The gut microbiota-bile acid axis links the positive association between chronic insomnia and cardiometabolic diseases
肠道微生物群-胆汁酸轴将慢性失眠与心脏代谢疾病之间的正相关联系起来

Chronic insomnia was associated with the diversity of the gut microbiota
慢性失眠与肠道微生物群的多样性有关

We first investigated the longitudinal association of chronic insomnia status of the GNHS participants with microbioal α-/β- diversity. We found that there were significant differences in the microbial β- diversity of the New-onset group and Long-term chronic insomnia group, compared to the Long-term healthy group (Fig. 1b). The α-diversity parameters (Observed species, Chao 1 index and ACE index) of the New-onset group and Long-term chronic insomnia group were significantly lower than those of the Long-term healthy group (Fig. 1b and Supplementary Fig. ¹). The α-diversity parameter the Shannon index of the Long-term chronic insomnia group was significantly lower than that of the Long-term healthy group (Supplementary Fig. ¹). The α-diversity parameter the Simpson index was not significantly different among the four chronic insomnia status groups (Supplementary Fig. ¹). Consistent results were observed by using three different statistical models (Fig. 1b and Supplementary Fig. ¹), suggesting that chronic insomnia was robustly associated with the gut microbiota structure.
我们首先研究了GNHS参与者的慢性失眠状态与微生物α/β多样性的纵向关联。我们发现，与长期健康组相比，新发组和长期慢性失眠组的微β生物多样性差异有统计学意义（图1b）。新发组和长期慢性失眠组的α多样性参数（观察到的物种、Chao 1指数和ACE指数）显著低于长期健康组（图1b和补充图1b）。 ¹ ）。长期慢性失眠组的α多样性参数Shannon指数显著低于长期健康组（补充图1）。 ¹ ）。α多样性参数辛普森指数在4个慢性失眠状态组中差异无统计学意义（补充图1）。 ¹ ）。使用三种不同的统计模型观察到一致的结果（图1b和补充图1b）。 ¹ ），表明慢性失眠与肠道微生物群结构密切相关。

We combined the New-onset group and Long-term chronic insomnia group into one Chronic insomnia group to increase the sample size, given that there was no significant difference in the microbiota structure between the two groups. Chronic insomnia in the GNHS participants was associated with lower levels of Observed species (p < 0.01), Shannon index (p < 0.05), Chao 1 index (p < 0.001) and ACE index (p < 0.001), respectively (Fig. 2a and Supplementary Fig. ²). Principal coordinate analysis of the gut microbial profiles in the GNHS showed a significant shift in the gut microbiota composition of the Chronic insomnia group compared to the Long-term healthy group (p < 0.01; PERMANOVA test with 999 permutations) (Fig. 2b). A similar pattern was observed for the shift in the structure of the gut microbiota in the GGMP (Fig. 2a, b and Supplementary Fig. ³).
鉴于两组之间的微生物群结构没有显著差异，我们将新发病组和长期慢性失眠组合并为一个慢性失眠组以增加样本量。GNHS参与者的慢性失眠分别与观察到的物种（p < 0.01）、Shannon指数（p < 0.05）、Chao 1指数（p < 0.001）和ACE指数（p < 0.001）水平较低有关（图2a和补充图）。 ² GNHS中肠道微生物谱的主坐标分析显示，与长期健康组相比，慢性失眠组的肠道微生物群组成发生了显着变化（p < 0.01; 具有 999 种排列的 PERMANOVA 检验）（图 2b）。 在GGMP中观察到肠道微生物群结构的变化也存在类似的模式（图2a，b和补充图2a，b）。 ³

Open in a separate window

Fig. 2 图2

Association of chronic insomnia with gut microbiota and bile acids.
慢性失眠与肠道微生物群和胆汁酸的关联。

a Observed species in the discovery cohort (n = 1809) and validation cohort (n = 6122). The results of Shannon index, Chao 1 index, ACE index and Simpson index are reported in Supplementary Fig. ² (discovery cohort) and Supplementary Fig. ³ (validation cohort). p value was calculated from multivariable-adjusted linear regression with three different models (Methods; model 1: *, model 2: +, model 3: #). Box plots indicate median and interquartile range (IQR). The upper and lower whiskers indicate 1.5 times the IQR from above the upper quartile and below the lower quartile. b β-diversity: principal coordinate analysis (PCoA) of genus-level Bray-Cutis distance in the discovery and validation cohorts. Permutational ANOVA (999 permutations) was used to identify the variation of β-diversity, adjusted for the same covariates as α-diversity. c Multivariate Analysis by Linear Models (MaAsLin) was used to identify the gut microbial biomarkers for chronic insomnia comparing Chronic insomnia group with Long-term healthy group. The q values (false discovery rate adjusted p value) were calculated using the Benjamini-Hochberg method (*q < 0.25, **q < 0.05). d Multivariable linear regression was used to assess the association of chronic insomnia with the gut microbial biomarkers in the discovery and validation cohorts, adjusted for the same covariates as α-diversity. Error bars are beta coefficient with 95% confidence intervals. e, Bile acid biomarkers of chronic insomnia in the GNHS (n = 954). Box plots indicate median and interquartile range. Orthogonal partial least squares discrimination analysis (OPLS-DA) (Supplementary Fig. ⁵) and multivariable-adjusted linear regression were used to identify potential bile acids associated with chronic insomnia (*p < 0.05, **p < 0.01, ***p < 0.001). f Partial correlation analysis was used to assess the interrelationships between the identified gut microbiota and bile acid biomarkers, adjusted for age, sex and BMI. Orange/sky blue circles indicate chronic insomnia-positive/negative biomarkers. Orange/sky blue lines indicate positive/negative associations. The Benjamini-Hochberg method was used to correct for multiple testing. All statistical tests were two-sided. Source data are provided as a ^{Source Data} file. α-MCA α-muricholic acid; β-HDCA β-hyodeoxycholic acid; HDCA hyodeoxycholic acid; HCA hyocholic acid; IsoLCA isolithocholic acid; LCA lithocholic acid; MCA Muro cholic acid; NorCA Nor cholic acid; NorDCA Nor deoxycholic acid; UDCA ursodeoxycholic acid.
a 在发现队列（n = 1809）和验证队列（n = 6122）中观察到的物种。香农指数、Chao 1指数、ACE指数和Simpson指数的结果见补充图。 ² （发现队列）和补充图。 ³ （验证队列）。p值是用三种不同的模型（方法;模型1：*，模型2：+，模型3：#）的多变量调整线性回归计算的。箱形图表示中位数和四分位距 （IQR）。上部和下部胡须表示上四分位数上方和下四分位数下方的 IQR 的 1.5 倍。b β多样性：发现和验证队列中属级Bray-Cutis距离的主坐标分析（PCoA）。使用排列方差分析（999 个排列）来识别β多样性的变化，并针对与α多样性相同的协变量进行调整。c 采用线性模型多变量分析（MaAsLin）鉴定慢性失眠的肠道微生物生物标志物，比较慢性失眠组和长期健康组。q值（错误发现率调整后的p值）使用Benjamini-Hochberg方法计算（*q < 0.25，**q < 0.05）。d 使用多变量线性回归评估慢性失眠与发现和验证队列中肠道微生物生物标志物的关联，并针对与α多样性相同的协变量进行调整。误差线是具有 95% 置信区间的 beta 系数。e， GNHS中慢性失眠的胆汁酸生物标志物（n = 954）。箱形图表示中位数和四分位距。正交偏最小二乘判别分析（OPLS-DA）（补充图1）。 ⁵ ）和多变量校正线性回归用于识别与慢性失眠相关的潜在胆汁酸（*p < 0。05， **p < 0.01， ***p < 0.001）。f 部分相关性分析用于评估已鉴定的肠道微生物群与胆汁酸生物标志物之间的相互关系，并根据年龄、性别和 BMI 进行调整。橙色/天蓝色圆圈表示慢性失眠阳性/阴性生物标志物。橙色/天蓝色线表示正/负关联。Benjamini-Hochberg 方法用于校正多重测试。所有统计检验都是双侧的。源数据以 ^{Source Data} 文件形式提供。α-MCA α-muricholic acid;β-HDCA β-玺脱氧胆酸;HDCA玖脱氧胆酸;HCA猪胆酸;异LCA异石胆酸;LCA石胆酸;MCA Muro胆酸;NorCA 也不是胆酸;NorDCA 也不是脱氧胆酸;UDCA熊去氧胆酸。

