Macronutrient, Energy, and Bile Acid Metabolism Pathways Altered Following a Physiological Meal Challenge, Relative to Fasting, among Guatemalan Adults

ABSTRACT

Background

The healthy human metabolome, including its physiological responses after meal consumption, remains incompletely understood. One major research gap is the limited literature assessing how human metabolomic profiles differ between fasting and postprandial states after physiological challenges.

Objectives

Our study objective was to evaluate alterations in high-resolution metabolomic profiles following a standardized meal challenge, relative to fasting, in Guatemalan adults.

Methods

We studied 123 Guatemalan adults without obesity, hypertension, diabetes, metabolic syndrome, or comorbidities. Every participant received a standardized meal challenge (520 kcal, 67.4 g carbohydrates, 24.3 g fat, 8.0 g protein) and provided blood samples while fasting and at 2 h postprandial. Plasma samples were assayed by high-resolution metabolomics with dual-column LC [C18 (negative electrospray ionization), hydrophilic interaction LC (HILIC, positive electrospray ionization)] coupled to ultra-high-resolution MS. Associations between metabolomic features and the meal challenge timepoint were assessed in feature-by-feature multivariable linear mixed regression models. Two algorithms (mummichog, gene set enrichment analysis) were used for pathway analysis, and P values were combined by the Fisher method.

Results

Among participants (62.6% male, median age 43.0 y), 1130 features (C18: 777; HILIC: 353) differed between fasting and postprandial states (all false discovery rate–adjusted q < 0.05). Based on differing C18 features, top pathways included: tricarboxylic acid cycle (TCA), primary bile acid biosynthesis, and linoleic acid metabolism (all P_combined < 0.05). Mass spectral features included: taurine and cholic acid in primary bile acid biosynthesis; and fumaric acid, malic acid, and citric acid in the TCA. HILIC features that differed in the meal challenge reflected linoleic acid metabolism (P_combined < 0.05).

Conclusions

Energy, macronutrient, and bile acid metabolism pathways were responsive to a standardized meal challenge in adults without cardiometabolic diseases. Our findings reflect metabolic flexibility in disease-free individuals.

Introduction

Cardiometabolic health, including the healthy human metabolome, remains incompletely understood (1–4). Cardiometabolic diseases are the leading cause of mortality worldwide (5). Elucidating the healthy metabolome is important to shed light on how to maintain more effectively optimal metabolic health over the life course.

One major research gap is characterizing the dynamic nature of systemic physiology in the metabolome of healthy individuals (3, 4, 6, 7). Energy homeostasis requires rapid adaptations of metabolic processes throughout heterogeneous energy states, including during meal consumption, exercise, and rest (1–4, 8). A solution is to use activity metabolomics (6, 9) to evaluate metabolic flexibility in a healthy state (3, 4).

Metabolic flexibility is a broad concept that describes physiological adaptability to differing energy and metabolic conditions (3, 4). A multitude of responses at the systemic, tissue, and cellular level in humans and other species have been characterized as different aspects of metabolic flexibility (3, 4). Two examples include the shifting energy metabolism of helminths in aerobic compared with anaerobic environments, and macronutrient fuel selection switching from lipids in a fasting state to glucose in a postprandial state in humans (3, 4).

Acknowledging the absence of a single standard definition for metabolic flexibility, in the present study we define metabolic flexibility as the changes in metabolomic features after a meal challenge, compared with while fasting. Metabolomic data reflect a plethora of influences, such as genetic and environmental factors, that affect physiological processes over the life course (10–13). Metabolic challenges, including meal consumption, represent a means to examine physiological perturbations in activity metabolomics studies in humans in vivo (14).

To date, few studies have assessed metabolomic changes before and after meal consumption in healthy individuals (Supplemental Table 1) (15–23). These previous studies have been characterized by small sample sizes (ranging from 9 to 21 participants) and extensive heterogeneity of mixed macronutrient composition meal challenges (15–23). Compared with a bolus dosage of glucose [oral-glucose-tolerance test (OGTT)], a mixed macronutrient intervention is more reflective of a physiological metabolic challenge, which is necessary to elucidate and characterize a healthy metabolome. Previous studies have shown differences between OGTTs and mixed meal challenges, including in insulin response (24), glucose concentration changes (25), and β-cell function (26). Our study objective was to evaluate high-resolution metabolomic profiles following a standardized meal challenge, relative to while fasting, in adults without cardiometabolic diseases.

