Biological Psychiatry 生物精神病学

Available online 1 March 2024
2024 年 3 月 1 日上网
Biological Psychiatry

Archival Report 档案报告
The Causal Relationships Between Gut Microbiota, Brain Volume, and Intelligence: A Two-Step Mendelian Randomization Analysis

https://doi.org/10.1016/j.biopsych.2024.02.1012 Get rights and content 获取权利和内容

Abstract 摘要

Background 背景介绍

Growing evidence indicates that dynamic changes in gut microbiome can affect intelligence; however, whether these relationships are causal remains elusive. We aimed to disentangle the poorly understood causal relationship between gut microbiota and intelligence.

Methods 方法

We performed a 2-sample Mendelian randomization (MR) analysis using genetic variants from the largest available genome-wide association studies of gut microbiota (N = 18,340) and intelligence (N = 269,867). The inverse-variance weighted method was used to conduct the MR analyses complemented by a range of sensitivity analyses to validate the robustness of the results. Considering the close relationship between brain volume and intelligence, we applied 2-step MR to evaluate whether the identified effect was mediated by regulating brain volume (N = 47,316).
我们利用现有最大的肠道微生物群(18,340 个)和智力(269,867 个)全基因组关联研究中的遗传变异进行了双样本孟德尔随机化(MR)分析。在进行MR分析时采用了反方差加权法,并辅以一系列敏感性分析,以验证结果的稳健性。考虑到脑容量与智力之间的密切关系,我们采用了两步磁共振分析法,以评估所发现的效应是否通过调节脑容量来介导(样本数=47,316)。

Results 成果

We found a risk effect of the genus Oxalobacter on intelligence (odds ratio = 0.968 change in intelligence per standard deviation increase in taxa; 95% CI, 0.952–0.985; p = 1.88 × 10−4) and a protective effect of the genus Fusicatenibacter on intelligence (odds ratio = 1.053; 95% CI, 1.024–1.082; p = 3.03 × 10−4). The 2-step MR analysis further showed that the effect of genus Fusicatenibacter on intelligence was partially mediated by regulating brain volume, with a mediated proportion of 33.6% (95% CI, 6.8%–60.4%; p = .014).
我们发现牛杆菌属对智力有风险效应(类群每增加一个标准差,智力变化的几率比=0.968;95% CI,0.952-0.985;p = 1.88 × 10 −4 ),而镰刀菌属对智力有保护效应(几率比=1.053;95% CI,1.024-1.082;p = 3.03 × 10 −4 )。两步磁共振分析进一步表明,镰刀菌属对智力的影响部分是通过调节脑容量来介导的,介导比例为 33.6% (95% CI, 6.8%-60.4%; p = .014)。

Conclusions 结论

Our results provide causal evidence indicating the role of the microbiome in intelligence. Our findings may help reshape our understanding of the microbiota-gut-brain axis and development of novel intervention approaches for preventing cognitive impairment.

Keywords 关键词

Brain volume
Cognitive ability
Genetic variants
Gut microbiome
Mendelian randomization


Intelligence, also known as cognitive ability, is a robust predictor of educational and socioeconomic achievement and has been broadly associated with lifestyle behaviors and health resource advantages across the life span (1, 2, 3, 4). Establishing causality and prioritizing targets responsible for individual differences in intelligence is one of the key challenges in psychological and brain sciences. Currently, emerging evidence recognizes gut microbiota as an essential component of normal physiology, with an important role in both brain development and function (5, 6, 7).
智力又称认知能力,是教育和社会经济成就的有力预测指标,并与生活方式行为和一生中的健康资源优势广泛相关(1, 2, 3, 4)。确定因果关系并优先考虑造成智力个体差异的目标是心理和脑科学面临的主要挑战之一。目前,新出现的证据表明,肠道微生物群是正常生理的重要组成部分,在大脑发育和功能方面发挥着重要作用(5、6、7)。

The gut microbiome is a highly complex and diverse hidden kingdom and plays a fundamental role in gut-brain communication. Growing evidence indicates that alterations in the gut microbiome can affect neurodevelopment and cognitive ability (8,9). Early-life antibiotic exposure is associated with subsequent worse neurocognitive outcomes (10,11). In contrast, the administration of probiotic strains has yielded controversial results in terms of cognitive changes. For example, probiotic ingestion has been reported to improve sustained attention and working memory in older adult participants (12), while an early study reported potential cognitive impairments of probiotic consumption (13). For the specific taxa, a multi-omics integration analysis revealed that 3 genera (Odoribacter, Butyricimonas, and Bacteroides) exhibited a positive association with improved cognitive performance (14). Additionally, the abundance of Odoribacter was linked to several important features of brain structure and volumes (14). Furthermore, a recent metagenomic association analysis found that bacteria with the ability to produce short-chain fatty acids, including Bacteroides massiliensis and Fusicatenibacter saccharivorans, were found to be positively correlated with improved cognitive performance (15). Despite growing evidence linking gut microbiome composition and cognitive ability, evidence about causal relationships is still scarce. Moreover, current conclusions are mainly based on conventional observational studies, which can be affected by a variety of confounding factors, such as diet. It is critical to explore the potential causal relationship between gut microbiome composition and intelligence.
肠道微生物组是一个高度复杂和多样化的隐蔽王国,在肠道与大脑的交流中扮演着重要角色。越来越多的证据表明,肠道微生物组的改变会影响神经发育和认知能力(8,9)。早年接触抗生素与随后神经认知能力下降有关(10,11)。与此相反,服用益生菌株对认知能力的改变却产生了有争议的结果。例如,有报道称摄入益生菌可改善老年人的持续注意力和工作记忆(12),而一项早期研究则称摄入益生菌可能会损害认知能力(13)。就特定类群而言,一项多组学整合分析显示,3 个菌属(Odoribacter、Butyricimonas 和 Bacteroides)与认知能力的提高呈正相关(14)。此外,Odoribacter 的丰度与大脑结构和体积的几个重要特征有关(14)。此外,最近的一项元基因组关联分析发现,具有产生短链脂肪酸能力的细菌(包括 Bacteroides massiliensis 和 Fusicatenibacter saccharivorans)与认知能力的提高呈正相关(15)。尽管有越来越多的证据表明肠道微生物组的组成与认知能力有关,但有关因果关系的证据仍然很少。此外,目前的结论主要基于传统的观察性研究,而这些研究可能会受到饮食等多种混杂因素的影响。探索肠道微生物组组成与智力之间的潜在因果关系至关重要。

Randomized controlled trials of gut microbiota have the potential to establish causal relationships. However, most randomized controlled trials are expensive and time-consuming, and more importantly, gut microbiome composition cannot be randomly allocated in practice. Alternatively, Mendelian randomization (MR) uses genetic variants as instrumental variables to investigate the causal associations between modifiable exposures and outcomes (16). Genetic variants are distributed randomly during meiosis and fertilization, making them largely independent of self-selected behaviors, thereby circumventing bias from confounding factors and reverse causality. Large-scale genome-wide association studies (GWASs) on the gut microbiome and intelligence provide the opportunity for MR analysis with significantly improved statistical power (17,18).
肠道微生物群随机对照试验有可能建立因果关系。然而,大多数随机对照试验既昂贵又耗时,更重要的是,肠道微生物组的组成在实践中无法随机分配。另外,孟德尔随机化(Mendelian randomization,MR)利用遗传变异作为工具变量来研究可改变的暴露与结果之间的因果关系(16)。遗传变异在减数分裂和受精过程中随机分布,因此在很大程度上与自选行为无关,从而避免了混杂因素和反向因果关系造成的偏差。关于肠道微生物组和智力的大规模全基因组关联研究(GWASs)为 MR 分析提供了机会,并显著提高了统计能力(17,18)。

