The gut-microbiota-brain axis in a Spanish population in the aftermath of the COVID-19 pandemic: microbiota composition linked to anxiety, trauma, and depression profiles

Abstract

The prevalence of anxiety and depression soared following the COVID-19 pandemic. To effectively treat these conditions, a comprehensive understanding of all etiological factors is needed. This study investigated fecal microbial features associated with mental health outcomes (symptoms of anxiety, depression, or posttraumatic stress disorder (PTSD)) in a Spanish cohort in the aftermath of the COVID-19 pandemic. Microbial communities from stool samples were profiled in 198 individuals who completed validated, self-report questionnaires. 16S ribosomal RNA gene V3-4 amplicon sequencing was performed. Microbial diversity and community structure were analyzed, together with relative taxonomic abundance. In our cohort of N=198, 17.17% reported depressive symptoms, 37.37% state anxiety symptoms, 40.90% trait anxiety symptoms, and 8.08% PTSD symptoms, with high levels of comorbidity. Individuals with trait anxiety had lower Simpson's diversity. Fusicatenibacter saccharivorans was reduced in individuals with comorbid PTSD + depression + state and trait anxiety symptoms, whilst an expansion of Proteobacteria and depletion of Synergistetes phyla were noted in individuals with depressive symptoms. The relative abundance of Anaerostipes was positively correlated with childhood trauma, and higher levels of Turicibacter sanguinis and lower levels of Lentisphaerae were found in individuals who experienced life-threatening traumas. COVID-19 infection and vaccination influenced the overall microbial composition and were associated with distinct relative taxonomic abundance profiles. These findings will help lay the foundation for future studies to identify microbial role players in symptoms of anxiety, depression, and PTSD and provide future therapeutic targets to improve mental health outcomes.

Keywords: COVID-19; Microbiome; anxiety; depression; gut-microbiota-brain axis; mental health; posttraumatic stress disorder; trauma.

Figure 1.

Sankey diagrams to illustrate the comorbid states in (a) the depressive symptom cohort, (b) state anxiety symptom cohort, (c) trait anxiety symptom cohort and (d) the PTSD symptom cohort.

Figure 2.

Simpson’s diversity index was significantly lower in individuals with (a) trait anxiety symptoms (mdn = 0.90) compared to those without (mdn = 0.92) (Wilcoxon rank-sum test, p = 0.016, r = 0.19, n = 198) and (b) compared to healthy controls (mdn = 0.92) (Wilcoxon rank-sum test, p = 0.02, r = 0.17, n = 198). The solid line indicates the median; the top and bottom of the boxes indicate the third and first quartiles, respectively. Whiskers indicate the 1.5 interquartile range (IQR) beyond the upper and lower quartiles and dots represent individual data points. Significance * for p ≤ 0.05.

Figure 3.

Cumulative effect sizes of variables on microbiome community variation (left blue bars; stepwise distance-based redundancy analysis (dbRDA) on genus-level Aitchison distance); individual effect sizes (assuming covariate Independence) (right Orange bars); variables with non-significant p-values for individual analyses (right gray bars) and the one variable (COVID-19 vaccination), that did not enter the dbRDA model (left gray bar).

Figure 4.

(a) F. saccharivorans was significantly lower in individuals with comorbid symptoms of PTSD + depression + state and trait anxiety (mdn = 1.12) (after correcting for main covariates) compared to individuals without these comorbid symptoms (mdn = 2.78) (GLM, p = 0.0001, r = 0.24, n = 198) and (b) compared to healthy controls (mdn = 2.82) (correction for confounding variables not possible) (Wilcoxon rank-sum tests, q = 0.1, n = 114). (c) Proteobacteria was higher in individuals with depressive symptoms (mdn = 3.42) (GLM, p = 0.02, n = 198) compared to those without (mdn = 3.02) whilst (d) Synergistetes were lower in those with depressive symptoms (mdn = −2.98) (GLM, p = 0.004, n = 198) compared to those without (mdn = −2.80). Sample sizes: PTSD + depression, state- and trait-anxiety symptoms Yes n = 8, PTSD + depression, state- and trait-anxiety symptoms No n = 190. Depressive symptoms Yes n = 32, Depressive symptoms No n = 166, Healthy controls n = 106. Solid lines indicate the median; the tops and bottoms of boxes indicate the third and first quartiles, respectively. Whiskers indicate the 1.5 IQR beyond the upper and lower quartiles. Dots represent individual data points. Abbreviations: clr – centered log-ratio, r = effect size. Significance * for p ≤ 0.05, ** for p ≤ 0.005, *** for p ≤ 0.0001, ^# for q ≤ 0.1. Fusicatenibacter saccharivorans – F. saccharivorans, posttraumatic stress disorder – PTSD, generalized linear model – GLM.

