Relationship between dementia and gut microbiome-associated metabolites: a cross-sectional study in Japan

Dysregulation of the gut microbiome is associated with dementia. However, the relationship between microbiome-associated metabolites and dementia has yet to be identified. Outpatients visiting a memory clinic in Japan enrolled in this cross-sectional study; 107 subjects were eligible for the study, 25 of which had dementia. We collected demographics, activities of daily living, risk factors, cognitive function, and brain imaging data. The gut microbiome was assessed using terminal restriction fragment length polymorphism analysis. Concentrations of faecal metabolite were measured. We used multivariable logistic regression analyses to identify whether metabolites were independently related to dementia. The concentrations of metabolites were significantly different between subjects with and those without dementia. Every 1 standard deviation increment in faecal ammonia concentration was associated with around a 1.6-fold risk for the presence of dementia. A higher faecal lactic acid concentration was related to a lower risk of dementia, by around 60%. A combination of higher faecal ammonia and lactic acid concentrations was indicative of the presence of dementia, and had a similar predictive value as traditional biomarkers of dementia. Thus, faecal ammonia and lactic acid are related to dementia, independently of the other risk factors for dementia and dysregulation of the gut microbiome.

Dementia vs. no-dementia group. Compared with the no-dementia group, the dementia group included significantly more women (Dementia vs. no-dementia: female sex, 80.0% vs. 52.4%, P = 0.02) and had impaired instrumental ADL (76.0% vs. 35.4%, P < 0.001), and a significantly lower cognitive function (median MMSE score, 19 vs. 27, P < 0.001; median CDR-SB score, 4.5 vs. 1.0, P < 0.001). Further, patients with dementia scored significantly lower on the Lawton and Brody scale, the Mini-Nutritional Assessment-Short Form (MNA-SF), Alzheimer's Disease Assessment Scale-cognitive subscale (ADAS-cog), Frontal Assessment Battery (FAB), Raven's Coloured Progressive Matrices, and Logical Memory subtests I and II of the Wechsler Memory Scale-Revised. Cerebral small vessel diseases such as silent lacunar infarcts (SLIs) and cerebral microbleeds (CMBs) and high VSRAD scores (the voxel-based specific regional analysis system for Alzheimer's disease: VSRAD) were frequent on MRI scans of subjects with dementia (Table 1, S1). The dementia group had fewer enterotype I microbes (a lower prevalence of Bacteroides) and more enterotype III microbes (a higher prevalence of other bacteria) than the no-dementia group (20.0% vs. 46.3%, 80.0% vs. 47.6%, P = 0.014, respectively, Table 2). Although constipation is a microbiome-related factor 10 , there was no significant relationship between constipation and cognitive function. There were also no significant differences in metabolites between patients with and without constipation (Table S2).
Metabolites. The concentrations of some metabolites in faeces such as ammonia, phenol, and p-cresol were significantly higher in the dementia group compared with the no-dementia group (median ammonia: 0.83 vs. 0.65 mg/g, P = 0.026; median phenol: 1.97 vs. 1.05 μg/g, P = 0.029; median p-cresol: 57.5 vs. 0.29 μg/g, P = 0014, respectively, Table 2). Univariable analyses to identify the metabolites that most significantly contributed to the presence of dementia showed that the highest OR was of faecal ammonia concentration (OR = 1.6, 95% CI 1.0-2.5, P = 0.033), the lowest OR was of faecal lactic acid concentration (OR = 0.3, 95% CI 0.02-1.0, P = 0.048), and OR range of the other metabolites was 0.7-1.6 ( Fig. 1, Table S3). Because medication is a microbiome metabolite-related factor 11 , we conducted bivariable analyses between each microbiome metabolite and a range of medications. The use of some medications (anti-thrombotic drug, statin, anti-hyperglycaemic drug, and aperient) was associated with faecal metabolite levels such as lactic acid, acetic acid, iso-valeric acid, n-butyric acid, phenol, 4-Ethylphenoll, and indolel (Table S4).

Multivariable analysis.
Multivariable logistic regression analysis revealed that ammonia concentration was associated with the presence of dementia, independent of age, sex, education years, ApoE ε4, enterotypes (either enterotype I or III), and risk factors (Table 3, S5). The overall ORs of ammonia after adjustment were at least 1.6 ( Fig. 2A, Table S6). Multivariable logistic regression analysis revealed that lactic acid concentration was associated with the presence of dementia, independent of age, sex, education years, ApoE ε4, risk factors, and brain MRI and/or SPECT abnormalities (OR = 0.1, 95% CI 0.01-0.5, P = 0.001, Table S7). Lactic acid concentration had a tendency to exhibit an inverse association with the presence of dementia (Table 3, S5). The overall ORs of lactic acid after adjustment were at most 0.4 (Fig. 2B, Table S7).

