Advantages of Studying the Metabolome in Response to Mixed-Macronutrient Challenges and Suggestions for Future Research Designs

ABSTRACT

Evaluating the postprandial response to a dietary challenge containing all macronutrients—carbohydrates, lipids, and protein—may provide stronger insights of metabolic health than a fasted measurement. Metabolomic profiling deepens the understanding of the homeostatic and adaptive response to a dietary challenge by classifying multiple metabolic pathways and biomarkers. A total of 26 articles were identified that measure the human blood metabolome or lipidome response to a mixed-macronutrient challenge. Most studies were cross-sectional, exploring the baseline and postprandial response to the dietary challenge. Large variations in study designs were reported, including the macronutrient and caloric composition of the challenge and the delivery of the challenge as a liquid shake or a solid meal. Most studies utilized a targeted metabolomics platform, assessing only a particular metabolic pathway, however, several studies utilized global metabolomics and lipidomics assays demonstrating the expansive postprandial response of the metabolome. The postprandial response of individual amino acids was largely dependent on the amino acid composition of the test meal, with the exception of alanine and proline, 2 nonessential amino acids. Long-chain fatty acids and unsaturated long-chain acylcarnitines rapidly decreased in response to the dietary challenges, representing the switch from fat to carbohydrate oxidation. Studies were reviewed that assessed the metabolome response in the context of obesity and metabolic diseases, providing insight on how weight status and disease influence the ability to cope with a nutrient load and return to homeostasis. Results demonstrate that the flexibility to respond to a substrate load is influenced by obesity and metabolic disease and flexibility alterations will be evident in downstream metabolites of fat, carbohydrate, and protein metabolism. In response, we propose suggestions for standardization between studies with the potential of creating a study exploring the postprandial response to a multitude of challenges with a variety of macronutrients.

Introduction

Metabolic health is maintained by balancing caloric intake and usage through nutrient absorption, trafficking, utilization, and storage (1). Chronic alterations in this energy balance are often manifested by changes in cellular metabolism that result in insulin resistance (IR) and inflammation in many tissues, which is ultimately responsible for an increased risk of type 2 diabetes (T2D), cardiovascular disease, cancer, and a variety of other diseases (2). Differences in individual responses to increased nutrient intake are affected by age, sex, genetics, body composition, oxidative capacity, and additional phenotypic variations (3). Identifying individuals who are at risk of metabolic dysfunction and the underlying cause for variations in response between populations and individuals is important to selectively intervene to prevent progression to clinically manifested cardiometabolic disease. Most individuals in Western societies spend the majority of their lives in the fed state (∼18 h/d) (4), however, most physiological studies to assess metabolism in health and disease occur in the fasted state. Therefore, classifying the postprandial response to a nutrient challenge may provide more novel insights into metabolic risk than a fasted measurement (5).

Metabolomics, the quantification of low-molecular weight intermediates in metabolic pathways, is widely used to identify underlying differences in biochemistry between individuals and to classify risk of developing diseases. The application of metabolomics to fasting blood samples has been successful in identifying biomarkers that differ between obese and lean individuals (6) and metabolic pathways associated with future IR, T2D (7, 8), and cardiovascular disease (9). Increasingly, metabolomics has been used to quantify the dynamic response to a nutrient challenge, providing insights into whole-body metabolism. Single-macronutrient challenges, including the oral-glucose-tolerance test (OGTT) and oral-lipid-tolerance test (OLTT), have been used to investigate the metabolic response to carbohydrates, lipids, or protein (1, 10). The most common single-macronutrient challenge is an OGTT, in which a drink containing 75 g of glucose is administered after an overnight fast and the metabolic response is classified for the following 2 h. An OGTT triggers insulin-dependent processes as insulin secretion increases from pancreatic β-cells recruiting GLUT4 to the plasma membrane to increase cellular glucose uptake and storage in muscle (11), while at the same time repressing glucose production by the liver and to a lesser extent from the kidney and potentially the small intestine (12). These alterations are paired with a rapid inhibition of adipose tissue lipolysis (13) and muscle proteolysis (14). The response of the plasma metabolome to an OGTT has previously been reviewed (15), highlighting an acute increase of glycolytic intermediates, indicating increased use of glucose as fuel and a reduction in free fatty acid (FFA) utilization. Reduced FFA utilization is reflected in decreases in acylcarnitines (AC) (generated in the mitochondria during β-oxidation), observed by reductions in medium-chain AC, which dropped by 60–70% from the fasted state to 120 min of the OGTT (16).

During an OLTT, the metabolic response is classified for a longer period, usually 8 h, and metabolome alterations are contingent on the type of fatty acids (FA) (e.g. saturated or polyunsaturated) within the challenge. Blood concentrations of triglycerides and lipoproteins typically peak at 4 h postingestion (10). Two articles have reported changes in the metabolome in response to an OLTT, demonstrating dynamic changes in bile acids (17) and the influence of aerobic fitness level on the response of the lipidome in healthy adults (18).

Single-macronutrient challenges are important to classify specific metabolic pathways and physiological responses; however, they are not representative of a typical meal. A mixed-macronutrient challenge represents a meal, triggering a combination of postprandial responses due to the presence of glucose, lipids, and amino acids. These challenges are used to assess the response of multiple organs, such as the gut, liver, adipose, pancreas, vasculature, muscle, and kidney (19), creating a more comprehensive view of metabolic health and an application to daily life. Here, we present a critical review of the literature assessing the metabolome response to a mixed-macronutrient challenge, containing carbohydrates, lipids, and protein, exploring differences in the study designs and alterations in amino acids, glucose, lipids, and ketones. Factors that influence the response of the metabolome are highlighted, including obesity, T2D, and weight loss, and we discuss important metabolites to profile for the classifications of metabolic health. Lastly, we present suggestions for consistency within study designs with the potential of creating a study exploring the postprandial response to a multitude of challenges with a variety of macronutrients.

Section 1. Study Design and Mixed-Macronutrient Challenge Composition

Supplemental Figure 1A details the study selection process using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flowchart. Articles included in this review were identified using Medline, EMBASE, and Google Scholar databases. Database searches were conducted between April and August 2020. Key search terms included “metabolomics,” “lipidomics,” “postprandial response,” “mixed meal tolerance test,” “mixed macronutrient tolerance test,” “oral lipid tolerance test,” and “dietary challenge.” Initial abstract review inclusion criteria were studies conducted in humans using a dietary challenge and measuring endogenous metabolism. Articles were excluded if they solely measured exogenous metabolites representing food intake (n = 7). All abstracts reviewed were published in English. Assessment of titles and abstracts generated by the literature search were conducted (n = 146, JLL) and full texts were obtained for any eligible articles (n = 63, JLL). Full texts were excluded in the final review due to dietary challenges not containing all macronutrients (n = 16, JLL), studies not conducting metabolomics (e.g. profiling lipid panel in response to a challenge) (n = 6, JLL), and studies conducted in infants using breastmilk and formula feeding (n = 1, JLL). The screening of full texts was checked and confirmed (CFB). Twenty-six articles were identified that measured the response of the human blood metabolome or lipidome to a mixed-macronutrient challenge (5, 20–44), briefly described in Table 1. The median number of participants is 29, with a range of 1 to 533 participants. The age range of the participants is 15–71 y. Full study details are provided in Supplemental Table 1.

