Cardiorenal metabolic biomarkers link early life stress to risk of non-communicable diseases and adverse mental health outcomes

Stress is one of the most critical determinants of lifetime health and increases the risk of chronic non-communicable diseases. To gain insight into underlying environment-gene interactions, we analyzed the cardiorenal metabolome of adult mice exposed to multidimensional early-life transportation stress. Using proton nuclear magnetic resonance (1H NMR) spectroscopy, we show that early life stress permanently programs metabolic pathways in somatic organs linked to cardiorenal and mental health disorders in later life. Heart and kidneys of stressed mice revealed robust metabolic markers linked to abnormal energy metabolism, branched-chain amino acid biosynthesis and degradation, methylhistidine metabolism, arginine and proline metabolism, glycine and serine metabolism, and aminoacyl-tRNA biosynthesis. These markers were strongly associated with anxiety-like behaviours. Dysregulation of energy and protein metabolism suggests an increased risk of metabolic diseases like insulin resistance, cardiorenal syndrome, diabetes, and obesity. These findings provide novel insights into the direct effects of early life stress on cardiorenal metabolism and are consistent with prior observations of increased non-communicable disease risk in stressed populations. Thus, stress-associated metabolic signatures in somatic organs may provide early predictors of health risks in later life and reveal new candidates for peripheral biomarker detection with diagnostic value.

Sample collection and preparation. Mice received an intraperitoneal overdose of sodium pentobarbital (150 mg/kg; Euthansol; Merck, QC, Canada). Kidney and heart tissues were extracted, weighed, and stored at -80 °C. To isolate water-soluble metabolites for NMR analysis, tissues were thawed at room temperature and subjected to methanol-based protein precipitation as well as chloroform-based lipid extraction. Finally, samples were centrifuged at 12,000 g for 5 min at 4 °C to precipitate any particulate matter, and 550 μL of the supernatant was transferred to a 5-mm NMR tube for NMR analysis. nMR data acquisition and processing. Data were acquired using a 700 MHz Bruker Avance III HD spectrometer (Bruker, ON, Canada). Spectra were obtained using a Bruker triple resonance TBO-Z probe with the outer coil tuned to the nuclei of 1 H, 31 P and 2 H and the inner coil tuned to the 13 C nucleus. The standard Bruker 1-D NOESY gradient water suppression pulse sequence 'noesygppr1d' was used with a mixing time of 10 ms. Each sample was acquired with 128 K data points, a sweep width of 20.5136 ppm, and a recycle delay of 4 s. Tissue samples were run for 512 scans to a total acquisition size of 128 k. The resulting spectra were then zero filled to 256 k, line-broadened by 0.3 Hz, transformed to the frequency domain, phased, and baseline-corrected. Spectral processing was performed using the Bruker Topspin software (version 3.2, patch level 6), after which spectra were exported to MATLAB (MathWorks, MA, USA) for spectral binning, data normalization, and scaling. Spectra were binned using Dynamic Adaptive Binning 32 . Datasets were normalized using the Constant Sum method 33 to remove effects of imperfect water signal suppression. The dataset was then Pareto-scaled (meancentered and divided by the square root of each variable's standard deviation). All peaks were referenced to TSP (0.00δ).
Statistical analyses. Spectral bins were first analyzed for all comparison groups and classified as either significant or non-significant using a decision tree algorithm and a Mann-Whitney U (MW) test. One hundred and seventy bins were initially included in the kidney analyses and 347 spectral bins were included in the heart analyses. All p-values obtained from these analyses were Bonferroni-Holm corrected for multiple comparisons. Variation in spectral data was visualized using Principal Component Analysis (PCA), Partial Least Squares Discriminant Analysis (PLS-DA), and Orthogonal Projection of Latent Structures Discriminant Analysis (OPLS-DA) using Metaboanalyst. Double cross-validation and permutation testing (2,000 iterations) were performed to validate statistically significant PLS-DA and OPLS-DA results. Variable importance in the projection analysis was performed and plots were made using the weighted sum of squares of the PLS loadings.
