Gut microbiota interacts with intrinsic brain activity of patients with amnestic mild cognitive impairment

Abstract Aims To explore the potential relationships among gut microbiota (GM), local brain spontaneous activity, and neuropsychological characteristics in amnestic mild cognitive impairment (aMCI) patients. Methods Twenty aMCI and 22 healthy control (HC) subjects were recruited. The GM composition was determined by 16S ribosomal RNA gene sequencing. Resting‐state functional magnetic resonance imaging scans were performed, and fractional amplitude of low‐frequency fluctuations (fALFF) was calculated across different frequencies. The Spearman or Pearson correlation analysis was used to analyze the relationship between spontaneous brain activity and cognitive function, and GM composition. Results aMCI patients had altered GM state and local spontaneous brain activity as compared with HC subjects. Correlation analysis showed that aMCI and HC groups had different “GM‐intrinsic brain activity interaction” patterns. In aMCI group, at the typical band (0.01‐0.08 Hz), the relative abundance (RA) of Bacteroides from phylum to genus level was negatively correlated with fALFF value of cerebellar vermis IV‐V, and the Ruminococcaceae RA was negatively correlated with fALFF values of left lenticular nucleus and pallidum. The Clostridiaceae RA and Blautia RA were positively correlated with the left cerebellum lobules IV‐V at the slow‐4 band (0.027‐0.073 Hz). The Veillonellaceae RA was positively correlated with fALFF values of left precentral gyrus at the slow‐5 band (0.073‐0.08 Hz). Correlation analysis showed that Clostridium members (Lachnospiraceae and Blautia) were positively, while Veillonellaceae was negatively, correlated with cognition test. Bacteroides was positively correlated with attention and computation, and negatively correlated with the three‐stage command score. Conclusions aMCI patients have a specific GM‐intrinsic brain activity‐cognitive function interaction pattern.


| INTRODUC TI ON
Alzheimer's disease (AD) is the commonest cause of dementia in the elderly and is clinically characterized by progressive cognitive impairment. Converging evidences from both genetic at-risk cohorts and clinically normal older individuals suggests that the pathophysiological process of AD begins years before the diagnosis of clinical dementia. [1][2][3] Both the underlying pathophysiological processes and clinical symptomatology suggest that AD should been considered as a continuum or a spectrum. 4 Mild cognitive impairment (MCI), recognized as the prodromal stage of AD, refers to a transitional period between normal aging and the dementia stage of this condition. 5 Among the subtypes of MCI, amnestic MCI (aMCI) carries a substantial risk for progression to AD, with a yearly transition rate of up to 25%. 6 To date, no effective medical treatment exists to prevent or slow AD progression, and early detection of individuals in the prodromal stage, especially of MCI patients, is of great importance.
Although Aβ peptide accumulation is considered to be a key early event in the pathogenesis of AD, precise pathologic mechanisms remain to be elucidated. 7 Ongoing work in the field of central nervous system diseases has yielded preliminary evidence of gut microbiota (GM), an important environmental factor, interacting closely with brain functions, including cognition, emotion, and social behavior, suggesting a potential role in AD pathogenesis. 8,9 Animal studies have shown that intestinal dysbiosis is involved in the initiation, development, and progression of AD, including chronic neuroinflammation, oxidative stress, and neuroimmune-neuroendocrine pathologies. 8,10 Recently, an important link has been established between gut-derived lipopolysaccharide (LPS) translocation to the perinuclear region and the AD-affected brain. 11 More recently, Vogt et al reported differences in GM of American patients with AD, including decreased Firmicutes, increased Bacteroidetes, and decreased Bifidobacterium. 12 However, there are few direct data on GM in aMCI patients. Our previous study found that fecal microbial composition of aMCI patients was altered, with increased Bacteroides and decreased several beneficial genus belong to phylum Firmicutes. 13 Given the difficulty of precisely elucidating human cerebral cellular networks, few studies have established direct links between the GM and cerebral function in vivo. 14 However, Tillisch et al discovered that intake of fermented milk product probiotics over a four-week period was associated with significant changes in the intrinsic activity of the resting brain during emotion recognition tasks in healthy female subjects as evaluated using functional magnetic resonance imaging (fMRI). 15 Furthermore, they grouped the healthy women into two bacterial genus-based clusters: one cluster with a greater abundance of Bacteroides and another with a greater abundance of Prevotella. The Prevotella group exhibited more negative emotional responses to negative images, and such responses were associated with reduced functional activation of the hippocampus. 16 On the other hand, because functional alterations often precede structural changes, analysis of intrinsic brain activity is essential for understanding the pathogenesis of aMCI, as well as its early detection. In contrast to task-based fMRI, resting-state fMRI (rs-fMRI) does not require complicated experimental designs that can remove some stimuli or task-related confounds. It provides a reliable measure of "baseline" brain activity and connectivity. The amplitude of low-frequency fluctuations (ALFF) technique was found to be reliable and useful in characterizing the intrinsic or spontaneous brain activity during rs-fMRI evaluation and has been widely used in patients with aMCI or AD. [17][18][19][20] The low-frequency range can be usually subdivided into four distinct bands: slow-5 (0.01-0.027 Hz), slow-4 (0.027-0.073 Hz), slow-3 (0.073-0.198 Hz), and slow-2 (0.198-0.25 Hz). Slow-3 and slow-2 bands mainly reflect white matter signals and high-frequency physiological noises, respectively, and slow-4 and slow-5 bands reflect gray matter signal. 20 Han et al demonstrated that the widespread abnormalities in intrinsic brain activity of aMCI patients are frequency-dependent (slow-4 versus slow-5). 21 However, ALFF was also reported to be sensitive to physiological noise. The fractional ALFF (fALFF) approach, relying on filtering, scaling, and normalization to control for physiological or random noise, as well as ventricular contamination and global individual differences, would thus significantly improve sensitivity and specificity in detection of regional spontaneous brain activity. 22,23 To the best of our knowledge, no study to date has detailed the interactions between intestinal microbiota and brain function as evaluated by fMRI and measures of cognition level in aMCI subjects.
Here, we aimed to study how intrinsic cerebral activity (as evaluated by fALFF of rs-fMRI) and cognitive function associates with the GM profiles in aMCI patients.