TABLE 1.

Mixed-macronutrient challenge study designs. Reviewed studies are sorted by the percentage of carbohydrates within the dietary challenge. Study IDs correspond to the percentage of kilocalories from carbohydrates. Studies received multiple IDs if they used >1 mixed-macronutrient challenge. Each macronutrient challenge is detailed, including the type of food, the size of the challenge measured by challenge, kilocalories/total daily, kilocalories (kilocalories ratio), or a predetermined size of the meal (kilocalories average), and the ratio of kilocalories from each macronutrient to total kilocalories within the challenge (carbohydrate, protein, and lipid ratio)

Study ID	Reference number	Mixed macronutrient challenge	Kilocalories ratio	Kilocalories average	Carbohydrate ratio	Protein ratio	Lipid ratio
1	(43)	Banana	0.15	n/a1	0.94	0.05	0.01
2a	(20)	300 mL, Nutridrink juice style, fat-free, Nutricia, Poland	n/a	450	0.89	0.11	0
3a	(26)	Refined rye bread	n/a	250	0.80	0.08	0.12
4	(33)	Refined wheat bread	n/a	283	0.71	0.13	0.16
3b	(26)	Refined wheat bread	n/a	283	0.71	0.13	0.16
5	(35)	237 mL of Ensure original therapeutic nutrition	n/a	250	0.64	0.14	0.22
3c	(26)	Whole-meal rye bread	n/a	315	0.64	0.14	0.22
6	(38)	Ensure plus	0.3	n/a	0.57	0.15	0.28
7a	(28)	Vegan—tea with oat-milk, soy-based yogurt with blueberries, 2 rye bread sandwiches (1 with lentil mash and green pepper topping and 1 with cashew butter), and banana	n/a	500–750	0.56	0.14	0.29
8	(21)	Indian breakfast—rice cake, chutney, milk tea, skimmed milk, table sugar, coconut oil	0.25	n/a	0.55	0.15	0.3
9a	(32)	Cod served with pasta, sauce, vegetables, and a cinnamon bun	n/a	717 (male) and 538 (female)	0.52	0.2	0.28
9b	(32)	Beef served with pasta, sauce, vegetables, and a cinnamon bun	n/a	717 (male) and 538 (female)	0.52	0.2	0.28
10	(39)	McDonald's Big Mac meal (Big Mac hamburger, 100 g French fries, and 400 g sucrose-sweetened Coca-Cola)	n/a	979	0.5	0.13	0.37
12	(22)	Liquid mixed meal (400 mL)	n/a	600	0.5	0.16	0.34
11a	(5)	Standard liquid diet—Fresubin energy drink, chocolate flavor	0.33	n/a	0.5	0.15	0.35
13	(23)	Standard meal	n/a	522	0.49	0.19	0.32
14a	(29)	Cereal breakfast—orange juice, oat puffs with milk, and a rye bread sandwich with hard cheese and fresh tomato	n/a	500–750	0.48	0.16	0.36
14b	(29)	Egg and ham breakfast sandwich—orange juice, scrambled eggs, white beans in tomato sauce, fried pork loin, tomato, and toasted white bread with orange marmalade	n/a	500–750	0.48	0.16	0.36
2b	(20)	360 mL, Cubitan, Nutricia, Poland	n/a	450	0.45	0.3	0.25
7b	(28)	Lacto-ovo vegetarian—tea with milk, fruit yoghurt (raspberry/rhubarb), 2 rye bread sandwiches (1 with cottage cheese and apple and 1 with hard cheese and tomato)	n/a	500–750	0.45	0.18	0.37
15	(40)	Two muffins and 300 mL 0% fat milk	n/a	1100	0.44	0.1	0.46
16	(30)	Cocoa-flavored drink (cocoa, corn oil, heavy cream, sucrose, and skim milk), cereal, banana, and skim milk. Drink included a palmitate tracer	0.38	n/a	0.43	0.07	0.5
17	(31)	Fast-food breakfast representative of the Western diet—2 sausage egg English muffins, 1 apple turnover, and ∼370 mL of concentrated orange juice	n/a	1330	0.42	0.13	0.45
7c	(28)	Omnivore—tea with milk, boiled egg, and caviar, 2 rye bread sandwiches, 1 topped with ham and red pepper, and 1 with liver pâté and cucumber	n/a	500–750	0.42	0.18	0.4
18	(44)	Oral lipid tolerance test	n/a	1080	0.4	0.13	0.47
19	(37)	Blended beverage—1 cup nonfat lactose-free milk, 1 cup low-fat vanilla-flavored yogurt (1% fat), 30 g 100% whey chocolate-flavored protein powder, 118 g banana, 22 g flax seed oil, 3 g borage oil, 3.6 g soy lecithin, and 3 g fish oil	0.4	n/a	0.4	0.19	0.41
20a	(24)	Substantial meal	n/a	1098	0.39	0.23	0.38
20b	(24)	Substantial meal	n/a	1323	0.38	0.2	0.42
21	(36)	Phenflex test (400 mL)—60 g palm olein, 75 g glucose, and 20 g dairy protein	n/a	950	0.34	0.09	0.57
22	(41)	Drink (400 mL)—320 mL tap water, 75 g glucose, 60 g palm oil, 20 g Protifar as a milk protein concentrate (Nutricia, Utrecht, The Netherlands), and 0.5 g artificial vanilla aroma	n/a	940	0.33	0.08	0.59
23	(27)	500 mL dairy shake—300 mL custard, 150 mL cream cheese, and 50 mL whipping cream	n/a	706	0.30	0.12	0.58
11b	(5)	3 parts Fresubin energy drink and 1 part Calogen (Nutricia, Zoetemeer, The Netherlands), long-chain triglyceride emulsion	n/a	35 g fat/m2	0.25	0.08	0.67
24b	(25)	Soy oil-based meal—cheddar-flavored soy cheese (100 g), soy beverage (300 mL), a nondairy spread (20 g), and toast (50 g)	n/a	790	0.24	0.15	0.61
24a	(25)	Dairy meal—cheddar cheese (60 g), butter (20 g), and extra creamy whole milk (300 mL) with toast (50 g)	n/a	750	0.19	0.16	0.65
25	(42)	Milkshake (500 mL)—53% whipping cream, 3% sugar, and 44% water	n/a	153 (per 100 g)	0.08	0.04	0.88
26a	(34)	SFA shake—low-fat yogurt, low-fat milk, strawberry flavor, 95 g palm oil	n/a	987	0.08	0.04	0.88
26b	(34)	MUFA shake—low-fat yogurt, low-fat milk, strawberry flavor, 95 g high oleic acid sunflower oil	n/a	987	0.08	0.04	0.88
26c	(34)	PUFA shake—low-fat yogurt, low-fat milk, strawberry flavor, 40 g palm oil, 55 g Marinol D-40s (40% DHA)	n/a	987	0.08	0.04	0.88

¹Not applicable.

