Serum Metabolomic Signatures Differentiate Fracture-Related Infections and Infected Nonunion in a Pathogen-Specific Manner

AIM Accurate diagnosis of fracture-related infection (FRI), particularly differentiating low-grade infection from aseptic non-union, remains a major clinical challenge. Current biomarkers lack specificity and fail to reliably distinguish infection from impaired healing. This study aimed to determine whether serum metabolomics can identify infection-specific host responses and distinguish infection from aseptic and infected non-union. METHODS Serum samples were analysed from infected and non-infected controls from three distinct sources. Firstly, in a sheep FRI model (n=21), a tibial defect was created and inoculated with S. aureus (106 CFU/mL), followed by DAIR revision surgery and antibiotic treatment. Serum samples were collected longitudinally from preoperative stages to euthanasia. Secondly, human serum samples from patients with orthopaedic device-related infections including S. aureus (n=36) and S. epidermidis (n=16), alongside non-infected controls undergoing implant removal (n=11). Finally, a prospective multicenter study (n=137) of femur and tibia non-union patients was conducted across eight trauma centers. Patients were classified as healed (HEAL, n=33), aseptic non-union (AS, n=62), or infected non-union (INF, n=43) based on microbiology and histopathology. Metabolites were extracted using methanol-based protocols and analysed by high-resolution liquid chromatography–mass spectrometry, followed by statistical and pathway analyses. RESULTS In the sheep FRI model, longitudinal metabolomic profiling revealed infection-associated increases in collagen degradation and oxidative stress markers, including hydroxyproline (p<0.0001) and cystine (p=0.0094), indicating active tissue damage and redox imbalance during S. aureus infection. In the human cohort with S. aureus or S. epidermidis FRI, infected patients showed a distinct metabolic profile compared with controls, characterized by altered glycolytic intermediates, increased TCA cycle metabolites (succinate, fumarate, malate), and changes in amino acid/dipeptide metabolism (glutamate, tyrosine, carnosine), consistent with a hypermetabolic inflammatory state. S. aureus and S. epidermidis infections largely overlapped, however, a limited number of metabolites, including methylhistidine (p=0.0027), phosphoenolpyruvate (p=0.0084), and α-ketoglutarate (p=0.0147), were significantly higher in the S. aureus group. In the non-union cohort, L-glutamic acid was significantly elevated in infected non-union compared to healed and aseptic cases (p=0.0017 and 0.0116, respectively), while L-lysine differed between aseptic and healed groups (p=0.0098), reflecting differences in inflammatory activity and extracellular matrix remodeling. CONCLUSIONS Serum metabolomics identifies a characteristic host metabolic response that robustly distinguishes infection from non-infected states. Pathogen-specific differences are limited, indicating a predominantly host-driven response. Metabolic signatures differ in the non-union setting, likely reflecting chronicity and low-grade infection. Integration of infection- and healing-associated metabolites enables discrimination between infected and aseptic non-union, supporting metabolomics for diagnosis and patient stratification.