OBJECTIVE

This randomized controlled-feeding trial aimed to determine the impact of fried meat intake on the gut microbiota and fecal cometabolites and whether such impacts influenced host glucose homoeostasis, intestinal endotoxin levels, and systemic inflammation.

RESEARCH DESIGN AND METHODS

A total of 117 overweight adults were randomized into two groups. Fifty-nine participants were provided fried meat four times per week, and 58 participants were restricted from fried meat intake, while holding food group and nutrient compositions constant, for 4 weeks. The gut microbiota was analyzed by 16S rRNA sequencing. Glucose and insulin concentrations at 0, 30, 60, and 120 min of an oral glucose tolerance test, fecal microbiota-host cometabolite levels, and intestinal endotoxin and inflammation serum biomarker levels were measured. The area under the curve (AUC) for insulin, insulinogenic index (IGI), and muscle insulin resistance index (MIRI) were calculated.

RESULTS

The participants who consumed fried meat had lower IGI values than the control subjects, but they had higher MIRI and AUC values of insulin and lipopolysaccharide (LPS), TNF-α, IL-10, and IL-1β levels (P < 0.05). Fried meat intake lowered microbial community richness and decreased Lachnospiraceae and Flavonifractor abundances while increasing Dialister, Dorea, and Veillonella abundances (P FDR <0.05), provoking a significant shift in the fecal cometabolite profile, with lower 3-indolepropionic acid, valeric acid, and butyric acid concentrations and higher carnitine and methylglutaric acid concentrations (P FDR <0.05). Changes in these cometabolite levels were significantly associated with changes in IGI and MIRI values and LPS, FGF21, TNF-α, IL-1β, and IL-10 levels (P < 0.05).

CONCLUSIONS

Fried meat intake impaired glucose homoeostasis and increased intestinal endotoxin and systemic inflammation levels by influencing the gut microbiota and microbial-host cometabolites.

A suboptimal diet is an important preventable risk factor for type 2 diabetes (1,2), and in recent years, interest in the association between fried food intake and the risk of type 2 diabetes has been increasing. A few prospective studies indicated that a higher frequency of fried food consumption was associated with a higher incidence of type 2 diabetes (3,4), which suggested that excessive energy intake and the production of hazardous material during frying probably mediated this association; in contrast, another prospective study from the Mediterranean region, in which foods were frequently consumed as part of a Mediterranean diet, did not support this association (5). Similarly, a dietary intervention study of 17 insulin-resistant women found that consumption of food fried in unsaturated fatty acids had beneficial effects on insulin sensitivity (6). These studies suggested that confounders, such as other nutrients and hazardous material production during frying, influenced the association between fried food and type 2 diabetes. It is still largely unknown whether and how fried food intake influences the development of type 2 diabetes if these confounders are rigorously controlled.

Furthermore, meat, as the major protein source, is commonly used for frying in daily life. Although protein in meat is largely digested in the upper intestine, ∼10% may reach the large bowel and become available for fermentation by the gut microbiota (7). It has been documented that the rate of protein digestibility in meat is decreased during frying (8,9), and the lower digestibility probably results in more undigested protein available for the gut microbiota. Recent studies also found that higher fried meat intake correlated with lower diversity of the gut microbial community in humans (10) and appears to influence the composition and activity of the gut microbiota in animals (7,11). In this study, our research team examined the effects of fried meat and selected cometabolites in a randomized controlled-feeding trial conducted in healthy overweight young adults. We also included information from our research using gastric gavage to examine the effects of selected cometabolites in mice (Supplementary Material).

Participant Recruitment

Participants in this randomized controlled-feeding trial were recruited among healthy overweight adults in Harbin from January to February 2020, and the trial was registered at chictr.org.cn, no. ChiCTR1900028562. The inclusion criteria were age 18–35 years old, BMI > 24 kg/m2, and consumption of fried food more than one time per week. Exclusion criteria were as follows: individuals who 1) had taken antibiotics, probiotics, and prebiotics within 3 months; 2) suffered from diabetes, dyslipidemia, or gastrointestinal disease; 3) had received surgical treatment within 3 months; 4) undertook frequent strenuous exercise or protein supplementation; or 5) had smoking or drinking habits. The volunteers who met the inclusion criteria were subjected to a physical condition survey and an electronic dietary questionnaire survey for collection of baseline information before the intervention. The study design was approved by the ethics committee of Harbin Medical University. The nature and potential risks of the study were explained to the participants before written informed consent was obtained.

Intervention Strategies

The participants were randomly assigned to the experimental group and control group through generation of random allocation sequences to ensure that there was no significant difference in the baseline characteristics of demographics, dietary intake, anthropometry, and biochemical indexes between the two groups. Detailed information on the method used to generate the random allocation sequence and allocation concealment mechanism is presented in Supplementary Material.

All the meals in this trial were finished in the canteen of Harbin Medical University under supervision of the trained personnel. The formula for dietary intervention was mainly based on the Alternate Health Eating Index (AHEI)-2010 (12) (Supplementary Table 1). The amount of food groups and composition of nutrients that were provided in the two groups were constant, with an AHEI score >85 (Supplementary Table 2), but differed in meat cooking methods, which was frying at 150°C for <3 min in the experimental group and boiling, steaming, or dressing with sauce at 100°C in the control group. Fried meat was provided four times per week in the experimental group. Before the intervention, a pre-experiment was performed to calculate the amount of oil and starch per gram of meat required for frying to ensure isocaloric feeding between the two groups (Supplementary Fig. 1). Trained personnel weighed and recorded the leftovers of each participant to calculate the actual dietary intake of each person, and the participants were not allowed to eat any food or beverages other than the uniform meal. The participants were asked to maintain their habitual daily physical activities, which were also monitored by collection of records of daily steps and a questionnaire of physical activities during the intervention.