Methods

Study population

We studied adult participants of the Institute of Nutrition of Central America and Panama (INCAP) Nutritional Supplementation Trial Longitudinal Study, which was established in Guatemala in 1969. Further details regarding the INCAP trial have been previously published (27, 28). In the present study, participants were enrolled in a follow-up data collection from 2015 to 2017 (29). Previous exclusion criteria included: 1) loss to follow-up (e.g., death, emigration from Guatemala, inability to contact, did not agree to voluntarily participate); 2) pregnancy or lactating; and 3) no data for the biological samples and key outcomes of interest. For this analysis, we additionally excluded any participants with obesity, hypertension, diabetes, metabolic syndrome, or any comorbidities of these cardiometabolic diseases, as defined elsewhere (24) (Supplemental Table 2).

Meal challenge

Following informed consent and an overnight fast, participants visited the research clinic in the morning. Participants were excluded from the meal challenge if their fasting blood glucose was ≥180 mg/dL or they self-reported having diabetes. Study participants received a standardized beverage, which was comprised of Incaparina (a vegetable protein mixture developed by INCAP), skim milk (lactose-free), safflower oil, and sugar. Each portion (259 mL) contained 520 kcal, 8.0 g protein, 24.3 g fat, and 67.4 g carbohydrate (Supplemental Table 3).

Data collection

Trained phlebotomists collected venous blood samples at baseline (fasting) and 120 min after the meal challenge (postprandial). Plasma samples for metabolomic analysis were collected in heparin tubes. Samples were stored on ice prior to centrifugation at 3000 rpm (906 × g) for 10 min, and subsequently at −80°C until assay.

Trained study staff interviewed study participants regarding sociodemographic and clinical information, including use of medications (29). Standard protocols were used for anthropometric measurements (29). Additional details regarding the 2015–2017 study data collection are in a prior publication (29).

High-resolution metabolomics

Plasma samples

Pairs of plasma samples (fasting, postprandial state) were sorted by sex, village, and birth year. A subset was randomly selected and ordered for metabolomic assay. Each sample was assayed in triplicate. After frozen samples were thawed, acetonitrile (2:1 v/v; HPLC grade; Millipore; Sigma Aldrich) was added to precipitate protein (30). Subsequently, samples were centrifuged (14,000 × g for 10 min at 4°C) and the supernatant was stored at 4°C in a refrigerated autosampler prior to assay (30, 31).

Fourteen commercially available, isotope-labeled internal standards were included: [¹³C₆]-d-glucose, [¹⁵N]-indole, [2-¹⁵N]-l-lysine dihydrochloride, [¹³C₅]-l-glutamic acid, [¹³C₇]-benzoic acid, [3,4-¹³C₂]-cholesterol, [¹⁵N]-l-tyrosine, [trimethyl-¹³C₃]-caffeine, [¹⁵N₂]-uracil, [3,3-¹³C₂]-cystine, [1,2-¹³C₂]-palmitic acid, [¹⁵N,¹³C₅]-l-methionine, [¹⁵N]-choline chloride, and 2′-deoxyguanosine-[¹⁵N₂,¹³C₁₀]-5′-monophosphate (30, 31). Human reference plasma from the National Institute of Standards and Technology (NIST) [standard reference material (SRM) 1950] and a pooled reference plasma (Q-std3) prepared from commercial human plasma samples (Equitech Bio) were also included for quality control (32).

LC-MS

Plasma samples were assayed by LC Fourier transform MS (LC-FT-MS) with 2 chromatographic columns: C18 (Higgins Analytical, Targa, 2.1 × 50 mm) with negative electrospray ionization, and hydrophilic interaction LC (HILIC; Waters BEH Amide 2.1 × 50 mm) with positive electrospray ionization (31, 33). A Switchos control valve (LC Packings) allowed alternation between the 2 columns (31). Data acquisition occurred by mass spectrometer (Orbitrap Fusion; Thermo Fisher Scientific) with specifications as previously described (31, 34, 35). The scan range for the detection of m/z scan was 85–1250 m/z, and mass resolution was 120,000 (36).

