Bringing doctors together

or

A free, private network for verified physicians

Leading voices

Dr. Dorairaj Prabhakaran

General Cardiology

View

Dr. Pugazhendhi Vijayaraman

EP · Geisinger

View

Dr. Robert Schaller

EP · HUP

View

Dr. Trey Woods

General Cardiology · Emory Healthcare

View

Dr. Brahmajee Nallamothu

Interventional Cardiology · UMHS

View

Dr. Felix Yang

EP · NYU Langone

View

Dr. Jennifer Rymer

Interventional & Peripheral Cards · Duke University

View

Dr. Ami Bhatt

Adult Congenital Heart Disease · MGH

View

Dr. Gregory Marcus

EP · UCSF

View

Dr. Connor O'Brien

Cardiac Critical Care · UCSF Medical Center

View
  • Discover your community

  • Stay current

  • A profile that represents you

  • Find your people

About RoonPrivacyTerms

    Overview

    N-Palmitoyl Glutamine Is a Candidate Mediator of Cardiorespiratory Fitness.

    Nov 12, 2025DOI: 10.1161/circulationaha.125.074187Journal ArticleAHA

    Written by: Jeremy M RobbinsMark BensonAnthony R P VerkerkeGaurav TiwariShuliang DengPrashant RaoUsman A TahirJulian Avila-PachecoXu ShiYuntian GuanFoje-Geh TendohJacob L BarberPatricia E MillerAndrew S PerryMichael E HallAlexis C WoodKent D TaylorWendy S PostStephen S RichMatthew NayorJames G WilsonGregory D LewisRavi V ShahJerome I RotterScott A SummersLaura M RaffieldShingo KajimuraClaude BouchardClary B ClishMark A SarzynskiRobert E Gerszten

    Find these authors on Roon

    PR
    Dr. Prashant Rao
    Sports Cardiology · BIDMC

    INTRODUCTION

    Cardiorespiratory fitness (CRF) reflects the body’s capacity to utilize oxygen in order to perform physical (locomotor) activity. This process requires integration across the cardiovascular, pulmonary, nervous, endocrine, and musculoskeletal systems in order to deliver fuel (oxygen) to the mitochondria for oxidative (aerobic) metabolism.1 Thus, CRF is truly a measure of systemic health.Moreover, CRF has emerged as one of the most powerful, independent predictors of survival.2–6 The ability of CRF to predict future risk of death from cardiovascular disease (CVD) and all-cause mortality outperforms traditional risk factors such as hypertension (HTN), obesity, type II diabetes mellitus as well as widely used composite risk scores, and its value and importance were recently highlighted in a Scientific Statement from the American Heart Association that recommended CRF be considered a clinical vital sign.7 Importantly, CRF can be increased through regular exercise and improvements in CRF are associated with increased survival.8,9 Despite its importance in human health, no targeted pharmacotherapies exist that improve CRF. Furthermore, while regular exercise may improve fitness, substantial inter-individual variability exists in CRF responses to exercise, even to standardized exercise training (ET) programs.10 Thus, identifying the biochemical pathways and molecular transducers that mediate CRF offers potential to improve human health.Mass spectrometry (MS)-based metabolite profiling of the circulation allows for the systematic identification of small molecules that reflect and/or contribute to health and disease states, and has proven to be a useful tool in characterizing metabolic responses to environmental stimuli such as physical activity and regular exercise.11,12 While plasma metabolites have long been known to act as physiologic signaling molecules13,14, there is also growing recognition of their role as mediators of the salutary effects of exercise as part of the emerging group of “exerkines”: compounds released by tissues in response to exercise that confer paracrine or endocrine health benefits.15,16 Further, advances in metabolomics technologies coupled with new bioinformatic approaches that integrate genetics and/or MS/MS spectral data have allowed for the identification of unknown MS peaks (“non-targeted metabolomics”), greatly enhancing the discovery potential of plasma metabolomics.17–19 Indeed, recent analyses suggest that >75% of the metabolite peaks associated with cardiometabolic traits represent unknown compounds that are often not well-correlated with known metabolites.18,20 Taken together, non-targeted metabolite profiling of the blood offers promise as a means to identify potential biomarkers and circulating mediators of CRF.Here, we perform non-targeted liquid chromatography (LC)-MS-based plasma metabolomics in 654 human participants with directly-measured maximal oxygen uptake (VO2max) in order to identify circulating metabolites related to CRF, and subsequently measure their responses to a 20-week supervised, endurance ET program. We identify several MS peaks highly associated with CRF, including a peak that we identify as N-palmitoyl glutamine (N-pal-gln), an N-acyl amino acid (NAA) that is positively associated with VO2max in two separate cohorts, and increases after aerobic ET. Because of the strong clinical association between CRF and survival, we test whether N-pal-gln is associated with mortality in two separate, community-based cohorts. Further, given the structural similarity between N-pal-gln and compounds that modulate mitochondrial function we tested its functional effects in model systems. Our studies highlight the value of non-targeted plasma metabolomics to identify circulating small molecules that may, in part, contribute to CRF, and identify N-pal-gln as a novel, candidate transducer.