Chronic insomnia was associated with specific gut microbes and bile acids
慢性失眠与特定的肠道微生物和胆汁酸有关

We used Multivariate Analysis by Linear Models (MaAsLin) adjusted for potential confounders to identify potential microbial biomarkers (as outcome variables) of chronic insomnia (as a predictor in the model). Ruminococcaceae UCG-002 and Ruminococcaceae UCG-003 were identified as the microbial biomarkers of chronic insomnia (Fig. 2c and Supplementary Table ¹). In the GNHS, Chronic insomnia group was associated with lower levels of Ruminococcaceae UCG-002 (β: −0.19, 95% CI: −0.32 to −0.06) and Ruminococcaceae UCG-003 (β: −0.20, 95% CI: −0.33 to −0.07), compared with the Long-term healthy group (Fig. 2d). These results were also observed in the GGMP (Fig. 2d). The results of the sensitivity analysis suggested that different covariate adjustments did not substantially affect the results in the GGMP (Supplementary Fig. ⁴). In addition, chronic insomnia had no interactions with age or sex for Ruminococcaceae UCG-002 and Ruminococcaceae UCG-003 (Supplementary Table ²). To assess the potential influence of the number of insomnia symptoms, we conducted a secondary analysis of the GGMP participants, for whom the data were available, and found that the per unit change in the chronic insomnia symptom score was inversely associated with a per 1-SD change in Ruminococcaceae UCG-002 (β: −0.04, 95% CI: −0.06 to −0.02, p < 0.001) and Ruminococcaceae UCG-003 (β: −0.04, 95% CI: −0.07 to −0.01, p = 0.002) (Supplementary Table ³).
我们使用线性模型多变量分析（MaAsLin）对潜在的混杂因素进行了调整，以确定慢性失眠的潜在微生物生物标志物（作为结果变量）（作为模型中的预测因子）。瘤胃球菌科UCG-002和瘤胃球菌科UCG-003被鉴定为慢性失眠的微生物生物标志物（图2c和补充表 ¹ ）。在GNHS中，与长期健康组相比，慢性失眠组与瘤胃球菌科UCG-002（β：-0.19,95%CI：-0.32至-0.06）和瘤胃球菌科UCG-003（β：-0.20,95%CI：-0.33至-0.07）的水平较低相关（图2d）。在GGMP中也观察到了这些结果（图2d）。敏感性分析的结果表明，不同的协变量调整对GGMP的结果没有实质性影响（补充图1）。 ⁴ ）。此外，瘤胃球菌科UCG-002和瘤胃球菌科UCG-003的慢性失眠与年龄或性别没有相互作用（补充表 ² ）。为了评估失眠症状数量的潜在影响，我们对GGMP参与者进行了二次分析，发现慢性失眠症状评分的单位变化与瘤胃球菌科UCG-002的每1-SD变化呈负相关（β：-0.04,95%CI：-0.06至-0.02， p < 0.001） 和瘤胃球菌科 UCG-003 （β： -0.04， 95% CI： -0.07 至 -0.01， p = 0.002） （补充表 ³ ）。

