Gut microbiota-associated metabolites and risk of ischemic stroke in REGARDS

Abstract

Several metabolite markers are independently associated with incident ischemic stroke. However, prior studies have not accounted for intercorrelated metabolite networks. We used exploratory factor analysis (EFA) to determine if metabolite factors were associated with incident ischemic stroke. Metabolites (n = 162) were measured in a case-control cohort nested in the REasons for Geographic and Racial Differences in Stroke (REGARDS) study, which included 1,075 ischemic stroke cases and 968 random cohort participants. Cox models were adjusted for age, gender, race, and age-race interaction (base model) and further adjusted for the Framingham stroke risk factors (fully adjusted model). EFA identified fifteen metabolite factors, each representing a well-defined metabolic pathway. Of these, factor 3, a gut microbiome metabolism factor, was associated with an increased risk of stroke in the base (hazard ratio per one-unit standard deviation, HR = 1.23; 95%CI = 1.15–1.31; P = 1.98 × 10⁻¹⁰) and fully adjusted models (HR = 1.13; 95%CI = 1.06–1.21; P = 4.49 × 10⁻⁴). The highest tertile had a 45% increased risk relative to the lowest (HR = 1.45; 95%CI = 1.25–1.70; P = 2.24 × 10⁻⁶). Factor 3 was also associated with the Southern diet pattern, a dietary pattern previously linked to increased stroke risk in REGARDS (β = 0.11; 95%CI = 0.03–0.18; P = 8.75 × 10⁻³). These findings highlight the role of diet and gut microbial metabolism in relation to incident ischemic stroke.

Keywords

Case-cohort exploratory factor analysis gut microbiome ischemic stroke metabolomics

Introduction

Traditional risk factors can explain a large proportion of the overall risk of ischemic stroke, and several preventative strategies, including lifestyle modifications and medications, successfully reduced stroke incidence.¹ However, the overall burden of stroke has remained high.² Improved prevention may depend on the identification of additional risk factors or a better understanding of known risk factors.^3–7

Blood-based biomarkers can provide insights into the mechanisms of disease progression and might provide valuable information about pathophysiological events leading to ischemic stroke.¹ Small molecule metabolites comprise an essential subset of blood-based biomarkers, which offer representations of phenotypic traits as they relate to cellular and metabolic processes. Several studies have identified metabolites associated with incident stroke. However, most studies were based on relatively small sample sizes,^8–10 assessed post-stroke metabolic disturbances,^11,12 were focused on European ancestry populations,^13,14 or assessed metabolite-stroke associations without accounting for underlying metabolite intercorrelations.^15,16 There remains a need to identify additional markers of ischemic stroke risk in multi-ethnic cohorts while accounting for broader pathways reflected in correlations between individual metabolites.

The REasons for Geographic and Racial Differences in Stroke (REGARDS) study is a large-scale bi-racial prospective cohort and provides a unique opportunity to address the above knowledge gaps. In this study, we sought to identify underlying metabolite correlations and their association with risk of ischemic stroke. Because candidate correlating metabolite patterns were related to gut microbiome metabolism, we also assessed for relationships between these metabolites and dietary patterns that have previously been established as risk factors for stroke in REGARDS.^5,17

Materials and methods

Study population

The REGARDS study is a national population-based cohort study which, between January 2003 and October 2007, enrolled 30,239 individuals with ≥45 years of age.³ The study was designed to oversample participants in the stroke belt and buckle regions of the United States and recruited 50% Black and 50% White adults. Demographics, socioeconomic factors, medical history, and verbal informed consent were obtained through a phone interview, which was followed by an in-home examination. During the in-home examination, written informed consent, blood pressures, anthropometric measurements, fasting blood samples, electrocardiogram (ECG), and medication inventory were obtained. Each participant, or their designated proxies, were then contacted every six months by telephone.

A committee of stroke physicians determined the designation of stroke by reviewing all suspected strokes and stroke-related symptoms. Stroke was defined as a focal neurological deficit lasting >24 hours or non-focal neurological symptoms with positive imaging findings.¹⁸ All stroke cases adjudicated in the follow up period between January 2003 and April 2019 were included.¹⁵ We used a nested case-cohort design¹⁹ in which the random control participants were selected from the entire cohort. Random selection was stratified based on age, gender, and race to ensure adequate representation of each demographic group. Participants of the random cohort who had an incident stroke during the follow up period (n = 68) were retained as contributors to the random cohort and included in the incident stroke group. Because of this overlap, formal P-values were not calculated to compare the random cohort and incident stroke groups. Further details regarding the methods and the case-cohort study design of the REGARDS have been described in related publications.^3,4,20–23

Informed consent was obtained from all participants included in the study. All procedures performed were in accordance with the ethical standards of the institutional and national research committees and with the 1975 Helsinki Declaration and its later amendments. The metabolomics analysis was conducted under IRB approval by the Mass General Brigham Human Research Committee (MGBHRC, protocol number: IRB 2016P001801).

Incident stroke cases and stroke subtyping

Medical records were retrieved and centrally reviewed by stroke neurologists to confirm the diagnosis, stroke type, and possible etiology. Adjudicators classified events by stroke type for every recorded event. In cases of disagreement, additional adjudicators reviewed the event. A stroke event was recorded if all reviewers agreed on the occurrence of stroke. Ischemic stroke cases (n = 1075) were further subtyped using the Trial of ORG 10172 in Acute Stroke Treatment (TOAST) classification system ²⁴ into cardioembolic, small vessel disease, large vessel disease, stroke of other determined etiology, and undetermined etiology.^20,24 Missing details in data collection lead to incomplete evaluations, and approximately one-half of the ischemic stroke cases did not have sufficient information to be subtyped. These cases were labeled as unclassified. There was no observable association between geographic region and missing stroke subtyping data.