Data extraction

Raw data (.raw files) were collected throughout the chromatographic separation and converted to .cdf files (Xcalibur software; Thermo Fisher Scientific). apLCMS (37) and xMSanalyzer (38) were used for the extraction and initial preprocessing of chromatographic data. Preprocessing included noise reduction, peak identification, retention time (RT) correction, peak alignment, feature quantification, weak signal detection, and batch effect adjustment with ComBat (37–40). A feature was defined as a unique combination of m/z and RT. The ion intensity (peak area) of each feature was integrated.

Standard operating procedures and quality control were based on prior studies, including via xMSanalyzer (38). For each sample, pairwise Pearson correlations between replicates were assessed and averaged. Prior to exclusions of participants and data filtering of features, the median averaged pairwise Pearson correlation between replicates was 98.7% (IQR: 97.0, 99.5%) across samples from C18 data, and 99.5% (IQR: 98.6, 99.8%) across all samples from HILIC data. Samples with mean Pearson correlation coefficients <0.75 across technical replicates were excluded.

In terms of data filtering, among data from each column (C18, HILIC), features observed in <80% of study participant samples were excluded (Supplemental Figure 1) (41). Subsequently, any feature peak area values that were missing or zero were assigned as half of the lowest observed value (limit of detection) of the feature peak area across all samples for each column.

Statistics and bioinformatics

Statistical analysis was conducted using R (version 3.5.1; R Foundation for Statistical Computing) and SAS (version 9.4; SAS Institute Inc). Statistical significance was based on 2-sided hypothesis tests, and α < 0.05. In feature-by-feature multivariable regressions with metabolomic data, a false discovery rate–adjusted q value <0.05 was considered significant. A complete case approach was used for all clinical and sociodemographic variables. Metabolic feature ion intensities (peak areas) were converted to log₂-normalized values.

Descriptive analysis

Continuous and categorical variables were reported as median (IQR) or n (percentage). Subgroups were compared by parametric (t) or nonparametric (Wilcoxon) test statistics.

Feature selection approach via regressions

We used a feature selection approach based on multivariable linear regressions with repeated measurements. The model equation was:

(1)

where each study participant was denoted as i, feature was j, and timepoint was t. We included features with a β coefficient of the meal challenge timepoint (β₁) of q < 0.05 in pathway enrichment analysis.

Functional pathway analysis and feature annotation

From the regression results (β coefficients, q values), functional pathway analysis was assessed by mummichog and gene set enrichment analysis (GSEA) algorithms [mummichog (version 1.0) in MetaboAnalyst] (42, 43). The Kyoto Encyclopedia of Genes and Genomes (KEGG) for Homo sapiens was the pathway reference library (44), and mass accuracy was 5 parts per million. To summarize the results from both algorithms (mummichog, GSEA), a combined P value (P_combined) was calculated via the Fisher method (42, 43).

Features in the top pathways were matched with annotations in the KEGG (44) and Human Metabolome Database (HMDB) (45). We used xMSannotator, which considers multiple criteria in its algorithm (46) and incorporates HMDB reference database information (45). The xMSannotator algorithm includes several steps, including: 1) pairwise correlation analysis of all features (m/z, RT, peak area); 2) network modularity analysis; 3) RT-based clustering; 4) mass defect analysis; 5) database matching; and 6) score assignment (46). An xMSannotator score is assigned, based on correlation strength, RT difference, number of matching adducts and isotopes, and weighting based on more probable adducts and isotopes (46). A nonzero xMSannotator score is only assigned if 3 conditions are satisfied, based on the initial steps (46). We also assigned identification confidence scores to some features, based on the 5-level system proposed by the Metabolomics Standards Initiative (MSI) (47). Identities of select metabolites from each of the pathways presented were confirmed by coelution and MS/MS fragmentation patterns relative to authentic standards.

Quantification of metabolites in top pathways

We used a reference standardization approach (32) to estimate concentrations for selected, confirmed metabolites in top pathways based on metabolite concentrations in Q-Std3. Triplicates of 2 Q-Std3 samples were included as the beginning, middle, or end of every batch of 40 samples. Metabolite concentrations were calibrated either by: 1) NIST SRM 1950 for human plasma, if metabolite concentrations were available; or 2) external calibration by the method of standard additions (32, 48–51).

Ethical conduct of research

The Institutional Review Boards at Emory University and INCAP approved the study protocol. All study participants provided their written informed consent.