We next used orthogonal partial least squares discrimination analysis (OPLS-DA) to identify potential fecal bile acids associated with chronic insomnia (Supplementary Fig. ⁵) and then used linear regression to confirm the chronic insomnia-bile acid associations in the GNHS (Fig. 2e). Chronic insomnia was associated with higher levels of muro cholic acid (MCA, p = 0.046) and nor cholic acid (NorCA, p = 0.046) and with lower levels of isolithocholic acid (IsoLCA, p = 0.029), lithocholic acid (LCA, p = 0.035) and ursodeoxycholic acid (UDCA, p = 0.039) (Fig. 2e). The results of the sensitivity analysis showed that adding dietary cholesterol intake and fiber intake as additional covariates did not substantially affect the association of chronic insomnia with bile acids (Supplementary Table ⁴). Co-occurrence network analysis based on the partial correlation coefficient showed that Ruminococcaceae UCG-002 and Ruminococcaceae UCG-003 were positively associated with secondary bile acids (IsoLCA, LCA, and UDCA), and inversely associated with primary bile acids (MCA and NorCA) (p < 0.001) (Fig. 2f). In addition, chronic insomnia was not associated with short-chain fatty acids, aromatic amino acids, or their derivatives (Supplementary Table ⁵). These results indicated that chronic insomnia might have a significant impact on the gut microbiota-bile acid axis.
接下来，我们使用正交偏最小二乘判别分析（OPLS-DA）来识别与慢性失眠相关的潜在粪便胆汁酸（补充图）。 ⁵ ），然后使用线性回归来确认GNHS中的慢性失眠-胆汁酸关联（图2e）。慢性失眠与较高的壁胆酸（MCA，p = 0.046）和胆酸（NorCA，p = 0.046）以及较低的异溶胆酸（IsoLCA，p = 0.029），石胆酸（LCA，p = 0.035）和熊去氧胆酸（UDCA，p = 0.039）水平有关（图2e）。敏感性分析的结果表明，添加膳食胆固醇摄入量和纤维摄入量作为额外的协变量，并没有显著影响慢性失眠与胆汁酸的关联（补充表 ⁴ ）。基于偏相关系数的共现网络分析表明，瘤胃球菌科UCG-002和瘤胃球菌科UCG-003与次生胆汁酸（IsoLCA、LCA和UDCA）呈正相关，与初级胆汁酸（MCA和NorCA）呈负相关（p < 0.001）（图2f）。 此外，慢性失眠与短链脂肪酸、芳香族氨基酸或其衍生物无关（补充表 ⁵ ).这些结果表明，慢性失眠可能对肠道微生物群-胆汁酸轴产生显着影响。

Chronic insomnia-related gut microbial features and bile acids were associated with CMD and risk factors
慢性失眠相关肠道微生物特征和胆汁酸与CMD和危险因素相关

To further investigate whether the chronic insomnia-related gut microbiota or bile acids play a role in CMD, we used multivariable logistic regression to examine the association of the chronic insomnia-related gut microbiota or bile acids with CMD. In the cross-sectional analysis of the GNHS participants, a per standard deviation (SD)-unit increment in Ruminococcaceae UCG-002 was associated with a 24% lower risk of metabolic syndrome (MetS) (OR: 0.76, 95% CI: 0.66–0.86), a 22% lower risk of T2D (OR: 0.78, 95% CI: 0.67–0.91), and 13% lower risk of dyslipidemia (OR: 0.87, 95% CI: 0.79–0.97) (Fig. 3a). Each per SD-unit increment in Ruminococcaceae UCG-003 was associated with a 23% lower risk of MetS (OR: 0.77, 95% CI: 0.67–0.88) (Fig. 3b). IsoLCA was inversely associated with T2D (OR: 0.74, 95% CI: 0.61–0.90) (Fig. 3c). MCA and NorCA were positively associated with MetS (OR: 1.33, 95% CI: 1.13–1.58; OR: 1.36, 95% CI: 1.15–1.61), respectively (Fig. 3c).
为了进一步研究慢性失眠相关肠道菌群或胆汁酸是否在CMD中起作用，我们使用多变量logistic回归来检查慢性失眠相关肠道微生物群或胆汁酸与CMD的关联。在GNHS参与者的横断面分析中，瘤胃球菌科UCG-002的每标准差（SD）单位增加与代谢综合征（MetS）风险降低24%相关（OR：0.76,95%CI：0.66-0.86），T2D风险降低22%（OR：0.78,95%CI：0.67-0.91），血脂异常风险降低13%（OR：0.87,95%CI： 0.79–0.97）（图3a）。瘤胃球菌科 UCG-003 中每 SD 单位的增加与 MetS 风险降低 23% 相关（OR：0.77,95% CI：0.67–0.88）（图 3b）。IsoLCA与T2D呈负相关（OR：0.74,95%CI：0.61–0.90）（图3c）。MCA和NorCA与MetS呈正相关（OR：1.33,95%CI：1.13–1.58;OR：1.36,95% CI：1.15–1.61），分别（图 3c）。

Open in a separate window

Fig. 3 图3

Association of the chronic insomnia-related gut microbiota-bile acid axis with cardiometabolic diseases.
慢性失眠相关肠道微生物群-胆汁酸轴与心脏代谢疾病的关联。