Metabolite set enrichment analysis and pathway topology analysis were performed using Metaboanalyst. Metabolic pathway analysis identified the most relevant pathways based on the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database (Mus musculus) and the Over Representation Analysis 33

Results
Stress altered the tissue phenotype. Early life stress led to abnormal kidney and heart weights. Specifically, absolute kidney weights were significantly increased [t(21) = -3.194, p < 0.01] and relative heart weights were significantly decreased [t(9.172) = 3.175, p < 0.01] in stressed animals compared to non-stressed controls ( Supplementary Fig. 1A,B).
Early life stress induced clearly separated metabolic profiles. The kidney and heart analyses included 170 and 347 spectral bins, respectively. A MW test was applied to each comparison group to identify which features led to univariate statistical differences between the groups. These analyses revealed 81 (47.6% of all spectral bins) and 241 (69.5% of all spectral bins) significantly altered features in kidney and heart tissues, respectively. Unsupervised multivariate PCA tests were initially performed using all features. In kidney tissues, separation of the groups was observed in the PCA scores plot, with principal components 1 and 2 accounting for 45.1% and 16.1% of the total variance, respectively ( Supplementary Fig. 2). In heart tissues, no separation of groups was observed when considering all bins. Supervised and unsupervised multivariate statistical tests were then performed using the bins identified as significant from the MW test. In both kidney and heart tissues, the unsupervised PCA scores plots showed clear group separation, with principle component 1 being equal to 41.4% and 50.3% of the total variance and principle component 2 being equal to 18.3% and 31.4% of the total variance, respectively (Fig. 1A,B). The scores plots from supervised PLS-DA (Fig. 1C,D) and OPLS-DA (Fig. 1E,F) tests also both displayed distinct group separation, supporting the PCA findings. Cross-validation and permutation tests validated the observed supervised separation results as a function of early life stress in both kidneys (PLS-DA: p < 0.01, R 2 = 90.2%, Q2 = 83.9%; OPLS-DA: p < 0.001, R 2 = 94.9%, Q2 = 85.9%) and hearts (PLS-DA: p < 0.01, R 2 = 74.4%, Q2 = 56.1%; OPLS-DA: p < 0.001, R 2 = 98.7%, Q2 = 95.5%).
Variable importance in the projection plots ( Fig. 2) indicate which variables contributed most to the group separation observed in the PLS-DA scores plot. In kidney tissues, dimethylamine, uridine, and malate contributed most to the separation ( Fig. 2A), with variable importance in the projection scores of 3.00, 1.61, and 1.58, respectively (Table 1). In heart tissues, glucuronate, adenosine, and uridine contributed most to the separation (Fig. 2B), with variable importance in the projection scores of 3.41, 2.80, and 2.66, respectively (Table 1). In kidneys, 22/44 (50.0%) unique, significantly altered metabolites (as determined by the MW test) were up-regulated in stressed animals compared to controls, with the remaining metabolites being down-regulated (Supplementary Table 1). In hearts, 54/82 (65.9%) unique, significantly altered metabolites were down-regulated in stressed animals compared to controls, with the remainder being up-regulated (Supplementary Table 1). early life stress permanently altered metabolic pathway function. A genome-wide network model of mouse metabolism was used to investigate metabolite sets altered as a result of early life stress (Figs. 3 and 4). In kidney tissues, early postnatal stress exposure most significantly affected arginine and proline metabolism (p < 0.001), methylhistidine metabolism (p < 0.001), as well as glycine and serine metabolism (p < 0.01) (Fig. 3A). Additionally, numerous energy metabolism systems were altered, including pathways in aminoacyl-tRNA biosynthesis (p < 0.00000001), arginine and proline metabolism (p < 0.001), alanine, aspartate, and glutamate metabolism (p < 0.001), and valine, leucine, and isoleucine degradation (p < 0.05) (Fig. 3B). In heart tissues, early life stress most significantly altered methylhistidine metabolism (p < 0.01), phosphatidylcholine biosynthesis (p < 0.01), as well as glycine and serine metabolism (p < 0.01) (Fig. 4A). Pathway topology analyses (Fig. 4B) revealed significant effects on aminoacyl-tRNA biosynthesis (p < 0.001), glycine, serine, and