of this initial broad, multi-domain study focused on taxonomic associations. Our planned next steps include exploring computationally inferred functions (e.g., using PICRUSt2\u003csup\u003e57\u003c/sup\u003e) on the existing dataset. However, for direct functional characterization, investigation of specific metabolic pathways, resolution of strain-specific taxonomic variations (including virulence factors), and identification of potential non-bacterial microbiome members, future studies utilizing high-resolution methods such as shotgun metagenomics will be essential to gain deeper mechanistic insights into the oral microbiome-brain axis.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eDrawing on a comprehensive investigation integrating multi-domain data with cutting-edge analysis techniques we characterized latent associations of periodontitis-related subgingival microbial composition with cognitive function, mental health, brain structural integrity, systemic inflammation, diet and vascular risk in predominantly healthy individuals. Notably, periodontitis-related oral dysbiosis was associated with lower cognitive performance, lower cortical and subcortical volume, and higher leukocytes. As this research field progresses, oral microbiome profiling could contribute to improved dementia risk stratification and guide preventive interventions.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eStudy design\u003c/h2\u003e \u003cp\u003eOur methodology is illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. This analysis integrates different data domains related to oral and brain health into a unified analysis framework: subgingival microbiome composition, oral health status, cognitive function, mental health status, brain structure, circulating inflammatory markers, dietary patterns, vascular risk and demographics. In brief, we employed topological data analysis, specifically the Mapper algorithm, to create a low-dimensional network representation of subgingival microbiome abundance data.\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e\u003c/sup\u003e This network positions individuals based on similarities in their subgingival microbiome composition. Subsequently, we conducted an enrichment analysis using Spatial Analysis of Functional Enrichment (SAFE).\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e This analysis statistically assesses regions of the network derived via Mapper to identify where specific phenotypes are significantly higher or lower than expected by chance. This approach allowed us to examine regional differences in phenotypes, revealing whether the network representation captures variance in specific participant traits related to oral and brain health. Finally, we performed a post-hoc group comparison between participants of two distinct participant clusters within the microbiome network, enabling us to assess the link of subgingival microbiome composition and clinical and lifestyle phenotypes while adjusting for relevant covariates.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eStudy population\u003c/h2\u003e \u003cp\u003ePAROMIND is a cross-sectional study nested within the Hamburg City Health Study (HCHS). The HCHS is a prospective, single-center, population-based cohort study investigating adults aged 45\u0026ndash;75 to enhance the detection of major chronic disease risks through extensive clinical and imaging phenotyping.\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e Participants were included in PAROMIND for periodontal examination and subgingival sampling if they reported no antibiotic use within the preceding three months, had no requirement for endocarditis prophylaxis, and possessed more than two remaining teeth. The study protocol includes assessments of oral health (including collection of gingival crevicular fluids for microbiome analyses), cognition, mental health, diet, vascular risk, brain MRI, and blood sampling. For each participant, the clinical assessments and biological sampling were generally completed during a single comprehensive baseline visit; the separate brain MRI appointment followed shortly thereafter, ensuring reasonable contemporaneity between these measures for cross-sectional analysis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003eEthics statement\u003c/h2\u003e \u003cp\u003e PAROMIND and the HCHS were approved by the local ethics committee of the Landes\u0026auml;rztekammer Hamburg (State of Hamburg Chamber of Medical Practitioners, PV5131). The conduct of PAROMIND is governed by ethical guidelines of Good Clinical Practice (GCP), Good Epidemiological Practice (GEP) and the Declaration of Helsinki.\u003csup\u003e\u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e\u003c/sup\u003e Written informed consent was obtained from all participants investigated in this work.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003eMolecular analysis and phenotyping of the oral microbiome\u003c/h2\u003e \u003cp\u003eOral microbiome phenotyping followed a standardized procedure targeting the subgingival environment within periodontal pockets. This specific environment was chosen because it contains the subgingival biofilms most directly implicated in periodontitis pathogenesis, distinguishing its microbial community from those in other oral sites like saliva or supragingival plaque. During the dental examination, gingival crevicular fluid samples were collected from periodontal pockets with sterile paper points. Two samples per participant were obtained: (1) from a single deep periodontal pocket and (2) one pooled sample (4 paper points) from the deepest periodontal pockets per quadrant. Each paper point remained in situ for 15 seconds. Subsequently, samples were placed in a sterile 2ml-Eppendorf tube and stored at -80\u0026deg;C in the HCHS biobank. For the following molecular analysis, only pooled samples were further processed.\u003c/p\u003e \u003cp\u003eBacterial composition was determined via 16S rRNA Illumina sequencing. DNA Isolation was performed using the DNA extraction kit from Innuprep Analytics (Analytik Jena AG, \u0026Uuml;berlingen, Germany). For the initial lysis with lysozyme and mutanolysin (3 mg lysozyme, 100 U mutanolysin, in 200 \u0026micro;l Tris EDTA buffer), all samples were incubated at 37\u0026deg;C for 10 min and further isolation was performed according to the manual with elution volumes of 100 \u0026micro;l. For 16S rDNA sequencing of isolated DNA, variable regions V3 and V4 of the 16S rRNA gene were amplified using the primer pair 341F (5\u0026lsquo;-CCTACGGGAGGCAGCAG-3\u0026lsquo;) and 806R (5\u0026lsquo;-GGACTACHVGGGTWTCTAAT-3\u0026rsquo;) 27F-338R in a dual barcoding approach.\u003csup\u003e\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e,\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e\u003c/sup\u003e Agarose gel electrophoresis was used to verify resulting PCR products before normalization using the SequalPrep Normalization Plate Kit (Thermo Fischer Scientific, Waltham, MA, USA), pooling and sequencing on the Illumina MiSeq with v3 2x300bp chemistry (Illumina Inc., San Diego, CA, USA). Demultiplexing after sequencing was based on 0 mismatches in the barcode sequences.\u003c/p\u003e \u003cp\u003eData processing was performed using the DADA2 workflow for big datasets (v. 1.10.42, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://benjjneb.github.io/dada2/bigdata.html\u003c/span\u003e\u003cspan address=\"https://benjjneb.github.io/dada2/bigdata.