Metabolomics-Based Identification of Metabolic Dysfunction in Frailty

Abstract

Dysregulation of energy producing metabolic pathways has been observed in older adults with frailty. In this study, we used liquid chromatography–mass spectrometry technology to identify aging- and frailty-related differences in metabolites involved in glycolysis, the tricarboxylic (TCA) cycle, and other energy metabolism-related pathways in the serum of a cohort of community-dwelling adults aged 20–97 (n = 146). We also examined the relationship between serum levels of metabolites and functional measures, physical frailty, and risk status for adverse health outcomes. We observed elevated levels of TCA cycle and glycolytic intermediates in frail subjects; however, the differences in the levels of ATP and other energy metabolites between young, nonfrail, and frail adults were not significant. Instead, we found that serum levels of neurotransmitters N-acetyl-aspartyl-glutamate, glutamate, and γ-aminobutyric acid were significantly elevated in older adults with frailty. These elevations of glycolytic and TCA cycle intermediates, and neurotransmitters may be part of the biological signature of frailty.

Physical frailty is a geriatric syndrome characterized by multisystem dysregulation and decline yielding decreased physiological reserves and increased vulnerability to stressors and adverse health outcomes. As the aging population is growing rapidly around the world, a better understanding of the biology that underlies physical frailty will likely yield improved prevention and treatment strategies for vulnerable older adults in the years ahead. While progress has been made in identifying and measuring frailty (1) and predicting adverse outcomes related to frailty (2), much remains to be learned about the specific molecular and physiological pathways that drive physical frailty and functional decline in older adults (3).

The complexity of the biology that underpins physical frailty and the need for greater diagnostic and treatment specificity warrants the use of novel high-throughput measurements to identify with more specificity the biomarkers that link metabolic pathways to frailty and functional decline. The development and implementation of metabolomic measurements over the past decade have enabled deeper dissection of these interrelated energy and inflammatory pathways on the molecular level in an efficient and high-throughput manner. This approach has already provided new insights into complex diseases including cardiovascular disease, diabetes, obesity (4), and renal disease (5) as well as identified metabolites which vary with age (6,7) and frailty status (8,9). Importantly, this approach could help identify more precise biological signatures that are characteristic of frailty, which may enable the development of improved diagnostic criteria beyond present frailty screening tools (10).

In order to better understand how chronic inflammation, aging, and frailty affect circulating pools of energy pathway-related metabolites, and to provide additional evidence for a biological “signature” of aging and frailty, we performed targeted and untargeted metabolomic profiling of the serum from a set of community-dwelling young and older adults. Previous studies have shown that changes in glucose metabolism and mitochondrial function are hallmarks of aging (11). Glucose intolerance as measured by oral glucose tolerance measurement and insulin dynamics have been closely linked to physical frailty in population studies (12), and mitochondrial dysregulation has been linked to physical frailty in both animal models and in human subjects (13). Further, chronic inflammatory pathway activation, a hallmark of physical frailty, impairs energy metabolism and glucose tolerance (14). Therefore, we hypothesized that there are significant differences in the metabolome of younger compared to older populations and older robust compared to older frail populations related to the glycolytic pathway, the tricarboxylic (TCA) cycle, and energy derivatives, such as NAD⁺ and ATP. The cytosolic glycolytic pathway and the mitochondrial TCA cycle are central energy producing metabolic pathways in eukaryotes. The TCA cycle, in particular, is a metabolic hub that processes inputs from the catabolism of carbohydrates, amino acids, and fatty acids, generates reduced intermediates for the electron transport chain, and provides precursors for biosynthesis including that of neurotransmitters. Recent studies indicate a prominent mechanism through which mitochondria communicate with the rest of the cell is through the release of TCA cycle intermediates which can alter cellular function (15). Due to the action of various mitochondrial and cellular transport mechanisms, the signaling capabilities of TCA cycle intermediates are not limited to the intracellular space, and these metabolites are now known as extracellular effector molecules that signal information on organismal physiological states (15).

Method

Population

One hundred and sixty-six community-dwelling adults, aged 20–93 years, living in the Baltimore, Maryland area were recruited with the goal of developing a discovery set of serum samples for comparisons between younger and older adults as well as comparisons between older robust and older physically frail adults. Frailty status was assessed using a commonly utilized and well-validated frailty phenotype screening tool (1). Study subjects were enrolled with informed consent and the study protocols had appropriate approval by the Johns Hopkins Medical Institutional Review Board.