Covariates

We included covariates as defined in the Framingham stroke risk score function.^18,25 Age, race, and smoking status were determined by self-report. During the in-home visits, systolic and diastolic blood pressure (SBP and DBP) were measured twice, and the average of these two baseline measurements were used. Hypertension was defined as SBP ≥140 mm Hg, DBP ≥90 mm Hg, or self-reported use of antihypertensive medications. Type 2 diabetes was defined as either current diabetes medication use or as blood glucose concentrations of ≥126 mg/dL and ≥200 mg/dL for fasted and non-fasted states, respectively. Cardiovascular disease was defined as self-reported history of myocardial infarction, coronary revascularization procedure, or baseline evidence of prior myocardial infarction on the study ECG. The ECG was also used to classify left ventricular hypertrophy and atrial fibrillation.

Metabolomics analysis

All plasma samples were collected at baseline and stored at −80°C until analysis. Targeted metabolomics analyses were carried out using liquid chromatography-triple quadrupole mass spectrometry (LC-QQQ) as reported previously.¹⁵ Metabolites were extracted using protein precipitation from 30 μL of EDTA plasma samples and chromatographed using Xbridge Amide columns (2.1 × 100 mm 3.5 µm, Waters, Milford, MA) and dual Infinity II 1290 HLPC pumps (Agilent, Santa Clara, CA). The sample preparation procedure was optimized to allow the use of low sample volumes with a high throughput extraction method, and recover metabolites with a wide range of polarity, structural moieties, and pKa. Mass spectra were detected using a 6495 QQQ tandem mass spectrometer (Agilent, Santa Clara, CA) with scheduled multiple reaction monitoring (MRM) events. The MRM transitions included a list of metabolites which were selected to cover the maximal range of compounds in key biochemical pathways that were compatible with the extraction procedure and column chemistry. Metabolite peaks (n = 162) were integrated using MassHunter QQQ Quantitative Analysis software (Agilent, Santa Clara, CA). Results were normalized using the nearest pooled plasma metabolite values based on standard approaches.¹² Rank-based inverse normal transformation was applied prior to statistical analyses.

To account for potential variability based on storage duration, sample storage time was calculated from baseline sample draw to the date of metabolomics analysis.

Statistical analyses

To generate the latent metabolite factors that reflected the underlying relationship among the measured metabolites, we used exploratory factor analysis (EFA).²⁶ EFA was applied using principal component factoring to the entire case-cohort metabolomics dataset, and correlations between the measured metabolites were calculated. A minimum eigenvalue threshold of 2 was used to determine the number of factors, and to retain a metabolite within a factor, a minimum threshold of 50% factor loading value was used. For improved interpretability, the factor matrix was rotated using varimax rotation. Fifteen metabolite factors were identified, and accordingly, a Bonferroni-corrected value of P < 3.33 × 10⁻³ was used to account for multiple comparisons.

Weighted Cox proportional hazards (Cox-PH) regression was used to evaluate associations between the identified EFA metabolite factors and incident ischemic stroke. Weightings were calculated based on age, race, and gender stratification, as described in prior studies.^22,27,28 Hazard ratios (HR) per one-unit standard deviation and 95% confidence intervals (CIs) were calculated for each metabolite factor after adjustment for age, race, gender, age-race interaction,²¹ and sample storage time ²⁹ (base model). A fully adjusted model additionally included the Framingham stroke risk factors: current smoking status, hypertension, type 2 diabetes, cardiovascular disease, left ventricular hypertrophy, and atrial fibrillation.²⁵ Factors associated with incident ischemic stroke were also divided into tertiles and treated categorically. Cox-PH analyses were performed for each factor tertile to calculate the HR of incident ischemic stroke in the base and in the fully adjusted models. Weighted Cox-PH regression was also used to test the association of metabolite factors with each incident ischemic stroke subtype.

Because the identified factors that passed the correction threshold were related to gut microbial metabolism, we also studied the relationship between these metabolite factors and dietary patterns, which had been previously shown to associate with stroke in REGARDS.¹⁷ The dietary patterns and metabolite factors were assessed using weighted linear regression models adjusting for age, race, and gender. Statistical analyses were performed using SAS version 9.4 (Cary, NC) and Stata version 15 (College Station, TX).

Analyses were reported in accordance with the STROBE guidelines.

Results

Characteristics of the study population

There were 1,404 stroke cases identified through April 2019 with an average follow-up time of 7.1 +/− 4.5 years. Individuals were excluded if they had a history of stroke at baseline (n = 211), developed incident hemorrhagic stroke (n = 122), did not have plasma samples available (n = 55), and had a missing record of follow-up time (n = 9). The random cohort sample included 1,127 participants. Individuals with a history of stroke at baseline (n = 97), no plasma samples available (n = 55), or a missing record of follow-up time (n = 7) were similarly excluded.

