Sex differences in body composition and serum metabolome responses to sustained, physical training suggest enhanced fat oxidation in women compared with men

Abstract

Sex differences in energy metabolism during acute, submaximal exercise are well documented. Whether these sex differences influence metabolic and physiological responses to sustained, physically demanding activities is not well characterized. This study aimed to identify sex differences within changes in the serum metabolome in relation to changes in body composition, physical performance, and circulating markers of endocrine and metabolic status during a 17-day military training exercise. Blood was collected, and body composition and lower body power were measured before and after the training on 72 cadets (18 women). Total daily energy expenditure (TDEE) was assessed using doubly labeled water in a subset throughout. TDEE was greater in men (4,085 ± 482 kcal/d) than in women (2,982 ± 472 kcal/d, P < 0.001), but not after adjustment for dry lean mass (DLM). Men tended to lose more DLM than women (mean change [95% CI]: −0.2[−0.3, −0.1] vs. −0.0[−0.0, 0.0] kg, P = 0.063, Cohen’s d = 0.50) and have greater reductions in lower body power (−244[−314, −174] vs. −130[−209, −51] W, P = 0.085, d = 0.49). Reductions in DLM and lower body power were correlated (r = 0.325, P = 0.006). Women demonstrated greater fat oxidation than men (Δfat mass/DLM: −0.20[−0.24, −0.17] vs. −0.15[−0.17, −0.13] kg, P = 0.012, d = 0.64). Metabolites within pathways of fatty acid, endocannabinoid, lysophospholipid, phosphatidylcholine, phosphatidylethanolamine, and plasmalogen metabolism increased in women relative to men. Independent of sex, changes in metabolites related to lipid metabolism were inversely associated with changes in body mass and positively associated with changes in endocrine and metabolic status. These data suggest that during sustained military training, women preferentially mobilize fat stores compared with men, which may be beneficial for mitigating loss of lean mass and lower body power.

NEW & NOTEWORTHY Women preferentially mobilize fat stores compared with men in response to sustained, physically demanding military training, as evidenced by increased lipid metabolites and enhanced fat oxidation, which may be beneficial for mitigating loss of lean mass and lower body power.

INTRODUCTION

Arduous military field training exercises provide opportunities for gaining insight into changes in human physiology under unique combinations of physical and environmental stress. For example, military training exercises lasting multiple days require high energy expenditures that elicit energy deficits that result in losses of total body mass, fat mass, and fat-free mass, and may contribute to decrements in physical performance (1–4). However, though generally considered to be well-established, these responses to military training have largely only been characterized in men (1).

Underrepresentation of women signifies a substantial knowledge gap given that sex differences in physiological and metabolic responses to physically demanding activities exist (5–7). During acute endurance exercise, women demonstrate greater fat oxidation compared with men, likely due to increased free fatty acid (FFA) availability, greater capacity to transport FFA, and/or differences in sex hormones (6, 8–10). In particular, 17-β-estradiol spares glycogen stores by shifting metabolism toward FFA at low-to-moderate exercise intensities, providing women a greater capacity to oxidize more lipids and less carbohydrates than men during prolonged aerobic exercise (11). To our knowledge, only two studies have investigated sex differences in changes in body composition during prolonged, strenuous military training (3, 12). In both studies, men demonstrated greater losses of total body mass and fat-free mass than women over the 7-day training exercises, with nearly all weight loss in women attributable to fat mass (3, 12). Although these studies support a higher rate of fat oxidation during prolonged, strenuous military training in women compared with men, sample sizes were small, and any impact of sex-specific metabolic responses to military training on physical performance remains elusive.

Metabolomics provides a comprehensive snapshot of metabolism that can detect metabolic signatures of physiological perturbations, such as energy deficit and physical activity (13–15). The approach can also provide insight into the regulation of metabolic pathways that may help interpret previous observations of sex differences in body composition changes during military training. Strict fasting has been shown to increase nonesterified fatty acids and branched-chain amino acids (16), whereas metabolomic responses to exercise vary based on modality, duration, and fitness level, but generally demonstrate changes in fatty acid metabolism, glycolysis, amino acid metabolism, and energy metabolism (13, 14, 17, 18). The combination of energy deficit and high energy expenditure during military training has been associated with marked changes in the plasma metabolome that reflected increases in energy metabolism, lipolysis, fatty acid oxidation, and ketogenesis in men (19). Whether the magnitude of these changes differs by sex is unclear. However, sex differences in metabolomic profiles at rest (20, 21) suggest that sex may impact changes in the metabolome during prolonged, strenuous training.

The increasing integration of women into combat occupations within the United States Armed Services and militaries worldwide makes improving knowledge of sex differences in metabolism and substrate utilization important for determining whether sex-specific recommendations during and following arduous military training and operations may be needed (5, 7). This improved knowledge would also inform guidelines in nonmilitary tactical populations (e.g., firefighters and other rescue professions) and ultra-endurance athletes where women continue to be understudied relative to men. The objective of this study was to identify sex differences within changes in the serum metabolome in relation to changes in body composition, physical performance, and circulating markers of endocrine and metabolic status during a 17-day military training exercise. We hypothesized that increases in markers of lipolysis and fatty acid metabolism would be greater in women than in men, whereas increases in markers of amino acid catabolism would be greater in men than women during the military training, and that these changes would be associated with changes in body composition and endocrine status.