| Gut microbiota analyses
Each participant was asked to collect a fresh fecal sample.
Stools were preserved at −80°C until analysis. Genomic DNA was extracted from stool samples using a DNA extraction kit (Qiagen, Valencia, CA) according to the manufacturer's protocol, as previously described. 28,29 The V3-V4 region of the 16S bacterial ribosomal RNA gene has been targeted by the following primers "forward fusion primers (5'-CAAGCAGAAGACGGC ATACGAGATGTGAC TGGAGT TCAGACGTGTGC TC T TCCGAT CTBARCODEACTCCTACGGGAGGCAGCAG-3') and the reverse fusion primer (5'-AATGATACGGCGACCACCGAGATCTACACTCTT  *P < .05, compared with HC; t, the P value was obtained by the twosample t test; χ 2 , the P value was obtained by the chi-square test.

| Data acquisition and preprocessing
Functional and structural imaging data were acquired using a 3.  in the slow-4 and slow-5 bands. 32 Here, the typical 0.01-0.08 Hz low-frequency range was divided into two sub frequency bands:

| fALFF analysis
slow-4 band (0.027-0.073 Hz) and slow-5 band (0.01-0.027 Hz). For standardization purposes, the fALFF value of each voxel was divided by the mean fALFF within the brain mask, and relevant frequencydependent fALFF values were calculated for each subject.

| Statistical analyses
Results are presented as numbers with percentages, means with standard deviations (mean ± SD), or medians with interquartile ranges (medians (IQR)) when appropriate. Statistical analyses were All statistical tests were two-sided, and difference achieving values of P < .05 were considered statistically significant.

| Demographic and clinical characteristics of enrolled subjects
Demographic and clinical parameters of aMCI and HC groups are summarized in

| Comparison of GM composition between aMCI patients and healthy controls
To assess differences in fecal microbiota of patients with aMCI compared with those of HC subjects, parallel pyrosequencing was We subsequently performed taxon-based analysis, comparing the prevalence of GM at phylum, class, order, family, and genus levels between aMCI and HC groups (see Figure 2). At the phylum level, no significant difference in relative abundance (RA)

| fALFF of group differences
Both increased and decreased regional function was revealed in aMCI subjects relative to controls ( precentral gyrus (see Figure 3 and Table 3).

| Associations between the cognitive tests, intrinsic brain activities, and gut microbiome
We subsequently explored the potential relationships between alterations in GM relative abundance (RA) and intrinsic brain activities (ie, fALFF values), and cognitive test scores. As shown in Figure 4, aMCI and HC groups had different "GM RA-intrinsic brain activity interaction" patterns. Specifically, we found that the GM with intergroup differences correlated significantly with fALFF values of several brain regions in both HC and aMCI patients.
In HC subjects, the RA of class Clostridia (r = .773, P = .024), order Correlation analyses were next performed to evaluate the association between RA of altered microbiomes and cognitive test findings. As shown in Figure 5, the RA of phylum Bacteroidetes, class