Overall, 1,075 incident ischemic stroke cases and 968 individuals from the random cohort were included in the final analysis (Supplementary Figure 1). Table 1 details the baseline characteristics of the study population. Ischemic stroke cases were subtyped based on the TOAST classification system and included cardioembolic stroke (n = 239), small vessel disease stroke (n = 148), large vessel disease stroke (n = 134), stroke of other determined etiology (n = 49) and stroke of undetermined etiology (n = 505). In total, stroke subtyping with a specific diagnosis was available for 570 participants (53% of all ischemic stroke cases), leaving 505 cases where the cause of stroke was undetermined due to incomplete evaluation.²⁰ The characteristics of the cohort by stroke subtypes are presented in Supplementary Table 1.

Table 1.

Baseline characteristics of the random cohort and incident ischemic stroke cases in REGARDS. In this case cohort study, 68 participants in the random cohort group became cases during the follow-up period, therefore accurate calculation of p- values are not possible and were omitted.

	Random cohort (n = 968)	Ischemic stroke (n = 1075)
Age (yrs. ± SD)	64 ± 9	70 ± 9
Female #. (%)	532 (55)	533 (50)
Black # (%)	387 (40)	439 (41)
Left ventricular hypertrophy # (%)	77 (8)	156 (15)
Atrial fibrillation # (%)	87 (9)	145 (14)
Smoking # (%)	135 (14)	176 (16)
Hypertension # (%)	697 (72)	907 (85)
Type 2 diabetes # (%)	194 (20)	306 (29)
Cardiovascular disease # (%)	155 (16)	303 (29)
Systolic blood pressure (mm Hg ± SD)	127.1 ± 16.5	132.4 ± 17.4

Metabolite factors

EFA identified fifteen metabolite factors that exceeded a minimum eigenvalue threshold of 2. Each factor represented a distinct metabolic pathway or process, such as metabolites associated with the carnitine shuttle, redox homeostasis, gut microbiome, and glyceric acid metabolism. Individual metabolites constituting each of the 15 factors, the associated metabolite processes, and the factor loadings are detailed in Supplementary Table 2.

Taken together, the 15 factors explained 83% of the variance in the metabolomics data. Factor 1 (carnitine shuttle metabolites) explained 12.1% of the variance; factor 2 (redox homeostasis) and factor 3 (gut microbial metabolism) each explained 9.6% of the variance; and factor 15 explained 3.3% of the variance (Supplementary Figure 2).

Metabolite factor associations with incident ischemic stroke

Associations of each identified metabolite factor with incident ischemic stroke were examined using Cox-PH models (Table 2). In the base model, the gut microbial metabolism factor (factor 3; HR = 1.23; 95% CI = 1.15–1.31; P = 1.98 × 10⁻¹⁰) and glyceric acid metabolism factor (factor 15; HR = 0.87; 95% CI =0.81–0.92; P = 1.0610⁻⁵) were both associated with incident stroke. In the fully adjusted model—which included smoking, atrial fibrillation, hypertension, type 2 diabetes, cardiovascular disease and left ventricular hypertrophy—the gut microbial metabolism factor remained significantly associated with incident ischemic stroke risk (factor 3; HR = 1.13; 95% CI = 1.06–1.21; P = 4.49 × 10⁻⁴), but the glyceric acid metabolism factor did not (factor 15; HR = 0.94; 95% CI = 0.88–1.01; P = 0.09; Table 2). The constituents of factor 3, representing gut microbial metabolism, are detailed in Table 3. The associations between each individual metabolite of factor 3 with incident ischemic stroke are reported in Supplementary Table 3. For example, these include the gut microbial metabolite trimethylamine-N-oxide (TMAO), previously implicated in incident cardiovascular disease.⁴²

Table 2.

Associations between metabolite factors and incident ischemic stroke (n = 1,075).

	Base model			Fully adjusted model
	HR	95% CI	p	HR	95% CI	p
Factor 1	1.04	0.98–1.11	1.82E−01	1.01	0.94–1.08	8.15E−01
Factor 2	1.01	0.95–1.07	7.51E−01	1.00	0.94–1.06	9.88E−01
Factor 3	1.23	1.15–1.31	1.98E−10*	1.13	1.06–1.21	4.49E−04*
Factor 4	0.98	0.90–1.08	7.55E−01	0.96	0.91–1.05	3.65E−01
Factor 5	0.99	0.93–1.06	8.25E−01	0.98	0.91–1.05	5.39E−01
Factor 6	0.92	0.86–0.97	4.90E−03	0.95	0.89–1.02	1.34E−01
Factor 7	1.04	0.98–1.10	2.30E−01	1.00	0.94–1.07	8.83E−01
Factor 8	1.01	0.95–1.07	8.70E−01	0.99	0.93–1.06	8.50E−01
Factor 9	1.00	0.93–1.07	9.73E−01	1.00	0.93–1.07	9.10E−01
Factor 10	1.02	0.96–1.08	5.50E−01	1.00	0.95–1.08	8.93E−01
Factor 11	1.01	0.95–1.07	6.97E−01	1.01	0.96–1.08	6.90E−01
Factor 12	1.03	0.97–1.10	2.97E−01	1.02	0.95–1.08	5.73E−01
Factor 13	1.06	1.00–1.13	4.01E−02	1.06	0.99–1.12	8.37E−02
Factor 14	1.08	1.02–1.15	1.19E−02	1.01	0.95–1.09	7.14E−01
Factor 15	0.87	0.81–0.92	1.06E−05*	0.94	0.88–1.01	9.00E−02

Hazard ratios (HR) with confidence intervals (CIs) were calculated for each metabolite factor and represent the HR per one unit of standard deviation. The base Cox model included age, race, gender, age by race interaction, length of sample storage time, and metabolite factor. Fully adjusted models included age, race, gender, age by race interaction, length of sample storage time, metabolite factor, smoking status, atrial fibrillation, hypertension, diabetes, cardiovascular disease, and left ventricular hypertrophy. Starred* values exceed the Bonferroni-corrected threshold (p < 3.33E−03).

