Brain age mediates gut microbiome dysbiosis-related cognition in older adults

doi:10.21203/rs.3.rs-4851828/v1

Background

Recent studies have focused on improving our understanding of gut microbiome dysbiosis and its impact on cognitive function. However, the relationship between gut microbiome composition, accelerated brain atrophy, and cognitive function has not yet been fully explored.

Methods

We recruited 292 participants from South Korean memory clinics to undergo brain magnetic resonance imaging, clinical assessments, and collected stool samples. We employed a pretrained brain age model– a measure associated with neurodegeneration. Using cluster analysis, we categorized individuals based on their microbiome profiles and examined the correlations with brain age, Mental State Examination (MMSE) scores, and the Clinical Dementia Rating Sum of Box (CDR-SB).

Results

Two clusters were identified in the microbiota at the phylum level that showed significant differences on a number of microbiota phylum. Greater gut microbiome dysbiosis was associated with worse cognitive function including MMSE and CDR-SB; this effect was partially mediated by greater brain age even when accounting for chronological age, sex, and education.

Conclusions

Our findings indicate that brain age mediates the link between gut microbiome dysbiosis and cognitive performance. These insights suggest potential interventions targeting the gut microbiome to alleviate age-related cognitive decline.

brain age

gut microbiome

cognition

The human gut microbiome, an intricate and dynamic ecosystem of microorganisms, has emerged as a key player in maintaining host health during diseases. Recent advances in sequencing technologies and bioinformatics have revealed the complexities of microbial communities and their pivotal roles in the metabolic, immunological, and neurological functions of host organisms. The concept of the “gut–brain axis” suggests a bidirectional communication network that links the enteric system and central nervous system (CNS). This axis encompasses several pathways, including neural, endocrine, immune, and humoral. It has been hypothesized that the microbiome influences brain health through these pathways, thereby affecting brain development, behavior, and cognitive function [1, 2]. Emerging research indicates that gut microbiome dysbiosis, a state of imbalance in microbial communities, may be associated with accelerated aging of gray matter in the brain. Dysbiosis has been linked to inflammation and increased intestinal permeability. This may lead to systemic and neural inflammation, ultimately affecting cognitive function [3]. Increasing evidence suggests that aging may trigger a cascade of mechanistic changes involving a reduction in gut microbial diversity. This includes decreased anti-inflammatory Bifidobacterium species, increased pro-inflammatory Enterococcus species, reduced immunological function [4], and increased release of inflammatory products [5]. These changes may subsequently lead to further accelerated aging, cognitive decline, and even the accumulation of amyloid and tau in the context of Alzheimer’s disease (AD) [6].

Gray matter aging is a natural process characterized by the progressive loss of neuronal density and volume, which affects cognitive abilities and increases the risk of neurodegenerative conditions, such as AD [7]. Several brain imaging studies have been conducted on gray matter volume. Brain age is a measure that utilizes an individual’s neuroimaging data, whether structural or functional, to estimate chronological age. If someone's estimated brain age is greater than their actual age, it may suggest they have experienced more environmental stressors, modifiable risk factors, or neurodegenerative processes over their lifetime. In our previous study, we trained a machine-learning algorithm to estimate brain age based on the gray matter volume and demonstrated its preliminary ability to detect future cognitive decline [8–11].

Clustering techniques are being introduced increasingly for analyzing microbiome dysbiosis. Saji et al. reported that microbiome clusters could distinguish between individuals with dementia and healthy controls and that significant differences in brain imaging characteristics were observed across these clusters [12]. Although microbiome dysbiosis has previously been demonstrated in patients with AD [13], it has not been determined whether it is directly associated with accelerated brain aging in humans.

Therefore, we aimed to explore the relationship between gut microbiome composition, brain aging, and cognitive function. Using cluster analysis, we categorized individuals based on their microbiome profiles and examined the correlations with brain age, the Mini-Mental State Examination (MMSE) score, and the score of Clinical Dementia Rating Sum of Boxes (CDR-SB), which measures cognitive function and dementia severity in older individuals.

Participants

This study was a part of the ongoing Biobank Innovations for chronic Cerebrovascular disease With ALZheimer's disease Study (BICWALZS) and the Center for Convergence Research on Neurological Disorders. BICWALZS was initiated in 2016 by the National Biobank of Korea and the Ajou University School of Medicine [14]. The original goal was to facilitate, regulate, and ensure the optimal use of human biological specimens for research using real-world data on subjective cognitive decline (SCD), mild cognitive impairment (MCI), AD, and subcortical vascular dementia (SVaD). BICWALZS is registered in the Korean National Clinical Trial Registry (Clinical Research Information Service; identifier: KCT0003391; Registration Date: 2018/07/04; http://cris.nih.go.kr/cris/en/use_guide/cris_introduce.jsp). This study was approved by the Institutional Review Board of Ajou University Hospital (AJOUIRB-SUR-2021-038). Written informed consent was obtained from all the participants and their caregivers. The participants from the BICWALZS were recruited from the memory clinics of seven university-affiliated hospitals and community geriatric centers in South Korea. However, we included participants who provided stool samples (N = 292). All participants were recruited from two sites: a memory clinic affiliated with the Ajou University Hospital and the Suwon Community Geriatric Mental Health Center. All participants were Koreans of Eastern Asian ethnicity. None of the participants in this study were part of the initial brain age training sample of our previously trained model [10].

