Probing the oral-brain connection: oral microbiome patterns in a large community cohort with anxiety, depression, and trauma symptoms, and periodontal outcomes

Abstract

The role of the oral microbiome in mental health has recently been appreciated within the proposed oral-brain axis. This study examined the structure and composition of the salivary microbiome in a large-scale population-based cohort of individuals reporting mental health symptoms (n = 306) compared to mentally healthy controls (n = 164) using 16S rRNA sequencing. Mental health symptoms were evaluated using validated questionnaires and included depression, anxiety, and posttraumatic stress disorder (PTSD), with accompanying periodontal outcomes. Participants also indicated current or previous diagnoses of anxiety, depression, periodontitis, and gingivitis. Mental and periodontal health variables influenced the overall composition of the oral microbiome. PTSD symptoms correlated with a lower clr-transformed relative abundance of Haemophilus sputorum and a higher clr-transformed relative abundance of Prevotella histicola. The clr-transformed relative abundance of P. histicola was also positively associated with depressive scores and negatively associated with psychological quality of life. Anxiety disorder diagnosis was associated with a lower clr-transformed relative abundance of Neisseria elongate and a higher clr-transformed relative abundance of Oribacterium asaccharolyticum. A higher clr-transformed relative abundance of Shuttleworthia and lower clr-transformed relative abundance of Capnocytophaga were evident in those who reported a clinical periodontitis diagnosis. Higher Eggerthia and lower Haemophilus parainfluenzae clr-transformed relative abundances were associated with reported clinical periodontitis diagnoses and psychotherapeutic efficacy. Functional prediction analysis revealed a potential role for tryptophan metabolism/degradation in the oral-brain axis, which was confirmed by lower plasma serotonin levels across symptomatic groups. This study sheds light on the intricate interplay between oral microbiota, periodontal and mental health outcomes, and a potential role for tryptophan metabolism in the proposed oral-brain axis, emphasizing the need for further exploration to pave the way for novel therapeutic interventions and predicting therapeutic response.

Introduction

Mental health disorders place a heavy burden on patients, families, societies, and global economies. In 2019, an estimated 418 million disability-adjusted life years (DALYs) could be attributable to mental disorders [1]. In 2017, depression was the leading cause of disability globally, with an estimated 322 million people living with depression [2], whilst about 260 million people suffered from anxiety disorders, and many suffered with additional comorbidities [3]. Mental disorders were the leading cause of the health-related burden of disease, and to worsen the situation, the COVID-19 pandemic has left in its wake a steep rise in the global prevalence of anxiety and depressive disorders [4].

Several factors, including economic insecurity, work-related stress, collective trauma, inequality, modern lifestyles, global events, and environmental factors, have likely contributed to the increased prevalence of mental health disorders [5]. Modern lifestyles, characterized by high stress levels, processed diets, excessive sanitation practices, and antibiotic use, alongside environmental changes like increased pollutants, climate change, and urbanization, have shifted human microbiota towards an industrialized state [6]. These microbiota alterations, coupled with the loss of specific functional attributes, may lead to suboptimal disease-promoting microbial communities, worsening compromised mental health [6].

The burden of mental health disorders is compounded by treatment limitations such as non-adherence, treatment resistance, and relapse [7–9], highlighting the necessity for innovative treatment modalities. The human holobiont, comprising the human host and its symbiotic microorganisms, plays a crucial role in health and disease [10–12]. While much attention has been given to the gut-brain axis, emerging evidence suggests that the oral microbiota, a less explored niche, may also influence the central nervous system (CNS) and behavior [13, 14].

The oral cavity hosts a diverse array of bacteria, and dysregulation can lead to disease. Periodontitis, a chronic bacterial infection affecting nearly half the global population [15], triggers systemic inflammation through pro-inflammatory cytokine release and invasion by periodontal keystone pathogens like Porphyromonas gingivalis (P. gingivalis) [16]. Periodontitis not only contributes to chronic inflammatory conditions like atherosclerosis, diabetes, and cardiovascular diseases [17, 18] but also shows associations with psychiatric disorders [19], suggesting involvement in the oral-brain axis. A longitudinal study spanning 10 years found a higher incidence of subsequent depression in individuals with periodontitis compared to those without [20], indicating a potential causal relationship between periodontitis and major depression. Additionally, recent research has identified specific bacterial taxa implicated in periodontal disease as well as anxiety, depressive disorders, and trauma-related disorders [21].

Pathogenic periodontal bacteria can impact the CNS through various pathways, both directly and indirectly [21]. Direct routes include bloodstream transmission or areas with compromised blood-brain barrier (BBB) integrity [22]. Indirectly, they induce pro-inflammatory cytokine production, activating endothelial cells expressing tumor necrosis factor (TNF)-α and interleukin-1 (IL-1) β receptors, which signal perivascular macrophages, leading to neuroinflammation [23]. Keystone pathogens widen intercellular spaces in periodontal pockets, causing epithelial rupture and a “leaky mouth” [24], facilitating lipopolysaccharide (LPS) access to circulation, activating the immune system and the hypothalamic pituitary adrenal (HPA) axis, influencing CNS function [23, 25]. Other entry points include circumventricular organs, the choroid plexus [22, 26], and olfactory/trigeminal nerves [27]. Brain-resident microglia can be influenced by periodontal bacteria via leptomeninges [25]. Periodontal pathogens also affect gut microbial composition/function directly via enteral or indirectly via hematogenous transmission [28].

Clinical data linking the oral microbiome to mental health disorders are limited. A study employing genetic association analysis and Mendelian randomization found significant associations and causal effects between salivary-tongue dorsum microbiome interactions and anxiety/depression, with Eggerthia notably linked to both conditions across multiple databases [29]. A clinical case study reported the pathogenic potential of Eggerthia catenaformis, which started as a submandibular abscess that spread hematogenously, subsequently causing a perihepatic abscess and severe clinical sepsis [30]. Another study involving adolescents (n = 66) observed differing abundances of Actinomyces, Spirochaetaceae, Fusobacterium, and Treponema in individuals with symptoms of anxiety and depression [31]. Wingfield et al. compared the oral microbial composition in depressed young adults (n = 40), identifying 21 bacterial taxa with varying levels compared to controls, including higher abundances of Neisseria spp. and Prevotella nigrescens [32]. P. nigrescens show moderate pathogenicity [33] and can produce mannose polysaccharides, which promote chronic inflammatory processes, including altered leukocyte phagocytosis and invasion of host barriers [34, 35].

Although limited mechanistic data is currently available to explain the links between these oral bacteria and mental health outcomes, an oral-brain axis has been proposed, where periodontal bacteria can directly reach and influence the brain via several pathways. Periodontitis can indirectly impact the CNS through pro-inflammatory cytokines [36], which activate endothelial cells expressing TNF-a and IL-1 receptors. This activation signals perivascular macrophages, which in turn activate microglia, leading to neuroinflammation [37, 38]. Additionally, periodontitis can cause a leaky periodontium, allowing LPS to enter the systemic circulation. This can activate the hypothalamic-pituitary-adrenal (HPA) axis, resulting in elevated stress hormones or neurotransmitters [39], thereby influencing mental health outcomes. Larger studies encompassing diverse age groups and gathering microbiome and mechanistic data longitudinally are needed to elucidate the role of the oral-brain axis in anxiety and depression.

This study aimed to enhance the limited understanding of how periodontal health can impact mental health through the oral microbiome. We characterized the salivary microbiome, predicted the functional potential of salivary taxa, and measured plasma levels of tryptophan metabolites to confirm these predictions in individuals exhibiting symptoms of anxiety, depression, and PTSD along with periodontal outcomes. These findings could lay the foundation for targeting the oral microbiome and improving periodontal health as a novel approach to enhancing mental health outcomes.

Methods

Study participant evaluation and enrollment

This cohort comprised two Spanish study populations from PsicoBioma (n = 186, March 2021 - Jan 2022) and TRIAD (n = 284, Nov 2021 - Dec 2022), both population-based microbiome projects. PsicoBioma recruited participants from Spain, while TRIAD recruited from Madrid, Barcelona, Vitoria, and Oviedo municipalities (allowing blood collection). Both cohorts provided saliva samples and completed similar online questionnaires. TRIAD also provided blood samples for plasma analysis and completed a more comprehensive periodontal health questionnaire. The research adhered to The Code of Ethics of the World Medical Association (Declaration of Helsinki) for human experiments, and data processing followed Spanish Organic Law 3/2018 on Personal Data Protection and Digital Rights Guarantee (BOE 16673 of 6 Dec 2018) and its 17th Additional Provision. Approval was obtained from the Ethics Committees of Hospital Clínico San Carlos (Madrid), Medical Research Ethics Committee of Asturias, Basque Medicine Research Ethics Committee, and Drug Research Ethics Committee of Hospital de la Santa Creu i Sant Pau (PSQ-19-2 C.I. 196/474-E). All research participants provided online, written informed consent.

