Exercise, FGF21, and SIRT3: A Diabetic Heart Hypothesis

Exercise helps protect the heart in type 2 diabetes. The interesting question is whether it does so partly by changing a specific mitochondrial control system: FGF21, AMPK, FOXO3, and SIRT3. The evidence supports pieces of that story in humans. It does not yet prove that exercise repairs mitochondrial function in the diabetic human heart through this chain.

That distinction matters. A plausible molecular pathway can explain a treatment effect, but it cannot substitute for cardiac tissue data or trials that measure cardiac outcomes. For now, the FGF21-to-SIRT3 idea is a useful hypothesis, not a reason to chase an FGF21 supplement or redesign an exercise program around a blood biomarker.

The proposed switch: FGF21 to SIRT3

FGF21 is a metabolic hormone. In preclinical diabetic-heart models, exercise is proposed to activate an FGF21 → AMPK → FOXO3 → SIRT3 sequence. SIRT3 is a mitochondrial protein involved in regulating cellular energy handling and stress responses. The proposed result is healthier cardiac mitochondria and less diabetic cardiomyopathy, the structural and functional heart disease associated with diabetes.

The appeal is obvious: exercise stops sounding like generic lifestyle advice and starts looking like a molecular intervention. But the complete pathway has been characterized in diabetic mice and induced-pluripotent-stem-cell-derived cardiomyocytes, not in human diabetic cardiac tissue. Those models can identify candidates worth testing; they cannot establish what happens in a living human heart exposed to years of hyperglycemia, inflammation, medication use, and other metabolic disease.

What the human exercise trial actually found

The most direct human evidence comes from a 2024 randomized controlled trial by Duan et al. in obese men with type 2 diabetes. The investigators compared structured resistance training with structured aerobic training and measured circulating FGF21 before and after the interventions. Both training modalities produced statistically significant changes in circulating FGF21.

That is an important causal result: structured exercise can alter the FGF21 axis in people with type 2 diabetes. It does not show that either program changed FGF21 signaling inside the heart, raised SIRT3 in cardiac tissue, or improved ejection fraction, diastolic function, or mitochondrial performance. The available abstract also does not provide usable group-specific effect estimates or clarify the direction of the FGF21 change.

That missing direction is more than a technical nuisance. People with type 2 diabetes can have chronically elevated circulating FGF21, a pattern consistent with reduced responsiveness to the hormone. Exercise might raise FGF21 in a way that resembles preclinical models, or it might lower an elevated level toward a more physiological range. Both possibilities count as a significant change in a blood test, but they imply different biology. The trial establishes modulation, not a settled explanation for it.

It also does not establish that resistance and aerobic exercise work through identical mechanisms. Both changed the measured biomarker in this population, but a shared biomarker result cannot reveal every downstream pathway.

SIRT3 has stronger clinical relevance than the full pathway

The human case for SIRT3 itself is firmer. A 2025 meta-analysis by Gupta et al. pooled clinical studies across cardiovascular disease states and found lower SIRT3 expression and activity, including in metabolic syndrome. That makes reduced SIRT3 a reproducible feature of human cardiometabolic disease rather than a rodent-only observation.

Still, this is disease association, not proof that low SIRT3 causes diabetic heart disease or that restoring it reverses disease. The meta-analysis did not test exercise, FGF21 signaling, or SIRT3 in human diabetic cardiac tissue. It identifies a credible mitochondrial target; it does not validate exercise as a SIRT3 treatment in the heart.

The gap between the two findings is where the hypothesis lives. Exercise changes circulating FGF21 in a randomized trial. SIRT3 is lower across human cardiometabolic disease studies. The necessary connecting evidence—exercise-induced restoration of cardiac SIRT3 through FGF21, AMPK, and FOXO3 in people with diabetes—has not been shown.

Human heart tissue adds a warning about variability

Human cardiac biology rarely behaves like a straight line on a pathway diagram. Barcena et al. reported in 2023 that AMPK phosphorylation varied by sex and age in cardiac tissue from patients with inflammatory cardiomyopathy. The observational study also assessed mitochondrial-homeostasis and inflammatory markers.

This is the only study here that measured a component of the proposed cascade in actual human heart tissue. Yet its population had inflammatory cardiomyopathy, not type 2 diabetes, and the investigators did not assess FOXO3 or SIRT3. It cannot confirm or refute the diabetic-heart pathway. It does show why claims about a uniform exercise signaling response deserve restraint: baseline cardiac AMPK activity may differ substantially across people.

The drug-development gap is real

FGF21 is not merely a laboratory curiosity. FGF21 analogs such as efruxifermin and pegozafermin are being studied for metabolic liver disease. But a 2024 network meta-analysis by Ayesh et al. found that the included trials of FGF21 analogs, resmetirom, and GLP-1 agonists in MASLD and MASH focused on liver steatosis and fibrosis. None included ejection fraction, diastolic-function measures, or mitochondrial biomarkers as endpoints.

That absence does not mean FGF21 analogs fail to help the heart. It means clinical trials have not tested the claim. Liver outcomes are easier to justify when a drug is developed for liver disease; they do not answer whether an FGF21 analog improves diabetic cardiomyopathy. A drug can have metabolically useful effects without proving a cardiac mitochondrial mechanism.

Exercise has a well-established role in diabetes care. The FGF21–SIRT3 story may explain part of that benefit someday, but it does not create a new evidence-based protocol today.

What this means in practice

For a person with type 2 diabetes, the practical conclusion remains conventional because the human cardiac evidence remains incomplete: structured exercise is worthwhile, but no trial here supports choosing aerobic over resistance training—or the reverse—to activate a heart-specific FGF21-SIRT3 program. The randomized trial supports the narrower claim that both modalities changed circulating FGF21 in obese men with type 2 diabetes.

There is also no human cardiac evidence here for SIRT3-activating supplements, FGF21-boosting products, or off-label FGF21-directed drug strategies. Blood FGF21 is not a DIY scoreboard for mitochondrial repair. In diabetes, a higher or lower result may reflect very different states of signaling sensitivity.

The next useful studies are concrete: cardiac tissue work showing whether AMPK, FOXO3, and SIRT3 are altered in diabetic hearts; exercise trials that pair serial FGF21 measurements with echocardiography or cardiac MRI; and FGF21-analog trials that include cardiac endpoints. Until then, the molecular model earns interest, not certainty.

Duan et al.'s randomized trial shows that exercise reaches the FGF21 system in human type 2 diabetes. Gupta et al.'s meta-analysis shows that SIRT3 is reduced across cardiometabolic disease. Connecting those findings inside the diabetic human heart is the experiment still missing.