In the current study, we performed a bidirectional 2-sample MR analysis to investigate the causal relationships between the composition of 211 gut microbiota (consisting of 131 genera, 35 families, 20 orders, 16 classes, and 9 phyla) and human intelligence. We identified 2 putative causal associations. Considering the close relationship between brain volume and human intelligence (19,20), we further conducted a 2-step MR analysis to explore whether the effect of the identified taxa on intelligence was mediated by regulating brain volume. Our findings may provide insight into early-stage interventions for cognitive ability at the gut-microbiome level.
在本研究中,我们进行了双向 2 样本 MR 分析,以研究 211 个肠道微生物群(包括 131 属、35 科、20 目、16 类和 9 门)的组成与人类智力之间的因果关系。我们发现了两种可能的因果关系。考虑到脑容量与人类智力之间的密切关系(19,20),我们进一步进行了两步磁共振分析,以探讨所发现的类群对智力的影响是否通过调节脑容量来介导。我们的发现可能会为在肠道微生物组水平上对认知能力进行早期干预提供启示。

Methods and Materials 方法与材料

Study Overview 研究概况

An overview of the study is shown in Figure 1. First, we used single nucleotide polymorphisms (SNPs) derived from the summary-level data as genetic instruments for the risk factor. Summary information on the data sources and sample sizes used in this study can be found in Table S1. Then, we performed a 2-sample bidirectional MR to assess the causal effect of each gut microbiome composition on intelligence and vice versa. Finally, we used a 2-step MR analysis to assess whether brain volume and neuroimaging phenotypes play a causal role in mediating the pathway linking the identified gut microbiome composition and intelligence. This study relied on summary-level data that have been made publicly available; ethical approval was obtained in all original studies.
研究概况见图 1。首先,我们使用从汇总级数据中得出的单核苷酸多态性(SNPs)作为风险因素的遗传工具。本研究中使用的数据来源和样本量的摘要信息见表 S1。然后,我们进行了 2 样本双向 MR,以评估每种肠道微生物组成分对智力的因果效应,反之亦然。最后,我们使用两步磁共振分析来评估脑容量和神经影像表型是否对已确定的肠道微生物组组成与智力之间的联系途径起着因果中介作用。本研究依赖于已公开的摘要级数据;所有原始研究均已获得伦理批准。

  1. Download : Download high-res image (536KB)
    下载 :下载高清图片 (536KB)
  2. Download : Download full-size image
    下载 :下载 : 下载全尺寸图片

Figure 1. Study workflow. GWAS, genome-wide association study.
图 1.研究工作流程。GWAS:全基因组关联研究。

Data Sources 数据来源

Gut Microbiome Composition

The genetic information for gut microbiome composition was obtained through the largest GWAS meta-analysis to date conducted by the MiBioGen consortium (17). The study involved the coordination of 16S ribosomal RNA gene sequencing and genetic profiling of 18,340 individuals from 24 cohorts, most of whom were of European ancestry (17). A total of 211 taxa at 6 levels (131 genera, 35 families, 20 orders, 16 classes, and 9 phyla) were ultimately analyzed.
肠道微生物组组成的遗传信息是通过 MiBioGen 联合体(17)进行的迄今为止最大的 GWAS 元分析获得的。这项研究对来自 24 个队列的 18,340 人进行了 16S 核糖体 RNA 基因测序和基因图谱分析,其中大多数人的祖先是欧洲人(17)。最终共分析了 6 个级别(131 属、35 科、20 目、16 类和 9 门)的 211 个类群。

Intelligence 情报

The genetic associations for intelligence were derived from a GWAS meta-analysis that encompassed 269,867 individuals of European ancestry across 14 independent cohorts (18). Intelligence was measured through various neurocognitive assessments, such as verbal and mathematical fluid intelligence tested in the UK Biobank (UKB) (21). However, different measures were operationalized to index a common latent factor, which was labeled general intelligence or Spearman’s g (22), also known as the positive manifold of cognitive ability or intelligence.
智力的遗传关联来自一项全球基因组研究荟萃分析(GWAS meta-analysis),该分析涵盖了 14 个独立队列中的 269 867 名欧洲血统个体(18)。智力通过各种神经认知评估进行测量,如英国生物库(UKB)测试的语言和数学流体智力(21)。然而,不同的测量方法被操作化为一个共同的潜在因子,该因子被标记为一般智力或斯皮尔曼 g(22),也被称为认知能力或智力的正向流形。

Brain Volume 脑容量

The genetic variants associated with brain volume were derived from meta-analysis results of brain volume in the UKB, as well as 2 additional GWASs on intracranial volume and head circumference, which are both considered proxy measures for brain volume (19). A total of 17,062 participants were included in the GWAS analyses in the UKB. The GWAS results of intracranial volume conducted by the ENIGMA (Enhancing Neuro Imaging Genetics through Meta Analysis) consortium included 11,373 participants, and the results for head circumference were based on a total of 18,881 participants, which resulted in a combined sample size of 47,316 (23,24).
与脑容量相关的遗传变异来自英国脑容量荟萃分析结果,以及另外两项关于颅内容量和头围的基因组研究,这两项研究都被认为是脑容量的替代指标(19)。共有17,062名参与者被纳入英国脑容量基因组研究分析。ENIGMA(通过元分析增强神经成像遗传学)联盟进行的颅内容积 GWAS 结果包括 11,373 名参与者,头围结果基于 18,881 名参与者,合计样本量为 47,316 个(23,24)。

Neuroimaging Phenotypes 神经影像表型

Summary statistics for neuroimaging phenotypes were obtained from Warrier et al. (25). This study conducted a GWAS of 13 magnetic resonance imaging–derived metrics. Each metric was measured at 180 regions in 36,843 individuals from the UKB and the ABCD (Adolescent Brain Cognitive Development) cohorts, resulting in 2347 neuroimaging phenotypes.
神经影像表型的汇总统计来自 Warrier 等人(25)。这项研究对 13 个磁共振成像衍生指标进行了 GWAS 分析。每项指标都在来自 UKB 和 ABCD(青少年大脑认知发展)队列的 36,843 人的 180 个区域进行了测量,得出了 2347 种神经影像表型。