The study recruited healthy controls, participants with a current/previous diagnosis of anxiety, depressive, or trauma-related disorders, or individuals who were experiencing these symptoms. Spanish residents, 18 years or older, who were proficient in reading and understanding Spanish were included. Individuals who used antibiotics within the previous six months, and those diagnosed with any other major psychiatric disorders including psychotic disorders, personality disorders, or neurodegenerative disorders, were excluded.

Demographic and clinical data

Demographic, health, and clinical data were collected using a secure online questionnaire. Psychological evaluations relied on standardized self-report questionnaires validated for the Spanish population; this study focused on symptoms rather than formal diagnoses. However, participants also indicated on the questionnaire whether they had a previous/current clinical diagnosis of anxiety or depression (“diagnosis” henceforth refers to clinical diagnoses, and “symptoms” refers to self-report questionnaire data). Depressive symptoms were assessed using the Center for Epidemiologic Studies Depression (CESD) scale; state and trait anxiety symptoms were evaluated using the state-trait anxiety inventory (STAI). The posttraumatic stress disorder (PTSD) Checklist for the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5, PCL-5), and the Childhood Trauma Questionnaire-Short Form (CTQ-SF) evaluated trauma exposure. Additionally, quality of life was measured using the World Health Organization Quality Of Life Questionnaire (WHOQOL).

Psychiatric symptoms were determined based on the following criteria

Depressive symptoms: CESD scores of 16 – 24 indicated mild and 25 – 55 severe depressive symptoms [40]. PTSD symptoms: PCL-5 score > 33 + more than 3 symptom clusters [41]. State anxiety symptoms: STAI-S scores > 41; trait anxiety symptoms: STAI-T scores > 45; CTQ-SF [42] total score was used to evaluate the severity of childhood maltreatment.

The TRIAD cohort completed a periodontal health questionnaire [43] (validated for the Spanish population [44]) to predict severe periodontitis, according to Montero et al. [44] Specifically, severe periodontitis was defined according to three criteria: (i) the Centers for Disease Control/American Academy of Periodontology (CDC/AAP) case definition [45], henceforth referred to as SeverePerioCDCAAP; (ii) the presence of ≥ 50% of teeth with clinical attachment level (CAL) ≥ 5 mm, henceforth referred to as TeethCAL5; (iii) the presence of ≥ 25% of teeth with probing pocket depth (PPD) ≥ 6 mm, henceforth referred to as TeethPPD6. In addition, participants also indicated whether they had a previous/current clinical diagnosis of periodontitis and/or gingivitis (“diagnosis” refers to a self-reported clinical diagnosis, and “predicted severe periodontitis” refers to self-reported questionnaire data).

Blood collection and processing

Whole blood (10 ml) was collected from 282 of the TRIAD participants, using BD Vacutainer® EDTA tubes. Blood was centrifuged at 1800 rpm for 10 minutes at room temperature, and the resulting supernatant (plasma) was transferred into clean 1.5 ml Eppendorf tubes for storage at −80 °C for later use.

Kynurenine, tryptophan, and serotonin quantification in plasma

Plasma levels of kynurenine (KYN), tryptophan (TRP), and serotonin (5-HT) were measured using high-performance liquid chromatography (HPLC). TRP and 5-HT were detected fluorometrically at excitation/emission wavelengths of 270/360 nm and 290/398 nm, respectively (Waters 2475, Multi fluorescence Detector; Waters, Milford, MA, USA). The ratios of KYN or 5-HT to TRP concentrations were calculated and used as a measure of TRP degradation (see Supplementary Materials for details).

Bacterial DNA extraction, 16S rRNA gene sequencing and analysis

Participants self-collected saliva samples in DNA/RNA Shield Safe Collect Saliva Collection tubes (Zymo Research, Irvine, California, USA), from which microbial DNA was extracted (ZymoBIOMICS DNA Miniprep Kit, Zymo Research). Bacterial 16S rRNA gene V3-4 amplicons were generated, using previously described primers [46] and sequenced (2 × 300 bp paired-end) (Center de Regulació Genòmica, Barcelona, Spain), on the Illumina NextSeq2000 platform (see Supplementary Material for details).

Quality control of FASTQ sequencing files was performed using fastqc and multiqc. Raw sequence reads were de-replicated and de-noised to combine identical reads into amplicon sequence variants (ASVs) [47], and construct consensus quality profiles for each combined set of sequences (dada2 version 3.11 [48]). Following chimera removal, a consensus paired-end reads file was generated for feature construction and downstream analysis. Taxonomic binning of classified sequences was built using a local copy of the Ribosomal Database Project (RDP) Classifier (Train Set 19 [49]), and normalized data were produced from the relative abundance of taxa present in each sample. A feature table of 54,817 unique ASVs with an average read length of 391 nucleotides in 470 samples was consequently constructed (following pre-processing, the minimum number of reads per sample was 18,868, and the average number of reads per sample was 95,763).

Statistical analyses

Data was analyzed using bioinformatics and statistical analysis packages in R [47], including the packages dada2 (version 3.18 [48]), vegan (version 2.6.4 [50]), phyloseq (version 1.46.0 [51]), ggplot2 (version 3.4.4 [52]), and CoDaSeq (version 0.99.7 [53, 54]. For clinical and demographic data, continuous variables were summarized as means (M) and standard deviations (SD) if normally distributed or as medians and interquartile ranges (IQRs) if non-normally distributed. To assess differences in the metadata variables between symptomatic and control groups, Student’s t-tests and Wilcoxon rank sum tests were used to assess differences between normally and non-normally distributed data (normality tested using Shapiro-Wilk Normality Test), respectively. Categorical data were summarized as counts (n) and percentages, and χ² or Fisher exact tests were used to assess differences between groups, where appropriate. Significance was defined as p ≤ 0.05.

The Simpson index (which takes both richness and evenness into account), Chao1 (estimator of species richness), and Pielou’s evenness index were used to evaluate α-diversity [55]. Taxa were agglomerated to genus level, assigning species-level where possible. Data was transformed to relative abundance out of 100 to account for differences in total depth per sample. Variance filtering was performed (genefilter function, version 1.84.0), which removed taxa with the lowest 40% variance. Abundance matrices were centered log-ratio (clr)-transformed, using the minimum proportional abundance detected for each taxon for the imputation of zeros. The ordination of community variation was visualized using multidimensional scaling (MDS) of genus-level Aitchison distances. The capscale function (vegan package) [50] was used to determine the contribution of metadata variables to microbiome community variation.

The ASV table was filtered to retain taxa observed in at least 15% of participants. Associations between taxonomic abundance and metadata variables were analyzed using a linear modeling approach (fw_glm function, Tjazi package) [56]) on the clr-transformed and filtered relative taxonomic abundances, whilst adjusting for covariates that were associated with taxonomic abundance in our dataset, including age, body mass index (BMI), smoking status, and cholesterol medication use. Other potential covariates were tested using linear modeling, but did not have an association with taxonomic abundance, these included sex, alcohol use, psychotropic medication use, weekly consumption of processed food, green vegetables, and fruits, and whether filtered or unfiltered water was consumed. We performed false discovery rate (FDR) correction using the Benjamini-Hochberg procedure and significance was defined as q ≤ 0.1. Throughout the manuscript, we refer to relative abundance, which denotes filtered, clr-transformed relative taxonomic abundances for brevity. Medians (mdn) and interquartile ranges (IQRs) are also reported.

We utilized PICRUSt2 [57] to predict the CNS-related functional potential of oral taxa, by focusing on gut-brain modules (GBMs) [58]. Associations between GBMs and mental health outcomes as well as periodontal outcomes were tested using the same linear modeling approach as previously described, with significance set at q ≤ 0.1.