Study objectives

Twelve studies (46%) recruited healthy subjects without obesity or impaired glucose tolerance (5, 20, 24, 26, 28, 29, 32, 33, 36, 37, 43, 44). Two of these studies only recruited male healthy subjects (5, 24) and 2 only recruited female healthy subjects (26, 33). Eleven studies (42%) used case-control study designs to assess differences in the postprandial response to a dietary challenge between cases, predominantly with metabolic disease or obesity, and healthy controls (20–22, 31, 34, 35, 38–40, 42, 44). Two studies (8%) assessed the postprandial response to a dietary challenge within overweight and obese participants, without a healthy control group (27, 30). Eight studies (31%) used multiple challenges to assess differences in the postprandial response of the metabolome to variations in type of macronutrient load (5, 20, 25, 26, 28, 29, 32, 34). Five studies (19%) classified the postprandial response of the metabolome before and after a short-term dietary intervention (4–12 wk, with n = 1 intervention of 1.5 y) (24, 27, 32, 42, 43). Four studies (15%) classified the postprandial response of the metabolome before and after weight loss via dietary interventions (35, 40, 41) or surgical interventions (23). Study objective breakdown is classified in Supplemental Figure 1B.

Variation in macronutrient composition and total kilocalories in dietary challenges

Carbohydrates, protein, and fat trigger different metabolic responses, emphasizing the importance of reporting the composition of the challenge. The ratio of caloric intake from each macronutrient to total caloric intake is shown (Figure 1A, Table 1). Large differences were observed in macronutrient ratios, reporting ranges of ratios for carbohydrates (0.08 to 0.94), fat (0.01 to 0.88), and protein (0.04 to 0.23). Total grams of macronutrients within the challenge are reported (Figure 1B, Table 1). Combined, Figure 1A and B shows the variation in the composition of macronutrients in each nutrient challenge, each challenge resulting in a varied response of the metabolome. Seventeen of the reviewed studies (65%) provided further details on the nutrient composition within the dietary challenge (5, 20, 24–32, 34–38, 41), such as type of FFA and micronutrient composition. To precisely measure a meal's nutrient composition, Badoud et al. (31) assessed the amino acid and fatty acid composition of the study's high calorie meal using MS and GC.

FIGURE 1.

Macronutrient content within mixed-macronutrient challenges. (A) Ratio of kilocalories from each macronutrient to total kilocalories within the challenge. (B) Total grams of macronutrients within dietary challenges. Multiple dietary challenges within a study are reported separately. Study IDs correspond to the percentage of kilocalories from carbohydrates (Table 1).

The number of kilocalories within the mixed-macronutrient challenges varied between studies. Six studies (23%) determined the size of the challenge via a percentage of each participant's daily caloric intake, ranging from 15 to 40% (5, 21, 30, 37, 38, 43) (Figure 2A, Table 1). Sixteen studies (62%) determined the size of the challenge using a fixed number of kilocalories for all participants (20, 22–27, 31, 33–36, 39–41, 44). Total kilocalories in these studies ranged from 250 to 1330 kcal (Figure 2B, Table 1). The remaining 4 studies used specific formulas or ranges for determining the challenge's caloric intake. Schmedes et al. (32) altered the number of kilocalories provided to males and females, 717 kcal and 538 kcal, respectively. Both studies by Rådjursöga et al. (28, 29) allowed participants to decide between consuming a “small” and “large” meal, 500 kcal and 750 kcal, respectively. Uniquely, Krug et al. (5) accounted for each participant's body surface area when determining the amount of liquid challenge to provide. Variations in the total macronutrient load may be due to different study objectives for the challenge (e.g. overloading the biological system compared with providing a typical meal), nevertheless, this presents challenges in comparing results between studies.

FIGURE 2.

Caloric content within mixed-macronutrient challenges. Size (kilocalories) of challenge was determined using (A) a ratio of the daily kilocalories required for subjects or (B) a predetermined size of the meal. Multiple dietary challenges within a study are reported separately. Study IDs correspond to the percentage of kilocalories from carbohydrates (Table 1).

Presenting the mixed-macronutrient challenge as a liquid or solid meal

Mixed-macronutrient challenges were administered as a solid meal (58%) or a liquid shake (42%). This variability in meal challenge choice may be due to outcomes of interest, however, it presents another potential difference in metabolomic response due to differences in rates of macromolecules entering the small intestine, gastric emptying, and possibly nutrient flux (45). Multiple meals were designed to be representative of a Western diet (31, 39) or local/regional diet (21, 24). Small variations in solid meals were done to test a nutrient or macronutrient's influence on the postprandial response, such as altering the source of protein (25, 26, 28, 29, 32). Many researchers did not provide rationale for their solid meal composition (23, 30, 33, 40, 43, 44). Six studies used commercially available liquid products for their dietary challenge (5, 20, 35, 36, 38, 41), ensuring standardization between participants and validation in future studies. One example of a commercially available dietary challenge is the PhenFlex test (36), which has been proposed as a standardized optimal nutritional challenge test due to its interaction with metabolic processes distributed over multiple organs, including the gut, liver, adipose tissue, pancreas, muscle, and kidney (10). Several liquid shakes were specifically created for studies (22, 27, 34, 37, 42), including multiple studies that used a mixture of whipping cream, sugar, and water to create a milkshake. Lastly, to alter the fatty acid profile, studies incorporated varieties of oils into their dietary challenges, including palm oil (34, 36, 41), coconut oil (21), corn oil (30), sunflower oil (34), fish oil (34, 37), and borage oil (37).

Preparation for the challenge using a standard meal and overnight fast

Ten studies (38%) (5, 25, 27–31, 34–36) incorporated a standardized meal the night before the nutrient challenge, supported by evidence that standardized diets normalize interindividual differences in the metabolome (46). The standard meal by Krug et al. (5) was adjusted for the study participant's resting metabolic rate (RMR) multiplied by a factor of 1.3 for low physical activity. Six studies provided weight-maintenance standardized meals (25, 28–31, 35). All studies required the participants arrive in the fasted state for the nutrient challenge, as there are differential metabolic responses when consuming a meal in the fasted or fed state (47). Most studies used an overnight fast, ranging from 10 to 15 h. The study design by Pellis et al. (27) had the shortest fast before the dietary challenge (4 h). Participants arrived after an overnight fast and consumed a light standardized breakfast. After a minimum of a 4-h period without food and drinks, participants began the macronutrient challenge (27). The study design by Ramos-Roman et al. (30) had the longest fast before the dietary challenge—18 h. The authors discuss the reason for the extended fast being to bring the subjects to a moderate rate of lipolysis, as observed in previous work (48). The study design by Mathew et al. (24) was unique as participants fasted for 13–18 h during the day for Ramadan before receiving their nutrient challenge, rather than overnight, potentially introducing the impact of circadian rhythm alterations in the metabolome (49).

Section 2. Variations in Metabolomics Methodology

Blood samples (plasma or serum) were collected at various times after the mixed-macronutrient challenge, with samples collected ≤480 min after challenge (Figure 3) with all studies collecting a fasted blood sample. The most frequent blood sampling occurred at 60 min (69%) (5, 20, 21, 23, 25–27, 30, 32, 33, 35–37, 39–43) and 120 min (65%) (5, 20, 21, 23–25, 27, 30–32, 34–36, 39–42). Six studies analyzed the plasma or serum samples using a metabolomics platform at fasting and only 1 postprandial time point, limiting the ability to interpret variations in a wide range of metabolite classes with differential responses (22, 24, 28, 29, 31, 44).

