Abstract Background Lifestyle factors toward diet and physical activity (PA) may directly influence the pathophysiology of dyslipidemia. However, the associations of the specific macronutrient‐to‐PA ratio with dyslipidemia, and the underlying mechanisms regarding gut microbiota and metabolites, remain largely unexplored. Methods Dietary and PA information from 273 participants with or at risk of metabolic syndrome was collected via a food frequency questionnaire and the International Physical Activity Questionnaire. Gut microbial genera and fecal metabolites were profiled through 16S rRNA sequencing and untargeted LC–MS metabolomics, respectively. Machine‐learning algorithms were applied to identify gut microbiome features of macronutrient‐to‐PA ratios and to construct microbiome risk score. Results Higher macronutrient‐to‐PA ratios, especially for high saturated fatty acid intake, were associated with increased risks of dyslipidemia, with adjusted odds ratio (95% CIs) of 2.87 (1.41–5.99) for hypercholesteremia, 2.21 (1.11–4.48) for hypertriglyceridemia, and 2.52 (1.26–5.16) for high low‐density lipoprotein cholesterol. Microbiome risk scores were significantly associated with elevated levels of total cholesterol, triglycerides, and low‐density lipoprotein cholesterol. Additionally, for each macronutrient‐to‐PA ratio, a core group of gut microbial genera were identified (eg, Phocaeicola, Lachnoclostridium, Limosilactobacillus, and Tyzzerella), exhibiting positive associations with lipid disorders and superior discrimination capacities for hypercholesterolemia, hypertriglyceridemia, and high low‐density lipoprotein cholesterol. Furthermore, we identified 9 metabolites (eg, acetyl phosphate, glycerol, and pyruvic acid), predominantly enriched in dyslipidemia‐related pathways and associated with both core gut microbial taxa and macronutrient‐to‐PA ratios. Conclusions This study identified varied associations between macronutrient‐to‐PA ratios and dyslipidemia and depicted the potential modulatory roles of gut microbiota and fecal metabolites. Keywords: dietary macronutrient, dyslipidemia, gut microbiota, metabolome, physical activity Subject Categories: Lifestyle, Exercise __________________________________________________________________ NONSTANDARD ABBREVIATIONS AND ACRONYMS MRS microbiome risk score PA physical activity TC total cholesterol Clinical Perspective. What Is New? * Our results reveal that high macronutrient‐to‐physical activity (PA) ratios are associated with elevated risks of hypercholesterolemia, hypertriglyceridemia, and high low‐density lipoprotein cholesterol, particularly among participants with higher fat or low‐density lipoprotein cholesterol ‐to‐PA ratios. * Notably, for each macronutrient‐to‐PA ratio, we identified a core set of gut bacterial genera, which also exhibited strong correlations with dyslipidemia. Furthermore, we identified 9 metabolites that were enriched in dyslipidemia‐related pathways, which showed robust associations with the core gut microbial taxa. * Our further integrated multi‐omics analyses revealed that macronutrient‐to‐PA ratios‐related alterations in gut microbiota and metabolites might potentially contribute to the pathogenesis of dyslipidemia. What Are the Clinical Implications? * This study provides novel insight into the impact of the macronutrient‐to‐PA ratios on dyslipidemia risk, underscoring the necessity of maintaining both a healthy, high‐quality macronutrient intake and adequate PA levels in lifestyle interventions to prevent lipid disorders. * By elucidating the potential mechanisms underlying macronutrient‐PA‐dyslipidemia covariations, our results offer opportunities for future precision nutrition research, through modulating the gut microbiota and related microbial metabolites, to mitigate the burden of dyslipidemia. Dyslipidemia, characterized by lipid metabolism disorders, is a well‐established risk factor for atherosclerosis cardiovascular disease, which remains the leading cause of disability and mortality worldwide.[50] ^1 Lifestyle modifications, such as adopting a balanced diet and engaging in regular physical activity (PA), are primary strategies for preventing and managing dyslipidemia.[51] ^2 , [52]^3 However, instead of focusing solely on the quantity of the diet, emphasizing the quality of macronutrients may provide more specific evidence for optimizing circulating lipid homeostasis.[53] ^4 , [54]^5 For example, reducing fat intake could decrease low‐density lipoprotein cholesterol (LDL‐C) levels without altering total energy intake.[55] ^6 Moreover, the bidirectional interplay between PA, nutrient intake, and metabolism of macronutrient remains complex and controversial.