Post hoc power analysis

Following the completion of the study, a post hoc power analysis was conducted to evaluate the statistical power achieved with the final sample size. The analysis was performed using the Power & Sample Sizes Tool for Case-Control Microbiome Studies [59], which incorporates Monte Carlo simulations with 100 replications and an alpha level of 0.1. Firstly, we computed the power across a range of sample sizes to understand how varying the sample size influences the power of the study. Secondly, we specified the number of controls as n = 164 (mentally healthy controls) and the number of cases as n = 306 (the symptomatic cohort). The outcomes of the simulation are reported in terms of the Wald statistic, the Wilcoxon–Mann–Whitney (WMW) test, and the average Wilcoxon–Mann–Whitney (WMW avg) test.

Results

Clinical and demographic characteristics

Clinical and demographic characteristics of the depressive (n = 148), state (n = 256) and trait anxiety (n = 281), and PTSD (n = 73) symptomatic cohorts (according to criteria described in the Methods and Materials), and healthy controls (no significant mental health symptoms) (n = 164) are described in Tables 1–4. In our cohort of 470 individuals, 306 presented with at least one or a combination of the aforementioned symptoms. Females constituted 72% of our cohort, and the median age was 40 years. Comorbidity of these psychiatric symptoms was common; of the 232 individuals who had both state and trait anxiety symptoms, 133 of them also had depressive symptoms (57.3%), 67 had comorbid PTSD symptoms (28.9%), and 52 (22.4%) had PTSD and depressive symptoms. Of the 148 individuals with depressive symptoms, 144 (97.3%) also had trait anxiety symptoms, and 54 (36.5%) had symptoms of PTSD.

Table 2.

Comparative statistics of continuous and categorical variables for the trait anxiety symptom vs. healthy control cohorts.

	Total cohort (n = 470)		Healthy controls (n = 164)		Trait anxiety symptoms cohort (n = 281)		p-value
	median	IQR	median	IQR	median	IQR
	n	%	n	%	n	%
Age	40.0	31.0–50.0	38.5	30.0–47.5	40.0	31.0–49.0	NS
BMI	22.8	20.7–25.4	22.8	20.9–25.4	22.8	20.4–25.4	NS
CTQ total score	34.0	29.0–44.0	29.0	27.0–34.0	39.0	31.0–49.0	<0.001
WHOQOL Physical health	14.3	12.0–16.6	16.6	14.9–17.9	13.1	10.9–14.9	<0.001
WHOQOL Psychological health	13.0	10.0–15.3	16.0	14.0–17.3	11.3	9.3–13.3	<0.001
WHOQOL Social relationships	12.7	9.3–14.7	14.7	13.3–17.3	10.7	8.0–13.3	<0.001
WHOQOL Environment	15.0	12.5–16.5	16.5	15.0–18.0	13.5	11.5–15.5	<0.001
WHOQOL Total score	7.0	5.0–8.0	8.0	7.0–9.0	6.0	4.0–7.0	<0.001
PCL Total score	17.5	5.0–35.0	5.0	1.0–11.5	26.0	13.0–41.8	<0.001
CESD total score	16.0	8.0–30.0	6.0	2.0–10.0	25.0	17.0–35.0	<0.001
STAIS total score	44.0	35.0–54.0	32.0	27.8–36.0	52.0	44.0–58.0	<0.001
STAIT total score	49.0	40.0–58.0	37.0	33.0–41.0	56.0	51.0–61.0	<0.001
LPS concentration	0.2	0.1–0.2	0.2	0.1–0.2	0.2	0.1–0.2	NS
KYN (log)	0.3	0.2–0.5	0.4	0.2–0.5	0.3	0.1–0.5	NS
5HT (log)	−0.7	−2.2–0.2	−0.5	−1.1–0.0	0.4	0.1–0.8	<0.005
TRP (log)	3.8	3.6–3.9	3.8	3.6–3.9	3.8	3.6–3.9	NS
KYN/TRP (log)	3.5	3.3–3.7	3.5	3.3–3.8	3.4	3.3–3.7	NS
5HT/TRP (log)	2.5	1.0–3.0	2.6	2.1–3.2	2.4	0.7–2.9	<0.05
Female	341	73%	110	67%	213	76%	<0.05
Gingivitis diagnosisa	30	11%	6	9%	24	12%	NS
Periodontitis diagnosis	34	7%	11	7%	22	8%	NS
IBS/Coeliac/Crohn's diagnosis	78	17%	18	11%	55	20%	<0.05
Psychotherapeutic Response
Unresponsive	29	9%	3	4%	26	12%	<0.001
Responsive	195	61%	65	77%	119	53%
Somewhat responsive	98	30%	16	19%	78	35%
Psychoactive medicationa	117	44%	12	21%	100	52%	<0.001
Smoking status (yes)	120	26%	32	20%	79	28%	0.050
PTSD symptoms	73	16%	0	0%	72	26%	<0.001
State anxiety symptoms	256	54%	0	0%	232	83%	<0.001
Trait anxiety symptoms	281	60%	0	0%	281	100%	<0.001
Predicted periodontitis
TeethPPD6a	81	41%	21	49%	56	39%	NS
SeverePerioCDCAAPa	106	54%	27	63%	75	52%	NS
TeethCAL5a	93	47%	24	56%	65	45%	NS

TeethPPD6, SeverePerioCDCAAP, TeethCAL5 data available for n = 196 of the total cohort, n = 43 of HC, and n = 143 of the trait anxiety symptoms cohort. Gingivitis diagnosis data available for n = 284 of the total cohort, n = 66 of HC, and n = 204 of the trait anxiety symptoms cohort.

5HT serotonin, BMI body mass index, CESD Center for Epidemiologic Studies Depression scale, CTQ childhood trauma questionnaire, HC healthy controls, IBS Irritable bowel syndrome, IQR interquartile range, KYN kynurenine, LPS lipopolysaccharide, PCL-5 Posttraumatic Stress Disorder Checklist for DSM-5, PTSD posttraumatic stress disorder, SeverePerioCDCAAP Centers for Disease Control/American Academy of Periodontology (CDC/AAP) case definition, sd standard deviation, STAI state-trait anxiety inventory, TeetchCAL5 probable severe periodontitis based on the “≥50% of teeth with clinical attachment level (CAL) ≥5 mm” criteria, TeethPPD6 probable severe periodontitis based on the “≥25% of teeth with probing pocket depth (PPD) ≥6 mm” criteria, TRP tryptophan, WHOQOL World Health Organization Quality of Life assessment.

^aPsychoactive medication data available for n = 256 of the total cohort, n = 58 of HC, and n = 194 of the trait anxiety symptoms cohort. Bold values indicate statistical significance.

Table 3.

Comparative statistics of continuous and categorical variables for the state anxiety symptom vs. healthy control cohorts.