Clinical and biological assessment

The clinical diagnosis criteria used for this study were as follows: SCD criteria included self- and/or informant reports of cognitive decline but no objective impairment in cognitive tasks (no less than − 1.5 SD in each of the neurocognitive test domain and Clinical Dementia Rating [CDR] = 0) [15]. Patients with MCI were evaluated based on a CDR score of 0.5 and the expanded Mayo Clinic criteria [16, 17]. Patients with AD dementia were evaluated using the National Institute on Aging-Alzheimer’s Association core clinical probable AD dementia criteria [18]. SVaD was evaluated based on above-moderate white matter hyperintensity (WMH) and vascular dementia criteria, following the Diagnostic Statistical Manual of Mental Disorders, fifth edition [19]. Patients with a history of neurological or medical conditions, such as territorial cerebral infarction, intracranial hemorrhage, Parkinson’s disease, heart failure, renal failure, or other medical conditions that could interfere with the study, were excluded. General cognitive function was evaluated using the MMSE [20]. Dementia severity was measured using the CDR-SB scores [16]. Cognitive function was assessed using the Seoul Neuropsychological Screening Battery, a standardized neuropsychological test that evaluates language, visuospatial abilities, memory, and executive functions [21].

For the laboratory assessments, morning blood samples were collected via venipuncture after an overnight fast. Samples were drawn into serum separation and dipotassium ethylenediaminetetraacetic acid tubes. Baseline blood tests included complete blood count, blood urea nitrogen, creatinine, albumin, liver function tests, fasting serum glucose, glycated hemoglobin (HbA1c), serum lipids, total protein, folic acid, high-sensitivity C-reactive protein, fibrinogen, venereal disease research laboratory test, Treponema pallidum hemagglutination, electrolyte analysis, vitamin B12, homocysteine, thyroid function tests, and apolipoprotein E (APOE). Along with laboratory assessments, surveys were conducted to gather basic psychosocial status (alcohol consumption, smoking, nutrition, depression, anxiety and sleep) and medical histories (hypertension, hyperlipidemia, and other past physical illnesses) [14].

Microbiome

Stool samples were collected at the Ajou University Hospital biobank the day before clinical assessment using a sterilized stool container and stored at -20°C until further processing. Bacteria within the stool were purified, and DNA was extracted for sequencing. The extracted bacterial 16s rDNA from each sample was assigned a unique barcode sequence for library preparation. Next-Generation Sequencing was then conducted on all the isolated bacterial 16s samples using the Illumina MiSeq platform technology [22]. After all microbiome samples were sequenced, taxonomic profiling using SILVA DB and downstream analyses were conducted [23].

Magnetic resonance imaging (MRI) analysis and brain age estimation

We have previously validated a brain age estimation algorithm that predicts chronological age using gray matter volume [24]. This algorithm was developed with the Pattern Recognition for Neuroimaging Toolbox [25]. Whole-brain, voxel-wise gray matter volume maps were mean-centered and used to calculate a similarity matrix kernel [26], which was then input into a Gaussian process regression to predict chronological age. The training set, consisting of 757 magnetic resonance images of adult individuals without any psychiatric or neurological disorders, as well as Alzheimer's pathology as measured by positron emission tomography, has been previously described [24]. Although WMHs are likely factors that influence brain aging, our brain age model primarily utilized gray matter data and not white matter. Thus, our brain age marker could be more accurately described as “gray matter” age [27]. We independently validated the utility of our brain age model using BICWALZS data, which may offer benefits in discriminating and tracking cognitive impairment as well as amyloid deposition status in older adults [28].

Methods used for MRI acquisition and structural MRI processing are described in Additional file 1. MRI sequence parameters are also presented in Additional file 1.

Statistical analysis

Based on the bacterial genus abundance, we used the Dirichlet multinomial mixtures (DMMs) algorithm to identify microbial community types [29]. The DMM was applied to 16S rRNA gene sequencing, and the microbiota composition was analyzed at the phylum level. The DMM can efficiently cluster samples based on the relative abundance of the identified microbiota. Based on the lowest Laplace approximation index, appropriate microbial community type numbers (DMM clusters) were determined. The DMM methodology effectively manages the large data dimensionality associated with microbiome analyses. This facilitated the use of sequencing results, including bacterial community types and clinical variables, for multivariable analyses. The analyses described above were performed using the R package Dirichlet Multinomial v1.36.0.

We compared the baseline characteristics, clinical assessment, biological assessment, and brain age between the cluster groups using independent sample t-tests and chi-square analyses for parametric and non-parametric variables, respectively. To understand what factors were associated with cluster groupings, we conducted elastic net regression on the groups (as outcome) with the following sets of features: sex, education, chronological age, brain age_residual, body mass index, global amyloid standardized uptake value, the MMSE score, the CDR-SB score, neurocognitive test results (Digit span test, Boston Naming Test, Rey complex figure test, Seoul verbal learning test delayed recall test, and Controlled Oral Word Association Test), psychological and behavioral symptoms (Beck Anxiety Inventory, Montgomery–Asberg Depression Rating Scale, Pittsburgh Sleep Quality Index, and alcohol consumption), blood test results (hemoglobin, hematocrit, mean corpuscular hemoglobin, mean corpuscular hemoglobin concentration, mean corpuscular volume, blood urea nitrogen/creatinine, thyroid stimulating hormone, free thyroxine, albumin, total cholesterol, triglycerides, high-density lipoprotein [HDL], low-density lipoprotein, fasting glucose, HbA1c, vitamin B12, and homocysteine), history of hypertension, diabetes, thyroid disease, brain MRI WMH, lacune and intracranial volume, APOE status, blood amyloid and tau level, and study site. We conducted multiple imputations (five imputations) with all features included, using the mice package in R with the random forest approach. We then conducted an elastic net regression using the eNetXplorer package in R. We included all the features listed above to predict the cluster groupings. We employed five-fold cross-validation, optimizing α values from 0 to 1 and λ values. This process was repeated 500 times, and the cluster groups were then permuted with 250 permutations to estimate null bounds. We then used these to estimate the p-values for the models and each variable. We chose the model with the best performance and lowest regularization (e.g., α closest to 0). The p-values were then combined using Fisher’s method [30–32].

