Long-Distance Trail Running Induces Inflammatory-Associated Protein, Lipid, and Purine Oxidation in Red Blood Cells

Abstract

Ultra-endurance running imposes extreme demands on oxygen transport, yet how red blood cells (RBCs) respond at the molecular level remains poorly defined. We integrated plasma and RBC multi-omics with hematology and hemorheology in athletes sampled before and after two trail races of distinct duration: a 40-km marathon (MCC) and a 171-km ultramarathon (UTMB). Both races elicited systemic inflammation, but UTMB was distinguished by marked IL-6 and kynurenine increases, acute-phase protein induction, and profound lipid remodeling. In RBCs, acylcarnitine accumulation, pantothenate depletion, and oxidized lipid species indicated Lands cycle activation, while purine salvage and carboxylate metabolism reflected redox-sensitive rerouting of energy pathways. Proteomics revealed non-random oxidation, particularly methionine oxidation of antioxidant enzymes, metabolic proteins, and proteasome components, correlating with impaired deformability as gleaned by testing of rheological properties. Elevated copper provided an additional correlate of reduced RBC mechanics. Despite minimal signatures of intravascular hemolysis, plasma bilirubin and hypoxanthine rose, consistent with extravascular clearance of damaged RBCs. Collectively, these results demonstrate that ultra-running accelerates RBC aging through inflammatory and oxidative pathways beyond mechanical trauma, linking systemic cytokine responses to molecular lesions, biomechanical dysfunction, and splenic sequestration. These findings not only identify actionable biomarkers of exercise-induced hemolysis but also provide translational insight into oxidative lesions that similarly limit RBC survival in transfusion and inflammatory disease settings.

Visual abstract

INTRODUCTION

Endurance exercise places significant demands on the oxygen transport system, and red blood cells (RBCs) are at the forefront of meeting these challenges. In addition to delivering oxygen and removing carbon dioxide, RBCs contribute to pH buffering and release signaling molecules - such as ATP and nitric oxide - that facilitate vasodilation and enhance muscle perfusion^1,2. While moderate exercise is associated with improved cardiovascular function and metabolic health, prolonged endurance events, like marathons and ultramarathons, impose additional mechanical, metabolic, and oxidative stresses. These stressors can result in acute cellular damage particularly in RBCs, where the capacity for repair is limited owing to the lack of nuclei and organelles that results in the incapacity to replace damaged protein and lipid components through de novo synthesis³.

Historically, investigations into exercise physiology have largely focused on skeletal muscle⁴, serum⁵, and plasma⁶ biomarkers to gauge the systemic impact of prolonged running. However, emerging evidence suggests that the RBC compartment provides critical insights into the cellular adaptations and injury responses elicited by extreme exercise^7,8. Yet, classic concepts like foot-strike hemolysis - also known as march hemoglobinuria, first described by Fleischer in 1881⁹ - have been recently challenged in comprehensive reviews¹⁰, which indicates an incomplete understanding of the role of RBC biology in exercise tolerance and performance. Recent advances in multiomics technologies have opened new avenues for comprehensively mapping the molecular responses to strenuous exercise. Simultaneous interrogation of the metabolome, lipidome, proteome, and metallome, enables holistic understanding of how tissues and circulating cells respond to exercise-induced stress. Studies have highlighted that transient cytokine release (e.g., interleukin-6) during endurance exercise triggers metabolic remodeling and tissue repair^11,12, in addition to oxidative modifications in RBC, shifts in membrane lipid composition, and, ultimately, compromised RBC deformability⁷.

Despite these advances, the specific molecular responses of RBCs to prolonged endurance exercise remain insufficiently characterized. This knowledge gap is particularly relevant in the context of long-distance trail running, where additional challenges such as variable terrain, altitude, and prolonged mechanical stress further compound the injury and repair processes in RBCs. Indeed, long distance running results in appreciable damage to RBC including decreased deformability and increased markers of senescence^13,14 in the face of increased inflammation¹⁵. Previous omics-based studies have predominantly focused on plasma responses to acute exercise or short-duration laboratory tests¹⁶, ultramarathon treadmill simulations¹⁷, or in situ^18,19, which cannot recapitulate the unique stressors encountered during ultra-endurance events in the outdoors²⁰. Moreover, while plasma-based analyses provide valuable systemic information, they do not capture the direct impact on RBCs - cells that lack nuclei and organelles and are thus uniquely dependent on metabolic pathways to cope with oxidative damage.

In light of these considerations, our study was designed to integrate multi-omics approaches to delineate the impact of long-distance trail running on both plasma and RBC molecular signatures. By comparing pre- and post-race samples from athletes competing in the World Class Martigny-Combes à Chamonix race (MCC, 40 km, elevation gain: 2,300 m) and Ultra-Trail du Mont Blanc race (UTMB^©, 171 km, elevation gain: 10,000 m), we aimed to elucidate the distinct metabolic, inflammatory, and oxidative modifications that occur with increasing exercise duration. We hypothesized that while both marathon and ultramarathon events trigger systemic inflammatory responses and metabolic remodeling, the extent of cellular damage -particularly within the RBC compartment - would be more pronounced in ultramarathon running.