	Total cohort (n = 470)		Healthy controls (n = 164)		State anxiety symptoms cohort (n = 256)		p-value
	median	IQR	median	IQR	median	IQR
	n	%	n	%	n	%
Age	40.0	31.0–50.0	38.5	30.0–47.5	41.0	32.0–50.0	NS
BMI	22.8	20.7–25.4	22.8	20.9–25.4	22.9	20.7–25.2	NS
CTQ total score	34.0	29.0–44.0	29.0	27.0–34.0	39.0	31.0–51.0	<0.001
WHOQOL Physical health	14.3	12.0–16.6	16.6	14.9–17.9	12.6	10.9–14.3	<0.001
WHOQOL Psychological health	13.0	10.0–15.3	16.0	14.0–17.3	10.7	8.7–12.7	<0.001
WHOQOL Social relationships	12.7	9.3–14.7	14.7	13.3–17.3	10.7	8.0–13.3	<0.001
WHOQOL Environment	15.0	12.5–16.5	16.5	15.0–18.0	13.5	11.5–15.5	<0.001
WHOQOL Total score	7.0	5.0–8.0	8.0	7.0–9.0	6.0	4.0–7.0	<0.001
PCL Total score	17.5	5.0–35.0	5.0	1.0–11.5	28.0	14.0–42.0	<0.001
CESD total score	16.0	8.0–30.0	6.0	2.0–10.0	26.0	17.0–35.0	<0.001
STAIS total score	44.0	35.0–54.0	32.0	27.8–36.0	53.0	47.0–59.0	<0.001
STAIT total score	49.0	40.0–58.0	37.0	33.0–41.0	56.0	51.0–61.0	<0.001
LPS concentration	0.2	0.1–0.2	0.2	0.1–0.2	0.2	0.1–0.2	NS
KYN (log)	0.3	0.2–0.5	0.4	0.2–0.5	0.4	0.2–0.5	NS
5HT (log)	−0.7	−2.2–0.2	−0.5	−1.1–0.0	-0.5	−1.1–0.0	<0.005
TRP (log)	3.8	3.6–3.9	3.8	3.6–3.9	3.8	3.6–3.9	NS
KYN/TRP (log)	3.5	3.3–3.7	3.5	3.3–3.8	3.5	3.3–3.8	NS
5HT/TRP (log)	2.5	1.0–3.0	2.6	2.1–3.2	2.6	2.1–3.2	<0.05
Female	341	73%	110	67%	192	75%	NS
Gingivitis diagnosisa	30	11%	6	9%	21	11%	NS
Periodontitis diagnosis	34	7%	11	7%	18	7%	NS
IBS/Coeliac/Crohn's diagnosis	78	17%	18	11%	55	21%	<0.05
Psychotherapeutic Response
Unresponsive	29	9%	3	4%	25	12%	<0.001
Responsive	195	61%	65	77%	105	52%
Somewhat responsive	98	30%	16	19%	72	36%
Psychoactive medicationa	117	44%	12	21%	87	50%	<0.001
Smoking status (yes)	120	26%	32	20%	76	30%	<0.05
PTSD symptoms	73	16%	0	0%	68	27%	<0.001
State anxiety symptoms	256	54%	0	0%	256	100%	<0.001
Trait anxiety symptoms	281	60%	0	0%	232	91%	<0.001
Predicted periodontitis
TeethPPD6a	81	41%	21	49%	55	41%	NS
SeverePerioCDCAAPa	106	54%	27	63%	70	53%	NS
TeethCAL5a	93	47%	24	56%	63	47%	NS

TeethPPD6, SeverePerioCDCAAP, TeethCAL5 data available for n = 196 of the total cohort, n = 43 of HC, and n = 133 for state anxiety symptoms cohort; Gingivitis diagnosis data available for n = 284 of the total cohort, n = 66 of HC, and n = 186 of the state anxiety symptoms cohort.

5HT serotonin, BMI body mass index, CESD Center for Epidemiologic Studies Depression scale, CTQ childhood trauma questionnaire, HC healthy controls, IBS Irritable bowel syndrome, IQR interquartile range, KYN kynurenine, LPS lipopolysaccharide, PCL-5 Posttraumatic Stress Disorder Checklist for DSM-5, PTSD posttraumatic stress disorder, SeverePerioCDCAAP Centers for Disease Control/American Academy of Periodontology (CDC/AAP) case definition, sd standard deviation, STAI state-trait anxiety inventory, TeetchCAL5 probable severe periodontitis based on the “≥50% of teeth with clinical attachment level (CAL) ≥5 mm” criteria, TeethPPD6 probable severe periodontitis based on the “≥25% of teeth with probing pocket depth (PPD) ≥6 mm” criteria, TRP tryptophan, WHOQOL World Health Organization Quality of Life assessment.

^aPsychoactive medicationdata available for n = 256 of the total cohort, n = 58 of HC, and n = 175 of state anxiety symptoms cohort. Bold values indicate statistical significance.

Table 1.

Comparative statistics of continuous and categorical variables for the depressive symptom vs. healthy control cohorts.

	Total cohort (n = 470)		Healthy controls (n = 164)		Depressive symptom cohort (n = 148)		p-value
	median	IQR	median	IQR	median	IQR
	n	%	n	%	n	%
Age	40.0	31.0–50.0	38.5	30.0–47.5	40.0	33.0–51.0	NS
BMI	22.8	20.7–25.4	22.8	20.9–25.4	22.9	20.4–26.0	NS
CTQ total score	34.0	29.0–44.0	29.0	27.0–34.0	43.5	34.0–55.3	<0.001
WHOQOL Physical health	14.3	12.0–16.6	16.6	14.9–17.9	12.0	9.7–13.1	<0.001
WHOQOL Psychological health	13.0	10.0–15.3	16.0	14.0–17.3	9.3	8.0–11.3	<0.001
WHOQOL Social relationships	12.7	9.3–14.7	14.7	13.3–17.3	9.3	6.7–12.0	<0.001
WHOQOL Environment	15.0	12.5–16.5	16.5	15.0–18.0	13.0	11.0–14.6	<0.001
WHOQOL Total score	7.0	5.0–8.0	8.0	7.0–9.0	5.0	4.0–6.0	<0.001
PCL Total score	17.5	5.0–35.0	5.0	1.0–11.5	35.0	21.0–47.0	<0.001
CESD total score	16.0	8.0–30.0	6.0	2.0–10.0	35.0	31.0–41.0	<0.001
STAIS total score	44.0	35.0–54.0	32.0	27.8–36.0	57.0	50.0–62.0	<0.001
STAIT total score	49.0	40.0–58.0	37.0	33.0–41.0	59.0	55.0–64.0	<0.001
LPS concentration	0.2	0.1–0.2	0.2	0.1–0.2	0.2	0.1–0.3	NS
KYN (log)	0.3	0.2–0.5	0.4	0.2–0.5	0.3	0.1–0.5	NS
5HT (log)	−0.7	−2.2–0.2	−0.5	−1.1–0.0	−1.3	−2.8–0.4	<0.001
TRP (log)	3.8	3.6–3.9	3.8	3.6–3.9	3.8	3.6–3.9	NS
KYN/TRP (log)	3.5	3.3–3.7	3.5	3.3–3.8	3.4	3.3–3.6	NS
5HT/TRP (log)	2.5	1.0–3.0	2.6	2.1–3.2	1.9	0.3–2.7	<0.001
Female	341	73%	110	67%	113	76%	NS
Gingivitis diagnosisa	30	11%	6	9%	12	10%	NS
Periodontitis diagnosis	34	7%	11	7%	10	7%	NS
IBS/Coeliac/Crohn's diagnosis	78	17%	18	11%	29	20%	<0.05
Psychotherapeutic Response
Unresponsive	29	9%	3	4%	19	15%	<0.001
Responsive	195	61%	65	77%	65	51%
Somewhat responsive	98	30%	16	19%	44	34%
Psychoactive medicationa	117	44%	12	21%	68	61%	<0.001
Smoking status (yes)	120	26%	32	20%	45	30%	<0.05
PTSD symptoms	73	16%	0	0%	54	36%	<0.001
State anxiety symptoms	256	54%	0	0%	136	92%	<0.001
Trait anxiety symptoms	281	60%	0	0%	144	97%	<0.001
Predicted periodontitis
TeethPPD6a	81	41%	21	49%	29	36%	NS
SeverePerioCDCAAPa	106	54%	27	63%	37	46%	NS
TeethCAL5a	93	47%	24	56%	35	43%	NS

TeethPPD6, SeverePerioCDCAAP, TeethCAL5 data available for n = 196 of the total cohort, n = 43 of HC, and n = 81 of the depressive symptom cohort; Gingivitis diagnosis data available for n = 284 of the total cohort, n = 66 of HC, and n = 117 of the depressive symptom cohort.

5HT serotonin, BMI body mass index, CESD Center for Epidemiologic Studies Depression scale, CTQ childhood trauma questionnaire, HC healthy controls, IBS Irritable bowel syndrome, IQR interquartile range, KYN kynurenine, LPS lipopolysaccharide, PCL-5 Posttraumatic Stress Disorder Checklist for DSM-5, PTSD posttraumatic stress disorder, SeverePerioCDCAAP Centers for Disease Control/American Academy of Periodontology (CDC/AAP) case definition, sd standard deviation, STAI state-trait anxiety inventory, TeetchCAL5 probable severe periodontitis based on the “≥50% of teeth with clinical attachment level (CAL) ≥5 mm” criteria, TeethPPD6 probable severe periodontitis based on the “≥25% of teeth with probing pocket depth (PPD) ≥6 mm” criteria, TRP tryptophan, WHOQOL World Health Organization Quality of Life assessment.

^aPsychoactive medication data available for n = 256 of the total cohort, n = 58 of the HC, and n = 111 of the depressive symptom cohort. Bold values indicate statistical significance.

Table 4.

Comparative statistics of continuous and categorical variables for the PTSD symptom vs. healthy control cohorts.