Serum Collection and Cytokine Measurement Procedures

Blood collection visits were scheduled in the morning. Serum cytokines (interleukin-6 [IL-6], tumor necrosis factor α [TNFα], TNF α receptor 1 [TNFαR1], and IL-1β) were assayed using a quantitative sandwich enzyme-linked immunosorbent assay (Mesoscale Diagnostics, Rockville, MD) following the manufacturer’s protocol. Patients were not fasted in this study. Due to sample volume limitations, some metabolite measurements were not available for 3-frail, 12-nonfrail, and 5 young subjects.

Metabolomic Analysis of Patient Serum Samples

Metabolites from 146 patient serum samples were extracted using 80% methanol and resuspended in 50% acetonitrile following speed vacuuming and lyophilization (16). Metabolomic data were acquired using an Agilent 6490 triple quadrupole (QQQ) mass spectrometer with an Agilent 1260 HPLC system (Agilent Technologies, Santa Clara, CA) and a Thermo Scientific Q Exactive Plus Orbitrap Mass Spectrometer with a Vanquish Horizon UPLC system at the Metabolomics Facility at Johns Hopkins Medical Institutions (Thermo Fisher Scientific, Inc., Waltham, MA). Both the Agilent 1260 HPLC autosampler system and the Vanquish UPLC autosampler system were used to draw 2 µL of each sample kept at 4 °C. The reverse-phase chromatography used for the Agilent 1260 HPLC had a mobile aqueous phase of 0.1% formic acid in MS grade water, and a mobile organic phase of 0.1% formic acid in 98% acetonitrile (Sigma-Aldrich, St. Louis, MO). The reverse-phase chromatography used for the Vanquish UPLC had a mobile aqueous phase of 0.1% formic acid in MS grade water, and a mobile organic phase of 0.1% formic acid in 100% acetonitrile. The total runtime for each sample for the Agilent 1260 system was 50 minutes. The total run time for each sample for the Vanquish system was 13 minutes with 11-minute data acquisition time and 2-minute column re-equilibration time. A Discovery HS F5 HPLC Column (Sigma) with a guard column (Sigma) kept at 35 °C was used for both systems. For the Agilent QQQ mass spectrometer, mass calibration was performed during data acquisition continuously using reference calibrant mix (Agilent Technologies). For the Thermo Scientific Q Exactive Plus Orbitrap Mass Spectrometer, a mass calibration was performed prior to every data acquisition using designated calibrant mix to ensure sensitivity and accuracy of data acquired.

The data acquired using the Agilent QQQ mass spectrometer were analyzed using the Agilent Quantitative Analysis Software (Agilent Technologies) (16). The data acquired using the Thermo Scientific Q Exactive Plus Orbitrap Mass Spectrometer were analyzed using Thermo Scientific Compound Discoverer software and Thermo Scientific TraceFinder software (Thermo Fisher Scientific, Inc.). The raw intensities were obtained by integrating the chromatographic peaks. Relative intensity was calculated by dividing the intensity of each sample by the average intensity of the young group.

Statistical Analysis

Using human subjects, comparisons of individual metabolite values between young and older adults were carried out by 2-tailed t test without assuming standard deviation with statistical significance determined using the Holm–Bonferroni and false discovery rate methods for multiple comparisons correction with maximum family-wise error rate at 0.05. A p value less than .05 was considered significant. Linear regression models were used to estimate the association between metabolite levels and cytokines (IL-6, IL-1, IFN, TNFα, and TNFαR1) and functional measures including walking speed and grip strength after adjusting for age. Statistical analyses were performed using GraphPad Prism 6, SAS version 9.2, and MetaboAnalyst 5.0.

Results and Discussion

In this study, we investigated changes in metabolic pathways that occur with aging and frailty in a cohort of young and older men and women. The effect of age on metabolite levels was analyzed by comparing young (mean age 25.8 years, n = 45) and older subjects (mean age = 77.8 years, n = 101) (patient demographics are given in Supplementary Table 1). We identified significantly elevated levels of malate, N-acetyl-aspartyl-glutamate (NAAG), α-ketoglutarate (α-KG), glucose, beta l-citryl-glutamate, γ-aminobutyric acid (GABA), cis-aconitate, succinic semialdehyde, fumarate, ADP, succinate, and dihydroxyacetone phosphate and significantly decreased NAD⁺/NADH ratio and pyruvate levels when we used the false discovery rate method of multiple comparison correction (Table 1). The Holm–Bonferroni method of correction for multiple comparisons generated similar results (Supplementary Table 2).

Table 1.