METHODS

Volunteers and Experimental Design

Cadets enrolled in Cadet Leader Development Training (CLDT) in July 2021 at the United States Military Academy (USMA), West Point, New York were recruited for the study. Exclusion criteria included any injury or health condition that limited full participation in CLDT, or self-reported pregnancy within the past year. Ninety-three cadets volunteered and were determined eligible to participate. Nineteen withdrew for undisclosed reasons before any data being collected, one withdrew immediately after the PRE data collection time point, and one was withdrawn during training due to a family emergency. Of the 72 cadets (54 men and 18 women) who completed the study, one provided only partial data at the POST time point due to an injury sustained during training.

The 17-day training is mandatory for cadets finishing their 3rd year at the USMA, though cadets finishing their 2nd year can also enroll. CLDT is designed to develop and assess leadership capabilities through skills training and a tactical field exercise. The first 7 days of the 2021 CLDT course were focused on military skills training (e.g., rifle marksmanship and motorized operations), conducting military movements (e.g., ambushes, raids, defenses, and patrols), and completing an abbreviated 2-day field training exercise to practice the military skills and movements learned during the first 7 days, before completing the full-field training exercise. Cadets then completed 1 day of preparation for field training, followed by a 9-day field training exercise designed to be both physically and cognitively demanding. During the 9-day field training, cadets slept in the field 4–6 h/night, carried ∼18–23 kg, traversed ∼60 km, were provided three Meal, Ready-to-Eat military rations daily, and completed several training events meant to simulate military attacks, defenses, ambushes, raids, motor operations and movements to the enemy. Major data collection activities occurred under resting conditions, before the 2-day field training exercise (PRE), and immediately following the completion of the 9-day field exercise (POST) (Fig. 1). Time of day for data collection activities varied based on each volunteer’s schedule and availability at PRE and as volunteers completed the field exercise at POST.

Figure 1.

Experimental design. The 17-day military training was composed of 7 days of military skills training (Skills), including an abbreviated 2-day field training exercise (Mini-FTX), followed by 1 day of preparation (Prep), and culminated with a 9-day field training exercise (Full-FTX). Urine samples were collected throughout training in a subset of participants to assess total daily energy expenditure using doubly labeled water. Blood collection, bioelectrical impedance, and vertical jump measurements were conducted before (PRE) and upon completion (POST) of training. [Image created with BioRender.com and published with permission].

This study was approved by the Headquarters U.S. Army Medical Research and Development Command (Ft. Detrick, Fredericksburg, MD). All subjects provided written informed consent before participating in any research activities. Investigators adhered to the policies regarding the protection of human subjects as prescribed in Army Regulation 70-25.

Body Composition

Body mass and composition were assessed at PRE and POST using bioelectrical impedance (InBody 770, Cerritos, CA). Participants were instructed to remove socks, clean their feet and hands using cleansing wipes, step on the electrode of the foothold, and hold the electrode handlebars. The measurement took ∼ 5 min to complete and has been documented as a valid and accurate measurement of body composition (22). Measures of body composition included: fat mass, body fat percentage, dry lean mass (DLM), and total body water. Fat oxidation was calculated as the change in fat mass from PRE to POST divided by the initial amount of DLM, as reported by Hoyt et al. (12). Vertical height was assessed in duplicate at PRE to the nearest 0.1 cm using a stadiometer (Seca: Creative Health Products, Plymouth, MI).

Energy Expenditure

Energy expenditure was determined using the doubly labeled water (DLW) method in a subset of 60 participants (42 men and 18 women). Of those, three male cadets could not be included in the analysis due to study withdrawal (n = 1) or incomplete sample collections (n = 2). Dosing occurred at 2100–2200. Immediately before drinking the DLW (0.23 g H₂ ¹⁸O/kg body mass and 0.15 g ²H₂O/kg body mass: Sigma Aldrich, St. Louis, MO), participants provided a urine sample to determine the natural abundance of ¹⁸O and ²H. After consuming the DLW, participants fasted overnight and then collected their first-morning void to determine peak enrichment. Two male participants were randomly chosen to consume locally available water to control for changes in ²H and ¹⁸O abundance, and local water sources were analyzed to determine isotopic enrichments. Participants were instructed to collect their first-morning void approximately every other day during CLDT to determine the elimination rates over time. If the first-morning void was missed, participants were asked to collect a sample later that same day. Total body water was calculated by determining the regression line for the elimination of ²H and ¹⁸O and extrapolated to a maximum enrichment. Enrichments of ²H and ¹⁸O were measured using isotope ratio mass spectroscopy (Finnigan Mat 252, Thermo Fisher Scientific, Waltham, MA). The ²H and ¹⁸O isotope elimination rates (k_H and k_O) were calculated by linear regression, which allowed for the inclusion of participants who failed to provide urine samples on one or more collection days. Determination of CO₂ production to calculate energy expenditure was determined according to Schoeller et al. (23):

$${\text{rCO}}_{\text{2}}(\text{moL·}{\text{d}}^{-\text{1}})=\left(\text{N}/2.078\right)\left(1.10{\text{K}}_{\text{O}}-\hspace{0.17em}1.04{\text{K}}_{\text{H}}\right)-0.0{\text{246rH}}_{\text{2}}{\text{O}}_f$$

where N is total body water; K_O and K_H are ¹⁸O and ²H isotope disappearance rates, respectively; and rH₂O_f is the rate of fractionated evaporated water loss, estimated to be 10.5 N × (1.01 K_O – 1.04 K_H). Energy expenditure was calculated using the energy equivalent of CO₂ for a respiratory quotient of 0.86.