Mediation analyses were performed to assess whether brain age mediated the relationship between the microbiome cluster and MMSE scores or between the cluster and CDR-SB scores in the full sample. Explorative analyses for mediation effects of plasma amyloid A40, A42, p-tau 217 and peak width of skeletonized mean diffusivity (PSMD) were conducted additionally (Additional file 1). Mediation analyses were conducted using the R package lavaan v0.6-11 [33]. We calculated 95% bias-corrected and accelerated bootstrap confidence intervals for indirect effects using at least 5,000 bootstrap samples. Confidence intervals excluding zero indicated significant bootstrapped mediation effects. Hayes introduced the estimation of mediating effects using PROCESS for the SPSS Macro software [34]. Our model aimed to assess the effect of microbiome clusters on the MMSE and CDR-SB scores, mediated by brain age. To examine our proposed model, we used PROCESS Model 7. The same bootstrap method was used to calculate confidence intervals. Statistical significance was evaluated at a 5% significance level (p < 0.05), and all analyses were performed using R (version 4.1.0) and its open-source statistical packages.

Demographic characteristics

In total, 288 patients with normal cognitive function, SCD, MCI, AD, or SVaD were included. The characteristics of the samples are listed in Additional file 1. The mean age of participants at baseline was 72.53 ± 6.86 years, and their estimated brain age was 75.58 ± 4.79 years. The proportions of participants with a clinical diagnosis of MCI and dementia were 69.2% and 14.9%, respectively. Additionally, 83.0% showed cognitive impairment below the global CDR of 0.5.

Identification and characteristics of microbial community types by DMM

Two DMM clusters were identified in the microbiota at the phylum level (Figure 1). The cluster 1 group was characterized by six genera with predominant commensal bacteria: Firmicutes, Actinobacteriota, Proteobacteria, Bacteroidota, Verrucomicrobiota, and Euryarchaeota. The cluster 2 group was characterized by seven genera with predominant commensal bacteria: Firmicutes, Bacteroidota, Actinobacteriota, Proteobacteria, Verrucomicrobiota, Euryarchaeota, and Cyanobacteria. The microbial community was mainly composed of Firmicutes, Bacteroidota, Actinobacteriota, and Proteobacteria, which is consistent with the results of previous studies [35]. However, Bacteroidota and Proteobacteria were more abundant in the cluster 2 group than in the cluster 1 group.

Figure 1. Dirichlet multinomial mixtures (DMM) clustering using microbiome data

The cluster 1 group is characterized by six genera with predominant commensal bacteria: Firmicutes, Actinobacteriota, Proteobacteria, Bacteroidota, Verrucomicrobiota, and Euryarchaeota. The cluster 2 group is characterized by seven genera with predominant commensal bacteria: Firmicutes, Bacteroidota, Actinobacteriota, Proteobacteria, Verrucomicrobiota, Euryarchaeota, and Cyanobacteria. Bacteroidota,and Proteobacteria are more abundant in the cluster 2 group than in the cluster 1 group * p<0.05

The characteristics between the cluster groups did not differ significantly in terms of sex, the CDR score, diagnosis, or clinical assessment (Table 1). However, the cluster 2 group was older (p=0.020), had lower MMSE scores (p=0.038), higher CDR-SB scores (p=0.028), higher brain age_residual (p=0.026), and a higher Bacteroidetes/Firmicutes ratio (p<0.001) than the cluster 1 group did.

Table 1. Clinical characteristics of participants according to the data-driven cluster groups using microbiome data

	Group 1 (N = 191)		Group 2 (N = 97)		X or T	P value
	N, or mean	%, or SD	N, or mean	%, or SD	X or T	P value
Age	71.84	7.06	73.84	6.30	-2.34	0.020
Sex (female)	132	68.40	71	72.40	0.51	0.502
CDR					4.28	0.233
0	6	3.10	14	1.00
0.5	160	82.90	75	76.50
1	23	11.90	19	19.40
2	4	2.10	3	3.10
Dx					1.52	0.468
SCD	11	5.70	8	8.20
MCI	139	72.00	64	65.30
Dementia	43	22.30	26	26.50
CDR-SB	2.25	0.15	2.86	0.23	4.89	0.028
MMSE	24.11	0.33	22.85	0.49	4.38	0.038
BMI	23.99	0.30	24.13	0.46	0.06	0.801
MNA	20.74	0.32	21.08	0.48	0.33	0.567
Brain age residual	-0.15	0.09	0.20	0.13	5.04	0.026
Bacteroidota/Firmicutes ratio	0.027	0.075	0.473	0.475	-9.188	< 0.001

Abbreviations: SD, standard deviation; CDR-SB, Clinical Dementia Rating Sum of Box; MMSE, Mini-Mental State Examination; BMI, body mass; MNA, Mini Nutritional Assessment; SCD, subjective cognitive decline; MCI, mild cognitive impairment

Elastic Net Predictions

The best model achieved an out-of-bag accuracy of 0.66 predicting groups (α=0, λ=68.9, p<0.001). The results for each individual factor are presented in Table 2, which shows the pooled p-values across the imputations. The cluster 1 group was associated with higher anxiety, lifetime alcohol consumption, and a higher likelihood of thyroid disorders, compared with the cluster 2 group. However, individuals with the cluster 1 group were younger, had lower brain age, lower global amyloid, higher overall cognitive function (working memory, spatial memory, word recall, and overall cognitive function), and higher hematocrit, hemoglobin, HDL, and vitamin B12 levels, compared with individuals with the cluster 2 group.