FIGURE 3.

Blood sampling time course. Number of studies that collected blood samples at each time point (min).

Each study reviewed incorporated a variety of untargeted and targeted metabolomics platforms to measure the postprandial response. Untargeted metabolomics, often described as a “discovery-based” tool, uses 2 main analytical platforms. The first platform uses GC or LC to separate the metabolites, typically by polarity, and is paired with a mass spectrometer for ionization and detection. Six of the articles reviewed used chromatography and MS to quantify the untargeted metabolome (20, 21, 33, 39, 43, 44). Untargeted metabolomics platforms have been adapted and modified to quantify the lipidome, the complete lipid profile. Three manuscripts used chromatography and MS for untargeted lipidomics (25, 32, 37). Proton NMR, the second platform for untargeted metabolomics, was used by 7 of the studies (5, 23, 28, 29, 32, 33, 43). In contrast, targeted metabolomics focuses on measuring specific metabolites and groups of metabolites with the capacity for absolute quantification when comparing to internal standards. The most frequently used targeted metabolomics platforms quantified amino acids (5, 22, 24, 26–29, 31, 35, 36, 38, 40, 41), FFAs (27–29, 31, 35, 41, 42), and ACs (5, 22, 24, 26, 30, 35, 38, 40, 41) (Supplemental Table 1).

Section 3. Exploring the Postprandial Response of the Metabolome in Healthy Participants

The following section reports the response of the metabolome to the mixed-macronutrient challenge in healthy individuals; those without obesity or impaired glucose tolerance. Results were synthesized from studies that recruited healthy subjects (5, 24, 26, 28, 29, 32, 33, 36, 37, 43) and studies using healthy controls (20–22, 31, 34, 35, 38–40, 42, 44). Metabolites summarized were grouped by 4 types of postprandial responses—amino acid metabolism, glucose metabolism, lipid metabolism, and ketogenesis—as suggested by Shaham et al. (50) and reported in Fazelzadeh et al. (40). In line with Pellis et al. (27), metabolites are classified by their speed of postprandial response; rapid (maximum change within 1–2 h) and slow (maximum changes after 6 h) and whether their response is decreasing or increasing.

Amino acids

Amino acids were the most widely profiled metabolite group in healthy individuals, utilizing both targeted and untargeted metabolomics platforms. Blood concentrations of amino acids tended to peak 60–90 min after the test meal (24, 26, 31–33, 35, 38–40, 43), returning to baseline by 240 min after the meal (26, 32, 33, 38, 40, 43). During a hyperinsulinemic-euglycemic clamp, skeletal muscle proteolysis is reduced in healthy individuals (51), which is reflective of the influence of insulin on muscle proteolysis during the postprandial response. The postprandial response of individual amino acids is largely dependent on the amino acid composition of the test meal and the gastric emptying rate. To support this, Bos et al. (52) used ¹⁵N-labeling of dietary protein to assess the appearance of amino acids in the blood after the ingestion of specific dietary proteins, observing plasma postprandial patterns mainly reflect differences in digestion kinetics and a potential influence by the actual dietary amino acid composition. Badoud et al. (31) suggested that the amino acid postprandial response is not entirely explained by the abundance of amino acids in the test meal. Leucine, glutamic acid, and proline were the most abundant amino acids in the test meal, however, proline and alanine seemed to elicit the largest postprandial response (31). This suggests that the dietary amino acid profile may be modified during digestion, absorption, and transportation, supported by previous work observing similar dietary and plasma profiles of essential amino acids, with discrepancies between the dietary and plasma profiles of nonessential amino acids (53).

Branched-chain amino acids (BCAAs) (leucine, isoleucine, and valine), aromatic amino acids (tyrosine, phenylalanine, and tryptophan), glycine, alanine, proline, and methionine were consistently elevated at 60–90 min, returning to baseline at later time points (26, 32, 33, 35, 38, 40, 43). Interestingly, aspartate and glutamate immediately decreased in response to the mixed-macronutrient challenge, observed in 2 challenges, both with a high percentage of kilocalories from carbohydrates (64% and 71%) (33, 35). One potential explanation for the drop of glutamate is an increased flux of glutamate into the tricarboxylic acid (TCA) cycle as an anaplerotic substrate to support carbohydrate oxidation. The entry of glutamate into the TCA cycle forming α-ketoglutarate is regulated by the enzyme glutamate dehydrogenase (GDH). In the switch from the fasted to fed state, positive regulators activate GDH to support glutamate flux into the TCA cycle (54).

Glucose metabolism

During an OGTT in a healthy individual, glucose rapidly increases during the first 30 min, dropping to baseline by 120 min due to clearance by the liver, supported by the rise in insulin. Using metabolomics, this rise and fall of glucose was observed (5, 21, 26, 31, 32, 35, 38, 40), demonstrating an initial insulin response to a range of carbohydrates in a mixed-macronutrient load. The challenge by Kardinaal et al. (42) that was designed to mimic an OLTT did not produce a rise in glucose concentrations, potentially because only 8% of the kilocalories were from carbohydrates. This suggests a minimum amount of carbohydrates within a mixed-macronutrient meal may be required to elicit a rise in glucose and the glucose-induced response of insulin.

Several studies profiled pyruvate and lactate, demonstrating an increase in these glycolytic products indicating the utilization of glucose via glycolysis (32, 33). Other studies profiled TCA cycle intermediates (27, 32, 33, 35), observing vast differences in intermediate responses, as the TCA cycle represents the intersection between amino acid, carbohydrate, and lipid metabolism. For instance, Schmedes et al. (32) observed a postprandial increase in citrate (at 30, 60, 240, 360 min), Shrestha et al. (33) observed a postprandial decrease in citrate (at 180 min), and Thonusin et al. (35) observed no postprandial change in citrate. The 3 studies used mixed-macronutrient challenges with a wide range of kilocalories [538–717 kcal (32), 283 kcal (33), and 250 kcal (35)] and range of percentage kilocalories from carbohydrates [52% (32), 71% (33), and 64% (35)]. These results may suggest differential anaplerotic reactions refueling the TCA cycle depending upon the challenge composition. Elevated concentrations of plasma citrate may indicate increased levels of TCA cycle flux, as evidenced by a more robust response to a higher caloric load in the mixed-macronutrient challenge (32). Increased concentrations of citrate may inhibit phosphofructokinase, a major regulation point in glycolysis, under conditions that favor glycogen storage (55). The decline in citrate, observed by Shrestha et al. (33), was hypothesized to be due to the formation of malonyl CoA or reduced TCA cycle flux. To fully assess the kinetics and flux behind these study discrepancies, isotope tracers should be considered in the mixed-macronutrient challenge.