Lower Body Power

Lower body power was assessed at PRE and POST using a field expedient vertical jump test with Vertec (Sunnyvale, CA) and the Just Jump System (Perform Better, West Warwick, RI). Participants stood on the Just Jump System mat and the Vertec was placed directly above the participant as a target for aim. Participants were instructed to perform a small counter-movement followed by an explosive jump and arm swing to maximize vertical projection. Instruction included a demonstration and verbal feedback focused on technique. Participants conducted up to two familiarization jumps not used for data analysis, followed by three jumps for maximal height used for analysis. Participants were provided at least 1 min of rest between jumps. Jump height was determined based on flight time, measured by microswitches within the mat that are sensitive to lift-off and landing of the feet, using the following equation. (24):

$$\text{Jump height}=[t^{\text{2}}\mathrm{\times }g]/\text{8}$$

where g = 9.81 m/s^s (acceleration due to gravity) and t = flight time. Peak lower body power was determined as the highest power generated from a single jump based on jump height and body mass (25):

$$\text{Peak power }\left(\text{watts}\right)=[\text{6}0.\text{7}\times \text{jump height }\left(\text{cm}\right)]+[\text{45}.\text{3}\times \text{body mass }\left(\text{kg}\right)]-\text{2},0\text{55}$$

Average lower body power was determined as the average power generated from three jumps.

Endocrine and Metabolic Status

Blood samples were collected via venipuncture at PRE and POST, and immediately processed, transported on dry ice, and stored frozen until analysis. Serum estradiol (sensitivity: 20 pg/mL), progesterone (0.2 ng/mL), and total testosterone (20 ng/dL) were assessed via immunoassays (Siemens Immulite 2000, Siemens Medical Solutions, Malvern, PA). Serum-free testosterone (0.17 pg/mL) was calculated according to the method of Vermeulen et al. (26). Serum glycerol (0 mg/L) was measured by colorimetric assay and free fatty acids (FFAs; 0 µM) were measured by fluorometric assay (Cayman, Ann Arbor, MI). Serum leptin (0.4 ng/mL) was assessed using a radioimmunoassay (Millipore Corporation, St. Louis, MO). All samples were run in duplicate with intraassay coefficients of variation of 10% or less, with the exceptions of estradiol (11.4%) and FFA (17.0%).

Metabolomics

Blood collected into serum tubes was processed, then stored frozen until shipped on dry ice to Metabolon Inc. (Durham, NC) for analysis. A detailed description of metabolomics analysis has been previously reported (19, 27). Briefly, samples were analyzed by applying two separate reverse-phase (RP)/ultra-high performance liquid chromatography (UPLC)-tandem mass spectrometry (MS-MS) methods with positive ion mode electrospray ionization (ESI), an RP/UPLC-MS/MS method with a negative ion mode ESI, and a hydrophilic interaction chromatography (HILIC)/UPLC-MS/MS with a negative ion mode ESI. All analyses were conducted using a Waters ACQUITY UPLC (Waters Corp., Milford, MA) and a Thermo Scientific Q-Exactive high-resolution/accurate mass spectrometer interfaced with a heated ESI-II source and Orbitrap mass analyzer operated at 35,000 mass resolution. To ensure quality control, several recovery standards, internal standards, technical replicates, and blanks were analyzed with experimental samples. Raw data were extracted, and peaks were identified based on comparing retention times, mass-to-charge ratios, and chromatographic data stored within a library managed by Metabolon that contains entries of purified standards and recurrent unknown entities. Peaks were quality control processed using proprietary hardware and software (Metabolon, Inc.) and quantified using the area under the curve, which was used for statistical analyses.

Statistical Analysis

Analyses were conducted using SPSS v.26 (IBM Analytics; Armonk, NY) and MetaboAnalyst v.5.0 (28). A priori sample size calculations using means, standard deviations, and effect sizes from previous reports (12, 29, 30) estimated that 29 men and 29 women would be needed to detect large effect sizes for sex differences in adjusted TDEE and changes in lower body power, respectively.

Energy expenditure and changes in body composition, lower body power, and circulating hormones during CLDT were compared between men and women by independent sample t test or Mann–Whitney U test (p_MW) if assumptions of normality were not met. Cohen’s d was calculated as a measure of effect size for sex differences in changes in body composition, lower body power, and circulating hormones. Prior to the analysis of metabolites, data were median-scaled, missing values were imputed using 1/5 of the minimum observed median-scaled value for each compound, and metabolites demonstrating near-constant values across samples were filtered from the data set using robust estimates of interquartile ranges. For sex comparisons, the log₂ fold change [log₂(post/pre)] of each metabolite was calculated and range scaled (mean-centered and divided by the value range of each variable). For PRE versus POST comparisons, data were log₁₀ transformed. Metabolite profiles were analyzed using principal components analysis and orthogonal projections to latent structures squares discriminant analysis (OPLS-DA). Due to suspected unequal variances between men and women resulting from differences in sample size, nonparametric statistical analyses were used for group comparisons. Sex differences in log₂ fold changes of individual metabolites were determined by Mann–Whitney U test, and PRE versus POST differences in individual metabolites were determined by the Wilcoxon rank test.

Associations between metabolites within lipid, carbohydrate, amino acid, peptide, and energy superpathways that significantly changed PRE to POST (Q < 0.2) and changes in circulating hormones, metabolic markers, body composition, and energy expenditure were assessed using Spearman’s rank correlation. Associations between changes in DLM and lower body power were assessed using Pearson’s correlation. The false discovery rate was controlled using the Benjamini–Hochberg procedure for all metabolomics analyses with adjusted P values reported as Q-values. Significance was set at P ≤ 0.05 and Q ≤ 0.2, and P ≤ 0.10 was considered as evidence of a tendency toward an effect.