Table 2. Factors associated with cluster groups identified by elastic net regression

	fcoeff average	fcoeff SD	P vales pooled	Compared to cluster 2, cluster 1 has
Chronological age	0.0010	1.77E-05	7.200E-06	lower age
Brain age_residual	0.0009	1.86E-06	3.520E-05	lower brain age
Global_amyloid_SUVR	0.0008	6.95E-05	0.001	lower amyloid
CDR-SB	0.0006	7.54E-06	0.019	better overall cognitive function
BAI	-0.0010	1.98E-06	2.090E-06	higher anxiety
Alcohol consumption	-0.0009	2.48E-06	1.080E-05	higher lifetime drinking
RCFT_DR_ z score	-0.0007	2.43E-05	0.002	higher spatial memory
RCFT_Copy_z score	-0.0006	6.43E-06	0.031	higher spatial memory
BNT_ z score	-0.0007	1.42E-05	0.007	higher naming function (word recall)
COWAT_Animal_z score	-0.0008	1.09E-05	0.001	higher working memory
Hematocrit	-0.0007	2.00E-05	0.002	higher hematocrit
Hemoglobin	-0.0007	2.06E-05	0.006	higher hemoglobin
HDL	-0.0007	3.31E-05	0.007	higher HDL
Vitamin B12	-0.0007	2.47E-05	0.009	higher Vitamin B12
History of thyroid disease	-0.0007	1.35E-06	0.014	higher likelihood of thyroid disorder

Abbreviations: SD, standard deviation; SUVR, standardized uptake value; CDR-SB, clinical dementia rating sum of box; BAI, Beck anxiety inventory; SCD, subjective cognitive decline; RCFT_DR, Rey complex ﬁgure delayed recall test; BNT, Boston naming test; COWAT, controlled oral word association test; HDL, High-Density Lipoprotein

Mediational analyses

We explored the mediating roles of brain age between group clusters and cognitive function (measured using MMSE and CDR-SB). For the mediation analysis, we divided the data into two clusters, coding cluster 1 as “0” and cluster 2 as “1.” Mediation analyses results are shown in Figure 2, Table 3, and Table 4. Table 3 presents the estimated regression coefficients for the clusters, brain age, and the MMSE score. As observed, individuals with the cluster 2 group exhibited greater brain age than did those with the cluster 1 group (ß =0.283, 95% confidence interval [CI]=0.044 to 0.523, p=0.021). Furthermore, when controlling for the cluster group, as a mediator, brain age significantly affected the MMSE score (ß=-1.244, 95% CI=-1.800 to -0.687, p<0.001). Although the direct effect was insignificant, the indirect effect of the cluster group on the MMSE score through brain age was significant (ß=-0.352, 95% CI=-0.690 to -0.015, p=0.041). The mediation analysis, adjusted for sex, age, and education, was consistent with the unadjusted analysis.

Figure 2. Mediation analysis among cluster group, brain age, and cognitive function measures

Individuals with the cluster 2 group exhibit greater brain age than do those with the cluster 1 group (ß =.436, 95% confidence interval [CI]=0.118 to 0.883, p=0.024). When controlling for the cluster group, as a mediator, brain age significantly affects the MMSE score (ß=-1.288, 95% CI=-1.793 to -0.783, p<0.001). Although the direct effect was insignificant, the indirect effect of the cluster group on the MMSE score through brain age is significant (ß=-0.370, 95% CI=-0.714 to -0.026, p=0.035). For the CDR-SB score outcome variable, individuals with the cluster 2 group exhibit greater brain age than do those with the cluster 1 group (ß=0.436, 95% CI=0.118 to 0.883, p=0.024). When controlling for the cluster group, as a mediator, brain age significantly affects the CDR-SB score (ß=0.425, 95% CI=0.189 to 0.661, p<0.001). Both direct and indirect effects of the cluster group on the CDR-SB score through brain age are not significant; however, the indirect effect shows the same tendency as the MMSE score outcome at the p value of 0.052 (direct: ß=0.301, 95% CI=-0.198 to 0.801, p=0.238; indirect: ß=0.122, 95% CI=-0.001 to -0.246, p=0.052, respectively). Sex, age, and education are adjusted for in the analysis.

