Relationship Between Plasma Neurofilament Light Chain,Gut Microbiota,and Dementia: A Cross-Sectional Study

Abstract

Background:

Previous studies have demonstrated associations between gut microbiota, microbial metabolites, and cognitive decline. However, relationships between these factors and neurofilament light chain (NfL; a disease-nonspecific biomarker of neural damage) remain controversial.

Objective:

To evaluate the associations between plasma NfL, gut microbiota, and cognitive function.

Methods:

We performed a cross-sectional sub-analysis of data from our prospective cohort study that was designed to investigate the relationship between gut microbiota and cognitive function. Patients who visited our memory clinic were enrolled and demographics, dementia-related risk factors, cognitive function, brain imaging, gut microbiomes, and microbial metabolites were assessed. We evaluated the relationships between the gut microbiome, microbial metabolites, and plasma NfL. Moreover, the relationships between plasma NfL and cognitive function were assessed using multivariable logistic regression analyses.

Results:

We analyzed 128 participants (women: 59%, mean age: 74 years). Participants with high (above the median) plasma NfL concentrations tended to be older, women, and hypertensive and have a history of stroke, chronic kidney disease, and dementia. Plasma NfL was also associated with cerebral small vessel disease. However, plasma NfL levels were not significantly correlated with gut microbial metabolites. Multivariable analyses revealed that a higher plasma NfL concentration was independently associated with the presence of dementia (odds ratio: 9.94, 95% confidence interval: 2.75–48.2, p < 0.001).

Conclusion:

High plasma NfL concentration was independently associated with the presence of dementia as previously reported. However, plasma NfL levels were not significantly correlated with gut microbial metabolites in this preliminary study.

Keywords

Biomarkers cerebral small vessel disease cognitive decline dementia gut microbiota

INTRODUCTION

Dementia is a significant global healthcare problem; in 2015, 47 million people worldwide were living with dementia [1]. The number of people with dementia is expected to increase to 66 million by 2030 and 131 million by 2050 [2]. Additionally, the global cost of dementia was estimated at almost US$820 billion in 2015, and this figure has continued to increase [1, 2]. Japan is also facing financial pressure on its healthcare system because of the increasing proportion of people with dementia [3]. Thus, a comprehensive strategy for dementia research has been introduced in Japan to improve the healthcare system [3].

Recently, associations between gut microbiota and cognitive decline have gained increasing attention [4 –6] because these associations may reveal the mechanisms underlying the onset of dementia. Previous studies have suggested that disruption of the neuroinflammatory system [7], vascular inflammation [8], and remote relationships driven by various metabolites [9] are mechanisms underlying cognitive decline caused by the gut microbiome. Furthermore, Cryan et al. summarized multiple direct (e.g., vagus nerve) and indirect (e.g., short-chain fatty acids, cytokines, and key dietary amino acids, such as tryptophan) gut–brain pathways and proposed that the gut microbiota modulates the gut–brain axis via these pathways [6]. Specifically, microbes locally synthesize neurotransmitters in the gut, which is an important mode of communication. Neuroactive bacterial metabolites and metabolites from the diet can modulate the brain and behavior, which include the influence of epithelial cells on gut-barrier function, enteroendocrine cells on hormone release, and dendritic cells on immune and microglial function, all of which play fundamental roles in aging and neurological disorders [6]. However, knowledge regarding the effects of gut microbiota on cognitive function remains limited. Examining blood biomarkers that indicate neural damage and analyzing the links between such biomarkers, gut microbiota, and cognitive function may fill this knowledge gap.

Neurofilament light chain (NfL) is an emerging, non-invasive, disease-non-specific marker of neurodegeneration and neuronal injury [10]. Neurofilaments are intermediate filament proteins of the cytoskeleton that are integral to the structure and function of axons. NfL is one of the subunits of neurofilaments. During neurodegeneration, neurons break down, and this neural breakdown releases NfL into the blood and cerebrospinal fluid (CSF) [11]. Previous studies have identified relationships between NfL and several neurological disorders, such as dementia [10, 11] and cerebral small vessel disease (SVD) [12]. NfL is associated with poorer cognitive performance across most neurological disorders and cognitive domains [11] and also reflects the severity and predicts the progression of cerebral SVD [12]. Additionally, Vogt et al. revealed that gut microbial-derived metabolite trimethylamine N-oxide in the CSF of people with Alzheimer’s disease (AD) is positively correlated with CSF biomarkers of neuronal degeneration indicated by tau and NfL, which suggests the presence of the gut–brain axis [13]. However, associations between NfL in blood samples, gut microbiota, and cognitive function have not been comprehensively investigated. Such analysis will clarify the link between gut microbiota and damage to the brain parenchyma, such as that observed in patients with cerebral SVD or brain atrophy.

We have recently conducted an observational study that was originally designed to investigate the relationship between gut microbiota and cognitive function. In this study, gut microbial dysregulation was revealed to be associated with cognitive decline [14, 15], vascular risk factors [16], and cerebral SVD, such as white matter hyperintensity [17]. These findings suggest links among gut microbiota, vascular risk factors, cerebral SVD, and cognitive decline. Furthermore, gut microbial metabolites and dietary components play important roles in these associations [18]. More specifically, gut microbial metabolites are associated with cognitive decline, independent of the gut microbiome [18]. Furthermore, adherence to a Japanese-style diet is associated with gut microbial metabolites and is inversely associated with cognitive decline [19, 20]. Thus, these potential factors may also affect the diet–microbiome–dementia cascade.

The present study aimed to evaluate the relationships between NfL, gut microbiota, and cognitive function by performing a sub-analysis of the data from our ongoing clinical study. We hypothesized that plasma NfL would be associated with cognitive decline in our cohort data and that there would be an association between plasma NfL, the gut microbiome, and microbial metabolites.