[56] ^7 , [57]^8 While some studies investigated the combined effects of dietary patterns and PA on dyslipidemia,[58] ^9 , [59]^10 most have primarily focused on joint associations or interactions, without providing a comprehensive quantification of their interplay. Meanwhile, the notion of “eat less, exercise more” is often advocated as a general principle for improving metabolic health; however, it remains unclear which specific macronutrient reduction would be most beneficial for improving circulating lipid profiles. Ratio‐based indicators have been proposed as credible markers of metabolic disorder risk, such as dietary sodium‐to‐potassium ratio, the triglyceride to high‐density lipoprotein cholesterol (HDL‐C) ratio, and the triglyceride glucose index.[60] ^11 , [61]^12 , [62]^13 Therefore, we introduced the macronutrient‐to‐PA ratio as a simple and direct instrument to capture the combined influence of dietary intake and PA on lipid metabolism and dyslipidemia risk. Additionally, given the importance of balancing energy intake and expenditure, it would be valuable to investigate whether different macronutrient‐to‐PA ratios exert differential effects on circulating lipid levels. Emerging evidence suggests that PA and diet directly shape the composition and function of gut microbiota, which in turn affects host lipid metabolism and contributes to dyslipidemia.[63] ^14 , [64]^15 However, the gut microbiota features reflecting the strength of macronutrient relative to PA was not known, nor was their potential functional capacities on dyslipidemia. Moreover, most previous studies have typically considered the effects of specific bacterial taxa separately, overlooking the overall impact of the gut microbiota.[65] ^16 , [66]^17 In fact, the human gut microbial taxa function as an ecological system, and their combined effects may play fundamental roles in modulating physiological performance. Machine‐learning approaches recently have facilitated the identification of microbiota risk scores of given exposures or outcomes[67] ^18 and thus may accelerate the characterization of overall microbiota power underling the epidemiological relations. Moreover, recent animal studies have demonstrated that microbiota‐derived metabolites, such as short‐chain fatty acids, bile salt hydrolase, and branched‐chain amino acids, may mediate the effects of gut microbiota on lipid homeostasis, by serving as precursors or ligands that modulate host physiological processes.[68] ^19 , [69]^20 Nevertheless, it remains unclear how specific macronutrient‐to‐PA ratio influences the gut microbiome and metabolites, and further contributes to dyslipidemia. The present study aimed to evaluate the associations of macronutrient‐to‐PA ratios with dyslipidemia, and to investigate the potential mechanisms underlying the associations of the macronutrient‐to‐PA ratio with dyslipidemia via gut microbiota and metabolites, by applying machine‐learning selections. METHODS The data that support the findings of this study are available from the corresponding author upon reasonable request. Study Design and Participants This study was conducted using baseline data from 2 nutrition intervention studies aimed at reducing plasma lipids and controlling blood pressure. Briefly, 136 participants with dyslipidemia (blood lipid levels: 3.4≤LDL‐C<4.9 mmol/L, 5.18≤total cholesterol [TC]<7.2 mmol/L, and 1.7≤triglycerides <5.6 mmol/L) were recruited between April and September 2022 to evaluate the effects of Lactobacillus plantarum on lipid profiles, and 172 participants with prehypertension (systolic blood pressure: 120–139 mm Hg; or diastolic blood pressure: 80–89 mm Hg) were included from March to June 2023 to examine the impact of early nutritional interventions on cardiovascular health. Therefore, we combined the baseline data of the 2 studies as a cross‐sectional sample of subjects with or at risk of metabolic syndrome to identify gut microbiota features in relation to cardiometabolic risk factors and their dietary habits. In the 2 original studies, the same standardized protocols were used to collect REDCap‐based online questionnaires ([70]https://www.wcrcnet.cn/redcap/), anthropometric measurements, and biological samples. The study protocols were approved by the Ethical Committee of Xi’an Jiaotong University Second Affiliated Hospital or the Ethical Committee of Xi’an Daxing Hospital. Both trials were registered at [71]ClinicalTrials.gov (ChiCTR2300069460 and ChiCTR2300069460). All participants provided written informed consent. In the present study, we excluded individuals with repeated participation in both trials (n=12), those with outlier values of PA (n=22), and those with missing values of the food frequency questionnaire (n=1). Finally, a total of 273 participants were included in the current study. Assessment of the Macronutrient‐to‐PA Ratio Dietary information was collected via a self‐reported, 87‐item food frequency questionnaire with satisfying reliability.[72] ^21 Participants were required to report the frequency (daily, weekly, or monthly) and precise quantity of each food item consumed over the past year. Daily intake of total energy and nutrients was calculated by standardizing the frequency of each food item into daily consumption values and applying the corresponding nutrient composition from the Chinese Food Composition Tables (2018).