	Total cohort (n = 470)		Healthy controls (n = 164)		PTSD symptoms cohort (n = 73)		p-value
	median	IQR	median	IQR	median	IQR
					mean	± sd
	n	%	n	%	n	%
Age	40.0	31.0–50.0	38.5	30.0–47.5	40.6	± 12.5	NS
BMI	22.8	20.7–25.4	22.8	20.9–25.4	21.7	19.7–24.7	<0.05
CTQ total score	34.0	29.0–44.0	29.0	27.0–34.0	45.0	34.0–59.0	<0.001
WHOQOL Physical health	14.3	12.0–16.6	16.6	14.9–17.9	11.9	± 2.9	<0.001
WHOQOL Psychological health	13.0	10.0–15.3	16.0	14.0–17.3	10.2	± 2.9	<0.001
WHOQOL Social relationships	12.7	9.3–14.7	14.7	13.3–17.3	10.7	8.0–13.3	<0.001
WHOQOL Environment	15.0	12.5–16.5	16.5	15.0–18.0	13.0	11.0–15.5	<0.001
WHOQOL Total score	7.0	5.0–8.0	8.0	7.0–9.0	6.0	4.0–6.0	<0.001
PCL Total score	17.5	5.0–35.0	5.0	1.0–11.5	46.0	38.0–54.0	<0.001
CESD total score	16.0	8.0–30.0	6.0	2.0–10.0	32.0	± 2.9	<0.001
STAIS total score	44.0	35.0–54.0	32.0	27.8–36.0	56.0	± 2.9	<0.001
STAIT total score	49.0	40.0–58.0	37.0	33.0–41.0	59.0	± 2.9	<0.001
LPS concentration	0.2	0.1–0.2	0.2	0.1–0.2	0.2	0.1–0.2	NS
KYN (log)	0.3	0.2–0.5	0.4	0.2–0.5	0.3	0.1–0.4	NS
5HT (log)	−0.7	−2.2–0.2	−0.5	−1.1–0.0	−2.0	−2.9–0.5	<0.001
TRP (log)	3.8	3.6–3.9	3.8	3.6–3.9	3.7	3.6–3.8	NS
KYN/TRP (log)	3.5	3.3–3.7	3.5	3.3–3.8	3.5	3.3–3.6	NS
5HT/TRP (log)	2.5	1.0–3.0	2.6	2.1–3.2	1.3	0.2–2.8	<0.01
Female	341	73%	110	67%	63	86%	<0.01
Gingivitis diagnosisa	30	11%	6	9%	9	16%	NS
Periodontitis diagnosis	34	7%	11	7%	6	8%	NS
IBS/Coeliac/Crohn's diagnosis	78	17%	18	11%	15	21%	NS
Psychotherapeutic Response
Unresponsive	29	9%	3	4%	5	8%	<0.05
Responsive	195	61%	65	77%	35	54%
Somewhat responsive	98	30%	16	19%	25	38%
Psychoactive medicationa	117	44%	12	21%	34	62%	<0.001
Smoking status (yes)	120	26%	32	20%	22	30%	NS
PTSD symptoms	73	16%	0	0%	73	100%	<0.001
State anxiety symptoms	256	54%	0	0%	68	93%	<0.001
Trait anxiety symptoms	281	60%	0	0%	72	99%	<0.001
Predicted periodontitis
TeethPPD6a	81	41%	21	49%	14	36%	NS
SeverePerioCDCAAPa	106	54%	27	63%	21	54%	NS
TeethCAL5a	93	47%	24	56%	19	49%	NS

TeethPPD6 SeverePerioCDCAAP, TeethCAL5 data available for n = 196 of the total cohort, n = 43 of HC, and, n = 39 of the PTSD symptoms cohort; Gingivitis diagnosis data available for n = 284 of the total cohort, n = 66 of HC, and n = 57 of the PTSD cohort

5HT serotonin, BMI body mass index, CESD Center for Epidemiologic Studies Depression scale, CTQ childhood trauma questionnaire, HC healthy controls, IBS Irritable bowel syndrome, IQR interquartile range, KYN kynurenine, LPS lipopolysaccharide, PCL-5 Posttraumatic Stress Disorder Checklist for DSM-5, PTSD posttraumatic stress disorder, SeverePerioCDCAAP Centers for Disease Control/American Academy of Periodontology (CDC/AAP) case definition, sd standard deviation, STAI state-trait anxiety inventory, TeetchCAL5 probable severe periodontitis based on the “≥50% of teeth with clinical attachment level (CAL) ≥5 mm” criteria, TeethPPD6 probable severe periodontitis based on the “≥25% of teeth with probing pocket depth (PPD) ≥6 mm” criteria, TRP tryptophan, WHOQOL World Health Organization Quality of Life assessment.

^aPsychoactive medication data available for n = 256 of the total cohort, n = 58 of the HC, and n = 55 of the PTSD symptoms cohort. Bold values indicate statistical significance.

A total of 34 (7%) individuals had a current clinical diagnosis of periodontitis, and 30 (11%) had a gingivitis diagnosis. The self-reported periodontal health questionnaire administered to most of the TRIAD cohort (n = 196) (unavailable for the PsicoBioma cohort) showed 81 individuals (41%) had probable severe periodontitis using TeethPPD6 criteria, 93 (47%) had it based on the TeethCAL5 criteria, and 106 (54%) had it based on the Centers for Disease Control (CDC)/American Academy of Periodontology (AAP) criteria (SeverePerioCDCAAP).

Post hoc power analysis

The power analysis conducted over a range of sample sizes (30 to 200 participants) indicated that a sample size of 200 participants would achieve 92% power to detect differences. Secondly, using the specified sample sizes of 164 controls and 306 cases, the simulation yielded the following outcomes: Wald Statistic = 0.98, Wilcoxon–Mann–Whitney (WMW) Test = 1, Average Wilcoxon-Mann-Whitney (WMW avg) Test = 0.798. Indicating that that the power to detect differences in microbiome composition between the case and control groups is 98%. The Wilcoxon-Mann-Whitney test value of 1 also indicates a high level of power, confirming the robustness of the test in detecting significant differences. The average WMW test value of 0.798 reflects the average performance of the WMW test across the simulations.

Periodontal and mental health variables influence oral microbiome community composition

The Simpson alpha diversity index showed no significant differences between the mental health symptomatic groups and controls. Individuals who experienced higher levels of childhood physical neglect had lower Pielou’s evenness scores (mdn = 0.713, IQR = 0.05 versus mdn = 0.72, IQR = 0.05, highest neglect quartile versus lowest) (Spearman’s rank correlation test, p = 0.03, rho = −0.10, n = 470). Individuals with receding gums had a lower Simpson alpha diversity index (mdn = 0.94, IQR = 0.03 versus mdn = 0.95, IQR = 0.02) (Wilcoxon rank-sum tests, p = 0.005, r = 0.17, n = 284), and Pielou’s evenness scores (mdn = 0.70, IQR = 0.06 versus mdn = 0.72, IQR = 0.05) (Wilcoxon rank-sum tests, p = 0.02, r = 0.13, n = 284) compared to those without. Individuals with self-reported malocclusion had lower Simpson’s diversity (mdn = 0.93, IQR = 0.02 versus mdn = 0.95, IQR = 0.02) (Wilcoxon rank-sum tests, p = 0.014, r = 0.18, n = 183), Chao1 richness (mdn = 122, IQR = 19.5 versus mdn = 129.5, IQR = 22.25) (Wilcoxon rank-sum tests, p = 0.005, r = 0.2, n = 183), and Pielou’s evenness scores (mdn = 0.70, IQR = 0.04 versus mdn = 0.72, IQR = 0.06) (Wilcoxon rank-sum tests, p = 0.05, r = 0.15, n = 183) compared to those without. Higher Chao1 richness scores were evident in those reporting bleeding and inflamed gums (mdn = 132, IQR = 27 versus mdn = 125, IQR = 19.5) (Wilcoxon rank-sum tests, p = 0.003, r = 0.17, n = 284), and also for those with a clinical diagnosis of gingivitis (mdn = 135.5, IQR = 26.75 versus mdn = 128, IQR = 21.75) (Wilcoxon rank-sum tests, p = 0.013, r = 0.15, n = 284).

Several variables influenced the overall oral microbial composition (β-diversity), with smoking status eliciting the largest effect, followed by age, living environment (city, town, or rural setting), and alveolar bone loss. The most significant subset (R² ≥ 0.002 and q ≤ 0.1) of variables is illustrated in Fig. 1 (Supplementary Table 1 contains the full set of significant variables).

