Product-specific mechanisms are necessary because 'MSC therapy' and 'exosome therapy' are not single, uniform interventions. Tissue source, donor, cell population, passage, culture conditions, inflammatory priming, oxygen tension, harvest method, purification, formulation, storage, viability, extracellular-vesicle composition, dose, and route can alter biological activity. A mechanism demonstrated for one product cannot automatically be assigned to another product carrying the same broad label. Mechanistic interpretation should therefore begin with a defined product identity and validated characterization. Without that foundation, apparent disagreements between studies may reflect non-equivalent products rather than contradictory biology.
Figure 1 placeholder — Product-specific mechanisms
Disease-specific mechanisms
Disease-specific mechanisms reflect differences in pathophysiology, tissue architecture, inflammatory state, vascular supply, extracellular matrix, disease duration, and standard treatment. A mechanism relevant to acute ischemic injury may be unimportant in chronic fibrosis; immunomodulation potentially useful in an inflammatory disorder may be undesirable during uncontrolled infection. Even within one organ, disease stage and molecular phenotype can change the relevant target. Mechanistic evidence should therefore be matched to the exact indication and should not be generalized from a biologically convenient model to a clinically different disease.
Figure 2 placeholder — Disease-specific mechanisms
Dose dependence
Dose dependence means that biological effects can change with the amount of cells, extracellular vesicles, protein, particles, or other active material administered. Dose-response relationships may be nonlinear, plateau, reverse, or vary by route and recipient. Higher dose is not necessarily better: increasing cell numbers can alter vascular trapping, complement or coagulation interactions, while increasing EV exposure can change biodistribution and off-target effects. A mechanistic claim is stronger when the proposed pathway changes coherently across an experimentally justified dose range and when dose units permit meaningful product comparison.
Figure 3 placeholder — Dose dependence
Timing
Timing is part of mechanism, not merely logistics. The biological environment after injury or during disease evolves over hours, days, months, or years. Inflammatory mediators, immune-cell populations, vascular permeability, cell death, fibrosis, and repair programs change with time. An intervention protective during an early inflammatory phase may be ineffective or harmful later, while a remodeling mechanism may require a later window. Preclinical treatment given before or immediately after induced injury may therefore have limited relevance to patients with established disease. Mechanistic claims should specify the treatment window and disease stage.
Figure 4 placeholder — Timing
Patient heterogeneity
Patient heterogeneity includes differences in age, genetics, sex, comorbidities, medications, immune status, prior treatment, disease severity, organ reserve, microbiome, and other biological variables. These factors can alter recipient response, biodistribution, clearance, inflammatory signaling, and tissue repair. A mechanism detected as an average study effect may operate strongly in one subgroup and weakly or oppositely in another. Mechanistic biomarkers and stratified analyses can help identify responders, but subgroup findings require adequate design and validation to avoid false discovery.
Figure 5 placeholder — Patient heterogeneity
Why biological plausibility is not clinical proof
Why biological plausibility is not clinical proof is the central evidentiary boundary. A proposed mechanism may be internally coherent, supported by cell culture, reproduced in animal models, and associated with favorable biomarkers while still failing to improve meaningful human outcomes. Conversely, a clinical effect can sometimes precede complete mechanistic understanding. Clinical efficacy requires indication-specific human evidence with appropriate comparators, endpoints, follow-up, and bias control. Mechanism strengthens interpretation and can guide product development, but it cannot substitute for evidence that patients actually benefit.
Figure 6 placeholder — Why biological plausibility is not clinical proof
Clinical takeaway
Mechanism of action is conditional rather than universal. Product identity, disease biology, dose, timing, and patient characteristics determine whether a proposed pathway is relevant.
Mechanistic hierarchy
Evidence can progress from molecular interaction to cellular response, tissue effect, organism-level physiology, biomarker change, and clinical outcome. Evidence at an earlier level should not be presented as proof at a later level.
Model validity
Cell culture and animal models are valuable for isolating mechanisms, but their relevance depends on how closely they reproduce the human product, disease, dose, route, timing, and outcome.
Mechanistic biomarkers
A biomarker can support target engagement or pathway activity. It is most informative when analytically valid, biologically linked to the proposed mechanism, and prospectively connected to clinically meaningful outcomes.
Causal evidence
Causal inference is strengthened when manipulating the proposed pathway changes the biological effect, when alternative mechanisms are excluded, and when independent methods converge on the same conclusion.
Negative and null findings
Failure to confirm a proposed mechanism is informative and should not be hidden. Null mechanistic results can reveal incorrect assumptions, inadequate exposure, wrong timing, nonresponsive subgroups, or an irrelevant target.
Regulatory and evidentiary separation
A jurisdiction's regulatory classification or authorization status is not itself proof for or against a biological mechanism or clinical effect. Mechanistic evidence, clinical evidence, and regulatory status should remain separate.
Clinical interpretation
A mechanistically elegant product can fail clinically, and a statistically positive trial can still leave mechanism uncertain. Both questions should be evaluated independently.
Knowledge gaps
For many MSC and EV interventions, dominant human mechanisms, dose-response relationships, treatment windows, responder phenotypes, and validated mechanistic biomarkers remain uncertain.
Bottom-line evidence statement
The most defensible mechanism-of-action claim is product-specific, disease-specific, dose-aware, time-aware, patient-aware, causally tested, and explicitly separated from proof of clinical efficacy.
Additional controlled terminology
A mechanism of action should begin with product identity and demonstrate target engagement. A dose-response relationship may include a nonlinear dose response, and the relevant treatment window depends on disease stage. Recipient response can be altered by biodistribution and clearance. A mechanistic biomarker may help identify a responder subgroup, but stratified analysis requires validation. A preclinical model has translational value only when model validity is adequate. A surrogate endpoint should not be confused with a clinical endpoint. Causal inference requires testing an alternative mechanism as well as the favored pathway. A null finding can refine the hypothesis. Clinical efficacy and regulatory status remain distinct from mechanistic plausibility, and translational relevance must be demonstrated rather than presumed.