MATERIALS and METHODS

Sample Collection, Hematological, and Hemorheological Measurements:

Blood samples were collected from endurance-trained participants before and immediately after either a 40 km trail race (MCC) or a 171 km ultra-trail (UTMB). The protocol was approved by the CPP Ouest VI Ethics Committee and conducted in accordance with the Declaration of Helsinki. Standardized assessments included hematological and hemorheological parameters (blood viscosity, red blood cell [RBC] deformability and aggregation, plasma free hemoglobin, microparticle release, phosphatidylserine exposure) and inflammatory markers such as IL-6, as previously described.¹⁴

Omics analyses:

Metabolomics and lipidomics,^21,22 ICP-MS²³ and proteomics analyses²⁴ were performed as previously described and extensively detailed in Supplementary Methods. Graphs, heat maps and statistical analyses (either T-Test or ANOVA), multivariate analyses including Principal Component Analysis (PCA), Partial Least Squares-Discriminant Analysis (PLS-DA), hierarchical clustering analysis (HCA), and metabolite pathway enrichment analysis are performed using MetaboAnalyst 5.0²⁵ DSPC IL-6 network filtered by degree (1.5), betweenness (1.0), and correlation (p < 0.001) were plotted using CytoScape.²⁶ Proteomics GO networks were prepared using ShinyGO 0.82.^27,28

RESULTS

Molecular responses to ultra-running in plasma and RBCs

We profiled paired plasma and RBC samples from 23 runners (MCC, 40 km: n=11; 5 female, 6 male, 35.7 ± 8.6 years; UTMB, 171 km: n=12; 4 female, 8 male, 38 ± 6.4 years) using multi-omics (metabolomics, lipidomics, proteomics, metallomics), and correlated omics results to hemorheology measurements (re-elaborated from Robert et al.¹⁴). This dataset encompassed 440 or 1105 proteins, 659 or 647 lipids, 197 or 271 metabolites, and 8 or 6 trace elements in plasma or RBCs, respectively (Figure 1A; Supplementary Figure1A–B; Supplementary Tables 1–2).

Figure 1. Multiomics of a 40 km or 171 km Run.

(A) Plasma and RBC were taken before (Pre) and after (Post) a 40-km (Martigny-Combes à Chamonix race, MCC) or 141-km (Ultra-Trail du Mont Blanc, UTMB^©) footrace. (B) Proteomics, lipidomics, metabolomics using liquid chromatography mass spectrometry (LC-MS), and metalomics using ion coupled plasmon (ICP)-MS were collected. (B) Partial Least Squares Discriminant Analysis (PLS-DA) of the plasma fraction during the MCC and (C) UTMB along with associated volcano plots highlighting all analytes with fold change (FC) > 1.5 and p-value < 0.05 from a paired T-test. (D) PLS-DA of the RBC fraction during the MCC and (E) UTMB along with associated volcano plots highlighting all analytes with fold change (FC) > 1.5 and p-value < 0.05 from a paired T-test. (F) Average speed (in kilometers per hour) for all runners in each race. (G) Linear mixed model adjusted for age, sex, BMI of runner speed (km/h) in the plasma, or (H) RBC fraction.

Pre- and post-race plasma profiles separated cleanly in both races (Figure 1B–C). The MCC was characterized by increases in acylcarnitines and fatty acids, consistent with enhanced β-oxidation and mobilization of lipid substrates. By contrast, UTMB plasma displayed decreases in lysophospholipids and higher levels of hypoxanthine, creatine kinase (KCRM), and acute-phase proteins (CRP, SAA1/2), pointing to metabolic stress and tissue damage. RBCs paralleled these differences: MCC samples accumulated acylcarnitines and fatty acids while pantothenate decreased, suggesting higher flux through CoA-dependent repair pathways; UTMB samples contained hydroxylated acylcarnitines, the oxidized linoleate derivative 13(S)-HODE, and kynurenine (Figure 1D–E). Of note, RBC samples after UTMB, but not MCC, showed increases in biliverdin and bilirubin (Figure 1E). Since RBCs are devoid of heme oxygenase, these observations are consistent with non-enzymatic heme breakdown by oxidation into both biliverdin IXα and β isoforms²⁹, which can then be converted by BLVRB – abundant in the RBC cytosol³⁰ – into bilirubin. However, it must be acknowledged that our high-throughput method does not resolve biliverdin isomers.

