Acute undernutrition
Most metabolic phenotyping studies investigating early-life undernutrition have focused on acute undernutrition in humans and various animal models (Table
2). During acute malnutrition, there is insufficient food intake to meet the energy and nutrient demands of the body. As a result, autophagy can occur where body fat and muscle is broken down resulting in wasting, the metabolic rate is reduced, kidney function is disrupted and the immune system is impaired. In acutely malnourished children the rates of death from diarrhea, malaria, pneumonia and measles are greatly increased [
45]. Bartz et al. conducted one of the first comprehensive metabolomic studies to investigate the metabolic derangements associated with SAM in children from Uganda aged 6 months to 5 years [
46]. Plasma metabolomes were characterized by a combination of LC–MS, targeting amino acids and acetylcarnitines, and an untargeted GC–MS-based approach. At baseline, acutely malnourished children were reported to have high circulating amounts of non-esterified fatty acids (NEFAs or free fatty acids), ketones, and even-chain acylcarnitines and low amounts of albumin and amino acids. However, a matched nourished control group was not included in this study and so it is assumed that these observations are derived from comparisons to age-matched reference values. High circulating amounts of free fatty acids, ketones and even-numbered acylcarnitines indicate that the hydrolysis of lipid stores and the oxidation of fatty acids form part of a key adaptive response to acute childhood malnutrition. In-hospital mortality remains high (10–30% of all treated cases) in children with SAM after nutritional rehabilitation following WHO-established protocols and following discharge (8.7% within 3 months of discharge [
47]). These circulating metabolites were sharply reduced after the nutritional rehabilitation of these infants, as were myo-inositol and ethanolamine, consistent with a lower turnover of complex lipids. Several plasma amino acids (alanine, isoleucine, leucine, proline, valine, ornithine, threonine, cysteine, tryptophan) were also found to increase after treatment. The authors hypothesize that nutritional therapy increased the availability of dietary protein resulting in a greater oxidation of amino acids and a reduction in fatty acid oxidation. As a result, fat deposition is increased, and weight gain is promoted. Leptin, a hormonal marker of adipose tissue reserve and a modulator of immune function, was found to be a strong predictor of mortality prior to and during nutritional rehabilitation.
Two clinical phenotypes of SAM exist, marasmus and kwashiorkor with an intermediate state of marasmic-kwashiorkor. Marasmus is characterized by severe wasting (WAZ < − 3), whereas edema (swelling caused by excess fluid trapped in tissues) is characteristic of kwashiorkor. Hepatic steatosis, hypoalbuminemia and loss of hair pigmentation is also seen with kwashiorkor. As such, marasmus is considered an adaptive response to starvation while kwashiorkor is considered a maladaptive response to this stress. Typically, marasmus results from an insufficient intake of energy required to meet the body’s demands, whereas kwashiorkor can result from an insufficient intake of protein despite adequate intake of carbohydrates. To understand the pathophysiologic mechanisms underlying the clinical phenotypes of severe acute undernutrition, Giovanni et al. investigated the serum metabolomes of Malawian children aged 9–29 months [
48]. These investigators used a targeted LC–MS/MS approach to measure 141 metabolites including sugars, amino acids, biogenic amines, sphingomyelins, acylcarnitines, and phosphatidylcholines. Hospitalized children with kwashiorkor and marasmus were compared to non-stunted and stunted community children at baseline and 3 days after nutritional stabilization. Metabolic distinctions were observed between the two SAM phenotypes at baseline with most metabolites (128 out of 141) being lower in sera from kwashiorkor children. This included lower circulating amounts of amino acids in children with kwashiorkor compared to those with marasmus, including the essential amino acids aspartate, tryptophan and its derivative kynurenine. These findings are consistent with inadequate protein intake with kwashiorkor and previous observations that protein breakdown in response to food deprivation is different between these phenotypes [
49]. Manary et al. previously noted an accelerated rate of protein turnover in children with marasmus and acute infections compared to those with kwashiorkor and acute infections [
50]. Other biochemical differences included lower acylcarnitines with kwashiorkor and a lower acylcarnitine to free carnitine ratio compared to marasmus indicative of reduced fatty acid transport into the mitochondria for β-oxidation in the kwashiorkor state. Carnitine synthesis requires lysine and methionine and both amino acids were lower in the sera of kwashiorkor children vs those with marasmus. Differences in the bioavailability of these precursors could limit the flexibility in energy strategies. Nutritional therapy was observed to improve the abundance of several metabolites in these children including circulating amino acids, biogenic amines, and lysophosphatidylcholines. Both SAM phenotypes were found to be metabolically distinct from the non-stunted and stunted community controls, even after nutritional stabilization. This included variation in the circulating abundance of amino acids, biogenic amines, sphingomyelins, acylcarnitines, and phosphatidylcholines. The persistence of these metabolic disruptions indicates the long-lasting effects of these metabolic adaptations and may contribute to the poor long-term outcomes of these individuals.