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), resulting in abundance tables of amplicon sequence variants (ASVs). For this, all sequencing runs were handled separately and finally collected in a single abundance table per dataset, which underwent chimera filtering. ASVs underwent taxonomic annotation using the Bayesian classifier provided in DADA2 and using the expanded Human Oral Microbiome Project (eHOMD) version V15.23. Samples with less than 10,000 sequences were not considered for further analysis. Downstream analyses were performed at the genus-level because this taxonomic rank provided a suitable balance, enhancing statistical power across the large cohort, improving interpretability and comparability with existing literature, and reducing data dimensionality while still providing sufficient resolution for the study's objectives. Genera that were observed with a frequency of less than 0.1% of all genera detected in a sample were discarded.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003eOral health assessment\u003c/h2\u003e \u003cp\u003eA certified study nurse assessed probing depth and gingival recession at six sites of the tooth (mesio-buccal, buccal, disto-buccal, disto-palatinal, palatinal and mesio-palatinal). The clinical attachment loss (CAL) was calculated (CAL\u0026thinsp;=\u0026thinsp;probing depth\u0026thinsp;+\u0026thinsp;gingival recession).\u003csup\u003e\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e\u003c/sup\u003e The BOP index was determined by probing 2 sites per tooth (vestibular, oral) and expressed as a percentage of bleeding sites. The number of bleeding sites on probing is a reliable and consistent measure to assess the degree of gingival inflammation. Measurements were completed with a standard periodontal probe (PCP 15, Hu-Friedy, Chicago, IL, USA). Additionally, the DMFT index was calculated. The oral health assessment followed a standardized protocol for the reporting of the prevalence and severity of periodontal diseases.\u003csup\u003e\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003eCognitive and mental health assessments\u003c/h2\u003e \u003cp\u003eCognitive testing was conducted using the extended version of the Consortium to Establish a Registry for Alzheimer\u0026rsquo;s Disease Neuropsychological Assessment Battery (CERAD-NP/Plus).\u003csup\u003e65\u003c/sup\u003e A trained study nurse administered all tests. For this analysis, we considered cognitive scores measuring executive function (Trail Making Test B), information processing speed (Trail Making Test A), memory (Word List Recall Test), reasoning (Multiple Choice Vocabulary Intelligence Test B), verbal fluency (Animal Naming Test), and the Mini Mental State Exam. To ensure higher scores indicated better cognitive performance across all tests, we inverted the results of Trail Making Test A and B. Subsequently, we performed a principal component analysis (PCA) on all individual test scores. Following previous procedures, the first principal component, which accounted for the greatest variance (40.5%), was defined as a measure of general cognitive ability (g) (for details see \u003cem\u003esupplementary figure S9\u003c/em\u003e).\u003csup\u003e\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e\u003c/sup\u003e According to the principal component loadings, higher values of this measure corresponded to lower cognitive performance; thus, it was also inverted. Furthermore, participants underwent mental health assessments via established questionnaires for depression (PHQ-9, Geriatric Depression Scale), somatic symptom severity (PHQ-15), and anxiety (GAD-7).\u003csup\u003e\u003cspan additionalcitationids=\"CR68\" citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003eNeuroimaging of brain macro- and microstructure\u003c/h2\u003e \u003cp\u003eNeuroimaging markers representing different aspects of macro- and microstructural brain integrity were computed based on T1-weighted and diffusion-weighted MRI following previous procedures.\u003csup\u003e\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e\u003c/sup\u003e After cortical surface reconstruction and subcortical segmentation based on T1-weighted images with FreeSurfer (v. 6.0.1), cortical thickness and subcortical volume were estimated representing morphometric measures of neurodegenerative processes.\u003csup\u003e\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e,\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e\u003c/sup\u003e The mean cortical thickness and mean subcortical volume were z-scored. Subsequently, the two resulting z scores were averaged to obtain a single summary measure reflecting cortical thickness and subcortical volume. Following preprocessing of the diffusion-weighted images, free-water imaging was employed to compute free-water which quantifies the amount of extracellular water as well as tissue fractional anisotropy reflecting neurite architecture and integrity.\u003csup\u003e\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e,\u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e\u003c/sup\u003e The free-water and tissue fractional anisotropy values were then averaged across cortical and subcortical gray matter voxels, as well as white matter voxels, to obtain global measures of gray and white matter, respectively. For a detailed account on the acquisition protocol, quality assessment, preprocessing and computation procedures on the different imaging measures see \u003cem\u003esupplementary text S10\u003c/em\u003e.\u003c/p\u003e \u003cdiv id=\"Sec25\" class=\"Section3\"\u003e \u003ch2\u003eDiet\u003c/h2\u003e \u003cp\u003eThe dietary behavior of all participants was assessed using a validated food frequency questionnaire with 102 items, developed for the European Prospective Investigation into Cancer and Nutrition Study (EPIC).\u003csup\u003e\u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e75\u003c/span\u003e\u003c/sup\u003e The adherence to different dietary patterns was measured based on the food frequency questionnaire scores, including the Mediterranean diet (MEDAS diet), the Dietary Approaches to Stop Hypertension (DASH diet), and the Mediterranean-DASH Intervention for Neurodegenerative Delay (MIND diet).\u003csup\u003e\u003cspan additionalcitationids=\"CR49\" citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e Adherence to the Mediterranean diet was determined using the German version of the Mediterranean Diet Adherence Screener, which assigns a score of 0 or 1 to 14 specific food items, producing a total adherence score ranging from 0 (no adherence) to 14 (maximum adherence).\u003csup\u003e\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e The DASH diet was evaluated using a previously established scoring method that assigns a score of 0, 0.5, or 1 to each of 10 items, resulting in an adherence score between 0 (no adherence) and 10 (maximum adherence).\u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e Finally, adherence to the MIND diet was calculated according to standard procedures: scores of 0, 0.5, or 1 to are assigned 10 healthy and 5 unhealthy food items, culminating in a total adherence score ranging from 0 (no adherence) to 15 (maximum adherence).