Genetic Instruments Selection

Selecting Genetic Instruments

The genetic instruments employed had to fulfill 3 assumptions (26): 1) the genetic variants should be strongly associated with the exposure, 2) the genetic variants should not be associated with any potential confounding factors, and 3) the genetic variants should not affect the outcome independently of exposure. We removed variants with minor allele frequency < 0.01 in the GWAS dataset and used the clump function in PLINK software to identify independent SNPs for each exposure, using the 1000 Genomes European data as the reference (27). A strict cutoff of r2 < 0.001, a window of 10,000 kb, and a p < 5 × 10−8 were used for clumping. It is worth noting that we used a relaxed p-value threshold of 1 × 10−5 for gut microbiome composition, similar to previous MR studies (17,28,29) because SNPs below this threshold were found to have the largest explained variance on microbial features (30). To evaluate whether these variants could capture the genetic association within the region, we performed statistical fine-mapping by applying FINEMAP (31) (see the Supplement) to each genomic window covering the association signals. To ensure consistency, we harmonized the effects of SNPs on both exposure and outcome by aligning the beta values to the same alleles. Where shared SNPs between exposure and outcome were not available, we replaced them with proxy SNPs (r2 > 0.8) that were significantly associated with the exposure.
采用的基因工具必须满足 3 个假设(26):1)基因变异应与暴露密切相关;2)基因变异不应与任何潜在混杂因素相关;3)基因变异不应独立于暴露而影响结果。我们剔除了 GWAS 数据集中小等位基因频率小于 0.01 的变异,并以欧洲 1000 基因组数据为参考(27),使用 PLINK 软件中的 clump 功能来识别每种暴露的独立 SNP。聚类时使用了严格的 r 2 < 0.001 临界值、10,000 kb 窗口和 p < 5 × 10 −8 。值得注意的是,我们对肠道微生物组组成采用了 1 × 10 −5 的宽松 p 值阈值,这与之前的 MR 研究(17,28,29)相似,因为研究发现,低于该阈值的 SNPs 对微生物特征的解释方差最大(30)。为了评估这些变异是否能捕捉到该区域内的遗传关联,我们对覆盖关联信号的每个基因组窗口应用 FINEMAP (31)(见附录)进行了统计精细映射。为确保一致性,我们通过将贝塔值与相同等位基因对齐来协调 SNP 对暴露和结果的影响。如果没有暴露和结果之间的共享 SNP,我们就用与暴露显著相关的替代 SNP(r 2 > 0.8)来替代。

Removing Confounders 消除混杂因素

To avoid potential confounding, we removed SNPs that were significantly associated with plausible confounders in the PhenoScanner database in European participants (32,33). Four potential confounders were taken into account, including diet, socioeconomic status, drinking, and smoking behavior. These traits have been reported to affect both gut microbiome composition (34,35) and intelligence or cognitive ability (36, 37, 38, 39).
为了避免潜在的混杂因素,我们剔除了欧洲参与者 PhenoScanner 数据库中与可能的混杂因素显著相关的 SNPs(32,33)。我们考虑了四个潜在的混杂因素,包括饮食、社会经济地位、饮酒和吸烟行为。据报道,这些特征会影响肠道微生物组的组成(34,35)和智力或认知能力(36,37,38,39)。

Quality Control of Genetic Instruments

We excluded palindromic SNPs with intermediate allele frequencies (>0.42), which would introduce potential strand-flipping issues. To enhance the accuracy and robustness of the remaining genetic instruments, we also removed outlier pleiotropic SNPs detected by RadialMR (40). RadialMR identified outlier pleiotropic SNPs using a heterogeneity test (modified Q statistics) with a nominal significance level of .05. F statistics were calculated to estimate the strength of genetic instruments (see the Supplement). Only SNPs with F statistics > 10 were included in the MR analysis. The statistical power of the remaining genetic instruments was then calculated according to the method described by Burgess (41). After removing confounders and quality control, all 211 taxa on which we performed MR analyses had at least 3 SNPs.
我们排除了具有中等等位基因频率(>0.42)的等位基因SNP,因为它们会带来潜在的链翻转问题。为了提高其余遗传工具的准确性和稳健性,我们还剔除了 RadialMR(40)检测到的离群多向性 SNP。RadialMR 使用异质性检验(修正的 Q 统计量)识别离群的多向性 SNP,标称显著性水平为 0.05。计算 F 统计量是为了估计遗传工具的强度(见附录)。只有 F 统计量大于 10 的 SNP 才被纳入 MR 分析。然后根据 Burgess(41)描述的方法计算其余遗传工具的统计能力。去除混杂因素和质量控制后,我们进行 MR 分析的所有 211 个类群都至少有 3 个 SNPs。

Statistical Analyses 统计分析

Two-Sample MR 双样本 MR

Analyses were performed to test whether gut microbiome composition causally affects intelligence and whether intelligence can causally affect gut microbiome composition. We used the inverse-variance weighted (IVW) method based on a multiplicative random-effects model as the primary causal inference (42). Because the IVW estimates can be biased if pleiotropic instrumental variables are introduced (43), we estimated causality using 4 additional methods in parallel to enhance the reliability of our results (Supplement), including robust adjusted profile score (MR RAPS) (44), weighted median (45), weighted mode (46), and MR-Egger (47). The p values from the IVW MR test were adjusted using Benjamini-Hochberg false discovery rate correction for multiple testing; for the resulting q value, the significance threshold was set to .05. Because of the large number and hierarchical structure of the taxa used in our study, the multiple comparison adjustments may be excessive. We also report nominally significant results (p < .05) in the Supplement.
我们进行了分析,以检验肠道微生物组的组成是否会对智力产生因果影响,以及智力是否会对肠道微生物组的组成产生因果影响。我们使用基于乘法随机效应模型的逆方差加权(IVW)方法作为主要的因果推断方法(42)。由于如果引入褶状工具变量(43),IVW 估计值可能会出现偏差,因此我们同时使用了另外 4 种方法来估计因果关系,以提高结果的可靠性(补编),包括稳健调整剖面得分(MR RAPS)(44)、加权中位数(45)、加权模式(46)和 MR-Egger(47)。IVW MR 检验的 p 值使用本杰明-霍奇伯格多重检验假发现率校正法进行调整;对于得出的 q 值,显著性阈值设为 0.05。由于我们研究中使用的分类群数量庞大且具有层次结构,多重比较调整可能会过度。我们还在补编中报告了名义上显著的结果(p < .05)。

Sensitivity Analysis 敏感性分析

We conducted a series of sensitivity analyses to address the potential issue of pleiotropy in the causal estimates. First, we used MR-Egger regression to assess the presence of horizontal pleiotropy based on its intercept term; deviation from 0 (p < .05) was considered evidence for directional pleiotropic bias (47). Second, we used MR-PRESSO to detect the presence of pleiotropy (p < .05) (48). MR-PRESSO compares the observed distance of all the variants to the regression line with the expected distance under the null hypothesis of no horizontal pleiotropy. Third, we assessed heterogeneity using Cochran’s Q statistic (49), which is produced by different genetic variants in the fixed-effect variance weighted analysis; a p value of <.05 indicated the presence of pleiotropy. We conducted a leave-one-out analysis to determine whether the causal association was driven by an individual variant. In addition to bidirectional MR, the MR Steiger test of directionality was performed to estimate whether the assumption that exposure causes outcome is valid.
我们进行了一系列敏感性分析,以解决因果关系估计中潜在的多向性问题。首先,我们使用 MR-Egger 回归根据截距项来评估是否存在水平多向性;偏离 0(p < .05)被认为是定向多向性偏倚的证据(47)。其次,我们使用 MR-PRESSO 检测是否存在多向性(p < .05)(48)。MR-PRESSO 将观察到的所有变体与回归线的距离与无水平多向性的假说下的预期距离进行比较。第三,我们使用 Cochran's Q 统计量(49)评估了异质性,该统计量是由固定效应方差加权分析中的不同遗传变异产生的;P 值小于 0.05 表示存在多向性。我们进行了撇除分析,以确定因果关联是否由单个变异体驱动。除了双向 MR 外,我们还进行了 MR Steiger 方向性检验,以估计暴露导致结果的假设是否成立。

Mediation Analysis 调解分析

A 2-step MR analysis was performed to evaluate whether the effect of identified taxa on intelligence was mediated by regulating brain volume. In the first step, we estimated the causal effect of specific gut microbiome composition on brain volume. In the second step, we assessed the causal effect of brain volume on intelligence. The indirect effect of identified gut microbiome composition on intelligence through brain volume was evaluated using the product of coefficients method (50). To determine the proportion of the effect of the contribution of identified taxa on intelligence that was mediated by regulating brain volume, we divided the indirect effect by the total effect. Standard errors for the indirect effect were obtained using the delta method (51).