Sample Size and Power Analysis

Sample size calculation (assuming a 7% difference between the groups in the area under the curve [AUC] for insulin levels during an oral glucose tolerance test, with an SD for the insulin level AUC = 20 mU/L * h) showed that a total sample size of 39 participants per group was necessary to reach statistical significance with a power level of 1 – β = 0.8 and a significance level of α = 0.05 (13). For the glucose level AUC, a sample size of 25 participants per group was required to detect a 1.20 mmol/L * h significant difference, assuming that the SD was 1.5 mmol/L * h (14). We therefore recruited 130 participants to the study (65 in each intervention group), allowing for a 10% dropout rate.

Sample Collection

After the participants entered the trial, they had a 1-week balance period and then began a 4-week dietary intervention. Blood and morning urine samples were collected at 0, 2, and 4 weeks of the intervention, while fecal samples were collected at 0 and 4 weeks (Supplementary Fig. 2A). The participants underwent a 2-h 75-g oral glucose tolerance test after a 12-h overnight fast, and blood samples were collected at 0, 30, 60, and 120 min. The detailed procedure of blood and fecal sample collection can be found in Supplementary Material.

Compliance Indicator Measurement

BMI was calculated as weight in kilograms divided by the square of the height in meters. The concentrations of 2-amino-3,8-dimethylimidazo[4,5-f]quinoxaline (MeIQx) and 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP) in urine samples, as biomarkers of heterocyclic amines, were measured by ultra-performance liquid chromatography–tandem mass spectrometry (UPLC-MS/MS) (ACQUITY UPLC-Xevo TQ-S, Waters Corp., Milford, MA). The nitrogen contents in food and feces were detected by the automatic Kjeldahl apparatus method, and the apparent protein digestibility was calculated. Serum advanced glycation end products (AGEs) were measured with an ELISA kit (Cusabio, College Park, MD). Targeted analysis of serum fatty acids was performed with the TRACE GC/PolarisQ-MS system (Thermo Finnigan, Austin, TX) and DB-WAX capillary column (J&W Scientific, Folsom, CA).

Main Outcome Measurement

Serum glucose was quantified with an Automatic Biochemical Analyzer (Roche Diagnostics, Mannheim, Germany). Serum insulin was measured with an automatic microparticle chemiluminescence immunoassay system (Beckman Coulter, Brea, CA). HbA1c levels were measured with an automatic glycosylated hemoglobin analyzer (HA-8380; ARKRAY, Tokyo, Japan). Serum C-peptide levels were measured with an ELISA kit (Cusabio). Total AUCs for glucose and insulin levels were calculated with the trapezoidal method (15). The insulinogenic index (IGI) was calculated, Δinsulin (0–30 min, µU/mL)/Δglucose (0–30 min, mg/mL) (16), and the muscle insulin resistance index (MIRI) was calculated as follows: slope of the blood glucose concentration decrease (60–120 min)/mean insulin concentration (17).

Secondary Outcome Measurement

Serum levels of lipopolysaccharide (LPS), LPS binding protein (LBP), soluble LPS receptor CD14 (sCD14), adiponectin, and fibroblast growth factor 21 (FGF21) were measured with ELISA kits (Cusabio). All serum inflammatory factors, including TNF-α, IFN-γ, IL-1β, IL-2, IL-4, IL-6, IL-8, and IL-10, were quantified with use of Luminex assay technology (R&D Systems, Minneapolis, MN).

16S rRNA Sequencing

Detailed methods for DNA extraction, PCR, and amplification sequencing are provided in Supplementary Material. FASTQ files were demultiplexed, merge paired, and quality filtered using Quantitative Insights Into Microbial Ecology (QIIME) software (version 1.9.0). Sequences were clustered into operational taxonomic units (OTUs) with 97% similarity and annotated based on the SSU rRNA database of SILVA 132.

Measurement of Cometabolite Levels

A total of 174 fecal microbial-host cometabolites were quantified by a UPLC-MS/MS system (Metabo-Profile, Shanghai, China). Detailed information about standards, procedures, conditions, and methods is provided in Supplementary Material.

Statistical Analysis

Baseline characteristics are presented as the mean (SD) for continuous variables and number (percentage) for categorical variables. A generalized linear mixed model was established to analyze the effect of the intervention on the main outcomes, in which participants were random effects and the intervention models were fixed effects. General linear models and χ2 tests were used to compare baseline characteristics, the differences in the compliance indicators, and the secondary outcomes before and after intervention or between groups. A subgroup analysis was also performed, and its aim and detailed procedure are provided in Supplementary Material.

The Abundance-based Coverage Estimator (ACE) and Chao1 estimator and Shannon index were used to analyze the richness and diversity of the microbial community (18). Principal coordinate analysis based on Bray-Curtis distance was performed to compare the global microbiota composition between groups at the OTU levels. Linear discriminant analysis (LDA) and effect size were used to emphasize statistical significance and biological relevance and find biomarkers with significant differences between groups. The set value of the LDA score was 2 (19). Orthogonal partial least squares discriminant analysis (OPLS-DA) was performed to examine the overall distribution of fecal metabolites between groups, and the cross-validation permutation test was conducted to validate the fitting situation of this model. Metabolites with variable importance in projection >1 were identified to be analyzed in univariate analysis.