    METHODS

    Written informed consent was provided for all human studies (HERITAGE Family Study, FHS, JHS, and MESA) and all analyses were approved by the Institutional Review Board at Beth Israel Deaconess Medical Center. A detailed description of the GWAS of plasma metabolites in JHS and HERITAGE has been published17 and whole genome association study summary statistics have been deposited in the GWAS catalog under accession code GCST90104476. JHS metabolomics data is available at dbGaP (phs000964.v5.p1). Data in support of the findings of this study are available from the corresponding author upon reasonable request. Detailed methods are provided in the Supplemental Material.Clinical characteristics of participants are reported as mean (± SD), proportions, or median (interquartile range). Two-sample Student’s t test, chi-square test, and Mann-Whitney U test, were used to compare differences, respectively. Metabolite levels were natural logarithmically transformed and standardized within batches. N-pal-gln was visually inspected to confirm a normal data distribution. Linear regression was performed to determine the relationship between metabolite values and VO2max (ml·kg−1·min−1). Metabolite levels were standardized to mean=0 and multiples of 1 SD. Paired student’s t tests were performed to quantify changes after ET. We used the Benjamini-Hochberg procedure to correct for multiple comparisons. Pearson correlation coefficients were calculated to compare the relationship between plasma metabolite levels.Mortality rates were estimated using the Kaplan-Meier method, with differences between tertiles of N-pal-gln levels assessed using the log-rank test. In JHS and MESA, the risk of death was quantified using multivariable Cox proportional hazards regression to estimate hazard ratios (HR) of batch-adjusted residuals of metabolite values after logarithmic transformation, standardization by batch and rank normalization across batch.The Student’s t test was used for pair-wise comparisons, and two-way ANOVA was used for isolated mitochondrial experiments.All statistical analyses were performed using either SAS 9.4 (SAS Institute, Cary, NC) or R version 4.1.2 (R Core Team, R Foundation for Statistical Computing, Vienna, Austria).