Lipid metabolism

Multiple studies profiled FFAs in healthy participants (24, 27–29, 31, 35, 41–43). During fasting, FFAs are released from adipose tissue via lipolysis, increasing their concentrations in the plasma. All studies observed a decrease in FFAs between baseline and 120 min after the mixed-macronutrient challenge, representing the insulin-induced suppression of lipolysis in the fed state (56). Studies with longer biological sampling observed FFAs returning to baseline 180 to 240 min after the meal (27, 42, 43). While studying overweight, healthy individuals, Pellis et al. (27) observed differences in the response of FFAs dependent on their chain length and number of double bonds. Medium-chain FFAs (FA 10:0, 12:0, and 14:0) rapidly increased in response to the challenge, due to their direct entry into the plasma rather than being re-esterified into triglycerides (57). Long-chain FFAs (FA 16:0, 16:1, 17:0, 18:0, 18:1, 18:2, and 20:4) began to increase after 180 min, highlighting the switch from the fed back to the fasted state. Interestingly, DHA (C22:6), an essential ω-3 FFA, remained decreased for a longer period in response to the dietary challenge, potentially suggesting specific metabolism of FFA depending on the number of double bonds (27).

Widely profiled in the challenges (5, 22, 24, 26, 33, 35, 38), ACs are carnitine esters derived from FFAs or amino acids that enter the mitochondria. As with FFAs, the postprandial response of ACs is dependent on the chain length and number of double bonds. Medium-chain and long-chain ACs decreased in response to the dietary challenge through 120–180 min (5, 22, 24, 33, 35, 38). This pattern is representative of a decrease in FFAs entering the mitochondria due to increases in postprandial glucose and amino acid metabolism for energy. Thonusin et al. (35) observed discrepancies in long-chain AC responses, as AC 16:0 and AC 18:0 did not decrease in response to the test meal, emphasizing the stability of plasma saturated and long-chain ACs (58). Short-chain AC metabolites of BCAAs, such as AC 3:0 and AC 5:0, tended to increase in response to the dietary challenge (24, 35), though some of these short-chain ACs may be derived from branched-chain fatty acids (BCFAs) (59). To model β-oxidation flux through the mitochondria, investigators used a ratio of a shorter-chain AC to a longer-chain AC (5, 40), with elevations in their fasting concentrations in adults related to IR and cardiovascular disease (60). Fluctuations in AC ratios represent switches between anabolic and catabolic states, as a decrease in the AC ratio symbolizes less FFA flux through β-oxidation. Krug et al. (5) suggests that AC ratios provide a more accurate description of β-oxidation, accounting for individual differences in absolute plasma AC concentrations. These ratios are used to determine how IR and adiposity influence FFA flux through β-oxidation, which will be further described in Section 4.

Accompanying the advancements of bioinformatics and instrumental analysis, 2 studies profiled the lipidome to classify the postprandial response of lipid classes, such as ceramides, diacylglycerols, lysophospholipids, phospholipids, sphingomyelins, and triacylglycerols (37, 39). The majority of lipid classes rapidly decreased in response to the mixed-macronutrient challenge (39). Despite the small sample size, Zivkovic et al. (37) quantified variations in lipid class postprandial response across individuals, conducting the mixed-macronutrient challenge 3 times in each individual. They observed that interindividual variation in the lipidome response was greater than intraindividual variation, observing stark interindividual differences in the responses of linoleic acid (18:2) and palmitoleic acid (16:1) in the triglyceride fraction and α-linolenic (18:3) in the phosphatidylcholine (PC) fraction (37). Other studies classified specific lipids within targeted or untargeted metabolomics platforms (20–22, 24, 26, 35, 44). Overall, a macronutrient challenge elicits an insulin-induced suppression of lipolysis resulting in lower concentrations of lipids, however, current studies are unable to decipher postprandial response differences between lipid classes due to limitations in the data generated from the studies.

Ketogenesis

The plasma concentrations of ketone bodies—acetoacetate, 3-hydroxybutyrate, and acetone—are produced in the liver with increased availability of FFAs in the face of lower insulin/increased glucagon concentrations. Most studies observed a rapid drop in ketone metabolites, paralleling FFA changes, which began to increase at 180–240 min postchallenge (27, 28, 33, 43). The postprandial decrease in ketone bodies is due to the inhibition of ketogenesis by insulin and the decreased availability of FFAs due to the inhibition of lipolysis. However, Thonusin et al. (35) and Fazelzadeh et al. (40) both observed a rise in acetoacetate in response to the dietary challenge (15 and 60 min, respectively), even though both mixed-meal challenges had a sizeable percentage of kilocalories from carbohydrates [64% (35) and 44% (40)].

Section 4. Metabolic Disease Influences on the Postprandial Metabolome Response

To expand upon the classification of the postprandial response of the metabolome to a mixed-macronutrient challenge in healthy subjects, we reviewed alterations in metabolite response due to obesity and metabolic diseases, providing insight on how weight status and disease influence the ability to cope with a nutrient load and return to homeostasis. We reviewed studies comparing the response of the metabolome between controls and individuals with obesity (30, 31, 34, 35, 38–40) and individuals with impaired fasting glucose (IFG) and/or T2D (21, 22, 31, 42). Additionally, we reviewed studies assessing if dietary intervention-based weight loss influences the trajectories of metabolites in response to a mixed-macronutrient challenge (40, 41).

Influence of obesity

As obesity (BMI ≥30 kg/m²) is associated with changes in the fasting metabolome (6), multiple studies have classified the influence of obesity on the postprandial response to a mixed-macronutrient challenge (30, 31, 34, 35, 38–40). These challenges used mixed-macronutrient meals consisting of a range of percentage kilocalories from carbohydrates (8–64%), lipids (22–88%), and protein (4–15%). Only 1 study specified recruiting nondiabetic obese/overweight individuals (30). In fasting plasma samples, obesity is associated with elevations in lipids, branched-chain and aromatic amino acids, and ACs (61, 62), signifying the overload of amino acid and FA flux into the mitochondria. Studies recruiting both obese and lean individuals observed patterns in the fasting metabolome mimicking previous studies (61, 62), with elevations in BCAAs (39, 40), carbohydrate metabolites (39), and ACs (40) in adults with obesity.

Comparing obese and lean individuals consuming a 1100 kcal meal challenge (44% carbohydrates, 46% fat), Fazelzadeh et al. (40) observed 19 metabolites differing at fasting and 8 metabolites with differential postprandial responses, potentially suggesting the metabolome postprandial response is tightly controlled irrespective of obesity. Metabolites with a differential postprandial response included alanine, proline, methylmalonic acid, threonine, histidine, methionine, phosphocholine, and 2-hydroxyisovalerate (directionality not specified) (40). In response to the consumption of an Ensure PLUS shake (57% carbohydrates, 28% fat), Bastarrachea et al. (38) observed a larger postprandial increase of many amino acids in lean individuals, compared to individuals with obesity, however, these findings will need to be replicated as they conducted a small pilot study (n = 8 lean, 8 obese). Bondia-Pons et al. (39) used a unique study design profiling differences in the response to a McDonald's Big Mac meal challenge (50% carbohydrates, 37% fat) between twins that were discordant for BMI, allowing for the assessment of the contribution of obesity, independent of genetic liability. Results from this study revealed converging and diverging postprandial patterns within lean and heavy twins. The converging pattern showed significantly higher concentrations in heavy twins compared with lean twins at fasting including 3-hydroxybenzoic acid, leucine, and oleic acid. The diverging pattern consisted of metabolites that were not differential at fasting and diverged by 120 min postprandial, with lower concentrations of bile acids, such as cholic acid and glycine lithocholic acid, in heavy twins compared with lean twins. Bondia-Pons et al. postulated that differences in the distal small intestine and colon microbiota between lean and heavy twins may influence bile acid metabolism in response to the Big Mac meal challenge (63). No other studies accessed bile acid response in the context of obesity.