MATERIALS AND METHODS

Study design

We performed a cross-sectional sub-analysis of data from a hospital-based prospective cohort study, the Gerontological Investigation of Microbiome: a Longitudinal Estimation Study (the Gimlet study), conducted at the National Center for Geriatrics and Gerontology (NCGG) in Japan [14]. Briefly, patients who visited the Memory Clinic at the NCGG and agreed to undergo both a medical assessment of their cognitive function and a fecal examination were enrolled. Activities of daily living and cognitive functions of participants were assessed annually after enrollment. This study was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the Institutional Review Board of the NCGG (no. 1191-3). Written informed consent was obtained from all patients and their families before their participation in the study. The Gimlet study is ongoing and is registered on the UMIN Clinical Trials Registry (UMIN000031851). Detailed information regarding the Gimlet study and the present sub-analysis is provided in the Supplementary Material.

Participants

We enrolled consecutive patients who had subjective memory impairment and visited the Memory Clinic at the NCGG between March 2016 and March 2017. Participants in the Gimlet study were eligible for this sub-study if they met the following criteria: 1) able to undergo brain magnetic resonance imaging (MRI); 2) blood samples were successfully preserved; and 3) provided written informed consent. Patients who had potential confounding factors or effect modifiers for the variables of interest (e.g., recent use of antibiotics) were excluded at the time of enrollment in the Gimlet study.

Assessments

All participants underwent a comprehensive geriatric assessment based on the following features: 1) demographic characteristics; 2) risk factors, such as hypertension, dyslipidemia, diabetes mellitus, ischemic heart disease, chronic kidney disease, and a history of stroke; 3) activities of daily living; 4) global cognitive function, which was assessed using the Mini-Mental State Examination (MMSE) and Clinical Dementia Rating (CDR) scales; 5) neuropsychological testing using evaluations such as the Alzheimer’s Disease Assessment Scale-Cognitive Subscale and Frontal Assessment Battery; 6) behavioral and psychological symptoms; 7) burden on caregivers; 8) depression status, which was assessed using the Geriatric Depression Scale; 9) laboratory parameters, such as apolipoprotein E ɛ4 and C-reactive protein; 10) ankle brachial index and pulse wave velocity as indicators of arteriosclerosis [21] and the ‘impact’ of pulse [22], respectively; 11) results of brain imaging, such as MRI and single-photon emission-computed tomography; and 12) dietary assessments. Clinical samples and data were provided by the NCGG Biobank, which collects clinical data for research. In this sub-analysis, we used a single dataset that was obtained at enrollment in the Gimlet study.

Brain imaging

Participants underwent a 1.5-T brain MRI (Philips Ingenia, Eindhoven, Netherlands). The presence of cerebral SVD and its components, such as silent lacunar infarcts, white matter hyperintensity, cerebral microbleeds, and enlarged periventricular space, were assessed. The voxel-based specific regional analysis system for Alzheimer’s Disease (VSRAD) software (Eisai Co., Ltd., Tokyo, Japan) was used to quantify cortical and hippocampal atrophy. A high VSRAD score indicated the presence of AD [23]. Participants also underwent N-isopropyl-p-[¹²³I]-iodoamphetamine single-photon emission-computed tomography, in which low blood flow in the area of the posterior cingulate gyrus and/or precuneus was regarded as a surrogate marker of AD [24].

Dietary assessments

We evaluated the Japanese Diet Index (JDI₁₂), which is a questionnaire on dietary composition that consists of 12 items: 10 beneficial components (rice, miso, fish and shellfish, green and yellow vegetables, seaweed, pickles, green tea, soybeans and soybean-derived foods, fruit, and mushrooms) and two less beneficial components (beef and pork, and coffee) [19, 20]. Participants were assigned one point if their daily beneficial intake was equal to or greater than the sex-specific median intake and their daily less beneficial intake was below the sex-specific median intake [19, 20].

Classification of cognitive function

Dementia was defined as an MMSE score < 20 and/or a CDR score≥1 [14]. Participants who did not have dementia were further categorized as having either mild cognitive impairment (MCI) or normal cognition (NC). MCI was defined as an MMSE score≥20 and a CDR score = 0.5; this implies possible, very mild dementia, and suggests a higher risk of dementia [15]. NC was defined as an MMSE score≥20 and a CDR score = 0.

Plasma neurofilament light chain

Blood samples were collected, processed onsite to isolate plasma, aliquoted, and frozen at –81°C in the NCGG Biobank. Plasma NfL concentrations were measured according to manufacturer instructions using the NF-Light Advantage Kit on a highly sensitive single-molecule array assay Simoa HD-1 platform (Quanterix, Lexington, MA, USA) [25, 26]. Plasma samples were measured at a 1 : 4 dilution and were run in duplicate by a trained technician. All samples were measured blinded. All NfL values were within the linear ranges of the assays.

Gut microbiome

Fecal samples were collected, frozen, and stored at –81°C in the NCGG Biobank. The gut microbiome of each participant was analyzed by Techno Suruga Laboratory (Shizuoka, Japan) using terminal restriction fragment length polymorphism (T-RFLP) analysis [27]. The T-RFLP analysis was used to classify gut microbes. By referencing the Human Fecal Microbiome T-RFLP profile, each gut microbiome was categorized as representing one of three enterotypes: enterotype I, which included Bacteroides at > 30%; enterotype II, which included Prevotella at > 15%; and enterotype III, which comprised other combinations of microorganisms. The Firmicutes/Bacteroidetes (F/B) ratio was also calculated because a high ratio is indicative of dysbiosis [28].

Analysis of microbial metabolites in feces

The fecal concentrations of organic acids, short-chain fatty acids, ammonium ions, indoles, phenol, skatole, and p-cresol were measured, as previously described [18]. The concentrations of organic acids and short-chain fatty acids (acetic, propionic, butyric, iso-butyric, succinic, lactic, formic, valeric, and iso-valeric acids) were measured using high-performance liquid chromatography. Ammonium ion concentrations were quantified using ion chromatography, and fecal concentrations of indoles, phenol, skatole, and p-cresol were quantified using gas chromatography/mass spectrometry.