Performance metrics also diverged. Runners averaged 6.1 km/h in MCC vs 4.4 km/h in UTMB, mirroring the longer distance in the latter race (p=0.009; Figure 1F). A linear mixed model identified 51 plasma molecules significantly associated with speed, spanning multiple lipid classes and tryptophan catabolites, while only RBC ADP correlated with pace (Figure 1G–H). Volcano plots of correlations to speed, unadjusted or adjusted by elevation gains (vertical speed) are shown in Supplementary Figure 1. Thus, plasma captured dynamic adjustments to workload, whereas RBCs recorded the integrated oxidative and metabolic burden of the entire race

Inflammatory remodeling dominates the plasma response

Plasma lipidomics revealed decreases in diacylglycerides across both races, with additional UTMB-specific decreases in lysophosphatidylcholines (LPC), lysophosphatidylethanolamines (LPE), and cardiolipins (CL), alongside robust acylcarnitine accumulation (Figure 2A) These changes are consistent with sustained mitochondrial lipid remodeling and altered membrane turnover. Pathway modeling highlighted distinct remodeling strategies: MCC runners had increased fatty acid transporter and immunomodulator CD36³¹; UTMB favored phosphatidylserine and ether-lipid axes, whereas MCC emphasized phosphatidylcholine synthesis (Figure 2B–C; Supplementary Table 3).

Figure 2. Lipidomics, Proteomics and correlates to the inflammation marker IL-6 after the MCC and UTMB.

(A) Total lipid class quantities shown as log₁₀(Peak area sum) before and after the MCC and UTMB. (B) BioPAN models of lipid class flux during the MCC (left) and UTMB (right). Lipid classes are indicated by node shape where filled nodes indicate significantly modified class. Fluxes normalized by Z-score with positive fluxes shown in teal and negative fluxes shown in purple. (C) A volcano plot showing the intra-subject proteomic fold changes with respect to each race. (D) IL-6 measurements before and after the MCC (blue) and UTMB (red). (E) A Debiased Sparse Partial Correlation (DSPC) network of IL-6. An expanded view is available in Supplementary Figure 2. (F) BioPAN model of lipid pathway flux for lipids significantly associated with IL-6 levels. (G) Metabolite set enrichment analysis for metabolites lipids significantly associated with IL-6 levels. (H) Gene Ontology analyses for proteins positively (left) of negatively (right) significantly associated with IL-6 levels. (I) KEGG pathway enrichment based on proteomics data. (J) Kynurenine correlations between plasma (y-axis) and RBC (x-axis) compartment. (K) RBC molecular correlates with plasma IL-6 levels.

Plasma proteomics showed broader induction after UTMB than MCC. Creatine kinase (KCRM) rose in UTMB runners - reflecting ongoing muscle injury³² - who also had higher increases in acute-phase proteins amyloid proteins (SAA1/2) and Lipopolysaccharide-Binding Protein (LBP) (Figure 2C; Supplementary Figure 1). Circulating cytokines increased after both races¹⁵, but IL-6 rose more strongly after UTMB, especially in males (Figure 2D; Supplementary Figure 1C).

An IL-6-centered network integrated these signatures (Figure 2E). IL-6 correlated with acute-phase proteins (CRP, SAA1, SAA4, S100A9), white blood cell counts, tricarboxylic acid intermediates (citrate, succinate, fumarate), purine catabolites (hypoxanthine, urate), fatty acids including arachidonic acid [FA(20:4)] and heterohydroxyeicosatetraenoic acids (HETE), and remodeling of diacylglycerides and lysophosphatidylcholines into phosphatidylcholines (Figure 2F; Supplementary Figure 2). Gene-ontology analyses supported this view, with acute inflammatory response enriched among positively correlated proteins (Figure 2G–H), whereas cholesterol trafficking and complement activation were suppressed (Figure 2I).

Kynurenine – a marker of RBC hemolytic fragility observed in RBC blood bank storage³³ – emerged as a central metabolite. Its levels correlated strongly with IL-6 across compartments, but RBC kynurenine was the single strongest correlate of IL-6 in the entire dataset (Figure 2J–K; Supplementary Figures 3–4), similar to prior work in SARS-CoV-2^34–36 and Sickle Cell Disease.³⁷ This supports the idea that RBCs act as reservoirs for inflammation-linked metabolites, buffering circulating levels but also becoming more vulnerable to stress.³³

RBCs accumulate damage markers with distance and time under load

Hematological parameters revealed reductions in mean cell volume and increases in immature reticulocytes after both races. While RBC only after the MCC had enhanced phosphatidylserine (PS) exposure indicative of eryptosis³⁸, only UTMB triggered a significant increase in RBC-derived extracellular vesicles and a measurable decrease in hematocrit (Figure 3A). Of note, RBC counts and hematocrits decreased significantly after UTMB – but not MCC – both in male and female runners (Figure 3B–C). These observations suggest enhanced eryptosis^38–40 and splenic clearance of damaged cells in the longer race.

Figure 3. RBC Multiomics of Running.

(A) Hematologic parameters of each race is shown. (B) RBC counts (C) and hematocrit by race and sex. (D) PLS-DA of intra-subject normalized molecular values for each race. (E) The top 25 molecules based on Variable Importance in Projection (VIP) for the PLS-DA are shown (left) along with relative values of intra-subject fold changes for each race in the heat map (right).