RESULTS

PRE height, body mass, DLM, fat mass, body fat percentage, and concentrations of estradiol, total and free testosterone, and leptin differed by sex (Table 1). Of the 18 women in this study, six reported using hormonal contraceptives. Three women reported using an etonogestrel implant, one reported using an oral contraceptive, one reported using a hormonal intrauterine device, and one reported using a vaginal ring.

Table 1.

Baseline characteristics of cadets enrolled in Cadet Leader Development Training

	Men (n = 54)	Women (n = 18)
Demographics
Age, yr	21.7 ± 1.4	21.4 ± 1.2
Height, cm	177.9 ± 7.1	164.7 ± 5.1*
Body mass, kg	84.7 ± 11.1	71.6 ± 10.6*
Dry lean mass, kg	19.5 ± 2.3	14.2 ± 1.3*
Fat mass, kg	12.2 ± 5.6	18.6 ± 7.6*
BMI, kg/m2	26.7 ± 3.1	26.2 ± 3.8
Body fat, %	14.2 ± 5.1	25.3 ± 7.0*
Endocrine status
Estradiol, pg/mL	17.3 ± 10.2	49.6 ± 45.1*
Progesterone, ng/mL	0.5 ± 0.2	1.0 ± 2.0
Total testosterone, ng/dL	291.3 ± 109.6	19.1 ± 14.6*
Free testosterone, pg/mL	10.1 ± 4.7	2.8 ± 1.7*
Metabolic status
Leptin, ng/mL	6.7 ± 4.6	23.1 ± 14.9*
Glycerol, mg/L	8.7 ± 4.6	6.2 ± 2.3*
Free fatty acids, µ/M	90.5 ± 114.8	100.5 ± 104.2

Values are represented as means ± SD. BMI, body mass index. Values compared by Mann–Whitney U test. *Significantly different between men and women (P < 0.05).

Energy Expenditure and Body Composition

TDEE during CLDT was greater in men compared with women (P < 0.001, Fig. 2A); however, TDEE relative to initial DLM did not differ by sex (P = 0.725, Fig. 2B). During CLDT, men lost more total body mass than women (P = 0.030, d = 0.57, Fig. 3A, Supplemental Table S1, https://doi.org/10.6084/m9.figshare.21701951) and demonstrated a tendency to lose more DLM (P = 0.063, d = 0.50, Fig. 3B). Changes in fat mass did not differ between men and women (p_MW = 0.672, d = 0.07, Fig. 3C), but women lost more fat per kilogram of initial DLM compared with men (p_MW = 0.012, d = 0.64, Fig. 3D). The proportion of total body mass loss attributable to fat mass did not differ between sexes (mean [95% CI]: men: 74.5 [66.3, 83.7] %, women: 85.8 [76.2, 95.3] %, p_MW = 0.219, d = 0.41). However, men demonstrated a tendency to lose a greater proportion of total body mass from DLM (men: 9.7 [6.4, 13.0] %, women: 4.3 [1.4, 7.2] %, p_MW = 0.059, d = 0.50). Changes in total body water did not differ between sexes (P = 0.151, d = 0.40, Fig. 3E) or over time within the full cohort (0.1 [−0.4, 0.2] kg, P = 0.290).

Figure 2.

Total daily energy expenditure (TDEE) during Cadet Leader Development Training (CLDT). TDEE was assessed via doubly labeled water in men (n = 37) and women (n = 18). Independent sample t test was used to compare energy expenditure between men and women. Bars represent mean and 95% confidence interval. Men had a significantly (*P < 0.001) greater TDEE compared with women (A). TDEE relative to initial dry lean mass (DLM) did not differ between men and women (B).

Figure 3.

Changes in body composition from pre- to post-Cadet Leader Development Training (CLDT). Body composition was assessed in men (n = 54) and women (n = 18) pre- and post-CLDT. Independent sample t test was used to compare changes in body composition between men and women. Mann-Whitney U test was used if assumptions of normality were not met. Lines and error bars represent mean change and 95% confidence interval. Individual data points are presented in gray. *Significantly different between men and women (P < 0.05). Men lost significantly more total body mass (A) and demonstrated a tendency to lose more dry lean mass (DLM) (B) than women. Changes in fat mass did not differ between sexes (C), though fat oxidation per kilogram (kg) of DLM, calculated as Δ fat mass/initial DLM, was greater in women than in men (D). Men and women exhibited minimal changes in total body water (E).

Lower Body Power

Changes in lower body power following CLDT did not differ between sexes, though men demonstrated a tendency to have greater declines in peak lower body power (−244 [−314, −174] vs. −130 [−209, −51] W, P = 0.085, d = 0.49) and average lower body power (−264 [−321, −208] vs. −169 [−243, −95] W, P = 0.079, d = 0.50) compared with women (Supplemental Table S1, https://doi.org/10.6084/m9.figshare.21701951). Greater declines in peak and average lower body power were associated with greater declines in dry lean mass during CLDT (peak: r = 0.325, P = 0.006, Fig. 4A; average: r = 0.266, P = 0.026, Fig. 4B).

Figure 4.

Associations between declines in dry lean mass (DLM) and lower body power were assessed using Pearson’s correlation. Greater decreases in DLM were significantly associated (P < 0.05) with greater declines in peak lower body power (A) and average lower body power (B) following Cadet Leader Development Training (CLDT). Individual data points represent a single individual (n = 70). The solid green line represents line of best fit based on simple linear regression and shaded green region represents the 95% confidence interval.