Multivariable logistic regression was used to estimate the association of the chronic insomnia inverse-related microbial biomarkers Ruminococcaceae UCG-002 (a) and Ruminococcaceae UCG-003 (b) with different CMD in the discovery and validation cohorts, respectively. The effect estimates from the discovery and validation cohorts were pooled using random effects meta-analysis. c Multivariable logistic regression was used to estimate the association of the chronic insomnia-related bile acid biomarkers with CMD in the discovery cohort. d The prospective associations of the above identified gut microbiota biomarkers (measurement of gut microbiota data at the second follow-up) with the incidence of CMD (dyslipidemia) at the third follow-up using multivariable logistic regression, adjusted for potential confounders. Error bars in a–d are odds ratios with 95% confidence intervals. e Associations of the identified gut microbiota and bile acid biomarkers with CMD-related risk factors (BMI, DBP, SBP, waist circumference, fasting serum TG, TC, HDL, LDL, glucose, insulin, and HbA1c) using multivariable linear regression model in the GNHS, adjusted for potential confounders. f Parallel coordinate chart showing the association among gut microbes, bile acid biomarkers and CMD outcomes. The left panel shows the microbial biomarkers, the middle panel shows the bile acid biomarkers, and the right panel shows the CMD outcomes. The red lines across panels indicate the positive association. The green lines across panels indicate the inverse association. g The chronic insomnia inverse-related microbial biomarker Ruminococcaceae UCG-002 affects risk of MetS and T2D though specific bile acid biomarkers, respectively. h The chronic insomnia inverse-related microbial biomarker Ruminococcaceae UCG-003 affects the risk of MetS and T2D through specific bile acid biomarkers, respectively. The gray lines indicate the associations, with corresponding normalized beta values and p values. The red arrowed lines indicate the microbial effects on CMD mediated by specific bile acid biomarkers, with the corresponding mediation p values. p value < 0.05 is considered significantly different. Throughout the above analyses, FDR from multiple testing was controlled by the Benjamini-Hochberg method. All statistical tests were two-sided. Source data are provided as a ^{Source Data} file. CMD cardiometabolic disease; T2D type 2 diabetes; MetS metabolic syndrome; SBP systolic blood pressure; DBP diastolic blood pressure; TG triglycerides; TC total cholesterol; HDL high-density lipoprotein cholesterol; LDL low-density lipoprotein cholesterol; HbAlc glycated hemoglobin.
采用多变量logistic回归法估计慢性失眠反向相关微生物生物标志物瘤胃球菌科UCG-002（a）和瘤胃球菌科UCG-003（b）与发现队列和验证队列中不同CMD的相关性。使用随机效应荟萃分析合并发现和验证队列的效应估计值。c 使用多变量logistic回归来估计发现队列中慢性失眠相关胆汁酸生物标志物与CMD的关联。d 使用多变量逻辑回归，对上述确定的肠道微生物群生物标志物（第二次随访时肠道微生物群数据的测量）与第三次随访时 CMD（血脂异常）发生率的前瞻性关联，并针对潜在的混杂因素进行调整。a–d 中的误差线是具有 95% 置信区间的比值比。e 使用 GNHS 中的多变量线性回归模型将已鉴定的肠道微生物群和胆汁酸生物标志物与 CMD 相关危险因素（BMI、DBP、SBP、腰围、空腹血清 TG、TC、HDL、LDL、葡萄糖、胰岛素和 HbA1c）关联，并针对潜在的混杂因素进行调整。f 显示肠道微生物、胆汁酸生物标志物和 CMD 结果之间关联的平行坐标图。左图显示微生物生物标志物，中图显示胆汁酸生物标志物，右图显示 CMD 结果。面板上的红线表示正相关。面板上的绿线表示反向关联。g 慢性失眠逆相关微生物生物标志物瘤胃球菌科 UCG-002 分别通过特异性胆汁酸生物标志物影响 MetS 和 T2D 的风险。 h 慢性失眠逆相关微生物生物标志物瘤胃球菌科 UCG-003 分别通过特异性胆汁酸生物标志物影响 MetS 和 T2D 的风险。灰线表示关联，并具有相应的归一化 beta 值和 p 值。红色箭头线表示由特定胆汁酸生物标志物介导的微生物对 CMD 的影响，以及相应的介导 p 值。p 值< 0.05 被认为存在显著差异。 在上述分析中，来自多个测试的 FDR 由 Benjamini-Hochberg 方法控制。 所有统计检验都是双侧的。 源数据以 ^{Source Data} 文件形式提供。CMD心脏代谢疾病;T2D 2 型糖尿病;MetS代谢综合征;SBP收缩压;DBP舒张压;TG甘油三酯;TC总胆固醇;高密度脂蛋白胆固醇;低密度脂蛋白低密度脂蛋白胆固醇;HbAlc糖化血红蛋白。

The majority of the results from the GNHS could be also observed in the GGMP. Meta-analysis of results from the two cohorts consistently showed that Ruminococcaceae UCG-002 and Ruminococcaceae UCG-003 were significantly inversely associated with MetS (Pooled OR: 0.82, 95% CI: 0.72–0.93; Pooled OR: 0.82, 95% CI: 0.77–0.88), T2D (Pooled OR: 0.84, 95% CI: 0.75–0.95; Pooled OR: 0.87, 95% CI: 0.81–0.95), dyslipidemia (Pooled OR: 0.91, 95% CI: 0.87–0.95; Pooled OR: 0.88, 95% CI: 0.84–0.92) and coronary heart disease (CHD) (Pooled OR: 0.88, 95% CI: 0.82–0.94; Pooled OR: 0.93, 95% CI: 0.87–0.99), respectively (Fig. 3a, b). Consistent results were observed in the sensitivity analyses (Supplementary Fig. ⁶). Ruminococcaceae UCG-002 was interacted with sex in the risk of dyslipidaemia (p_interaction = 0.003) (Supplementary Table ⁶). The stratified analyses by sex showed that the inverse association of Ruminococcaceae UCG-002 with dyslipidaemia was significant among the men participants (Pooled OR: 0.87, 95% CI: 0.81–0.93, p < 0.001), but not among the women participants (Pooled OR: 0.96, 95% CI: 0.88–1.05, p = 0.394) (Supplementary Table ⁶).
GNHS的大部分结果也可以在GGMP中观察到。对两个队列结果的荟萃分析一致表明，瘤胃球菌科 UCG-002 和瘤胃球菌科 UCG-003 与 MetS 呈显著负相关（合并 OR：0.82,95% CI：0.72–0.93;合并 OR：0.82,95% CI：0.77–0.88）、T2D（合并 OR：0.84,95% CI：0.75–0.95;合并OR：0.87,95% CI：0.81-0.95）、血脂异常（合并OR：0.91,95% CI：0.87-0.95;合并OR：0.88,95%CI：0.84-0.92）和冠心病（CHD）（合并OR：0.88,95%CI：0.82-0.94;合并OR：0.93,95%CI：0.87–0.99），分别（图3a，b）。在敏感性分析中观察到一致的结果（补充图1）。 ⁶ ）。瘤胃球菌科 UCG-002 在血脂异常的风险中与性别相互作用 （p _interaction = 0.003）（补充表 ⁶ ）。按性别分层分析显示，瘤胃球菌科UCG-002与血脂异常的负相关在男性参与者中显著（合并OR：0.87,95%CI：0.81-0.93，p < 0.001），但在女性参与者中则不显著（合并OR：0.96,95%CI：0.88-1.05，p = 0.394）（补充表 ⁶ ）。