| D ISCUSS I ON
Here, we first report cognitive and intrinsic brain activity differences related to intestinal microbial composition in aMCI subjects.
In the present study, although there were no significant differences in global diversity or evenness of the complex microbial communities between aMCI and HC subjects, the RA of several gut bacteria in aMCI subjects was distinct from age-and sex-matched control individuals. Moreover, the associations between RA of specific bacterial taxa with the intrinsic activity of the resting brain paralleled changes of corresponding cognition domains. Such a chronic inflammatory state is considered to be a key contributor to AD pathology. 34 Thus, it is necessary to determine the host health state and populations of specific symbionts in order to understand the significance of certain physiological changes in the gut and/or the brain.
The cerebellum has traditionally been considered to be primarily dedicated to motor functions. However, earlier pathologic studies have reported insults in the cerebellum of AD patients. 35 Clinical and neuroimaging studies have demonstrated that the cerebellum greatly influences thoughts, emotions, as well as language and cognitive processes by functionally connecting and interacting with cerebral networks (ie, dorsal and ventral attention, frontoparietal, default mode, and salience networks). 36 Indeed, the cerebellum is greatly involved in the pathophysiology of AD. Of note, an insidious decline in the accuracy, speed, and consistency of information processing and cognitive performance in aMCI patients also supports a cerebellar role in the pathophysiology of AD. 37  tumor necrosis factor-α and interleukin-1β in the brain of APP/ PS1 mice. 38 There are a number of various acetate and butyrate producers among Blautia, and their roles in amelioration of obesity and insulin resistance in rats have previously been reported. 39 Importantly, Blautia spp. were found to correlate with a reduction in mortality caused by acute graft-versus-host disease after blood/ marrow transplantation due to the anti-inflammatory effects they exert. 40 It has been proposed that insulin resistance is related to a higher risk for AD. 41 Willette et al have reported that, in asymptomatic late middle-aged individuals at risk for AD, higher insulin resistance predicts temporal and frontal amyloid deposition. 42 Meanwhile, neuroinflammation plays a significant role in the AD pathogenesis. 43 As is consistent with our results, preclinical studies have previously confirmed decreased RA of Ruminococcus in both APP/PS1 and 5 × FAD mice. 7 In addition to SCFA production, Ruminococcus also modulates gut mucin expression and degradation, which is likely responsible for compromised intestinal permeability in AD. 7,44,45 Increases in resting-state activity were found in the left middle occipital gyrus, right lingual gyrus, right cuneus, and left medial superior frontal gyrus in aMCI patients as compared with controls, consistent with previous data. 19 The first three areas were involved in visuospatial perception and the visual network. And the left medial superior frontal gyrus belongs to prefrontal cortex, which is important for working memory, executive function. 46 Impairment of the above cognitive function was reported to occur in both aMCI and AD patients. One possible interpretation of this phenomenon is that a compensatory mechanism exists throughout a limited spectrum of along the aging-MCI-AD continuum that results in hyperactivation among aMCI patients. These brain regions are activated to compensate for a reduction in function of other brain regions. 18,47 Here, we found that aMCI and health controls showed different association pattern between intrinsic brain activities, and GM. Bajaj et al demonstrated that elderly populations had a complex interaction between cognition, fMRI activation patterns, and GM abundance. Subjects with both amnestic and non-amnestic impairment required greater neuronal recruitment of the visuospatial network to achieve the same response compared those with amnestic impairment alone, or unimpaired subjects under task-based fMRI. 48 Those findings suggest that a potential mechanism regarding the brain activities may be attributable to the GM, either directly or indirectly.
As mentioned above, the metabolites, toxins, and pro-inflammatory factors from different GM may be a kind of key regulator of the brain activity. Thereafter, longitudinal studies examining the association between intrinsic brain activity and GM metabolites will shed light on this issue.
The results of our study also showed aMCI had altered intrinsic brain activity in specific frequency bands compared with healthy control. Previous studies have demonstrated that pattern of intrinsic brain activity was sensitive to specific frequency bands. 17,21 And it has been reported that abnormalities of intrinsic brain activity in specific brain regions were frequency-dependent (particularly, the slow-4 and slow-5 bands) both in AD and aMCI patients. 21,49 Though there are no consistent results, different frequency-dependent patterns may suggest different oscillations and frequency-dependent brain spontaneous activities for specific disease state. Further studies combined with electroencephalographic scalp recordings will be helpful to identify the neurophysiological mechanism of the signals located at specific frequency bands.
One limitation of this study is small sample size and preliminary cross-sectional study. Due to the small study size, the adjustment for multiple comparisons has not been done. The results should be interpreted with caution. Likewise, although stool samples enabled us to detect a wide range of intestinal microflora, it was not possible to differentiate between the luminal and mucosal environment, much less local microenvironments, or regional differences throughout the gut. We did not evaluate bacterial functional changes (ie, metabolic and inflammatory), and our brain scans only studied resting-state activity. Multi-modal neuroimaging examinations, including those evaluating brain microstructure and functional connectivity, may reveal interactions between intestinal flora and the brain function more comprehensively.
In summary, we found aMCI subjects have distinct GM compared with age-and sex-matched HC subjects. Changes in resting-state brain activity and cognitive function run in parallel to specific bacterial taxal populations comprising the GM.

ACK N OWLED G M ENTS
We thank Dr Erik for editing the English version of a draft of this manuscript.

F I G U R E 5
Heatmap showing correlations between fecal microbiome and cognitive test scores of all subjects. Red means positive correlation and blue means negative correlation; *P < .05, **P < .01