Statistical analysis

Continuous, ordinal, and categorical variables are expressed as means±standard deviations, medians and interquartile ranges, or frequencies and proportions (percentage), respectively. Data were compared using Student’s unpaired t-test, Wilcoxon rank-sum test, and χ² test, respectively. Natural logarithm-transformed plasma NfL (Ln-NfL) was occasionally used to reduce the right skewness of plasma NfL concentrations for subsequent analysis. First, we compared clinical characteristics between participants with high concentrations of plasma NfL and those with low concentrations, who were divided according to the median value of plasma NfL. Second, relationships between the gut microbiome, microbial metabolites, and plasma NfL concentrations were evaluated. Furthermore, we compared the differences in dietary composition between participants with high concentrations of plasma NfL and those with low concentrations. Third, participants were allocated to groups according to the presence or absence of dementia, the presence or absence of enterotype I, and MCI or NC status (among those who did not have dementia), to compare plasma NfL concentrations. Additionally, participants were allocated to four subgroups according to high or low plasma NfL concentrations divided by the median value and the presence or absence of dementia-related items. Dementia-related items consisted of items, such as age, sex, years of education, and risk factors. The presence of dementia-related items was categorized using the median value of the continuous variables and was based on our previous findings (i.e., being a woman, higher age, higher VSRAD score, higher pulse wave velocity, higher F/B ratio, fewer years of education, lower Mini-Nutritional Assessment-Short Form score, lower ankle brachial index, lower JDI₁₂ score, and having categorical items). Finally, multivariable logistic regression model was used to identify independent associations between the presence of dementia and plasma NfL concentrations. Backward stepwise multivariable logistic regression analyses were performed adjusting for age, sex, years of education, risk factors, enterotype I, F/B ratio, and brain imaging. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated. All comparisons were two-tailed, and p < 0.05 represented statistical significance. Data were analyzed using the JMP 14.3 software package (SAS Institute Inc., Cary, NC, USA) and R software (http://www.r-project.org).

RESULTS

Participant characteristics

A total of 128 participants were enrolled in this study, as previously reported (women: 59%, mean age: 74.2±8.7 years). The levels of cognitive function and enterotypes were as follows: 34 participants (26.6%) were classified as having dementia, 71 (55.5%) as having MCI, and 23 (18.0%) as having NC; 47 (36.7%) participants were classified as having enterotype I, 5 (4%) as having enterotype II, and 76 (59.4%) as having enterotype III.

Clinical data and plasma NfL

Compared with participants with low concentrations of plasma NfL, those with high concentrations tended to be older (high versus low: 77 versus 73 years, p < 0.001), women (69.8% versus 46.9%, p = 0.012), and hypertensive (74.6% versus 50.0%, p = 0.006) and have a history of stroke (14.3% versus 3.1%, p = 0.030), chronic kidney disease (44.4% versus 20.3%, p = 0.005), and dementia (47.6% versus 4.7%, p < 0.001). They also tended to have cerebral SVD, higher scores in the VSRAD (median score: 1.6 versus 0.8, p < 0.001), and reduced cognitive function (Tables 1 and 2 and Supplementary Table 1).

Table 1

Comparisons of participant background information between participants with high neurofilament light chain (NfL) concentrations and those with low NfL concentrations

	High NfL (n = 63)	Low NfL (n = 64)	p
Demographic factor
Age, y	77, 75–82	73, 66.3–78	< 0.0001
Women, n (%)	44 (69.8)	30 (46.9)	0.012
Education, y	12, 9–12	12, 12–14	0.0005
Risk factor
Hypertension, n (%)	47 (74.6)	32 (50.0)	0.006
Diabetes mellitus, n (%)	14 (22.2)	6 (9.4)	0.054
Dyslipidemia, n (%)	30 (47.6)	29 (45.3)	0.860
IHD, n (%)	7 (11.1)	5 (7.8)	0.560
Stroke, n (%)	9 (14.3)	2 (3.1)	0.030
CKD, n (%)	28 (44.4)	13 (20.3)	0.005
Smoking habit, n (%)	9 (14.3)	23 (35.9)	0.007
Alcohol consumption, n (%)	18 (28.6)	31 (48.4)	0.029
Apolipoprotein E ɛ4 carrier, n (%)	22 (34.9)	16 (25.0)	0.249
Dementia, n (%)	30 (47.6)	3 (4.7)	< 0.0001
Comprehensive geriatric assessment
Barthel Index	100, 95–100	100, 100–100	0.009
IADL impairment, n (%)	41 (65.1)	17 (26.6)	< 0.0001
DBDS	12, 5–18.3	6, 3–11.8	0.0002
GDS	3, 1–5	2, 1–5	0.578
ZBI	15, 6–24	7, 3–19	0.007
MNA-SF	12, 11–13	13, 12–14	0.015
Arterial stiffness
Ankle brachial index	1.11, 1.04–1.15	1.08, 1.06–1.15	0.842
Pulse wave velocity, m/s	18.0, 16.1–20.5	19.3, 15.8–23.3	0.275
Laboratory finding
eGFR, mL/min/1.73 m²	60.9, 50.2–72.3	72.4, 61.3–84.8	< 0.0001

Data are presented as medians, interquartile ranges, or number of patients (%). Wilcoxon rank-sum test and χ² test were used. CKD, chronic kidney disease; DBDS, Dementia Behavior Disturbance Scale; eGFR, estimated glomerular filtration rate; GDS, Geriatric Depression Scale; IADL, instrumental activities of daily living; IHD, ischemic heart disease; MNA-SF, Mini-Nutritional Assessment-Short Form; ZBI, Zarit Caregiver Burden Interview.