[73] ^22 PA level was assessed based on the self‐reported International Physical Activity Questionnaire, collecting information on intensity, frequency, and duration of PA, and then converted into metabolic equivalent tasks in hours per day. The macronutrient‐to‐PA ratio was calculated by dividing the daily intake of each macronutrient by the PA level. Biochemical Measurements and Definition of Dyslipidemia Blood samples were collected after overnight fasting by trained nurses, centrifuged at 2400g for 15 minutes, allocated as plasma, buffy coat, and erythrocyte samples, and then immediately stored at −80 °C until laboratory assays. Plasma levels of TC, triglycerides, HDL‐C, and LDL‐C were measured by an automatic biochemical analyzer (Beckman Coulter AU680). Participants who met at least 1 of the following criteria were considered as dyslipidemia[74] ^23 : (1) TC ≥5.2 mmol/L (hypercholesteremia); (2) triglycerides ≥1.7 mmol/L (hypertriglyceridemia); (3) HDL‐C <1.0 mmol/L (low HDL‐C); and (4) LDL‐C ≥3.4 mmol/L (high LDL‐C). Fecal Sample Collection and 16S rRNA Profiling Fecal samples were self‐collected by participants in a sterile tube (Real‐Bio Technology, Shanghai) following a standard protocol, immediately frozen on ice after defecation, and then transferred to a −80 °C freezer for further analysis. Bacterial DNA was extracted using the PF Mag‐Bind Stool DNA Kit (Omega Bio‐tek, Georgia, USA). The DNA extract was checked on 1% agarose gel, and DNA concentration and purity were determined with a NanoDrop 2000 UV–vis spectrophotometer (Thermo Scientific, Wilmington, USA). The bacterial 16S rRNA genes were amplified by the universal bacterial primers 27F (5′‐AGRGTTYGATYMTGGCTCAG‐3′) and 1492R (5′‐RGYTACCTTGTTACGACTT‐3′). Purified products were pooled in equimolar, and the DNA library was constructed using the SMRTbell prep kit 3.0 (Pacific Biosciences, CA, USA) according to PacBio's instructions. Purified SMRTbell libraries were sequenced on the Pacbio Sequel IIe System (Pacific Biosciences) at the Majorbio Bio‐Pharm Technology Co. Ltd. (Shanghai, China). A total of 222 bacterial genera were identified, and only those genera (n=82) represented in >10% of the total samples were included in the subsequent analysis. Metabolomics Profiling The metabolite was extracted from a 50‐mg fecal sample with 400 μL of methanol (4:1, vol/vol) containing 0.02 mg/mL l‐2‐chlorophenylalanine, followed by ultrasonic extraction at low temperature and centrifugation. The supernatant was then transferred to an injection vial for LC–MS analysis, which was performed on a Thermo UHPLC‐Q Exactive HF‐X system with an ACQUITY HSS T3 column (100 mm × 2.1 mm id, 1.8 μm; Waters, USA), at a flow rate of 0.40 mL/min and a column temperature of 40 °C. The mass spectrometric data were collected using a Thermo UHPLC‐Q Exactive HF‐X Mass Spectrometer equipped with an electrospray ionization source operating in positive mode and negative mode. For the metabolite data, missing values were imputed by the minimum non‐zero values. Metabolite of QC samples with relative SD ≤30% were included and log10 logarithmic zed for further analysis. Statistical Analysis Means (SDs) or percentages were used to present the characteristics of the study participants across tertiles of the macronutrient‐to‐PA ratios. Between‐group comparisons were computed using ANOVA or the Kruskal–Wallis test, as appropriate. Adjusted means of lipids were assessed using general linear model, with adjustment for age, sex (men and women), education attainment (primary school, middle school, high school, college or university, postgraduate and above), marital status (married, divorced, widowed, never married), household income (<1000, 1000–2999, 3000–4999, 5000–7999, 8000–9999, ≥10 000 CNY/mo), smoking status (never smoker, current smoker), alcohol status (never drinker, current drinker), and total energy intake (except for the energy‐to‐PA ratio). Logistic regression was performed to estimate odds ratio and corresponding 95% CI for the associations between the macronutrient‐to‐PA ratios and dyslipidemia, adjusting for covariates as mentioned above. Further analyses using alternative stricter definitions of dyslipidemia were conducted to assess the robustness of the associations between each macronutrient‐to‐PA ratio and dyslipidemia.[75] ^24 Microbial richness (α‐diversity) was estimated using Chao 1 estimators, and microbial dissimilarity (β‐diversity) was assessed by permutational multivariate analysis of variance based on Bray‐Curtis distances. After evaluating 3 machine‐learning algorithms (LightGBM, XGBoost, and CatBoost) in parallel, the best‐performing model was subsequently used to predict the macronutrient‐to‐PA ratio based on demographic data (age and sex) and genus‐level microbiota. SHapley Additive exPlanations were applied to elucidate the impact of features on the model. The microbiome risk score was constructed based on the SHapley Additive exPlanations value of the top 20 features contributing to the model predictions, following the conventional method.