Another study applied an untargeted multi-platform (GC– and LC–MS) approach to measure the stool and plasma metabolomes of Nigerian children aged 6–48 months with SAM (21 with marasmus and 26 with kwashiorkor) and compared them to 11 hospital control children [
51]. While no SAM-associated metabolic variation was identified in the fecal profiles the SAM and control children could be discriminated based on their plasma metabolomes. Consistent with the findings of Bartz et al., children with SAM exhibited a marked reduction in amino acids (glutamine, arginine, tyrosine, BCAA, and the tryptophan metabolite kynurenine) and an increase in even-chain acylcarnitines [
46]. Several phospholipids belonging to the phosphatidylcholine and phosphatidylethanolamine families were significantly decreased with SAM as were several oxylipins. Oxylipins are bioactive lipids derived from the oxidation of long-chain polyunsaturated fatty acids (LCPUFA), including arachidonic acid-derived eicosanoids. These lipids perform important functions such as immune regulation, tissue repair and blood clotting. LCPUFA deficiency has been previously observed in children with SAM and is hypothesized to be a product of both dietary deficiencies and increased β-oxidation of fatty acids during starvation (indicated in this study by increased even-chain acylcarnitines) [
52]. In addition, dihexoses, lactate, decanoylcarnitine, angiotensin I, free iron (III) heme, and a sphingoid base were found to be higher in the plasma of SAM children compared to nourished children. Disaccharides are not readily absorbed and increased plasma dihexoses suggest increased movement of these sugars through the intercellular space of the intestinal mucosa indicating impaired gut function with reduced intestinal barrier function. Interestingly, no differences were observed in the plasma or fecal metabolomes between edematous and non-edematous malnutrition. However, the relatively low sample numbers and wide age range may have been a factor in obscuring such differences.
The temporal biochemical impact of SAM has been studied on the plasma metabolome of juvenile pigs [
53]. In this model, symptoms consistent with marasmus were induced by feeding 4-week-old weaned pigs a nutrient-deficient maize diet for 7 weeks. This diet had a comparable energy content to the reference diet but contained 61% less digestible protein. Maize-based diets are common in low-income countries and are known to be deficient in lysine, tryptophan, and vitamins A and B. Weekly profiling of the plasma metabolome using an untargeted LC–MS strategy revealed the evolution of a variety of biochemical disruptions. Circulating essential (valine, leucine, lysine, threonine, methionine, phenylalanine, tryptophan) and non-essential (glutamine, glutamate, proline, tyrosine) amino acids decreased in the undernourished pigs compared to the reference pigs. This was in-line with previous findings in marasmic children [
54]. Many of these changes occurred rapidly after the first and second weeks of undernutrition and became more pronounced as the study progressed. Histidine was the only plasma amino acid observed to increase with undernutrition. Carnosine is a dipeptide of histidine and β-alanine and is present in high concentrations in skeletal muscle. Increased histidine may reflect an adaptive response to maintain or recover homeostasis through muscle degradation. This was supported by increases in pseudouridine from the first week of acute malnutrition, which is released from the catabolism of tRNA and rRNA, indicating greater turnover of muscle cells. Pseudouridine, uric acid and allantoin are metabolites involved in nucleoside metabolism and all were increased following SAM. Interestingly, uric acid, a product of purine metabolism, changes in this marasmus model but is not observed to change in kwashiorkor children [
55]. This raises the potential for plasma uric acid to be a biomarker for distinguishing edematous and non-edematous SAM. Other circulating metabolites altered by SAM were those derived from the gut microbiota (hippurate, phenylacetylglycine, and
p-cresol glucuronide) suggesting that gut microbial–host metabolic exchange is disrupted with undernutrition.