Results

Of the 123 study participants, 62.6% were men (Table 1). The median age was 43.0 y (IQR: 40.0, 46.0 y), and ages ranged between 37.0 and 52.0 y (Table 1). Median BMI was lower in men (24.7 kg/m²; IQR: 23.1, 26.6) than in women (25.6; IQR: 24.3, 27.9) (P < 0.05) (Table 1).

TABLE 1.

Sociodemographic and cardiometabolic indicators in Guatemalan adults (n = 123)¹

	Overall (n = 123)	Males (n = 77)	Females (n = 46)	P
Sociodemographic
Age at follow-up,5 y	43.0 (40.0, 46.0)	44.0 (40.0, 47.0)	43.0 (40.0, 45.0)	0.422
Anthropometric
BMI, kg/m2	25.0 (23.3, 27.0)	24.7 (23.1, 26.6)	25.6 (24.3, 27.9)	<0.052
Biochemical
Glucose profile
Fasting blood glucose, mg/dL	94.9 (90.9, 98.8)	96.0 (93.0, 99.6)	93.2 (89.7, 95.8)	<0.012
Postprandial glucose, mg/dL	99.4 (90.9, 112)	94.7 (88.3, 104)	105 (98.9, 118)	<0.012
HbA1c, %	5.6 (5.4, 5.8)	5.6 (5.4, 5.8)	5.6 (5.4, 5.8)	0.333
Plasma lipid concentrations
Triglycerides, mg/dL	108 (83, 142)	115 (87, 157)	102 (73, 125)	0.062
Total cholesterol, mg/dL	167 (146, 194)	162 (146, 187)	174 (147, 199)	0.243
HDL cholesterol, mg/dL	42.7 (36.8, 49.0)	40.6 (35.9, 46.8)	47.0 (40.4, 54.2)	<0.014
Non-HDL cholesterol,6 g/dL	0.13 (0.11, 0.14)	0.12 (0.10, 0.14)	0.13 (0.11, 0.15)	0.923
Clinical
Systolic blood pressure, mm Hg	115 (108, 121)	117 (110, 122)	110 (102, 119)	<0.014
Diastolic blood pressure, mm Hg	67.5 (63.0, 73.0)	68.0 (64.0, 73.5)	66.5 (61.5, 70.5)	0.023

¹Study participants did not have any of the assessed cardiometabolic diseases, that is, obesity (52), hypertension (53), diabetes (54), or metabolic syndrome (55), including comorbidities (see Supplemental Table 2). Data values are median (IQR). HbA1c, glycated hemoglobin.

² P values based on Wilcoxon rank-sum tests (for nonnormally distributed continuous variables).

³ P values from t tests (for normally distributed continuous variables) with equal variance (pooled).

⁴ P values from t tests (for normally distributed continuous variables) with unequal variance (Satterthwaite).

⁵At study visit date (of biological sample collection) in 2015–2017 data collection.

⁶Non-HDL cholesterol (mg/dL) calculated as the difference between total (mg/dL) and HDL cholesterol (mg/dL) plasma concentrations.

Overall, 9849 (C18) and 13,908 (HILIC) metabolomic features were observed across all study participant samples. After data filtering, 5085 C18 and 7444 HILIC features remained eligible for the feature selection process.

Metabolomic profiling in a standardized meal challenge

In total, 1130 features (C18: 777; HILIC: 353) differed following the meal challenge, compared with a fasting state (q < 0.05; Table 2). Among these features, 554 (49.4%) had putative annotations (MSI levels 2 and 3), based on xMSannotator (Table 2). Specifically, 43.6% of C18 features and 62.0% of HILIC features had annotations from xMSannotator (MSI levels 2 and 3; Table 2).

TABLE 2.

Summary of metabolomic features differing by meal challenge¹

	Meal challenge
	LC-FT-MS columns
	C18	HILIC
No. of differing features	(n = 123)	(n = 123)
Feature selection
Stage 1: overall linear regression2
q < 0.053	777	353
Annotations
Features with putative annotations	339	219
Total no. of putative annotations	1710	2463

¹Unless otherwise stated, values indicate the number of features with log₂-normalized peak areas, which differed by the meal challenge timepoint (fasting, postprandial). FDR, false discovery rate; HILIC, hydrophilic interaction liquid chromatography; LC-FT-MS, liquid chromatography Fourier transform mass spectrometry.