[76] ^18 Partial Spearman's correlations between lipid levels and genera in each microbiome risk score of the macronutrient‐to‐PA ratio were calculated to identify microbial features significantly associated with lipid levels, which were subsequently used to form core microbial genera groups for each ratio. The random forest classification algorithm was then applied to discriminate dyslipidemia using the core genera groups for each macronutrient‐to‐PA ratio. The machine‐learning algorithm was implemented using 5‐fold cross‐validation with grid search and an 80/20 random split of the data set. Specifically, 80% of the data was used for training and hyperparameter tuning, while the remaining 20% was strictly held out for final model evaluation. The best‐performing continuous model was determined by the correlation coefficient between predicted and target values, and the classification model examined by the area under curve.[77] ^25 , [78]^26 Additionally, linear regression analyses were conducted to calculate the explained variance (R^2) of each identified genus and core microbial group in relation to lipid disorders, for each macronutrient‐to‐PA ratio. Differential metabolites in dyslipidemia were identified using partial least squares discriminant analysis and Wilcoxon rank‐sum tests. Pathway enrichment analysis of differential metabolites was performed using the Kyoto Encyclopedia of Genes and Genomes database. Integrated analysis networks were constructed based on partial Spearman's correlations among microbiota, metabolites, and the macronutrient‐to‐PA ratios, and were visualized using Cytoscape version 3.10.2. All statistical analyses were conducted using R version 4.4.1 or Python version 3.11. Statistical significance was set at P <0.05 with a 2‐sided test. The Benjamini‐Hochberg false discovery rate correction was appropriately applied for multiple testing corrections. RESULTS Characteristics of Study Participants The present study included 273 participants with a mean (SD) age of 40.2 (11.9) years and 49.8% men. Despite significant differences in macronutrient intake, PA level, and the macronutrient‐to‐PA ratio (P <0.001), other characteristics among the 3 groups were comparable ([79]Table, Supplementary Table [80]S1). In addition, the characteristics of participants in the hypercholesterolemia, hypertriglyceridemia, low HDL‐C, and high LDL‐C groups, as well as their corresponding reference groups, are presented in Supplementary Table [81]S2. Table 1. Characteristics of the Study Participants by the Tertiles of the Energy‐to‐Physical Activity Ratio Characteristic Overall Tertiles of the energy‐to‐physical activity ratio P value Tertile 1 Tertile 2 Tertile 3 No. of participants 273 91 91 91 Age (y) 40.2 (11.9) 44.3 (11.7) 40.8 (12.5) 41.0 (11.2) 0.087 Male, n (%) 136 (49.8) 42 (46.2) 47 (51.6) 47 (51.6) 0.693 Married, n (%) 206 (75.5) 71 (78.0) 65 (71.4) 70 (76.9) 0.542 Middle school and above, n (%) 228 (83.5) 69 (75.8) 79 (86.8) 80 (87.9) 0.052 Household income ≥5000 CNY/mo, n (%) 153 (56.0) 47 (51.6) 52 (57.1) 54 (59.3) 0.560 Current smoker, n (%) 55 (20.1) 23 (25.3) 15 (16.5) 17 (18.7) 0.306 Weekly drinker, n (%) 23 (8.4) 8 (8.8) 6 (6.6) 9 (9.9) 0.717 Body mass index (kg/m^2) 25.1 (3.3) 25.3 (3.5) 24.6 (3.1) 25.3 (3.2) 0.231 Waist circumference (cm) 86.8 (9.5) 86.8 (9.2) 85.6 (9.2) 88.0 (10.0) 0.245 Total energy intake (kcal/d) 2349.3 (1421.3) 1593.7 (735.4) 2230.2 (1123.5) 3224.2 (1713.7) <0.001 Physical activity (MET‐h/d) 9.4 (8.0) 16.4 (8.8) 8.1 (5.1) 3.6 (2.5) <0.001 Total energy/physical activity 530.0 (689.9) 111.6 (44.6) 294.3 (77.8) 1184.3 (874.3) <0.001 [82]Open in a new tab Values are mean (SD) or numbers (%) as indicated. Between‐group difference was computed by ANOVA or Kruskal–Wallis test for normal continuous or categorical variables, respectively. MET indicates metabolic equivalent. Association Between the Macronutrient‐to‐PA Ratios and Dyslipidemia After multivariable adjustment, higher macronutrient‐to‐PA ratios were associated with incremental levels of TC, triglycerides, and LDL‐C (Supplementary Figures [83]S1 and [84]S2). Moreover, after adjusting for covariates, participants with higher macronutrient‐to‐PA ratios were associated with elevated risks of dyslipidemia, but no significant association was noted for energy‐to‐PA ratio (Figure [85]1). For example, the adjusted odds ratios (95% CIs) of high LDL‐C in the highest tertile were 2.19 (1.05–4.62) for the carbohydrate‐to‐PA ratio, 2.26 (1.11–4.71) for the protein‐to‐PA ratio, 2.30 (1.12–4.81) for the fat‐to‐PA ratio, and 2.52 (1.26–5.16) for the saturated fatty acid‐to‐PA ratio, as compared with their