In the univariate analysis, the Student t test or Mann-Whitney U test was used to examine the significant differences in the levels of individual genera, microbiota predicted pathways, and fecal metabolites before and after intervention or between groups, and the P values were adjusted using the Benjamini-Hochberg false discovery rate (P FDR). Partial Spearman rank correlation analyses between changes in abundant bacterial genera and fecal metabolites and altered outcome variables were used with adjustment for differences in BMI before and after intervention.

OPLS-DA was performed with SIMCA-P software (14.0), and other statistical analyses were performed in R 3.6.1. A two-sided P value <0.05 was considered statistically significant for the generalized linear mixed models or general linear models, and a P FDR <0.05 was considered statistically significant for Student t test or Mann-Whitney U test in the univariate analysis of individual genera and fecal metabolites.

Feeding Trial Process and Baseline Characteristics

Of the 150 participants assessed for eligibility, 130 were randomized. During the intervention stage, 7 dropped out from the control group, 6 dropped out from the fried meat group, and 117 participants completed the study (Supplementary Fig. 2B). The baseline characteristics regarding demographics, dietary intake, anthropometry, and biochemical indexes were similar between the two groups (all P > 0.05) (Table 1).

Table 1

Baseline characteristics of the participants (n = 117)

CharacteristicsControl group (n = 58)Fried meat group (n = 59)P
Age, years 21.73 (2.97) 21.13 (2.66) 0.110 
Female, n (%) 31 (53.45) 33 (55.93) 0.332 
BMI, kg/m2 26.39 (2.80) 26.06 (2.38) 0.498 
Body fat rate, % 26.11 (6.33) 26.93 (6.05) 0.773 
Waist-to-hip ratio, % 0.87 (0.16) 0.86 (0.15) 0.420 
Laboratory measurements    
 HbA1c    
  % 5.37 (0.20) 5.36 (0.22) 0.848 
  mmol/mol 35.19 (2.20) 35.09 (2.40) 0.803 
 Fasting glucose, mmol/L 4.68 (0.41) 4.64 (0.39) 0.232 
 Fasting insulin, μU/mL 4.42 (2.10) 4.68 (1.89) 0.214 
 IGI, μU/mmol * 103 20.93 (13.09) 22.53 (13.87) 0.526 
 MIRI, mmol/min ⋅ μU * 10−5 1.64 (2.38) 1.56 (2.40) 0.848 
 AUC of glucose, mmol * min/L 679.13 (93.40) 662.92 (90.11) 0.389 
 AUC of insulin, μU * min/mL 3,605.75 (1,574.29) 4,036.86 (1,907.20) 0.225 
Dietary intake    
 Energy, kcal/day 2,669 (586.52) 2,634 (605.48) 0.341 
 Total fat, %TE 26.62 (7.39) 26.28 (6.98) 0.724 
 Carbohydrate, %TE 61.73 (10.39) 61.41 (11.39) 0.981 
 Protein, %TE 11.65 (3.70) 12.31 (3.74) 0.480 
Component of AHEI-2010 score    
 Vegetables 5.32 (2.94) 5.03 (3.04) 0.725 
 Fruit 4.88 (2.80) 5.40 (3.14) 0.524 
 Whole grains 4.84 (2.91) 5.43 (3.05) 0.466 
 Polyunsaturated fatty acids 5.36 (2.96) 5.00 (3.03) 0.659 
 Nuts and legumes 5.36 (3.01) 5.00 (2.98) 0.659 
 Long-chain n-3 fats 5.08 (3.03) 5.23 (2.98) 0.851 
 Red and processed meat 4.8 (2.92) 5.47 (3.04) 0.413 
 Sugar-sweetened drinks 5.32 (2.84) 5.03 (3.13) 0.725 
 Sodium 4.64 (2.93) 5.60 (2.99) 0.234 
 Alcohol 10.00 10.00 
 Total score 55.60 (11.50) 57.20 (10.84) 0.598 
Physical exercise habits, n (%) 39 (67.24) 40 (67.80) 0.896 
Frequency of consumption of fried foods, n (%)   0.814 
 1–3 times per week 31 (53.45) 34 (57.62) 0.653 
 4–6 times per week 25 (43.10) 22 (37.29) 0.525 
 ≥7 times per week 2 (3.45) 3 (5.08) 0.665 
CharacteristicsControl group (n = 58)Fried meat group (n = 59)P
Age, years 21.73 (2.97) 21.13 (2.66) 0.110 
Female, n (%) 31 (53.45) 33 (55.93) 0.332 
BMI, kg/m2 26.39 (2.80) 26.06 (2.38) 0.498 
Body fat rate, % 26.11 (6.33) 26.93 (6.05) 0.773 
Waist-to-hip ratio, % 0.87 (0.16) 0.86 (0.15) 0.420 
Laboratory measurements    
 HbA1c    
  % 5.37 (0.20) 5.36 (0.22) 0.848 
  mmol/mol 35.19 (2.20) 35.09 (2.40) 0.803 
 Fasting glucose, mmol/L 4.68 (0.41) 4.64 (0.39) 0.232 
 Fasting insulin, μU/mL 4.42 (2.10) 4.68 (1.89) 0.214 
 IGI, μU/mmol * 103 20.93 (13.09) 22.53 (13.87) 0.526 
 MIRI, mmol/min ⋅ μU * 10−5 1.64 (2.38) 1.56 (2.40) 0.848 
 AUC of glucose, mmol * min/L 679.13 (93.40) 662.92 (90.11) 0.389 
 AUC of insulin, μU * min/mL 3,605.75 (1,574.29) 4,036.86 (1,907.20) 0.225 
Dietary intake    
 Energy, kcal/day 2,669 (586.52) 2,634 (605.48) 0.341 
 Total fat, %TE 26.62 (7.39) 26.28 (6.98) 0.724 
 Carbohydrate, %TE 61.73 (10.39) 61.41 (11.39) 0.981 
 Protein, %TE 11.65 (3.70) 12.31 (3.74) 0.480 
Component of AHEI-2010 score    
 Vegetables 5.32 (2.94) 5.03 (3.04) 0.725 
 Fruit 4.88 (2.80) 5.40 (3.14) 0.524 
 Whole grains 4.84 (2.91) 5.43 (3.05) 0.466 
 Polyunsaturated fatty acids 5.36 (2.96) 5.00 (3.03) 0.659 
 Nuts and legumes 5.36 (3.01) 5.00 (2.98) 0.659 
 Long-chain n-3 fats 5.08 (3.03) 5.23 (2.98) 0.851 
 Red and processed meat 4.8 (2.92) 5.47 (3.04) 0.413 
 Sugar-sweetened drinks 5.32 (2.84) 5.03 (3.13) 0.725 
 Sodium 4.64 (2.93) 5.60 (2.99) 0.234 
 Alcohol 10.00 10.00 
 Total score 55.60 (11.50) 57.20 (10.84) 0.598 
Physical exercise habits, n (%) 39 (67.24) 40 (67.80) 0.896 
Frequency of consumption of fried foods, n (%)   0.814 
 1–3 times per week 31 (53.45) 34 (57.62) 0.653 
 4–6 times per week 25 (43.10) 22 (37.29) 0.525 
 ≥7 times per week 2 (3.45) 3 (5.08) 0.665 