    RESULTS

    Baseline characteristics of the HERITAGE Study cohort are summarized in Table 1. The mean (SD) age of the HERITAGE participants (N = 654) was 35 (14) years; 361 (55%) were female, and 230 (35%) were self-identified Black. Mean (SD) baseline VO2max (mlkg−1min−1) was 31.5 (9.0). We measured 1886 unknown metabolite peaks in resting, fasted plasma samples taken at baseline and related them to VO2max (mlkg−1min−1) after adjustment for age, sex, race, and lean body mass. A total of 244 unknown MS peaks were associated with VO2max at an FDR-adjusted p-value (q-value) < 0.05 (Figure 1A and Supplemental Table 1).We focused our attention on circulating MS peaks that were positively associated with VO2max given their potential role as beneficial mediators of CRF. The metabolite with the strongest, positive association with VO2max was QI9825 (m/z: 385.3056, RT: 3.69; β = 1.29, q-value = 5.3 x 10−6, Figure 1A), a MS peak predicted to be N-palmitoyl glutamine (N-pal-gln) using MS/MS data according to SIRIUS+CSI:Finger ID version 4.7.2.21 We subsequently confirmed its identity using a synthetic chemical standard (ChemScene catalogue CS-0227092, CAS 58725-34-1, Figure 1B). Given known sex- and age-based differences in VO2max,22 we next stratified our analyses of this previously unstudied metabolite among males (N = 293) and females (N = 361), and by the parent (N = 221) and offspring (N = 433) generations, and tested for sex and generation by N-pal-gln interactions on VO2max. There were no interactions by sex (β = 1.11 and 1.36, p-value = 4.5 x 10−5 and 2.6 x 10−4 among females and males, respectively; p for interaction = 0.25) or generation (β = 1.32 and 1.06, p-value = 7.3 x 10−6 and 4.3 x 10−3 among offspring and parents, respectively; p for interaction = 0.45) on N-pal-gln’s association with VO2max (Figure 2).A subset of Framingham Heart Study (FHS) offspring cohort participants underwent cardiopulmonary exercise testing (CPET) and plasma metabolomics profiling using the same method performed in HERITAGE.12 We identified N-pal-gln in this dataset based on its chromatographic and mass spectrometry characteristics; it was previously annotated as an “unknown peak.” We tested the relationship between N-pal-gln and VO2max in 408 FHS participants with complete measures of N-pal-gln levels, VO2peak (a surrogate for VO2max), and lean body mass. Clinical characteristics of the FHS sample are summarized in Supplemental Table 2. The relationship between N-pal-gln and VO2peak was strikingly similar among FHS participants as compared to HERITAGE: N-pal-gln was strongly, positively associated with VO2peak in a linear regression model adjusted for age, sex, and lean body mass (β=1.2, p=3.8x10−5). There was no difference in the N-pal-gln-VO2peak relationship in men and women in FHS (p for interaction = 0.51).We next tested whether levels of N-pal-gln changed in response to the HERITAGE endurance training intervention. N-pal-gln levels increased substantially [log fold-change (FC)=0.22, q=5.3 x 10−12]. These increases occurred among both parents and offspring (logFC=0.29 and 0.18; p=0.0003 and 0.0002, respectively), and males and females (logFC=0.23 and 0.20; p=0.0001 and 0.0003, respectively).CRF has a strong, inverse, and independent relationship with all-cause mortality2–6,23, thus we subsequently tested whether N-pal-gln was in turn associated with survival in two community samples. Long-term health outcomes are not available in the HERITAGE Study, thus we tested this relationship in the Jackson Heart Study (JHS) and the Multi-Ethnic Study of Atherosclerosis (MESA), where N-pal-gln was previously measured - but not annotated - using the same plasma metabolomic profiling method as that used in HERITAGE.18 Clinical characteristics of JHS and MESA participants with complete data on N-pal-gln and covariates are shown in Supplemental Table 3.Among 2,495 available JHS participants, 599 (24%) died over 36,773 person-years of follow-up. In Cox proportional hazards analyses adjusted for age, sex, BMI, estimated glomerular filtration rate (eGFR), HTN, fasting plasma glucose (FPG), total cholesterol, HDL cholesterol, and tobacco use, N-pal-gln was inversely associated with incident all-cause mortality (HR: 0.91 per 1 standard-deviation [SD] increase; p-value = 0.029). Among 3,113 MESA participants, over 56,088 person-years of follow-up, 283 (9%) participants died. In Cox proportional hazards analyses adjusted for the same covariates as JHS, N-pal-gln was also inversely associated with incident all-cause mortality (HR: 0.65; p-value = 0.028). The risk of death decreased with increasing tertile of N-pal-gln level in both JHS and MESA; compared to individuals with the lowest tertile of N-pal-gln, participants with the highest tertile of N-pal-gln had a 30% and 37% lower risk of death in JHS and MESA, respectively (Figure 3).To further examine N-pal-gln’s relationship to palmitic acid and glutamine, we first correlated plasma levels measured on targeted LC-MS methods in HERITAGE (C18- and HILIC+ methods, respectively). N-pal-gln was weakly correlated to palmitic acid (Pearson’s correlation coefficient [r] = −0.07, p-value = 0.10) and moderately correlated with glutamine (r = 0.41, p-value = 6.5 x 10−30). To next assess whether glutamine may be, in turn, related to VO2max, we tested its relationship in the same linear model as N-pal-gln. Glutamine was not associated with VO2max (β = 0.21, p-value = 0.37); similarly, N-pal-gln’s relationship to VO2max was unchanged after the addition of glutamine as a covariate (β = 1.40, q-value = 3.57 x 10−6).We then tested whether N-pal-gln could form non-enzymatically in