²After data filtering, 5085 (C18) and 7444 (HILIC) features remained eligible for the feature selection approach. A complete case approach was used in multivariable regressions. For each feature, a linear mixed model with repeated measurements was used (Proc Mixed in SAS). The model equation was: Y (log₂-normalized feature peak area)_{i, j, t} = β_{0i, j, t} + β₁X₁ [timepoint (postprandial, baseline)] _{i, j, t} + β₂X₂ (age at study visit)_i + β₃X₃ (sex)_i, where each study participant was denoted as _i, feature was _j, and timepoint was _t. The numbers of features with a β coefficient of the meal challenge timepoint (β₁) with an FDR-adjusted P value (q) < 0.05 were included in this table.

³These features were subsequently eligible for annotations and functional pathway analysis.

C18 features that responded to the meal challenge were associated with 5 pathways, based on both GSEA and mummichog algorithms (all P_combined < 0.05; Figure 1). These pathways were primary bile acid biosynthesis, citrate cycle [tricarboxylic acid (TCA) cycle], fatty acid metabolism, linoleic acid metabolism, and biotin metabolism (all P_combined < 0.05; Figure 1B; Supplemental Table 4). A pathway for valine, leucine, and isoleucine biosynthesis was significant according to the mummichog algorithm (P_mummichog = 0.02), although not according to GSEA (P_GSEA > 0.05; Figure 1A; Supplemental Table 4).

FIGURE 1.

Top pathways associated with metabolomic feature peak areas (C18) responsive to standardized meal challenge. Features were assayed by LC-Fourier transform-MS (C18 with negative electrospray ionization). Pathway analysis [mummichog and gene set enrichment algorithms via MetaboAnalyst (42, 43)] was conducted among features selected via a multivariable linear mixed regression model screening approach (q < 0.05; stage 1 in feature selection approach). (A) In this pathway analysis plot, KEGG pathways are visually represented. Each dot represents a pathway. The size and location of every dot reflect the enrichment factor and P value. Circle sizes indicate the most represented pathways, based on enrichment factor. The enrichment factor is the ratio between the number of hits in the pathway and the expected hits in the pathway. Quadrants indicate significant pathways based on the mummichog (y-axis, light gray), GSEA (x-axis, light gray), or both algorithms (dark gray). (B) Top pathways and their respective P values (mummichog and GSEA algorithms, combined) are shown. Pathways were ranked by the combined P values from the mummichog and GSEA algorithms, based on the selected features in stage 1 regressions and the Fisher method. GSEA, gene set enrichment analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes; TCA, tricarboxylic acid cycle.

From HILIC features, linoleic acid metabolism was a significant pathway, according to both the mummichog (P_mummichog < 0.05) and GSEA (P_GSEA < 0.05) algorithms (P_combined < 0.05; Figure 2A; Supplemental Table 4). Primary bile acid biosynthesis was a significant pathway based on the mummichog algorithm (P_mummichog = 0.03), although not GSEA (P_GSEA > 0.05; Supplemental Table 4). Conversely, 1-carbon pool by folate, cyanoamino acid metabolism, sphingolipid metabolism, ascorbate and aldarate metabolism, fatty acid elongation in mitochondria, and phenylalanine metabolism were significant pathways based on the GSEA algorithm (all P_GSEA < 0.05), but not mummichog (P_mummichog > 0.05; Supplemental Table 4).

FIGURE 2.

Top pathways associated with metabolomic feature peak areas (HILIC) responsive to standardized meal challenge. Features were assayed by LC-Fourier transform-MS (HILIC with positive electrospray ionization). Pathway analysis [mummichog and gene set enrichment algorithms via MetaboAnalyst (42, 43)] was conducted among features selected via a multivariable linear mixed regression model screening approach (q < 0.05; stage 1 in feature selection approach). (A) In this pathway analysis plot, KEGG pathways are visually represented. Each dot represents a pathway. The size and location of every dot reflect the enrichment factor and P value. Circle sizes indicate the most represented pathways, based on enrichment factor. The enrichment factor is the ratio between the number of hits in the pathway and the expected hits in the pathway. Quadrants indicate significant pathways based on the mummichog (y-axis, light gray), GSEA (x-axis, light gray), or both algorithms (dark gray). (B) Top pathways and their respective P values (mummichog and GSEA algorithms, combined) are shown. Pathways were ranked by the combined P values from the mummichog and GSEA algorithms, based on the selected features in stage 1 regressions and the Fisher method. GSEA, gene set enrichment analysis; HILIC, hydrophilic interaction liquid chromatography; KEGG, Kyoto Encyclopedia of Genes and Genomes.