Endocrine and Metabolic Status

Changes in concentrations of estradiol (p_MW = 0.321, d = 0.10, Fig. 5A) and progesterone (p_MW = 0.189, d = 0.71, Fig. 5B) during CLDT did not differ by sex. Men demonstrated increases in total (p_MW = 0.009, d = 0.57, Fig. 5C) but not free testosterone (P = 0.989, d = 0.003, Fig. 5D) relative to women (Supplemental Table S1, https://doi.org/10.6084/m9.figshare.21701951). Changes in glycerol (p_MW = 0.750, d = 0.04, Fig. 5E) and FFA (p_MW = 0.634, d = 0.06, Fig. 5F) did not differ between the sexes. Women had a greater decline in circulating leptin compared with men (p_MW < 0.001, d = 1.54, Fig. 5G).

Figure 5.

Changes in circulating sex hormones, free fatty acids (FFAs), glycerol, and leptin from pre- to post-Cadet Leader Development Training (CLDT). Hormones were assessed in men (n = 54) and women (n = 17) pre- and post-CLDT. Independent sample t test was used to compare changes in hormones between men and women. Mann-Whitney U test was used if assumptions of normality were not met. Lines and error bars represent mean change and 95% confidence interval. Individual data points are presented in gray. *Significantly different between men and women (P < 0.05). Men and women demonstrated no differences in changes of circulating estradiol (A) or progesterone (B). Men exhibited an increase in testosterone (T) (C), but not free testosterone (D), compared with women. No differences were observed in changes in glycerol (E) or free fatty acids (FFAs) (F). A greater decrease in leptin (G) was observed in women compared with men.

Metabolomics

Principal components analysis of the metabolome PRE and POST (Fig. 6A) as well as the log₂ fold change of all metabolites (Fig. 6B) did not reveal any noticeable separation between men and women. Univariate analysis of the log₂ fold change in the metabolome indicated that 78 metabolites differed between men and women (P < 0.05) before false discovery rate adjustment, of which 39 metabolites (50%) were related to lipid metabolism (Supplemental Table S2, https://doi.org/10.6084/m9.figshare.21702017). However, principal component analysis of all metabolites within the lipid superpathway PRE and POST (Fig. 6C) and the log₂ fold change of lipid metabolites (Fig. 6D) demonstrated no separation between men and women. Univariate analysis of log₂ fold changes revealed that changes in 18 of the 524 lipid metabolites demonstrated sex differences (P < 0.05, Q < 0.2; Table 2). The largest difference was observed for dihomo-linoleoylcarnitine (C20:2), a metabolite related to acylcarnitine polyunsaturated fatty acid metabolism, which increased in women relative to men. Increases in metabolites of the endocannabinoid, lysophospholipid, phosphatidylcholine (PC), phosphatidylethanolamine (PE), and plasmalogen lipid subpathways were also observed in women relative to men. In contrast, metabolites related to acylcarnitine dicarboxylate and acyl glycine fatty acid metabolism, as well as ceramides, were decreased in women following CLDT relative to men.

Figure 6.

Changes in plasma metabolites during military training. Metabolites were assessed in men (n = 54) and women (n = 17) before (pre) and after (post) Cadet Leader Development Training (CLDT). A: principal component (PC) analysis of all metabolites. B: PC plots of the log₂ fold change from pre- to post-CLDT for all metabolites. C: PC plot of lipid metabolites. D: PC plots of the log₂ fold change from pre- to post-CLDT for lipid metabolites. Individual data points represent the metabolome or changes in the metabolome within a single individual. Data points in closer proximity to one another are more similar.

Table 2.

Significant differences in lipid metabolism between men (n = 54) and women (n =17) following military training

Subpathway	Biochemical Name	Log2FC Men	Log2FC Women	Q-Value
Ceramides	Ceramide (d18:1/14:0, d16:1/16:0)*	0.03 ± 0.15	−0.10 ± 0.18	0.176
Endocannabinoid	N-Stearoyltaurine	–0.04 ± 0.22	0.12 ± 0.24	0.176
	N-Palmitoyltaurine	–0.05 ± 0.30	0.15 ± 0.24	0.183
Fatty acid metabolism (acylcarnitine, dicarboxylate)	Pimeloylcarnitine/3-methyladipoylcarnitine (C7-DC)	0.04 ± 0.21	–0.12 ± 0.18	0.162
Fatty acid metabolism (acyl carnitine, polyunsaturated)	Dihomo-Linoleoylcarnitine (C20:2)*	–0.05 ± 0.26	0.16 ± 0.22	0.176
Fatty acid metabolism (acyl glycine)	cis-3,4-Methyleneheptanoylglycine	0.05 ± 0.19	–0.15 ± 0.21	0.123
Lysophospholipid	1-Arachidoyl-GPC (20:0)	–0.04 ± 0.18	0.13 ± 0.18	0.123
	1-Stearoyl-GPE (18:0)	–0.04 ± 0.20	0.12 ± 0.18	0.176
	1-Eicosapentaenoyl-GPE (20:5)*	–0.04 ± 0.23	0.11 ± 0.18	0.176
	1-Behenoyl-GPC (22:0)	–0.03 ± 0.18	0.10 ± 0.19	0.176
	1-Palmitoyl-GPI (16:0)	–0.04 ± 0.20	0.11 ± 0.15	0.176
	2-Palmitoyl-GPC (16:0)*	–0.04 ± 0.23	0.14 ± 0.23	0.176
Phosphatidylcholine (PC)	1-Myristoyl-2-arachidonoyl-GPC (14:0/20:4)*	–0.04 ± 0.23	0.11 ± 0.17	0.176
Phosphatidylethanolamine (PE)	1-Palmitoyl-2-docosahexaenoyl-GPE (16:0/22:6)*	–0.04 ± 0.17	0.12 ± 0.09	0.015
	1-Stearoyl-2-docosahexaenoyl-GPE (18:0/22:6)*	–0.03 ± 0.16	0.08 ± 0.06	0.065
	1-Stearoyl-2-arachidonoyl-GPE (18:0/20:4)	–0.03 ± 0.20	0.11 ± 0.12	0.162
	1-Palmitoyl-2-oleoyl-GPE (16:0/18:1)	–0.04 ± 0.21	0.12 ± 0.19	0.183
Plasmalogen	1-(1-Enyl-Stearoyl)-2-arachidonoyl-GPE (P-18:0/20:4)*	–0.05 ± 0.23	0.15 ± 0.21	0.176