\u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section3\"\u003e \u003ch2\u003eTopological data analysis\u003c/h2\u003e \u003cp\u003eWe implemented a topological data analysis pipeline that integrates two key components: (1) the Mapper algorithm, which performs an unsupervised reconstruction of a topological network based on genus-level abundance data, capturing microbiome composition similarity, and (2) SAFE, which conducts statistical tests to examine the relationship between the network's structure and different phenotypes.\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e,\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e This method effectively combines dimensionality reduction with topological insights, offering a powerful tool for understanding the intrinsic geometry of high-dimensional microbiome data and its relationships with other data domains. The analysis was performed in \u003cem\u003epython\u003c/em\u003e v3.8.1 based on the packages \u003cem\u003eNetworkX\u003c/em\u003e v2.2 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/networkx/networkx\u003c/span\u003e\u003cspan address=\"https://github.com/networkx/networkx\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), \u003cem\u003esafepy\u003c/em\u003e (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/baryshnikova-lab/safepy\u003c/span\u003e\u003cspan address=\"https://github.com/baryshnikova-lab/safepy\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), \u003cem\u003escikit-learn\u003c/em\u003e v1.5.1 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/scikit-learn/scikit-learn\u003c/span\u003e\u003cspan address=\"https://github.com/scikit-learn/scikit-learn\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) and \u003cem\u003etmap\u003c/em\u003e v1.2 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/GPZ-Bioinfo/tmap\u003c/span\u003e\u003cspan address=\"https://github.com/GPZ-Bioinfo/tmap\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) as well as \u003cem\u003eR\u003c/em\u003e v4.4.0 based on the package \u003cem\u003evegan v2.6-6.1\u003c/em\u003e.\u003csup\u003e\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e,\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e Data visualization was based on \u003cem\u003eplotly\u003c/em\u003e v5.22 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/plotly/plotly.py\u003c/span\u003e\u003cspan address=\"https://github.com/plotly/plotly.py\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e) and iTOL v6 (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://itol.embl.de/\u003c/span\u003e\u003cspan address=\"https://itol.embl.de/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). HTML versions of many presented plots can be found on OSF allowing interactive data exploration (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://osf.io/vqj8m/\u003c/span\u003e\u003cspan address=\"https://osf.io/vqj8m/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section3\"\u003e \u003ch2\u003eMapper: Reconstruction of the microbiome similarity network\u003c/h2\u003e \u003cp\u003eGenus-level subgingival microbiome abundance data served as input to the Mapper algorithm, a topological data analysis technique that simplifies complex high-dimensional data by constructing a topological network capturing essential relationships and patterns in the data. This network preserves the data's underlying topological and geometric structure by positioning participants with similar subgingival microbiome profiles nearby. Conceptually, this representation is analogous to a topographical map that reveals the essential features of a landscape. Importantly, the network can represent non-linear associations that conventional linear techniques might miss. Mapper has previously been used to analyze the dynamic organization of brain function,\u003csup\u003e\u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e76\u003c/span\u003e,\u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e77\u003c/span\u003e\u003c/sup\u003e the shape of genetic data in breast-cancer patients,\u003csup\u003e\u003cspan citationid=\"CR78\" class=\"CitationRef\"\u003e78\u003c/span\u003e\u003c/sup\u003e biomolecular folding pathways,\u003csup\u003e\u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e79\u003c/span\u003e\u003c/sup\u003e brain structure in patients with fragile X syndrome,\u003csup\u003e\u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e80\u003c/span\u003e\u003c/sup\u003e and neuronal data from the visual cortex.\u003csup\u003e\u003cspan citationid=\"CR81\" class=\"CitationRef\"\u003e81\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eThe applied Mapper pipeline comprises multiple analysis steps to reconstruct the topological network (for an illustration see Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb): filtering, covering, clustering and network reconstruction.\u003csup\u003e\u003cspan citationid=\"CR82\" class=\"CitationRef\"\u003e82\u003c/span\u003e\u003c/sup\u003e First, as the filtering step, we conducted a principal coordinate analysis of the Aitchison distance of genus-level abundance (n\u003csub\u003eparticipants\u003c/sub\u003e x n\u003csub\u003egenera\u003c/sub\u003e).\u003csup\u003e\u003cspan citationid=\"CR83\" class=\"CitationRef\"\u003e83\u003c/span\u003e\u003c/sup\u003e By that, we obtained two components \u0026ndash; also called lenses \u0026ndash; capturing the major axes of variation in the microbial community composition across the participants. Based on these axes, overlapping covers were defined (overlap\u0026thinsp;=\u0026thinsp;1.5, resolution\u0026thinsp;=\u0026thinsp;30) to segment the data into overlapping bins, each representing a local region of inter-individual variation. Unsupervised clustering of datapoints within each bin was performed using Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN, epsilon threshold\u0026thinsp;=\u0026thinsp;0.95).\u003csup\u003e84\u003c/sup\u003e By this, nodes are obtained that represent participant groups with similar configuration of the subgingival microbiome. Participants can belong to multiple nodes with the number varying per participant. Lastly, the network reconstruction was accomplished by connecting clusters sharing common participants. Of note, not all datapoints are retained by Mapper resulting in the omission of some participants (n\u003csub\u003enot retained\u003c/sub\u003e = 109).\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section2\"\u003e \u003ch2\u003eSAFE: Enrichment analysis\u003c/h2\u003e \u003cp\u003eSAFE is an annotation technique for biological networks that enables the computation and statistical assessment of local enrichment for specific phenotypes.\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e In our work, we performed SAFE on the microbiome similarity network derived using the Mapper algorithm to identify regions within the network that are significantly enriched for specific participant traits. Specifically, we investigated the enrichment of genus-level microbiome abundance, oral health measures (clinical attachment loss, plaque index, bleeding on probing index, DMFT index), cognitive scores (general cognitive ability, Animal Naming Test, Mini Mental State Exam, Multiple Choice Vocabulary Intelligence Test B, Trail Making Tests A and B, Word List Recall), mental health scores (PHQ-9, PHQ-15, Geriatric Depression Scale, GAD-7), imaging measures (cortical thickness and subcortical volume, gray matter free-water, white matter free-water, gray matter tissue fractional anisotropy, white matter tissue fractional anisotropy), circulating inflammatory markers (hsCRP, leukocytes), diet scores (MEDAS score, DASH score, MIND score), vascular risk factors (systolic and diastolic blood pressure, body mass index, smoking behavior, blood triglycerides, cholesterol, low density lipoprotein, high density lipoprotein, HbA1c), and demographics (age, sex, education).