TCA

In C18 data, 8 spectral features were in the TCA cycle, including fumaric acid, l-malic acid, oxoglutaric acid, l-lysine, and citric acid (Supplemental Figure 2, Table 3, Figure 3, Supplemental Table 5A). Median plasma concentrations of l-malic acid (5.25 μM compared with 4.12 μM) and citric acid (327.58 μM compared with 266.22 μM) were higher following the prandial challenge, compared with baseline (both P ≤ 0.02).

TABLE 3.

Metabolomic features differing by meal challenge in top pathways¹

	Feature						xMSannotator3				Annotation score
Pathway2	m/z	RT, s	Adduct	Log2 peak area (median; IQR)	Concentration, μM (median; IQR)4	Column	Chemical compound	Formula	HMDB monoisotopic mass	HMDB ID	xMSannotator confidence score	In-house reference library	MSI level
Linoleic acid metabolism	281.2473	25.6	M + H	16.71 (16.28, 17.14)	—	HILIC+	Linoleic acid	C18H32O2	280.24023	HMDB00673	3	✓	2
	279.2335	235.8	M-H	30.19 (29.55, 30.79)	7820.72 (5194.41, 11,637.75)	C18−	Linoleic acid	C18H32O2	280.24023	HMDB00673	3	✓	2
Primary bile acid synthesis	124.0073	24.3	M-H	20.51 (20.19, 20.85)	7.12 (5.85, 8.74)	C18−	Taurine	C2H7NO3S	125.014664	HMDB00251	3	✓	2
	407.2812	29.8	M-H	14.93 (7.02, 16.51)	0.03 (0.02, 0.08)	C18−	Cholic acid	C24H40O5	408.287574	HMDB00619	3	—	3
	433.3329	235.3	M-H	14.73 (13.72, 15.51)	—	C18−	3α,7α-Dihydroxycoprostanic acid	C27H46O4	434.33961	HMDB00359	2	—	3
	433.3329	235.3	M-H	14.73 (13.72, 15.51)	—	C18−	3α,7α,12α-Trihydroxy-5β-cholestan-26-al	C27H46O4	434.33961	HMDB03533	2	—	3
TCA cycle	115.004	18.8	M-H	16.99 (16.35, 17.35)	14.81 (11.65, 18.65)	C18−	Fumaric acid	C4H4O4	116.010959	HMDB00134	2	✓	3
	133.0143	21.9	M-H	20.41 (19.46, 21.05)	4.67 (3.11, 6.11)	C18−	l-Malic acid	C4H6O5	134.021523	HMDB00156	3	✓	2
	145.0143	20.6	M-H	21.80 (20.85, 22.39)	—	C18−	Oxoglutaric acid	C5H6O5	146.021523	HMDB00208	2	—	3
	145.0982	28.2	M-H	16.83 (14.23, 17.65)	176.31 (130.17, 264.13)	C18−	l-Lysine	C6H14N2O2	146.105528	HMDB00182	2	✓	3
	173.0085	23.2	M-H2O-H	19.91 (19.37, 20.49)	—	C18−	Citric acid	C6H8O7	192.027003	HMDB00094	3	—	3
	191.0199	22.4	M-H	24.37 (23.90, 24.87)	286.35 (184.76, 416.27)	C18−	Citric acid	C6H8O7	192.027003	HMDB00094	3	—	3

¹C18-, C18 column with negative electrospray ionization; GSEA, gene set enrichment analysis; H, hydrogen; HILIC+, hydrophilic interaction liquid chromatography column with positive electrospray ionization; HMDB, Human Metabolome Database; KEGG, Kyoto Encyclopedia of Genes and Genomes; M, analyte molecule; MSI, Metabolomics Standards Initiative; RT, retention time; TCA, tricarboxylic acid.

²Top pathways based on mummichog and GSEA algorithms (P_combined < 0.05) in MetaboAnalystR (mummichog, v 1.0).

³Features in top pathways. Annotations based on feature matches in KEGG (mummichog) and HMDB (xMSannotator).