Serum Metabolite Comparison of Young and Old Human Subjects

	Young (n = 45)	Old (n = 101)
Metabolite	Relative Area		Direction	p Value
Malate	836767.0	1309410.0	↑	3.04E-07
NAAG	62523.4	134707.0	↑	5.08E-07
α-Ketoglutarate	978682.0	1630059.0	↑	7.77E-07
Glucose	34094060.0	55468720.0	↑	.000003
Beta l-citryl-glutamate	179765.0	312645.0	↑	.000023
GABA	5879346.0	9588539.0	↑	.0007
NAD+/NADH	1.67	0.91	↓	.0008
cis-Aconitate	6032.2	23282.7	↑	.0024
Succinic semialdehyde	7093654.0	9749627.0	↑	.0027
Fumarate	694697.0	1040745.0	↑	.0037
Pyruvate	4001974.0	2400886.0	↓	.0064
ADP	4887.7	7422.9	↑	.0065
Succinate	18695200.0	24808530.0	↑	.0091
Dihydroxyacetone phosphate	277721.0	394037.0	↑	.0195

Notes: GABA = γ-aminobutyric acid; NAAG = N-acetyl-aspartyl-glutamate. Metabolites shown were significantly altered using 2-sample t test without assuming equal variance with the false discovery rate method of correction for multiple comparisons, with (Q) value = 5.0%.

Given that within the general population, frail older adults experience increased vulnerability to adverse health outcomes and have metabolic pathway alterations compared to nonfrail older adults, we stratified our older patient groups based on physical frailty status (frail [average age 79.7, n = 29] and nonfrail [average age 77.1, n = 72]) (patient demographics in Supplementary Table 3). We then compared serum metabolite levels between young, nonfrail older, and frail older subjects. In addition to dysregulation of glucose metabolism, which is known to be associated with physical frailty (12), we see that the dysregulation extends to the TCA cycle as well. Among the glycolysis-related metabolites (Figure 1), glucose was elevated in frail subjects compared to both nonfrail and young subjects and was elevated in nonfrail older subjects compared to young subjects. Glucose-6-phosphate levels were elevated in nonfrail subjects compared to frail subjects. Both nonfrail and frail older adults had elevated dihydroxyacetone phosphate levels compared to young adults. Glycerol-3-phosphate, which is positioned at an intermediate step in glycolysis that bridges to lipid metabolism, was lower in frail subjects compared to both nonfrail and young subjects. Pyruvate was lower in nonfrail subjects compared to young and frail subjects. For TCA cycle metabolites (Figure 1), cis-aconitate was elevated in frail subjects compared to nonfrail and young subjects. α-KG, succinate, and fumarate were all significantly elevated in nonfrail and frail individuals compared to young subjects. α-KG is an important TCA cycle checkpoint molecule (17) as well as a metabolite with actions outside of this pathway. Recently, a study showed dietary supplementation with α-KG extended life span and compressed morbidity in mice and suggested that plasma α-KG levels decline as much as 10-fold between the ages of 40 and 80 in humans (18); however, this decline was not apparent in our cohort. Elevated succinate levels were also apparent in the older groups. Succinate is recognized as an inflammatory signal (19) which plays a role in signaling cell trauma and accordingly, elevated succinate is seen in obesity, type 2 diabetes (20), and cancer (21). Malate was elevated with age in our cohort as well. Malate and fumarate have been shown to positively correlate with frailty, cardiovascular risk, and mortality in other studies (7). Oxaloacetate was also elevated with frailty. In addition, beta l-citryl-glutamate, a direct product of citrate, was increased in nonfrail and frail subjects compared to young subjects and in frail subjects compared to nonfrail subjects. TCA cycle metabolites can be transported from the mitochondria through the dicarboxylic acid transporter and other transport mechanisms to the cytosol and plasma, thus signaling cellular status to the entire organism. We also investigated the correlation between the TCA cycle intermediates within individual subjects of the same group. All 3 groups showed distinct patterns of correlation and clustering of the metabolites. Among the TCA cycle intermediates, only fumarate showed some correlation with malate and succinate, while cis-aconitate, oxaloacetate, and α-KG did not show much correlation with any other TCA cycle metabolites (Supplementary Figures 1 and 2), potentially because of the differential expression and activity of the TCA cycle enzymes. There is also no significant correlation observed between the TCA cycle intermediates with other metabolites mentioned above. In contrast to the overall elevations of glycolytic and TCA cycle intermediates, there were no significant alterations observed in the levels of energy metabolites produced except for a significantly decreased NAD⁺/NADH ratio in nonfrail subjects compared to young subjects and a significantly elevated ADP level in frail subjects compared to nonfrail and young individuals (Figure 1).

Figure 1.