In the longitudinal analysis among GNHS, we found that Ruminococcaceae UCG-002 and Ruminococcaceae UCG-003 were inversely associated with the incidence of dyslipidemia (OR: 0.72, 95% CI: 0.57–0.90; OR: 0.74, 95% CI: 0.60–0.91), respectively (Fig. 3d). In addition, Ruminococcaceae UCG-002 was significantly inversely associated with BMI (β: −0.28, 95% CI: −0.43 to −0.12) (Fig. 3e).
在GNHS的纵向分析中，我们发现瘤胃球菌科UCG-002和瘤胃球菌科UCG-003与血脂异常的发生率呈负相关（OR：0.72,95%CI：0.57–0.90;OR：0.74,95% CI：0.60–0.91），分别为 （Fig. 3d）。此外，瘤胃球菌科UCG-002与BMI呈显著负相关（β：-0.28,95%CI：-0.43至-0.12）（图3e）。

To evaluate whether bile acids can mediate the relationship between the gut microbiota and CMD (Fig. 3f), we applied mediation analysis, which showed that the inverse association of the chronic insomnia-related gut microbial biomarkers with MetS and T2D were mediated by some specific bile acids (Fig. 3g, h). MCA mediated the association of Ruminococcaceae UCG-002 and Ruminococcaceae UCG-003 with the risk of MetS (50.8%, p_mediation < 0.001; 39.5%, p_mediation < 0.001, respectively, Fig. 3g, h). NorCA mediated the association of Ruminococcaceae UCG-002 and Ruminococcaceae UCG-003 with the MetS risk (32.9%, p_mediation < 0.001; 31.9%, p_mediation < 0.001, respectively, Fig. 3g, h). In addition, IsoLCA mediated the association of Ruminococcaceae UCG-002 and Ruminococcaceae UCG-003 with the risk of T2D (41.7%, p_mediation = 0.040; 53.2%, p_mediation < 0.001, respectively, Fig. 3g, h). Sensitivity analysis for mediation effects indicated that the results of the above mediation analysis were relatively robust to the possible existence of an unmeasured confounder (Supplementary Table ⁷).
为了评估胆汁酸是否可以介导肠道微生物群与CMD之间的关系（图3f），我们应用了中介分析，结果表明慢性失眠相关肠道微生物生物标志物与MetS和T2D的负相关是由一些特定的胆汁酸介导的（图3g，h）。MCA介导瘤胃球菌科UCG-002和瘤胃球菌科UCG-003与MetS风险的关联（分别为50.8%，p _mediation <0.001;39.5%，p _mediation <0.001，图3g，h）。 NorCA介导瘤胃球菌科UCG-002和瘤胃球菌科UCG-003与MetS风险的关联（分别为32.9%，p _mediation < 0.001;31.9%，p _mediation < 0.001，图3g，h）。此外，IsoLCA介导瘤胃球菌科UCG-002和瘤胃球菌科UCG-003与T2D风险的关联（分别为41.7%，p _mediation = 0.040;53.2%，p _mediation <0.001，图3g，h）。 对中介效应的敏感性分析表明，上述中介分析的结果对可能存在未测量的混杂因素相对可靠（补充表 ⁷ ）。

Metadata collection in the GNHS
GNHS 中的元数据收集

For the GNHS, during the on-site face-to-face questionnaire interviews, we collected information on sociodemographic, lifestyle, dietary factors, and medical history. Anthropometric parameters, including weight, height, waist circumference, and hip circumference, were measured by trained staff. Total energy intake was calculated according to the Chinese Food Consumption Table, 2002⁷⁷. Physical activity was assessed as the total metabolic equivalent for task hours per day based on a questionnaire for physical activity⁷⁸.
对于GNHS，在现场面对面问卷访谈中，我们收集了有关社会人口学、生活方式、饮食因素和病史的信息。人体测量参数，包括体重、身高、腰围和臀围，由训练有素的工作人员测量。总能量摄入量是根据2002年中国食品消费量表计算 ⁷⁷ 的。根据身体活动 ⁷⁸ 问卷，将身体活动评估为每天任务小时数的总代谢当量。

Fasting venous blood samples were taken at the recruitment and follow-up visits and were aliquoted and stored in a −80 °C freezer prior to analysis. FBG, TG, TC, HDL, and LDL were measured by colorimetric methods using a Roche Cobas 8000 c702 automated analyzer (Roche Diagnostics GmbH, Shanghai, China). Insulin was measured by electrochemiluminescence immunoassay methods using a Roche Cobas 8000 e602 automated analyzer (Roche Diagnostics GmbH, Shanghai, China). High-performance liquid chromatography was used to measure the HbA1c level using the Bole D-10 Hemoglobin A1c Program on a Bole D-10 Hemoglobin Testing System.
在招募和随访时采集空腹静脉血样，并在分析前等分并储存在-80°C冰箱中。FBG、TG、TC、HDL 和 LDL 使用罗氏 Cobas 8000 c702 自动分析仪（Roche Diagnostics GmbH，中国上海）通过比色法测量。使用罗氏 Cobas 8000 e602 自动分析仪（罗氏诊断有限公司，中国上海）通过电化学发光免疫测定法测量胰岛素。使用高效液相色谱法在 Bole D-10 血红蛋白测试系统上使用 Bole D-10 血红蛋白 A1c 程序测量 HbA1c 水平。

Fecal sample collection, DNA extraction, and 16S rRNA gene sequencing in the GNHS
GNHS中的粪便样本采集、DNA提取和16S rRNA基因测序

During a follow-up visit to the study center, the participants were given a stool sampler and provided detailed instructions for stool sample collection. Briefly, each participant collected their stool sample after defecation and gave the sample to the staff immediately. The stool samples with ice bags were transported to the research laboratory and stored in a −80 °C freezer within 4 h. Detailed information regarding DNA extraction and gut microbiota 16S rRNA gene sequencing in the GNHS is provided in the ^{Supplementary methods}.
在对研究中心的后续访问中，参与者获得了粪便采样器，并提供了粪便样本采集的详细说明。简而言之，每个参与者在排便后收集粪便样本，并立即将样本交给工作人员。将带有冰袋的粪便样品运送到研究实验室，并在4小时内储存在-80°C冰箱中。有关 GNHS 中 DNA 提取和肠道微生物群 16S rRNA 基因测序的详细信息，请参见 ^{Supplementary methods} .