Table 2

Comparisons of participant characteristics between participants with high neurofilament light chain (NfL) concentrations and those with low NfL concentrations

	High NfL (n = 63)	Low NfL (n = 64)	p
Cognitive function
MMSE	22, 18–24	27.5, 24–29	< 0.0001
CDR-GB			< 0.0001
0, n (%)	2 (3.2)	21 (32.8)
0.5, n (%)	43 (68.3)	42 (65.6)
≥1, n (%)	18 (28.6)	1 (1.6)
CDR-SB	3.5, 2–4.5	0.5, 0.5–1.9	< 0.0001
ADAS-cog	13.2, 8–17.3	6.3, 4.7–10	< 0.0001
RCPM	24.5, 20.8–28	31, 28–33.5	< 0.0001
FAB	10, 8–11	13, 10–14	< 0.0001
LM-WMSR I	4, 2–9	12, 7–21	< 0.0001
LM-WMSR II	0, 0–3	7, 1–14	< 0.0001
Brain MRI finding
SLI, n (%)	11 (17.5)	2 (3.1)	0.009
WMH, n (%)	22 (34.9)	12 (18.8)	0.047
CMB, n (%)	20 (31.8)	8 (12.5)	0.011
EPVS, n (%)	43 (68.3)	53 (82.8)	0.065
VSRAD	1.6, 0.9–2.3	0.8, 0.5–1.2	< 0.0001
SPECT
Low blood flow^*	44 (73.3)	41 (68.3)	0.688

Data are presented as medians, interquartile ranges, or number of patients (%). Wilcoxon rank-sum test and χ² test were used. ^*Blood flow reduction in the posterior cingulate gyrus and/or precuneus, as seen in SPECT images. ADAS-cog, Alzheimer’s Disease Assessment Scale-Cognitive Subscale; BP, blood pressure; CDR-GB, Clinical Dementia Rating Global Score; CDR-SB, Clinical Dementia Rating-Sum of Boxes; CMB, cerebral microbleed; EPVS, enlarged periventricular space; FAB, Frontal Assessment Battery; LM-WMSR, Logical Memory subtests I and II of the Wechsler Memory Scale-Revised; MMSE, Mini-Mental State Examination; MRI, magnetic resonance imaging; RCPM, Raven’s Coloured Progressive Matrices; SLI, silent lacunar infarct; SPECT, single-photon emission-computed tomography; VSRAD, voxel-based specific regional analysis system for Alzheimer’s disease; WMH, white matter hyperintensity.

Plasma NfL concentrations were significantly higher in participants with cognitive decline and those with cerebral SVD (Fig. 1). Higher levels of Ln-NfL were significantly associated with greater cognitive decline, as indicated by MMSE, CDR-Sum of Boxes (CDR-SB), Alzheimer’s Disease Assessment Scale-Cognitive Subscale, and Frontal Assessment Battery scores (Fig. 2). Higher levels of Ln-NfL were also associated with increased VSRAD and decreased JDI₁₂ scores (Fig. 3). Cognitive function, VSRAD scores, and JDI₁₂ scores were significantly associated with Ln-NfL levels in both the unadjusted analyses and analyses adjusted for age (Supplementary Table 2). However, there were no significant correlations between levels of Ln-NfL and gut microbial metabolites.

Fig. 1

Comparisons of plasma neurofilament light chain (NfL) concentrations and (A) cognitive function or (B) total cerebral small vessel disease (SVD) scores. The x-axes show: (A) cognitive function stratified into normal cognition (NC), mild cognitive impairment (MCI), and dementia groups; (B) total cerebral SVD scores stratified as scores of 0, 1–2, or 3–4. The y-axes show plasma NfL concentrations. The plasma NfL concentrations increased with increasing degrees of cognitive decline and total cerebral SVD scores (Kruskal-Wallis test). Post hoc analysis performed using Dunnett’s test. ^*p < 0.05; ^**p < 0.0001.

Fig. 2

Comparisons of plasma neurofilament light chain (NfL) concentrations and cognitive function. The x-axes show: (A) Mini-Mental State Examination (MMSE) scores; (B) Clinical Dementia Rating-Sum of Boxes (CDR-SB) scores; (C) Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-cog) scores; (D) Frontal Assessment Battery (FAB) scores. The y-axes show the levels of the natural logarithm-transformed plasma NfL (Ln-NfL). The levels of Ln-NfL increased with increasing degrees of cognitive decline (single regression analysis).

Fig. 3

Comparisons of plasma neurofilament light chain (NfL) concentrations and clinical biomarkers. The x-axes show: (A) voxel-based specific regional analysis system for Alzheimer’s disease (VSRAD); (B) Firmicutes/Bacteroidetes ratio (F/B ratio); (C) C-reactive protein (CRP); (D) Japanese Diet Index (JDI₁₂). The y-axes show the levels of the natural logarithm-transformed plasma NfL (Ln-NfL). The levels of Ln-NfL increased with increasing VSRAD scores, which suggested more cortical and hippocampal atrophy. They also decreased with increasing adherence to a traditional Japanese diet (single regression analysis).

Cutoff values of plasma NfL

The plasma NfL cutoff value for the detection of dementia was 24.2 pg/mL, with 91% sensitivity and 65% specificity (area under the receiver operating characteristic curve [AUC] = 0.80, p < 0.0001). In participants without dementia, the plasma NfL cutoff value for the detection of MCI was 21.0 pg/mL, with 64% sensitivity and 83% specificity (AUC = 0.74, p < 0.0001).

Dietary assessments and NfL

Compared with participants with high plasma NfL concentrations, those with low concentrations tended to consume more fish and shellfish (71.7% versus 42.1%, p = 0.008), mushrooms (65.2% versus 39.5%, p = 0.028), fruit (89.1% versus 73.7%, p = 0.088), soybeans and soybean-derived foods (93.5% versus 71.1%, p = 0.008), and coffee (80.4% versus 60.5%, p = 0.055), although some of these differences were not significant (Supplementary Table 3). Participants with low plasma NfL concentrations also tended to have higher JDI₁₂ scores than participants with high plasma NfL concentrations (median score: 8 versus 7, p = 0.036).

