Cardiorespiratory fitness (CRF) significantly impacts metabolic, cardiovascular, pulmonary, and neurological health. However, challenges in measuring CRF have limited its clinical use. This study aimed to develop and validate a proteomic signature of CRF, addressing the limitations of traditional CRF measurement methods. The study's importance lies in potentially providing easily accessible, training-responsive biomarkers of CRF, which could enhance clinical decision-making and identify pharmacological targets mimicking exercise's effects. Existing research has shown links between molecular surrogates of CRF and clinical prognosis, but these studies often involved single populations, limited follow-up, and effect sizes not significantly additive over standard risk factors. This research addresses these gaps by using a large, diverse international population and focusing on the creation of a clinically applicable proteomic score.

The literature extensively documents the prognostic value of CRF, linking it to improved health, quality of life, and longevity. Measuring CRF is crucial in several disease conditions, yet widespread clinical assessment is hampered by test availability, cost, and limitations in performing maximal effort exercise. Research on molecular surrogates of CRF and training responses has shown promising results, but most studies lack the breadth and depth necessary for broad clinical applicability. The authors cite previous work demonstrating acute metabolic changes during exercise, impacting various pathways like tissue regeneration, fibrosis, muscle structure, mitochondrial dysfunction, insulin resistance, and inflammation. However, effect sizes in those studies were often not significantly additive to existing clinical risk factors. This study builds upon the previous literature by leveraging a large international dataset and sophisticated statistical methods to create a more robust and clinically relevant proteomic score of CRF.

This international, population-based study encompassed 14,145 individuals from four cohorts (CARDIA, Fenland, BLSA, and HERITAGE) with diverse CRF assessment methods. The CARDIA study (n=2238) served as the discovery cohort to define and validate a proteomic signature of CRF using penalized regression (LASSO). The UK Biobank (n~22,000) was used to assess the association of the proteomic signature with various clinical outcomes (death, cardiovascular, metabolic, malignancy, neurological conditions). The HERITAGE study evaluated the proteomic signature's modifiability with a 20-week exercise training program. Different methods of CRF assessment were used across cohorts (e.g., symptom-limited treadmill test, submaximal treadmill test with extrapolation, cycle ergometer with CPET, treadmill CPET), contributing to differences in CRF distributions. The SomaScan platform (aptamer-based technology) was used for proteomic quantification in CARDIA, Fenland, HERITAGE, and BLSA, while the Olink Explore 1536 panel was used in the UK Biobank. Statistical analysis involved LASSO regression for score development, recalibration for external validation across cohorts, and Cox regression for survival analysis in the UK Biobank. Interactions between the proteomic CRF score and polygenic risk scores were also examined. A simplified 21-protein score was developed for potential clinical translation.

A proteomic CRF score was successfully developed and validated across four cohorts. In the UK Biobank, a higher proteomic CRF score was associated with a nearly 50% lower hazard of all-cause mortality (HR=0.53, 95% CI 0.50–0.56, P<0.0001) after adjusting for clinical risk factors and fat mass. Similar protective associations were observed for cardiovascular, metabolic, and neurological outcomes, but not consistently for cancers. The proteomic CRF score improved risk prediction beyond standard risk factors, with improved discrimination and reclassification. The study also found substantial additive effects between the proteomic CRF score and polygenic risk scores for various diseases. Even a more parsimonious 21-protein panel showed consistent results. Finally, the proteomic CRF score increased significantly after a 20-week exercise training program in the HERITAGE cohort, and changes in the score were correlated with changes in peak VO2. A higher baseline proteomic CRF score was associated with a greater increase in peak VO2 with training.

This study provides substantial evidence for a biologically plausible and clinically relevant proteomic biomarker of CRF. The proteomic CRF score's strong association with multiple clinical outcomes, alongside its additive effect with genetic risk, highlights its potential to improve clinical risk prediction and personalize interventions. The successful validation across diverse populations and CRF assessment methods, combined with its modifiability through exercise training, strengthens the score's clinical utility. The findings advance precision medicine approaches by integrating genomic and proteomic information for more comprehensive risk assessment. The development of a simplified 21-protein panel makes clinical translation more feasible. The results highlight the value of population-based proteomics for understanding the complex biological mechanisms underlying CRF and its relationship to overall health.

This large-scale study successfully developed and validated a proteomic signature of CRF that is strongly associated with mortality and multisystem disease risks. The score's additive effect with genetic risk, its modifiability with exercise training, and the feasibility of a simplified 21-protein panel suggest substantial clinical potential for risk stratification and personalized exercise recommendations. Future research could focus on larger, more diverse populations, exploring the score's applicability across different ethnicities and ages, and investigating potential sex-specific effects. Further work is also needed to translate this finding into a clinically available test.

The CRF assessment methods varied across cohorts, potentially influencing the results. There was a time gap between proteomic and CRF assessment in CARDIA. The UK Biobank outcome data relied on administrative data, potentially introducing ascertainment biases. Although the study included a diverse population, it might not fully represent all ethnicities and age groups. The use of different proteomic platforms could have introduced some variability. Finally, while the 21-protein panel simplifies clinical translation, its effect sizes were slightly lower than those of the full panel.