Discussion
The main finding of this sub-analysis study was a relationship between gut microbiome-associated metabolites and dementia. More specifically, every 1 SD increment in faecal ammonia concentration was associated with around a 1.6-fold increased risk for the presence of dementia. We also found that every 1 SD decrement in faecal lactic acid concentration tended to be associated with a reduced risk of dementia. A combination of faecal ammonia and lactic acid concentration was indicative of the presence of dementia and had the same or slightly lower predictive accuracy as traditional biomarkers of dementia, including ApoE ε4 and VSRAD scores.
The independent relationship between the faecal metabolite concentration and dementia shown in our study is novel. Existing literature indicates that gut microbial metabolites provide a functional readout of microbial activity and can be used as an intermediate phenotype mediating host-microbiome interactions 12 . Furthermore, metabolites such as phenol, p-cresol, indole, and ammonia are typically considered to be potentially harmful 13 . The levels of these metabolites in our study were higher in demented patients than in non-demented patients. Our finding is in line with previous studies 12,13 . To the best of our knowledge, this is the first clinical study to demonstrate the association of gut microbiome metabolites -ammonia and lactic acid concentration -with cognitive function in the elderly. We found that faecal ammonia was positively related to the presence of dementia. A high blood ammonia concentration is known to be a risk factor for cognitive impairment 14 and Alzheimer's disease 15 , and our findings are in line with these reports. Higher faecal ammonia concentrations may be related to higher blood ammonia concentrations 13 . Furthermore, faecal ammonia concentrations in children with autism spectrum disorders are higher than those in control children 16 . Ammonia alters nucleic acid synthesis, changes morphology and intermediary metabolism, and reduces the life span of intestinal cells 17 . These previous findings also support our findings in the current study. Short-chain fatty acids (SCFAs) were also associated with dementia in the present study. SCFAs produced by gut microbes can regulate the biosynthesis of neurotransmitters, such as serotonin 18 and leptin 19 . Although we did not identify any neurotransmitters related to SCFAs that might regulate host cognitive function in our study, dysbiosis may affect the biosynthesis of neurotransmitters, affecting brain function. Previous findings of associations between the gut microbiome and depression 20 or attention deficit hyperactivity disorder 21 support this speculation.
Inversely, our multivariable analysis revealed that the OR of lactic acid for dementia was at most 0.4, although there was not a significant independent association with dementia. This means that lactic acid may have a protective effect against dementia. Indeed, several previous studies have reported Lactobacillus 22 and/or Bifidobacterium to have protective effects against cognitive deterioration 23,24 . The relationship between lactic acid and dementia may be due to the direct activation of lactic acid-producing bacterium such as Lactobacillus or Bifidobacterium,  www.nature.com/scientificreports www.nature.com/scientificreports/ or the secondary benefit of consuming food and drink containing a lactic acid fermentation product. We did not identify the origin of the faecal lactic acid (endogenous or exogenous) and this should be addressed in future studies.
The mechanism underlying the relationship between faecal metabolite concentration and dementia has yet to be clarified. However, we found an independent association between the two on the basis of our systematic assessment of cognitive function using a comprehensive geriatric assessment, various faecal metabolites, and potential biomarkers for dementia. The graphical modelling also showed such relationship. This relationship might be caused by functional disorders of the neuro-inflammatory system 25 , microvascular inflammation 8 , or remote effects driven by various metabolites 26 . Specifically, previous studies have speculated on the potential mechanisms connecting faecal metabolites and dementia. Gut microbiome disorder promotes the production of toxic metabolites and inflammation-mediated cytokines, and a reduction in beneficial substances such as anti-inflammatory factors. There are functional disorders of the gut epithelial barrier, with concomitant activation of the immune system, as well as the dysregulation of enteric neurons and glia 27 . These events subsequently lead to bloodbrain barrier dysfunction, triggering neuro-inflammatory reactions and apoptotic neuronal and glial cell death,     www.nature.com/scientificreports www.nature.com/scientificreports/ particularly in the hippocampus and cerebral cortex, which underlie the development of dementia 27 . Recent studies have also reported that microbiome metabolites, such as SCFA and trimethylamine/trimethylamine N-oxide, contribute to life-threatening diseases such as heart failure and atherosclerosis 8 . Our comprehensive assessment can provide detailed evidence to fill the knowledge gap regarding the mechanism between dementia and the gut microbiome. Furthermore, analysing metabolites of the gut microbiome longitudinally could tell us more about the relationship between gut microbiome dysfunction and dementia, and could reveal the unidentified mechanism for the onset of dementia.
Our study has limitations. The causal relationship between metabolites and dementia could not be tested because this is a cross-sectional study. We are in the process of conducting a longitudinal assessment in the schema of the Gimlet study and this assessment will clarify the causal relationship. The relatively small number of subjects may mean that there was a low statistical power. There might be a selection bias because of a single hospital-based cohort. There might be possible differences in collection and storage of faecal samples among study subjects. High-throughput DNA sequencing technology would be useful to identify the specific genera or species of microbes compared with the terminal restriction fragment length polymorphism (T-RFLP) method 4 . The types of gut microbiome in an individual can be affected by age 28 , diet 29 , and medication 6,30,31 . We have recently assessed patients' diet patterns and will analyse the associations between cognitive function, gut microbiome, and diet patterns in a forthcoming sub-analysis. Regarding medication, although the use of some drugs was associated with metabolites in our bivariable analyses, it is difficult to find clinically consistent interpretations of these associations. The use of any anti-thrombotic drugs may be associated with underlying cardiovascular or cerebrovascular diseases. In addition, the use of aperient may be associated with constipation. Furthermore, the use of gastric acid-suppressing drugs might affect the gut microbiome, because a previous study has reported that proton pump inhibitors reduce gastric acid secretion and may modulate gut microbiota composition 30 . Thus, we did not include medication in the multivariable analyses because the effects of medication on cognitive function are more complex than our study allows. Frailty 32 may also affect the gut microbiome, and should be investigated in further studies. Amyloid β precursor protein may also be important to assess, because a high serum concentration of amyloid β precursor protein is indicative of inflammatory endothelial dysfunction and increased the risk of cognitive impairment 33 . Furthermore, subtypes of dementia, such as Alzheimer's disease and frontotemporal lobar degeneration, were not considered in the current sub-analysis, because our sample size was relatively small; subdividing patients by dementia subtypes would lead to a small number of subjects in each group and a resulting statistical underestimation. The main aim of the present study was to determine the association between the gut microbiome and dementia, which was defined by simple categorisation on the basis of MMSE and CDR scores. However, the absence of any analysis of dementia subtypes limits the utility of our findings, and will be corrected in future studies.
Although this sub-analysis study included a small number of subjects and we did not show a robust and statistically significant relationship regarding the association between lactic acid and dementia in the multivariable analyses, our findings, at least regarding faecal ammonia, support the idea that there is a relationship between the gut microbiome and dementia. Further studies are warranted to clarify the mechanism underlying such relationship.

