Abstract

Background

   The association between gut microbes and Alzheimer's disease (AD) has
   not been entirely elucidated.

Objective

   We aimed to demonstrate the association between gut microbes and AD and
   to further investigate the pathogenesis of microbes with a causal
   relationship to AD.

Methods

   Mendelian randomization analyses were used to determine the significant
   causal relationship between gut microbes and AD. Protein-protein
   interaction (PPI) network was used to identify the hub genes.
   Functional enrichment analysis was used to reveal the pathogenesis
   theoretically between gut microbes and AD.

Results

   In the present study, a total of 32 microbes were identified that were
   significantly associated with AD. Subsequently, DLGAP2, NRXN3, NEGR1,
   NTNAP2, MYH9, and SCN3A were identified as hub genes. The genes NRXN3,
   NEGR1, and NTNAP2 were enriched in the cell adhesion molecules (CAMs)
   signaling, and the taxons of gut microbes that corresponded to these
   were Bifidobacterium adolescentis, Actinomycetales, and Intestinimonas
   massiliensis.

Conclusions

   Bifidobacterium adolescentis, Actinomycetales, and Intestinimonas
   massiliensis may promote the progression of AD through the regulation
   of the CAMs signaling pathway-mediated synaptic function. Hence, the
   in-depth study of gut microbes may increase the efficiency of screening
   and diagnosis of AD.

   Keywords: Alzheimer's disease, bioinformatics, cell adhesion molecules
   signaling, gut microbes, mendelian randomization

Introduction

   Alzheimer's disease (AD) is a neurodegenerative encephalopathy caused
   by damage to neurons in the brain, leading to the deficits in language,
   movement, and thinking. The Alzheimer's Association (2024) announced
   that approximately 10 to 12 million people (>65 years old in America)
   will suffer from AD or other forms of cognitive disorders.^ [31]1
   Between 2000 and 2021, the mortality of AD rose significantly and
   AD-related deaths doubled.^ [32]2 Some studies have reported that the
   life expectancy after a diagnosis of AD in older people aged 65 and
   over is generally four to eight years on average, with some lasting as
   long as 20 years.^[33]3–[34]5 There was no doubt that the length of
   illness before death had a serious impact on patients, families, and
   even the social health care system, ultimately leading to a heavy
   economic burden on families and society. Until now, there was no
   definitive prevention or cure for AD, which means that a broader and
   deeper understanding of AD was needed, including its etiology and how
   to prevent, manage, and cure it. The entire pathogenesis for AD was
   unclear, with the mainstream direction focus on neurofibrillary tangles
   and senile plaques.^ [35]6

   It is noteworthy that the central nervous system (CNS) appears to be a
   subtle link with the gut microbe, based on the concept of the gut-brain
   axis.^ [36]7 Approximately 1000 species of microorganisms are involved
   in a variety of metabolic functions within the human gut.^ [37]8 Due to
   the inter-individual variability in age, lifestyle, and environment,
   microbial diversity varies between individuals. Abnormalities in the
   gut microbe may contribute to an inflammatory response, which can
   result in damage to the intestinal mucosal barrier and the subsequent
   progression of inflammatory bowel disease.^ [38]9 Previous studies have
   indicated that Escherichia, Klebsiella, Haemophilus, and Proteus are
   associated with this progression.^[39]10,[40]11

   Furthermore, microbial metabolites have been demonstrated to contribute
   to host immunity and cerebral function. These include short-chain fatty
   acids, butyric acid, amino acids, and lipopolysaccharides
   (LPS).^[41]12–[42]14 A noteworthy study suggested that tryptamine is a
   derivative of tryptophan, and that certain bacteria in the gut (such as
   Clostridium sp.) are capable of producing tryptamine. These bacteria
   contain NADH: ubiquinone reductase (NQR) sequences have been associated
   with AD. Elevated levels of tryptamine could induce the cytotoxicity
   and neurotoxicity before the onset of AD, which means that tryptamine
   in feces may serve as a potential early warning sign of AD.^ [43]15 In
   summary, the gut-brain axis is involved in the pathways between the
   CNS, endocrine, metabolism, and the immune system.^ [44]16 The health
   of the host is contingent upon the abundance and diversity of the gut
   microbe. Recent studies have suggested that gut microbial diversity may
   be an important factor in the pathogenesis of AD.^[45]17–[46]20 In
   particular, the use of bioinformatics to explore common pathogenic
   mechanisms within the gut-brain axis system has become an increasingly
   established approach.