Figure 5.

Positive relationship between the relative abundance of the Anaerostipes genus and CTQ total score (Spearman r_s = 0.23; p ≤ 0.01, n = 198). (b) The relative abundance of T. sanguinis was higher in individuals who experienced life-threatening traumas (mdn = −0.65 versus mdn = −2.37) (GLM, p ≤ 0.001, r = 0.24, n = 198), and the relative abundance of Lentisphaerae was lower (GLM, p = 0.002, r = 0.20, n = 198) (mdn = −2.52 versus mdn = −2.09) compared to individuals unexposed to such traumas. Y-axes show the clr-transformed relative abundances of the taxa. For box plots, solid lines indicate the median; the top and bottom of boxes indicate the third and first quartiles, respectively. Whiskers indicate the 1.5 IQR beyond the upper and lower quartiles. Dots represent individual data points. Significance * for p ≤ 0.05, ** for p ≤ 0.005. Abbreviations: centered log-ratio – clr, Childhood Trauma Questionnaire – CTQ, Turicibacter sanguinis – T. sanguinis, effect size – r.

Figure 6.

(a) Associations between a previous COVID-19 infection and the relative abundance of Escherichia-Shigella (mdn = 0.15 versus mdn = −1.36) (GLM, p = 0.004, r = 0.23, n = 198), Parasutterella excrementihominis (mdn = 2.85 versus mdn = 0.48) (GLM, p = 0.0003, r = 0.25, n = 198), Flavonifractor plautii (mdn = 1.52 versus mdn = 0.43) (GLM, p = 0.002, r = 0.21, n = 198) and Holdemania (mdn = −0.60 versus mdn = −1.29) (GLM, p = 0.0003, r = 0.24, n = 198). (b.) Associations between COVID-19 vaccine administration and the relative abundances of Clostridium sensu stricto (mdn = −0.79 versus mdn = 0.65) (GLM, p = 0.005, r = 0.22, n = 198), Intestinibacter bartlettii (mdn = −1.58 versus mdn = −1.33) (GLM, p ≤ 0.002, r = 0.3, n = 198), Romboutsia (mdn = −0.48 versus mdn = 0.22) (GLM, p = 0.01, r = 0.22, n = 198) and the Clostridiales order (mdn = 4.61 versus mdn = 3.77) (GLM, p = 0.01, r = 0.25, n = 198). Y-axes show the clr-transformed relative abundances of the taxa. The solid line indicates the median, lower and upper bounds of boxes indicate the first and third quartiles, respectively; whiskers indicate the 1.5 IQR beyond the upper and lower quartiles. Dots represent outlier data points. Sample sizes: previous COVID-19 infection YES n = 42, previous COVID-19 infection NO n = 156. COVID-19 vaccine administered YES n = 90, COVID-19 vaccine administered NO n = 108. Significance * for p ≤ 0.05, ** for p ≤ 0.005.

Figure 7.

(a) Positive correlation between Monoglobus abundance and WHOQOL domain 1 scores (Spearman r_s = 0.26; GLM p = 0.01, n = 198). (b) Negative associations between recent alcohol use and the relative abundance of Barnesiella (mdn = 2.32 versus mdn = 2.8) (GLM, p = 0.03, r = 0.2, n = 198), and (c) between periodontitis diagnosis (current and/or previous) and the relative abundances of Dysosmobacter (mdn = 1.13 versus mdn = 2.06) (GLM, p = 0.002, r = 0.3, n = 198). (d) Lower relative abundance of Verrucomicrobia in individuals with a current/prior diagnosis of IBD/IBS/CeD (mdn = 1.75 versus mdn = 0.23) (GLM, p = 0.03, r = 0.2, n = 198). Y-axes show the clr-transformed relative abundances of the taxa. The solid line indicates the median, lower and upper bounds of boxes indicate the first and third quartiles, respectively; whiskers indicate the 1.5 interquartile range IQR beyond the upper and lower quartiles. Dots represent individual data points. Sample sizes: Alcohol intake YES n = 146, Alcohol intake NO n = 52, Periodontitis diagnosis YES n = 44, Periodontitis diagnosis NO n = 154. IBD/IBS/CeD YES n = 33, IBD/IBS/CeD NO n = 165. Significance * for p ≤ 0.05. Celiac disease – CeD, inflammatory bowel disease – IBD, irritable bowel syndrome – IBS, World Health Organization Quality Of Life scores for domain 1 (physical health) – WHOQOL_DOM1.