Continuous variables are presented as mean (SD), and categorical variables are presented as n (%). %TE, percentage of total energy.

Compliance Assessment

During the intervention, the energy intake was reduced, whereas the AHEI score was increased, and the daily physical activity was similar between the two groups (Supplementary Fig. 3A–C). The levels of BMI, saturated fatty acids, n-3 fatty acids, n-6 fatty acids, MeIQx, PhIP, and AGEs and the digestibility of protein in the two groups were similar at baseline (Supplementary Fig. 3D–K). After the intervention, the participants in the control group had lower BMI and AGE levels, whereas they had higher levels of n-3 fatty acids and protein digestibility than the participants in the experimental group (all P < 0.05).

Changes in the Main and Secondary Outcomes

During the intervention, the rates of change in HbA1c and C-peptide levels and the AUC of glucose levels in the two groups did not differ significantly (Fig. 1A, B, and E). The IGI value significantly increased in both groups; however, it was significantly higher in the control group than in the experimental group after the intervention (29.75 vs. 24.75 µU/mmol * 103, P = 0.006) (Fig. 1C). Meanwhile, the MIRI and insulin level AUC values significantly decreased in both groups, and they were significantly lower in the control group than in the experimental group (0.82 vs. 1.67 mmol/min ⋅ µU * 10−5, P = 0.008 for MIRI; 3,310.5 vs. 2796.0 µU/mL ⋅ min, P < 0.001 for insulin levels) (Fig. 1D and F).

Figure 1

The differences for the main outcomes in terms of HbA1c (A), C-peptide (B), IGI (C), MIRI (D), and AUC for glucose (E) and insulin (F). P values refer to the differences for the changing rate of biomarkers being different from 0.

Figure 1

The differences for the main outcomes in terms of HbA1c (A), C-peptide (B), IGI (C), MIRI (D), and AUC for glucose (E) and insulin (F). P values refer to the differences for the changing rate of biomarkers being different from 0.

Close modal

The biomarkers of intestinal endotoxin levels and systemic inflammation were similar between the two groups at baseline (Supplementary Fig. 4). After the intervention, the LPS, LBP/sCD14, TNF-α, IL-1β, and IL-10 levels in the control group were significantly lower than those in the experimental group, whereas FGF21 levels were significantly greater in the control group than in the experimental group (all P < 0.05). The IL-2, IL-4, and IL-8 levels were all significantly decreased, and they did not differ significantly between the two groups.

Subgroup Analysis

A total of 45 participants who habitually consumed fried food infrequently before they participated in the trial were identified. The baseline characteristics were similar between the two groups (Supplementary Table 3). The results in the subsample were consistent with those of the total sample (Supplementary Fig. 5), which further supported that the differences for the main and secondary outcomes were due to the effect of fried meat intake rather than from the restriction of fried food.

Changes in the Gut Microbiota and Fecal Metabolomic Profiles

The microbial community richness indicated by ACE and the Chao1 estimator was significantly increased in both groups, and ACE value in the control group was greater than that in the experimental group (Fig. 2A and B). Additionally, the community diversity estimated by the Shannon index was significantly increased in both groups, and it did not differ significantly between the two groups (Fig. 2C). Moreover, there were significant differences in the overall composition of the gut microbiota at the OTU level between the two groups at the end point (R2 = 0.016, P = 0.041) (Fig. 2E). The microbiota predicted pathways showed that the LPS biosynthesis protein (ko01005), insulin resistance (ko04931), adipocytokine signaling (ko04920), and AMPK signaling (ko04152) pathways were significantly different between the two groups (Supplementary Fig. 6A). At the phylum level, the ratio of Firmicutes and Bacteroidetes was decreased in both groups, and it was significantly lower in the control group than in the experimental group (2.87 ± 2.31 vs. 4.29 ± 3.89, P = 0.020) (Supplementary Fig. 6B). At the genus level, five differentially abundant genera of bacteria were identified (Fig. 2F). Compared with the control group, the experimental group had a greater abundance of Dialister, Dorea, and Veillonella but a lower abundance of Lachnospiraceae and Flavonifractor (all P FDR <0.05) (Supplementary Fig. 7).