Analytical Tools 分析工具

We used R for data analysis. All the analyses in this study were conducted using R packages TwoSampleMR (version 0.4.26), MRPRESSO (version 1.0), and RadialMR (version 1.0).
我们使用 R 进行数据分析。本研究的所有分析均使用 R 软件包 TwoSampleMR(0.4.26 版)、MRPRESSO(1.0 版)和 RadialMR(1.0 版)进行。

Results 成果

Causal Effects of Gut Microbiota Composition on Intelligence

We conducted a 2-sample MR analysis to investigate the impact of gut microbiome abundance on intelligence. The IVW analyses revealed 29 taxa causally associated with intelligence at a nominally significant level (p < .05) (Table S2). More importantly, the genetic liability for 2 specific taxa, namely the genus Oxalobacter and the genus Fusicatenibacter, achieved statistical significance after false discovery rate correction. An abundance of genus Oxalobacter was negatively associated with intelligence (IVW odds ratio [OR] = 0.968; 95% CI, 0.952–0.985; p = 1.88 × 10−4) (Table 1). The estimates were similar in size in MR RAPS. We also found causal evidence that the abundance of genus Fusicatenibacter was positively associated with intelligence (IVW OR = 1.053; 95% CI, 1.024–1.082; p = 3.03 × 10−4) (Table 1). These results were further supported by MR RAPS and weighted median. The confidence intervals of the weighted mode and MR-Egger methods were wider than other methods, which could be attributed to their lower statistical power when compared with IVW (46). Scatterplots of SNP effects on these 2 taxa versus intelligence are presented in Figure 2, with colored lines representing the slopes of different MR analyses. Forest plots of individual and combined SNP MR-estimated effect sizes are also presented.
我们进行了双样本 MR 分析,以研究肠道微生物组丰度对智力的影响。IVW分析显示,29个类群与智力有因果关系,且具有名义上的显著性水平(p < .05)(表S2)。更重要的是,2 个特定类群(即牛杆菌属和 Fusicatenibacter 属)的遗传责任在经过错误发现率校正后达到了统计学意义。牛杆菌属的丰度与智力呈负相关(IVW 比值比 [OR] = 0.968; 95% CI, 0.952-0.985; p = 1.88 × 10 −4 )(表 1)。MR RAPS 的估计值大小相似。我们还发现了Fusicatenibacter属丰度与智力正相关的因果关系证据(IVW OR = 1.053; 95% CI, 1.024-1.082; p = 3.03 × 10 −4 )(表1)。MR RAPS 和加权中位数进一步证实了这些结果。加权模式和 MR-Egger 方法的置信区间比其他方法宽,这可能是由于与 IVW 相比,它们的统计能力较低(46)。SNP 对这两个类群的影响与智力的散点图见图 2,彩色线代表不同 MR 分析的斜率。图中还显示了单个和组合 SNP MR 估计效应大小的森林图。

Table 1. Significant MR Results for the Relationship Between Gut Microbiota and Intelligence
表 1.肠道微生物群与智力关系的显著 MR 结果

MethodNumber of SNPs SNPs 数量F Statistic F 统计OR (95% CI)p Value p 值q Value q 价值
Effect of Genus Oxalobacter on Intelligence
IVW1099.90.968 (0.952–0.985) 0.968 (0.952-0.985)1.88 × 10−4 1.88 × 10 −4 .040
MR RAPS 先生 RAPS0.968 (0.950–0.986) 0.968 (0.950-0.986)6.85 × 10−4 6.85 × 10 −4
Weighted Median 加权中位数0.982 (0.959–1.005) 0.982 (0.959-1.005).115
Weighted Mode 加权模式0.985 (0.949–1.022) 0.985 (0.949-1.022).377
MR Egger 埃格先生0.999 (0.918–1.087) 0.999 (0.918-1.087).973
Effect of Genus Fusicatenibacter on Intelligence
Fusicatenibacter 菌属对智力的影响
IVW1424.81.053 (1.024–1.082) 1.053 (1.024-1.082)3.03 × 10−4 3.03 × 10 −4 .032
MR RAPS 先生 RAPS1.053 (1.021–1.087) 1.053 (1.021-1.087)1.01 × 10−3 1.01 × 10 −3
Weighted Median 加权中位数1.050 (1.010–1.091) 1.050 (1.010-1.091).014
Weighted Mode 加权模式1.015 (0.948–1.087) 1.015 (0.948-1.087).643
MR Egger 埃格先生1.002 (0.896–1.121) 1.002 (0.896-1.121).964

p Values from the IVW MR test were adjusted using the Benjamini-Hochberg false discovery rate correction; for the resulting q value, the threshold was set to .05.
使用本杰明-霍奇伯格错误发现率校正法对 IVW MR 检验的 p 值进行调整;对于得出的 q 值,阈值设为 0.05。

IVW, inverse-variance weighted; MR, Mendelian randomization; OR, odds ratio; RAPS, robust adjusted profile score; SNP, single nucleotide polymorphism.

  1. Download : Download high-res image (640KB)
    下载 :下载高清图片(640KB)
  2. Download : Download full-size image
    下载 :下载 : 下载全尺寸图片

Figure 2. Mendelian randomization (MR) plots for the relationship of gut microbiome composition to intelligence. (A) Scatterplot of single nucleotide polymorphism (SNP) effects on genus Oxalobacter vs. intelligence, with the slope of each line corresponding to the estimated MR effect per method. Data are expressed as raw β values with 95% CI. (B) Forest plot of individual and combined SNP MR-estimated effect sizes for genus Oxalobacter on intelligence. (C) Scatterplot of SNP effects on genus Fusicatenibacter vs. intelligence. (D) Forest plot of individual and combined SNP MR-estimated effects sizes for the effect of genus Fusicatenibacter on intelligence. IVW, inverse-variance weighted; RAPS, robust adjusted profile score.
图 2.肠道微生物组组成与智力关系的孟德尔随机化(MR)图。(A) 单核苷酸多态性(SNP)对牛杆菌属的影响与智力的散点图,每条线的斜率对应于每种方法的估计 MR 影响。数据以原始 β 值和 95% CI 表示。(B) 单个和组合 SNP MR 对牛杆菌属智力影响大小的森林图。(C) SNP 对 Fusicatenibacter 属与智力的影响散点图。(D) Fusicatenibacter 属对智力影响的单个和合并 SNP MR 估计效应大小的森林图。IVW:逆方差加权;RAPS:稳健调整的特征得分。

