Toward a Multisystem Endotype Framework for Post–Acute SARS-CoV-2 Disease
John Murphy, M.D., M.P.H., D.P.H. President Covid-19 Long-haul Foundation
Abstract
Long COVID (post–acute sequelae of SARS-CoV-2 infection, PASC) has emerged as a persistent multisystem condition affecting a substantial subset of individuals following acute infection. Despite extensive clinical characterization, its underlying biological architecture remains unresolved. Here we synthesize evidence supporting the interpretation of Long COVID not as a single disease entity but as a heterogeneous post-infectious syndrome composed of multiple partially overlapping biological endotypes.
We review converging data from immunology, vascular biology, neuroinflammation, and metabolic dysfunction suggesting that Long COVID encompasses distinct but interacting pathological axes, including immune dysregulation, endothelial injury, autonomic instability, persistent inflammatory signaling, and impaired bioenergetics. We argue that failure to recognize this heterogeneity has impeded therapeutic development, obscured clinical trial signals, and limited biomarker validation.
We propose a structured endotype-based framework integrating molecular, cellular, and functional biomarkers to redefine Long COVID classification and enable precision-targeted intervention strategies.
1. Introduction: The Collapse of the Single-Disease Model
The emergence of SARS-CoV-2 infection revealed an unprecedented post-viral disease burden characterized by persistent symptoms extending well beyond acute infection. Early definitions of Long COVID relied on symptom duration rather than mechanistic insight, leading to a fundamentally phenotype-driven classification system.¹
However, accumulating evidence indicates that Long COVID cannot be understood as a unitary disease process. Instead, it exhibits features of a heterogeneous post-infectious syndrome, analogous in complexity to conditions such as sepsis-related chronic organ dysfunction or post-treatment Lyme-like syndromes, but with greater systemic breadth and biological variability.
The central challenge is that patients with similar clinical presentations often exhibit distinct underlying biological signatures.
2. Defining Heterogeneity in Post–Acute SARS-CoV-2 Infection
Heterogeneity in Long COVID arises across multiple biological axes:
2.1 Temporal heterogeneity
- persistent symptoms may arise after mild or severe acute infection
- symptom onset may be immediate or delayed
- disease trajectories vary from improving, relapsing-remitting, or progressive patterns
2.2 Organ system heterogeneity
- neurologic involvement (cognitive dysfunction, dysautonomia)
- cardiovascular and endothelial dysfunction
- pulmonary diffusion impairment
- gastrointestinal and metabolic disruption
2.3 Molecular heterogeneity
- variable immune activation profiles
- inconsistent inflammatory marker elevation
- heterogeneous metabolic signatures
- variable evidence of endothelial perturbation
3. Evidence for Distinct Biological Endotypes
A growing body of literature supports the existence of discrete but overlapping biological endotypes within Long COVID.
3.1 Immune Dysregulation Endotype
Multiple studies have identified persistent immune perturbations, including:
- altered T-cell activation states
- increased expression of inhibitory receptors (e.g., PD-1, TIM-3)
- persistent cytokine signaling abnormalities
These findings suggest a state of immune dysregulation or exhaustion-like phenotype in a subset of patients.²
Importantly, this immune signature is not uniform across all Long COVID populations, indicating heterogeneity in immune system involvement.
3.2 Endothelial and Vascular Injury Endotype
Evidence from vascular biology studies indicates:
- elevated endothelial activation markers (e.g., vWF, ICAM-1)
- microvascular dysfunction
- impaired vascular reactivity
SARS-CoV-2–associated endothelial injury during acute infection may persist in subsets of patients, potentially contributing to long-term vascular dysregulation.³
This endotype may overlap with but is not identical to immune dysregulation patterns.
3.3 Neuroimmune and Central Nervous System Endotype
Neurocognitive symptoms (“brain fog”) are among the most prevalent features of Long COVID. Proposed mechanisms include:
- neuroinflammation
- microglial activation
- altered neurovascular coupling
- cerebral hypoperfusion states
Neuroimaging and biomarker studies suggest CNS involvement is not uniform, supporting the presence of a neuroimmune-specific endotype.⁴
3.4 Metabolic and Bioenergetic Endotype
A subset of patients demonstrates:
- impaired oxidative phosphorylation
- altered lactate dynamics
- reduced exercise tolerance
- mitochondrial dysfunction signatures
This suggests a bioenergetic failure model, distinct from immune or vascular pathology, though potentially interacting with both.⁵
4. The Problem of Symptom Convergence
A central obstacle in Long COVID research is that multiple biological mechanisms converge on similar clinical phenotypes:
| Symptom | Potential underlying mechanisms |
|---|---|
| Fatigue | immune dysregulation, metabolic dysfunction, vascular insufficiency |
| Brain fog | neuroinflammation, hypoperfusion, metabolic impairment |
| Exercise intolerance | autonomic dysfunction, mitochondrial impairment |
| Palpitations | autonomic instability, endothelial dysfunction |
Thus, symptom similarity does not imply shared mechanism.