Table 3.

List of metabolites associated with Factor 3 (gut microbial metabolism).

Metabolite	Loading %	Metabolic function
Guanosine	74	The complementary nucleoside of cytidine.
Dimethylguanidino valeric acid (DMGV)	69	Nutritional marker of sugary beverages; associated with incident coronary artery disease, type 2 diabetes, and liver fat.^30,31
Cystathionine	62	A transsulfuration pathway metabolite.
Kynurenic acid	61	A host-microbiome cometabolite associated with bacterial tryptophan metabolism.³²
D-Gluconic acid	58	Oxidized derivative of glucose, a host-microbiome cometabolite.³³
Indole-3-lactic acid	58	A degradation product of tryptophan, exclusively produced by gut microbial metabolism.³⁴
Acetyl-neuraminic acid	57	Nine-carbon sugar salicylic acid forms mucus layers of the gastrointestinal tract, associated with intestinal inflammation and infection by the microbiome.³⁵
Kynurenine	57	A host-microbiome cometabolite from the degradation product of the aromatic amino acid tryptophan.^36,37
S-Adenosyl-homocysteine (SAH)	56	A methylation cycle metabolite and host-microbiome cometabolite.^32,38
Cytidine	54	The complementary nucleoside of guanine.
Trimethylamine-N-oxide (TMAO)	54	The degradation product of trimethylamine that is exclusively produced by gut microbial metabolism.³⁹
S-Adenosyl-methionine (SAM)	53	A methylation cycle metabolite and methyl donor,⁴⁰ SAM-enzymes are key in bacterial host colonization.⁴¹
Hydroxyphenyl pyruvic acid	53	A gut microbial degradation product of the aromatic amino acid tyrosine.³⁴

Participants in the third vs. first tertile of the gut microbial metabolite factor had a 45% increased risk of stroke (factor 3; HR = 1.45; 95%CI = 1.25–1.70; P = 2.24 × 10⁻⁶; Figure 1(a)) in the base model and a 23% increase in risk for stroke in the fully adjusted model (factor 3; HR = 1.23; 95%CI = 1.04–1.44; P = 1.42 × 10⁻²). In contrast, the highest tertile of glyceric acid metabolism factor (vs. lowest) was associated with a 25% reduction in incident ischemic stroke risk (factor 15; HR = 0.75; 95%CI = 0.64–0.87; P =1.92 × 10⁻⁴, Figure 1(b)). However, this inverse association was not significant in the fully adjusted model (factor 15; HR = 0.87; 95%CI = 0.74–1.02; P = 0.09).

Figure 1.

Stroke-free survival estimates of the low, medium, and high factor tertiles for factors 3 and 15. (a) Stroke-free survival probability for participants with different levels of the gut microbiome metabolism factor (factor 3). Survival probability decreased from low to medium to high levels; P = 2.24 × 10⁻⁶ and (b) The glyceric acid metabolism factor (factor 15) showed the inverse trend, in which the survival probability decreased from high to medium to low levels; P = 1. 42 × 10⁻². The data input for the survival curves were the baseline adjusted Cox-PH regression models.

Metabolite factor associations with incident ischemic stroke subtypes

We next examined whether factor 3 or factor 15 were associated with one or more TOAST stroke subtypes (Table 4). In the base model, gut microbial metabolism was associated with an increased risk of incident cardioembolic stroke (n = 239; factor 3; HR = 1.36; 95%CI = 1.18–1.55; P = 1.31 × 10⁻⁵). However, In the fully adjusted model, factor 3 did not remain associated with a risk of cardioembolic stroke (factor 3; HR = 0.89; 95%CI = 0.78–1.02; P = 0.60.

Table 4.

Associations between incident ischemic stroke subtypes and metabolite factor 3 (gut microbial metabolism) and factor 15 (glyceric acid metabolism).

	Factor 3			Factor 15
Stroke subtype	HR	95% CI	p	HR	95% CI	P
Cardioembolic (n = 239)	1.36	1.18–1.55	1.31E−05*	0.89	0.78–1.02	9.29E−02
Small vessel disease (n = 148)	1.12	0.94–1.34	2.16E−01	0.79	0.67–0.94	7.03E−03
Large vessel disease (n = 134)	1.25	1.04–1.51	1.58E−02	0.87	0.73–1.05	1.45E−01
Other (n = 49)	1.31	0.97–1.76	8.33E−02	0.91	0.67–1.23	6.27E−01
Undetermined (n = 505)	1.32	1.21–1.45	2.13E−09*	0.86	0.78–0.94	8.62E−04*

Hazard ratios (HRs) with confidence intervals (CIs) represent the HR for each stroke type per unit of the standard deviation of the baseline level of each metabolite factor. Cox models were adjusted for age, race, gender, age byrace interaction, and length of sample storage time. Starred* values exceed the Bonferroni-corrected threshold (p < 5.00E−3).