threonine metabolism (p < 0.001), histidine metabolism (p < 0.01), and valine, leucine, and isoleucine biosynthesis (p < 0.05). early life stress-induced behavioural phenotype was associated with metabolic function. Open field locomotor profiles revealed increased anxiety-like behaviour in stressed animals compared to non-stressed controls, as reflected by the number of squares traversed in the centre of the open field arena [t(21) = 3.069, p = 0.0060]. Sixteen and 117 features correlated significantly with increased anxiety-like behaviours in kidney and heart tissues, respectively (Supplementary Table 2). Of interest, the relationship between the number of central squares traversed and the relative concentrations of methionine, malate, and serine indicated a positive correlation for serine (r = 0.45, p = 0.029) and negative correlations for methionine (r = -0.66, p = 0.00066) and malate (r = -0.64, p = 0.0011) in kidney tissues (Fig. 5A, Table 2). The relationship between the number of central squares traversed and the relative concentrations of malonate, lysine, and creatine indicated a negative correlation for creatine (r = -0.52, p = 0.011) and positive correlations for malonate (r = 0.57, p = 0.0050) and lysine (r = 0.49, p = 0.017) in heart tissues ( Fig. 5B, Table 2).  www.nature.com/scientificreports/

Discussion
The perinatal environment is a critical determinant of lifelong NCD risk. Here, we found that multidimensional early life stress was linked to chronic cardiorenal metabolic pathologies in adulthood based on stressed animals having altered (i) organ weights, (ii) affective state, and (iii) metabolites and/or metabolic pathways linked to adverse mental health outcomes and metabolic illness, such as cardiorenal syndrome, insulin resistance, diabetes, and obesity. Experience-dependent metabolic markers in heart and kidney are involved in energy metabolism, branched-chain amino acid biosynthesis and degradation, methylhistidine metabolism, arginine and proline metabolism, and glycine and serine metabolism. Moreover, many of the metabolites found to be significantly altered by stress belong to metabolic pathways involved in aminoacyl-tRNA biosynthesis, supporting a link between early environmental insults and NCD risk later in life. Previous studies have linked perinatal stress to abnormal HPA axis function in the offspring, along with neuronal and receptor loss in brain areas associated with the control of stress response and affective state 34 . Interestingly, these hormonal changes are associated with epigenetic changes, including altered microRNA expression, along with HPA axis reprogramming 35 . The present findings support these earlier studies by linking early life stress to cardiorenal remodeling. The observed changes in heart and kidney weights suggest a permanent reprogramming of physiological, metabolic, and epigenetic processes manifesting as downstream alterations in cardiorenal phenotype and function. Indeed, the development of renal pathology was associated with increased kidney weight. Absolute kidney weight in animals and humans is attributable to endothelial cell proliferation and intrarenal lipid deposition that impair intrarenal hemodynamics 36,37 . Stress-induced kidney hypertrophy is also correlated with higher blood pressure and subsequent hypertensive renal damage 38 . Similarly, cardiac remodeling and hypertrophy are pathologically associated with NCDs such as diabetes, obesity, and CVDs 39 . By contrast, relative heart weight is decreased in diabetes due to myocyte loss and inadequate reactive hypertrophy of the remaining cells 40 , which is also related to deteriorated heart contractile function 41 and impaired insulin signaling 42 . Thus, the present data suggest an association between early life stress and cardiac atrophy and risk of metabolic disorders such as diabetes.
Early life stress significantly changed 81 metabolites in kidney tissues and 241 metabolites in heart tissues. In both heart and kidney, stress most drastically altered pathways involved in aminoacyl-tRNA biosynthesis. Thus, while alterations in individual metabolites were not identical across tissue types, aggregate stress-induced metabolic changes ultimately led to a shared cardiorenal phenotype. Accordingly, previous work demonstrated mitochondrial dysfunction, cardiorenal pathologies, and impaired aminoacyl-tRNA biosynthesis in response to environmental insults 26,39,43,44 .