Abbreviations: CDR-SB, Clinical Dementia Rating Sum of Box; MMSE, Mini-Mental State Examination; * p<0.05; **p<0.01

Table 3. Mediation analysis among cluster groups, brain age, and MMSE

	Coeff (ß)	SE	T	P value	LLCI	ULCI
Unadjusted analysis
Outcome: Brain age residual
Cluster group	0.283	0.122	2.316	0.021*	0.044	0.523
Outcome: MMSE
Cluster group	-0.195	0.595	-0.328	0.743	-1.361	0.971
Brain age residual	-1.244	0.284	-4.382	<0.001*	-1.800	-0.687
Indirect effect	-0.352	0.172	-2.047	0.041*	-0.690	-0.015
Direct effect	-0.195	0.595	-0.328	0.743	-1.361	0.971
Total effect	-0.547	0.609	-0.899	0.368	-1.740	0.646
Adjusted analysis
Outcome: Brain age residual
Cluster group	0.436	0.193	2.262	0.024*	0.118	0.883
Sex	-0.004	0.127	-0.035	0.972	-0.254	0.245
Age	-0.002	0.009	-0.196	0.845	-0.019	0.015
Education	0.005	0.013	0.360	0.719	-0.020	0.029
Outcome: MMSE
Cluster group	-0.195	0.545	-0.358	0.720	-1.264	0.873
Brain age residual	-1.288	0.258	-5.000	<0.001*	-1.793	-0.783
Sex	-0.035	0.557	-0.063	0.950	-1.126	1.056
Age	0.047	0.038	1.245	0.213	-0.027	0.122
Education	0.427	0.055	7.777	<0.001*	0.319	0.534
Indirect effect	-0.370	0.176	-2.109	0.035*	-0.714	-0.026
Direct effect	-0.195	0.545	-0.358	0.720	-1.264	0.873
Total effect	-0.565	0.563	-1.004	0.315	-1.669	0.538

Abbreviations: MMSE, mini mental status examination; SE, standard error; LLCI, lower limit confidence interval; ULCI, upper limit confidence interval. * p<0.05

Table 4. Mediation analysis among cluster groups, brain age, and CDR-SB

	Coeff (ß)	SE	T	P value	LLCI	ULCI
Unadjusted analysis
Outcome: Brain age residual
Cluster group	0.283	0.122	2.316	0.021*	0.044	0.523
Outcome: CDR-SB
Cluster group	0.288	0.252	1.140	0.254	-0.207	0.783
Brain age residual	0.425	0.120	3.524	<0.001*	0.188	0.661
Indirect effect	0.120	0.062	1.935	0.053	-0.002	-0.242
Direct effect	-0.195	0.595	-0.328	0.743	-1.361	0.971
Total effect	-0.547	0.609	-0.899	0.368	-1.740	0.646
Adjusted analysis
Outcome: Brain age residual
Cluster group	0.436	0.193	2.262	0.024*	0.118	0.883
Sex	-0.004	0.127	-0.035	0.972	-0.254	0.245
Age	-0.002	0.009	-0.196	0.845	-0.019	0.015
Education	0.005	0.013	0.360	0.719	-0.020	0.029
Outcome: CDR-SB
Cluster group	0.301	0.255	1.182	0.237	-0.198	0.800
Brain age residual	0.425	0.120	3.533	<0.001*	0.189	0.661
Sex	0.108	0.260	0.417	0.677	-0.401	0.618
Age	-0.006	0.018	-0.332	0.740	-0.041	0.029
Education	-0.014	0.026	-0.550	0.583	-0.064	0.036
Indirect effect	0.122	0.063	1.943	0.052	-0.001	0.246
Direct effect	0.301	0.255	1.181	0.238	-0.198	0.801
Total effect	0.423	0.258	1.641	0.101	-0.082	0.929

Abbreviations: CDR-SB, clinical dementia rating sum of box; SE, standard error; LLCI, lower limit confidence interval; ULCI, upper limit confidence interval. * p<0.05

For the CDR-SB score outcome variable in Table 4, individuals with the cluster 2 group exhibited greater brain age than did those with the cluster 1 group (ß=0.283, 95% CI=0.044 to 0.523, p=0.021). Furthermore, when controlling for the cluster group, as a mediator, brain age significantly affected the CDR-SB score (ß=0.425, 95% CI=0.188 to 0.661, p<0.001). In contrast, both direct and indirect effects of the cluster group on the CDR-SB score through brain age were not significant; however, the indirect effect showed the same tendency as the MMSE score outcome at the p value of 0.053 (direct: ß=-0.195, 95% CI=-1.361 to 0.971, p=0.743; indirect: ß=0.120, 95% CI=-0.002 to -0.242, p=0.053, respectively). The results of the mediation analysis, adjusted for sex, age, and education, was consistent with those of the unadjusted analysis.

Explorative analyses for mediation effects of plasma amyloid A40, A42, p-tau 217 and PSMD did not show any significant results (Additional file 1).

In this study, we investigated whether brain age impacted the relationship between gut microbiome dysbiosis and cognitive performance in a cohort of older adults with normal cognition, MCI, AD, or SVaD. Our analysis revealed that brain age significantly mediated the relationship between microbiome clusters and cognitive performance, including the MMSE and CDR scores. Moreover, this association for the MMSE score was independent of the chronological age, sex, and education. The intricate relationship between the gut microbiome and brain health represents a paradigm shift in our understanding of neurodegenerative diseases. The concept that dysbiosis contributes to accelerated gray matter aging is not only innovative but also poses a substantial challenge to traditional therapeutic approaches. In this discussion, we aim to delve into the mechanisms by which dysbiosis could potentially influence brain aging and discuss the implications for future research and therapy.