vitro in the presence of glutamine and palmitic acid. First, to confirm the purity of glutamine and palmitic acid authentic standards, both compounds were dissolved in optima-grade water or methanol and analyzed by LC-MS/MS. We found no glutamine/N-pal-gln signal in the palmitic acid sample, and no palmitic acid/N-pal-gln signal in the glutamine sample, thus confirming that the N-pal-gln signal in our samples was not from impurity of the standards (Supplemental Figure 1A). We next tested whether the addition of 100 uM of palmitic acid to varying concentrations of glutamine (100–500 uM) in water led to N-pal-gln synthesis at 0 min, 30 min, 60 min, 120 min and 240 min. LC-MS of the reaction sample showed an undetectable N-pal-gln signal. We subsequently spiked 100 uM, 200 uM and 500 uM of 15N2D5-glutamine in human pooled plasma and performed LC-MS at the same time intervals and were unable to detect 15N2D5-palmitoyl-glutamine, whereas endogenous N-pal-gln was present (Supplemental Figure 1B).Prior work has shown that several structurally similar metabolites and NAAs stimulate cellular respiration and energy expenditure in model systems,24,25 thus we sought to test N-pal-gln’s effects on cellular respiration and mitochondrial function to begin to understand its relationship to CRF. We first approximated N-pal-gln’s physiologic concentrations in human plasma by generating a standard curve using a synthetic standard spiked into plasma samples from healthy volunteers (N = 20) and determined the mean (SD) concentration to be 1.9 (0.36) ng/mL (Supplemental Figure 2). We next directly tested the effects of N-pal-gln administration on cellular respiration in differentiated C2C12 cells after 4 days of incubation with N-pal-gln (10 ng/mL). N-pal-gln treated cells demonstrated increased basal respiration and a trend towards increased maximally-stimulated (FCCP) respiration (Figures 4A and 4B). Given that increased basal respiration may represent either increased mitochondrial content or more efficient mitochondria, we next performed two lines of study in isolated mitochondrial experiments.First, we isolated mitochondria from differentiated C2C12 cells using a control (0 ng/mL), low (2.5 ng/mL), or high dose (10 ng/mL) N-pal-gln for 4 days, and found a dose-dependent increase in mitochondrial content among N-pal-gln treated cells compared to control (Supplemental Figure 3A). To confirm N-pal-gln’s effects on mitochondrial content, we quantified the effects of N-pal-gln on the mitochondrial DNA (mtDNA):nuclear DNA (nDNA) ratio and found a dose dependent increase in mtDNA/nDNA in N-pal-gln treated C12C12 myotubes relative to control after 4 days of incubation (15 and 20% increase among 2.5 ng/mL and 10 ng/mL N-pal-gln doses, respectively, Figure 4C). We found no difference in relative respiration in the N-pal-gln treated cells (Supplemental Figure 3B). Palmitic acid (26 nM), by contrast, had no effect on mtDNA/nDNA compared to control [data not shown; glutamine is present in DMEM (Gibco: 119650920) at 4mM].We next tested N-pal-gln’s effects on mitochondrial efficiency using high-resolution respirometry. Using varying titrations of ADP, we found that there was no difference in ATP production between N-pal-gln (26 nM) treated C2C12 myotubes compared to control (Figure 5A). However, there was a small but consistent reduction in mitochondrial respiration (Figure 5B), which resulted in increased ATP synthesis per oxygen molecule consumed (P/O ratio; Figure 5C), suggesting more efficient mitochondria.Finally, to further assess where N-pal-gln might be mediating its cellular effects, we performed LC-MS on fractionated C2C12 cells that had been treated with control, low- and high-dose N-pal-gln. We observed that N-pal-gln was found in higher relative concentrations in mitochondrial fractions than the nuclear and cytosolic compartments (Supplemental Figure 4).To begin to understand the genetic regulation of N-pal-gln, we leveraged a genome-wide association study (GWAS) of plasma N-pal-gln levels from the JHS (N=2,466) using whole genome sequencing.17 Overall, the heritability estimate for N-pal-gln was 0.49. We identified several variants that were highly associated with N-pal-gln, including the lead single nucleotide polymorphism (SNP), a missense variant in the ASAH1 locus (rs10103355, Chr8:18061425:A:G; beta estimate for effect allele, G: −0.49, p-value = 3.72 x 10−62; Figure 6). ASAH1 encodes a lysosomal ceramidase that hydrolyzes sphingolipid ceramides into sphingosine and free fatty acids, and has been shown to regulate energy expenditure, insulin and glucose homeostasis, and lipid metabolism with its effects likely mediated through ceramide regulation.26 Given that ASAH1 is not known to regulate N-pal-gln and its locus is in proximity to other genes that encode enzymes (e.g. NAT1 and NAT2), we further tested which SNPs were potentially causally related to N-pal-gln by fine-mapping 1 Mb from the transcriptional start site of our lead SNP, rs10103055, using SuSiE. We found one credible set (CS) that contained rs10103055 (posterior inclusion probability [PIP] = 1) suggesting that our lead SNP in the ASAH1 locus is likely causally related to N-pal-gln. While the SNP is common in African populations [minor allele frequency (MAF)=0.39], it is a rare variant in non-Finnish European populations (MAF=0.008) according to gnomAD v4.1.0.27 In addition, we performed GWAS of N-pal-gln in Black HERITAGE participants17 (N=658 participants with plasma N-pal-gln and genotyping available, 38% self-identified Black) and found a consistent direction of effect between rs10103355 and N-pal-gln (beta: −0.40, p-value = 0.00008).