All SNPs selected for inclusion and exclusion for these 2 taxa are presented in Tables S3 to S6 for replication. After selection, 10 and 14 SNPs significantly associated with Oxalobacter and Fusicatenibacter, respectively, were used in the MR analyses (see the Supplement). The F statistics for the genetic instruments indicated an absence of weak instrument bias (Tables S3 and S5). With the current sample size and given an estimated 5.2% (Oxalobacter) and 1.9% (Fusicatenibacter) of the phenotypic variance of gut microbiome composition, our study had sufficient power (>90%) when the expected causal effect for intelligence was ≥0.03 and ≥0.05, respectively (Table S7). We performed statistical fine-mapping and found that half of the selected genetic instruments were prioritized as potential causal variants (Table S8). Sensitivity analyses did not address any pleiotropy in the causal estimates (Table S9). Leave-one-out analysis revealed that none of the SNPs were responsible for driving the MR results (Figure S1), and funnel plots indicated that causal associations were less likely to be influenced by potential biases with SNPs symmetrically distributed (Figure S2). We found no evidence of reverse causality across the analyses in the MR Steiger test (Table S10).
表 S3 至表 S6 列出了这两个类群中所有被选入和排除的 SNPs,以供参考。经过筛选,分别有 10 个和 14 个与 Oxalobacter 和 Fusicatenibacter 显著相关的 SNPs 被用于 MR 分析(见附录)。遗传工具的 F 统计表明不存在弱工具偏差(表 S3 和 S5)。根据目前的样本量,并考虑到肠道微生物组组成的表型变异估计分别为 5.2%(牛杆菌)和 1.9%(镰刀菌),当智力的预期因果效应分别≥0.03 和≥0.05 时,我们的研究具有足够的功率(>90%)(表 S7)。我们进行了统计精细映射,发现所选遗传工具中有一半被优先列为潜在的因果变异(表 S8)。敏感性分析没有发现因果关系估计值中存在任何褶积现象(表 S9)。剔除分析表明,没有一个 SNPs 是导致 MR 结果的原因(图 S1),漏斗图显示,在 SNPs 对称分布的情况下,因果关联不太可能受到潜在偏差的影响(图 S2)。在 MR Steiger 检验中,我们在所有分析中都没有发现反向因果关系的证据(表 S10)。

Causal Effects of Intelligence on Gut Microbiota Composition

With genetic liability for intelligence as the exposure, we performed MR analyses to explore the causal effect of intelligence on the abundance of the gut microbiome. The SNPs that were included and excluded for intelligence are presented in Tables S11 to S13 to replicate our findings. We found no evidence of causal relationships for intelligence with Oxalobacter (IVW OR = 1.017; 95% CI, 0.838–1.234 p = .864) or Fusicatenibacter (IVW OR = 1.044; 95% CI, 0.944–1.155; p = .400) (Table 2; Figure S3). Similar effect patterns were observed across other MR methods. Additionally, the F statistics of the genetic instruments suggested a lack of weak instrument bias (Table 2; Table S11). Sensitivity analyses did not address potential pleiotropy in the causal estimates (Table S9; Figures S4 and S5). Although there was no causal evidence after multiple testing correction, we found that the genetic liability for intelligence made a causal contribution to the abundance of 15 gut microbiomes at a nominally significant level (p < .05) (Table S14).
以智力的遗传责任作为暴露,我们进行了磁共振分析,以探讨智力对肠道微生物组丰度的因果效应。表 S11 至 S13 列出了因智力而纳入和排除的 SNPs,以复制我们的研究结果。我们没有发现智力与牛杆菌(IVW OR = 1.017; 95% CI, 0.838-1.234 p = .864)或镰刀菌(IVW OR = 1.044; 95% CI, 0.944-1.155; p = .400)之间存在因果关系的证据(表 2;图 S3)。其他 MR 方法也观察到类似的效应模式。此外,遗传工具的 F 统计表明缺乏微弱的工具偏倚(表 2;表 S11)。敏感性分析并未解决因果关系估计中潜在的多义性问题(表 S9;图 S4 和 S5)。虽然多重检验校正后没有因果关系证据,但我们发现智力的遗传因子对 15 个肠道微生物组的丰度有因果关系,且具有名义上的显著性水平(p < .05)(表 S14)。

Table 2. Bidirectional MR Results for the Relationship Between Intelligence and Gut Microbiota
表 2.智力与肠道微生物群关系的双向 MR 结果

MethodNumber of SNPs SNPs 数量F Statistic F 统计OR (95% CI)p Value p 值q Value q 价值
Effects of Intelligence on Genus Oxalobacter
IVW14442.81.017 (0.838–1.234) 1.017 (0.838-1.234).8641
MR RAPS 先生 RAPS1.025 (0.838–1.254) 1.025 (0.838-1.254).811
Weighted Median 加权中位数1.169 (0.884–1.545) 1.169 (0.884-1.545).273
Weighted Mode 加权模式1.868 (0.847–4.118) 1.868 (0.847-4.118).121
MR Egger 埃格先生1.263 (0.430–3.711) 1.263 (0.430-3.711).669
Effects of Intelligence on Genus Fusicatenibacter
IVW14741.71.044 (0.944–1.155) 1.044 (0.944-1.155).4001
MR RAPS 先生 RAPS1.055 (0.949–1.173) 1.055 (0.949-1.173).320
Weighted Median 加权中位数1.026 (0.892–1.181) 1.026 (0.892-1.181).715
Weighted Mode 加权模式1.002 (0.621–1.617) 1.002 (0.621-1.617).993
MR Egger 埃格先生0.948 (0.523–1.718) 0.948 (0.523-1.718).859

p Values from the IVW MR test were adjusted using Benjamini-Hochberg false discovery rate correction; for the resulting q value, the threshold was set to .05.
使用本杰明-霍奇伯格假发现率校正法对 IVW MR 检验的 p 值进行了调整;所得 q 值的临界值定为 0.05。

IVW, inverse-variance weighted; MR, Mendelian randomization; OR, odds ratio; RAPS, robust adjusted profile score; SNP, single nucleotide polymorphism.

Mediation Analyses 调解分析

Considering the close relationship between brain volume and human intelligence (19,20), we performed a 2-step MR analysis to investigate whether the effect of identified taxa on intelligence was mediated by regulating brain volume. First, we estimated the causal effect of identified taxa on brain volume using genetic instruments specific to Oxalobacter and Fusicatenibacter. Summary information on brain volume for the SNPs associated with these 2 taxa is listed in Tables S15 and S16. We identified that increased Fusicatenibacter was associated with increased brain volume (IVW beta = 0.086; 95% CI, 0.019–0.153; p = .012) (Figure 3A; Figure S6). We found no evidence of causal relationships for Oxalobacter on brain volume (IVW beta = 0.006; 95% CI, −0.033 to 0.046; p = .745). Then, we assessed the causal effect of brain volume on intelligence using genetic instruments that were associated with brain volume (Tables S17 and S18). We found strong evidence for causal effects of brain volume on intelligence (IVW beta = 0.199; 95% CI, 0.163–0.236; p = 4.32 × 10−27) (Figure 3B; Figure S7). Sensitivity analyses did not address any pleiotropy in the causal estimates (Table S19; Figures S8–S10). Because both brain volume and intelligence included subjects from the UKB, sample overlap between the exposure and outcome GWASs may bias the estimate toward the confounded estimate. We estimated the bias due to participant overlap using the method described by Burgess et al. (52). We found that a maximum of 6.3% of brain volume participants overlapped with the intelligence GWAS (17,062/269,867), and the estimated bias was <0.82% in the MR study, indicating that our MR results were less likely to be affected by sample overlap between brain volume and intelligence. Finally, we revealed that the genus Fusicatenibacter indirectly affected intelligence by regulating brain volume. Specifically, the mediation effect was estimated to be 0.017 (95% CI, 0.003–0.031; p = .014) with a mediation proportion of 33.6% (95% CI, 6.8%–60.4%) (Table 3).
考虑到脑容量与人类智力之间的密切关系(19,20),我们分两步进行了磁共振分析,以研究确定的类群对智力的影响是否通过调节脑容量来介导。首先,我们利用牛杆菌和镰刀菌特有的遗传工具估算了已识别类群对脑体积的因果效应。表 S15 和表 S16 列出了与这两个类群相关的 SNPs 脑容量信息摘要。我们发现,Fusicatenibacter 的增加与脑容量的增加有关(IVW beta = 0.086; 95% CI, 0.019-0.153; p = .012)(图 3A;图 S6)。我们没有发现 Oxalobacter 对脑容量有因果关系的证据(IVW beta = 0.006;95% CI,-0.033 至 0.046;p = .745)。然后,我们利用与脑容量相关的遗传工具评估了脑容量对智力的因果效应(表 S17 和 S18)。我们发现了脑容量对智力的因果效应的有力证据(IVW beta = 0.199; 95% CI, 0.163-0.236; p = 4.32 × 10 −27 )(图 3B;图 S7)。敏感性分析并没有解决因果关系估计中的任何多义性问题(表 S19;图 S8-S10)。由于脑容量和智力都包括了来自 UKB 的受试者,暴露和结果 GWAS 之间的样本重叠可能会使估计值偏向混杂估计值。我们使用 Burgess 等人(52)描述的方法估计了参与者重叠造成的偏差。我们发现,最多有 6.3% 的脑容量参与者与智力 GWAS(17,062/269,867)重叠,而 MR 研究中的估计偏差小于 0.82%,这表明我们的 MR 结果受脑容量和智力样本重叠影响的可能性较小。最后,我们发现镰刀菌属通过调节脑容量间接影响智力。具体来说,中介效应估计为 0.017(95% CI,0.003-0.031;p = .014),中介比例为 33.6%(95% CI,6.8%-60.4%)(表 3)。