Mechanisms of Microbiome Influence on Brain Aging:

Recent research has elucidated several pathways through which the gut microbiota impacts the CNS. First, inflammatory pathways are implicated, because dysbiosis can induce systemic inflammation. This occurs through the release of microbial byproducts that function as pro-inflammatory agents, potentially leading to neuroinflammation and contributing to neurodegenerative processes in the CNS [29]. Second, the metabolic products produced by the gut microbiota, such as short-chain fatty acids, play a critical role. These metabolites significantly affect brain chemistry and neurophysiology by modulating neuroinflammation, oxidative stress, and neuroplasticity [36]. Third, the vagus nerve serves as a direct communication pathway between the gut and brain. The gut microbiota can influence brain function through this neural link, impacting areas associated with stress response, anxiety, and memory [37]. Collectively, these mechanisms underscore the intricate interplay between the gut microbiota and brain function and highlight the gut–brain axis as a crucial component in understanding neuroinflammatory and neurodegenerative disorders.

Impact of gut microbiome dysbiosis on brain aging and cognitive function

Disruption of the balance of the gut microbiota is often referred to as dysbiosis and has been associated with several conditions, including obesity, metabolic syndrome, and inflammatory disorders [38]. Gut dysbiosis has significant implications for neurodegeneration and brain aging. An imbalance in gut microbiome composition has been linked to an increased production of pro-inflammatory cytokines. These cytokines can cross the blood-brain barrier and contribute to neuroinflammation, which is a recognized factor in accelerating brain aging and the development of neurodegenerative diseases [39]. Regarding the specific composition of dysbiosis, there was a significant decrease in the proportion of the phylum Firmicutes and a substantial increase in Proteobacteria in patients with AD relative to controls [40]. Another study has reported a reduction in Firmicutes and an increase in Bacteroidetes species in AD [41]. In our clustering results, consistent with those of previous studies, cluster 2 exhibited a lower proportion of Firmicutes and higher proportions of Proteobacteria and Bacteroidota than did cluster 1. Furthermore, cognitive function, including that measured using MMSE and CDR-SB, was significantly worse and brain age was significantly higher, with significant gut dysbiosis, in the cluster 2 group than in cluster 1 group. In the factor analysis between the groups using elastic nets, brain age showed the highest correlation, second only to chronological age. Additionally, there were differences in specific cognitive function measures, such as spatial memory, naming function, and working memory. Notably, a high HDL level was identified as a factor that contributed less to gut dysbiosis, which aligns with existing explanations regarding neuroinflammation through metabolic inflammation [38]. Considering the close relationships among gut dysbiosis, brain age, and cognitive function, our study shed light on the role of gut dysbiosis in cognitive performance, with brain age as a mediating factor. A key finding of our study was that the link between severe gut dysbiosis and poor cognitive performance was partly mediated by the brain age.

Clinical and therapeutic implications

Understanding the relationship between microbiome profiles and brain aging may lead to the development of microbial biomarkers that predict the rate of gray matter aging, cognitive decline, and the risk of dementia [40]. Probiotics, prebiotics, and dietary interventions can be explored as potential strategies to modulate the gut microbiome, thereby mitigating neurodegenerative processes and promoting cognitive health in elderly populations [42].

Establishing a causal relationship between the clusters and brain aging could lead to microbiome-based interventions. For example, modifying the diet to improve microbiome dysbiosis may help reduce neuroinflammation and slow gray matter aging. Probiotics and prebiotics designed to optimize microbiome composition could potentially serve as adjuvant therapies for the management of neurodegenerative diseases.

Limitations

This study has some limitations. First, the current study employed a cross-sectional design, which provides only a snapshot in time. Longitudinal studies are needed to establish causal relationships between microbiome changes and brain aging. Such studies would offer insights into the dynamic interactions between the gut microbiome and brain health over extended periods, enhancing our understanding of the long-term effects. Second, this study was conducted exclusively with East Asian participants. The gut microbiome is highly variable among individuals and influenced by factors such as diet, environment, and genetics. Given these circumstances, the current results should be interpreted in the context of race, and further studies with greater diversity are needed. Third, we focused solely on an older population with varying degrees of cognitive function. As the results may vary by age group and cognitive function level, further studies involving different age groups and disease populations are required.

We found that brain age mediated the relationship between cluster-related gut microbiome dysbiosis and cognitive performance. Our results suggest that the relationship between increased gut microbiome dysbiosis and worsened cognitive performance may be partially mediated by brain age. The implications of these findings could pave the way for novel interventions targeting the gut microbiome to mitigate age-related cognitive decline and improve the quality of life in older adults.

MMSE: Mini-Mental State Examination

CDR-SB: Clinical Dementia Rating Sum of Box

AD: Alzheimer’s disease

BICWALZS, Biobank Innovations for chronic Cerebrovascular disease With ALZheimer's disease Study

SCD: subjective cognitive decline

MCI: mild cognitive impairment

SVaD: subcortical vascular dementia

CDR: Clinical Dementia Rating

WMH: white matter hyperintensity

HbA1c: glycated hemoglobin

APOE: apolipoprotein E

DMM: Dirichlet multinomial mixtures

HDL: high-density lipoprotein

PSMD: peak width of skeletonized mean diffusivity

CNS: central nervous system

ETHICS APPROVAL AND CONSENT TO PARTICIPATE

BICWALZS was registered in the Korean National Clinical Trial Registry (KCT0003391|| Registration Date: 2018/07/04|| http://cris.nih.go.kr/cris/en/use_guide/cris_introduce.jsp). The research protocol was approved by the Institutional Review Boards of Ajou University Hospital (AJOUIRB-SUR-2021-038) and conducted in accordance with the current version of the Declaration of Helsinki. Written informed consent was obtained from all the participants and caregivers.