lowest tertiles, respectively. Similar association patterns were also observed for hypercholesterolemia and hypertriglyceridemia, with the corresponding adjusted odds ratios (95% CIs) comparing extreme tertiles of saturated fatty acid‐to‐PA ratio were 2.87 (1.41–5.99) and 2.21 (1.11–4.48), respectively. No association was found between macronutrient‐to‐PA ratios and the risk of low HDL‐C. Moreover, the associations were not materially altered when stricter definitions of dyslipidemia were applied (Figure [86]1). Figure 1. Associations of the macronutrient‐to‐PA ratios with dyslipidemia. Figure 1 [87]Open in a new tab Results were presented as OR (95% CI), with adjustment for age, sex, education attainment, household income, marital status, smoking status, alcohol status, and total energy (except for energy‐to‐PA ratio). The horizontal lines represent the corresponding 95% CIs. HDL‐C indicates high‐density lipoprotein cholesterol; LDL‐C, low‐density lipoprotein cholesterol; OR, odds ratio; PA, physical activity; SFA, saturated fatty acid; TC, total cholesterols; and TG, triglyceride. *Hypercholesteremia, TC ≥5.2 mmol/L; hypertriglyceridemia, TG ≥1.7 mmol/L; Low HDL‐C, HDL‐C <1.0 mmol/L; High LDL‐C, LDL‐C ≥3.4 mmol/L. ^†Hypercholesteremia, TC ≥6.2 mmol/L; hypertriglyceridemia, TG ≥2.3 mmol/L; High LDL‐C, LDL‐C ≥4.1 mmol/L. Associations of the Gut Microbiota Features of Macronutrient‐to‐PA Ratios With Dyslipidemia We then investigated individual microbiome features related to macronutrient‐to‐PA ratio status and the potential associations with dyslipidemia (Figure [88]2). Gut microbial richness and dissimilarity were similar among the tertiles of macronutrient‐to‐PA ratio groups (Figure [89]S3). Based on XGBoost and CatBoost models (Supplementary Tables [90]S3 and [91]S4), the top 20 features contributing to each macronutrient‐to‐PA ratio prediction were identified (Figures [92]S4 and [93]S5). The MRSs, based on the identified gut microbiota features, were highly correlated with the macronutrient‐to‐PA ratios and inversely correlated with PA. No correlation was observed between the MRS of total energy‐to‐PA ratio with lipid levels. Notably, most of the MRSs were significantly correlated with elevated levels of TC, triglycerides, and LDL‐C, except for the MRSs of protein‐to‐PA ratio and fat‐to‐PA ratio, which showed no correlation with TG (Figure [94]2A). Figure 2. Associations of macronutrient‐to‐physical‐activity‐ratios–related gut microbial and dyslipidemia. Figure 2 [95]Open in a new tab A, Correlations between microbiome risk scores (MRSs) of macronutrient‐to‐PA ratios and lipid profiles were presented as partial Spearman's correlation coefficients, adjusted for age, sex, education attainment, household income, marital status, smoking status, alcohol status, and total energy (except for energy‐to‐PA ratio). The MRS was constructed based on microbial genera from the top 20 features contributing to the prediction model. The color gradient from red to blue indicated the direction and strength of the correlation, from positive to negative. B, Different colors represent identified gut microbiota responsible for macronutrient‐to‐PA ratios, and the numbers on the overlapped parts were the number of genera that contributed to 2 or more prediction models of macronutrient‐to‐PA ratios simultaneously. C, Correlations between lipid levels and microbiota involved in MRS of each macronutrient‐to‐PA ratio. Partial Spearman's correlation coefficients were adjusted for age and sex. P values were corrected using the FDR method. D, AUC and 95% CI values for discriminating dyslipidemia using random forest models. The model predictors included the core genera group for each macronutrient‐to‐PA ratio, which consisted of microbes that met both the top 20 features and have significant associations with lipid levels. E, Explained variance of each genus and the core microbial‐group‐to lipid disorders. The size of the circle represented the explained variance of each genus or core microbial group in relation to lipid disorders for each macronutrient‐to‐PA ratio, with the color gradient from purple to yellow indicating increasing explained variance. The regression model included age, sex, education attainment, household income, marital status, smoking status, alcohol consumption, and each genus or core microbial group contributing to the macronutrient‐to‐PA ratio. Two hundred fifty‐six participants who met the inclusion criteria and had 16S rRNA sequencing data were included in Figure [96]2 analysis. *P for FDR <0.05; **P for FDR <0.01. ^†Due to the limited space in (E), Oscillospiraceae was used to represent unclassified_f__Oscillospiraceae, and Lachnospiraceae was used to represent unclassified_f__Lachnospiraceae. FDR indicates false discovery rate; HDL‐C, high‐density lipoprotein cholesterol; LDL‐C, low‐density lipoprotein cholesterol; PA, physical activity; SFA, saturated fatty acid; TC, total