  1. Download : Download high-res image (410KB)
    下载 :下载高清图片 (410KB)
  2. Download : Download full-size image
    下载 :下载 : 下载全尺寸图片

Figure 3. Mediation analysis of effects of gut microbiome composition on intelligence via brain volume. (A) Summary Mendelian randomization (MR) estimates derived from the inverse-variance weighted (IVW), weighted median, weighted mode, robust adjusted profile score (MR RAPS), and MR-Egger methods for effects of genus Oxalobacter on brain volume (left) and effects of genus Fusicatenibacter on brain volume (right). (B) Summary MR estimates for the effects of brain volume on intelligence.
图 3.肠道微生物组组成通过脑容量对智力影响的中介分析。(A) 用逆方差加权法(IVW)、加权中位数法、加权模式法、稳健调整剖面得分法(MR RAPS)和 MR-Egger 法得出的孟德尔随机化(MR)估计值摘要,用于分析 Oxalobacter 属对脑容量的影响(左)和 Fusicatenibacter 属对脑容量的影响(右)。(B) 脑容量对智力影响的 MR 估计摘要。

Table 3. The Mediation Effect of Gut Microbiota on Intelligence via Affecting Brain Volume
表 3.肠道微生物群通过影响脑容量对智力的中介效应

ExposureTotal Effect 总效果Direct Effect A 直接影响 ADirect Effect B 直接影响 BMediation Effect 调解效应Mediated Proportion (95% CI)
调解比例(95% CI)
β (95% CI)β (95% CI)β (95% CI)β (95% CI)p Value p 值
Genus Oxalobacter 牛杆菌属−0.032 (−0.049 to −0.015)
-0.032(-0.049 至 -0.015)
0.006 (−0.033 to 0.046) 0.006(-0.033 至 0.046)0.199 (0.163 to 0.236) 0.199(0.163 至 0.236)0.001 (−0.006 to 0.009) 0.001(-0.006 至 0.009).758
Genus Fusicatenibacter Fusicatenibacter 属0.051 (0.023 to 0.079) 0.051(0.023 至 0.079)0.086 (0.019 to 0.153) 0.086(0.019 至 0.153)0.017 (0.003 to 0.031) 0.017(0.003 至 0.031).01433.6% (6.8% to 60.4%) 33.6%(6.8% 至 60.4)

Total effect: indicates the effect of gut microbiota on intelligence; direct effect A: the effect of gut microbiota on brain volume; direct effect B: the effect of brain volume on intelligence; mediation effect: the effect of gut microbiota on intelligence through affecting brain volume. The total effect, direct effect A, and direct effect B were derived using the inverse-variance weighted method; the mediation effect was derived by using the delta method. All statistical tests were 2-sided. p < .05 was considered significant.
总效应:表示肠道微生物群对智力的影响;直接效应 A:肠道微生物群对脑量的影响;直接效应 B:脑量对智力的影响;中介效应:肠道微生物群通过影响脑量对智力的影响。总效应、直接效应 A 和直接效应 B 采用反方差加权法得出;中介效应采用 delta 法得出。所有统计检验均为双侧检验,P < .05 为显著。

Considering the bidirectional relationships between brain volume and intelligence (18,19), we further performed a 2-step MR analysis to evaluate whether the genus Fusicatenibacter could indirectly regulate brain volume by affecting intelligence (Table S20). We found that Fusicatenibacter could affect intelligence and then regulate brain volume, and the mediation effect was estimated to be 0.008 (95% CI, 0.003–0.014, p = .004) with a mediation proportion of 9.6% (95% CI, 3.1%–16.2%) (Tables S21 and S22; Figures 11 and 12). Our results indicate that brain volume and intelligence are mutually reinforcing in a virtuous cycle, with the genus Fusicatenibacter exerting a positive causal effect in both directions. There was no evidence for reverse causality of the mediators (brain volume or intelligence) on the exposure (genus Fusicatenibacter) (Tables S10 and S23); thus, the preconditions for causal mediation analyses were satisfied in our study.
考虑到脑容量与智力之间的双向关系(18,19),我们进一步进行了两步磁共振分析,以评估Fusicatenibacter属是否可以通过影响智力来间接调节脑容量(表S20)。我们发现,Fusicatenibacter 可影响智力,进而调节脑容量,其中介效应估计为 0.008 (95% CI, 0.003-0.014, p = .004),中介比例为 9.6% (95% CI, 3.1%-16.2%) (表 S21 和 S22;图 11 和 12)。我们的研究结果表明,脑容量和智力在良性循环中相互促进,Fusicatenibacter 属在两个方向上都产生了积极的因果效应。没有证据表明中介因子(脑容量或智力)与暴露因子(镰刀菌属)之间存在反向因果关系(表 S10 和 S23);因此,我们的研究满足了因果中介分析的先决条件。

In addition, given that intelligence is also related to a wide range of imaging phenotypes (4), we further investigated whether the subdivisions of the human cerebral cortex (25) could also act as potential mediators. Although there was no evidence for causal relationships after multiple testing correction, we found that the genetic liability for Oxalobacter and Fusicatenibacter made a causal contribution to 158 and 190 neuroimaging phenotypes, respectively, at a nominally significant level of p < .05 (Table S24). Then, we assessed the causal effect of these neuroimaging phenotypes on intelligence and found 50 mediating pathways from Oxalobacter or Fusicatenibacter to intelligence (Table S25). For example, the most significant mediator, the surface area of posterior insular area 2 [regions are based on the Human Connectome Project parcellation scheme (53)], was causally affected by Oxalobacter, which then influenced intelligence, and the mediation effect was estimated to be −0.009 (95% CI, −0.015 to −0.003; p = .003).
此外,考虑到智力也与多种影像表型相关(4),我们进一步研究了人类大脑皮层的分支(25)是否也可以作为潜在的中介。尽管多重检验校正后没有证据表明存在因果关系,但我们发现,Oxalobacter 和 Fusicatenibacter 的遗传责任分别对 158 种和 190 种神经影像表型有因果关系,且 p < .05 的水平具有名义显著性(表 S24)。然后,我们评估了这些神经影像表型对智力的因果效应,发现了 50 条从草酸杆菌或镰刀菌到智力的中介途径(表 S25)。例如,最显著的中介因子--后岛叶区2的表面积[区域基于人类连接组计划的划分方案(53)]受Oxalobacter的因果影响,进而影响智力,中介效应估计为-0.009 (95% CI, -0.015 to -0.003; p = .003)。