Strokes of undetermined etiology (n = 505) were also associated with both factor 3 (HR = 1.32; 95%CI = 1.21–1.45; P = 2.13 × 10⁻⁹) and factor 15 (HR =0.86; 95%CI = 0.78–0.94; P = 8.62 × 10⁻⁴). This association remained significant in the fully adjusted model for factor 3 (HR = 1.23; 95%CI = 1.12–1.36; P = 3.16 × 10⁻⁵) but not for factor 15 (HR = 0.93, 95%CI = 0.81–1.03; P = 0.15).

Effect modification by race, gender and age

To assess whether there were racial differences in the risk of stroke, effect modification was examined by including an interaction term. However, there were no evidence of effect modification by race for either the gut microbial metabolism factor (factor 3; P = 0.71) or glyceric acid metabolism factor (factor 15; P = 0.81). A complete list of interactions with all factors is presented in Supplementary Table 4.

We also examined for effect modification by gender and age. There was no evidence for an effect by gender, but age demonstrated an interaction with factor 3 (P = 5.5 × 10⁻⁵). The main association between factor 3 and incident stroke remained significant after the inclusion of factor 3 by age interaction term in the base model (HR = 2.98; 95%CI = 1.93–4.58; P = 7.12 × 10⁻⁷) and fully adjusted model (HR = 2.46; 95%CI = 1.56–3.87; P = 9.93 × 10⁻⁵). The strongest effect of factor 3 on incident stroke risk was observed among younger participants (<65 years old; see Supplementary Table 4 and Supplementary Figure 3).

Metabolite factors are associated with dietary patterns

We next examined the association between gut microbial factor and glyceric acid metabolism factor with dietary patterns (Table 5). A previous REGARDS study⁵ had found that a plant-based diet pattern and a Southern dietary pattern had opposing associations with stroke risk. Accordingly, a plant-based diet was inversely associated with the gut microbial metabolism factor (factor 3; β = −0.10; 95%CI = −0.18– −0.03; P = 5.28 × 10⁻³) and positively associated with the glyceric acid metabolism factor (factor 15; β = 0.21; 95%CI = 0.13–0.28; P = 7.83 × 10⁸; Table 5). In contrast, the Southern dietary pattern was positively associated with the gut microbial metabolism factor (factor 3; β = 0.11; 95%CI = 0.03–0.18; P = 8.75 × 10⁻³) and inversely associated with glyceric acid metabolism factor (factor 15; β = −0.22; 95%CI = −0.31– −0.13; P = 1.98 × 10⁻⁶; Table 5). A complete list of metabolite factor associations with a plant-based and Southern dietary patterns is provided in Supplementary Table 5.

Table 5.

Associations between dietary patterns related to ischemic stroke risk and metabolite factor 3 (gut microbial metabolism) and factor 15 (glyceric acid metabolism).

	Plant based				Southern
	β	95% CI		p	β	95% CI		p
Factor 3	−0.10	−0.18	−0.03	5.3E−03	0.11	0.03	0.18	8.8E−03
Factor 15	0.21	0.13	0.28	7.8E−08	−0.22	−0.31	−0.13	2.0E−06

Beta values (β) with confidence intervals (CIs) represent changes in diet scores for each one-unit increase in the factors. Weighted linear regression models were adjusted for age, race, and gender.

Discussion

Association of gut microbiota with circulating metabolites

In this study, we assessed whether correlated networks of metabolites were associated with the risk of incident stroke. Using EFA, we identified 15 underlying metabolite factors which were related to specific metabolic processes. Of these 15 metabolite factors, factor 3 was enriched for gut microbiome-related metabolites, and only this factor was also associated with an increased risk of ischemic stroke in both the base and fully adjusted models. We further demonstrated that factor 3 was associated with a cardioembolic stroke subtype in a base model and positively associated with a Southern dietary pattern. Collectively, these observations accord with studies that highlight the interaction between gut microbial metabolism and the central nervous system⁴³ and implicate the gut-brain axis in association with stroke risk.^44,45

The gut microbiome factor was comprised of metabolites that include aromatic amino acid derivatives (e.g., hydroxyphenyl pyruvic acid, indole-3-lactic acid, kynurenic acid, and kynurenine; Table 3). The aromatic amino acids tyrosine, tryptophan, and phenylalanine and their derivatives play a central role in metabolism among human gut microflora, which in turn metabolize these substrates into phenyl carboxylic acids.^36,37,46 Altered abundance of these metabolites has previously been implicated in acute stroke, including elevated kynurenine levels⁴⁷ and an increased ratio of kynurenine and tryptophan.^48,49 Other metabolites are synthesized exclusively by gut microbes and not endogenously produced by the host. These metabolites include trimethylamine N-oxide (TMAO), which is produced through gut microbial metabolism of choline and its related metabolites.⁵⁰ TMAO has also been implicated in prior studies as a promoter of cardiovascular disease risk, atherosclerosis,^42,50–52 and stroke severity.^39,53–56

In contrast, the glyceric acid metabolism factor was associated with a reduction in the risk of stroke, though this association was not significant in the fully adjusted model. Our data suggest a relationship between dietary patterns, gut microbial metabolism, and metabolite markers associated with stroke risk and support the hypothesis that diet may impact stroke risk by altering the composition of the gut microbiome and the resulting metabolite products entering host circulation. Prior studies have extensively catalogued host-gut microbe interactions and the diversity of circulating metabolites produced as a consequence,^34,57,58 including the ones measured in this study. However additional work is needed to clarify the causal links between gut microbial species, the metabolites they produce, and how dietary modifications may influence them.