\u003c/p\u003e \u003cp\u003eFollowing previous analyses leveraging SAFE,\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e,\u003cspan citationid=\"CR85\" class=\"CitationRef\"\u003e85\u003c/span\u003e\u003c/sup\u003e the microbiome similarity network was spring-embedded, i.e., nodes in the network were positioned so that those with connections are placed closer together, while repulsive forces push non-connected nodes apart, resulting in a visually balanced and interpretable representation of the network's structure.\u003csup\u003e\u003cspan citationid=\"CR86\" class=\"CitationRef\"\u003e86\u003c/span\u003e\u003c/sup\u003e Next, we computed node attributes for all phenotypes by averaging the values of each variable across all participants within a node. Subsequently, enrichment scores were calculated for each node following a four-step process (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ec). First, we defined the local neighborhood by identifying all nodes within a maximum distance threshold of 0.75 from the central node. The distance was measured using the map-weighted shortest path length (MSPL).\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e Second, we calculated a neighborhood score by summing the attribute values of neighboring nodes. Third, we computed a p-value by comparing the empirical neighborhood score against a distribution derived from 5000 permutations. Permutations were performed by randomly reassigning attributes to nodes while preserving the network topology.\u003csup\u003e\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e\u003c/sup\u003e The resulting p-value was corrected for multiple comparisons across all phenotypes. Finally, we assigned an enrichment score to the neighborhood center by applying a -log10 transformation to the corrected p-value. Given the 5000 permutations, the maximal enrichment score is -log10(1/5000)\u0026thinsp;=\u0026thinsp;3.70, with -log10(0.05)\u0026thinsp;=\u0026thinsp;1.30 indicating significance. Positive enrichment scores indicate that observed values are higher than the permuted distribution, negative enrichment scores indicate that they are lower than the permuted distribution. This procedure is repeated for each node of the network, resulting in an enrichment map indicating where attributes are higher or lower than expected by chance.\u003c/p\u003e \u003cp\u003eTo understand the primary taxonomic patterns shaping the network topology, we performed a dominance analysis alongside examining individual phenotype enrichments. This involved labeling each network node by the single bacterial genus with the highest positive enrichment score. Visualizing this node-level dominance helps to reveal major taxonomic transitions across the network. While highlighting the most strongly enriched genus in each region provides valuable pointers to key drivers, this is a simplification; each node still represents a complex microbial community, not just the dominant genus identified.\u003c/p\u003e \u003cp\u003eTo measure how strongly the microbiome similarity network reflects a specific phenotype we computed the enrichment ratio as the number of significantly enriched nodes (corrected p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) divided by the total number of nodes. A higher enrichment ratio indicates that more nodes are significantly enriched, suggesting that the network\u0026rsquo;s topology captures a greater extent of a phenotype\u0026rsquo;s variance. As a complement, we used three different statistical methods from the \u003cem\u003evegan\u003c/em\u003e package to test the linear association between non-microbiome phenotypes and microbiome configuration: 1) \u003cem\u003eenvfit\u003c/em\u003e (n\u003csub\u003epermutations\u003c/sub\u003e = 10,000) testing the linear association between each phenotype and the PCoA ordination (all components) of the Aitchison distance matrix; 2) adonis (n\u003csub\u003epermutations\u003c/sub\u003e = 5,000) testing the proportion of variance in the Aitchison distance matrix explained by each phenotype individually, and capscale (n\u003csub\u003epermutations\u003c/sub\u003e = 5,000) testing the extent to which the ordination based on the Aitchison distance matrix is constrained by each phenotype individually. All p-values of the complementary approaches were false discovery rate-corrected. The complementary linear analyses were only performed for non-microbiome phenotypes and not for abundance of individual genera.\u003c/p\u003e \u003cp\u003eTo determine whether specific phenotypes co-enrich, i.e., exhibit similar enrichment patterns, we performed an ordination of the enrichment scores using principal component analysis retaining the first two principal components and assessed the pairwise Spearman correlation between the enrichment scores.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec29\" class=\"Section2\"\u003e \u003ch2\u003eGroup analysis\u003c/h2\u003e \u003cp\u003eTo further examine the relationship between the microbiome similarity network and clinical as well as lifestyle phenotypes, we performed a post-hoc group analysis which allowed us to adjust for relevant covariates. Therefore, nodes of the microbiome similarity network were clustered in two non-overlapping groups using k-Means clustering of the node positions. We chose k-Means as it is widely used and arguably represents the simplest unsupervised clustering technique.\u003csup\u003e\u003cspan citationid=\"CR87\" class=\"CitationRef\"\u003e87\u003c/span\u003e\u003c/sup\u003e k-Means requires to predefine the number of clusters to assign datapoints to. Given that enrichment analysis indicated that most participant traits varied along the microbiome similarity network in a linear left-right trajectory, i.e., participants on the left end differed from those on the right end, the number clusters was predefined as n\u003csub\u003eclusters\u003c/sub\u003e = 2 post-hoc.\u003c/p\u003e \u003cp\u003eAfter the clustering, participants were categorized based on the resulting groupings. Importantly, individuals present in both groups \u0026ndash; due to being assigned to nodes in both groups \u0026ndash; were excluded from the analysis (n\u0026thinsp;=\u0026thinsp;137). Given that not all datapoints are retained by the Mapper algorithm during the network reconstruction step, participants that were not represented in the microbiome similarity network were not considered for this analysis. The groups were statistically compared for the phenotypes using multiple linear regression and age, sex and education and vascular risk factors (systolic and diastolic blood pressure, body mass index, smoking behavior, triglycerides, cholesterol, LDL, HDL, HbA1c) were included as covariates:\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:Phenotype\\:\\sim\\:Group+Age+Sex+Education+Vascular\\:risk\\:factors$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eCovariates were selected a priori based on literature demonstrating their potential influence on both oral health/microbiome and brain health indices.