   The MiBioGen study,^ [47]21 which had 18,340 participants from 24
   cohorts (across Europe, the Middle East, and East Asia), was a
   large-scale international scientific research cooperation project. The
   reporting of 211 microbial taxa for genome-wide association studies has
   frequently been employed to elucidate the influence of gut microbes on
   AD, and this subset of data has considerable research value. However,
   we elected not to utilize this particular subset of data, as the
   annotation of the alignment of their gut microbiota demonstrated
   considerable heterogeneity in microbial composition across cohorts. In
   order to maintain the racial consistency of the data, we elected to
   analyze the most recent gut microbial data available, with summary
   statistics for 473 microbial taxa by Mendelian randomization (MR). The
   combination of bioinformatics analysis with the aforementioned data
   provides compelling evidence for further in-depth analysis.

Methods

Study design

   The detailed flowchart of this study was shown in [48]Figure 1. First,
   the gut microbe was used as the exposure and AD as the outcome, and
   instrumental variables (IVs) were sought to explore the causal
   relationship between the two. Second, the gene set was built by
   matching the IVs and the National Center for Biotechnology Information
   (NCBI). Finally, the Gene Expression Omnibus (GEO) dataset was used to
   identify the differentially expressed genes (DEGs), and the overlaps
   were identified between DEGs and the gene set, which was further
   analyzed by protein-protein interaction (PPI) analysis and functional
   enrichment analysis. The study population was all European.

Figure 1.

   [49]Figure 1.
   [50]Open in a new tab

   The detailed flowchart of the study.

Exposure and outcome data

   The latest summary statistics data on gut microbe was selected for the
   present study. Exposure data source from a single large
   population-based cohort of 5959 genotyped individuals with matched gut
   microbial met genomes. A total of 473 genome-wide summary statistics
   were found in the NHGRI-EBI GWAS Catalog
   ([51]https://www.ebi.ac.uk/gwas/) from accessions GCST90032172 to
   GCST90032644.^ [52]22 The outcome data (ieu-b-5067) was selected as the
   AD dataset.

GEO dataset source

   The AD GEO dataset selected for this study was obtained from the GEO
   database ([53]https://www.ncbi.nlm.nih.gov/geo/). The following
   criteria were used to screen the dataset: (1) including cases and
   controls. (2) The minimum number of samples was 20. The [54]GSE28146 of
   AD was selected to screen for DEGs, and the “GEO query” package of the
   R software (version 4.3.2, [55]http://rproject.org/) was used to
   download the sample source from the GEO database. The dataset was
   downloaded from publicly available databases, and no additional ethical
   approval was required.

Data processing and differentially expressed gene screening

   First, the data matrix of gene expression profiling was log2
   transformed. Second, the “Limma” package in the R software was used to
   identify AD DEGs, which were defined as the |logFC|> (mean (abs
   (logFC)) + 2*sd (abs (logFC)) and p-values < 0.05. Third, to visualize
   the DEGs as heatmaps and volcano plots, the packages “ggplot2” and
   “pheatmap” were used, respectively.