RBC metabolomes clearly separated by race (Figure 3D). MCC RBCs were enriched in medium chain acylcarnitines and urate, consistent with accelerated lipid turnover and purine catabolism. UTMB RBCs contained long-chain acylcarnitines, dicarboxylic acids (fumarate, methylmalonate), carnitine, and elevated copper (Figure 3E). Together, these findings suggest that distance dictates whether RBCs rely primarily on rapid catabolism (MCC) or engage longer-term oxidative stress adaptations (UTMB).

Membrane repair via the Lands cycle is a shared RBC response

RBC lipid classes reflected widespread remodeling. Acylcarnitines increased and lysophospholipids decreased after both races (Figure 4A), consistent with activation of the Lands cycle (Figure 4B), a process by which RBCs replace oxidatively damaged membrane lipid acyl chains using a pool of acyl-CoA molecules that are interconverted with acylcarnitines based on demand for lipid repair.⁴¹ Free fatty acids and acylcarnitines increased in both cohorts (Figure 4C–D), while lysophospholipid:phospholipid ratios declined (Figure 4E). The CoA precursor pantothenate decreased in both races, whereas carnitine rose only after UTMB suggesting ongoing utilization of the acyl-CoA pool rather than conversion into acylcarnitines for storage of acyl chains (Figure 4B, F). These changes highlight the reliance of RBCs on acyl pools to replace oxidized membrane lipids.

Figure 4. RBC Lipidomics.

(A) The average peak areas of lipids identified from untargeted analysis were summed according to lipid class and plotted as a heat map. p-values from Two-tailed paired T-test for MCC pre/post comparisons are indicated as *, p<0.05; **, p<0.01, ***, p<0.001; ****, p<0.0001 and p-values for UTMB pre/post comparisons are indicated as +, p<0.05; ++, p<0.01, +++, p<0.001; ++++, p<0.0001. (B) A cartoon of the Lands cycle for red blood cell membrane repair is shown. (C) Average peak areas from semi-targeted analysis for fatty acids and (D) acylcarnitines are shown as heat maps. (E) The ratio of lysophospholipid-to-phospholipid by group from untargeted analysis (F) The levels of pantothenate and carnitine are plotted as violin plots. (G) BioPAN or RBC during the MCC and (H) UTMB are shown. (I) Violin plots of significant oxylipins before and after the MCC (blue) and UTMB (red).

BioPAN lipid networks revealed race-specific pMCC emphasized phosphatidylcholine and ether-lipid remodeling, pathways that support ferroptosis resistance, whereas UTMB showed stronger redistribution toward phosphatidylserine (Figure 4G–H; Supplementary Table 4). Oxidized lipid species such as hydroxy- and hydroperoxy-derivatives of linoleate and linolenate increased significantly, especially after MCC (Figure 4I; Supplementary Table 2). Thus, while both races engaged repair mechanisms, UTMB RBCs appeared to preserve phosphatidylserine asymmetry – likely via ATP-dependent flippases or by selective removal of PS-exposing cells – whereas MCC RBCs accumulated oxidized lipid species that may predispose to hemolysis.

Redox-sensitive rerouting toward purine salvage and alternative carboxylate metabolism

Because RBCs lack mitochondria, glycolysis is their sole source of ATP. Lactate accumulated significantly after MCC, while UTMB elicited broader changes in glycolytic intermediates with altered phosphoenolpyruvate:pyruvate ratios suggestive of redox-sensitive pyruvate kinase inhibition⁴² (Supplementary Figure 5). Such inhibition could theoretically impair ATP production but preserve 2,3-BPG, an allosteric effector that enhances oxygen unloading.

The pentose phosphate pathway, essential for NADPH generation⁴³, was more active after UTMB, supporting oxidative stress management. Purine metabolism bifurcated by distance: MCC RBCs accumulated inosine and urate, consistent with catabolism, whereas UTMB RBCs accumulated IMP and hypoxanthine, consistent with salvage via AMPD3⁴⁴ (Figure 5A). This salvage relies on aspartate, which was also consumed specifically in UTMB RBCs, and was accompanied by increased malate, consistent with NAD+ recycling via the GOT1/MDH axis⁴⁵ (Figure 5B). Pyruvate:lactate ratios were higher both at baseline and after UTMB compared with MCC, consistent with enhanced methemoglobin reductase activity under oxidant load (Figure 5B), similar to observations in people with G6PD deficiency^46,47. Collectively, these data support a shift from purine degradation during the shorter MCC to purine salvage and alternative carboxylate metabolism during the longer UTMB distance.

Figure 5. Purine Salvage and Carboxylate Metabolism in RBC.

Metabolites in the (A) purine salvage or (B) carboxylate metabolism pathways are shown. Individual samples from MCC (O) at Pre (light blue) and Post (dark blue), or UTMB (□) at Pre (light red) or Post (dark red) are shown within violin plots. p-values from Two-tailed homoscedastic T-tests are indicated as *, p<0.05; **, p<0.01, ***, p<0.001; ****, p<0.0001. The pathway map (center) was created with Biorender.com.