Environmental and social factors may also influence risk for stroke, with recognition of the impact of commensal microorganisms on health and disease.^59,60 Emerging gut-brain-microbiome research using murine models, humanized mouse models, and human clinical trials^61,62 has shown that microbiome-related metabolic changes can be induced by stress,⁶³ exercise,⁶⁴ and even changes in the circadian rhythm. These alterations in metabolite composition have been shown across different brain regions, peripheral organs, and in circulation.⁶⁵ Therefore, understanding the cross-talk between the gut-microbiome and additional metabolic pathways is critical to unraveling the potential impact on clinical outcomes.

This study has several strengths. First, the REGARDS cohort is a large and well characterized observational study designed to investigate risk factors of stroke. Second, we used EFA to identify underlying or latent metabolite factors that explain the covariance within the data. We demonstrated that EFA can be successfully applied to metabolomics datasets and can uncover complex relationships such as metabolites altered by gut microbial metabolism. However, there are also several limitations. Since the metabolite factors are empirically derived, EFA may not be directly replicable in independent datasets. Nevertheless, EFA is well-suited to highlight important insights into metabolite networks and their relationship to disease. Another limitation is the small number of identified ischemic stroke subtypes, as TOAST etiology was undetermined in approximately half of the incident ischemic stroke cases. This may have contributed to some null findings in the fully adjusted models. Finally, direct associations with specific gut microbial species could not be assessed since plasma but not stool samples were collected and analyzed. However, prior studies have highlighted that several metabolites identified in this study are not produced by the host and are exclusive metabolic products of gut microbes.

Taken together, our findings contribute to the understanding of relationships between diet, the gut microbiome, and circulating metabolites. They support the role for gut microbial metabolism in influencing host metabolite concentrations and risk of ischemic stroke. These findings raise the possibility of developing gut microbiome-targeted interventions for the prevention of stroke.

Supplemental Material

sj-pdf-1-jcb-10.1177_0271678X231162648 - Supplemental material for Gut microbiota-associated metabolites and risk of ischemic stroke in REGARDS

Supplemental material, sj-pdf-1-jcb-10.1177_0271678X231162648 for Gut microbiota-associated metabolites and risk of ischemic stroke in REGARDS by Zsuzsanna Ament, Amit Patki, Varun M Bhave, Ninad S Chaudhary, Ana-Lucia Garcia Guarniz, Naruchorn Kijpaisalratana, Suzanne E Judd, Mary Cushman, D Leann Long, M Ryan Irvin and W Taylor Kimberly in Journal of Cerebral Blood Flow & Metabolism

Footnotes

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and publication of this article: this work was supported by the National Institutes of Health (NIH) R01 NS099209 (WTK), American Heart Association (AHA) 17CSA33550004 (WTK). REGARDS is supported by cooperative agreement U01 NS041588 co-funded by the National Institute of Neurological Disorders and Stroke (NINDS) and the National Institute on Aging (NIA), NIH, Department of Health and Human Service.

Acknowledgements

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NINDS or the NIA. Representatives of the NINDS were involved in the review of the manuscript but were not directly involved in the collection, management, analysis or interpretation of the data. The authors thank the other investigators, the staff, and the participants of the REGARDS study for their valuable contributions. A full list of participating REGARDS investigators and institutions can be found at: .

Data availability

Qualified investigators may request access to obtain de-identified data under institutional data sharing agreements.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Authors’ contributions

MRI and WTK had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Data acquisition and analysis: ZA, WTK. Statistical analysis: ZA, AP, VMB, MRI, SEJ, LL. Manuscript draft: ZA WTK. Critical revision and important intellectual content: VMB, NSC, AGG, NK, SEJ, MC, LL, MRI. All authors revised, edited and approved the final version of the manuscript.

ORCID iD

Zsuzsanna Ament

Supplemental material

Supplemental material for this article is available online.

References

Montellano

Ungethüm

Ramiro

, et al. Role of blood-based biomarkers in ischemic stroke prognosis: a systematic review. Stroke 2021; 52: 543–551.

GBD 2019 Stroke Collaborators. Global, regional, and national burden of stroke and its risk factors, 1990-2019: a systematic analysis for the global burden of disease study 2019. Lancet Neurol 2021; 20: 795–820.

Howard

Cushman

Pulley

, et al. The reasons for geographic and racial differences in stroke study: objectives and design. Neuroepidemiology 2005; 25: 135–143.

Zakai

Judd

Alexander

, et al. ABO blood type and stroke risk: the REasons for Geographic And Racial Differences in Stroke Study. J Thromb Haemost 2014; 12: 564–570.

Judd

Gutiérrez

Newby

, et al. Dietary patterns are associated with incident stroke and contribute to excess risk of stroke in black Americans. Stroke 2013; 44: 3305–3311.

Bang

Ovbiagele

Kim

JS.

Nontraditional risk factors for ischemic stroke. Stroke 2015; 46: 3571–3578.

Jenny

Callas

Judd

, et al. Inflammatory cytokines and ischemic stroke risk. Neurology 2019; 92: e2375–e2384.

Wang

Zhang

Sun

, et al. Changes of metabolites in acute ischemic stroke and its subtypes. Frontiers in Neuroscience 2021; 0: 1346.

Chumachenko

Waseem

Fedorovich

SV.