Several metabolites that obtained a significant variable importance in the projection score in both heart and kidney tissues are involved in global hemodynamics. Abnormal levels of metabolites may mediate vasoconstriction and vasodilation, such as uridine and adenosine, and impair hemodynamics and hypertension and are risk factors for non-communicable cardiovascular and kidney diseases. Uridine, which has vasoconstrictive actions, was up-regulated in both heart and kidney, while adenosine, an inhibitor of vasoconstriction, was down-regulated. Indeed, hypertension is characterized by a sustained increase in total peripheral vascular resistance, suggestive of a primary cause in hypertension 45 . In kidneys, dimethylamine, a product of the hydrolysis of asymmetric dimethylarginine 46 , contributed to group separations. Kidneys are a major extraction site for asymmetric dimethylarginine 47 and renal dysfunction promotes accumulation of asymmetric dimethylarginine and reduced dimethylamine 46 .   www.nature.com/scientificreports/ playing a key role in the pathophysiological underpinnings of anxiety disorders. Oxidative stress, involved in the pathogenesis of mental health disorders 48 , cardiovascular and kidney diseases 49 , is caused by altered mitochondrial energy pathways leading to overabundance of oxidative stress compounds. In both animals and humans, stress-induced anxiety alters levels of Krebs cycle intermediates, which subsequently exacerbates oxidative damage.
In heart tissues, glucuronate, a product of the oxidative cleavage of myo-inositol catalyzed by myo-inositol oxygenase, was highly significant in contributing to unsupervised and supervised separations. Both glucuronate  www.nature.com/scientificreports/ and myo-inositol are involved in inositol metabolism, supporting the involvement of this biochemical pathway in the mediation of stress-induced metabolic changes. Abnormalities in inositol metabolism have been implicated in insulin resistance and long-term microvascular complications in diabetes 50 . In renal tissues, myo-inositol depletion has been associated with diabetic nephropathy through the activation of fibronectin 50 . Myo-inositol has also been found to decrease in the prefrontal cortex and cerebrospinal fluid in affective disorders 51,52 . Interestingly, glucuronate is markedly reduced in the urine of perinatally-stressed rats with atherosclerosis, suggesting a link to atherogenesis following early life stress 43 .  Increased anxiety-like behaviours in stressed animals correlated with the relative concentrations of serine and methionine in their kidneys. Stress down-regulated serine, a co-agonist of ionotropic N-methyl-D-aspartic acid receptor activation and involved in affective behaviour, with serine-depleted mice exhibiting more anxious behaviour and impaired cognitive function 53 . Conversely, elevated brain serine is linked to anxiety-like behaviours, with chronic dietary serine supplementation also having anxiolytic consequences 54 . Interestingly, plasma serine levels correlate with glomerular filtration ratio 55 . The renal reabsorption of serine is sensitive to the presence of chronic kidney disease, with the combination of plasma serine and urinary dynamics effectively distinguishing chronic kidney disease from non-diseased 55 .
Stress also up-regulated methionine, a sulfur-containing donor of methyl groups. Elevated methionine levels have been implicated in the pathogenesis of disorders linked to oxidative stress. In mice, a high-methionine diet is associated with oxidative stress in cardiac tissues due to increased levels of lectin-like oxidized low-density lipoprotein receptor-1 (LOX-1) and superoxide dismutase 1 (SOD1) 44 . Abnormally high levels of methionine also Table 2. Kidney and heart metabolites found to be most significantly correlated to anxiety-like behaviour. Pearson correlations were used to assess the relationship between behaviours indicative of heightened anxiety (i.e., more central squares entered in the open field) and relative concentrations of metabolites found to be significantly altered by stress in a Mann-Whitney U test (n = 9 control animals; n = 14 stressed animals). Top 15 r values, shown in descending order, correspond to Fig. 5. In hearts, the bottom three entries correspond to additional metabolites shown to have an association with adverse mental health outcomes. Positive correlations indicate that a higher anxiety-like state was linked to lower metabolite concentrations, while negative correlations indicate that a higher anxiety-like state was linked to higher metabolite concentrations. Metabolites for which more than one NMR resonance peak was identified are represented as metabolite.1, metabolite.2, … metabolite.n. ‡ Indicates metabolites that were significantly correlated to anxious behaviour in both kidney and heart tissues. www.nature.com/scientificreports/ have anxiogenic effects mediated by oxidative stress 56 . Thus, stress-induced methionine increases may provide a mechanism underlying the comorbidity of cardiorenal and affective disorders. The increased prevalence of anxiety-like behaviours in stressed animals correlated with low concentrations of lysine in heart tissues. In both animals and humans, prolonged dietary lysine inadequacy can elevate stressinduced anxiety while lysine fortification can reduce chronic anxiety 57 . In a rural Syrian population dietary lysine supplementation was found to lower blood cortisol levels and sympathetic arousal in response to stress 57 . Anxiogenic consequences of lysine deficiency may be mediated by serotonin alterations in the central amygdala 58 , with lysine acting as a partial serotonin receptor 4 antagonist. Since abnormal concentrations of lysine have been associated with diabetes 59 , abnormal altered lysine metabolism may also be a factor in chronic disease etiology.