Proteome oxidation preferentially targets proteostasis, metabolism, and the cytoskeleton

Proteomic analysis revealed widespread oxidative post-translational modifications. PCA at the peptide/PTM level separated pre- and post-race samples (Figure 6A). Methionine oxidation increased after both races, whereas cysteine carbamidomethylation (reflecting reduced thiols) trended lower after UTMB (Figure 6B). These modifications correlated strongly with rheological parameters: methionine oxidation associated with reduced deformability and an oxidized glutathione ratio, while cysteine carbamidomethylation correlated negatively with IL-6 and adiposity (Figure 6C).

Figure 6. RBC Proteomics demonstrates increased protein oxidation.

(A) Principal Component Analysis (PCA) of RBC peptide-level data. (B) Distribution of post translational modification (PTM) across the proteome. (C) Hive plot of significant Pearson correlations (R) between physiological, hematological, and rheological parameters are shown with R > 0.7 shown in red and R < −0.7 shown in blue. On the right, a labeled plot showing Pearson correlations for values measured at each race/time point (R > 0.7, R < −0.7) with methionine oxidation and cysteine carbamidomethylation. (D) Heat map showing relative peptide abundance across MCC and UTMB. (E) Volcano plot comparing peptide levels in the post MCC and UTMB samples. (F) KEGG Pathway enrichment of proteins with statistically significant (ANOVA Fisher’s LSD FDR < 0.05) enrichment of methionine oxidation.

Oxidation was more pronounced in UTMB RBCs when normalized to total peptide signal (heat map and volcano plot in Figure 6D–E, respectively). Pathway enrichment indicated non-random targeting of antioxidant enzymes - including those involved in protein isoaspartyl-damage repair⁴⁸ (PRDX6, PCMT1, GSTO, GCLM), metabolic enzymes (BPGM, PGD, BLVRB, ACLY), proteasome and ubiquitination machinery⁴⁹ (UBA1, UBE2V2, PSMC3), and structural proteins (spectrin, ankyrin, band 3) (Figure 6F; Supplementary Figures 6–7). Notably, proteins shared with neurodegenerative pathways (Parkinson, Huntington, Alzheimer) were enriched, underscoring convergence between RBC oxidative stress and broader proteostatic failure. In an anucleate cell, the inability to replace oxidized proteins magnifies reliance on the proteasome, which itself showed oxidative impairment.

Deformability falls when methionine oxidation and copper rise

RBC deformability, measured as elongation index across shear stresses,¹⁴ declined significantly after UTMB but not after MCC (Figure 7A). Correlation analysis revealed that nitrogen-linked metabolites (lysine, proline, xanthine) tracked positively with deformability, whereas glycolytic intermediates (2/3-phosphoglycerate, 2,3-BPG) and hydroxy-acylcarnitines tracked negatively (Figure 7B). Proteins associated with deformability were enriched for proteasome assembly (Figure 7C). While total proteasome subunit levels were unchanged (Figure 7D), methionine oxidation on proteasome components increased during ultrarunning and inversely correlated with deformability (Figure 7E), consistent with oxidation-induced impairment of proteostasis.

Figure 7. Molecular Correlates with RBC Deformability.

(A) Total deformability decreases specifically after UTMB (see Sup Fig for deform on all Pressures). (B) Spearman correlations with Deformability AUC. (C) GO Enrichment terms with deformability. (D) Heatmap showing total levels of proteasome subunits. (E) Dot plot showing the Pearson correlation with RBC deformability on the x-axis. The size of each dot is proportional to the Post/Pre fold change and the dot outline indicates with the fold change increased (red) or decreased (blue) after the race. The significance of the correlation is indicated by color. (F) Pearson correlation at each shear stress pressure for metals, along with (G) abundance before and after each race. (H) Pearson correlation plots of hematocrit and plasma levels of bilirubin and hypoxanthine. Individual races and time points are indicated. (I) A model of RBC damage accumulation during long distance trail running.

Metals provided another dimension. RBC potassium, magnesium, and iron showed variable correlations, but copper levels notably rose after UTMB and consistently correlated inversely with deformability across shear stresses (Figure 7F–G). Copper accumulation is a known feature of the acute-phase response,⁵⁰ and free copper can catalyze lipid peroxidation and impair membrane fluidity in RBC.⁵¹ These findings suggest that copper may be both a biomarker and a mediator of impaired RBC mechanics during ultra-endurance stress.

Plasma readouts favor extravascular clearance over intravascular hemolysis

To explain the fall in hematocrit after UTMB, while accounting for the documented expansion of plasma fluid volumes during ultrarunning⁵², plasma proteomics were re-analyzed without cross run normalization to assess impact of potential changes in total protein abundance. Abundant proteins (albumin, ApoA-I) and electrolytes (sodium, iron, calcium) remained stable, but transferrin decreased after MCC and haptoglobin decreased after both races (Supplementary Figure 8A). A linear mixed model of plasma metabolites linked hematocrit inversely to acetoacetate, but most strongly to bilirubin and hypoxanthine (Figure 7H, Supplementary Figure 8B). Both metabolites increased after UTMB (Figure 7H), consistent with splenic clearance of damaged RBCs rather than frank intravascular hemolysis.^53,54

RBC creatinine, a proxy for circulatory age^55,56, rose after both races, more prominently after MCC (Supplementary Figure 8D). This result may reflect preferential clearance of younger, less dense cells during the longer distances of the UTMB. Together, these data suggest a model in which prolonged running accelerates oxidative and structural damage, stiffens RBCs, and favors extravascular clearance, as captured in plasma bilirubin and hypoxanthine (Figure 7I).