Gut microbiota, metabolites, and NfL

There were no significant correlations between levels of Ln-NfL and concentrations of gut microbiome metabolites (Supplementary Figure 1). Regarding the enterotypes, there were no significant differences between participants with high and low plasma NfL concentrations (Supplementary Table 4).

Plasma NfL was significantly different between participants with and without dementia, and between participants with MCI and those with NC among participants without dementia (Supplementary Table 5). F/B ratios were significantly different between participants with and without dementia, and between participants with enterotype I and those with other enterotypes (Supplementary Table 5). However, serum C-reactive protein was not significantly different between any of the groups.

Subgroup analyses stratified by plasma NfL and dementia-related items

There were significant correlations among the four groups stratified by plasma NfL and dementia-related items (Supplementary Table 6). Those with both high plasma NfL and dementia-related items, such as fewer years of education and higher VSRAD score, tended to have a higher prevalence of dementia compared with those who had no dementia-related items. Regarding the gut microbiome, having both high plasma NfL and high F/B ratio also tended to be associated with the presence of dementia (Supplementary Table 6). Additionally, those who had high plasma NfL who presented with enterotype III tended to be more associated with higher prevalence of dementia compared with that of those who presented with enterotype I (prevalence of dementia: 63.2% versus 18.2%; Supplementary Table 6).

Compared with the AUC of NfL alone for the detection of dementia (AUC [95% CI]: 0.81 [0.73–0.88]), those of NfL+F/B ratio, NfL+enterotype I, and NfL+enterotype III were equal or slightly higher (0.81 [0.74–0.89], 0.83 [0.74–0.91], and 0.84 [0.75–0.92], respectively); however, there were no significant differences (Supplementary Figure 2).

Multivariable analyses

Multivariable analyses revealed that plasma NfL was independently associated with the presence of dementia. Specifically, compared with participants with low plasma NfL concentrations, those with high concentrations had a higher OR (OR: 9.94, 95% CI: 2.75–48.2, p < 0.001) calculated in the stepwise multivariable logistic regression analyses after adjusting for other confounding covariables (Table 3 and Supplementary Table 7). Similarly, the stepwise multivariable logistic regression analysis revealed that higher plasma NfL (categorized as a continuous variable) was more likely than lower plasma NfL to be associated with the presence of dementia; however, this result was not significant (per 10 pg/mL, OR: 1.26, 95% CI: 0.94–1.79; p = 0.139; Supplementary Table 8).

Table 3

Multivariable logistic regression analyses for the presence of dementia

	OR	95% CI	p
High concentration of plasma NfL^*	9.94	2.75–48.2	< 0.001
Enterotype I	0.11	0.03–0.39	< 0.001
Alcohol consumption	0.47	0.13–1.51	0.206
CMBs	3.22	0.98–11.0	0.054
VSRAD	2.48	1.26–5.58	0.007

The dependent variable was the presence of dementia. ^*Participants were allocated to two groups according to their plasma NfL concentrations: high NfL group if the NfL value was above the median; low NfL group if the NfL value was below the median. Model: backward stepwise multivariable logistic regression analyses adjusted for age, sex, years of education, and risk factors, such as hypertension, dyslipidemia, diabetes mellitus, ischemic heart disease, chronic kidney disease, smoking, alcohol consumption, apolipoprotein E, enterotype I, F/B ratio, silent lacunar infarct, white matter hyperintensity, cerebral microbleeds, enlarged perivascular space, VSRAD score, single-photon emission-computed tomography findings (presence or absence of low blood flow in the posterior cingulate gyrus and/or precuneus). CI, confidence interval; CMBs, cerebral microbleeds; NfL: neurofilament light chain; OR, odds ratio; VSRAD; voxel-based specific regional analysis system for Alzheimer’s disease.

DISCUSSION

The main finding of the present study was that plasma NfL was independently and significantly associated with the presence of dementia, as previously reported. Moreover, plasma NfL was associated with dementia-related items, such as cognitive function, total SVD score, and JDI₁₂ score. Furthermore, those with high plasma NfL and dementia-related items tended to have a higher prevalence of dementia compared with those with no dementia-related items. Regarding the gut microbiome, high plasma NfL presenting with either enterotype III or high F/B ratio also tended to be associated with the presence of dementia. However, there were no significant correlations between the levels of NfL and gut microbial metabolites.

NfL is an intermediate filament in the axonal cytoskeleton, and NfL changes in biological fluids, such as blood and CSF, have been proposed as a biomarker for brain damage and atrophy in multiple neurological disorders, including neurodegenerative diseases [25, 26]. Both amyloid-β and tau are considered robust biomarkers of dementia, such as AD [29]. In addition, recent studies have proposed the A/T/N system, which comprises three biomarkers (Amyloid-β, Tau, and Neurodegeneration or neural injury) to describe and categorize multidomain biomarkers findings [29]. Both amyloid-β and tau are more specific indicators of AD than is NfL an indicator of neurodegeneration. Therefore, we speculated that, of these three biomarkers, NfL is useful for estimating disease-non-specific neural injury or neurodegeneration caused by the gut microbiome and/or microbial metabolites. In our dataset, plasma NfL was related to cognitive impairment and was associated with several clinical biomarkers, such as VSRAD and cerebral SVD scores. The cutoff value of plasma NfL to detect the presence of dementia in the present study (24.2 pg/mL) was similar to that of a previous study to detect the presence of AD (25.7 pg/mL) [30]. Thus, these associations are reasonable and consistent with the results of previous studies [10 –12].