To understand the wide-reaching systemic effects of early-life protein–energy undernutrition, Preidis et al., investigated the metabolic derangements occurring in the neonatal mouse model of undernutrition [
56]. In this model, protein–energy undernutrition was induced by timed separation of neonates from lactating dams. Here, 5-day-old CD1 mice were separated from dams for 4, 5, and 12 h on days 5, 6, and 7–15 days of life, respectively. Control mice had unrestricted access to nursing. Following 11 days of caloric restriction, urine, stool, plasma, ileal fluid, cecal fluid and liver tissue were collected. Untargeted profiling using a combination of GC–MS and LC–MS approaches identified 695 metabolite pairs significantly different between the experimental groups. Consistent with the initiation of autophagy, amino acid and lipid catabolites represented 22, 21, and 16 of the 30 most important urine, plasma, and stool metabolites, respectively. Similarly, undernourished mice also had greater amounts of monoacylglycerols and free fatty acids in the liver and greater amounts of carnitine in the urine than their nourished equivalents. Carnitine conjugates (isobutyrylcarnitine, 2-methylbutyrylcarnitine, isovalerylcarnitine) derived from the catabolism of BCAA were also present in higher amounts in the urine, plasma and liver of the undernourished animals. All bile acids were found in lower concentrations in the undernourished mice except for tauromuricholate (plasma) and taurolithocholate 3-sulfate (ileum, cecum, and colon). The metabolite with the greatest fold-change in stool was the secondary bile acid, deoxycholic acid, which was present in lower amounts in the undernourished feces. Bile acids can inhibit the growth of intestinal bacteria, including pathogenic species, and may influence the susceptibility of the host to enteric infections [
57]. Metabolites indicative of inflammation and oxidative stress, features commonly observed in undernourished infants, were altered across different sample types with undernutrition. Collectively, this study highlights the profound impact of early-life undernutrition on the biomolecular landscape of the individual and the global attempts of the metabolic system to adapt and compensate for nutrition insufficiencies. This same model was also used to investigate whether the metabolic derangements were resolved after refeeding and catch-up growth [
58]. In total, 287 metabolites (98 in urine, 73 in plasma and 116 in stools) differed in relative concentrations between undernourished and control mice immediately after nutrient restriction. The most significant group of metabolites were microbial-derived products of aromatic amino acid metabolism. Plasma concentrations of the tryptophan derivative, serotonin were lower in undernourished mice, whereas several tryptophan catabolites (5-hydroxyindoleacetate, indolelactate, and
C-glycosyltryptophan) were increased in the urine compared to control animals. Following 3 weeks of refeeding the deficits in body weight and length of the undernourished mice had recovered. Similarly, the urinary and plasma metabolomes also became more comparable to the nourished profiles with only 20 urinary and 15 plasma metabolites differing between the two groups at this sampling point. Despite this systemic improvement and catch-up growth, the fecal metabolome and the intestinal microbiota of the undernourished mice remained markedly distinct from the control mice. After refeeding, the metabolites that remained most disrupted were the aromatic amino acid phenylalanine and the derivatives of aromatic amino acids, phenyllactate, and
p-cresol sulfate.