Several measurements discussed above showed significant correlations with both absolute speed and vertical-efficiency speed (i.e., speed normalized by elevation gain) across the two races (Supplementary Figure 1). Faster runners exhibited higher hematocrit, increased PS⁺ RBCs, and greater microparticle abundance, consistent with race-associated hemolysis and RBC membrane stress. Speed was also positively associated with plasma hemoglobin, as well as metabolites such as histidine and anthranilate, reflecting enhanced muscle protein turnover and tryptophan–kynurenine pathway activation. In contrast, negative correlations were observed for several proteins of likely muscle origin (e.g., ACTN1, CAP1, FLNA, TPM3, TPM4), suggesting that higher performance levels are accompanied by greater skeletal muscle disruption independent of RBC injury. These trends further support the possibility that circulating biliverdin/bilirubin partially derives from degradation of muscle heme-containing proteins – even though myoglobin (MB) itself was not detected in our plasma proteomics, potentially due to its rapid renal clearance given its monomeric size (16 kDa, i.e., <45 kDa kidney filtration threshold).

In summary, running distance – and thus cumulative inflammatory and oxidative stress – dictated the depth of RBC remodeling. Both races triggered activation of the Lands cycle and proteome oxidation, but only UTMB reached a threshold where methionine oxidation, copper accumulation, and proteostasis bottlenecks translated into measurable declines in deformability and extravascular hemolysis. The integrated plasma and RBC datasets thus connect inflammation to membrane damage, protein oxidation, mechanics, and clearance in vivo, providing a molecular explanation for why ultra-endurance running accelerates RBC aging.

DISCUSSION

This study demonstrates that marathon and ultramarathon trail running elicit a cascade of molecular and rheological changes in red blood cells (RBCs) that extend well beyond mechanical trauma. By integrating multi-omics with hemorheology, we show that systemic inflammation, oxidative stress, and metabolic remodeling act as central drivers of accelerated RBC aging and clearance⁵⁷. Even a 40 km race (MCC) induced inflammatory signals – leukocytosis, IL-6 elevation, acute-phase protein induction, and kynurenine accumulation – with further amplification during the 171 km ultramarathon (UTMB). These signatures were accompanied by early phosphatidylserine (PS) exposure after MCC and by impaired deformability, cell shrinkage, structural protein oxidation, microparticle generation, and reticulocytosis after UTMB. Although both races elicited inflammatory and metabolic remodeling, the depth and nature of these responses diverged markedly with distance. The shorter MCC distance primarily induced transient metabolic activation characterized by increases in lactate and the early signatures of RBC membrane remodeling revealed by increased oxidative lipid damage, elevated acylcarnitines and fatty acids, and depleted pantothenate. In contrast, the UTMB produced a far more extensive systemic and cellular perturbation, with disproportionately higher IL-6 and kynurenine levels, stronger acute-phase protein induction, and widespread plasma lipid remodeling involving depletion of lysophospholipids. RBCs from UTMB runners accumulated long-chain and hydroxylated acylcarnitines, dicarboxylates, and copper, and exhibited more pronounced activation of the pentose phosphate pathway, redirection toward purine salvage, and increased aspartate consumption; features consistent with sustained oxidative stress. Proteomics further revealed that only UTMB triggered extensive methionine oxidation across antioxidant enzymes, metabolic proteins, and cytoskeletal components, and the proteasome, correlating with impaired deformability. Correspondingly, rheological measurements showed a significant decline in elongation index after UTMB but not MCC, and plasma bilirubin and hypoxanthine rose selectively after UTMB, indicating enhanced extravascular clearance of damaged RBCs. Together, these findings demonstrate that while both events impose metabolic stress, only the ultra-endurance load of the UTMB crosses a threshold at which oxidative injury, proteostatic failure, membrane remodeling, and biomechanical dysfunction converge to accelerate RBC aging and clearance.