In our previous study, the F/B ratio was associated with the presence of dementia. Specifically, the F/B ratio was higher in participants with dementia compared with those without dementia, which indicated that advanced dysbiosis may be related to the onset of dementia [14]. However, both the F/B ratio and gut microbiome, which we categorized according to enterotypes, did not significantly differ between different plasma NfL concentrations. This negative finding may be a result of our methodological limitations regarding the T-RFLP analysis and the F/B ratio. The use of next-generation sequencing can reveal more detailed microbial data than can the T-RFLP method. Furthermore, several biomarkers, other than the F/B ratio, such as lipopolysaccharides [31] and trimethylamine N-oxide [13], are attributed to the gut microbiome and may affect the brain. Lipopolysaccharides are the major outer membrane components of gram-negative bacteria and are capable of triggering systemic inflammation and the release of pro-inflammatory cytokines [31]. Trimethylamine N-oxide is a small molecule produced by the meta-organismal metabolism of dietary choline and has been implicated in human disease pathogenesis, including as a risk factor for AD [13]. Analyzing these biomarkers in addition to NfL may reveal more precise associations between the gut microbiome, microbial metabolites, blood biomarkers, and cognitive function. Furthermore, these associations may also clarify the presence of the gut–microbiome–brain axis.

We found that plasma NfL concentrations were not significantly correlated with gut microbial metabolite concentrations. Previously, we reported that several gut microbial metabolites are associated with dementia [18], where several metabolites were positively associated with dementia, whereas others were inversely associated with dementia, which suggested a multiplex linkage among the gut microbiome, microbial metabolites, and cognitive function. Alkasir et al. reported the presence of a brain–gut link and proposed a novel strategy for the management of dementia by modulating microbiota [32]. Furthermore, Rosa et al. found a significant correlation between gastrointestinal disease and high CDR score, which indicated the presence of a brain–gut link [33]. A recent study reported that older individuals with subjective cognitive decline had a significant reduction in the genus Faecalibacterium, which suggested an alteration of anti-inflammatory gut bacteria [34]. Another study also demonstrated that some microbial metabolites, such as indole and propionic acid, promote brain health [35]; moreover, microbiota-derived short-chain fatty acids are critical mediators along the gut–brain axis and promote amyloid-β deposition [36]. These findings support the presence of a gut–microbiome–brain axis. However, our sub-analysis revealed no significant correlations between microbial metabolites and plasma NfL. Nevertheless, associations between metabolites and blood biomarkers, other than NfL, may exist, and further comprehensive analyses that include these factors will likely reveal the nature of the gut–microbiome–brain axis.

Another notable result of the present study was that fish and shellfish, mushrooms, fruits, soybeans and soybean-derived foods, and coffee may have protective effects on cognitive decline. We found that greater consumption of these foods was associated with lower concentrations of plasma NfL, which indicated that these foods may alter gut microbiota and suppress cognitive decline. A previous large-scale study also reported an association between gut microbiota and healthy dietary habits, which was related to favorable cardiometabolic markers [37]. In addition, we revealed that a healthier diet pattern indicated by JDI₁₂ was associated with lower concentrations of plasma NfL. This will be investigated in more detail in the future.

The present study has several strengths. First, we revealed novel relationships among gut microbiota, microbial metabolites, plasma NfL, and cognitive decline. Specifically, plasma NfL was strongly associated with cognitive decline. Second, we also revealed that NfL levels were not correlated with gut microbial metabolites. Although we did not observe a positive association, this is the first study to analyze such associations and thus, offers new research directions in this emerging field. Third, we confirmed the utility of plasma NfL for detecting the presence of cognitive decline as well as several biomarkers for cognitive function, such as VSRAD and cerebral SVD scores, which are consistent with previous studies. Finally, we systematically evaluated cognitive function using a comprehensive geriatric assessment and a range of neuropsychological tests. Our findings encourage further focus on the relationships between blood biomarkers and cognitive function through examinations of the gut microbiome.

The present study also has several limitations. We did not assess amyloid-β or tau in the present study because the Gimlet study did not perform CSF testing or positron emission tomography to identify these biomarkers. The small number of participants and large number of potential variables may have also statistically underpowered our study. However, because this was a sub-analysis, sample size was not calculated. Furthermore, selection bias may also exist because our sample was a single hospital-based cohort. Another limitation of our study is that the distribution of enterotypes was unbalanced; however, enterotypes were likely unbalanced according to the characteristics of the cohort. Moreover, the specific effects of each microbial metabolite on blood biomarkers have not yet been determined. Thus, a comprehensive assessment that includes other biomarkers, such as amyloid-β precursor protein and non-neurodegenerative inflammatory biomarkers, may be useful given that they are associated with the risk of cognitive impairment [38].

Although this sub-study was a preliminary analysis that included a small number of patients, we provided evidence for relationships between plasma NfL, gut microbiota, and cognitive function. Detailed assessments of these relationships should be conducted in future studies to determine the underlying mechanisms.

Conclusions

High plasma NfL concentration was independently and significantly associated with the presence of dementia and was also associated with cerebral SVD, as previously reported. However, plasma NfL levels were not significantly correlated with gut microbial metabolites in this preliminary study. Further studies are warranted to examine these associations in relation to the microbiome–gut–brain axis.

Footnotes

ACKNOWLEDGMENTS

This study was supported by a Grant-in-Aid for Scientific Research (C), JSPS KAKENHI (20k07861), grants from the Research Funding of Longevity Sciences from the NCGG (19–24), the NARO Bio-Oriented Technology Research Advancement Institution Project (Advanced Integration Research for Agriculture and Interdisciplinary Fields), the Danone Institute of Japan Foundation, the Honjo International Scholarship Foundation (to Dr. Saji), and the Japan Agency for Medical Research and Development (AMED) under Grant Number JP20dk0207042 (to Dr. Sato). We would like to thank Yukie Ohsaki, Maki Yamamoto, Hana Saito, Ayaka Suzuki, and Tomomi Sato (NCGG) for their technical and secretarial assistance. We also thank the BioBank and NCGG for quality control of the clinical samples and data. Finally, we thank Petar Milovanovic, PhD, Bronwen Gardner, PhD, and Sarina Iwabuchi, PhD from Edanz () for editing a draft of this manuscript.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Livingston

, Sommerlad

, Orgeta

, Costafreda

, Huntley

, Ames

, Ballard

, Banerjee

, Burns

, Cohen-Mansfield

, Cooper

, Fox

, Gitlin

, Howard

, Kales

, Larson

, Ritchie

, Rockwood

, Sampson

, Samus

, Schneider

, Selbæk

, Teri

, Mukadam

(2017) Dementia prevention, intervention, and care, Lancet 390, 2673–2734.