Statistical analyses

   The single nucleotide polymorphisms (SNPs) were used as IVs to perform
   the MR analysis. Three assumptions are crucial for the validity of MR
   studies: (1) the SNPs must be significantly associated with the
   exposure; (2) the SNPs must be independent of confounding factors on
   the exposure and outcome; (3) the SNPs only related to the outcome by
   the exposure under study. Therefore, ‘'p < 1e-05'’ was used as a
   significant threshold for filtering out SNPs associated with the
   exposure. Then, this study used ‘'R2 = 0.001'’ and ‘'clump_kb = 10000'’
   for linkage disequilibrium (LD) pruning. After excluding LD, the
   exposure data and outcome data were harmonized. To avoid potential weak
   instrumental variables, the F value was used to filter out weak
   instrumental variables. If F > 10, the instrumental variables were
   considered strong enough to support the results of the MR analysis.
   Eventually, the results were analyzed using the methods of MR analysis,
   such as inverse variance weighted (IVW), MR Egger, weighted median, and
   weighted mode. In particular, the causal relationship between exposure
   and outcome was held for p-value of IVW less than 0.05, and further to
   detect the heterogeneity and horizontal pleiotropy. All statistical
   analyses were based on the Two-Sample MR package in the R software.

Hub genes identification

   First, two datasets were built, named ‘'gene set'’ and ‘'AD DEGs’’,
   respectively. The “gene set” was formed by matching the SNPs that
   supported the MR analyses with NCBI one by one. More specifically, we
   searched the rsID one by one on the NCBI
   ([56]https://www.ncbi.nlm.nih.gov/snp/), identifying genes to form the
   “gene set”. The rsID was a kind of numbering form of the MR analysis
   results, which was based on the “BSgenome.Hsapiens.NCBI.GRCh38” package
   in the R version 4.4.0 software platform. The “AD DEGs” was built by
   using the ‘'Limma package'’ in the R software. Second, the overlap
   between the “gene set” and the “AD DEGs” was obtained. Third, the hub
   genes were confirmed by using the K-means clustering to filter the
   overlaps. The K-means clustering was based on the STRING database (v
   12.0, [57]https://string-db.org), and the PPI network was constructed
   with a confidence = 0.4.

Functional enrichment analysis

   The Gene Ontology (GO) analysis and Kyoto Encyclopedia of Genes and
   Genomes (KEGG) pathway analysis were performed using the “Cluster
   Profiler” package in the R software. Dotplot function was used to
   visualize the results of the GO analysis, and the barplot function was
   used to visualize the results of the KEGG pathway. p < 0.05 was
   considered statistically significant.

Results

The associations between gut microbes and AD

   The results of this study showed that a total of 32 microbes were
   identified from 473 microbial taxa as having a causal relationship with
   AD (p < 0.05). The 32 taxa included one phylum, one order, two
   families, 10 genera, and 18 species. The horizontal pleiotropy and
   heterogeneity effects were shown in [58]Supplemental Table 1. The
   principal results are presented in [59]Figure 2. A total of 16 microbes
   were identified as weak protective factors with odds ratios > 1, and
   the reminding microbes were identified as weak risk factors with odds
   ratios < 1. The weak protective factors included Aureimonas, Bacillus,
   Bacillus Velezensis, CAG-81 sp000435795, CAG-194, sp002441865, CAG-349,
   CAG-485 sp002362485, CHKCI006 sp900018345, Corynebacterium, Eubacterium
   callanderi, Haloplasmatales, Intestinimonas massiliensis,
   Lachnospiraceae, Parabacteroides johnsonii, Turicibacteraceae,
   Turicibacter, Turicibacter sp001543345. The weak risk factors included
   Actinobacteriota, Actinomycetales, AR31, Bifidobacterium,
   Bifidobacterium angulatum, Bifidobacterium adolescentis,
   Bifidobacterium ruminantium, Bifidobacterium kashiwanohense, CAG-449,
   CAG-826, Clostridium M clostridioforme, KLE1615, Lawsonibacter
   sp900066645, RUG472 sp900319345, UBA9475 sp002161235.

Figure 2.

   [60]Figure 2.
   [61]Open in a new tab

   The associations between gut microbes and AD. The microbes were
   identified as weak protective factors by odds ratio > 1, and the
   microbes were identified as weak risk factors by odds ratio > 1.

Identification of two datasets

   In the present study, two datasets were built. The first dataset of
   1316 differentially expressed genes of [62]GSE28146 was named ‘'AD
   DEGs’’. The visualization results were presented in [63]Figure 3 by
   heatmap (3A) and volcano (3B). The second dataset of 314 genes,
   depicted in [64]Figure 4(a), was matched by IVs and NCBI.