The detailed metabolic effects of early-life protein deficiency on the urinary metabolome of mice has been resolved by
1H NMR spectroscopy [
59]. In this model, C57Bl6 mice were weaned onto a protein deficient (2% protein, 84.2% carbohydrates) diet and compared to mice weaned onto a standard diet (20% protein, 64% carbohydrates). Following 13 days of protein deficiency, an up-regulation of the TCA cycle was observed, indicated by the higher excretion of TCA cycle intermediates,
N-methylnicotinamide, and nicotinamide-
N-oxide. The higher carbohydrate content of the protein-deficient diet (84.2%) compared with the standard chow diet (64.1%) may be a factor driving these alterations. Intriguingly, and consistent with observations in children and animals, undernutrition was also found to modify the functional capacity of the gut microbiome. Several amino acid catabolites derived from the gut microbiota were excreted in lower amounts compared to the nourished mice. These catabolites included isobutyrate, isovalerate, polyamines (cadaverine and putrescine), and BCAA catabolites (2-oxoisocaproate, 2-methyl-2-oxovalerate, 2-oxoisovalerate). The gut microbial metabolites of choline (dimethylamine, trimethylamine, and trimethylamine-
N-oxide) were observed to increase in the urine of protein malnourished mice, with choline itself reduced compared to the nourished mice, indicative of enhanced microbial choline utilization following protein deprivation. Other microbial–mammalian co-metabolites that were disrupted by the protein deficient diet included 2-hydroxyisobutyrate, hippurate, phenylacetylglycine (PAG), 4-hydroxyphenylacetate,
m-hydroxyphenylpropionylsulfate, and cinnamate derivatives. To further explore the impact of protein deficiency on the community structure of the gut microbiota, microbial profiling using 16S rRNA gene sequencing was performed on stool samples collected at various times of protein deficiency. The microbial and metabolic profiles were then integrated to identify alterations in the cross-talk between the intestinal microbes and the host following protein malnutrition. From these analyses protein deficiency was observed to restrict the maturation of the gut microbiome as well as its functionality and biochemical exchange with the host. Given the known importance of the gut microbiota on host health, including indigestion and the gut–brain axis, these alterations could have profound effects on host development.
As a significant component of mammalian biocomplexity and an influential factor on host metabolism, digestion and immunological maturation, microbial–host metabolic cross-talk has been investigated further. Changes in the gut microbiota were explored in the CD1 neonatal mouse model of undernutrition. In these animals, protein–energy undernutrition reduced the diversity of the intestinal microbiota and depleted the microbiome of genes that extract energy from non-digestible components [
60]. However, no differences were observed in the fecal microbiota of Nigerian children with SAM compared to nourished hospital controls and no fecal metabolites derived from the gut microbiota were observed to change [
51]. The limited sample size in this study may have prevented such differences from being detected. To further investigate the role of the gut microbiota in the pathogenesis of acute undernutrition Smith et al. transplanted kwashiorkor-associated fecal microbiota from children into C57BL/6J germ-free mice [
61]. In this model, stool samples were collected from twin pairs discordant for kwashiorkor (healthy well-nourished twin vs kwashiorkor twin). Both groups were freely fed a regional Malawian diet (low caloric density, nutrient-deficient diet) for 21 days and compared to their respective controls fed a standard mouse chow diet. Targeted GC–MS of SCFA and untargeted GC–MS metabolomic analyses were performed on fecal and cecal samples, while untargeted
1H NMR spectroscopic analyses were used to measure the urinary and plasma metabolic profiles of the mice. The combination of the Malawian diet and the kwashiorkor microbiome resulted in marked weight loss in the recipient mice compared to the mice receiving the healthy co-twin microbiota on the same diet. These mice excreted higher amounts of allantoin (an end-product of purine metabolism in non-primates, equivalent to uric acid in humans), creatine and creatinine (markers of muscle turnover) and gut microbial-derived metabolites of choline (methylamine, dimethylamine, and trimethylamine-
N-oxide) and amino acid metabolism (PAG, 3-indoxyl sulfate, and hippurate). In addition, fecal methionine and cysteine (involved in sulfur metabolism) concentrations were significantly lower in mice harbouring the kwashiorkor microbiota compared with those carrying healthy co-twin microbiota when consuming a Malawian diet. Selective inhibition of the TCA cycle by the kwashiorkor associated gut microbiota was also observed. These findings support the notion that the gut microbiota participate in the pathogenesis of SAM.