Beyond the RBC-intrinsic signatures, the integrated plasma multi-omics data provide essential context for interpreting the cellular phenotypes observed in both races. Plasma metabolomics and lipidomics revealed robust changes consistent with escalating inflammatory and metabolic load, the UTMB-specific depletion of lysophospholipids and cardiolipins, broad acylcarnitine accumulation, and increases in purine degradation products such as hypoxanthine. These circulating shifts paralleled, and likely contributed to, the oxidative and metabolic stress observed within RBCs. For example, the pronounced rise in IL-6 and kynurenine in plasma during UTMB mirrored the accumulation of kynurenine inside RBCs, which emerged as the strongest correlate of IL-6 across the entire dataset thus reinforcing the role of RBCs as reservoirs and buffers of inflammation-linked metabolites. Acute-phase protein induction in plasma (SAA1/2, LBP, CRP) also aligned with increased methionine oxidation of antioxidant enzymes and cytoskeletal proteins within RBCs, suggesting coordinated systemic and cellular oxidation. Likewise, elevated plasma hypoxanthine and bilirubin, consistent with increased purine turnover and extravascular clearance, matched the observed decline in deformability and accumulation of oxidation products in RBCs. These integrated plasma/RBC relationships emphasize that the cellular lesions captured in RBCs do not arise in isolation but reflect a broader systemic inflammatory and metabolic environment shaped by race duration and workload.

The most evident systemic change was a reduction in hematocrit. While some studies have suggested RBC are removed substantially due ultrarunning^58,59, use of more direct carbon-monoxide rebreathing techniques that assess total hemoglobin mass and total red blood cell volume after ultrarunning (including the UTMB) have attributed hematocrit alterations to plasma volume expansion of 18–20%^52,60. Although hemodilution accounts for part of this decrease, our findings indicate that splenic sequestration and extravascular hemolysis of oxidatively damaged and less deformable RBCs are significant contributors, consistent with increases in bilirubin and decreases in haptoglobin despite minimal intravascular hemolysis, and in keeping with similar findings in the context of transfusion of RBCs after prolonged storage⁶¹. Consistent with this hypothesis, plasma levels of bilirubin increased in accordance with decreased hematocrit, and while cell free hemoglobin remained unchanged, haptoglobin was decreased. Once RBCs accumulate sufficient oxidative or mechanical damage, they are cleared by hepatic and splenic macrophages⁶². During periods of excess RBC removal, indirect bilirubin is released back into circulation rather than removed through the biliary duct and intestine⁵³. Therefore, these results illustrate how extended endurance running exercise results in an accelerated aging process within the RBC population, thereby leading to enhanced removal of damaged cells from circulation.

Running-induced hemolysis is often attributed to repetitive foot-strike trauma.⁶³ Our results highlight additional, non-mechanical lesions, consistent with evidence for supraphysiological oxidant stress in RBCs during exercise⁶⁴. These mechanisms include alteration of the membrane lipidome^7,65, cytoskeleton^59,66, and depletion of antioxidant systems⁶⁷, which occur in conjunction with increased osmotic fragility due, in part, to prolonged exposure to lactate acidosis and subsequent dehydration at higher exercise intensities^68–72. Indeed, RBCs here decreased in volume and deformability, and while glutathione did not appear to be depleted, protein and lipid oxidation increased substantially.

We observed multiple impacts of running on the RBCs membrane lipidome. RBCs repair membrane lipid damage using the Lands Cycle, which relies on the generation of acyl-CoA substrates to replace damaged acyl chains⁷³. This process is buffered by acylcarnitines, which serve as an ATP-independent source of acyl chains^74–76. Therefore, carnitine availability can bolster this process and improve RBC function under extreme oxidative stress environments of blood storage⁷⁷. In this study, however, we observed carnitine accumulation specifically after UTMB, which might suggest that RBCs are utilizing the acyl-CoA pool at an accelerated rate and have less need for storing acyl chains in the acylcarnitine pool. Consistent depletions in both races of the CoA precursor pantothenate (Vitamin B5), which is metabolized by RBCs^78,79, suggests alternative strategies to maintain the Lands Cycle during exercise provided that intestinal absorption is optimized^80,81.

Double bonds in acyl chains are susceptible to hydroxylation via alpha-oxidation to generate hydroxyacylcarnitines. Inverse correlations between hydroxyacylcarnitines and RBC deformability seen here were also observed in cyclists⁷, who do not experience similar mechanical stress in the foot, thereby indicating alternative sources of oxidative lipid stress. Notably, these molecules increase in RBC as a function of cardiorespiratory workload in a controlled cardiopulmonary exercise test (CPET) in healthy subjects⁸². Acylcarnitine hydroxylation may occur in conjunction with increased desaturation of fatty acids, which are subsequently susceptible to lipid peroxidation⁸³. Indeed, fatty acid desaturase activity is upregulated in part to recycle NAD⁺⁸⁴. While the role of this pathway has been established in ex vivo RBCs storage⁸⁵, future work should look at the role this pathway plays in response to exercise in vivo.

The extent of oxidation on proteasome subunits inversely correlated with RBC elongation index, suggesting that the loss of ability to degrade damaged proteins may be linked to loss of deformability, or that deformability itself may influence proteasome activity. Indeed, loss of proteasome activity in conjunction with increasing oxidative stress and decreasing RBC deformability has been observed in ultra-trail running races on La Reunion Island.⁸⁶ In addition, the levels of copper were consistently inverse correlated with elongation index at all shear stresses. Copper accumulation is known to occur as part of the acute phase response⁵⁰ and can directly decrease RBC deformability⁵¹ through interaction with and subsequent oxidation of membrane phospholipids⁸⁷. Future work should focus on understanding the source of this copper (i.e. extracellular versus loss of conjugation to copper-containing proteins such as superoxide dismutase, for example), its direct role in RBC function, and how exercise or other stressors influence concentration dynamics.