conclusions
We have shown a strong relationship between gut microbiome-associated metabolites such as ammonia and dementia, independent of the traditional risk factors and the gut microbiome. Furthermore, faecal lactic acid concentration may be inversely associated with the presence of dementia.

Methods
Study design. This study was a sub-analysis of data from the gerontological investigation of microbiome: a longitudinal estimation study (the Gimlet study) at the National Center for Geriatrics and Gerontology (NCGG) 4 . The present sub-analysis study investigated the cross-sectional association between metabolites of the gut microbiome and dementia. The study complied with the tenets of the Declaration of Helsinki and was approved by the Institutional Review Board at the NCGG (No. 1191). Informed consent was obtained from all patients and their families before participating in this study. The study is registered with the UMIN Clinical Trials Registry (UMIN000031851). Detailed information regarding the Gimlet study and the study methods is provided in the supplementary file and the previous report 4 .

Subjects.
Between March 2016 and March 2017, we enrolled consecutive patients visiting the memory clinic at the NCGG who agreed to undergo medical assessment of their cognitive function and faecal examination. Patients were eligible for participation in the Gimlet study if they met the following criteria: (1) able to undergo brain magnetic resonance imaging (MRI); (2) provided written informed consent; (3) provided informed consent for the NCGG Biobank to store their clinical data, blood, and faecal samples; and (4) were accompanied by a study partner who could assess the daily condition of the patient (patient's family or a primary caregiver). The exclusion criteria were as follows: (1) the MRI scan could not be evaluated because of movement-related artefacts; (2) local lesions, such as cerebral infarction, that were detected by MRI before enrolment, which could significantly affect cognitive functioning; (3) a history of a major psychological disorder or current serious or unstable alcohol or drug abuse; (4) ≤6 years of education, because a low level of education is a robust risk factor for dementia; (5) a history of cancer of the digestive tract; (6) brain tumour, encephalitis/meningitis, normal pressure hydrocephalus, subdural hematoma, or lower cognitive function due to head injury; and/or (7) were unable to provide sufficient faecal samples for metabolite analysis.
www.nature.com/scientificreports www.nature.com/scientificreports/ Baseline assessment. All participants underwent a comprehensive geriatric assessment 34 that measured the following: (1) demographic characteristics; (2) potential risk factors for dementia, such as hypertension, dyslipidaemia, diabetes mellitus, ischemic heart disease, chronic kidney disease, a smoking habit, a history of stroke, and alcohol consumption; (3) the Barthel Index to assess basic ADL 35 , and the Lawton and Brody scale to assess instrumental ADL 36 ; (4) global cognitive function, using the Mini-Mental State Examination (MMSE) 37 and Clinical Dementia Rating (CDR) scale 38 ; (5) neuropsychological tests, including the Alzheimer's Disease Assessment Scale-cognitive subscale 39 , Frontal Assessment Battery 40 , Raven's Coloured Progressive Matrices 41 , and Logical Memory subtests I and II of the Wechsler Memory Scale-Revised 42 ; (6) laboratory variables, including ApoE ε4 as a risk factor for AD; (7) ankle brachial index and pulse wave velocity as indicators of arteriosclerosis 43 , and the 'impact' of pulse 44 ; (8) brain imaging such as MRI and single photon emission-computed tomography (SPECT); (9) assessment of other clinical features such as frailty 45 and subjective hearing loss; (10) social and lifestyle factors, such as nutritional status, which was measured using the MNA-SF 46 , and constipation, which was defined as a bowel movement once in 3 days and/or the taking of anti-constipation drugs (aperients); and (11) assessment of current medication (anti-dementia drugs, anti-hypertensive drugs, statins, proton pump inhibitors/ H2 blockers, anti-thrombotic drugs, anti-hyperglycaemic drugs, and aperients). The clinical samples and data were provided by the NCGG Biobank, which collects clinical data for research.