Data are represented as means ± SD log₂ fold change (Log₂FC). Mann–Whitney U test was used to compare changes in lipid metabolites between men and women following military training. P values were adjusted using the Benjamini–Hochberg correction (Q). Only significantly (P < 0.05, Q < 0.2) different lipid metabolites are shown. Metabolites in bold are those that increased in women relative to men; all others are significantly increased in men relative to women. *Compound that has not been confirmed based on a standard but identified with high confidence.

Principal components analysis (Fig. 6, A and C) and subsequent OPLS-DA demonstrated clear separation between PRE and POST samples. OPLS-DA differentiated PRE from POST samples with 90.4% accuracy (P < 0.01, Fig. 7A) and explained 97.6% of the variance (P < 0.01). The top 22 metabolites responsible for the greatest separation are shown in Fig. 7B. Eleven of the top 22 metabolites (50%) were related to lipid metabolism, two to amino acid metabolism, two to nucleotide metabolism, 2 to tricarboxylic acid (TCA) metabolism, and 4 were unidentified metabolites (Fig. 7B).

Figure 7.

Changes in plasma metabolite profiles during Cadet Leader Development Training (CLDT). Orthogonal Projections to Latent Structures Squares Discriminant Analysis (OPLS-DA) (n = 71) (A) was used to identify the top 22 metabolites ranked based on the variable importance in projection (VIP) score (B). Arrows indicate direction of change from pre- to post-CLDT. Volcano plot (C) of significantly different metabolites having a fold change threshold of ≥ 2 (x axis) and t test Q-value threshold of ≤ 0.2 (y axis). Of the 860 metabolites identified, 160 metabolites significantly increased (green) and 114 significantly decreased (purple) from pre- to post-CLDT. ¹Compound that has not been confirmed based on a standard but identified with high confidence.

Among all 1,439 metabolites analyzed, 114 metabolites (8%) decreased and 160 metabolites (11%) increased following CLDT (P < 0.05, Q < 0.2; Fig. 7C, Supplemental Table S3, https://doi.org/10.6084/m9.figshare.21702020). The largest changes PRE to POST were observed within the lipid superpathway, which accounted for 49% of the metabolites that differed PRE to POST. Metabolites indicative of fatty acid metabolism and long and medium-chain saturated fatty acids comprised a majority of those differences. In addition, all ketone bodies measured were increased following CLDT. Partially characterized molecules or unidentified metabolites accounted for 28% of all significantly altered metabolites, and xenobiotics, cofactor and vitamin metabolites accounted for 17% of the metabolites. Minimal changes were observed in metabolites of carbohydrate, amino acid, peptide, nucleotide, and energy metabolism, which collectively accounted for 12% of the metabolites that changed significantly PRE to POST (Supplemental Table S3, https://doi.org/10.6084/m9.figshare.21702020).

Associations between Changes in Metabolites, Endocrine and Metabolic Status, and Body Composition

Significant associations between circulating hormones, metabolites, and body composition are presented in Fig. 8. Increases in glycerol and FFA (P < 0.05, Q < 0.2) were moderately to strongly (Spearman’s ρ: 0.366 to 0.889) associated with increases in metabolites of fatty acid metabolism and long and medium chain saturated fatty acids. Weaker, positive associations (Spearman’s ρ: 0.235 to 0.446) were observed between changes in progesterone and total testosterone and lipid metabolites. Decreases in body mass and fat mass were inversely associated with changes in glycerol and FFA. Fat oxidation per kilogram of DLM (Δ fat mass/initial DLM) was inversely associated with changes in progesterone and glycerol concentrations and positively correlated with changes in leptin. There were no significant correlations between changes in estradiol and changes in metabolites or body composition (Q > 0.2, data not shown).

Figure 8.

Changes in circulating hormones correlate with changes in metabolites, changes in body composition (Body Comp), and energy expenditure (EE). Changes (Δ = Post – Pre) in circulating hormone concentrations were correlated with significant log₂ fold changes in metabolites of amino acid, carbohydrate, energy, lipid, and peptide pathways, as well as changes in body composition and energy expenditure during CLDT (n = 71) using Spearman’s correlation (ρ). P values were adjusted using the Benjamini–Hochberg correction (Q). Data presented are statistically significant correlations (P < 0.05, Q < 0.2). Inverse correlations indicate that a decrease in circulating hormones during CLDT was associated with a greater log₂ fold change in metabolites. No significant correlations were identified for ΔEstradiol (data not included in the figure). ¹Compound that has not been confirmed based on a standard but identified with high confidence. BF%, body fat percentage; DLM, dry lean mass; FFA, free fatty acid; FM, fat mass; FM/DLM, fat mass relative to initial dry lean mass; PRG, progesterone; T, testosterone; TBM, total body mass; TDEE, total daily energy expenditure; TDEE/TBM, total daily energy expenditure adjusted for initial total body mass.