Figure 3.

   [65]Figure 3.
   [66]Open in a new tab

   Visualization results of differential analyses on AD. (A) The heatmap
   of differentially expressed genes for AD. The clusters on the left
   y-axis mean the dendrogram on the cluster analysis of gene expression
   patterns. (B) The volcano of differentially expressed genes for AD. The
   genes in the blue and red boxes represent top genes, while the genes in
   the black box are the hub genes.

Figure 4.

   [67]Figure 4.
   [68]Open in a new tab

   The PPI network of 14 overlapped DEGs, and the clusters based on the
   K-means clustering. (A) Two datasets overlapped a total of 14 DEGs. (B)
   Cluster 1 based on the K-means clustering. (C) Cluster 2 based on the
   K-means clustering.

Identification of hub genes

   A total of 14 overlaps of these datasets were identified and visualized
   by Venn diagram. The interactions of the 14 overlapping genes were
   further explored using the STRING database (v 12.0) ([69]Figure 4(a)).
   The criterion confidence = 0.4 was used to filter out unconnected
   nodes, and two clusters were established by the K-means clustering.
   Cluster 1 had four nodes and five edges ([70]Figure 4(b)), and cluster
   2 had two nodes and one edge ([71]Figure 4(c)). Finally, a total of six
   genes were confirmed as the hub genes and presented in [72]Figure 4.

The associations between hub genes and chromosomes

   The association between SNPs and various categories of gut microbes is
   visualized by a Manhattan plot ([73]Figure 5). A total of 14 overlaps
   were clearly marked on the corresponding chromosomes, and cluster 1 and
   cluster 2 were marked in red and green tag. In addition, 1e-05 and
   5e-08 were used as thresholds for marking in the Manhattan plot. The
   hub genes were mainly distributed in chr 1, chr 2, chr 7, chr 8, chr
   14, and chr 22.

Figure 5.

   [74]Figure 5.
   [75]Open in a new tab

   Manhattan plot showed the number of SNPs that associated with the
   various categories of gut microbes. The circumference was the x-axis,
   which was arranged from chromosomes 1 to 22. The radius was the y-axis,
   which represented the results of the relational degree by -log10
   (p-value).

Enrichment analyses for hub genes

   The six hub genes were used to perform GO analyses and KEGG pathway
   analyses. The top three in biological process (BP) were behavior, cell
   adhesion, and plasma membrane bounded cell projection organization. The
   top three in cell component (CC) were synapse, cell junction, and cell
   projection. The top three in molecular function (MF) were cell adhesion
   molecule binding, neuroligin family protein binding, and voltage-gated
   sodium channel activity. Cell adhesion molecules were the top KEGG
   pathway. The results were presented in [76]Figure 6. A dotplot was used
   to visualize the results of the GO analysis ([77]Figure 6(a)) and a
   barplot was used to visualize the results of the KEGG pathway
   ([78]Figure 6(b)).

Figure 6.

   [79]Figure 6.
   [80]Open in a new tab

   Enrichment analyses for hub genes by GO analyses and KEGG pathway
   analyses. (A) The GO enrichment analysis of hub genes. (B) The KEGG
   pathway enrichment analysis of hub genes.

Correspondence between overlapped genes and microbes

   We identified 32 microbial taxa with causal relationships with AD by
   Mendelian randomization, resulting in a total of 544 SNPs. We then
   searched NCBI for each of these SNPs and constrained a ‘gene set’
   dataset. We then used bioinformatic analysis to identify 14 overlapping
   genes. Final, we matched genes with microbial taxa by identical rsIDs.
   The results showed that (Actinobacteriota,Bifidobacterium-DLGAP2),
   (Actinomycetales-NEGR1), (Bifidobacterium adolescentis-NRXN3),
   (CAG-449-TCF12), (Actinomycetales-TLDC2), and (Bifidobacterium
   ruminantium-CFAP46) showed risk factors for statistical significance
   (P < 0.05, OR < 1); (Intestinimonas assiliensis-CNTNAP2),
   (Turicibacter_sp001543345-MYH9), (Parabacteroides johnsonii-SCN3A),
   (Haloplasmatales, Turicibacter_sp001543345, Turicibacter-SNX27),
   (Lachnospiraceae-UBE3C), (CAG-485_sp002362485-ATP10A),
   (Corynebacterium-ARHGEF3), and (Eubacterium callanderi-EIF1B-AS1)
   showed protective factors for statistical significance (p < 0.05,
   OR > 1). The results are shown in [81]Table 1.