    DISCUSSION

    CRF – and its gold-standard measurement, VO2max – captures multiple physiologic inputs from the heart, lungs, skeletal muscle, and nervous systems, and as such, provides a powerful snapshot of cardiopulmonary and metabolic health. Its value in risk stratifying individuals with and without disease regarding long-term health outcomes is firmly established,2,3,5 yet much remains unknown about the biochemical pathways and molecular transducers that underlie this complex trait. Here, we perform non-targeted plasma metabolomics profiling in individuals with directly-measured VO2max and identify a previously unannotated MS peak as N-palmitoyl glutamine, a lipidated amino acid whose functional effects have not previously been studied. N-pal-gln is positively associated with VO2max in two separate cohorts. We show that this circulating small molecule increases in response to regular aerobic exercise, and that its levels are inversely associated with future risk of death. We subsequently demonstrate that N-pal-gln stimulates mitochondrial biogenesis and increases mitochondrial efficiency in cell models. Taken together, our findings highlight the potential of non-targeted metabolomics profiling to discover new potential mediators of CRF, and characterize a previously unstudied lipidated amino acid as a candidate molecule.The identification of N-pal-gln as it relates to human CRF and exercise builds on prior work studying the role of NAAs as potential signaling molecules. NAAs constitute a large, structurally and physiologically diverse group of circulating, lipidated amino acids in which an amide bond links an amino acid to the acyl moiety of a long-chain fatty acid.28 Prior studies have shown that NAAs and N-acyl amides can serve as endogenous signaling moieties, including ligands for transient receptor potential vanilloid (TRPV) channels, G protein coupled receptors (GPCRs), and cannabinoid receptors that in turn are involved in nociception, vascular homeostasis, and sleep regulation in model systems.29–31 More recently, Long et al. demonstrated that several NAAs increased cellular oxygen consumption and uncoupling in differentiated white adipocytes in a UCP1-independent pathway, and that chronic treatment of N-Oleoyl leucine (C18:1-Leu), N-Oleoyl phenylalanine (C18:1-Phe), and N-arachidonoyl-glycine (C20:4-Gly) in diet-induced obese mice led to increases in indirect calorimetry-derived VO2 as well as reduced fat mass and improved glucose homeostasis.24While our group has recently shown that circulating levels of several NAAs are inversely associated with future risk of coronary heart disease and heart failure even after adjustment for free amino acid levels,32 to our knowledge these are the first data to show that a circulating NAA is related to human CRF, metabolism, and exercise response. Our findings that: 1) N-pal-gln is highly associated with VO2max even after adjustment for lean body mass and 2) N-pal-gln increases mitochondrial content and efficiency but not relative cellular respiration suggest that, in contrast to other NAAs described previously24, N-pal-gln’s effects on CRF do not proceed through mitochondrial uncoupling but rather through biogenesis and improved bioenergetics. Prior studies have shown that a closely related compound, N-palmitoyl ethanolamide, acts via nuclear receptors, including PPAR signaling pathways.33 Our findings that N-pal-gln is present in both nuclear and mitochondrial fractions may offer a possible means through which N-pal-gln stimulates mitochondrial biogenesis through the peroxisome proliferator-activated receptor-γ coactivator 1α (PGC-1α) pathway, itself a well-known exercise-induced biochemical pathway.34 Additional studies to understand whether N-pal-gln acts in a PGC-1α dependent manner, and how N-pal-gln improves mitochondrial efficiency are needed to further characterize its functional effects on CRF. In addition, several outstanding questions remain including what the tissue sources for circulating N-pal-gln are both at rest and during exercise, how acute exercise and resistance exercise affect plasma N-pal-gln levels, and whether changes in N-pal-gln levels