Prince

, Wimo

, Guerchet

, Ali

, Wu

, Prina

(2015) World Alzheimer Report 2015. The global impact of dementia: An analysis of prevalence, incidence, cost and trends. Alzheimer’s Disease International, London.

Saji

, Sakurai

, Suzuki

, Mizusawa

, Toba

, ORANGE investigators (2016) ORANGE’s challenge: Developing wide-ranging dementia research in Japan, Lancet Neurol 15, 661–662.

Vogt

, Kerby

, Dill-McFarland

, Harding

, Merluzzi

, Johnson

, Carlsson

, Asthana

, Zetterberg

, Blennow

, Bendlin

, Rey

(2017) Gut microbiome alterations in Alzheimer’s disease, Sci Rep 7, 13537.

Zhou

, Wang

, Quan

, Zhao

, Jia

(2021) Gut microbiota changes and their correlation with cognitive and neuropsychiatric symptoms in Alzheimer’s disease, J Alzheimers Dis 81, 583–595.

Cryan

, O’Riordan

, Sandhu

, Peterson

, Dinan

(2020) The gut microbiome in neurological disorders, Lancet Neurol 19, 179–194.

, Cao

, Chang

, Juang

(2017) Intestinal microbial dysbiosis aggravates the progression of Alzheimer’s disease in Drosophila, Nat Commun 8, 24.

Tang

, Kitai

, Hazen

(2017) Gut microbiota in cardiovascular health and disease, Circ Res 120, 1183–1196.

Wang

, Klipfell

, Bennett

, Koeth

, Levison

, Dugar

, Feldstein

, Britt

, Fu

, Chung

, Wu

, Schauer

, Smith

, Allayee

, Tang

, DiDonato

, Lusis

, Hazen

(2011) Gut flora metabolism of phosphatidylcholine promotes cardiovascular disease, Nature 472, 57–63.

10.

Moscoso

, Grothe

, Ashton

, Karikari

, Lantero Rodríguez

, Snellman

, Suárez-Calvet

, Blennow

, Zetterberg

, Schöll

, Alzheimer’s Disease Neuroimaging Initiative (2021) Longitudinal associations of blood phosphorylated tau181 and neurofilament light chain with neurodegeneration in Alzheimer disease, JAMA Neurol 78, 396–406.

11.

Ramani

, Berard

, Walker

LAS

(2021) The relationship between neurofilament light chain and cognition in neurological disorders: A scoping review, J Neurol Sci 420, 117229.

12.

, Tan

, Shen

, Li

, Cui

, Tan

, Dong

, Yu

, Alzheimer’s Disease Neuroimaging Initiative (2021) Association of plasma neurofilament light with small vessel disease burden in nondemented elderly: A longitudinal study, Stroke 52, 896–904.

13.

Vogt

, Romano

, Darst

, Engelman

, Johnson

, Carlsson

, Asthana

, Blennow

, Zetterberg

, Bendlin

, Rey

(2018) The gut microbiota-derived metabolite trimethylamine N-oxide is elevated in Alzheimer’s disease, Alzheimers Res Ther 10, 124.

14.

Saji

, Niida

, Murotani

, Hisada

, Tsuduki

, Sugimoto

, Kimura

, Toba

, Sakurai

(2019) Analysis of the relationship between the gut microbiome and dementia: A cross-sectional study conducted in Japan, Sci Rep 9, 1008.

15.

Saji

, Murotani

, Hisada

, Tsuduki

, Sugimoto

, Kimura

, Niida

, Toba

, Sakurai

(2019) The relationship between the gut microbiome and mild cognitive impairment in patients without dementia: A cross-sectional study conducted in Japan, Sci Rep 9, 19227.

16.

Saji

, Hisada

, Tsuduki

, Niida

, Toba

, Sakurai

(2019) Proportional changes in the gut microbiome: A risk factor for cardiovascular disease and dementia? Hypertens Res 42, 1090–1091.

17.

Saji

, Murotani

, Hisada

, Tsuduki

, Sugimoto

, Kimura

, Niida

, Toba

, Sakurai

(2021) The association between cerebral small vessel disease and the gut microbiome: A cross-sectional analysis, J Stroke Cerebrovasc Dis 30, 105568.

18.

Saji

, Murotani

, Hisada

, Kunihiro

, Tsuduki

, Sugimoto

, Kimura

, Niida

, Toba

, Sakurai

(2020) Relationship between dementia and gut microbiome-associated metabolites: A cross-sectional study in Japan, Sci Rep 10, 8088.

19.

Zhang

, Otsuka

, Tomata

, Shimokata

, Tange

, Tomida

, Nishita

, Matsuyama

, Tsuji

(2019) A cross-sectional study of the associations between the traditional Japanese diet and nutrient intakes: The NILS-LSA project, Nutr J 18, 43–10.

20.

Saji

, Tsuduki

, Murotani

, Hisada

, Sugimoto

, Kimura

, Niida

, Toba

, Sakurai

(2022) Relationships between the Japanese-style diet, gut microbiota, and dementia: A cross-sectional study, Nutrition 94, 111524.

21.

Saji

, Kimura

, Yagita

, Kawarai

, Shimizu

, Kita

(2015) Comparison of arteriosclerotic indicators in patients with ischemic stroke: Ankle-brachial index, brachial-ankle pulse wave velocity and cardio-ankle vascular index, Hypertens Res 38, 323–328.

22.

Saji

, Toba

, Sakurai

(2016) Cerebral small vessel disease and arterial stiffness: Tsunami effect in the brain? Pluse 3, 182–189.

23.