\u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e,\u003cspan additionalcitationids=\"CR89\" citationid=\"CR88\" class=\"CitationRef\"\u003e88\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR90\" class=\"CitationRef\"\u003e90\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eSensitivity analyses\u003c/h3\u003e\n\u003cp\u003eTo verify the robustness of our results against variations in the sample as well as the analysis pipeline, we conducted a comprehensive sensitivity analysis.\u003c/p\u003e \u003cp\u003eThis involved repeating the entire analysis across random subsamples of different sizes. Specifically, we varied the sample sizes from 100\u0026ndash;10% of the total dataset, in 1% decrements. For each decrement, we randomly sampled subsets 100 times, resulting in a total of 9,000 iterations (90 different sample sizes \u0026times; 100 random samples per steps). For each iteration, we assessed the robustness of our findings by comparing the results from the subsamples to the original results. The stability of the enrichment analysis results was evaluated by calculating the Spearman correlation of enrichment ratios for the non-microbiome phenotypes and genus-level abundance. In addition to this, we measured the agreement of the participant-group assignments resulting from k-Means clustering used for group analysis employing the Adjusted Rand Index (ARI) which ranges from 0 (no agreement) to 1 (full agreement).\u003c/p\u003e \u003cp\u003eThe parameters and components of our analysis pipeline were selected based on established practices, including strategies to optimize sample coverage as described in the tmap documentation, or utilized default software settings. Recognizing that pipeline parameters can influence resulting network characteristics (e.g., node count, connectivity), we conducted a sensitivity analysis using alternative configurations to verify that our results were not biased by these initial design choices. We systematically explored variations in parameter values and pipeline components, altering the Mapper cover overlap from the original 1.5 to 1, 1.2, 1.4, 1.6, 1.8 and 2; the Mapper cover resolution from 30 to 20, 25, 35, 40 and 45; and the Mapper epsilon threshold from 0.95 to 0.99, and 0.90. Additionally, we adjusted the SAFE distance threshold from 0.75 to 0.5 and 0.99, and the SAFE neighborhood radius from 0.1 to 0.05 and 0.15. We adjusted one parameter from the original pipeline per iteration, resulting in a total of 15 iterations. The robustness of our results was evaluated by comparing the findings from the original configuration to those from alternative setups. Matching the approach for the previous sensitivity analysis, we calculated the Spearman correlation of enrichment ratios for clinical and genus abundance phenotypes and the ARI for group assignments from k-Means clustering to assess stability.\u003c/p\u003e "},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHCHS data can be obtained by qualified researchers on reasonable request to the study\u0026rsquo;s steering committee. The analysis code for this work is publicly available on GitHub (https://github.com/csi-hamburg/oral_microbiome_brain_health). Interactive versions of the plots can be found on OSF (https://osf.io/vqj8m/).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors would like to acknowledge all participants, all cooperation partners, patrons and the Deanery from the University Medical Center Hamburg-Eppendorf for supporting the HCHS. Special thanks are due to the staff at the Population Health Research Department for conducting the study. The publication of its results has been approved by the Steering Board of the HCHS. We also thank the staff of the microbiome and sequencing laboratories of the Institute of Clinical Molecular Biology Kiel for their excellent support.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eJG has\u0026nbsp;received speaker fees from Lundbeck, Janssen-Cilag,\u0026nbsp;Lilly, Otsuka and Boehringer outside the submitted work. JF reported receiving personal fees from Acandis, Cerenovus, Microvention, Medtronic, Phenox, and Penumbra; receiving grants from Stryker and Route 92; being managing director of eppdata; and owning shares in Tegus and Vastrax; all outside the submitted work. RT is\u0026nbsp;a co-inventor of an international patent on the use of a computing device to estimate the probability of myocardial infarction (PCT/EP2021/073193, International Publication Number WO2022043229A1). RT is\u0026nbsp;shareholder of the company ART-EMIS GmbH Hamburg. GT has received fees as consultant or lecturer from Acandis, Alexion, Amarin, Bayer, Boehringer Ingelheim, BristolMyersSquibb/Pfizer, Daichi Sankyo, Portola, and Stryker outside the submitted work. The remaining authors declare no conflicts of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEach author has made a significant contribution to the manuscript and all authors read and approved its final version. We describe contributions to the paper using the CRediT contributor role taxonomy. M.P.: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Visualization, Writing\u0026mdash;original draft, Writing\u0026mdash;review \u0026amp; editing; C.W.: Conceptualization, Data curation, Investigation, Methodology, Resources, Writing\u0026mdash;original draft, Writing\u0026mdash;review \u0026amp; editing; K.B.: Data curation, Methodology, Investigation, Resources, Software, Writing\u0026mdash;review \u0026amp; editing; G.H.: Resources, Writing\u0026mdash;review \u0026amp; editing; T.B.: Resources, Writing\u0026mdash;review \u0026amp; editing; M.A.: Resources, Writing\u0026mdash;review \u0026amp; editing; C.M.: Resources, Writing\u0026mdash;review \u0026amp; editing; F.L.N.: Data curation, Resources, Writing\u0026mdash;review \u0026amp; editing; B.Z.: Resources, Writing\u0026mdash;review \u0026amp; editing; J.F.: Resources, Writing\u0026mdash;review \u0026amp; editing; J.G.: Resources, Writing\u0026mdash;review \u0026amp; editing; S.K.: Resources, Writing\u0026mdash;review \u0026amp; editing; R.T.: Resources, Writing\u0026mdash;review \u0026amp; editing; C.B.: Methodology, Writing\u0026mdash;review \u0026amp; editing; G.T.: Supervision, Funding, Writing\u0026mdash;review \u0026amp; editing; B.C.: Conceptualization, Funding, Project administration, Resources, Supervision, Writing\u0026mdash;original draft, Writing\u0026mdash;review \u0026amp; editing; G.A.: Conceptualization, Funding, Project administration, Resources, Supervision, Writing\u0026mdash;original draft, Writing\u0026mdash;review \u0026amp; editing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation): PAROMIND \u0026ndash; 514762487 \u0026ndash; GZ AA 93/9-1 and GZ CH 2631/4-1 (B.C., G.A.); Sonderforschungsbereich 936 \u0026ndash; 178316478 \u0026ndash; C2 (G.T., B.C.); and Schwerpunktprogramm 2041 \u0026ndash; 454012190 (G.T.). Microbiome sequencing received infrastructure support from the DFG Research Unit 5042 \u0026bdquo;miTarget\u0026quot; and the DFG Excellence Cluster 2167 \u0026quot;Precision Medicine in Chronic Inflammation\u0026quot; (PMI).