CONSENT FOR PUBLICATION

Not applicable

AVAILABILITY OF DATA AND MATERIALS

The datasets analyzed during the current study are available from the corresponding author on reasonable request.

COMPETING INTERESTS

The authors declare that they have no competing interests.

FUNDING

This study was conducted using biospecimens and data from the Biobank Innovations for Chronic Cerebrovascular Disease With ALZheimer’s Disease Study (BICWALZS) consortium, which was funded by the Korea Disease Control and Prevention Agency for the Korea Biobank Project (#6637-303). This research was supported by grants from the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health and Welfare, Republic of Korea (HR21C1003 and HR22C1734). The biospecimens and data used in this study were provided by the Biobank of Ajou University Hospital, a member of the Korea Biobank Network. This work was supported by the National Research Foundation of Korea (NRF), funded by the Ministry of Science and ICT (NRF-2019R1A5A2026045). This research was supported by a grant from the Korea Dementia Research Project through the Korea Dementia Research Center (KDRC), funded by the Ministry of Health & Welfare and the Ministry of Science and ICT, Republic of Korea (RS-2024-00339665).The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication.

AUTHORS’ CONTRIBUTIONS

Conceptualization/Funding acquisition: SJS, CHH, and HWR; Data generation: SJS, CHH, HWR, TP, PF, BB, BP, NK, JWC, JYC, YKK, TSS, CUK, COK, SYY; Data Curation: SJS, HWR; Formal Analysis: DYL, HTK, SJS; Original Draft: SJS, DYL, HTK; and Reviewing and Editing: ML, AK, HA, CA, EJ, YHC, SH, YJN

ACKNOWLEDGEMENTS

We thank the staff of the BICWALZS and the Suwon Geriatric Mental Health Center for their involvement in data acquisition.