Several studies assessed the influence of obesity on AC postprandial response (30, 35, 38), as the transition from the fasted to fed state is characterized by a decrease in FFA metabolism, which may be influenced by obesity-associated adipose tissue expansion, adipocyte capacity for lipid storage versus lipolysis, and changes in mitochondrial fatty acid oxidation rates (64). In overweight and obese individuals, Ramos-Roman et al. (30) quantified fasting and postprandial ACs using fatty acid stable isotope tracers (either d₃₁ or ¹³C₁₆ palmitate potassium salt) and identified subject characteristics influencing AC concentrations, including lean body mass (LBM), fat mass, glucose oxidation, and fatty acid oxidation rates. At fasting, they observed that LBM was positively associated with short-chain ACs (AC 3:0, AC 4:0, AC 5:0, and AC 5:1), however, fat mass was not associated with any fasting AC metabolites. Using indirect calorimetry, they observed that fasting fat oxidation rates are positively associated with long-chain saturated and unsaturated ACs (AC 12:1, AC 14:0, AC 14:1, AC 14:2, and AC 16:1), suggesting that at fasting, those with lower AC concentrations have the lowest fat oxidation rates. In response to the meal, saturated long-chain ACs (AC 14:0, AC 16:0, and AC 18:0) did not change postprandially, however, the unsaturated ACs decreased significantly (21–46% decrease). These results were mirrored by Thonusin et al. (35) finding that saturated long-chain ACs (AC 14:0, 16:0, and 18:0) did not have a postprandial response to the nutrient challenge, however, unsaturated ACs significantly decreased. These results suggest that variations in AC concentrations are dependent on the chain length and saturation, even though both saturated and unsaturated FFAs decrease in response to the nutrient challenge. Lastly, Ramos-Roman et al. (30) identified subject characteristics influencing the postprandial response of ACs. An individual's minimum concentration of long-chain ACs (nadir long-chain AC) is positively associated with the minimum concentration of FFA (nadir FFA) in response to the nutrient challenge, independent of fat mass levels. These results may suggest that the lack of suppression of lipolysis is associated with the postprandial response of ACs, not necessarily the amount of adipose tissue.

Influence of weight loss

Two studies analyzed differences in the response to a mixed-macronutrient challenge before and after weight loss interventions (40, 41). In the study by Fazelzadeh et al. (40), abdominally obese male subjects (n = 29, baseline BMI = 30.3) received a mixed-macronutrient challenge prior to and after a weight-loss intervention consisting of an 8-wk very low energy diet. The average weight loss during the intervention was 9.3 kg. At fasting, 3 metabolites differed before and after the weight-loss intervention, including glycine, which was higher with weight loss, and creatinine and glutamic acid, which were lower with weight loss. Weight loss influenced the postprandial response of 11 metabolites, including glutamine, histidine, creatine, pyroglutamic acid, glucose, choline, and 5 oxylipins derived from the oxidation of arachidonic acid (directionality not specified). After weight loss, there was an improvement in the HOMA-index, however, no differences were seen in AC ratios. In the study by Fiamoncini et al. (41), overweight, metabolically healthy individuals (n = 70, baseline BMI = 29.7) were comprehensively phenotyped and received a mixed-macronutrient challenge. Two distinct groups (Metabotype A and B) were classified based on fasting levels and the response of plasma metabolites to the challenge within metabolic pathways such as lipolysis, FA β-oxidation, and ketogenesis. Metabotype A consisted of subjects with lower fasting FFAs and lower medium-to-long-chain AC ratios. Individuals in Metabotype A had a lower glucose peak, less insulin secreted, and a stronger suppression of lipolysis and fatty acid oxidation in response to the challenge, demonstrating that Metabotype A is more metabolically healthy than Metabotype B. All individuals were stratified into a weight-loss intervention (n = 40) with reduced caloric intake by 20% and a no-weight change control group (n = 30). The average weight loss in the intervention group was 5.6 kg. Only a few significant differences in the metabolome were observed between the weight-loss and control groups (not described). However, there were major differences in the effect of the intervention between Metabotype A and B. After the intervention, individuals in the Metabotype B, the metabolically unhealthy group, had greater improvements in their glycemic response and decreases in BCAAs and short- and medium-chain ACs, with concentrations converging to individuals in Metabotype A. The study results indicate that classifying the response to a mixed-macronutrient challenge may reveal distinct metabolite profiles predicting response to weight loss.

Influence of IFG and T2D

Multiple studies profiled the postprandial response of the metabolome in the context of IFG and T2D (21, 22, 31, 42), using challenges consisting of a range of percentage kilocalories from carbohydrates (8–55%), lipids (30–88%), and protein (4–16%). Across all studies, individuals with IFG and T2D had a heightened glucose postprandial response to the mixed-macronutrient challenge (21, 22, 31, 42) as anticipated from what is seen in OGTTs. Using an untargeted metabolomics platform, Kumar et al. (21) observed differences in metabolites that changed postprandially (60 and 120 min), finding that less metabolites responded to the nutrient challenge in the prediabetic group (96 metabolites) compared with the overweight group (157 metabolites) and controls (154 metabolites). These results may suggest that individuals who have prediabetes already have marked alterations in postprandial metabolic pathways, prior to advancing to T2D. The authors do not elaborate on the classes of metabolites that respond to the nutrient challenge in controls but not prediabetics, therefore limited biological interpretations can be drawn.

In a similar study design, Li-Gao et al. (22) explored the ability of the postprandial response to distinguish T2D and normal glucose tolerance (NGT) individuals. Using lasso regression, the authors initially identified the most parsimonious set of metabolites at fasting (t0) (12 metabolites), the postprandial time point (t150) (4 metabolites), and the response [log(t150)-log(t0)] (16 metabolites). These 28 metabolites were further compared between T2D and NGT individuals at each time point. At fasting, 9 of the 12 metabolites were differential, including higher concentrations of AC 16:1, AC 4:1, tyrosine, and leucine in T2D and lower concentrations of glycine, lysophosphatidylcholine (LPC) 18:1, LPC 17:0, PC 44:4, and PC 44:5 in T2D. At the postprandial time point, all 4 of the selected metabolites were differential, including higher concentrations of AC 16:1 and AC 4:1 in T2D and lower concentrations of glycine and LPC 17:0 in T2D. Lastly, in the response profile, only 4 of the 16 selected metabolites were differential, including an elevated response of AC 10:0 in T2D and a decreased response of methionine, serine, and valine in T2D. Li-Gao et al. concludes that the fasting and postprandial metabolome provides similar distinguishing power between NGT and T2D individuals, however, these results illustrate an unfinished picture. Further work should apply an untargeted metabolomics platform, rather than a 163 metabolite targeted platform, and analyze the metabolome across multiple time points postprandially.