\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eBaker JL, Welch M, Kauffman JL, McLean KM, J. S., He X (2024) The oral microbiome: diversity, biogeography and human health. Nat Rev Microbiol 22:89\u0026ndash;104\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen C-K, Wu Y-T, Chang Y-C (2017) Association between chronic periodontitis and the risk of Alzheimer\u0026rsquo;s disease: a retrospective, population-based, matched-cohort study. Alz Res Therapy 9:56\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNilsson H, Sanmartin Berglund J, Renvert S (2018) Longitudinal evaluation of periodontitis and development of cognitive decline among older adults. J Clin Periodontol 45:1142\u0026ndash;1149\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIwasaki M et al (2019) Periodontitis, periodontal inflammation, and mild cognitive impairment: A 5-year cohort study. J Periodontal Res 54:233\u0026ndash;240\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChoi S et al (2019) Association of Chronic Periodontitis on Alzheimer\u0026rsquo;s Disease or Vascular Dementia. J Am Geriatr Soc 67:1234\u0026ndash;1239\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKamer AR, Craig RG, Niederman R, Fortea J, de Leon (2020) M. J. Periodontal disease as a possible cause for Alzheimer\u0026rsquo;s disease. Periodontol 2000 83:242\u0026ndash;271\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDioguardi M et al (2020) The Role of Periodontitis and Periodontal Bacteria in the Onset and Progression of Alzheimer\u0026rsquo;s Disease: A Systematic Review. J Clin Med 9:495\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDemmer RT, Papapanou PN (2010) Epidemiologic patterns of chronic and aggressive periodontitis. Periodontol 2000 53:28\u0026ndash;44\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSocransky Ss, Haffajee Ad, Cugini Ma, Smith C, Kent Jr. (1998) R. L. Microbial complexes in subgingival plaque. J Clin Periodontol 25:134\u0026ndash;144\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBelstr\u0026oslash;m D et al (2017) Microbial profile comparisons of saliva, pooled and site-specific subgingival samples in periodontitis patients. PLoS ONE 12:e0182992\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDominy SS et al (2019) Porphyromonas gingivalis in Alzheimer\u0026rsquo;s disease brains: Evidence for disease causation and treatment with small-molecule inhibitors. Sci Adv 5:eaau3333\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLei S et al (2023) Porphyromonas gingivalis bacteremia increases the permeability of the blood-brain barrier via the Mfsd2a/Caveolin-1 mediated transcytosis pathway. Int J Oral Sci 15:1\u0026ndash;12\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHajishengallis G, Chavakis T (2021) Local and systemic mechanisms linking periodontal disease and inflammatory comorbidities. Nat Rev Immunol 21:426\u0026ndash;440\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang RP-H, Ho Y-S, Leung WK, Goto T, Chang R (2019) C.-C. Systemic inflammation linking chronic periodontitis to cognitive decline. Brain Behav Immun 81:63\u0026ndash;73\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSparks Stein P et al (2012) Serum antibodies to periodontal pathogens are a risk factor for Alzheimer\u0026rsquo;s disease. Alzheimers Dement 8:196\u0026ndash;203\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLaugisch O et al (2018) Periodontal Pathogens and Associated Intrathecal Antibodies in Early Stages of Alzheimer\u0026rsquo;s Disease. J Alzheimers Dis 66:105\u0026ndash;114\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eL\u0026rsquo;Heureux JE et al (2025) Oral microbiome and nitric oxide biomarkers in older people with mild cognitive impairment and APOE4 genotype. PNAS Nexus 4:pgae543\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTroci A et al (2024) Disease- and stage-specific alterations of the oral and fecal microbiota in Alzheimer\u0026rsquo;s disease. PNAS Nexus 3:pgad427\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNoble JM et al (2014) Serum IgG Antibody Levels to Periodontal Microbiota Are Associated with Incident Alzheimer Disease. PLoS ONE 9:e114959\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRubinstein T et al (2024) Periodontitis and brain magnetic resonance imaging markers of Alzheimer\u0026rsquo;s disease and cognitive aging. Alzheimer\u0026rsquo;s Dement 20:2191\u0026ndash;2208\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJagodzinski A, Koch-gromus U, Adam G, Anders S, Augustin M (2019) Rationale and Design of the Hamburg City Health Study. Eur J Epidemiol. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s10654-019-00577-4\u003c/span\u003e\u003cspan address=\"10.1007/s10654-019-00577-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEscapa IF et al (2018) New Insights into Human Nostril Microbiome from the Expanded Human Oral Microbiome Database (eHOMD): a Resource for the Microbiome of the Human Aerodigestive Tract. \u003cem\u003emSystems\u003c/em\u003e 3, e00187-18\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCurtis MA, Diaz PI, Van Dyke T (2020) E. The role of the microbiota in periodontal disease. \u003cem\u003ePeriodontol 2000\u003c/em\u003e 83, 14\u0026ndash;25\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSedghi LM, Bacino M, Kapila YL (2021) Periodontal Disease: The Good, The Bad, and The Unknown. Front Cell Infect Microbiol 11\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoriuchi A, Kokubu E, Warita T, Ishihara K (2020) Synergistic biofilm formation by \u003cem\u003eParvimonas micra\u003c/em\u003e and \u003cem\u003eFusobacterium nucleatum\u003c/em\u003e. Anaerobe 62:102100\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMacuch PJ, Tanner ACR (2000) Campylobacter Species in Health, Gingivitis, and Periodontitis. J Dent Res 79:785\u0026ndash;792\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHajishengallis G, Darveau RP, Curtis MA (2012) The Keystone Pathogen Hypothesis. Nat Rev Microbiol 10:717\u0026ndash;725\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAntezack A, Etchecopar-Etchart D, La Scola B, Monnet-Corti V (2023) New putative periodontopathogens and periodontal health-associated species: A systematic review and meta-analysis. J Periodontal Res 58:893\u0026ndash;906\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCostalonga M, Herzberg MC (2014) The oral microbiome and the immunobiology of periodontal disease and caries. Immunol Lett 162:22\u0026ndash;38\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi X, Liu Y, Yang X, Li C, Song Z (2022) The Oral Microbiota: Community Composition, Influencing Factors, Pathogenesis, and Interventions. Front Microbiol 13\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou P, Manoil D, Belibasakis GN, Kotsakis GA (2021) Veillonellae: Beyond Bridging Species in Oral Biofilm Ecology. Front Oral Health 2:774115\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHajishengallis G (2015) Periodontitis: from microbial immune subversion to systemic inflammation. Nat Rev Immunol 15:30\u0026ndash;44\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHajishengallis G (2014) Immunomicrobial pathogenesis of periodontitis: keystones, pathobionts, and host response. Trends Immunol 35:3\u0026ndash;11\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFogelholm N et al (2023) Subgingival microbiome at different levels of cognition. J Oral Microbiol 15:2178765\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSaid-Sadier N et al (2023) Association between Periodontal Disease and Cognitive Impairment in Adults. Int J Environ Res Public Health 20:4707\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCs D et al (2018) Tooth loss is associated with accelerated cognitive decline and