Cryan JF, O’riordan KJ, Cowan CSM, Sandhu KV, Bastiaanssen TFS, Boehme M, et al. The microbiota-gut-brain axis. Physiol Rev. 2019;99:1877-2013.
Sharon G, Sampson TR, Geschwind DH, Mazmanian SK. The central nervous system and the gut microbiome. Cell. 2016;167:915-32.
Zhuang ZQ, Shen LL, Li WW, Fu X, Zeng F, Gui L, et al. Gut microbiota is altered in patients with Alzheimer’s disease. J Alzheimers Dis. 2018;63:1337-46.
Corrêa-Oliveira R, Fachi JL, Vieira A, Sato FT, Vinolo MA. Regulation of immune cell function by short-chain fatty acids. Clin Transl Immunology. 2016;5:e73.
Franceschi C, Garagnani P, Parini P, Giuliani C, Santoro A. Inflammaging: a new immune–metabolic viewpoint for age-related diseases. Nat Rev Endocrinol. 2018;14:576-90.
Zhang Y, Geng R, Tu Q. Gut microbial involvement in Alzheimer’s disease pathogenesis. Aging (Albany NY). 2021;13:13359-71.
Jack CR, Knopman DS, Jagust WJ, Petersen RC, Weiner MW, Aisen PS, et al. Tracking pathophysiological processes in Alzheimer’s disease: an updated hypothetical model of dynamic biomarkers. Lancet Neurol. 2013;12:207-16.
Karim HT, Aizenstein HJ, Mizuno A, Ly M, Andreescu C, Wu M, et al. Independent replication of advanced brain age in mild cognitive impairment and dementia: detection of future cognitive dysfunction. Mol Psychiatry. 2022;27:5235-43.
Karim HT, Ly M, Yu G, Krafty R, Tudorascu DL, Aizenstein HJ, et al. Aging faster: worry and rumination in late life are associated with greater brain age. Neurobiol Aging. 2021;101:13-21.
Ly M, Yu GZ, Karim HT, Muppidi NR, Mizuno A, Klunk WE, et al. Improving brain age prediction models: incorporation of amyloid status in Alzheimer’s disease. Neurobiol Aging. 2020;87:44-8.
Yu GZ, Ly M, Karim HT, Muppidi N, Aizenstein HJ, Ibinson JW. Accelerated brain aging in chronic low back pain. Brain Res. 2021;1755:147263.
Saji N, Niida S, Murotani K, Hisada T, Tsuduki T, Sugimoto T, et al. Analysis of the relationship between the gut microbiome and dementia: a cross-sectional study conducted in Japan. Sci Rep. 2019;9:1008.
Vogt NM, Kerby RL, Dill-McFarland KA, Harding SJ, Merluzzi AP, Johnson SC, et al. Gut microbiome alterations in Alzheimer’s disease. Sci Rep 2017;7:13537.
Roh HW, Kim N-R, Lee D, Cheong JY, Seo SW, Choi SH, et al. Baseline clinical and biomarker characteristics of biobank innovations for chronic cerebrovascular disease with Alzheimer’s disease study: BICWALZS. Psychiatry Investig, 2022;19:100-9.
Molinuevo JL, Rabin LA, Amariglio R, Buckley R, Dubois B, Ellis KA, et al. Implementation of subjective cognitive decline criteria in research studies. Alzheimers Dement. 2017; 13: 296-311.
Morris JC. The Clinical Dementia Rating (CDR): current version and scoring rules. Neurology 1996;43:2412-4.
Winblad B, Palmer K, Kivipelto M, Jelic V, Fratiglioni L, Wahlund LO, et al. Mild cognitive impairment--beyond controversies, towards a consensus: report of the International Working Group on Mild Cognitive Impairment. J Intern Med. 2004;256:240-6.
McKhann GM, Knopman DS, Chertkow H, Hyman BT, Jack CR Jr, Kawas CH, et al. The diagnosis of dementia due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement. 2011;7:263-9.
American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, 5th ed. American Psychiatric Publishing, Inc; 2013.
Oh E, Kang Y, Shin JH, Yeon BK. A validity study of K-MMSE as a screening test for dementia : comparison against a comprehensive neuropsychological evaluation. Dement Neurocogn Disord. 2010;9:8-12.
Ryu HJ, Yang DW. The Seoul neuropsychological screening battery (SNSB) for comprehensive neuropsychological assessment. Dement Neurocogn Disord, 2023;22:1-15.
Hall M, Beiko RG. 16S rRNA Gene Analysis with QIIME2. Methods Mol Biol. 2018; 1849: 113-129.
Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P, et al. The SILVA ribosomal RNA gene database project: improved data processing and web-based tools. Nucleic Acids Res. 2013;41:D590-6.
Ly M, Yu GZ, Karim HT, Muppidi NR, Mizuno A, Klunk WE, et al. Improving brain age prediction models: incorporation of amyloid status in Alzheimer’s disease. Neurobiol Aging. 2020;87:44-8.
Schrouff J, Rosa MJ, Rondina JM, Marquand AF, Chu C, Ashburner J, et al. PRoNTo: Pattern recognition for neuroimaging toolbox. Neuroinformatics. 2013;11:319-37.
LaConte S, Strother S, Cherkassky V, Anderson J, Hu X. Support vector machines for temporal classification of block design fMRI data. Neuroimage. 2005;26:317-29.
Karim HT, Ly M, Yu G, Krafty R, Tudorascu DL, Aizenstein HJ, et al. Aging faster: worry and rumination in late life are associated with greater brain age. Neurobiol Aging. 2021;101:13-21.
Karim HT, Aizenstein HJ, Mizuno A, Ly M, Andreescu C, Wu M, et al. Independent replication of advanced brain age in mild cognitive impairment and dementia: detection of future cognitive dysfunction. Mol Psychiatry. 2022;27:5235-43.
Holmes C, Cunningham C, Zotova E, Woolford J, Dean C, Kerr S, et al. Systemic inflammation and disease progression in Alzheimer disease. Neurology. 2008;73:768-74.
Brown MB. 400: A method for combining non-independent, one-sided tests of significance. Biometrics. 1975;31:987-92.
Cinar O, Viechtbauer W. The poolr package for combining independent and dependent p values. J Stat Softw. 2022;101:1-42.
Fisher RA. Statistical Methods for Research Workers, 4th ed. Edinburgh: Oliver and Boyd;1932.
Rosseel Y. lavaan: An R Package for Structural Equation Modeling. J Stat Softw. 2012;48:1-36.
Hayes AF. PROCESS: A versatile computational tool for observed variable mediation, moderation, and conditional process modeling. [White paper]. Retrieved from http://www.afhayes.com/ public/process2012.pdf.
Vandeputte D, Kathagen G, D’Hoe K, Vieira-Silva S, Valles-Colomer M, Sabino J, et al. Quantitative microbiome profiling links gut community variation to microbial load. Nature. 2017;551:507-11.
El Aidy S, Stilling R, Dinan TG, Cryan JF. Microbiome to brain: Unravelling the multidirectional axes of communication. Adv Exp Med Biol. 2016;874:301-36.
Bonaz B, Bazin T, Pellissier S. The vagus nerve at the interface of the microbiota-gut-brain axis. Front Neurosci. 2018;12:49.
Łuc M, Misiak B, Pawłowski M, Stańczykiewicz B, Zabłocka A, Szcześniak D, et al. Gut microbiota in dementia. Critical review of novel findings and their potential application. Prog Neuropsychopharmacol Biol Psychiatry. 2021;104:110039.
Simpson CA, Mu A, Haslam N, Schwartz OS, Simmons JG. Feeling down? A systematic review of the gut microbiota in anxiety/depression and irritable bowel syndrome. J Affect Disord, 2020;266:429-46.
Liu P, Wu L, Peng G, Han Y, Tang R, Ge J, et al. Altered microbiomes distinguish Alzheimer’s disease from amnestic mild cognitive impairment and health in a Chinese cohort. Brain Behav Immun. 2019;80:633-43.
Vogt NM, Romano KA, Darst BF, Engelman CD, Johnson SC, Carlsson CM, et al. The gut microbiota-derived metabolite trimethylamine N-oxide is elevated in Alzheimer's disease. Alzheimers Res Ther. 2018;10:124.
Hill C, Guarner F, Reid G, Gibson GR, Merenstein DJ, Pot B, et al. Expert consensus document: The international scientific association for probiotics and prebiotics consensus statement on the scope and appropriate use of the term probiotic. Nat Rev Gastroenterol Hepatol. 2014;11:506-14.

No competing interests reported.

Supplementsubmit.docx

Brain age mediates gut microbiome dysbiosis-related cognition in older adults

Status:

Version 1

Abstract

Figures

BACKGROUND

METHODS

Participants

Clinical and biological assessment

Microbiome

Magnetic resonance imaging (MRI) analysis and brain age estimation

Statistical analysis

RESULTS

DISCUSSION

CONCLUSIONS

ABBREVIATIONS

DECLARATIONS

References

Additional Declarations

Supplementary Files

Status:

Version 1