Two studies assessed the switch from a catabolic to anabolic state following a mixed-macronutrient challenge in metabolically healthy and unhealthy individuals (31, 42). Badoud et al. (31) classified obese individuals as metabolically healthy (MHO) and metabolically unhealthy (MUO), observing that MUO individuals have elevated glucose and insulin AUC in response to an OGTT. Using a carnitine-to-AC ratio, they were able to estimate FA β-oxidation flux and the ability to switch from a catabolic to anabolic state, as the catabolic state is reflected by higher concentrations of ACs. They compared the postprandial percent change [(120 min – 0 min/0 min) × 100] of the carnitine-to-AC ratio between MUO, MHO, and controls. They observed the greatest percentage change in the controls, within a significantly diminished percentage change in MHO, and an intermediate percentage change in MUO. These results suggest a greater switch from a catabolic to anabolic state in MUO and MHO individuals. In parallel, Kardinaal et al. (42) assessed fat and carbohydrate oxidation rates using indirect calorimetry during the mixed-macronutrient challenge. They demonstrated that individuals with metabolic syndrome have a lower respiratory quotient (RQ) compared with controls in response to the liquid milkshake (180 min). Although they did not profile ACs, the authors hypothesize that diminishing metabolic health and the progression to T2D is marked with reduced metabolic flexibility in response to a meal (1).

Section 5. Discussion

Twenty-six articles were identified that measured the response of the human blood metabolome to a mixed-macronutrient challenge (Supplemental Figure 1) (5, 20–44). The metabolome is plastic and fluctuations within glycolytic, lipid, proteolytic, and ketogenic metabolic pathways occur in response to a mixed-macronutrient challenge. Changes in postprandial concentrations and the response of metabolites are dependent on the caloric content and macronutrient distribution of the challenge and may be influenced by metabolic diseases, such as obesity and T2D. This critical review demonstrates the promise in studying the postprandial response to a nutrient challenge, however, sheds light on the obstacles the field must overcome to understand the physiology.

Postprandial response of the metabolome in the context of metabolic health

Conclusions from this review highlight metabolites with differential responses to a mixed-macronutrient challenge between metabolically healthy and unhealthy individuals, providing insights on altered substrate flux, metabolic flexibility (65), and future disease risk. We have identified potential gaps in the current literature, such as a limited number of comparable studies assessing the postprandial response of the metabolome in the context of metabolic health and the lack of multiple postprandial samples for analysis. Results from this review emphasize that metabolites which are differential by obesity or metabolic health at fasting [e.g. leucine (39)] may normalize between groups in response to a challenge. In response, we postulate that quantifying metabolites approaching the convergence of fat, carbohydrate, and protein metabolism may yield insight on the ability to respond to a nutrient challenge in the context of metabolic health (Figure 4A). The flexibility to respond to a substrate load will be evident in downstream metabolites of fat, carbohydrate, and protein metabolism. Although we recommend the application of untargeted metabolomics and lipidomics to observe global metabolome differences postprandially, we detail key metabolites to quantify to observe their postprandial response in the context of metabolic health, including nonessential amino acids, TCA cycle intermediates, ACs, and other fatty acid oxidation intermediates, such as dicarboxylic fatty acids (Figure 4B).

FIGURE 4.

Profiling metabolites at the convergence of fat, carbohydrate, and protein metabolism. (A) In response to a dietary challenge, macronutrients are metabolized reaching a convergence point of fat (yellow), carbohydrate (red), and amino acid (blue) substrates, represented by dotted rectangle. Profiling downstream metabolites within this convergence point in response to a dietary challenge may provide insights on metabolic health. (B) Several pathways to profile in response to a mixed-macronutrient dietary challenge to determine if metabolic diseases, such as obesity and T2D, alter flux (represented by red exploding star). Elevated medium-chain acylcarnitines in response to a meal may be representative of overloaded or impaired β-oxidation in the mitochondria. Downstream branched-chain amino acid metabolites (represented by valine) such as 2-hydroxyisovalerate and downstream aromatic amino acid metabolites (represented by tryptophan) such as kynurenine represent impaired flux through essential amino acid metabolic pathways. Overload of the TCA cycle may be represented by alterations in upstream metabolites such as alanine, serine, and glycine or glutamate and proline. Many of these metabolic pathways reside in the mitochondria of the cell, however, extra-mitochondrial pathways, such as ω-oxidation forming dicarboxylic fatty acids in the endoplasmic reticulum, may yield insight on the ability to respond to a nutrient challenge. Created with BioRender.com. AC, acylcarnitine; FA-COOH, dicarboxylic fatty acid; FA, fatty acid; LC, long-chain; MC, medium-chain; PDH, pyruvate dehydrogenase; TCA, tricarboxylic acid cycle; T2D, type 2 diabetes; VLC, very long-chain.

Nonessential amino acids represent downstream protein metabolism and TCA cycle overload and metabolic health may alter their postprandial response, as supported by (22, 31, 40). Badoud et al. (31) demonstrated that essential amino acid abundance in the test meal was correlated with concentrations in the postprandial plasma, implying that essential amino acids represent input, rather than the ability to metabolize the nutrient load. These authors demonstrated that nonessential amino acids, such as proline and alanine, elicited the largest response to the mixed-macronutrient challenge and represent modification and metabolism of the input substrates. As we hypothesize that the ability to respond to a substrate load will provide insights on metabolic health, profiling nonessential amino acids, in addition to amino acid metabolites [e.g. 2-hydroxyisovalerate (40)], is necessary. Furthermore, assessment of all TCA cycle intermediates will indicate flux into the cycle and points of overload. For instance, elevations in alanine may indicate a rescue pathway from the increased production of pyruvate due to a high carbohydrate nutrient challenge. It will be necessary to consider the ratios of carbohydrates, fat, and protein in the input nutrient challenge, as we have observed differences in TCA cycle flux dependent on the load [e.g. citrate (32, 33, 35)].

Supported by multiple studies (5, 30, 31, 42), individuals with obesity and metabolic disease have a delayed switch from a catabolic to anabolic state, as evidenced by ratios of carnitine to ACs (31) and indirect calorimetry (42). Profiling ACs will provide insights on metabolic flexibility, in addition to an assessment of the flux of FA from a high-fat macronutrient challenge. Interestingly, the results from 2 studies (30, 35) have suggested that unsaturated long-chain ACs have a larger response to a nutrient challenge than saturated ACs, presenting the rationale to profile ACs with a variety of chain lengths and number of double bonds. Finally, profiling metabolites from extra-mitochondria FA oxidation pathways (e.g. dicarboxylic fatty acids from ω-oxidation), in particular during a high-fat macronutrient challenge, will provide insights on the ability to cope with excess FAs via rescue oxidation pathways (66).