Metabolomics and metabolites in ischemic stroke. Rev Neurosci 2022; 33: 181–205.

10.

Pan

C-W

Zhang

, et al. Metabolomics facilitates the discovery of metabolic biomarkers and pathways for ischemic stroke: a systematic review. Metabolomics 2019; 15: 1–21.

11.

Nelson

Ament

Wolcott

, et al. Succinate links atrial dysfunction and cardioembolic stroke. Neurology 2019; 92: e802–e810.

12.

Ament

Bevers

Wolcott

, et al. Uric acid and gluconic acid as predictors of hyperglycemia and cytotoxic injury after stroke. Transl Stroke Res 2021; 12: 293–302.

13.

Floegel

Kühn

Sookthai

, et al. Serum metabolites and risk of myocardial infarction and ischemic stroke: a targeted metabolomic approach in two german prospective cohorts. Eur J Epidemiol 2018; 33: 55–66.

14.

Lind

Salihovic

Ganna

, et al. A multi-cohort metabolomics analysis discloses sphingomyelin (32:1) levels to be inversely related to incident ischemic stroke. J Stroke Cerebrovasc Dis 2020; 29: 1–8.

15.

Ament

Patki

Chaudhary

, et al. Nucleosides associated with incident ischemic stroke in the REGARDS and JHS cohorts. Neurology 2022; 98: e2097–e2107.

16.

Sun

Tiedt

, et al. A prospective study of serum metabolites and risk of ischemic stroke. Neurology 2019; 92: e1890–e1898.

17.

Judd

Letter

Shikany

, et al. Dietary patterns derived using exploratory and confirmatory factor analysis are stable and generalizable across race, region, and gender subgroups in the REGARDS study. Front Nutr 2014; 1: 29.

18.

Howard

Cushman

Kissela

, et al. Traditional risk factors as the underlying cause of racial disparities in stroke: Lessons from the half-full (empty?) glass. Stroke 2011; 42: 3369–3375.

19.

Prentice

RL.

A case-cohort design for epidemiologic cohort studies and disease prevention trials. Biometrika 1986; 73: 1–11.

20.

Cushman

Judd

Howard

, et al. N-terminal pro-B-type natriuretic peptide and stroke risk: the reasons for geographic and racial differences in stroke cohort. Stroke 2014; 45: 1646–1650.

21.

Howard

Kleindorfer

Judd

, et al. Disparities in stroke incidence contributing to disparities in stroke mortality. Ann Neurol 2011; 69: 619–627.

22.

Chaudhary

Bridges

Saag

, et al. Severity of hypertension mediates the association of hyperuricemia with stroke in the REGARDS case cohort study. Hypertension 2020; 75: 246–256.

23.

Cummings

Patil

Long

, et al. Does the association between hemoglobin A1c and risk of cardiovascular events vary by residential segregation? The REasons for Geographic And Racial Differences in Stroke (REGARDS) study. Diabetes Care 2021; 44: 1151–1158.

24.

Adams

Bendixen

, et al. Classification of subtype of acute ischemic stroke. Stroke 1993; 23: 35–41.

25.

Wolf

D'Agostino

Belanger

, et al. Probability of stroke: a risk profile from the Framingham study. Stroke 1991; 22: 312–318.

26.

Gorsuch

RL.

Factor analysis: classic edition. Milton Park: Routledge, 2014.

27.

Onland-Moret

van der A

van der Schouw

, et al. Analysis of case-cohort data: a comparison of different methods. J Clin Epidemiol 2007; 60: 350–355.

28.

Olson

Cushman

Judd

, et al. Associations of coagulation factors IX and XI levels with incident coronary heart disease and ischemic stroke: the REGARDS study. J Thromb Haemost 2017; 15: 1086–1094.

29.

Bhave

Ament

Patki

, et al. Plasma metabolites link dietary patterns to stroke risk. Ann Neurol 2023; 93: 500–510.

30.

Ottosson

Ericson

Almgren

, et al. Dimethylguanidino valerate: a lifestyle‐related metabolite associated with future coronary artery disease and cardiovascular mortality. J Am Heart Assoc 2019; 8: e012846.

31.

Wali

Koay

Chami

, et al. Nutritional and metabolic regulation of the metabolite dimethylguanidino valeric acid: an early marker of cardiometabolic disease. Am J Physiol Endocrinol Metab 2020; 319: E509–E518.

32.

Popkov

Zharikova

Demchenko

, et al. Gut microbiota as a source of uremic toxins. Int J Mol Sci 2022; 23: 483.

33.

Ramachandran

Fontanille

Pandey

, et al. Gluconic acid: properties, applications and microbial production. Food Technol Biotechnol 2006; 44: 185–195.

34.

Dodd

Spitzer

Van Treuren

, et al. A gut bacterial pathway metabolizes aromatic amino acids into nine circulating metabolites. Nature 2017; 551: 648–652.

35.

Han

Thuy-Boun

Pfeiffer

, et al. Identification of an N-acetylneuraminic acid-presenting bacteria isolated from a human microbiome. Sci Rep 2021; 11: 4763.

36.

Kennedy

Cryan

Dinan

, et al. Kynurenine pathway metabolism and the microbiota-gut-brain axis. Neuropharmacology 2017; 112: 399–412.

37.

Colpo

Venna

McCullough

, et al. Systematic review on the involvement of the kynurenine pathway in stroke: pre-clinical and clinical evidence. Front Neurol 2019; 10: 778.