⁴Concentrations calculated based on the observed feature peak area, Q-std3 peak area, and known metabolite concentration. Only among features that matched with available calibrated reference data for metabolite.

FIGURE 3.

(A–F) Plasma concentrations of metabolites, stratified by prandial challenge. Horizontal lines in each boxplot indicate the median, 25th and 75th percentile values. Examples of selected features were included in this figure. Features were matched with calibrated reference data for metabolite. Metabolite plasma concentrations were calculated based on the observed feature peak area, Q-std3 peak area, and known metabolite concentration. The comparisons between median plasma concentrations at fasting compared with postprandial timepoints were based on Wilcoxon test statistics. RT, retention time.

Fatty acid metabolism

From C18 data, there were 4 features in fatty acid metabolism and 2 in linoleic acid metabolism (Supplemental Tables 4 and 5A). Linoleic acid was observed in both the C18 (m/z = 279.2335; RT = 235.8 s) and HILIC (m/z = 281.2473; RT = 25.6 s) columns (Table 3, Supplemental Tables 5A and 5B).

Bile acid metabolism

Observed C18 features in primary bile acid biosynthesis included: taurine, cholic acid, 3α,7α-dihydroxycoprostanic acid, and 3α,7α,12α-trihydroxy-5β-cholestan-26-al (Table 3, Figure 3, Supplemental Table 4). Median plasma concentration of taurine (m/z = 124.0073; RT = 24.3 s) was 7.58 μM while fasting, and 6.66 μM in a postprandial state (P < 0.01).

Discussion

In Guatemalan adults in our study, >1100 metabolomic features differed following a standardized meal challenge compared with the fasting state. From functional analysis, these features were associated with pathways of energy (TCA cycle), macronutrient, and bile acid metabolism. These changes in metabolomic profiling following a meal challenge reflect short-term metabolic flexibility in individuals without cardiometabolic diseases.

Metabolomic profiles differed between fasting and postprandial states, including in terms of functional pathways and individual features. Prior studies have elucidated differences between fasting compared with nonfasting states. For example, in the diagnoses of metabolic diseases such as diabetes and metabolic syndrome, the cutoff values and utility of key biomarkers (e.g., glucose, HDL cholesterol, triglycerides) depend on whether blood samples were collected while fasting or nonfasting (54, 56). Previous literature has raised questions about the strength of evidence for recommendations of fasting blood samples in cardiometabolic disease screening and diagnoses (57–60). Nonfasting triglycerides have been found to be predictive of cardiovascular disease outcomes (59, 61), and in 1 study to be more associated than fasting triglycerides (59). Based on our findings, future metabolomic studies should account for fasting or postprandial blood collection in study designs and harmonizing results across studies.

In our study participants, the TCA cycle was among the top pathways associated with features that changed following the standardized meal challenge. Two other studies also found that the TCA cycle (15) and energy metabolism (19) pathways differed after meal challenges. Previous evidence has demonstrated the role of the hepatic TCA cycle in the mitochondrial matrix, in which pyruvate oxidation occurs (62). Alterations in TCA cycle metabolism–related indicators have also been observed in people with metabolic diseases (63).

Fatty acid metabolism, particularly linoleic acid metabolism, was among the top pathways that were affected by the meal challenge. Median linoleic acid concentration in our participants was higher than in prior studies (64, 65), although possible explanations include higher dietary consumption of foods rich in linoleic acids (e.g., corn or maize oils) in Guatemala (66). Two prior metabolomic studies have shown that free fatty acids (16) and other lipids [sphingolipids (21)] responded to a meal challenge. However, direct comparisons with our results are difficult. One long-term study provided participants with olive oil, peanuts, Gruyère cheese, and butter, and had an initial overfeeding diet phase (days 5–34) followed by ad libitum food consumption (days 34–62) (16). The second study provided a single meal challenge comprised of refined wheat bread (50 g carbohydrates, 9 g protein, 4.2 g fat, 2.7 g dietary fiber), cucumber (40 g), and a noncaloric orange beverage (300 mL) (21).

Other previous studies have also found that amino acid pathways, including branched-chain amino acids and phenylalanine, responded to meal challenges (15–17, 21). Further metabolomic and mechanistic studies are needed to evaluate the extent that other key factors [e.g., dietary composition (carbohydrate-to-protein ratio), hepatic AMP-activated kinase activity (67)] affect macronutrient metabolism after meal consumption, given that these could provide insights regarding how to maintain metabolic flexibility over the life course.