Discussion 讨论

The role that our microbiome plays in health and disease is one of the biggest scientific challenges (54). In this study, by using summary statistics obtained from the largest GWAS meta-analysis of gut microbiota and intelligence, we conducted a bidirectional 2-sample MR analysis to disentangle the causal relationship. We observed causal evidence indicating a risk effect of Oxalobacter and a protective effect of Fusicatenibacter on intelligence. In contrast, we found no evidence of a causal relationship between intelligence and the abundance of the gut microbiome. More interestingly, we conducted a mediation analysis and showed that the effect of genus Fusicatenibacter on intelligence was partially mediated by regulating brain volume. These findings may have implications for public health interventions that seek to enhance individual intelligence.
我们的微生物群在健康和疾病中扮演的角色是最大的科学挑战之一(54)。在本研究中,我们利用从最大规模的肠道微生物群与智力的 GWAS 元分析中获得的汇总统计数据,进行了双向 2 样本 MR 分析,以厘清其中的因果关系。我们观察到的因果关系证据表明,Oxalobacter 对智力有风险作用,而 Fusicatenibacter 对智力有保护作用。相比之下,我们没有发现智力与肠道微生物组丰度之间存在因果关系的证据。更有趣的是,我们进行了一项中介分析,结果表明镰刀菌属对智力的影响部分是通过调节脑容量来中介的。这些发现可能会对旨在提高个人智力的公共卫生干预措施产生影响。

The gut microbiota potentially affects the development and function of the immune, metabolic, and nervous systems through bidirectional communication along the gut-brain axis (55), which is believed to be involved in the intelligence/cognitive ability of the host (9). For example, compared with normal mice, germ-free mice showed impairments in tests of memory and reductions in hippocampal brain-derived neurotrophic factor (56), which is a neurotrophin crucial for neuronal development and survival, synaptic plasticity, and cognitive function. In contrast, the administration of probiotic strains can promote memory behavior through their production of lactate and the promotion of GABA (gamma-aminobutyric acid) accumulation in the hippocampus (57,58), which provides solid evidence for the role of the microbiome in intelligence. The association between the gut microbiome and cognitive ability in animals also occurs in humans, as mentioned previously. However, current conclusions rely predominantly on disparities in gut microbiome composition and the results of trials that involved the transplant of gut microbiota into gnotobiotic mice (59, 60, 61), which can be affected by various confounding factors. Because the gut microbiome is considered to be highly dynamic, causal association has been an unresolved issue in the field.
肠道微生物群可能会通过肠道-大脑轴的双向交流影响免疫、代谢和神经系统的发育和功能(55),这被认为与宿主的智力/认知能力有关(9)。例如,与正常小鼠相比,无菌小鼠在记忆测试中表现出障碍,海马脑源性神经营养因子减少(56),这种神经营养因子对神经元的发育和存活、突触可塑性和认知功能至关重要。相反,服用益生菌菌株可通过产生乳酸和促进 GABA(γ-氨基丁酸)在海马中的积累来促进记忆行为(57,58),这为微生物组在智力中的作用提供了确凿的证据。如前所述,肠道微生物组与动物认知能力之间的联系在人类中也同样存在。然而,目前的结论主要依赖于肠道微生物组组成的差异以及将肠道微生物群移植到非生物小鼠体内的试验结果(59、60、61),而这些结果可能会受到各种混杂因素的影响。由于肠道微生物群被认为是高度动态的,因此因果关系一直是该领域的一个悬而未决的问题。

To our knowledge, we have reported the first MR analysis to investigate the potential causal relationship between gut microbiota and intelligence. We found that the genetic liability for 2 taxa, Oxalobacter and Fusicatenibacter, reached statistical significance after false discovery rate correction. Oxalobacter is one of the key taxa involved in the gut microbiome diversity of individuals (62), and previous studies have shown that Oxalobacter formigenes plays an important role in oxalate absorption and secretion pathways in the gut (63). Observational studies have found controversial associations between Oxalobacter and cognitive ability. Some studies reported a negative association between the abundance of Oxalobacter and cognitive ability estimated by the Mini-Mental State Examination score (64). In contrast, another study reported that the reduced abundance of Oxalobacter was associated with mild cognitive impairment in older adults (65). Our MR analysis produced causal evidence indicating a risk effect of Oxalobacter on intelligence. We also provided evidence for a protective effect of Fusicatenibacter on intelligence. A recent cross-sectional analysis focusing on species-level features associated with cognition found that certain bacteria capable of producing short-chain fatty acids, such as Fusicatenibacter saccharivorans, were positively associated with better cognitive performance (15). Specifically, a higher abundance of Fusicatenibacter saccharivorans was linked to better scores on both the Mini-Mental State Examination and the Montreal Cognitive Assessment (15). Some taxa that only showed nominal statistical significance have also been reported to be associated with intelligence, such as Ruminococcaceae UCG003, Ruminococcus, and Phascolarctobacterium (see the Supplement) (66,67).
据我们所知,我们首次报告了研究肠道微生物群与智力之间潜在因果关系的磁共振分析。我们发现,经过误发现率校正后,2 个类群,即 Oxalobacter 和 Fusicatenibacter 的遗传责任达到了统计学显著性。牛杆菌是参与个体肠道微生物组多样性的关键类群之一(62),先前的研究表明,形牛杆菌在肠道草酸盐吸收和分泌途径中发挥着重要作用(63)。观察性研究发现,牛杆菌与认知能力之间的关系存在争议。一些研究报告称,草酸杆菌的丰度与以迷你精神状态检查(Mini-Mental State Examination)评分估算的认知能力之间存在负相关(64)。相反,另一项研究报告称,牛分枝杆菌丰度的降低与老年人的轻度认知障碍有关(65)。我们的磁共振分析提供的因果关系证据表明,牛分枝杆菌对智力有风险影响。我们还提供了证据,证明镰刀菌对智力有保护作用。最近一项侧重于与认知相关的物种水平特征的横断面分析发现,某些能够产生短链脂肪酸的细菌,如囊酵母菌(Fusicatenibacter saccharivorans),与更好的认知表现呈正相关(15)。具体来说,较高的 Fusicatenibacter saccharivorans 丰度与较高的迷你精神状态检查和蒙特利尔认知评估得分有关(15)。据报道,一些仅显示名义统计意义的类群也与智力有关,如反刍球菌科(Ruminococcaceae)UCG003、反刍球菌(Ruminococcus)和Phascolarctobacterium(见补编)(66,67)。