Table 1.

   The correspondence between overlapped genes and microbes.
   Overlapped genes SNP Gut microbes OR OR_lci95 OR_uci95
   Cluster 1
    DLGAP2 rs2158787 Actinobacteriota 1.002243 1.0000656 1.004425
   Bifidobacterium 1.0006591 1.0000506 1.001268
    CNTNAP2 rs73170308 Intestinimonas assiliensis 0.9980859 0.996611
   0.999563
    NEGR1 rs17091560 Actinomycetales 1.00111 1.0002277 1.001993
    NRXN3 rs716064 Bifidobacterium adolescentis 1.000559 1.0001569
   1.000962
   Cluster 2
    MYH9 rs117155525 Turicibacter_sp001543345 0.9980859 0.996611 0.999563
    SCN3A rs1899014 Parabacteroides johnsonii 0.9992465 0.9985467
   0.9999468
   Unconnected
    SNX27 rs72692745 Haloplasmatales 0.99921 0.9984588 0.9999618
   Turicibacter_sp001543345 0.9992555 0.9986553 0.9998562
   Turicibacter 0.9985984 0.9978868 0.9993105
    UBE3C rs57210743 Lachnospiraceae 0.9984218 0.9969177 0.9999281
    ATP10A rs117094938 CAG-485_sp002362485 0.9991982 0.998489 0.999908
    TCF12 rs76014583 CAG-449 1.001088 1.0002523 1.001925
    TLDC2 rs116897774 Actinomycetales 1.00111 1.0002277 1.001993
    CFAP46 rs10870212 Bifidobacterium ruminantium 1.000987 1.0002762
   1.001698
    ARHGEF3 rs76907118 Corynebacterium 0.9975662 0.9953066 0.9998308
    EIF1B-AS1 rs11712107 Eubacterium callanderi 0.9986354 0.9973077
   0.9999649
   [82]Open in a new tab

Discussion

   In the present study, the most recent data on the gut microbiome, with
   summary statistics for 473 microbes, were selected for analysis. First,
   we identified 32 gut microbes that were statistically significant in
   the MR analyses, involving one phylum, one order, two families, 10
   genera, and 18 species. Next, a total of 14 overlapping genes were
   identified, six of which were divided into two clusters and considered
   to be hub genes that interact with each other. The mechanisms of gut
   microbiota influence on AD were then revealed by molecular function and
   KEGG pathway enrichment. The pathway with the most enriched genes was
   the Cell Adhesion Molecules (CAMs) pathway, including NRXN3, NEGR1, and
   CNTNAP2. The corresponding taxon of gut microbes was identified as
   Bifidobacterium adolescentis, Actinomycetales, and Intestinimonas
   massiliensis. Finally, we attempted to elucidate the relationships
   between hub genes, gut microbes, and AD.