Total body mass loss during CLDT was associated with changes in 87 metabolites, of which 83 were metabolites from the lipid superpathway (P < 0.05, Q < 0.2, Fig. 9). Inverse correlations between total body mass loss and metabolites of fatty acid metabolism demonstrated that a greater loss of total body mass was associated with a greater increase in metabolites of fatty acid metabolism, long and medium chain fatty acids, as well as ketone bodies. Decreases in metabolites of acyl choline fatty acid metabolism, lysophospholipid, PC, PE, and diacylglycerol subpathways were associated with decreases in fat mass during CLDT. Fat oxidation per kilogram of DLM and change in body fat percentage correlated with changes in six metabolites. There were no significant correlations identified between changes in metabolites and changes in DLM, TDEE, TDEE adjusted for initial DLM, or TDEE adjusted for initial body mass (Q > 0.2, data not shown).

Figure 9.

Changes in body composition correlate with changes in metabolites. Changes in body composition (Δ = Post – Pre) were correlated with significant log₂ fold changes in metabolites of amino acid, carbohydrate, energy, lipid, and peptide pathways during CLDT (n = 71) using Spearman’s correlation (ρ). P values were adjusted using the Benjamini–Hochberg correction (Q). Data presented are statistically significant correlations (P < 0.05, Q < 0.2). Inverse correlations indicate that a decrease in total body mass (TBM), fat mass (FM), fat mass relative to initial dry lean mass (FM/DLM), or body fat percentage (BF%) during CLDT was associated with a greater log₂ fold change in metabolites. No significant correlations were identified for ΔDLM, TDEE, TDEE adjusted for initial DLM, or TDEE adjusted for initial total body mass; data for these variables are not included in the figure. ¹Compound that has not been confirmed based on a standard but identified with high confidence. ²Compound that has not been confirmed based on a standard but identified with moderate confidence.

DISCUSSION

The primary findings from this longitudinal, observational study were that decreases in body mass and fat mass along with changes in circulating hormones and the metabolome were consistent with an increase in lipolysis and fatty acid oxidation to meet energy demands during CLDT, with some evidence of differential lipolysis patterns observed in men and women. Despite no sex differences in changes in circulating glycerol or FFA, women demonstrated preservation of DLM and greater fat oxidation per kilogram DLM, accompanied by increases in metabolites derived from lysophospholipid, PC, PE, and acylcarnitine polyunsaturated fatty acid metabolism. In addition, declines in DLM during CLDT, which tended toward being higher in men than women, were positively associated with declines in peak and average lower body power. Taken together, these findings suggest that women display preferential fat metabolism compared with men during CLDT as evidenced by increased lipid metabolites and enhanced fat oxidation, enabling women to spare lean mass and possibly better maintain lower body power during physically demanding military training.

The observed sex-based differences in body composition changes align with previous reports in which men exhibited greater losses of total body mass and fat-free mass during military training compared with women, with most of the weight loss in women attributable to loss of fat mass (3, 12). Similar to the present results, Hoyt et al. (12) also reported that loss of fat mass did not differ by sex, but women lost more fat mass per kilogram of fat-free mass compared with men, indicating a higher rate of fat oxidation. Maintaining lean body mass may help preserve muscle strength and power, as suggested by the positive associations between loss of DLM and decreases in lower body power following CLDT. Similar correlations between fat-free mass and lower body power have been reported in men following an intense, 8-wk military training course (2). In contrast, Vikmoen et al. (3) did not observe significant correlations between loss of muscle mass and decreased performance in men and women following a 5.5-day field training exercise, though lower body strength recovered more quickly in women. A recent meta-analysis demonstrated that changes in body mass during military training are associated with changes in lower body power and strength, with loss in body mass driven by the combination of the magnitude and duration of energy deficit (31). This suggests that preservation of lean body mass may become more important for maintaining muscle strength and power as total energy deficit increases, in which case higher rates of fat oxidation may provide women an advantage to maintain muscle strength.

Sex-based differences in energy metabolism during acute, submaximal exercise are well established (8, 32, 33), with women demonstrating greater lipid utilization that may be in part attributable to sex hormones (9, 10). Specifically, evidence suggests estrogen can stimulate AMP-activated protein kinase (AMPK) activity, a key regulator of metabolic processes, and can regulate proteins that oxidize long-chain fatty acids (32, 34). Estrogen supplementation studies in men have demonstrated increased lipid oxidation and reduced carbohydrate utilization (10, 35). In addition to estrogen, leptin has also been reported to stimulate fatty acid oxidation via AMPK in skeletal muscle (36). Though changes in estradiol did not differ between men and women in this study, and despite a larger decrease in leptin in women compared with men, women had higher estradiol and leptin concentrations throughout CLDT. Thus, these findings appear to be consistent with an endocrine-mediated sex difference in lipid metabolism in response to physically demanding military training.