These results, however, do highlight the importance of oxidative stress management in circulation prior to RBC removal by the reticuloendothelial system. One process by which RBCs are selected for removal involves their ability to traverse the interendothelial slits within the spleen. With sub-1.5 μm diameters, this transit is thus invariably dependent on the RBCs ability to deform⁸⁸. Moreover, passing through gradients of varying tonicity in both the spleen and kidney requires that the RBCs can maintain resilience to osmotic stress. Interestingly, RBCs that possess high levels of kynurenine are also increasingly fragile to changes in osmotic gradients³³. Because kynurenine is not synthesized by RBCs, its abundance is therefore a reflection of the circulating environment. Individuals who donate RBC units with higher levels of kynurenine tend to be of older age or higher BMI, and the plasma proteome within these units is characterized by a sterile inflammatory environment³³. This signature is also present in interferon-associated inflammation of Trisomy 21⁸⁹. In an acutely inflamed environment such as COVID-19, kynurenine is strongly correlated with both IL-6 levels and disease severity^34,36,90. Notably, comparable damage to RBC membrane lipids and proteins has been documented in COVID-19 patients⁹¹, in line with decreased RBC deformability^92,93. These studies, however, were performed using blood from one or two collections and thus represented “snapshots” of individual biology at the time of collection. In this study, we were able to obtain a more dynamic view of this network after paired sampling in the same subjects. Collectively, the association between cell rheology and oxidative damage to proteins and lipids reinforces previous findings, though expands these from therapeutic settings such as blood storage and pathological settings such as COVID-19 to a new setting of prolonged endurance exercise.

The present findings not only confirm that extreme endurance exercise induces a complex cascade of RBC damage and repair processes but also offer mechanistic insights that have broader implications. Repeated exposure to mechanical trauma (e.g., foot-strike hemolysis) and additional circulatory shear stress compounded by metabolic perturbations, appear to trigger oxidative injury that is evident in the accumulation of oxidized proteins (e.g., methionine oxidation) and altered lipid profiles. In addition to these mechanistic insights, our study underscores important differences between running-based endurance activities and other modalities such as cycling. Running, particularly on hard surfaces, uniquely exacerbates mechanical damage to RBCs through repetitive impact forces. In contrast, non-weight-bearing activities predominantly induce shear stress without the added burden of direct mechanical trauma. This distinction is critical when designing training and recovery strategies, including targeted nutritional support (e.g., antioxidants, pantothenate, or carnitine supplementation) or tailored recovery protocols.

Moreover, the multi-omics signatures identified here have direct implications for athletic performance and recovery. The observed decreases in RBC deformability and alterations in metabolite and protein profiles suggest that prolonged endurance exercise challenges the balance between RBC damage and repair. In the short term, this imbalance may impair microcirculatory flow and reduce oxygen delivery, contributing to performance decline during ultramarathons. Over the longer term, repeated bouts of exercise-induced RBC turnover could drive hematological adaptations that, while beneficial (e.g., a younger, more resilient RBC pool), also risk cumulative damage if recovery is insufficient. These insights pave the way for future research aimed at defining an “optimal range” of hemolysis and RBC turnover during exercise – a concept that could be instrumental in developing personalized training and nutritional interventions.

Notably, the molecular signatures identified here parallel those observed in stored RBCs used for transfusion. Storage induces progressive oxidative damage, accumulation of oxidized lipids and proteins, impairment of deformability, and eventual extravascular clearance after transfusion – features strikingly similar to those we observed in ultramarathon RBCs. This convergence underscores that endurance running provides a unique “in vivo stress test” that recapitulates lesions central to transfusion medicine.

Like for recent advances in precision transfusion medicine,⁹⁴ inter-individual variability, driven by factors such as age, sex, training status, and genetic predisposition likely influences the degree of molecular damage observed, highlighting the need for personalized approaches in exercise regimens and recovery strategies. Moreover, the parallels between the molecular signatures observed in endurance athletes and those in other models of oxidative stress – such as sepsis, aging, or even storage lesions in transfused blood – offer broader biological insights that could extend to clinical settings, including transfusion medicine. Finally, while the current study provides a comprehensive snapshot of the acute molecular responses to long-distance running, future longitudinal studies are needed to assess the persistence of these changes and to directly link molecular alterations to functional outcomes. Such work would not only deepen our understanding of exercise-induced cellular stress but also inform interventions aimed at preserving RBCs functionality and overall cardiovascular health in athletes.

Key points:

• In vivo RBC damage resulting from ultra-running mirrors molecular damage accrued by RBC during ex vivo storage under blood bank conditions

• Ultra-running accelerates RBC aging via inflammation, IL-6, kynurenine, oxidation, copper

Data and materials availability:

All data are available in the main text or the supplementary materials.