Chronic undernutrition
The urinary metabolic phenotypes of Brazilian children aged 6 months–2 years enrolled in a case–control study were characterized using
1H NMR spectroscopy [
62]. In this study, similar metabolic associations were found with all measures of stunting (HAZ), underweight (WAZ) and wasting (WHZ). Stunted children excreted lower amounts of betaine and dimethylglycine, endogenous metabolites of choline. These changes indicate that choline and betaine bioavailability is lower in chronically undernourished children. As these are important methyl donors such differences may have downstream implications for various functions reliant on methylation including epigenetic processes such as DNA methylation. In addition, the gut bacterial-host co-metabolites of the amino acids phenylalanine (PAG), tyrosine (4-cresyl sulfate), and tryptophan (3-indoxyl sulfate) were excreted in greater amounts by the stunted children suggesting a functional modulation of the gut microbiome towards increased proteolytic activity. In addition to 3-indoxyl sulfate, other metabolic perturbations reflected disruptions to tryptophan metabolism. The majority of tryptophan is metabolized to kynurenine via the hepatic enzyme tryptophan 2,3-dioxygenase (TDO) or the enzyme indoleamine 2,3-dioxygenase (IDO), which is expressed in several extra-hepatic tissues. While TDO protects against toxic levels of tryptophan accumulating in plasma, IDO is increased in response to infection and inflammation [
63].
N-methyl-2-pyridone-5-carboxamide (2-PY) and
N-methylnicotinic acid are downstream metabolites of kynurenine and were excreted in higher amounts by stunted children. This indicates greater oxidation of tryptophan to kynurenine in these individuals. The downstream metabolites of kynurenine suppress the proliferation of activated T cells, effectively dampening the immune response to prevent the pathology associated with chronic inflammation [
64]. These changes are most likely in response to systemic inflammation as a result of increased intestinal permeability and persistent infections. This is supported by observations in these same children where the plasma kynurenine: tryptophan ratio was positively correlated with plasma levels of lipopolysaccharide-binding protein [
65]. Research in Peruvian and Tanzanian children also found plasma tryptophan and kynurenine to be associated with systemic markers of inflammation, growth and oral vaccine response [
66]. This metabolic adaptation to inflammation and enhanced microbial-utilization, coupled to reduced dietary intake of tryptophan and malabsorption, depletes the availability of tryptophan for the host. Tryptophan is an essential amino acid and is necessary for protein synthesis and growth, immunomodulation and the biosynthesis of the neurotransmitter, serotonin [
67]. In this study,
N-methyl-nicotinamide (NMND) and β-aminoisobutyric acid (BAIBA) were also identified as metabolic predictors of catch-up growth 6 months later in stunted children. NMND was positively associated with future growth while BAIBA was inversely associated with growth. Through different mechanisms, both metabolites reflect energy sparing strategies by the individual allowing greater energy availability for growth. Identification of such biomarkers may provide useful tools for detecting children at risk of further growth shortfalls allowing interventional strategies to be targeted towards those most in need.
Another study used a targeted approach to measure serum amino acids in children from rural Malawi [
68]. Here, stunted children aged 1–5 years had lower serum concentrations of a variety of amino acids including essential (isoleucine, leucine, valine, methionine, threonine, histidine, phenylalanine, lysine), conditionally essential (arginine, glycine, glutamine), and non-essential amino acids (asparagine, glutamate, serine), as well as citrulline. As with the Brazilian, Peruvian and Tanzanian children, tryptophan was also found to be lower in the stunted children. Additional changes included lower serum ornithine, taurine, dimethylarginine, carnitine and six sphingomyelins in stunted compared to non-stunted children.