Classification of cognitive function. We divided patients into the two following categories according to our previous report: 4 (1) the group without dementia (MMSE score ≥20 and a CDR score <1) and the group with dementia (MMSE score <20 and/or a CDR score ≥1). A CDR score of 0.5 indicates the presence of mild cognitive impairment (MCI) and possibly very mild dementia, which means that the patient has a higher risk of dementia 47 . Normal cognition was defined as MMSE ≥ 20 and CDR = 0.
Brain imaging. Patients underwent a 1.5 T MRI brain scan (Philips Ingenia, Eindhoven, the Netherlands).
MRI scans obtained included diffusion-weighted imaging, fluid-attenuated inversion recovery imaging, T2-weighted imaging, T2 * -weighted gradient echo imaging, 3D T1-weighted sagittal and axial coronal views, and 3D time-of-flight MR angiography scans. The presence and components of cerebral small vessel disease were categorised using the standards for reporting vascular changes on neuroimaging 48 , including SLIs, white matter hypersensitivity (WMH), CMBs, and cortical superficial siderosis (CSS). We used VSRAD advance software (Eisai Co., Ltd., Tokyo, Japan) to quantify cortical and hippocampal atrophy using a standardised z-score. An increased VSRAD score suggests the potential presence of AD because it is indicative of hippocampal atrophy, one of the characteristics of the AD brain 49 . Patients also underwent N-isopropyl-p-[ 123 I]-iodoamphetamine-SPECT to assess the presence or absence of a reduction in blood flow in the posterior cingulate gyrus and/or praecuneus as a surrogate marker of Alzheimer's disease 50 . Gut microbiome. Faecal samples were collected at home and were frozen and preserved at −81 °C at the NCGG Biobank. Patients were consuming their usual diet at the time of sampling. After all samples had been collected, the gut microbiome was analysed using T-RFLP analysis by the TechnoSuruga Laboratory (Shizuoka, Japan) 51 . T-RFLP analysis is one of the most well-established and reliable 16 S ribosomal RNA-based methods, which has a high throughput and reproducibility 51 . First, T-RFLP was used to classify gut microbes into the following 10 groups: Prevotella, Bacteroides, Lactobacillales, Bifidobacterium, Clostridium cluster IV, Clostridium subcluster XIVa, Clostridium cluster IX, Clostridium cluster XI, Clostridium cluster XVIII, and 'others' . Second, we stratified the gut microbiome into the three following enterotypes: enterotype I included Bacteroides at >30%, enterotype II included Prevotella at >15%, and enterotype III included the remaining bacteria, and this classification was made according to the Human Faecal Microbiome T-RFLP profile 52,53 . Third, we assessed the Firmicutes/ Bacteroidetes (F/B) ratio 53 . The phylum Firmicutes includes the Lactobacillales and the Clostridium clusters, and the phylum Bacteroidetes includes Bacteroides and Prevotella.
Statistical analysis. Continuous, ordinal, and categorical variables are expressed as the mean ± standard deviation (SD), median and interquartile range, and frequency or proportion (percentage), and were compared using the unpaired Student's t-test, Wilcoxon rank-sum test, and χ 2 test, respectively according to our previous study 4 . We compared clinical characteristics, composition of the gut microbiome, and metabolites between the no-dementia and dementia groups using the Wilcoxon rank-sum test and the χ 2 test. Second, we standardised actual measurement values of metabolites and performed univariable logistic regression analyses to identify the metabolites that were associated with the presence of dementia. Third, multivariable logistic regression models were used to identify the metabolites independently associated with dementia using the standardised metabolite