38.

Kim

J-Y

Suh

J-W

G-E.

Evaluation of S-adenosyl-L-methionine production by Bifidobacterium bifidum BGN4. Food Sci Biotechnol 2008; 17: 184–187.

39.

Zhu

Romano

, et al. Gut microbes impact stroke severity via the trimethylamine N-oxide pathway. Cell Host & Microbe 2021; 29: 1199–1208.e5.

40.

Vaccaro

Naser

SA.

The role of methyl donors of the methionine cycle in gastrointestinal infection and inflammation. Healthcare 2022; 10: 61.

41.

Benjdia

Berteau

Sulfatases and radical SAM enzymes: emerging themes in glycosaminoglycan metabolism and the human microbiota. Biochem Soc Trans 2016; 44: 109–115.

42.

Tang

WHW

Wang

Levison

, et al. Intestinal microbial metabolism of phosphatidylcholine and cardiovascular risk. N Engl J Med 2013; 368: 1575–1584.

43.

Durgan

Lee

McCullough

, et al. Examining the role of the microbiota-gut-brain axis in stroke. Stroke 2019; 50: 2270–2277.

44.

Spychala

Venna

Jandzinski

, et al. Age‐related changes in the gut microbiota influence systemic inflammation and stroke outcome. Ann Neurol 2018; 84: 23–36.

45.

Peh

O'Donnell

Broughton

BRS

, et al. Gut microbiota and their metabolites in stroke: a double-edged sword. Stroke 2022; 53: 1788–1801.

46.

Low-molecular weight bacterial metabolites in host-microbial interaction. Infect Non Infect Dis 2016; 2: 1–11.

47.

Cuartero

Ballesteros

De La Parra

, et al. L-kynurenine/aryl hydrocarbon receptor pathway mediates brain damage after experimental stroke. Circulation 2014; 130: 2040–2051.

48.

Darlington

Mackay

Forrest

, et al. Altered kynurenine metabolism correlates with infarct volume in stroke. Eur J Neurosci 2007; 26: 2211–2221.

49.

Huang

Xia

Influence of the gut microbiome on inflammatory and immune response after stroke. Neurological Sciences 2021 2021; 1: 1–15.

50.

Wang

Klipfell

Bennett

, et al. Gut flora metabolism of phosphatidylcholine promotes cardiovascular disease. Nature 2011; 472: 57–63.

51.

Koeth

Wang

Levison

, et al. Intestinal microbiota metabolism of l-carnitine, a nutrient in red meat, promotes atherosclerosis. Nat Med 2013; 19: 576–585.

52.

Zhu

Gregory

Org

, et al. Gut microbial metabolite TMAO enhances platelet hyperreactivity and thrombosis risk. Cell 2016; 165: 111–124.

53.

Rotondo

Ho-Palma

Remesar

, et al. Glycerol is synthesized and secreted by adipocytes to dispose of excess glucose, via glycerogenesis and increased acyl-glycerol turnover. Sci Rep 2017; 7: 8983.

54.

Zhang

Davies

SS.

Microbial metabolism of dietary components to bioactive metabolites: opportunities for new therapeutic interventions. Genome Med 2016; 8: 46.

55.

Amabebe

Robert

Agbalalah

, et al. Microbial dysbiosis-induced obesity: role of gut microbiota in homoeostasis of energy metabolism. Br J Nutr 2020; 123: 1127–1137.

56.

Miller

AW.

The role of the intestinal microbiome in oxalate homeostasis. The Role of Bacteria in Urology. Cham: Springer International Publishing, pp. 179–186.

57.

Wikoff

Anfora

Liu

, et al. Metabolomics analysis reveals large effects of gut microflora on mammalian blood metabolites. Proc Natl Acad Sci U S A 2009; 106: 3698–3703.

58.

Han

Van Treuren

Fischer

, et al. A metabolomics pipeline for the mechanistic interrogation of the gut microbiome. Nature 2021; 595: 415–420.

59.

Chelluboina

Kieft

Breister

, et al. Gut virome dysbiosis following focal cerebral ischemia in mice. J Cereb Blood Flow Metab 2022; 42: 1597–1602.

60.

Sampson

Debelius

Thron

, et al. Gut microbiota regulate motor deficits and neuroinflammation in a model of Parkinson’s disease. Cell 2016; 167: 1469–1480.

61.

Liang

, et al. Gut bacterial profiles in Parkinson’s disease: a systematic review. CNS Neurosci Ther 2023; 29: 140–157.

62.

Tonomura

Ihara

Friedland

RP.

Microbiota in cerebrovascular disease: a key player and future therapeutic target. J Cereb Blood Flow Metab 2020; 40: 1368–1380.

63.

Zhang

Dong

, et al. Hyperbaric oxygen improves depression-like behaviors in chronic stress model mice by remodeling gut microbiota and regulating host metabolism. CNS Neurosci Ther 2023; 29: 239–255.

64.

Kingsbury

Shear

Heyck

, et al. Inflammation-relevant microbiome signature of the stroke brain, gut, spleen, and thymus and the impact of exercise. J Cereb Blood Flow Metab 2021; 41: 3200–3212.

65.

Vallianatou

Lin

Bèchet

, et al. Differential regulation of oxidative stress, microbiota-derived, and energy metabolites in the mouse brain during sleep. J Cereb Blood Flow Metab 2021; 41: 3324–3338.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.68 MB