Serum glycolytic and tricarboxylic (TCA) cycle intermediates in young, old nonfrail, and old frail human subjects. Data for young subjects are shown in green. Data for old nonfrail subjects are shown in blue. Data for old frail subjects are shown in red. *p < .05, **p < .01, ***p < .001, ****p < .0001.

Due to the known connections between cognitive impairment and frailty (22), we next investigated pathways related to neurotransmitter metabolism that may be affected by alterations in the TCA cycle. As shown in Figure 2, we found that levels of glutamate, which can be produced from α-KG via glutamate dehydrogenase, were significantly elevated in frail subjects compared to both young and nonfrail subjects. Glutamate-related metabolites were also significantly altered with aging and frailty. GABA, the major inhibitory neurotransmitter and a product of glutamate via glutamic acid decarboxylase (GAD), was significantly elevated in frail subjects compared to nonfrail and young subjects and in nonfrail subjects compared to young subjects. Succinic semialdehyde, the direct product of GABA via GABA transaminase, was significantly elevated in frail adults as compared to nonfrail and young adults and in nonfrail adults as compared to young adults. NAAG, a neuropeptide of the central nervous system (23) and a product of glutamate via N-acetyl-aspartyl-glutamate synthetase A and B, was also significantly elevated in nonfrail and frail subjects compared to young subjects. Aspartate, a precursor for NAAG which can be produced from oxaloacetate via a transamination reaction, was also significantly elevated in frail adults as compared to nonfrail adults. This observation that the production of neurotransmitters, including glutamate, GABA, and NAAG, are elevated in frail subjects corroborate with previous findings showing that elevated levels of neurotransmitters are associated with neurological diseases, specifically neurodegenerative diseases which are common in frail older adults. Elevated levels of GABA are found in astrocytes in the dentate gyrus of Alzheimer’s disease (AD) patients as a result of increased GAD activity (24). This increased GABA level can result in the enhanced tonic neural inhibition observed in AD (24). Elevations in glutamate can lead to excitotoxicity, which has been shown to be involved in neurodegenerative diseases, such as Huntington’s disease and AD (25). Increased level of NAAG, which can serve as a reservoir for glutamate (16), can also contribute to increased glutamate level and cause neurodegeneration in frailty. While the increase in the levels of TCA cycle and glycolytic intermediates in frail adults could be explained by several different reasons, including reduced utilization of these pathways following reduced energy demand, or alterations in metabolic flux, or a combination of both, the observed increase in the levels of several neurotransmitters suggests that there is indeed metabolic perturbations. It is likely that increased levels of TCA cycle intermediates α-KG and oxaloacetate would lead to the increase in glutamate and aspartate levels, thus feeding into the synthesis of GABA and NAAG. The metabolism of GABA through the GABA shunt to succinate then further sustains the elevated levels of TCA intermediates.

Figure 2.

Serum neurotransmitter metabolism intermediates in young, old nonfrail and old frail human subjects. Data for young subjects are shown in green. Data for old nonfrail subjects are shown in blue. Data for old frail subjects are shown in red. Pathway involving N-acetyl-aspartyl-glutamate (NAAG) is shown in blue arrows. Pathway involving γ-aminobutyric acid (GABA) is shown in green arrows. *p < .05, **p < .01, ***p < .001, ****p < .0001.

We then used linear regression models to estimate the association between metabolite levels and cytokines (IL-6, IL-1β, IFNγ, TNFα, and TNFαR1) among young and older adults (Supplementary Table 4). After adjusting for age, the NAD⁺/NADH ratio and NAD⁺ levels were negatively correlated with IL-6 levels, while α-KG, NAAG, and pyruvate were positively correlated with IL-6. Aspartate, glutamate, dihydroxyacetone phosphate, succinate, ADP, and beta l-citryl-glutamate levels were all positively correlated with IL-1β levels. The NAD⁺/NADH ratio and NAD⁺ levels were negatively correlated with TNFα levels and NAAG was positively correlated with TNFα. NAAG, glutamate, GABA, oxaloacetate, glucose, cis-aconitate, aspartate, malate, and beta l-citryl-glutamate were positively correlated with TNFαR1 levels. Glutamate, pyruvate, aspartate, and dihydroxyacetone phosphate were negatively correlated with IFNγ levels, while glucose-6-phosphate was positively correlated with IFNγ levels. The associations between metabolite levels and cytokines among only the older adults in the study were largely similar (Supplementary Table 5).