Fig. 1. Effect sizes of variables that had a statistically significant effect on the oral microbiome community variation (distance-based redundancy analysis (dbRDA) on genus-level Aitchison distance) in our cohort (n = 470).

Color intensity is proportional to the q-values (False Discovery Rate (FDR) corrected p-values); adjusted R² effect sizes are indicated on the y-axis. Certain variables were not available for the entire cohort: gingivitis diagnosis (ever), oral bleeding or inflammation, loose teeth, consumption of dietary whole grains (n = 284); alveolar bone loss (n = 184), psychoactive medication use (n = 265), and having arthritis or gout (n = 282). BMI - body mass index, CESD - Center for Epidemiologic Studies Depression, Whole grains – weekly whole grain consumption, TeetchCAL5 criteria - probable severe periodontitis based on the “≥50% of teeth with clinical attachment level (CAL) ≥ 5 mm” criteria.

Oral taxa associated with trauma, mental health outcomes, and psychological quality of life

Our symptomatic cohort reported significantly higher levels of childhood trauma compared to controls. Individuals who reported high levels of emotional neglect had significantly higher relative abundance (clr-transformed) of Streptococcus mutans (mdn = −2.52, IQR = 2.12) compared to those with low levels or no emotional neglect (mdn = −2.68, IQR = 0.8) (GLM q = 0.07, β = −0.6, n = 470) (Fig. 2a). Individuals with PTSD symptoms (n = 73) had significantly lower clr-transformed relative abundance of Haemophilus sputorum (H. sputorum) (mdn = −1.92, IQR = 3.07 versus mdn = 0.62, IQR = 3.94) (GLM, q = 0.09, β = −1.2, n = 237) and higher clr-transformed relative abundance of Prevotella histicola (P. histicola) (mdn = −2.92, SD = 3.07, IQR = 5.80 versus mdn = −3.25, SD = 2.38, IQR = 3.46) (GLM, q = 0.09, β = 1.6, n = 237), compared to mentally healthy controls (n = 164) (Fig. 2b).

Fig. 2. The relative abundance of oral taxa associated with mental health outcomes, trauma, and well-being.

a Higher relative abundance of S. mutans in individuals who experienced childhood emotional neglect. b Lower relative abundance of H. sputorum and higher relative abundance of P. histicola in individuals with symptoms of PTSD compared to controls. c Individuals with a current anxiety disorder diagnosis had lower levels of N. elongata and higher levels of O. asaccharolyticum. d The relative abundance of P. histicola was positively associated with CESD depressive scores and (e) negatively associated with World Health Organization Quality Of Life (WHOQOL) scores for domain 2. f Summary graphic highlighting common taxa associated with mental health outcomes, trauma, and well-being (positive associations are indicated in yellow tones, negative associations in blue tones, color intensity is proportional to standardized GLM β coefficients, and point size is proportional to the q-values (FDR corrected p-values)). Horizontal lines on the violin plots indicate the median and the thicker part of the violin around the median represents the interquartile range (IQR). Significance q ≤ 0.1 (only statistically significant taxa are illustrated). Relative abundance is the clr-transformed and filtered relative abundance values. H. sputorum - Haemophilus sputorum, P. histicola - Prevotella histicola, N. elongata - Neisseria elongate, O. asaccharolyticum - Oribacterium asaccharolyticum, CESD - Center for Epidemiologic Studies Depression, PTSD - posttraumatic stress disorder, GLM - generalized linear model.

Those with a current anxiety disorder diagnosis (n = 134) harbored significantly lower clr-transformed relative abundance of Neisseria (mdn = 3.71, IQR = 2.15 versus mdn = 4.28, IQR = 1.74) (GLM, q = 0.001, β = −0.96, n = 470), specifically Neisseria elongate (N. elongate) (mdn = −0.52, IQR = 4.74 versus mdn = 0.39, IQR = 2.94) (GLM, q = 0.09, β = −0.94, n = 470) and significantly higher clr-transformed relative abundance of Oribacterium asaccharolyticum (O. asaccharolyticum) (mdn = 0.73, IQR = 1.52 versus mdn = 0.25, IQR = 2.19) (GLM q = 0.03, β = 0.53, n = 470) compared to those without a current diagnosis (n = 336) (Fig. 2c). Interestingly, higher relative abundance (clr-transformed) of P. histicola was also evident in individuals with higher CESD depressive scores (mdn = −3.17, IQR = 4.68) (GLM q = 0.05, β = 0.04, n = 470) (Fig. 2d) and those with poor psychological quality of life scores (mdn = −3.17, IQR = 4.68) (GLM q = 0.08, β = −0.2, n = 470) (Fig. 2e). Individuals with higher CESD scores also had a higher clr-transformed relative abundance of Lancefieldella (mdn = 1.98, IQR = 1.34) (GLM q = 0.05, β = 0.01, n = 470) and O. asaccharolyticum (mdn = 0.37, IQR = 2.01) (GLM q = 0.05, β = 0.02, n = 470), however, the effect sizes were relatively small. Supplementary Table 2 contains all statistical results and Fig. 2f provides a graphical summary.

Oral microbiome signatures related to periodontal health

Several oral taxa were associated with periodontal health variables (Fig. 3, Supplementary Table 3). The clr-transformed relative abundance of Shuttleworthia was higher in participants with a self-reported periodontitis diagnosis and those with predicted severe periodontitis based on the TeethCAL5 criteria. The clr-transformed relative abundance of Capnocytophaga was lower in participants with a self-reported clinical periodontitis diagnosis and those with predicted severe periodontitis based on the TeethPPD6 criteria. Several common taxa were differentially abundant in those with a self-reported clinical periodontitis diagnosis and those who reported loose teeth (a symptom of periodontitis), including a higher clr-transformed relative abundance of Tannerella forsythia, Metaprevotella, Fretibacterium fastidiosum, and lower clr-transformed relative abundance of Prevotellaceae and Haemophilus parainfluenzae. Three taxa had a higher relative abundance (clr-transformed) in participants with a self-reported clinical gingivitis diagnosis, Parvimonas, Gallibacter, and Eggerthia; Eggerthia was the only common taxon between periodontitis and gingivitis diagnosis, whose clr-transformed relative abundance was higher in both diagnoses.

Fig. 3. Summary graphic highlighting common taxa associated with periodontal outcomes (self-reported clinical diagnosis of current periodontitis/gingivitis available for all participants, n = 470; questionnaire data to predict severe periodontitis available for n = 196 [48%] of the participants).

Positive associations are in shades of yellow, negative associations in shades of blue, and color intensity is proportional to standardized GLM β coefficients, and point size is proportional to the q-values (FDR corrected p-values). Periodontitis diagnosis (n = 34, 7%) – participants reported having a current clinical diagnosis of periodontitis; Gingivitis diagnosis (n = 30, 11%) – participants reported having a current clinical diagnosis of gingivitis; TeetchCAL5 - probable severe periodontitis based on the “≥50% of teeth with clinical attachment level (CAL) ≥ 5 mm” criteria (n = 93, 47%); TeethPPD6 - probable severe periodontitis based on the “≥25% of teeth with probing pocket depth (PPD) ≥ 6 mm” criteria (n = 81, 41.3%).

Functional potential of the oral microbiome: possible implications for mental health outcomes

Functional prediction revealed lower tryptophan metabolism/degradation in individuals with PTSD symptoms (mdn = −0.68, IQR = 1.98 versus mdn = −0.42, IQR = 2.17) (GLM, q = 0.02, β = −0.8, n = 470), those who experienced higher levels of childhood trauma (mdn = -0.35, IQR = 2.22) (GLM, q = 0.09, β = −0.01, n = 470), and those with lower quality of life relating to personal relationships) (mdn = −0.35, IQR = 2.22) (GLM, q = 0.06, β = 0.06, n = 470). Supplementary Table 4 contains all statistical results for predicted GBMs associated with periodontal variables. Interestingly, lower metabolism/degradation of tryptophan was also predicted in individuals with predicted severe periodontitis based on the TeethCAL5 criteria (n = 93) (mdn = −0.40, IQR = 1.70 versus mdn = −0.01, IQR = 1.29) (GLM, q = 0.07, β = −0.52, n = 196) and TeethPPD6 criteria (n = 81) (mdn = −0.40, IQR = 1.67 versus mdn = −0.02, IQR = 1.37) (GLM, q = 0.02, β = -0.6, n = 196,). Supplementary Table 5 contains all statistical results for GBMs associated with periodontal variables and Fig. 4 illustrates the full set of predicted GBMs linked to severe periodontitis, mental health, childhood trauma, and quality of life.