   Previous studies have reported the associations between gut microbes
   and AD.^[83]23–[84]25 For example, a sequencing study reported that
   there were significant differences in several gut microbes between the
   AD and control brains. In particular, actinobacteria showed a
   relatively higher proportion in the AD group.^ [85]26 The underlying
   mechanism may be that cerebral amyloid deposition, which could induce
   AD, was influenced by manipulation of gut microbes.^ [86]27 Also,
   inflammation and neurotoxicity were induced by gut microbes-mediated
   tau phosphorylation, such as actinobacillus, tannerella forsythensis,
   and helicobacter pylori.^ [87]28 This evidence indicated that bacterial
   infections may be involved in the pathological processes of AD. In this
   study, actinobacteriota and actinomycetales were weak risk factors for
   AD, which was consistent with the evidence mentioned above. In
   addition, Bifidobacterium and Bifidobacterium adolescentis in this
   study also showed weak risk factors for AD, which seemed to be in line
   with some recent research. For example, a systematic review of 11
   studies reported that the abundance of Proteobacteria, Bifidobacterium
   and Phascolarctobacterium was higher in AD patients.^ [88]29 Moreover,
   some studies showed that the abundance of Bifidobacterium was higher in
   Parkinson's disease (PD) cases.^[89]30–[90]34 A systematic review of 42
   studies (26 PD and 16 AD) also shown that there were shared microbes
   serving between PD and AD, showing a higher abundance of which included
   Bifidobacterium, Lachnospiraceae, and Proteobacteria. As PD and AD are
   both belonged to neurodegenerative diseases, these overlapped microbes
   meant the possibility of shared pathogenic mechanisms, which required
   to be explored in depth.^ [91]35

   However, some contradictory results have also been reported. Several
   animal experiments showed that the cognitive dysfunction of Aβ-mediated
   AD model mice could be regulated by Bifidobacterium breve strain A1,
   and inflammation and immune response gene were also inhibited.^ [92]36
   Bifidobacterium longum NK46 was reported to be associated with LPS, and
   the cognitive dysfunction of 5xFAD-transgenic mice could be regulated
   by targeting at the LPS-mediated NF-κB activation.^ [93]37 A randomized
   controlled trial investigating the effect of probiotic on cognition
   showed that MMSE scores improved significantly with Bifidobacterium
   intervention.^ [94]38 In addition, Bifidobacterium belongs to the
   Actinobacteria, both of which have been reported to be less abundant in
   people with AD.^ [95]39 In another meta-analysis of randomized
   controlled trials, cognitive function was not significantly different
   between AD cases and controls in the Bifidobacterium and Lactobacillus
   interventions, meaning that there was no benefit of Bifidobacterium for
   AD.^ [96]40 Such conflicting evidence suggested that more research was
   needed to confirm the influence of Bifidobacterium on AD.

   In the present study, Turicibacter_sp001543345, Parabacteroides
   johnsonii, and intestinimonas massiliensis were identified as weak
   protective factors for AD. Although these microbes have not been
   directly linked to AD, and the abundance of which had some significant
   differences at the genus level. The results of 16S rRNA sequencing
   showed a decrease in Turicibacter in AD individuals,^ [97]39 and
   similar results were also observed in an animal study, further
   indicating that three species of Turicibacter were evidently less
   abundant in the 5xfAD mice model.^ [98]41 It was notable that a
   similarity of 98% sequence was demonstrated between Parabacteroides
   johnsonii and Parabacteroides merdae,^ [99]42 both of which belong to
   the genus Parabacteroides and have been associated with cardiovascular
   damage,^ [100]43 metabolic syndrome, and inflammatory bowel disease.^
   [101]44 Although there was no experimental evidence of an association
   between Parabacteroides and AD, a multivariable MR indicated a causal
   relationship between Parabacteroides and AD, and alanine serums
   mediated the causal effect of Parabacteroides on AD.^ [102]45 Some
   studies have also shown that Parabacteroides was a common bacterial
   genus among PD, stroke and multiple sclerosis,^ [103]46 and a decreased
   abundance of Parabacteroides has been shown in PD cases,^ [104]47 as it
   may block off the effects of Levodopa medication.^ [105]48
   Additionally, it also had an effect on tyrosine metabolism, which has
   been shown to be an effective intervention in stroke.^ [106]49 We
   tended to guess that Parabacteroides may promote the development of AD,
   based on the findings mentioned above, but this needed further
   verification. Similarly, no direct evidence seemed to have been
   searched to support the findings of our study with Intestinimonas. A
   study exploring the interactions between microbes and host immune
   responses in Huntington's disease was considered the most relevant
   literature to explain this finding, the results of which showed that
   Intestinimonas was positively associated with IL-4,^ [107]50 meaning
   that the anti-inflammatory effect associated with neurodegenerative
   diseases may be benefited by Intestinimonas. However, this hypothesis
   needed further verification.