Critique of study designs

We suggest all studies measuring the metabolome in response to mixed-macronutrient challenges consider incorporating the following guidelines into their study design to have consistencies that allow for comparison between studies and integrating data. Participant recruitment should take into consideration that phenotypes [e.g. obesity (40)], lifestyles [e.g. habitual dietary intake (43)], and intrinsic differences [e.g. aerobic capacity (18)] will influence the postprandial response to a challenge. Although research is still inconclusive (46), consideration should be made for incorporating a standard meal the night prior to the challenge. All participants should arrive at the clinic after an overnight fast (≥12 h). Weight and height should be collected from subjects recruited, however, distinguishing between fat and lean mass should be considered, as Ramos-Roman et al. (30) observed specific metabolites in the fasted and fed state associated with lean mass, rather than fat mass. Although not explored in this review, the influence of age (36), sex, race, and ethnicity on the metabolome postprandial response is uncertain, placing importance on collecting and considering these characteristics. Several studies have demonstrated differences in the fasting metabolome by sex in adults, including higher concentrations of plasma BCAAs (67) and ACs (67) in males compared with females. Less explored are differences in the fasting metabolome between races and ethnicities, particularly due to imbalances in race representation with studies. Patel et al. (67) observed higher concentrations of plasma long-chain ACs in Caucasians compared with African Americans. Interestingly, work by Lau et al. (68) explored determinants of the urinary and serum metabolomics in European children, finding that population-specific variance in the metabolome (age, sex, BMI, ethnicity, dietary intake, and country of origin) was more evident in the serum profile (9.0% variance) compared with the urine profile (5.1% variance). Although we hypothesize many metabolites will exhibit a converging pattern between sexes and demographic groups in response to a dietary challenge, future work is warranted.

Investigators should report the percentage of kilocalories from each macronutrient and/or the raw grams of each macronutrient within the challenge. Total caloric load and how the kilocalories were adjusted to each participant (e.g. energy expenditure adjusted) should be included. Only 17 of the reviewed studies (65%) provided details on the nutrient composition within the dietary challenge beyond macronutrients (5, 20, 24–32, 34–38, 41). It is necessary to consider more than the type of macronutrient as previous work has demonstrated that the type of fat has differing influence on the postprandial response of triglycerides (69). Precise details on the meal or homemade liquid shake should be reported to allow for future replication and considerations should be made for using a premade liquid shake for ease in validation.

Glucose and insulin should be profiled over the entire time course and blood sampling should occur ≥240 min postingestion of the mixed-macronutrient challenge, as the largest difference in postprandial triglycerides are observed after 240 min (70). Researchers should consider collecting biological samples for metabolomics at 0, 60, and 240 min to allow for comparison between studies. Although additional metabolites and metabolic pathways are responsive to a meal [e.g. oxylipins (34)], as discussed in this review, it is evident that the metabolites within glycolytic, proteolytic, lipid, and ketogenic pathways should be prioritized for profiling (40, 50). Many unannotated features may respond to a mixed-macronutrient challenge and correlate with degree of obesity or metabolic disease, placing a strong importance of utilizing metabolomics and lipidomics platforms that measure the global metabolome. Metabolomics raw AUC values should be reported in a supplemental table [e.g. (31)] and considerations should be made when calculating fold changes [e.g. postprandial – baseline or (postprandial – baseline)/baseline]. Studies should specify if they are analyzing postprandial values or fold changes, as Li-Gao et al. (22) observed that the postprandial metabolome differed more than the response metabolome in T2D compared with NGT.

Mitigating challenges in the creation of future studies

The culmination of this review has generated several important challenges to consider when pondering future directions for the field. The metabolome reflects the interaction between an organism's genome (e.g. genotype and epigenome) and environment (e.g. adiposity, physical activity, oxidative capacity, habitual dietary intake, microbiome, and chemical exposure), emphasizing interindividual variability in the metabolome (5, 37). For instance, Krug et al. (5) observed marked differences in the postprandial responses of 2 subjects (V13 and V14), which was not attributed to differences in anthropometric measures or RMRs. The authors suggest that the metabolic capacity to oxidize fatty acids, measured by AC ratios, may explain the differences observed between these 2 subjects. Furthermore, the PREDICT 1 study analyzed the postprandial response of metabolic markers in twins and unrelated healthy adults (71). They observed large interindividual variability in the postprandial response of blood triglycerides, glucose, and insulin following identical metabolic challenge test meals. Individual factors that contributed to variation in the postprandial response included meal composition, genetics, anthropometry, and the gut microbiome among several other measured factors. These studies emphasize the challenges and complexity in understanding host metabolism. To address these challenges, team science approaches should be considered, increasing funding and scientific knowledge to precisely profile an individual's genetics and environment.

The majority of reviewed studies (65%) used a targeted metabolomics platform to measure the postprandial response, limiting the exploration of less common metabolites. The benefit of these targeted assays are their price, rapid data accrual, and simplicity in data analysis strategies. However, we encourage the use of an untargeted metabolomics or lipidomics platform to classify the flow of metabolites through metabolic pathways. Since 2012, the NIH Common Fund Metabolomics Program increased access to high-dimensional metabolomics datasets by creating new tools, infrastructure, and bioinformatics capabilities for researchers. This program encourages institutional collaborations to obtain untargeted metabolomics datasets to overcome financial boundaries, one of the main barriers for researchers. Advances in bioinformatics have increased accessibility to statistical tools for high-dimensional data analysis, including user-friendly Java programs grouping untargeted metabolomics datasets into subnetworks based on partial correlations (72). Additional funding opportunities and interdisciplinary collaborations should be available.

In response to this review, we recommend creating a study that measures the metabolome response in participants to liquid mixed-macronutrient challenges, varying the amount of fat (10–60% kcal) and carbohydrates (25–75%), while initially keeping protein concentrations constant (15 ± 5% kcal). These variations in fat and carbohydrate will be representative of typical intakes in the population [e.g Ornish diet (low-fat) compared with Atkins diet (high-fat)], challenging different metabolic pathways. Recruitment should select participants with a lean BMI (18.5 < BMI ≤24.9), male/female, and across a variety of ages and races. Collaborating with geneticists, registered dietitian researchers, kinesiologists, and microbiologists, considerations should be made in identifying contributors to interindividual variations in the postpostprandial response of the metabolome. Once the metabolome response is classified in healthy individuals, we can consider classifying the metabolome response differences attributed to obesity and metabolic disease. The objectives of these studies would be to classify the genetic and environmental contributors to metabolism to understand pathophysiology of metabolic diseases. Potentially, metabolite biomarkers of disease development with limited interindividual variability may be identified. Such a study would fill the gaps identified in this critical review and provide a future direction for reliably evaluating individuals and assessing responses to these very critical mixed-macronutrient challenges.

Notes

Supported by the A. Alfred Taubman Medical Institute of the University of Michigan(CFB), the Robert C. and Veronica Atkins Foundation(CFB), the Michigan Regional Metabolomics Resource Core (R24 DK097153) (CFB), and the University of Michigan Training Program in Endocrinology and Metabolism (T32 DK007245) (JLL).

Author disclosures: The authors report no conflicts of interest.

Supplemental Figure 1 and Supplemental Table 1 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: AC, acylcarnitine; BCAA, branched-chain amino acid; BCFA, branched-chain fatty acid; FA, fatty acid; FFA, free fatty acid; GDH, glutamate dehydrogenase; IFG, impaired fasting glucose; IR, insulin resistance; LBM, lean body mass; LPC, lysophosphatidylcholine; MHO, metabolically healthy obese; MUO, metabolically unhealthy obese; NGT, normal glucose tolerance; OGTT, oral-glucose-tolerance test; OLTT, oral-lipid-tolerance test; PC, phosphatidylcholine; RMR, resting metabolic rate; TCA, tricarboxylic acid cycle; T2D, type 2 diabetes.