Fig. 4. Shared (solid line boxes) and unique (dotted line boxes) gut-brain modules (GBMs) associated with predicted severe periodontitis (stages III-IV) as well as mental health, childhood trauma, and quality of life variables.

Positive correlations are displayed in yellow tones and negative correlations in blue tones (white tones indicate values close to zero). Color intensity is proportional to standardized GLM β coefficients, and point size is proportional to the q-values (FDR corrected p-values).

Plasma measures

Analyses to confirm lower TRP metabolism/degradation revealed lower plasma levels of 5-HT and the ratio of 5-HT/TRP in individuals with depressive symptoms (Wilcoxon rank-sum test, p < 0.01, mean difference (MD) = 0.8, and p < 0.01, MD = 0.9, n = 282 respectively), state anxiety symptoms (p < 0.01, MD = 0.6, and p < 0.01, MD = 0.5, n = 282, respectively), trait anxiety symptoms (p < 0.01, MD = 0.6, and p < 0.01, MD = 0.6, n = 282 respectively), and PTSD symptoms (p < 0.01, MD = 1.0, and p < 0.01, MD = 0.9, n = 282 respectively) compared to healthy controls (Fig. 5).

Fig. 5. Log transformed plasma levels (nmol/ml) of serotonin (5-HT) and the ratio of serotonin/tryptophan (5-HT/TRP) in participants with self-reported mental health symptoms.

Lower plasma levels of 5-HT and 5-HT/TRP in participants with (a) depressive, (b) trait anxiety, (c) state anxiety, and (d) PTSD symptoms had lower plasma levels of 5-HT and 5-HT/TRP compared to healthy controls. Horizontal lines on the violin plots indicate the median, and the thicker part of the violin around the median represents the interquartile range (IQR). Significance p < 0.01. PTSD posttraumatic stress disorder.

Oral microbes and therapeutic response

Interestingly, two taxa associated with a current self-reported clinical diagnosis of periodontitis and/or gingivitis were also associated with self-reported efficacy of psychotherapy, namely Eggerthia and Haemophilus parainfluenza. Eggerthia was present at a higher relative abundance (clr-transformed) in those with a current self-reported clinical diagnosis of periodontitis and/or gingivitis and in individuals with poor self-reported psychotherapeutic efficacy (n = 29) (GLM q = 0.12, r = −0.62, n = 322), whereas the clr-transformed relative abundance of H. parainfluenza was lower in these individuals (GLM q = 0.12, r = 0.60, n = 322), although the association did not reach the threshold for statistical significance of q ≤ 0.1.

Discussion

This study represents one of the largest oral microbiome investigations in mental health, to date. The composition of the overall oral microbiome was significantly impacted by several mental health variables (including clinical diagnoses of anxiety disorders or depression), as well as periodontal symptoms, predicted severe periodontitis, and self-reported clinical diagnoses of gingivitis and/or periodontitis. These findings correlate with previous research highlighting the significant effects of mental [32] and periodontal [60, 61] health on the oral microbiome beta diversity. Various additional factors shaped the oral microbiome composition, aligning with earlier research emphasizing the impact of factors such as smoking [62], BMI [63], age [60], arthritis [64], gout [65], and geographic location [66] on the oral microbiome.

None of the mental health variables or self-reported periodontal outcomes influenced Simpson’s diversity. Earlier studies also failed to detect differences in alpha diversity between individuals with depression and anxiety compared to controls [31]. We detected lower Pielou’s evenness scores in individuals who experienced higher levels of childhood physical neglect. One study reported no significant difference between Simpson and Shannon diversity indices of the oral microbiome among those who experienced early life trauma [67], but these diversity metrics evaluate richness and evenness, which could explain the discrepant results.

Lower diversity was evident among individuals with self-reported receding gums (Simpson’s diversity and Pielou’s evenness), and malocclusion (Simpson’s diversity, Chao1 richness, and Pielou’s evenness), whilst higher Chao1 richness scores were noted in those reporting bleeding and inflamed gums and a reported clinical diagnosis of gingivitis. Receding gums could be a symptom of periodontitis, and bleeding and inflamed gums are the main symptoms of gingivitis. Alpha diversity findings in periodontitis patients have shown contradictory results, with some studies reporting no differences between periodontitis patients and controls [61], and others noting higher [68] or lower diversity [69] in periodontitis patients. The latter study reported the loss of diversity in oral microbiota between healthy individuals, patients with stable periodontitis, or patients with progressing periodontitis, reporting a strong association between lower alpha diversity in the oral microbiota and the progression of periodontitis, illustrating the potential of using oral microbial alpha diversity as a predictor of periodontitis [69]. These conflicting results require further investigation, particularly through studies involving large cohorts or meta-analyses of combined cohorts. Other studies have also reported higher alpha diversity (Chao1 and Shannon indices) in gingivitis patients compared to controls [70]. Concerning malocclusion, one recent study also reported lower Simpson’s diversity and Chao1 richness in individuals with malocclusion undergoing orthodontic treatment [71].

Several taxa were associated with mental health, trauma, and well-being. Due to the limited data currently available for the oral microbiome and mental health outcomes, we will discuss our findings in the context of the current literature available on the oral microbiome in mental health disorders. We recognize that mental health disorders are distinct from one another, although some share common symptoms. Precisely because of this, identifying commonalities in the abundance of oral taxa across different disorders could provide further insights into the transdiagnostic role of these taxa in mental health symptoms.

P. histicola is of particular interest, with a higher clr-transformed relative abundance in individuals with PTSD symptoms, those with higher CESD scores, and those with poorer interpersonal quality of life. Prevotella is the second most common bacteria dominating the oral cavity and this diverse genus includes several species. P. histicola is a facultative oral pathogen, which can cause pathologies such as caries and periodontitis [72, 73]. Although no data is currently available on the relationship between Prevotella species and PTSD or depression, a negative association was previously reported between the relative abundance of Prevotella and distress [74]. Importantly, Prevotella is a genus strongly associated with waking samples, and the majority of our samples were collected close to waking, which could explain the discrepancies between the findings. Furthermore, the majority of studies report on genus level, whereas our finding is for the species P. histicola, and research has shown that species from the same genus could fulfill vastly different roles [75, 76], therefore future studies should aim for higher resolution of taxonomic classification to disentangle taxonomic functions and disease associations.

We noted a lower clr-transformed relative abundance of N. elongata in individuals with a clinical diagnosis of anxiety disorders. Although no literature is available regarding the association of this taxon and anxiety disorders, similar observations were made in patients with schizophrenia and mania [77], whilst research in young adults linked depression with a higher relative abundance of Neisseria spp [32]. Neisseria species, integral members of the oropharyngeal flora [78], play crucial roles in maintaining oral [79] and cardiovascular health [79, 80]. Their presence correlates with good oral health, attributed to their aerobic, nitrite-reducing capabilities essential for gum health [79–81]. Moreover, Neisseria-dominated oral microbiomes exhibit a reduced likelihood of hosting the cariogenic pathogen S. mutans [82]. Notably, we detected a higher clr-transformed relative abundance of S. mutans in individuals reporting childhood emotional neglect, a known risk factor for mental health disorders. These findings underscore the intricate interplay between oral microbial composition, mental health outcomes, and early life adversity. The involvement of cross-feeding and interactions among microbial taxa adds complexity to understanding comorbidity and risk factors in mental health conditions.

Individuals with an anxiety disorder diagnosis and higher CESD scores harbored a higher clr-transformed relative abundance of O. asaccharolyticum. This correlates with the higher relative abundance reported in elderly people receiving treatment for anxiety, depression, and insomnia [83]. Furthermore, a higher relative abundance of the Odoribacter genus was also noted in the gut microbiomes of patients with depression [84], a preclinical model of depression [85], and individuals with periodontal disease [86]. A higher relative abundance of the Odoribacter genus was also detected in the oral microbiome of periodontitis patients [87]. These findings underscore the potential interconnectivity between oral and gut microbiomes and taxa implicated in both mental and periodontal health, with implications for the oral-gut-brain axis. Furthermore, in a randomized, double-blind placebo-controlled trial, synbiotics reduced both systemic inflammation and systemic lupus erythematosus disease activity, whilst simultaneously also depleting O. asaccharolyticum from the microbiome [88], suggesting pathogenicity and a potential therapeutic target to facilitate anti-inflammatory effects, which warrant further investigation.