Endocrine responses to CLDT were also accompanied by sex-specific changes in metabolites. Specifically, we observed increases in glycerophospholipids and lysophospholipids among women compared with men. Previous associations between the plasma lipidome and sex have shown that glycerophospholipids (i.e., PC and PE) are higher in women than men (37). Men have higher levels of some phospholipid-metabolizing enzymes, namely, lipoprotein-associated phospholipase A2 (Lp-PLA2), that break down PC and PE (37, 38). Estrogen is suspected to downregulate Lp-PLA2 expression and activity (39) and may have contributed to the increase in PC and PE metabolites observed in women following CLDT. In addition, sex-based differences in lipid metabolites following CLDT may be partly attributable to differences existing before training, provided changes in fat mass were similar between men and women. In support, Magkos et al. (40) demonstrated that FFA rate of appearance before exercise accounted for 48% of the variance in changes in FFA rate of appearance following exercise, whereas sex did not significantly contribute to the variance. Regardless, phospholipids have essential roles in cell membrane composition, cellular homeostasis, and cell signaling (41). Therefore, sex-based differences in cell signaling during exercise- and diet-induced energy deficit may warrant future research.

Although we observed sex differences in 3% of measured lipid metabolites, changes in metabolites within lipid and other superpathways were generally similar between men and women, suggesting the metabolic response in men and women during military training may be more alike than different. In the combined cohort, an increase in lipolysis was observed in association with total body mass loss. Those findings are consistent with previous studies of shorter duration military training exercises (4, 19) wherein changes in the metabolome demonstrated increases in lipolysis, fatty acid oxidation, and ketogenesis in association with a greater loss of body mass during a 7-day military training exercise (19). Similar observations have also been reported following periods of fasting (16, 42) and endurance exercise (43), suggesting metabolic shifts in the present study are likely a result of energy deficit due to a combination of inadequate energy intake and increased energy expenditure. However, in contrast to previous reports (44, 45), significant changes in protein and amino acid metabolic pathways were not observed. That may be attributable to both the duration of the training and energy balance. In support, during acute periods (days) of negative energy balance, proteolysis, and circulating amino acids are increased for energy production, but subsequently, proteolysis is downregulated with sustained negative energy balance (weeks), thereby mitigating endogenous protein loss (46, 47).

Though study findings are consistent with previous reports, several limitations require consideration. First, certain metabolites are expected to be sensitive to recent food intake (48) and exercise (14). Provided the POST blood was collected immediately following the final training event, changes in some metabolites may have reflected the acute training that preceded the blood draw more so than the cumulative effect of the 17-day training. That combined with the need to rely on nonfasted blood samples and the inability to standardize the timing of sample collection to avoid interfering with the training likely introduced variability that may have masked some sex differences in circulating metabolites. The body composition data should also be interpreted cautiously as changes in body composition were largely within the error of the measurement method, and both the lack of control over measurement timing and changes in hydration status could have introduced errors into the body composition measures. Notably, total body water measured by bioelectrical impedance consistently overestimated total body water measured using DLW (mean difference [95% CI] = 2.16 [1.57, 2.75] kg, P < 0.001), but the measures were strongly correlated (Spearman’s ρ = 0.96, P < 0.001) and no changes in total body water from pre- to post-CLDT were observed in either sex. Thus, while the bioelectrical impedance measurements at individual time points may have overestimated DLM and underestimated fat mass, the error should not differentially affect changes in body composition. Finally, the small sample of women (n = 18) relative to men (n = 54) limited statistical power for identifying sex differences. Results should be considered exploratory, but support and build upon an otherwise small evidence base, comprised of studies including only 16 (n = 6 women) (12) and 35 (n = 12 women) (3) military personnel.

In conclusion, both men and women exhibited increases in lipolysis and fatty acid oxidation in response to a 17-day military training exercise. However, women demonstrated evidence of higher fat oxidation and preservation of lean mass, along with increases in circulating levels of metabolites derived from several subpathways of lipid metabolism compared with men. Combined with the lack of sex differences in changes in glycerol or FFA following CLDT, those observations may suggest that the metabolomics approach used herein provides a more nuanced insight into sex-specific differences in lipid metabolism during sustained exercise- and diet-induced energy deficit. Though sex differences were observed in a small proportion of measured lipid metabolites, greater differences may be observed under more strenuous and prolonged exercise conditions. Future studies should investigate the potential physiological effects of these sex-based differences in lipid utilization, and determine whether differences in substrate utilization warrant sex-specific nutrition recommendations to optimize performance during military training and other endurance training events (5, 7).

DATA AVAILABILITY

Data are available upon reasonable request and pending ethical and legal approvals.

SUPPLEMENTAL DATA

Supplemental Table S1: https://doi.org/10.6084/m9.figshare.21701951.

Supplemental Table S2: https://doi.org/10.6084/m9.figshare.21702017.

Supplemental Table S3: https://doi.org/10.6084/m9.figshare.21702020.

GRANTS

This work was funded by the US Army Medical Research and Development Command, Military Operational Medicine Research Program, and appointment to the US Army Research Institute of Environmental Medicine administered by the Oak Ridge Institute for Science and Education through an interagency agreement between the US Department of Energy and the US Army Medical Research and Development Command.

DISCLAIMERS

The opinions or assertions contained herein are the private views of the authors and are not to be construed as official or reflecting the views of the Army or the Department of Defense. Any citations of commercial organizations and trade names in this report do not constitute an official Department of the Army endorsement or approval of the products or services of these organizations.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

N.B., L.M.M., and J.P.K. conceived and designed research; L.T., P.N.R., R.C., M.W., N.B., L.M.M., and J.P.K. performed experiments; M.E.B., P.N.R., M.W., and J.P.K. analyzed data; M.E.B., L.M.M., and J.P.K. interpreted results of experiments; M.E.B. prepared figures; M.E.B. drafted manuscript; M.E.B., P.N.R., M.W., L.M.M., and J.P.K. edited and revised manuscript; M.E.B., L.T., P.N.R., R.C., M.W., N.B., L.M.M., and J.P.K. approved final version of manuscript.