cholesterols; and TG, triglycerides. In addition, the identified gut microbial genera contributing to the MRSs for different macronutrient‐to‐PA ratios differed considerably, despite some overlaps (Figure [97]2B). Among those contributing to >3 macronutrient‐to‐PA ratios, 6 genera were positively associated with both TC and LDL‐C levels (eg, Lactobacillus, Coprococcus, Faecalibacterium, and Phocaeicola). Conversely, 2 genera were inversely associated with levels of TC or triglycerides (Streptococcus and unclassified_f__Oscillospiraceae) (P for false discovery rate <0.05, Figure [98]2C). Of note, Phascolarctobacterium, Lacrimispora, and Allisonella, which were uniquely related to the fat‐to‐PA ratio, were associated with elevated levels of TC, TG, and LDL‐C. Tyzzerella, specifically contributed to the carbohydrate‐to‐PA ratio, also exhibited strong associations with lipid disorders. However, no significant associations were detected between identified microbiota and HDL‐C level. Furthermore, random forest models based on the core group of microbial genera connecting with each macronutrient‐to‐PA ratio also exhibited satisfying discrimination value for dyslipidemia, except for low HDL‐C (Figure [99]2D, Supplementary Table [100]S5). Specifically, the microbial genera group related to carbohydrate‐to‐PA ratio showed outstanding capacity to distinguish hypertriglyceridemia (area under curve=0.830), while the protein‐to‐PA ratio genera group effectively identified hypercholesterolemia (area under curve=0.790). Additionally, for each macronutrient‐to‐PA ratio, the core microbial genera group also exhibited greater variance explained capacities in lipid disorders compared with individual genera (Figure [101]2E). Gut Microbiotas and Metabolites Modulate the Associations of the Macronutrient‐to‐PA Ratio With Dyslipidemia Compared with the corresponding reference group with normal lipid levels, significant metabolite dissimilarities were observed in the hypercholesterolemia, hypertriglyceridemia, and high LDL‐C groups, identifying 315, 189, and 247 differential metabolites, respectively (P for false discovery rate <0.05, Figure [102]3A, Supplementary Figure [103]S6). Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis revealed that the differential metabolites were primarily related to the endocrine system, carbohydrate metabolism, and amino acid metabolism (Figure [104]3B, Supplementary Figure [105]S7). We then identified 10 metabolites involved in pathways highly correlated with dyslipidemia, 9 of which were significantly associated with core genera responsible for macronutrient‐to‐PA ratios (Figure [106]3C). For example, acetyl phosphate, which is elevated in hypercholesterolemia and in participants with high LDL‐C, was positively associated with 13 genera contributing to the macronutrient‐to‐PA ratios. In contrast, pyruvic acid, reduced in dyslipidemia participants, was negatively associated with Tyzzerella and the macronutrient‐to‐PA ratios. However, the associations between metabolites and macronutrient‐to‐PA ratios were attenuated or even abolished after further false discovery rate correction. Figure 3. Dyslipidemia‐related differential metabolites associated with gut microbes. Figure 3 [107]Open in a new tab A, Differentially abundant metabolites in dyslipidemia. The x‐axis of the volcano plots represented the fold changes of metabolites compared with the corresponding normolipidemic group, while the y‐axis indicated the significance of the difference (q‐values were adjusted by FDR correction). Dots and annotations represented differentially abundant metabolites between normolipidemic participants and those with dyslipidemia (blue, decreased; red, increased; gray, not significant). The disorder groups for each lipid index were defined as follows: hypercholesterolemia (TC ≥5.2 mmol/L), hypertriglyceridemia (TG ≥1.7 mmol/L), low HDL‐C (HDL‐C <1.0 mmol/L), and high LDL‐C (LDL‐C ≥3.4 mmol/L). The reference groups for each lipid index consisted of participants who did not meet the criteria for the corresponding disorder. Two hundred fifty‐nine participants who met the inclusion criteria and had metabolites data were included in this analysis. B, KEGG pathway enrichment analysis of differential metabolites. The color gradient of the circles, ranging from red to blue, indicated the enrichment factor of each pathway, while the size of the circle reflects the number of metabolites contributing to that pathway. C, Correlation between selected metabolites in dyslipidemia‐associated pathways and genera significantly associated with lipid levels. The dyslipidemia‐correlated pathways included regulation of lipolysis in adipocytes, insulin secretion, insulin resistance, pyruvate metabolism, taurine and hypotaurine metabolism, and retrograde endocannabinoid signaling. The color of the circles indicates the direction of correlation between core genera and lipid disorders. The gradient of the lines, ranging