are associated with changes in CRF. Efforts to address these questions in N-pal-gln and other plasma metabolites may be addressed by the ongoing studies from the NIH’s Molecular Transducers of Physical Activity Consortium (MoTrPAC).35Our findings that over 200 unknown MS peaks - including N-pal-gln – are highly associated with CRF highlight the enormous discovery potential of non-targeted metabolomics. Increasingly, non-targeted metabolomics profiling in the circulation has produced new markers of future metabolic- and cardiovascular diseases, many of which lie in orthogonal biochemical pathways to known metabolites,17,18 offering an attractive platform to study human exercise and, in particular, search for exercise-responsive bioactive mediators of exercise’s salutary effects. While peak identification remains an analytical and time-consuming challenge, several approaches exist to help aid in this process.One approach leverages MS/MS spectral data and chemical biology structural databases such as SIRIUS, GNPS, CANOPUS to help elucidate MS peak identity.21,36,37 Here, we predicted the unknown MS peak to be N-pal-gln using SIRIUS and subsequently confirmed its identity with an authentic chemical standard, a necessary step in peak identification. While over 70 N-acyl amides have been identified in nature38, limited data exist about this family’s relationship to exercise and CRF. Annotating existing, non-targeted metabolomics datasets using authentic chemical standards may help further our understanding of this biochemical class as they relate to human fitness and metabolism, a task limited by the availability of specific standards.An alternative approach integrates human genetics with plasma metabolomics to inform a metabolite’s identification; for example, the discovery of genetic variants in or near a gene [i.e. metabolite quantitative trait loci (mQTL)] encoding an enzyme that is highly associated with an MS peak may point to it as a putative substrate or product.19,39,40 Our identification of several mQTLs in ASAH1 for N-pal-gln, including a missense mutation, is interesting in that ASAH1 is not known to regulate NAAs but rather hydrolyze sphingolipid ceramides into sphingosine and free fatty acids. ASAH1 has been shown to have bidirectional enzymatic activity in humans,41 and we hypothesize that ASAH1 serves as a transferase that moves the palmitoyl group from ceramides to other amino acids. Additional mechanistic work is needed to determine ASAH1’s enzymatic role in N-pal-gln metabolism and how plasma N-pal-gln is otherwise regulated at the tissue/cellular level. Further, our GWAS findings of N-pal-gln in the JHS highlights the importance of including multi-ethnic populations in genetic and genomic investigations to identify new biology. In this case, while the association between ASAH1 and N-palmitoyl glutamine in our cohort was enabled by the higher allele frequency in Black individuals, we believe the underlying biology and its relevance to cardiorespiratory fitness extends to all populations.There are limitations to our work. HERITAGE is a single-arm ET study; thus, changes in plasma metabolites after the training intervention are relative to a participant’s baseline and the potential remains for unmeasured factors to influence these results. However, HERITAGE was a supervised, in-lab ET program where participants demonstrated amongst the largest gains in VO2max (~17% increase) in any large-scale human study, a change that is accounted for by the robust ET stimuli. We cannot identify the tissue source of circulating N-pal-gln or other metabolites on the platform as mentioned above; tissue-specific profiling from the MoTrPAC Study may help inform our understanding in key, exercise-relevant tissues.42In summary, these data identify a novel, lipidated amino acid that is exercise responsive, associated with long-term human health outcomes, and a candidate molecular transducer of CRF.

    Supplementary Material

    Expanded MethodsFigures S1–4Tables S1–3References 43–55
    PRDr. Prashant Rao

    Dr. Prashant Rao

    Cardiology · Sports & Performance Cardiology
    Beth Israel Deaconess Medical Center
    prashant