Targeted fecal bile acid profiling in the GNHS
GNHS中的靶向粪便胆汁酸分析

Targeted bile acid profiling of fecal samples (n = 954) was performed with an ultra-performance liquid chromatography coupled to tandem mass spectrometry (UPLC-MS/MS) system (ACQUITY UPLC-Xevo TQ-S, Waters Corp., Milford, MA, USA) at Metabo-Profile Biotechnology Co., Ltd. (Shanghai, China) (^{Supplementary methods}).
在Metabo-Profile Biotechnology Co.， Ltd.（中国上海）（ ）使用超高效液相色谱联用串联质谱（UPLC-MS/MS）系统（ACQUITY UPLC-Xevo TQ-S，Waters Corp.，Milford，MA，USA）对粪便样品（n = 954）进行靶向胆汁酸分析 ^{Supplementary methods} 。

Description of the GGMP GGMP的说明

The GGMP is a large community-based cross-sectional cohort conducted between 2015 and 2016 including 7009 participants with high quality gut microbiome data³⁸. The GGMP participants were from 14 randomly selected districts or counties in Guangdong Province, China. In face-to-face questionnaire interviews, the host metadata, including socio-demographic features, disease status, lifestyle and dietary information, were collected. We excluded participants (1) without chronic insomnia information (n = 633); and (2) with missing covariates (age, sex, BMI, education, smoking status, alcohol status) (n = 254). Finally, we included 6122 participants (52.8 ± 14.7 y, 55.2% of women) from the GGMP in our analysis as an independent validation cohort. The characteristics of the included participants in the GGMP are presented in Table 2. Detailed information regarding the host metadata and stool sample collection and the 16S rRNA gene sequencing process in the GGMP have been reported previously³⁸.
GGMP 是 2015 年至 2016 年间进行的基于社区的大型横断面队列，包括 7009 名具有高质量肠道微生物组数据 ³⁸ 的参与者。GGMP参与者来自中国广东省14个随机选择的区或县。在面对面的问卷访谈中，收集了宿主元数据，包括社会人口学特征、疾病状况、生活方式和饮食信息。我们排除了（1）没有慢性失眠信息的受试者（n = 633）;（2）缺少协变量（年龄、性别、BMI、教育程度、吸烟状况、酒精状况）（n = 254）。最后，我们将来自GGMP的6122名参与者（52.8名±14.7岁，55.2%的女性）作为独立验证队列纳入我们的分析。表2列出了GGMP中纳入的参与者的特征。关于GGMP中宿主元数据和粪便样本采集以及16S rRNA基因测序过程的详细信息之前 ³⁸ 已经报道过。

Statistical analysis 统计分析

We compared differences between four groups using the chi-square test for categorical variables and ANOVA for continuous variables. In the GNHS, we examined the association of chronic insomnia with gut microbial α-diversity indices (Observed species, Shannon index, Chao 1 index, ACE index and Simpson index) among the four groups using a multivariable linear regression with three different statistical models. Model 1 was adjusted for age, sex, BMI, smoking status, alcohol status, physical activity, education, income and total energy intake at baseline. Model 2 was additionally controlled for hypertension, hyperlipidemia, MetS, T2D, CHD, stroke, and medication for T2D. Model 3 was further adjusted for dietary intake of vegetables, fruits, red and processed meat, fish, dairy products, coffee and tea. The association between chronic insomnia and β-diversity dissimilarity based on genus-level Bray-Curtis distance was examined using permutational ANOVA (PERMANOVA) (999 permutations).
我们使用分类变量的卡方检验和连续变量的方差分析比较了四组之间的差异。在GNHS中，我们使用三种不同统计模型的多变量线性回归检查了四组慢性失眠与肠道微生物α多样性指数（观察到的物种、香农指数、Chao 1指数、ACE指数和Simpson指数）的关联。模型 1 根据年龄、性别、BMI、吸烟状况、酒精状况、身体活动、教育程度、收入和基线时的总能量摄入进行了调整。模型 2 还控制了高血压、高脂血症、MetS、T2D、CHD、中风和 T2D 药物。Model 3进一步调整了蔬菜、水果、红肉和加工肉类、鱼类、乳制品、咖啡和茶的膳食摄入量。使用排列方差分析（PERMANOVA）（999 次排列）检查了基于属级 Bray-Curtis 距离的慢性失眠与β多样性差异之间的关联。

In the GNHS, there was a significant difference in the gut microbial structure for the New-onset group or Long-term chronic insomnia group, compared with the Long-term healthy group (Fig. 1b and Supplementary Fig. ¹). To identify robust microbial biomarkers of chronic insomnia and increase the sample size, we combined the New-onset group and Long-term chronic insomnia group into Chronic insomnia group. We used MaAsLin to identify potential chronic insomnia associated gut microbiota (q value < 0.25 was used as the threshold of significance in the exploratory analyses, as commonly used previously^26,79) using the above three different statistical models by comparing the Chronic insomnia group with the Long-term healthy group. The Benjamini-Hochberg method was used to control the false discovery rate (FDR).
在GNHS中，与长期健康组相比，新发组或长期慢性失眠组的肠道微生物结构存在显著差异（图1b和补充图1b）。 ¹ ）。为了确定慢性失眠的可靠微生物生物标志物并增加样本量，我们将新发组和长期慢性失眠组合并为慢性失眠组。我们使用MaAsLin通过比较慢性失眠组和长期健康组，使用上述三种不同的统计模型来识别潜在的慢性失眠相关肠道菌群（q值<0.25用作探索性分析中的显着性阈值，如前所述 ^26,79 ）。Benjamini-Hochberg方法用于控制错误发现率（FDR）。