Figure 2

Changes in gut microbiota and fecal metabolomics profiles. The microbial community richness was indicated by ACE (A) and the Chao1 estimator (B). The community diversity of gut microbiota was estimated by the Shannon index (C). D and E: The overall composition of the gut microbiota at OTU levels before (D) and after intervention (E). F: The differential gut genera identified by LDA effect size. G and H: The overall composition of the cometabolites between host and gut microbiota before (G) and after (H) intervention. IM: The differences for the fecal valeric acid (I), butyric acid (J), IPA (K), carnitine (L), and MGA (M) between the two groups. P values for the differences of the identified cometabolites were adjusted using the Benjamini-Hochberg FDR. R2X, the percentage of matrix information that the model can explain; Q2, the predictive power of the model calculated by cross validation; P, validation of the OPLS-DA from ANOVA of the cross-validated predictive residuals.

Figure 2

Changes in gut microbiota and fecal metabolomics profiles. The microbial community richness was indicated by ACE (A) and the Chao1 estimator (B). The community diversity of gut microbiota was estimated by the Shannon index (C). D and E: The overall composition of the gut microbiota at OTU levels before (D) and after intervention (E). F: The differential gut genera identified by LDA effect size. G and H: The overall composition of the cometabolites between host and gut microbiota before (G) and after (H) intervention. IM: The differences for the fecal valeric acid (I), butyric acid (J), IPA (K), carnitine (L), and MGA (M) between the two groups. P values for the differences of the identified cometabolites were adjusted using the Benjamini-Hochberg FDR. R2X, the percentage of matrix information that the model can explain; Q2, the predictive power of the model calculated by cross validation; P, validation of the OPLS-DA from ANOVA of the cross-validated predictive residuals.

Close modal

Differences in the overall composition of fecal metabolites between the two groups were also observed at the end point (P = 0.039) (Fig. 2K). Further stratification by metabolite categories showed that the metabolites of amino acids and short-chain fatty acids (SCFAs) were significantly different between the two groups (both P < 0.001) (Supplementary Fig. 8A and F). Individual metabolic analysis identified carnitine, 3-indolepropionic acid (IPA), valeric acid, butyric acid, and methylglutaric acid (MGA). Compared with those in the control group, carnitine and MGA were enriched, whereas valeric acid, butyric acid, and IPA were significantly depleted in the fried meat group (all P FDR <0.05) (Fig. 2I–M).

Correlation Analysis

The correlations among the changes in the gut microbiota, fecal cometabolite levels, and altered outcome variables are presented in Fig. 3. The changes in the abundance of Flavonifractor were positively correlated with changes in fecal valeric acid levels (r = 0.226), whereas the changes in the abundance of Dorea were negatively correlated with it (r = −0.336). Additionally, the changes in the abundances of Dialister, Dorea, and Veillonella were positively correlated with changes in fecal carnitine levels (r = 0.218 for Dialister, r = 0.395 for Dorea, r = 0.314 for Veillonella) (all P < 0.05).

Figure 3

Correlation analyses among gut microbiota, cometabolites, and biomarkers of glucose homoeostasis, gut health, and inflammations. Red and blue lines, respectively, refer to significant positive and negative correlation, and gray lines refer to nonsignificant correlation; the color of the lines from light to dark reflected the magnitude of the correlation coefficients. FGF21, fibroblast growth factor 21; LBP, lipopolysaccharide binding protein; sCD14, soluble lipopolysaccharide receptor CD14.

Figure 3

Correlation analyses among gut microbiota, cometabolites, and biomarkers of glucose homoeostasis, gut health, and inflammations. Red and blue lines, respectively, refer to significant positive and negative correlation, and gray lines refer to nonsignificant correlation; the color of the lines from light to dark reflected the magnitude of the correlation coefficients. FGF21, fibroblast growth factor 21; LBP, lipopolysaccharide binding protein; sCD14, soluble lipopolysaccharide receptor CD14.

Close modal

The changes in fecal levels of valeric acid, butyric acid, and IPA were negatively correlated with changes in AUC for insulin levels (r = −0.266 for valeric acid, r = −0.282 for butyric acid), MIRI values (r = −0.243 for butyric acid, r = −0.286 for IPA), LPS levels (r = −0.198 for butyric acid, r = −0.243 for IPA) and TNF-α levels (r = −0.204 for valeric acid, r = −0.352 for butyric acid, r = −0.436). The changes in fecal MGA levels were negatively correlated with IGI values (r = −0.285), whereas they were positively correlated with TNF-α levels (r = 0.310). The changes in fecal carnitine levels were positively correlated with MIRI values (r = 0.241) and LPS (r = 0.214) and LBP/sCD14 (r = 0.238) levels (all P < 0.05).

In this study, we found that fried meat consumption influenced the composition and function of the gut microbiota and microbial-host cometabolites, which were associated with impaired glucose homoeostasis and increased the levels of intestinal endotoxin and systemic inflammation in overweight young adults. Additionally, the effect of the identified cometabolites that were related to the fried meat intake on glucose homoeostasis was further demonstrated in the mice by giving these cometabolites through gastric gavage.

Although previous studies have investigated the association between fried food intake and the risk of type 2 diabetes, uncontrolled nutritional confounders made these findings inconsistent (36). In this study, we provided the participants with an isocaloric diet, for which the AHEI score was >85, and the amount of food groups and composition of nutrients were held constant. However, we still found that participants who consumed fried meat had impaired glucose homoeostasis. It has been reported that high-heat cooking is a potent promoter of advanced glycation, and high levels of AGEs are therefore found in many fried foods (20). Previous studies showed that a diet with high levels of AGEs was associated with insulin resistance (21). Consistent with these studies, we also found that the participants in the fried group had higher levels of AGEs, although the frying temperature was rigorously controlled, and aromatic hydrocarbons were barely produced.