from red to blue, illustrates the direction and strength of the significant correlations, with the correlations between microbe genera and metabolites adjusted for FDR correction (P for FDR <0.05), while the correlations between macronutrient‐to‐PA ratios and metabolites were not adjusted due to the smaller number of comparisons (P <0.05). Two hundred forty‐one participants who met the inclusion criteria and had both 16S rRNA sequencing and metabolite data were included in network analysis. FDR indicates false discovery rate; HDL‐C, high‐density lipoprotein cholesterol; KEGG, Kyoto Encyclopedia of Genes and Genomes; LDL‐C, low‐density lipoprotein cholesterol; PA, physical activity; TC, total cholesterols; and TG, triglycerides. DISCUSSION In the current study, we observed that increased macronutrient‐to‐PA ratios were associated with elevated risks of dyslipidemia, particularly for participants with high‐fat intake coupled with insufficient PA. In addition, for each higher macronutrient–PA ratio, we identified a core group of predominant gut microbial genera. These genera not only correlated with lipid profiles but also exhibited substantial discrimination for dyslipidemia. Moreover, we identified 9 metabolites enriched in pathways highly related to dyslipidemia and also showed significant associations with the core gut microbial taxa. Further integrated analysis suggested that macronutrient‐to‐PA ratios–driven alterations in gut microbiota and metabolites might be relevant to the development of dyslipidemia. Although PA and macronutrient intake were independently associated with the risk of dyslipidemia, the intricate interaction between macronutrient quality and energy expenditure remains insufficiently understood.[108] ^7 This complexity underscores the necessity of quantifying the interaction between macronutrient intake and PA, as well as evaluating their cumulative effects on dyslipidemia. In the present study, we developed the macronutrient‐to‐PA ratio, providing valuable insights into how these lifestyle determinants interact to influence lipid disorders. Consistent with previous studies,[109] ^27 , [110]^28 our data found no significant associations between the energy‐to‐PA ratio and dyslipidemia, suggesting that dietary quality, rather than total energy intake, may play a more pivotal role in the development of dyslipidemia. Notably, participants in the highest tertile of macronutrient‐to‐PA ratios, especially for those with increased intake of fat and saturated fatty acid, demonstrated elevated risks of hypercholesterolemia, hypertriglyceridemia, and high LDL‐C. This indicated that an imbalance characterized by high‐fat dietary intake coupled with insufficient PA might directly accelerate the risk of dyslipidemia. In addition, HDL‐C level was not significantly related to macronutrient‐to‐PA ratios. This may be partially explained by the relatively stable nature of HDL‐C in response to lifestyle modifications than other lipid parameters, as well as the limited sample size of our study.[111] ^29 Accordingly, our study highlights the necessity of maintaining both healthy, high‐quality macronutrient intake and adequate PA in future lifestyle interventions for maintaining healthy lipid profiles. Accumulating evidence has shown the critical role of gut microbiota in the progression of dyslipidemia.[112] ^20 We observed combined gut microbiome risks and identified predominant microbial genera that were highly responsive to varying macronutrient‐to‐PA ratios and strongly associated with lipid disorders. Consistent with previous animal studies, our study found that higher macronutrient‐to‐PA ratios were associated with probiotics such as Oscillospiraceae, which are known to mitigate the effects of a high‐fat diet by regulating adipose tissue accumulation and lipid metabolism.[113] ^30 In contrast, our study also observed that potentially pathogenic bacteria were responsible for higher macronutrient‐to‐PA ratios status (eg, Lactobacillus, Lachnoclostridium, and Parasutterella),[114] ^31 , [115]^32 , [116]^33 which were also associated with the increased risks of lipid disorders. Interestingly, Phocaeicola, primarily involved in polysaccharide metabolism, displayed inconsistent effects on metabolic diseases in previous studies[117] ^34 ; however, in our study, Phocaeicola contributed to 4 macronutrient‐to‐PA ratios and was significantly associated with TC and LDL‐C disorders. This disparity may arise from variations in microbial strains, which may cause opposing effects on host metabolism.