Next, we used OPLS-DA to identify potential bile acids associated with chronic insomnia. We further used linear regression, adjusted for the same covariates as above model 3, to confirm the association of chronic insomnia with the OPLS-DA selected bile acids. Given that dietary cholesterol intake and fiber intake might be potential confounders affecting the relationship between chronic insomnia and the bile acid pool, we further performed a sensitivity analysis by including dietary cholesterol intake and fiber intake as additional covariates in the above model 3. We examined the association of the above identified gut microbiota biomarkers with bile acid biomarkers using partial correlation analysis, adjusted for age, sex and BMI. In addition, we tested the association of chronic insomnia with another two important classes of gut microbial metabolites (short-chain fatty acids, aromatic amino acids and their derivatives) by using multivariable linear regression, adjusted for the same covariates as above in model 3. The Benjamini-Hochberg method was used to control FDR.
接下来，我们使用OPLS-DA来鉴定与慢性失眠相关的潜在胆汁酸。我们进一步使用线性回归，针对与上述模型 3 相同的协变量进行调整，以确认慢性失眠与 OPLS-DA 选择的胆汁酸的关联。鉴于膳食胆固醇摄入量和纤维摄入量可能是影响慢性失眠与胆汁酸库之间关系的潜在混杂因素，我们进一步进行了敏感性分析，将膳食胆固醇摄入量和纤维摄入量作为上述模型 3 中的额外协变量。我们使用部分相关性分析检查了上述已鉴定的肠道微生物群生物标志物与胆汁酸生物标志物的关联，并根据年龄、性别和 BMI 进行了调整。此外，我们通过使用多变量线性回归测试了慢性失眠与另外两类重要肠道微生物代谢物（短链脂肪酸、芳香族氨基酸及其衍生物）的关联，并针对上述模型 3 中的相同协变量进行了调整。Benjamini-Hochberg方法用于控制FDR。

To gain further mechanistic insight into the connection between the chronic insomnia and CMD, we investigated the correlation of the chronic insomnia-related microbial and bile acid biomarkers with different CMD and risk factors (BMI, DBP, SBP, waist circumference, and fasting serum levels of TG, TC, HDL, LDL, glucose, insulin, and HbA1c) using multivariable logistic regression and linear regression model in the GNHS, adjusted for age, sex, smoking status, alcohol status, physical activity, education, income, and total energy intake. We further examined the prospective association of the above identified gut microbiota biomarkers with the incidence of CMD outcomes at the third follow-up using multivariable logistic regression, adjusting for age, sex, smoking status, alcohol status, physical activity, education, income, and total energy intake. Throughout the above analyses, correction of multiple testing was conducted by using the Benjamini-Hochberg method.
为了进一步了解慢性失眠与CMD之间的联系，我们使用GNHS中的多变量logistic回归和线性回归模型研究了慢性失眠相关微生物和胆汁酸生物标志物与不同CMD和危险因素（BMI、DBP、SBP、腰围和空腹血清TG、TC、HDL、LDL、葡萄糖、胰岛素和HbA1c水平）的相关性。 根据年龄、性别、吸烟状况、酒精状况、身体活动、教育程度、收入和总能量摄入进行调整。我们使用多变量逻辑回归，调整年龄、性别、吸烟状况、酒精状况、身体活动、教育、收入和总能量摄入，进一步检查了上述确定的肠道微生物群生物标志物与第三次随访时 CMD 结果发生率的前瞻性关联。在上述分析中，使用Benjamini-Hochberg方法对多重检验进行校正。

Based on the biological plausibility of the associations among the gut microbiota, bile acids and CMD^54,59,80,81, and our above findings, we performed mediation analysis to evaluate whether bile acids could mediate the association of the chronic insomnia related-gut microbiota with CMD outcomes (gut microbiota → bile acids → CMD). The mediation analysis was performed to examine the mediating effect of bile acids on the association of the chronic insomnia-related gut microbiota with CMD outcomes^82,83. We defined three pathways in the mediation analysis: (1) exposure to mediator; (2) mediator to outcome; and (3) exposure to outcome. In the mediation analysis, the covariates included age, sex, BMI, smoking status, alcohol status, physical activity, education, income, and total energy intake. The mediation analysis was performed using the R package “mediation” with the same parameter settings (boot = “TRUE”, boot.ci.type = “perc”, conf.level = 0.95, sims = 1000). The total effect was obtained through the sum of a direct effect and a mediated (indirect) effect. The percentage of the mediated effect was calculated using the formula: (mediated effect/total effect) × 100. Sensitivity analysis was performed to test the robustness of the mediation effect and violation of the assumption (sequential ignorability) using the R package “medsens” with default parameters^84,85. The reporting of the mediation results followed the Guideline for Reporting Mediation Analyses (AGReMA) statement⁸⁶.
基于肠道微生物群、胆汁酸和 CMD ^54,59,80,81 之间关联的生物学合理性，以及我们的上述发现，我们进行了中介分析，以评估胆汁酸是否可以介导慢性失眠相关肠道微生物群与 CMD 结果（肠道微生物群→胆汁酸→ CMD）的关联。通过中介分析来检验胆汁酸对慢性失眠相关肠道菌群与CMD结局 ^82,83 关联的中介作用。我们在中介分析中定义了三种途径：（1）暴露于中介;（2）结果的中介;（3）暴露于结果。在中介分析中，协变量包括年龄、性别、BMI、吸烟状况、酒精状况、身体活动、教育程度、收入和总能量摄入。使用具有相同参数设置（boot = “TRUE”， boot.ci.type = “perc”， conf.level = 0.95， sims = 1000） 的 R 包“中介”执行中介分析。总效应是通过直接效应和介导（间接）效应的总和获得的。使用以下公式计算介导效应的百分比：（介导效应/总效应）× 100。使用具有默认参数 ^84,85 的 R 包“medsens”进行敏感性分析以测试中介效果的鲁棒性和违反假设（顺序可忽略性）。调解结果的报告遵循了《报告调解分析指南》（AGReMA）声明 ⁸⁶ 。

Finally, we used a linear regression model to determine the prospective association of dietary factors with the gut microbial and bile acid mediators of chronic insomnia and CMD, adjusted for age, sex, BMI, smoking status, alcohol status, physical activity, education, income, dietary intake of vegetables/fruits/red and processed meat/fish/dairy products/coffee/tea) (mutual adjustment for each other) and total energy intake. The analyses were conducted among the GNHS participants without chronic insomnia or CMD at baseline.
最后，我们使用线性回归模型来确定饮食因素与慢性失眠和 CMD 的肠道微生物和胆汁酸介质的前瞻性关联，调整年龄、性别、BMI、吸烟状况、酒精状况、身体活动、教育、收入、蔬菜/水果/红肉和加工肉类/鱼类/乳制品/咖啡/茶的膳食摄入量（相互调整）和总能量摄入。这些分析是在基线时没有慢性失眠或CMD的GNHS参与者中进行的。