Metabolite set enrichment analysis was used to identify patterns of metabolite concentration changes in a biologically meaningful framework 60 . The most significant pathway altered in heart tissues was methylhistidine www.nature.com/scientificreports/ metabolism, with 3 metabolite hits, whereas in kidney tissues, this was the case for arginine and proline metabolism, with 9 metabolite hits. Significant pathways in the kidney metabolite set enrichment analysis, apart from arginine and proline metabolism, included methylhistidine metabolism as well as valine, leucine, and isoleucine degradation. In the heart metabolite set enrichment analysis, significant pathways included glycine and serine metabolism, methionine metabolism, and arginine and proline metabolism. 3-Methylhistidine, a constituent of actin and myosin released via protein degradation, was up-regulated in response to stress. Methylhistidine metabolism serves as an indicator of muscle myofibrillar protein breakdown, with an increase in 3-methylhistidine release in muscle atrophy 61 . Disruptions in methionine, glycine and serine metabolisms support the finding that stressed animals are at a higher risk for developing cardiorenal and affective disorders. Furthermore, decreased arginine availability due to perturbed renal biosynthesis is causally implicated in nitric oxide deficiency, which contributes to cardiovascular events and kidney damage 62 . Consistent with metabolite set enrichment analysis results and variable importance in the projection scores, both heart and kidney pathway topology analyses revealed changes in valine, leucine, and isoleucine biosynthesis and degradation. This draws attention to a possible metabolic dysregulation of branched-chain amino acids in stressed animals. Abnormal branched-chain amino acid metabolism has been implicated in the pathogenesis of insulin resistance, obesity, and type 2 diabetes mellitus. Altered concentrations of branched-chain amino acids promotes oxidative stress and inflammation of peripheral blood mononuclear cells via mTORC1 activation 63,64 , thus contributing to the pro-inflammatory and oxidative status observed in many pathophysiological conditions.
Stress also altered the aminoacyl-tRNA biosynthesis pathway, which is central for biosynthesis and kinetics of mRNA translation. Accordingly, changes in aminoacyl-tRNA biosynthesis have been causally implicated in the pathogenesis of mitochondrial metabolic and congenital heart diseases. In diabetes, both hyperglycemia and hyperinsulinemia stimulate the accumulation of mutations in mitochondrial tRNAs, the substrates of aminoacyl-tRNA synthetases, by inducing oxidative stress 65 . Altogether, the present results suggest that multidimensional early life stress increases the risk for developing diabetes later in life.
The present findings provide novel insights into the mechanisms underlying early life stress-induced disease vulnerability by linking a cardiorenal metabolic stress phenotype to chronic heart and kidney dysfunction and impaired mental health. These data also suggest a role for transportation stress in the etiology of long-term health complications in laboratory animals and livestock [29][30][31] . The metabolic changes arguably reflect underlying epigenetic modifications and associated cellular functions that manifest as downstream alterations to the metabolome. Clearly recognizable metabolic biomarkers of disease open new personalized medicine strategies for early detection and diagnosis of chronic NCDs in clinical populations.