Moreover, we also observed that the participants in the experimental group had lower protein digestion and absorption rates. It has been reported that the protein digestion and absorption rate could influence the gut microbiota (22). In accordance, we also found that the participants in the experimental group had lower gut microbiota richness and that the overall structure of the gut microbiota at the OTU level was also different from that of the participants in the control group, suggesting that fried meat intake had a selective effect on the human gut microbiota. The microbiota predicted pathways also showed that those relating to glucose homoeostasis were also significantly changed during the intervention. Additionally, the ratio of Firmicutes and Bacteroidetes, the classical biomarker of type 2 diabetes (23), was significantly higher in the experimental group than in the control group, which further supported the effect of fried meat intake on glucose homoeostasis. At the genus level, fried meat intake decreased the abundances of Lachnospiraceae and Flavonifractor. Lachnospiraceae plays an important role in intestinal homeostasis (24), and it can prevent obesity and insulin resistance (25). Flavonifractor is an important species for intestinal health, and its abundance is negatively correlated with obesity (26). Moreover, the abundances of Dialister, Dorea, and Veillonella were increased in the experimental group. Dialister is a pathogenic bacterium, and its increased abundance is related to weight gain (27). Dorea abundance is positively correlated with obesity, and its increased abundance is found in patients with prediabetes (28).

Consistent with the gut microbiota results, fried meat intake also provoked significant shifts in fecal cometabolites, especially SCFA and amino acids metabolites. Fried meat intake significantly decreased the levels of some beneficial fecal cometabolites, including butyric acid, valeric acid, and IPA, whereas it also increased the levels of some deleterious cometabolites, including carnitine and MGA. A negative or positive association of these beneficial or deleterious cometabolites with the levels of biomarkers of glucose homoeostasis was also observed. Butyric acid is an SCFA, and its beneficial effect on glucose homoeostasis has already been demonstrated in previous studies (29,30). Therefore, we conducted the animal experiments to examine the effects of other identified cometabolites on glucose homoeostasis, and the beneficial effect of valeric acid and IPA, and deleterious effects of carnitine on glucose homoeostasis were further demonstrated, which were supported by previous studies. Valeric acid, as an SCFA, inhibits oxidative stress and neuroinflammation and modulates autophagy pathways (31). IPA can reduce plasma endotoxin levels, and consumption of an IPA-enriched diet can significantly lower fasting blood glucose levels and improve insulin resistance (32,33). Carnitine can be oxidized to trimethylamine N-oxide (TMAO), and high levels of plasma TMAO are associated with cardiovascular disease and diabetes (34).

We also identified several important bacterial genera as sensitive to fried meat intake based on the correlation between changes in abundance of the gut microbiota constituents and cometabolites. The results indicated that fried meat intake could decrease the abundance of Flavonifractor and increase the abundance of Dialister, Dorea, and Veillonella, which resulted in downregulated valeric acid levels and upregulated carnitine levels in feces. These results are supported by previous studies. It has been reported that Flavonifractor can produce SCFAs, Flavonifractor participates in the metabolism of valeric acid (35), and high abundance of Dorea could decrease the abundance of SCFA-producing bacteria, resulting in the reduction in the levels of fecal SCFAs, including valeric acid (36). Moreover, Dialister and Veillonella are involved in host amino acid metabolism, which could influence host TMAO levels through mediating carnitine production and metabolism (37,38).

In this randomized controlled-feeding trial, we applied 16S rRNA amplicon profiling coupled with quantitatively targeted bacterial metabolomics, which could provide additional knowledge for understanding the cross talk of host–gut microbiota metabolism in response to fried meat intake. However, we also recognize that our study had certain limitations. First, this study included only healthy overweight young adults because fried food is popular in this population (39), which could ensure compliance with this trial, and it has been documented that overweight during early adulthood is associated with a higher incidence of type 2 diabetes in late adulthood (40). Therefore, it is important to clarify the association between fried food intake and glucose metabolism in this population, which aids in establishing dietary intervention strategies for the early prevention of type 2 diabetes. However, these findings might not apply to other populations. Second, fecal sampling was conducted only at baseline and at the end of the trial. Future studies with more frequent sampling may provide a more complete picture of changes in the gut microbiota and metabolites. Third, for maintenance of the AHEI score at >85, this study provided only fish and chicken for the participants, and the temperature and duration during frying were rigorously controlled to limit the production of hazardous materials. With these approaches, the possibility that the type of meat and cooking method would influence our results cannot be excluded, and the gut microbiota could also be influenced by some local environment. Fourth, although it has been reported that IGI and MIRI correlate with the M value well, the euglycemic-hyperinsulinemic clamp, the gold standard, was not used to measure insulin sensitivity. Therefore, future studies examining other types of fried meat with different frying conditions in other regions and measurement of insulin sensitivity using the euglycemic-hyperinsulinemic clamp are still needed to provide additional evidence regarding the health impacts of fried food.

In summary, fried meat intake impaired glucose homoeostasis and increased the levels of intestinal endotoxin and systemic inflammation by influencing the gut microbiota and microbial-host cometabolites.

J.G. and X.G. contributed equally.

Clinical trial reg. no. ChiCTR1900028562, chictr.org.cn

See accompanying article, p. 1907.

This article contains supplementary material online at https://doi.org/10.2337/figshare.14731347.