The primary bile acid biosynthesis pathway was responsive to a meal challenge. Two previous small-scale studies have shown similar results (17, 68). In addition to the classical roles of bile acids in digestion, putative mechanisms explaining the link between bile acids and metabolism include endocrine signaling via bile acid–activated nuclear and membrane receptors (farnesoid X receptor, TGR5) (69–71), which can have downstream effects on glucagon-like peptide 1 and glucose metabolism (72). However, many preliminary studies have used murine models or have been in vitro experiments (70), and therefore further studies need to determine any clinical implications of the nonclassical roles of bile acids in metabolism (73). Separately, the potential influences of other factors, such as the microbiome (71) and tissue-specific receptor expression (69), on the link between bile acids and metabolism remain unclear.

This study had several limitations. One key challenge was metabolite identification, including the need for confirmation of putative feature annotations [MSI level 1 (47)]. Furthermore, improvements in reference databases are needed to address the substantial proportion of features with unknown identities. There is substantial diversity in metabolomic assay methodology, including the analytical technique (e.g., NMR, GC-MS, or LC-MS), instrumentation, sample collection and preparation protocols, preprocessing assumptions, data filtering, and bioinformatics (11). Our results should be interpreted while taking into account the specific methodological approaches we used. Metabolomic profiling encompasses small metabolites (generally <1.5 kDa) and related functional pathways; however, further studies with orthogonal techniques are needed to comprehensively characterize the healthy metabolome after meal consumption. For example, proteomic and lipidomic data are specific subsets of the metabolome, and based on different workflows and databases that are respectively optimized for proteins and lipids. We did not assess physiological factors, including circadian rhythms, genetics, and adipose and skeletal muscle tissues, that potentially affect metabolism following meal consumption.

One strength of this study was the standardized meal challenge, particularly given the heterogeneity of challenges in the literature to date. Additionally, we studied a larger sample size relative to study populations in prior studies (Supplemental Table 1). High-resolution untargeted metabolomics allows for more comprehensive coverage of features, compared with targeted metabolomics (11). In untargeted metabolomics, the wide range of observed metabolites have heterogeneous chemical properties (e.g., polarity, hydrophobicity) and interactions with different column types (11). Dual analytical columns (e.g., C18, HILIC) and different electrospray ionization modes (positive, negative) also increased feature coverage (31). We converted the relative ion intensities of selected features to plasma concentrations, based on a reference standardization approach. Reporting blood concentrations facilitates comparisons with other studies and addresses a major shortcoming of relative quantitation in many nontargeted metabolomic studies.

In conclusion, metabolomic features that differed following a standardized meal challenge were associated with energy (TCA cycle), macronutrient, and bile acid metabolism pathways in adults. Our results indicate short-term metabolic flexibility in disease-free individuals, which could suggest the need for consideration of metabolic flexibility in disease states such as cardiometabolic diseases.

Notes

This research was supported by the National Institutes of Health: HD075784 (ADS, Eunice Kennedy Shriver National Institute of Child Health and Human Development); HL007745 (EAY, National Heart, Lung, and Blood Institute); S10 OD018006 (DPJ, Office of the Director). DPJ was supported in part by R01 ES023485, U2C ES030163, and P30 ES019776 (National Institute of Environmental Health Sciences). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH, the Eunice Kennedy Shriver National Institute of Child Health and Human Development, the National Heart, Lung, and Blood Institute, or the National Institute of Environmental Health Sciences.

Author disclosures: The authors report no conflicts of interest.

ADS is an Associate Editor on the Journal and played no role in the Journal's evaluation.

Supplemental Tables 1–5 and Supplemental Figures 1 and 2 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/jn/.

Abbreviations used: C18, 18-carbon; GSEA, gene set enrichment analysis; HILIC, hydrophilic interaction liquid chromatography; HMDB, Human Metabolome Database; INCAP, Institute of Nutrition of Central America and Panama; KEGG, Kyoto Encyclopedia of Genes and Genomes; LC-FT-MS, liquid chromatography Fourier transform mass spectrometry; MSI, Metabolomics Standards Initiative; NIST, National Institute of Standards and Technology; OGTT, oral-glucose-tolerance test; RT, retention time; SRM, standard reference material; TCA, tricarboxylic acid.