Matsuda

, Mizumura

, Nemoto

, Yamashita

, Imabayashi

, Sato

, Asada

(2012) Automatic voxel-based morphometry of structural MRI by SPM8 plus diffeomorphic anatomic registration through exponentiated lie algebra improves the diagnosis of probable Alzheimer disease, AJNR Am J Neuroradiol 33, 1109–1114.

24.

Ito

, Mori

, Fukuyama

, Ishii

, Washimi

, Asada

, Mori

, Meguro

, Kitamura

, Hanyu

, Nakano

, Matsuda

, Kuwabara

, Hashikawa

, Momose

, Uchida

, Hatazawa

, Minoshima

, Kosaka

, Yamada

, Yonekura

, J-COSMIC Study Group (2013) Prediction of outcomes in MCI with (123)I-IMP-CBF SPECT: A multicenter prospective cohort study, Ann Nucl Med 27, 898–906.

25.

Rohrer

, Woollacott

, Dick

, Brotherhood

, Gordon

, Fellows

, Toombs

, Druyeh

, Cardoso

, Ourselin

, Nicholas

, Norgren

, Mead

, Andreasson

, Blennow

, Schott

, Fox

, Warren

, Zetterberg

(2016) Serum neurofilament light chain protein is a measure of disease intensity in frontotemporal dementia, Neurology 87, 1329–1336.

26.

Preische

, Schultz

, Apel

, Kuhle

, Kaeser

, Barro

, Gräber

, Kuder-Buletta

, LaFougere

, Laske

, Vöglein

, Levin

, Masters

, Martins

, Schofield

, Rossor

, Graff-Radford

, Salloway

, Ghetti

, Ringman

, Noble

, Chhatwal

, Goate

, Benzinger

TLS

, Morris

, Bateman

, Wang

, Fagan

, McDade

, Gordon

, Jucker

, Dominantly Inherited Alzheimer Network (2019) Serum neurofilament dynamics predicts neurodegeneration and clinical progression in presymptomatic Alzheimer’s disease, Nat Med 2, 277–283.

27.

Osborn

, Moore

, Timmis

(2000) An evaluation of terminal-restriction fragment length polymorphism (T-RFLP) analysis for the study of microbial community structure and dynamics, Environ Microbiol 2, 39–50.

28.

Spychala

, Venna

, Jandzinski

, Doran

, Durgan

, Ganesh

, Ajami

, Putluri

, Graf

, Bryan

, McCullough

(2018) Age-related changes in the gut microbiota influence systemic inflammation and stroke outcome, Ann Neurol 84, 23–36.

29.

Jack

Jr , Bennett

, Blennow

, Carrillo

, Dunn

, Haeberlein

, Holtzman

, Jagust

, Jessen

, Karlawish

, Liu

, Molinuevo

, Montine

, Phelps

, Rankin

, Rowe

, Scheltens

, Siemers

, Snyder

, Sperling

, Contributors (2018) NIA-AA Research Framework: Toward a biological definition of Alzheimer’s disease, Alzheimers Dement 14, 535–562.

30.

Lewczuk

, Ermann

, Andreasson

, Schultheis

, Podhorna

, Spitzer

, Maler

, Kornhuber

, Blennow

, Zetterberg

(2018) Plasma neurofilament light as a potential biomarker of neurodegeneration in Alzheimer’s disease, Alzheimers Res Ther 10, 71.

31.

Jiang

, Li

, Huang

, Liu

, Zhao

(2017) The gut microbiota and Alzheimer’s disease, J Alzheimers Dis 58, 1–15.

32.

Alkasir

, Li

, Jin

, Zhu

(2017) Human gut microbiota: The links with dementia development, Protein Cell 2, 90–102.

33.

Rosa

, Henriques

, Carvalho

, Oliveira

, da Cruz E Silva

(2017) Screening younger individuals in a primary care setting flags putative dementia cases and correlates gastrointestinal diseases with poor cognitive performance, Dement Geriatr Cogn Disord 43, 15–28.

34.

Sheng

, Lin

, Wang

, Han

, Liu

(2021) Altered gut microbiota in adults with subjective cognitive decline: The SILCODE study, J Alzheimers Dis 82, 513–526.

35.

Pappolla

, Perry

, Fang

, Zagorski

, Sambamurti

, Poeggeler

(2021) Indoles as essential mediators in the gut-brain axis. Their role in Alzheimer’s disease, Neurobiol Dis 156, 105403.

36.

Colombo

, Sadler

, Llovera

, Singh

, Roth

, Heindl

, Sebastian Monasor

, Verhoeven

, Peters

, Parhizkar

, Kamp

, Gomez de Aguero

, MacPherson

, Winkler

, Herms

, Benakis

, Dichgans

, Steiner

, Giera

, Haass

, Tahirovic

, Liesz

(2021) Microbiota-derived short chain fatty acids modulate microglia and promote Aβ plaque deposition, Elife 10, e59826.

37.

Asnicar

, Berry

, Valdes

, Nguyen

, Piccinno

, Drew

, Leeming

, Gibson

, Le Roy

, Khatib

, Francis

, Mazidi

, Mompeo

, Valles-Colomer

, Tett

, Beghini

, Dubois

, Bazzani

, Thomas

, Mirzayi

, Khleborodova

, Oh

, Hine

, Bonnett

, Capdevila

, Danzanvilliers

, Giordano

, Geistlinger

, Waldron

, Davies

, Hadjigeorgiou

, Wolf

, Ordovás

, Gardner

, Franks

, Chan

, Huttenhower

, Spector

, Segata

(2021) Microbiome connections with host metabolism and habitual diet from 1,098 deeply phenotyped individuals, Nat Med 27, 321–332.

38.

Saji

, Tone

, Murotani

, Yagita

, Kimura

, Sakurai

(2018) Cilostazol may decrease plasma inflammatory biomarkers in patients with recent small subcortical infarcts: A pilot study, J Stroke Cerebrovasc Dis 27, 1639–1645.