[118] ^35 Additionally, our results noticed that the overall microbiome pattern exhibited superior discriminative power and explained capacity for dyslipidemia, suggesting that the combined gut microbiome dysbiosis, rather than specific microbial taxa, exerts a more significant impact on host metabolic health. Taken together, the available evidence highlights the potential modulation role of gut microbiome in the effects of macronutrient‐to‐PA ratios on dyslipidemia. Further large‐scale validation studies are warranted to verify the precise effects of the identified microbes. By integrating gut microbiome and metabolomics data, our results shed light on the potential mechanisms underlying the association between macronutrient‐to‐PA ratios and dyslipidemia. Lachnospiraceae, belonging to the phylum Firmicutes, is the core symbiotic microorganism in the human gastrointestinal tract.[119] ^36 Consistent with our findings, Lachnospiraceae has been shown to be highly linked to high‐energy diets and may contribute to body adiposity, insulin sensitivity, and the pathogenesis of metabolic diseases.[120] ^36 Notably, our integrative analysis further observed that Lachnospiraceae was associated with multiple differential metabolites, including acetyl phosphate, Dg(20:2(11z,14z)/16:1(9z)/0:0), glycerol, o‐acetyl carnitine, and pyruvic acid, many of which are enriched in dyslipidemia‐related Kyoto Encyclopedia of Genes and Genomes pathways, such as regulation of lipolysis in adipocytes, insulin secretion, and insulin resistance.[121] ^37 , [122]^38 , [123]^39 , [124]^40 Our results detected a higher abundance of Lachnospiraceae associated with increased glycerol level, which, as a polyalcohol generated during the metabolism of TAG and other glycolipids, is directly involved in lipid metabolism in adipose tissues and the secretion of LDL‐C in vivo.[125] ^41 Moreover, our findings were also in line with previous studies indicating that a high‐energy diet often impaired pyruvate dehydrogenase activity via pyruvate dehydrogenase kinase upregulation.[126] ^42 We noticed that pyruvic acid was inversely associated with macronutrient‐to‐PA ratios and closely linked to Lachnospiraceae, suggesting that the modulation of Lachnospiraceae and pyruvic acid may be crucial in the progression from high macronutrient‐to‐PA ratios to dyslipidemia. Moreover, our data demonstrated that multiple gut microbial genera contributed to high macronutrient‐to‐PA ratios, including Lachnoclostridium, Roseburia, Streptococcus, and Parasutterella, were also positively associated with dyslipidemia‐related metabolites, yet further studies are needed to clarify the relationship. Collectively, our findings suggest that alterations in gut microbiota and metabolites driven by macronutrient‐to‐PA ratios may contribute to dyslipidemia, providing potential targets for therapeutic intervention. This study has several limitations. Firstly, because of the cross‐sectional design and limited sample size, the potential causality association could not be assessed and reverse causation is possible, because the alterations in gut microbiota and metabolites may result from dyslipidemia. Further longitudinal studies are needed to better clarify the directionality and underlying mechanisms. Secondly, the ascertainment of dietary intake and PA was based on a self‐report questionnaire, which could introduce recall bias. Thirdly, our study participants were selected from subjects with or at risks of metabolic syndrome, limiting the generalizability to other populations. Fourthly, our study only utilized gut microbial genera data, because species‐level gut microbes did not show significant associations with dyslipidemia risk in the present study. Furthermore, the associations between phyla and family‐level gut microbes and dyslipidemia risk still warrant investigation, given their crucial role in lipid metabolism. Additionally, nonbacterial intestinal microbes, such as fungi, viruses, and archaea, significantly impact host metabolism, and future studies should explore their potential role in dyslipidemia and metabolic health. Finally, since the macronutrient‐to‐PA ratio is a new metric introduced in this study to capture the combined influence of dietary patterns and PA, further large‐scale studies are needed to validate this ratio more robustly. In conclusion, our results demonstrate higher macronutrient‐to‐PA ratios were associated with elevated risks of dyslipidemia, suggesting that maintaining both high quality diet and balanced PA are crucial to prevent long‐term lipid disorders. Importantly, we reveal the potential mechanism underlying the associations of macronutrient‐to‐PA ratios and dyslipidemia through the modulation of gut microbiota and related metabolites. Understanding the modifiable nature of the gut microbiome and metabolites offers insights for future preventive strategies to mitigate the burden of dyslipidemia induced by unhealthy lifestyles. However, large‐scale longitudinal studies are still warranted to validate and expand our findings. Sources of Funding This study was supported by the National Natural Science Foundation of China (82173504, 82011530197). Disclosures There are no conflicts of interest. Supporting information Data S1 Tables S1–S5 Figures S1–S7 [127]JAH3-14-e040042-s001.pdf^ (1.4MB, pdf) Acknowledgments