volumetric brain differences: a population-based study. Neurobiol Aging 67\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang R-Q et al (2023) Poor Oral Health and Risk of Incident Dementia: A Prospective Cohort Study of 425,183 Participants. J Alzheimers Dis 93:977\u0026ndash;990\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMatsuyama Y et al (2022) Differences in Brain Volume by Tooth Loss and Cognitive Function in Older Japanese Adults. Am J Geriatr Psychiatry 30:1271\u0026ndash;1279\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSixin L, Stuart G, D., Rui Z (2023) Association Between Oral Bacteria and Alzheimer\u0026rsquo;s Disease: A Systematic Review and Meta-Analysis. J Alzheimer\u0026rsquo;s disease: JAD 91\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTang Z et al (2022) Treponema denticola Induces Alzheimer-Like Tau Hyperphosphorylation by Activating Hippocampal Neuroinflammation in Mice. J Dent Res 101:992\u0026ndash;1001\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKato Y et al (2024) Fusobacterium in oral bacterial flora relates with asymptomatic brain lesions. Heliyon 10\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBotelho J et al (2021) Periodontitis and circulating blood cell profiles: a systematic review and meta-analysis. Exp Hematol 93:1\u0026ndash;13\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKamer AR et al (2008) Inflammation and Alzheimer\u0026rsquo;s disease: Possible role of periodontal diseases. Alzheimer\u0026rsquo;s Dement 4:242\u0026ndash;250\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAdam HS et al (2022) The prospective association between periodontal disease and brain imaging outcomes: The Atherosclerosis Risk in Communities study. J Clin Periodontol 49:322\u0026ndash;334\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMayer C et al (2023) Association between periodontal disease and microstructural brain alterations in the Hamburg City Health Study. \u003cem\u003eJournal of Clinical Periodontology\u003c/em\u003e n/a\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMalan-M\u0026uuml;ller S et al (2024) Probing the oral-brain connection: oral microbiome patterns in a large community cohort with anxiety, depression, and trauma symptoms, and periodontal outcomes. Transl Psychiatry 14:1\u0026ndash;16\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSantonocito S et al (2022) A Cross-Talk between Diet and the Oral Microbiome: Balance of Nutrition on Inflammation and Immune System\u0026rsquo;s Response during Periodontitis. Nutrients 14:2426\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHebestreit K et al (2017) Validation of the German version of the Mediterranean Diet Adherence Screener (MEDAS) questionnaire. BMC Cancer 17:341\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMorris MC et al (2015) MIND diet slows cognitive decline with aging. Alzheimer\u0026rsquo;s Dement 11:1015\u0026ndash;1022\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFolsom AR, Parker ED, Harnack LJ (2007) Degree of Concordance With DASH Diet Guidelines and Incidence of Hypertension and Fatal Cardiovascular Disease*. Am J Hypertens 20:225\u0026ndash;232\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAltun E et al (2021) Association between Dietary Pattern and Periodontitis\u0026mdash;A Cross-Sectional Study. Nutrients 13:4167\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSeo Y et al (2024) Effect of MIND diet on cognitive function in elderly: a narrative review with emphasis on bioactive food ingredients. Food Sci Biotechnol 33:297\u0026ndash;306\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSantonocito S et al (2022) A Cross-Talk between Diet and the Oral Microbiome: Balance of Nutrition on Inflammation and Immune System\u0026rsquo;s Response during Periodontitis. Nutrients 14:2426\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWoelber JP et al (2019) The influence of an anti-inflammatory diet on gingivitis. A randomized controlled trial. J Clin Periodontol 46:481\u0026ndash;490\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJiang Y, Zhou X, Cheng L, Li M (2020) The Impact of Smoking on Subgingival Microflora: From Periodontal Health to Disease. Front Microbiol 11\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchwahn C et al (2021) Effect of periodontal treatment on preclinical Alzheimer\u0026rsquo;s disease\u0026mdash;Results of a trial emulation approach. Alzheimer\u0026rsquo;s Dement alz 12378. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1002/alz.12378\u003c/span\u003e\u003cspan address=\"10.1002/alz.12378\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDouglas GM et al (2020) PICRUSt2 for prediction of metagenome functions. Nat Biotechnol 38:685\u0026ndash;688\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiao T, Wei Y, Luo M, Zhao G-P, Zhou H (2019) tmap: an integrative framework based on topological data analysis for population-scale microbiome stratification and association studies. Genome Biol 20:293\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaryshnikova A (2016) Systematic Functional Annotation and Visualization of Biological Networks. Cell Syst 2:412\u0026ndash;421\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePetersen M et al (2020) Network Localisation of White Matter Damage in Cerebral Small Vessel Disease. Sci Rep 10:9210\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMuyzer G, de Waal EC, Uitterlinden AG (1993) Profiling of complex microbial populations by denaturing gradient gel electrophoresis analysis of polymerase chain reaction-amplified genes coding for 16S rRNA. Appl Environ Microbiol 59:695\u0026ndash;700\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCaporaso JG et al (2012) Ultra-high-throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. ISME J 6:1621\u0026ndash;1624\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEke PI, Page RC, Wei L, Thornton-Evans G, Genco RJ (2012) Update of the case definitions for population-based surveillance of periodontitis. J Periodontol 83:1449\u0026ndash;1454\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoltfreter B et al (2015) Standards for reporting chronic periodontitis prevalence and severity in epidemiologic studies: Proposed standards from the Joint EU/USA Periodontal Epidemiology Working Group. J Clin Periodontol 42:407\u0026ndash;412\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoms JC et al (1989) The Consortium to Establish a Registry for Alzheimer\u0026rsquo;s Disease (CERAD). Part I. Clinical and neuropsychological assesment of Alzheimer\u0026rsquo;s disease. Neurology 39:1159\u0026ndash;1159\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFawns-Ritchie C, Deary IJ (2020) Reliability and validity of the UK Biobank cognitive tests. PLoS ONE 15:e0231627\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKroenke K, Spitzer RL, Williams JBW (2001) The PHQ-9: Validity of a brief depression severity measure. J Gen Intern Med 16:606\u0026ndash;613\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKroenke K, Spitzer RL, Williams JBW (2002) The PHQ-15: Validity of a New Measure for Evaluating the Severity of Somatic Symptoms. Psychosom Med 64:258\u0026ndash;266\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSpitzer RL, Kroenke K, Williams JBW, L\u0026ouml;we B (2006) A Brief Measure for Assessing Generalized Anxiety Disorder: The GAD-7. Arch Intern Med 166:1092\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePetersen M et al (2023) Brain imaging and neuropsychological assessment of individuals recovered from a mild to moderate SARS-CoV-2 infection. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e 120, e2217232120\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFischl B, Dale AM (2000) Measuring the thickness of the human cerebral cortex from magnetic resonance images. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e 97, 11050\u0026ndash;11055\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFischl B et al (2002) Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron 33:341\u0026ndash;355\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eReveley C et al (2022) Diffusion MRI anisotropy in the cerebral cortex is determined by unmyelinated tissue features. Nat Commun 13:6702\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePasternak O, Sochen N, Gur Y, Intrator N, Assaf Y (2009) Free water elimination and mapping from diffusion MRI. Magn Reson Med 62:717\u0026ndash;730\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoeing H, Wahrendorf J, Becker N (1999) EPIC-Germany\u0026ndash;A source for studies into diet and risk of chronic diseases. European Investigation into Cancer and Nutrition. Ann Nutr Metab 43:195\u0026ndash;204\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSaggar M et al (2018) Towards a new approach to reveal dynamical organization of the brain using topological data analysis. Nat Commun 9:1399\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSaggar M, Shine JM, Li\u0026eacute;geois R, Dosenbach NUF, Fair D (2022) Precision dynamical mapping using topological data analysis reveals a hub-like transition state at rest. Nat Commun 13:4791\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNicolau M, Levine AJ, Carlsson G (2011) Topology based data analysis identifies a subgroup of breast cancers with a unique mutational profile and excellent survival. Proc Natl Acad Sci U S A 108:7265\u0026ndash;7270\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYao Y et al (2009) Topological methods for exploring low-density states in biomolecular folding pathways. J Chem Phys 130:144115\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRomano D et al (2014) Topological methods reveal high and low functioning neuro-phenotypes within fragile X syndrome. Hum Brain Mapp 35:4904\u0026ndash;4915\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSingh G et al (2008) Topological analysis of population activity in visual cortex. J Vis 8:11\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSingh G, Memoli F, Carlsson G (2007) Topological Methods for the Analysis of High Dimensional Data Sets and 3D Object Recognition. The Eurographics Association\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGloor GB, Macklaim JM, Pawlowsky-Glahn V, Egozcue JJ (2017) Microbiome Datasets Are Compositional: And This Is Not Optional. Front Microbiol 8\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcInnes L, Healy J, Astels S (2017) hdbscan: Hierarchical density based clustering. J Open Source Softw 2:205\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCostanzo M et al (2016) A global genetic interaction network maps a wiring diagram of cellular function. Science 353:aaf1420\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFruchterman TMJ, Reingold EM (1991) Graph drawing by force-directed placement. Software: Pract Experience 21:1129\u0026ndash;1164\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIkotun AM, Ezugwu AE, Abualigah L, Abuhaija B, Heming J (2023) K-means clustering algorithms: A comprehensive review, variants analysis, and advances in the era of big data. Inf Sci 622:178\u0026ndash;210\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlfaro-Almagro F et al (2021) Confound modelling in UK Biobank brain imaging. NeuroImage 224:117002\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePetersen M et al (2024) A latent clinical-anatomical dimension relating metabolic syndrome to brain structure and cognition. \u003cem\u003eeLife\u003c/em\u003e 12, RP93246\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSun Q, Li M (2025) Association between periodontitis and cognitive impairment in older adults: A cross-sectional study of the National Health and Nutrition Examination Survey. Clin Epidemiol Global Health 102020. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.cegh.2025.102020\u003c/span\u003e\u003cspan address=\"10.1016/j.cegh.2025.102020\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHow KY, Song KP, Chan KG (2016) Porphyromonas gingivalis: An Overview of Periodontopathic Pathogen below the Gum Line. Front Microbiol 7:53\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLetunic I, Bork P (2024) Interactive Tree of Life (iTOL) v6: recent updates to the phylogenetic tree display and annotation tool. Nucleic Acids Res 52:W78\u0026ndash;W82\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Oral microbiome, Periodontitis, Brain health, Cognition, Brain structure, Neuroimaging, Mental health, Inflammation, Nutrition","lastPublishedDoi":"10.21203/rs.3.rs-6580781/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6580781/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe relationship between oral microbiome composition and brain health in the general population remains poorly understood. In this study, we inferred a microbiome similarity network based on 16S rRNA sequencing data of crevicular fluid collected from 1,026 participants in the Hamburg City Health Study, which revealed a continuous disease gradient mirroring the microbial pathogenicity spectrum of periodontitis. Leveraging this network, we systematically examined associations between periodontal microbiome profiles and 37 brain health-related phenotypes, including cognitive function, brain structure, mental health, inflammatory biomarkers, diet, vascular risk factors, and demographics. Higher abundance of periodontitis-related microbial taxa was linked to poorer cognitive performance, elevated leukocyte counts and lower MIND diet adherence after covariate adjustment, but no significant associations were found for the remaining brain health phenotypes. Notably, we identified both previously known as well as novel microbial associations with brain health phenotypes. These findings advance the understanding of the oral microbiome-brain axis, highlighting potential pathways connecting periodontal health and brain function with potential implications for future causal and interventional studies.\u003c/p\u003e","manuscriptTitle":"Oral microbiome profiles relate periodontal disease and brain health - the PAROMIND Study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-14 09:38:42","doi":"10.21203/rs.3.rs-6580781/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"a6de9c0e-e5cd-410f-bbe0-d3a10f4c96ed","owner":[],"postedDate":"May 14th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":48388165,"name":"Biological sciences/Microbiology/Bacteria"},{"id":48388166,"name":"Health sciences/Neurology/Neurological disorders"}],"tags":[],"updatedAt":"2025-10-09T08:40:57+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-14 09:38:42","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6580781","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6580781","identity":"rs-6580781","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}