In the GGMP participants, we used a multivariable linear regression model to examine the association of chronic insomnia with the gut microbiota structure and the identified gut microbiota biomarkers, adjusting for age, sex, BMI, smoking status, alcohol status, education, dietary intake of vegetables, fruits, and red and processed meat. We conducted a secondary analysis to evaluate the association of the insomnia symptoms score (per unit change) with the identified gut microbiota biomarkers by using linear regression, adjusted for the same covariates. We also used logistic regression and linear regression to examine the association between the gut microbiota biomarkers and different CMD outcomes, adjusted for age, sex, smoking status, alcohol status, and education. For the GGMP participants, we did not include income in the statistical models due to a large number of missing values (income data were available among 3774 out of 6122 participants). We therefore performed a sensitivity analysis with further adjustment for income in the above analyses to examine the robustness of the models. Then, for each of the above linear regressions or logistic regressions, the effect estimates from the GNHS and the GGMP were pooled by random effects meta-analysis. In addition, we further performed additional interaction analysis and stratified analyses by age and sex to explore potential heterogeneity for the chronic insomnia-gut microbiota association and the gut microbiota-CMD association, and used random effects meta-analysis to pool the effect estimates from the GNHS and GGMP.
在GGMP参与者中，我们使用多变量线性回归模型来检查慢性失眠与肠道微生物群结构和已识别的肠道微生物群生物标志物的关联，调整年龄，性别，BMI，吸烟状况，酒精状况，教育，蔬菜，水果，红肉和加工肉类的饮食摄入量。我们进行了二次分析，通过使用线性回归评估失眠症状评分（单位变化）与已识别的肠道微生物群生物标志物的关联，并针对相同的协变量进行调整。我们还使用logistic回归和线性回归来检查肠道微生物群生物标志物与不同CMD结局之间的关联，并根据年龄、性别、吸烟状况、酒精状况和教育程度进行调整。对于GGMP参与者，由于大量缺失值，我们没有将收入纳入统计模型（6122名参与者中有3774名参与者的收入数据）。因此，我们进行了敏感性分析，并在上述分析中对收入进行了进一步调整，以检查模型的稳健性。然后，对于上述每个线性回归或逻辑回归，通过随机效应meta分析合并GNHS和GGMP的效应估计值。此外，我们进一步进行了额外的交互分析，并按年龄和性别进行了分层分析，以探索慢性失眠-肠道微生物群关联和肠道微生物群-CMD关联的潜在异质性，并使用随机效应meta分析来汇总GNHS和GGMP的效应估计。

In the GGMP, we also used multivariable linear regression to examine the association of dietary factors with gut microbial features of chronic insomnia and CMD, adjusted for age, sex, BMI, smoking status, alcohol status, education, dietary intake of vegetables/fruits/red and processed meat/tea/coffee (mutual adjustment for each other). The analyses were conducted among the GGMP participants without chronic insomnia or CMD. We also performed additional stratified analyses by tea consumption (yes versus no) using logistic regression in the GNHS to explore whether the associations between the chronic insomnia-related gut microbiota and CMD risk factors could be affected by tea consumption. We further investigated the association of tea consumption with the risk of chronic insomnia using logistic regression in the GNHS and GGMP and used random effects meta-analysis to pool the effect estimates from the GNHS and GGMP.
在GGMP中，我们还使用多变量线性回归来检查饮食因素与慢性失眠和CMD的肠道微生物特征的关联，并根据年龄，性别，BMI，吸烟状况，酒精状况，教育程度，蔬菜/水果/红肉和加工肉类/茶/咖啡的饮食摄入量进行调整（相互调整）。分析是在没有慢性失眠或CMD的GGMP参与者中进行的。我们还在GNHS中使用逻辑回归对茶饮量（是与否）进行了额外的分层分析，以探讨慢性失眠相关肠道微生物群与CMD危险因素之间的关联是否会受到茶叶消费的影响。我们进一步研究了茶叶消费与慢性失眠风险的关联，使用GNHS和GGMP的逻辑回归，并使用随机效应meta分析来汇总GNHS和GGMP的效应估计。

In the GNHS, we used the co-occurrence network analysis based on the above partial correlation coefficient to demonstrate the interaction of the above gut microbiota and bile acid biomarkers, and only the significant correlations (larger than 0.1 or smaller than −0.1) were used for network construction. The networks were further visualized in Cytoscape software version 3.7.2. We used R version 3.6.3 for statistical analysis unless otherwise specified, and p value < 0.05 was considered statistically significant.
在GNHS中，我们采用基于上述偏相关系数的共现网络分析来证明上述肠道菌群与胆汁酸生物标志物的相互作用，仅使用显著相关性（大于0.1或小于-0.1）进行网络构建。这些网络在 Cytoscape 软件版本 3.7.2 中进一步可视化。除非另有说明，否则我们使用 R 版本 3.6.3 进行统计分析，并且 p 值< 0.05 被认为具有统计学意义。

We thank the Westlake University Supercomputer Center for computational resources and related assistance and all the participants involved in the Guangzhou Nutrition and Health Study and the Guangdong Gut Microbiome Project. This study was funded by the National Natural Science Foundation of China (82073529, 81903316, 81773416), Zhejiang Ten-thousand Talents Program (2019R52039), Zhejiang Provincial Natural Science Foundation of China (LQ19C200005, LQ21H260002), Westlake Education Foundation and the 5010 Program for Clinical Research (2007032) of Sun Yat-sen University (Guangzhou, China).
我们感谢西湖大学超级计算机中心的计算资源和相关帮助，以及所有参与广州营养与健康研究和广东肠道微生物组项目的参与者。本研究由国家自然科学基金（82073529,81903316,81773416），浙江省万人计划（2019R52039），中国浙江省自然科学基金（LQ19C200005，LQ21H260002），西湖教育基金会和中山大学（中国广州）5010临床研究计划（2007032）资助。