The metabolic effects of long-term protein–energy malnutrition have been studied in Sprague–Dawley rats through 5 weeks of strict dietary restriction (deprived to 60% of the amount consumed by control rats) [
69]. Long-term starvation resulted in a range of biochemical derangements including alterations to the TCA cycle (
cis-aconitate, citrate, isocitrate, succinate), choline metabolism (betaine), and purine metabolism (uric acid). It also caused an increased excretion of metabolites involved in carnitine (3-hydroxyisovalerylcarnitine), muscle (creatine, creatinine) and tryptophan/nicotinamide metabolism (kynurenic acid, 2PY/4PY, dihydroxyquinoline). These findings were consistent with many of those seen in children and highlight the wide range of metabolic processes modulated by chronic nutritional restrictions.
Environmental enteric dysfunction
Environmental enteric dysfunction (EED), previously termed environmental enteropathy or tropical sprue, is a condition characterized by villous atrophy, chronic inflammation of the small intestinal mucosa, malabsorption and increased intestinal permeability. It is highly prevalent in South America, South Asia and sub-Saharan Africa and arises from the additive effects of a malnourishing diet and chronic exposure to enteric pathogens. This condition is considered a significant driver of childhood stunting, cognitive impairment and metabolic derangements that predispose individuals to later-life disorders such as obesity, cardiovascular diseases and diabetes. These have recently defined as “
HAZdrop”,
“COG-hit” and “
Met-syn”, respectively [
70]. As a major hallmark of EED, metabolites associated with villus damage were investigated in hospitalized children with SAM from Zambia [
71]. In this study, measures of villus health obtained from small intestinal biopsies were correlated with matched urinary metabolic phenotypes. Villus blunting was associated with a reduced excretion of many gut microbial-derived metabolites (4-hydroxyphenylacetate, phenylacetylglutamine, 3-indoxyl sulfate, acetamide, 4-hydroxyhippurate, dimethylamine, trimethylamine), as well as metabolites related to energy and muscle metabolism (succinate, creatine, β-hydroxy-β-methylbutyrate). The lower excretion of these metabolites is likely explained by malabsorption from the gut but decreases in the gut bacterially-derived metabolites could also stem from the reduction in epithelial surface area in the gut, and therefore, inhabitable niches for the microbes. In these children, sucrose was found to be excreted in significantly higher amounts by the two children with the shortest villi. In a healthy gut, sucrose is hydrolyzed by sucrases in the brush border to glucose and fructose. Increased excretion of sucrose with EED suggests that this enzyme activity is reduced and that paracellular absorption occurs due to a loss in barrier function. In agreement with this, sucrose excretion was positively correlated with the lactulose: rhamnose excretion ratio in these children.
Using a combination of malnourishing diets and enteric infections common to developing countries various rodent models of EED have been developed. The double hit of protein malnutrition and
Cryptosporidium infection in the mouse was found to amplify the biomolecular disruptions induced by protein deficiency alone and impact on a variety of biochemical processes [
72]. This included fluctuations in the energy-related pathways, microbial choline degradation, tryptophan metabolism, and biomarkers of inflammation and oxidative stress. The parasite also modified the community structure of the fecal microbiota. Interestingly, many of the infection-associated metabolic changes were not correlated with the microbiome suggesting that these may be products of parasite metabolism. Another study by Brown et al. found several metabolic changes in the small intestinal contents of mice fed a malnourishing diet (7% protein and 5% fat) compared to those on a nourished diet (20% protein and 15% fat) [
73]. In total, 420 metabolites differed between the two groups including dramatic shifts in the small intestinal bile acid pool (reduced tauro-conjugated bile acids in the malnourished mice). Pathways involved in linoleic acid metabolism, and the biosynthesis of phenylalanine, tyrosine, tryptophan, BCAA, and bile acids were all significantly under-represented in the malnourished animals while steroid and sphingolipid biosynthesis pathways were over-represented compared to their nourished counterparts. Furthermore, differences were noted in many vitamins in the small intestine. Consistent with other studies, malnutrition modulated the bacterial community structure of the small intestine, which assisted the acquisition of environmentally acquired microbes. The authors demonstrated that coupling this malnourishing diet with a set of commensal Bacteroidales and
E. coli strains synergistically triggered physiological changes consistent with EED. This included villus blunting and exacerbated the impact of the malnourishing diet on inflammation, intestinal permeability, and growth stunting.