Acknowledgments. The authors thank the participants for their cooperation.

Funding. This research was supported by funds from the National Natural Science Foundation (82030100 to Y.L., 81803227 to T.H.), the National Key R&D Program of China (2017YFC1307401 to C.S.), and the Young Elite Scientists Sponsorship Program by the China Association of Science and Technology (2019QNRC001).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.L., C.S., and T.H. conceived the study design. J.G., K.H., W.W., H.S., and H.W. participated in the execution of the study. J.G., X.G., and S.L. performed the statistical analysis. R.L., W.W., and Y.Z. repeated and validated the statistical analysis. K.H., X.L., W.W., and W.G. measured biomarker levels in biological samples. J.G., X.L., W.J., and W.W. performed mice tests and metabolites interventions. J.G., X.G., and T.H. wrote the manuscript. All authors provided critical revisions of the draft and approved the submitted draft. C.S. affirmed that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as originally planned (and, if relevant, registered) have been explained. C.S. is the guarantor of this work and, as such, 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.

1.
GBD 2017 Diet Collaborators
.
Health effects of dietary risks in 195 countries, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017
.
Lancet
2019
;
393
:
1958
1972
2.
Ley
SH
,
Hamdy
O
,
Mohan
V
,
Hu
FB
.
Prevention and management of type 2 diabetes: dietary components and nutritional strategies
.
Lancet
2014
;
383
:
1999
2007
3.
Liu
G
,
Zong
G
,
Wu
K
, et al
.
Meat cooking methods and risk of type 2 diabetes: results from three prospective cohort studies
.
Diabetes Care
2018
;
41
:
1049
1060
4.
Cahill
LE
,
Pan
A
,
Chiuve
SE
, et al
.
Fried-food consumption and risk of type 2 diabetes and coronary artery disease: a prospective study in 2 cohorts of US women and men
.
Am J Clin Nutr
2014
;
100
:
667
675
5.
Sayon-Orea
C
,
Martinez-Gonzalez
MA
,
Gea
A
,
Flores-Gomez
E
,
Basterra-Gortari
FJ
,
Bes-Rastrollo
M
.
Consumption of fried foods and risk of metabolic syndrome: the SUN cohort study
.
Clin Nutr
2014
;
33
:
545
549
6.
Farnetti
S
,
Malandrino
N
,
Luciani
D
,
Gasbarrini
G
,
Capristo
E
.
Food fried in extra-virgin olive oil improves postprandial insulin response in obese, insulin-resistant women
.
J Med Food
2011
;
14
:
316
321
7.
Shen
Q
,
Chen
YA
,
Tuohy
KM
.
A comparative in vitro investigation into the effects of cooked meats on the human faecal microbiota
.
Anaerobe
2010
;
16
:
572
577
8.
Murphy
RY
,
Marks
BP
.
Effect of meat temperature on proteins, texture, and cook loss for ground chicken breast patties
.
Poult Sci
2000
;
79
:
99
104
9.
Bax
ML
,
Aubry
L
,
Ferreira
C
, et al
.
Cooking temperature is a key determinant of in vitro meat protein digestion rate: investigation of underlying mechanisms
.
J Agric Food Chem
2012
;
60
:
2569
2576
10.
Partula
V
,
Mondot
S
,
Torres
MJ
, et al.;
Milieu Intérieur Consortium
.
Associations between usual diet and gut microbiota composition: results from the Milieu Intérieur cross-sectional study
.
Am J Clin Nutr
2019
;
109
:
1472
1483
11.
Zhou
Z
,
Wang
Y
,
Jiang
Y
, et al
.
Deep-fried oil consumption in rats impairs glycerolipid metabolism, gut histology and microbiota structure
.
Lipids Health Dis
2016
;
15
:
86
12.
Chiuve
SE
,
Fung
TT
,
Rimm
EB
, et al
.
Alternative dietary indices both strongly predict risk of chronic disease
.
J Nutr
2012
;
142
:
1009
1018
13.
Lund
SS
,
Tarnow
L
,
Stehouwer
CD
, et al
.
Impact of metformin versus repaglinide on non-glycaemic cardiovascular risk markers related to inflammation and endothelial dysfunction in non-obese patients with type 2 diabetes
.
Eur J Endocrinol
2008
;
158
:
631
641
14.
Bell
KJ
,
Fio
CZ
,
Twigg
S
, et al
.
Amount and type of dietary fat, postprandial glycemia, and insulin requirements in type 1 diabetes: a randomized within-subject trial
.
Diabetes Care
2020
;
43
:
59
66
15.
Kim
SH
,
Liu
A
,
Ariel
D
, et al
.
Effect of salsalate on insulin action, secretion, and clearance in nondiabetic, insulin-resistant individuals: a randomized, placebo-controlled study
.
Diabetes Care
2014
;
37
:
1944
1950
16.
Yeckel
CW
,
Taksali
SE
,
Dziura
J
, et al
.
The normal glucose tolerance continuum in obese youth: evidence for impairment in beta-cell function independent of insulin resistance
.
J Clin Endocrinol Metab
2005
;
90
:
747
754
17.
Matsuda
M
,
DeFronzo
RA
.
Insulin sensitivity indices obtained from oral glucose tolerance testing: comparison with the euglycemic insulin clamp
.
Diabetes Care
1999
;
22
:
1462
1470
18.
Wan
Y
,
Wang
F
,
Yuan
J
, et al
.
Effects of dietary fat on gut microbiota and faecal metabolites, and their relationship with cardiometabolic risk factors: a 6-month randomised controlled-feeding trial
.
Gut
2019
;
68
:
1417
1429
19.
Zheng
X
,
Huang
F
,
Zhao
A
, et al
.
Bile acid is a significant host factor shaping the gut microbiome of diet-induced obese mice
.
BMC Biol
2017
;
15
:
120
20.
Mark
AB
,
Poulsen
MW
,
Andersen
S
, et al
.
Consumption of a diet low in advanced glycation end products for 4 weeks improves insulin sensitivity in overweight women
.
Diabetes Care
2014
;
37
:
88
95
21.
Poulsen
MW
,
Bak
MJ
,
Andersen
JM
, et al
.
Effect of dietary advanced glycation end products on postprandial appetite, inflammation, and endothelial activation in healthy overweight individuals
.
Eur J Nutr
2014
;
53
:
661
672
22.
Wang
C
,
Huang
Z
,
Yu
K
, et al
.
High-salt diet has a certain impact on protein digestion and gut microbiota: a sequencing and proteome combined study
.
Front Microbiol
2017
;
8
:
1838
23.
Prattichizzo
F
,
Giuliani
A
,
Mensà
E
, et al
.
Pleiotropic effects of metformin: shaping the microbiome to manage type 2 diabetes and postpone ageing
.
Ageing Res Rev
2018
;
48
:
87
98
24.
Sun
D
,
Bai
R
,
Zhou
W
, et al
.
Angiogenin maintains gut microbe homeostasis by balancing α-Proteobacteria and Lachnospiraceae
.
Gut
2021
;
70
:
666
676
25.
Shively
CA
,
Register
TC
,
Appt
SE
, et al
.
Consumption of Mediterranean versus western diet leads to distinct mammary gland microbiome populations
.
Cell Rep
2018
;
25
:
47
56.e3
26.
Borgo
F
,
Garbossa
S
,
Riva
A
, et al
.
Body mass index and sex affect diverse microbial niches within the gut
.
Front Microbiol
2018
;
9
:
213
27.
Muñiz Pedrogo
DA
,
Jensen
MD
,
Van Dyke
CT
, et al
.
Gut microbial carbohydrate metabolism hinders weight loss in overweight adults undergoing lifestyle intervention with a volumetric diet
.
Mayo Clin Proc
2018
;
93
:
1104
1110
28.
Allin
KH
,
Tremaroli
V
,
Caesar
R
, et al.;
IMI-DIRECT consortium
.
Aberrant intestinal microbiota in individuals with prediabetes
.
Diabetologia
2018
;
61
:
810
820
29.
Dalile
B
,
Van Oudenhove
L
,
Vervliet
B
,
Verbeke
K
.
The role of short-chain fatty acids in microbiota-gut-brain communication
.
Nat Rev Gastroenterol Hepatol
2019
;
16
:
461
478
30.
Henagan
TM
,
Stefanska
B
,
Fang
Z
, et al
.
Sodium butyrate epigenetically modulates high-fat diet-induced skeletal muscle mitochondrial adaptation, obesity and insulin resistance through nucleosome positioning
.
Br J Pharmacol
2015
;
172
:
2782
2798
31.
Jayaraj
RL
,
Beiram
R
,
Azimullah
S
, et al
.
Valeric acid protects dopaminergic neurons by suppressing oxidative stress, neuroinflammation and modulating autophagy pathways
.
Int J Mol Sci
2020
;
21
:
7670
32.
Zhao
ZH
,
Xin
FZ
,
Xue
Y
, et al
.
Indole-3-propionic acid inhibits gut dysbiosis and endotoxin leakage to attenuate steatohepatitis in rats
.
Exp Mol Med
2019
;
51
:
1
14
33.
Abildgaard
A
,
Elfving
B
,
Hokland
M
,
Wegener
G
,
Lund
S
.
The microbial metabolite indole-3-propionic acid improves glucose metabolism in rats, but does not affect behaviour
.
Arch Physiol Biochem
2018
;
124
:
306
312
34.
Wang
Z
,
Klipfell
E
,
Bennett
BJ
, et al
.
Gut flora metabolism of phosphatidylcholine promotes cardiovascular disease
.
Nature
2011
;
472
:
57
63
35.
Gupta
A
,
Dhakan
DB
,
Maji
A
, et al
.
Association of Flavonifractor plautii, a flavonoid-degrading bacterium, with the gut microbiome of colorectal cancer patients in India
.
mSystems
2019
;
4
:
e00438
e19
36.
Rodriguez
J
,
Neyrinck
AM
,
Zhang
Z
, et al
.
Metabolite profiling reveals the interaction of chitin-glucan with the gut microbiota
.
Gut Microbes
2020
;
12
:
1810530
37.
Liu
F
,
Fan
C
,
Zhang
L
, et al
.
Alterations of gut microbiome in Tibetan patients with coronary heart disease
.
Front Cell Infect Microbiol
2020
;
10
:
373
38.
Leiva-Gea
I
,
Sánchez-Alcoholado
L
,
Martín-Tejedor
B
, et al
.
Gut microbiota differs in composition and functionality between children with type 1 diabetes and MODY2 and healthy control subjects: a case-control study
.
Diabetes Care
2018
;
41
:
2385
2395
39.
Taveras
EM
,
Berkey
CS
,
Rifas-Shiman
SL
, et al
.
Association of consumption of fried food away from home with body mass index and diet quality in older children and adolescents
.
Pediatrics
2005
;
116
:
e518
e524
40.
Zheng
Y
,
Manson
JE
,
Yuan
C
, et al
.
Associations of weight gain from early to middle adulthood with major health outcomes later in life
.
JAMA
2017
;
318
:
255
269
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/content/license.