Another interesting conclusion arising from the current study is that the protective effect of Fusicatenibacter abundance on intelligence was partially mediated by increasing brain volume. In the first step, we identified that an increased abundance of the Fusicatenibacter was associated with increased intelligence. Differences in gut microbial composition have been reported to be associated with brain structure (68, 69, 70, 71). Specifically, multimodal neuroimaging fusion biomarkers have been reported to mediate the association between gut microbiota and cognition (72). However, little is known about the effect of the Fusicatenibacter on brain volume. The second step provided evidence that genetically determined higher brain volume was associated with higher intelligence. Jansen et al. (19) found causal evidence of effects of genetically predicted brain volume on intelligence (beta = 0.154, p = 1.88 × 10−23) using the generalized summary-data-based MR package. This result supports our second-step estimate in terms of both direction and magnitude.
本研究得出的另一个有趣结论是,丰富的镰刀菌对智力的保护作用部分是通过增加脑容量来实现的。第一步,我们发现镰刀菌丰度的增加与智力的提高有关。据报道,肠道微生物组成的差异与大脑结构有关(68、69、70、71)。具体而言,多模态神经影像融合生物标志物据报道介导了肠道微生物群与认知之间的关联(72)。然而,人们对 Fusicatenibacter 对脑容量的影响知之甚少。第二步提供的证据表明,由基因决定的较高脑容量与较高智力有关。Jansen 等人(19)利用基于概括数据的 MR 软件包发现了遗传预测脑容量对智力影响的因果证据(β=0.154,p=1.88×10 −23 )。这一结果在方向和幅度上都支持我们的第二步估计。

Admittedly, several limitations should be acknowledged when interpreting the results of this study. First, although we used the largest GWAS meta-analysis for gut microbiome composition reported to date, the number of subjects in the gut microbiome composition GWAS is relatively small, and genetic factors can only explain a small proportion of the variance in gut microbiome features; thus, the power to detect a causal relationship was limited. Second, consistent with previous microbiota MR studies, we used a relaxed p-value threshold to select genetic instruments due to the limited variant number associated with gut microbiome composition (17,28,29), which may introduce weak instrument bias. Nevertheless, we tested instrument strength and found that all instruments had F statistics exceeding 10, which is a conventional cutoff for strong instruments (52). Third, due to the lowest taxonomic level being genus in the exposure dataset, we were unable to explore the causal association between gut microbiota and intelligence at the species level. Fourth, although we identified several neuroimaging phenotypes that could serve as potential mediators of the relationship between the gut microbiome and intelligence, no significant results were identified after multiple testing corrections. This is probably due to the correlations among neuroimaging phenotypes; the multiple comparison adjustments may be excessive. Therefore, causality cannot be completely ruled out in the mediation results. Moreover, the study primarily involved individuals of European ancestry, suggesting that future research is needed to determine whether the results can be generalized to other populations.
诚然,在解释本研究的结果时,应该承认有几个局限性。首先,虽然我们使用了迄今为止报告的最大的肠道微生物组组成 GWAS meta 分析,但肠道微生物组组成 GWAS 的受试者数量相对较少,遗传因素只能解释肠道微生物组特征变异的一小部分;因此,检测因果关系的能力有限。其次,与之前的微生物群 MR 研究一致,由于与肠道微生物群组成相关的变异数量有限(17,28,29),我们在选择遗传工具时使用了宽松的 p 值阈值,这可能会带来微弱的工具偏差。尽管如此,我们还是测试了工具强度,发现所有工具的 F 统计量都超过了 10,这是强工具的常规临界值(52)。第三,由于暴露数据集中的最低分类级别为属,我们无法在物种级别探讨肠道微生物群与智力之间的因果关系。第四,尽管我们发现了几种神经影像表型可以作为肠道微生物组与智力之间关系的潜在中介,但经过多重检验校正后,没有发现显著的结果。这可能是由于神经影像表型之间存在相关性;多重比较调整可能过度。因此,不能完全排除中介结果中的因果关系。此外,该研究主要涉及欧洲血统的个体,这表明未来的研究需要确定这些结果是否可以推广到其他人群。

Conclusions 结论

In summary, we used the largest GWAS meta-analysis of gut microbiota and intelligence to test for a causal association between the 2 variables. We found robust genetic evidence of effects of gut microbiome features on intelligence, and a protective effect of Fusicatenibacter on intelligence was partially mediated by regulating brain volume. Our findings may reshape our understanding of the microbiota-gut-brain axis and highlight the gut microbiota as a prospective target for treating and preventing cognitive impairment.
总之,我们利用最大规模的肠道微生物群和智力的 GWAS meta 分析来检验这两个变量之间的因果关系。我们发现了肠道微生物群特征对智力影响的可靠遗传证据,而且Fusicatenibacter对智力的保护作用部分是通过调节脑容量来介导的。我们的研究结果可能会重塑我们对微生物群-肠道-大脑轴的认识,并强调肠道微生物群是治疗和预防认知障碍的前瞻性目标。

Acknowledgments and Disclosures

This work was supported by the National Natural Science Foundation of China (Grant No. 82101601 [to SY] and Grant No. 32170616 [to T-LY]), the Innovation Capability Support Program of Shaanxi Province (Grant No. 2022TD-44 [to T-LY]), China Postdoctoral Science Foundation (Grant No. 2023T160517 [to SY]), an award from the Open Fund of Guangdong Key Laboratory of Age-Related Cardiac and Cerebral Diseases (2023) (to SY), and special guidance funds for the construction of world-class universities (disciplines) and characteristic development in central universities.
本研究得到国家自然科学基金(批准号:82101601[给SY]和批准号:32170616[给T-LY])、陕西省创新能力支持计划(批准号:2022TD-44[给T-LY])、中国博士后科学基金(批准号:2023T160517[给SY])、广东省老年性心脑疾病重点实验室开放基金(2023)(给SY)和广东省老年性心脑疾病重点实验室特别指导。2023T160517 [给 SY])、广东省老年性心脑疾病重点实验室开放基金(2023)奖励(给 SY)、世界一流大学(学科)建设和中央高校特色发展专项引导基金。

SY, T-LY, and YG designed the study. SY participated in the data collection. SY, JG, LQ, WS, and R-JZ performed the data analyses. SY, J-ZH, XW, HW, and J-HW prepared the tables and figures. SY wrote the paper. JG, S-SD, L-LC, and YW critically revised the content. T-LY and YG supervised the study. All authors contributed to editing the paper.
SY、T-LY 和 YG 设计了本研究。SY 参与了数据收集。SY、JG、LQ、WS 和 R-JZ 进行了数据分析。SY、J-ZH、XW、HW 和 J-HW 准备了表格和图表。SY 撰写论文。JG、S-SD、L-LC 和 YW 对内容进行了严格的修改。T-LY和YG指导了研究。所有作者都参与了论文的编辑工作。

The GWAS for gut microbiome composition can be obtained through the NHGRI-EBI (National Human Genome Research Institute-European Bioinformatics Institute) GWAS catalog, study accession nos. GCST90016908–GCST90017118. The GWAS summary statistics for brain volume and intelligence are available for download at https://ctg.cncr.nl/software/summary_statistics/. Summary GWAS statistics for neuroimaging phenotypes are available from https://portal.ide-cam.org.uk/overview/483.
肠道微生物组组成的 GWAS 可通过 NHGRI-EBI(美国国家人类基因组研究所-欧洲生物信息学研究所)的 GWAS 目录获取,研究登录号:gcst90016908-gcst90017118。GCST90016908–GCST90017118.脑容量和智力的 GWAS 统计摘要可在 https://ctg.cncr.nl/software/summary_statistics/ 上下载。神经影像表型的 GWAS 统计摘要可从 https://portal.ide-cam.org.uk/overview/483 下载。

A previous version of this article was published as a preprint on medRxiv: https://www.medrxiv.org/content/10.1101/2023.05.11.23289760v2 .

The authors report no biomedical financial interests or potential conflicts of interest.

Supplementary Material 补充材料

What’s this? 这是什么?

Download : Download spreadsheet (22KB)

Key Resources Table.

Download : Download Acrobat PDF file (2MB)


References 参考资料

Cited by (0)


SY, J-ZH, and JG contributed equally to this work.

View Abstract