We used logistic modeling to estimate the odds ratio of being frail versus nonfrail in relation to serum levels of each metabolite and linear regression models to estimate the association between metabolite levels and grip strength and walking speed (Table 2).Glutamate, pyruvate, glucose, beta l-citryl-glutamate, cis-aconitate, malate, aspartate, ADP, NAAG, succinic semialdehyde, oxaloacetate, NAD⁺, and GABA were all associated with increased risk of being frail. Succinic semialdehyde, glucose, pyruvate, and NAAG were negatively correlated with grip strength. Pyruvate, NAAG, cis-aconitate, glutamate, glucose, beta l-citryl-glutamate, and oxaloacetate were negatively associated with walking speed.

Table 2.

Odds Ratio From Logistic Regression of Frailty Status and Regression Coefficients From Linear Regression of Grip Strength and Walking Speed, and, Against Standardized Values of Metabolites, Individually, After Adjusting for Age, in Community-Dwelling Older Adults

Frailty Status			Grip Strength				Walking Speed
Metabolite	OR	p Value	Metabolite	Estimate	p Value	R 2	Metabolite	Estimate	p Value	R 2
Glutamate	2.32	.000	Succinic semialdehyde	−0.24	.006	.068	Pyruvate	−0.39	.002	.098
Pyruvate	2.89	.001	Glucose	−0.25	.006	.071	Glycerol 3-phosphate	0.29	.005	.089
Glucose	2.45	.002	Pyruvate	−0.30	.017	.052	NAAG	−0.35	.013	.072
BLCG	2.11	.002	NAAG	−0.26	.049	.047	cis-Aconitate	−0.26	.015	.113
cis-Aconitate	2.62	.002					Glutamate	−0.21	.021	.051
Malate	2.15	.003					Glucose	−0.21	.032	.050
Aspartate	2.03	.003					BLCG	−0.21	.039	.066
ADP	1.65	.015					Oxaloacetate	−0.18	.050	.059
NAAG	2.08	.016
Glycerol 3-phosphate	0.38	.019
Succinic semialdehyde	1.81	.021
Glucose 6-phosphate	0.45	.032
Oxaloacetate	2.15	.037
NAD+	1.59	.043
GABA	1.85	.047

Note: BLCG = beta l-citryl-glutamate; GABA = γ-aminobutyric acid; ADP = adenosine diphosphate; NAAG = N-acetyl-aspartyl-glutamate; OR = odds ratio.

While many metabolites were correlated with negative outcomes and inflammation, a few were associated with better physical performance and decreased frailty status. Glycerol 3-phosphate and glucose 6-phosphate were associated with a decreased odds ratio of becoming frail, and glycerol 3-phosphate was positively associated with walking speed.

A limitation of this study is that the results are cross-sectional, and no causality can be determined from the results. Despite the limitation, the results of this study suggest that there are specific and measureable differences in glucose metabolism, the TCA cycle, and neurotransmitter metabolism intermediates with aging and frailty. These results will likely further the goal of understanding dysregulation of metabolism in aging and physical frailty and improve frailty detection methods. It also provides more evidence for a biological signature of aging and frailty, which can be used to further study the metabolism-related mechanisms that underlie alterations in aging and in physical frailty. Summary statistics and pathway analysis for all compounds detected included in Supplementary Tables 6 and 7 and Supplementary Figures 3–5.

Funding

This work was supported by the Johns Hopkins University Claude D. Pepper Older Americans Independence Center funded by the National Institute on Aging of the National Institutes of Health (NIH) under award number P30AG021334 and NIH grants UH3 AG056933 (R.W. and P.M.A.), R01AG046441, and K23 AG035005; and Bright Focus Foundation Research Award (P.M.A.) and the Nathan W. and Margaret T. Shock Aging Research Foundation, Nathan Shock Scholar in Aging (P.M.A. and R.W.), The American Federation for Aging Research (R.W.), and the Shared Instrument Grant S10 (1S10OD025226-01) funded by the NIH (A.L.). Its contents are solely the responsibility of the authors and do not necessarily represent the official view of the NIH.

Conflict of Interest

None declared.

Author Contributions

R.W., A.L., J.W., and P.M.A. participated in study concept and design. P.M.A. is the PI of the study recruiting the subjects to the study and was responsible for the acquisition of the functional, demographic, and serum cytokine data. A.L., R.W., C.Z., and S.G. were responsible for the acquisition of the metabolomic data. Q.-L.X., R.W., A.L., J.W., P.M.A., C.Z., J.T., and S.G. participated in the analysis, interpretation of the data, and drafting the manuscript and approved the final version. R.W., Q.-L.X., A.L., J.W., P.M.A., and C.Z. were responsible for the critical revision of the manuscript for important intellectual content and approval of the final version. R.W., J.W., Q.-L.X., and P.M.A. had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.