Participants with PTSD symptoms had a lower clr-transformed relative abundance of H. sputorum, correlating with findings in young adults with depression [32]. Haemophilus is a nitrate-reducing genus, and therefore, higher abundances are associated with good oral health. This taxon is also depleted in the oral and gut microbiomes of individuals with rheumatoid arthritis (RA), which also correlated with higher levels of serum autoantibodies [89], suggesting a potential involvement in autoimmunity and inflammation. Interestingly, PTSD is associated with RA, with female PTSD patients having a 76% higher risk of developing RA [90]. Levels of Haemophilus in the oral and gut microbiomes could therefore be involved in this comorbidity, possibly via its immunomodulatory effects.

The clr-transformed relative abundance of several taxa was associated with periodontal outcomes; Shuttleworthia and Capnocytophaga were associated with a self-reported clinical diagnosis of periodontitis and predicted severe periodontitis. Previous studies reported similar findings in periodontitis [87, 91] and gingivitis patients [92]. Shuttleworthia and Capnocytophaga should therefore be investigated as potential non-invasive, salivary microbiome markers of periodontitis.

Eggerthia was more prevalent in those with a periodontitis/gingivitis diagnosis, correlating with previous reports in periodontitis [93], and gingivitis patients [70]. These findings suggest that salivary levels of Eggerthia should be investigated as an early, non-invasive indicator of periodontal health problems. We however did not detect differences in the clr-transformed relative abundance of certain keystone periodontitis-associated taxa, including Porphyromonas gingivalis, which could be attributed to analyzing saliva samples and not periodontal pocket samples [94].

While no shared oral taxa were associated with both mental and periodontal health, we found a common functional pathway: metabolism/degradation of TRP. This pathway was diminished in individuals with PTSD symptoms, those who experienced childhood trauma, those with poor interpersonal quality of life, and those with predicted periodontitis. Decreased degradation of TRP through the 5-HT pathway could result in lower 5-HT levels and higher TRP levels. Our data revealed reduced plasma levels of 5-HT and 5-HT/TRP ratios in all symptomatic groups compared to healthy controls. Decreased 5-HT/TRP ratios may result from lower 5-HT levels and increased TRP levels. Directly measuring 5-HT levels in plasma can be challenging due to factors like its short half-life and difficulties in accurately measuring relatively low 5-HT plasma levels [95]. Therefore, the 5-HT/TRP ratio allows for an indirect assessment of serotonin synthesis capacity, which may indicate alterations in serotonin function in the CNS [96, 97].

Lower levels of 5-HT align with previous research on depression [98] and PTSD [99]. Serotonin-mediated neurotransmission is implicated in anxiety disorders, although its relationship is complex due to the diversity of anxiety disorders [100]. Serotonin is crucial for CNS development and function [101, 102], yet it also affects oral health. Psychotropic drugs like selective serotonin reuptake inhibitors (SSRIs) can reduce the salivary flow rate and cause xerostomia (dry mouth) [103], affecting oral cleansing and tooth decay prevention [104]. While our study didn’t find statistically significant differences in 5-HT levels in those with clinical diagnosis or predicted severe periodontitis, altered serotonin levels could influence oral health.

Although serotonin is vital for mental health, most of its production ( ~ 90%) occurs in the gastrointestinal tract [105], influencing various physiological processes beyond the CNS, including colonic motility [106]. While our findings suggest potential oral microbiota involvement in TRP metabolism and systemic serotonin levels, systemic levels don’t directly reflect CNS levels due to the BBB. Psychotropic medications, like SSRIs, may have affected CNS serotonin levels in this cohort. Nevertheless, our study underscores the importance of the serotonergic system in mental and oral health, suggesting avenues for further research into oral microbiota, TRP metabolism, and serotonin production interplay. Understanding these relationships could lead to novel therapeutic approaches for mental health disorders associated with serotonin dysregulation.

Identifying treatment response markers is crucial to lighten the burden of disease and enhance treatment efficacy. Although oral taxa were not associated with psychoactive medication use, a higher clr-transformed relative abundance of Eggerthia and a lower clr-transformed relative abundance of H. parainfluenza was evident in individuals reporting poor psychotherapeutic efficacy and those with a self-reported periodontitis diagnosis, hinting at a potential effect of oral health on treatment efficacy. These associations narrowly missed statistical significance, and warrants further investigation. Data on oral microbiome and treatment response associations are limited, however, research suggests a causal effect of elevated levels of Eggerthia on anxiety and depression [29]; links between lower levels of H. parainfluenza and generalized anxiety disorder [107], and higher levels of this taxon in individuals with periodontitis + IBS [108]. These taxa are good candidates to explore in future longitudinal treatment outcome studies, especially in patients with periodontal health problems.

Limitations of this study include the absence of clinical assessments of anxiety, depression, PTSD, periodontitis, and gingivitis. Instead, validated questionnaires were used to assess symptoms, in addition, participants reported current diagnoses of any of these disorders/conditions. Mental health disorders are complex, with varying symptom presentations even among individuals diagnosed with the same disorder. Understanding these disorders in this context is crucial. Additionally, diagnoses and treatment strategies are informed by symptoms rather than rigid diagnostic criteria, with associations with biological markers often correlating more strongly with symptoms [109]. Further oral microbiome studies in well-defined clinical samples are warranted to compare findings to self-reported symptom cohorts.

Our analyses accounted for several covariates that showed associations with clr-transformed relative taxonomic abundance in our cohort. Although we tested an extensive list of potential confounders, there may be additional confounders that we were unable to test and incorporate into our models. Furthermore, it is important to be cautious of overfitting the models by including too many covariates, as this can reduce the ability to detect significant associations [110].

Although species-level assignment was possible for several taxa, it should be noted that species-level identification is less reliable for targeted 16S rRNA sequencing compared to full-length 16S rRNA sequencing or shotgun metagenomic sequencing.

Different oral niches harbor distinct microbiomes. This study investigated self-collected saliva samples as a proxy for the oral microbiome. Samples from the periodontal pocket would be ideal for studying the microbiome related to periodontitis. However, the aim of this study was mental health outcomes whilst considering self-reported periodontal health outcomes. Furthermore, studies have shown that saliva samples were the most stable within-subjects (temporal) as well as between-subjects [111]. Lastly, our cross-sectional study can only report on microbial associations with symptomology; future longitudinal studies are needed to infer causality between the oral microbiome and mental health symptoms.

This study reveals a compelling connection between the composition of oral microbiota, mental health conditions, early life experiences, as well as periodontal outcomes. By identifying taxa and functional pathways potentially involved in both mental and oral health, our findings contribute to the growing body of evidence linking oral microbiota to mental and oral health outcomes. These results suggest a complex interplay between microbial composition, systemic neuromodulators, and health outcomes, and highlight potential avenues for further research. Understanding these relationships offers promising avenues for integrated approaches to promote oral and psychological resilience, emphasizing the importance of considering both oral and mental health within a holistic framework of care.

Author contributions

SMM conceptualized the study, and performed microbiome and plasma-related experimentation, data curation, formal analysis, and writing of the manuscript. RV and EO performed HPLC analyses and manuscript editing. EM and EF performed periodontal health questionnaire scoring and assisted with manuscript writing and editing. IZ, JDDA, MC, MPGP, and AGP, assisted with participant recruitment, blood collection, sample preparation, and manuscript editing. JCL provided expert supervision and guidance, including project administration, conceptual direction, and manuscript writing and editing.

Funding

This research was supported by a 2018 NARSAD Young Investigator Grant from the Brain and Behavior Research Foundation (grant number: 27050), Una4Career grant (European Union’s Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement No 847635), a Knowledge Generation Grant from the Ministry of Science and Innovation (Spain) (PID2021-126468OA-I00), and Ministry of Health (Spain) research grant (PNSD 2022I033).

Data availability

The data that support the findings of this study are available from the corresponding author, [S Malan-Müller], upon reasonable request. The sequencing data have been deposited with links to BioProject accession number PRJNA1162741 in the NCBI BioProject database (https://www.ncbi.nlm.nih.gov/bioproject/).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41398-024-03122-4.