   Some evidence suggested these hub genes may be employed in the
   development of AD. A recent study showed that DLGAP2 was a regulator
   for AD dementia, mainly showing differential expression in declining
   cognition, AD diagnosis, and neuropathology of multiple cerebral
   regions.^ [108]51 And the expression of DLGAP2 may be elevated in the
   brain and served in the process of neuronal cells signaling and
   molecular organization of synapses.^ [109]52 In addition, a number of
   studies indicated that NRXN3,^[110]53,[111]54 NEGR1,^[112]55,[113]56
   and CNTNAP2^ [114]57 were associated with the neural cell communication
   and synapse function, which were the main genes of enrichment in the
   CAMs signaling pathway. Numerous studies have demonstrated that the
   CAMs played a crucial role in the pathogenesis and development of
   AD.^[115]55,[116]58

   It was worth mentioning that NEGR1 has been shown to target the same
   signaling pathways (such as ERK and AKT signaling) synergistically with
   FGFR2.^ [117]59 And more than one study has demonstrated that the ERK
   and the AKT signaling served a crucial role in AD
   neuroprotection.^[118]60–[119]62 For the genes in cluster 2 in this
   paper, although there was no direct evidence of regulatory action in
   AD, numerous studies have shown that MYH9 also targets the ERK and the
   AKT pathways.^[120]63,[121]64 And SCN3A has been reported to be
   involved in PD.^ [122]65 Remarkably, there is no direct experimental
   evidence of the targeting relationship between these microbes and hub
   genes, which needs to be further explored.

   In the present study, we used the Mendelian randomization to avoid the
   confounding factors, particularly environmental exposures and lifestyle
   habits. This approach provided direct evidence of a potential causal
   link between the gut microbes and AD, combining with bioinformatics
   analysis, and further identifying the hub genes linked to AD. These
   findings served as evidence for unraveling potential therapeutic
   targets. However, there were several limitations in this paper. The
   pathogenesis of AD was complex, involving the variation and interaction
   of multiple genes, and it was difficult to fully explain the variation
   of a single gene or SNP. Although we have identified hub genes, their
   specific functions and mechanisms within the context of AD were still
   unclear, what's more, the available data are hard to support the
   relationship between the gut microbes and AD completely, further
   functional validation and rigorous experimental exploration were
   needed. In addition, the availability of high-quality, large-scale
   genome-wide association study data, and the consistency across genetic
   variations and phenotypic profiles in Mendelian randomization analyses,
   these two both limited the depth and accuracy of study. These
   limitations underscore the vast potential for future study.

   In conclusion, the hub genes (including DLGAP2, NRXN3, NEGR1, CNTNAP2,
   MYH9, and SCN3A) were identified to associate with AD. The taxons of
   gut microbes that corresponded to these were Actinobacteriota,
   Actinomycetales, Bifidobacterium, Bifidobacterium adolescentis,
   Intestinimonas massiliensis, Turicibacter_sp001543345, and
   Parabacteroides johnsonii. Especially, NRXN3 (Bifidobacterium
   adolescentis), NEGR1 (Actinomycetales) and CNTNAP2 (Intestinimonas
   massiliensis) may promote the progress of AD through regulating the
   CAMs pathway-mediated synaptic function.

Supplemental Material

   sj-docx-1-alr-10.1177_25424823241310719 - Supplemental material for Gut
   microbiota may affect Alzheimer's disease through synaptic function
   mediated by CAMs pathway: A study combining Mendelian randomization and
   bioinformatics
   [123]sj-docx-1-alr-10.1177_25424823241310719.docx^ (22.9KB, docx)

   Supplemental material, sj-docx-1-alr-10.1177_25424823241310719 for Gut
   microbiota may affect Alzheimer's disease through synaptic function
   mediated by CAMs pathway: A study combining Mendelian randomization and
   bioinformatics by Ji-yun Liu, Cong-yan Tan, Li Luo and Xuan Yin in
   Journal of Alzheimer's Disease Reports

Acknowledgments