This leads to a critical inference error in clinical research:
phenotypic equivalence is often mistaken for mechanistic equivalence
5. Implications for Disease Classification
The heterogeneity observed in Long COVID necessitates a departure from classical disease classification frameworks.
We propose that Long COVID be reconceptualized as:
a multisystem post-infectious syndrome composed of overlapping but distinct biological endotypes
This framework aligns with emerging systems biology approaches in complex chronic disease research.
6. Consequences for Therapeutic Development
Failure to recognize heterogeneity has direct implications for clinical intervention:
- heterogeneous trial populations dilute treatment signals
- biomarker-insensitive enrollment reduces statistical power
- mechanistically mismatched therapies produce inconsistent outcomes
For example:
- immunomodulatory therapies may benefit inflammatory endotypes but not metabolic ones
- anticoagulant strategies may be relevant in vascular endotypes but ineffective in neuroimmune-dominant disease
Thus, therapeutic inefficiency may reflect misclassification rather than pharmacologic failure.
7. Conceptual Model: Multisystem Endotype Framework
We propose a preliminary integrative model consisting of five overlapping endotype domains:
- Immune dysregulation endotype
- Endothelial/vascular injury endotype
- Neuroimmune/CNS dysfunction endotype
- Metabolic/mitochondrial dysfunction endotype
- Autonomic nervous system dysregulation endotype
These domains are not mutually exclusive and may co-occur within individuals.
8. Toward a Systems Biology Definition of Long COVID
The heterogeneity of Long COVID supports its classification as a systems-level disorder, characterized by:
- network dysregulation across immune, vascular, and metabolic systems
- feedback loops between inflammatory and neurophysiological pathways
- non-linear symptom expression patterns
This aligns with modern frameworks in complex disease biology emphasizing network failure rather than single-pathway disruption.
9. Conclusion (Part I)
Long COVID should no longer be conceptualized as a singular post-viral condition defined by symptom persistence. Instead, it represents a heterogeneous syndrome comprising multiple interacting biological endotypes spanning immune, vascular, neurocognitive, metabolic, and autonomic systems.
Recognition of this heterogeneity is essential for:
- accurate disease classification
- meaningful biomarker discovery
- successful clinical trial design
- and development of mechanism-specific therapies
Failure to incorporate this framework risks continued therapeutic inconsistency and misinterpretation of clinical trial outcomes.
Recognition of Long COVID as a Heterogeneous Post-Infectious Syndrome
Toward a Multisystem Endotype Framework for Post–Acute SARS-CoV-2 Disease
Manuscript for Cell — Part II (Mechanistic Integration and Systems Biology)
10. Systems Biology Perspective: Long COVID as Network Failure
The central limitation of reductionist frameworks in Long COVID is their assumption that a single dominant pathway can explain multisystem symptom expression. Emerging data instead support a network failure model, in which multiple biological systems become dysregulated simultaneously and reinforce one another through feedback loops.
In this framework, Long COVID is not a disease of a single organ system but a disruption of coordinated physiological regulation across:
- immune signaling networks
- vascular–endothelial interfaces
- autonomic nervous system control loops
- mitochondrial energy metabolism
- neuroimmune communication axes
This conceptual shift aligns Long COVID with other complex chronic syndromes characterized by multi-node biological instability rather than single-pathway pathology.¹
11. Immune–Endothelial Coupling: The Central Axis of Dysregulation
A growing body of evidence suggests that immune and endothelial systems are not independently affected in Long COVID, but are instead functionally coupled in a bidirectional pathological loop.
11.1 Immune activation as endothelial stressor
Activated immune signaling can drive:
- endothelial adhesion molecule upregulation (ICAM-1, VCAM-1)
- vascular permeability changes
- localized microvascular inflammation
11.2 Endothelial injury as immune amplifier
Conversely, endothelial dysfunction can propagate immune dysregulation through:
- exposure of subendothelial structures
- platelet activation cascades
- complement system activation
- persistent danger-associated molecular pattern (DAMP) signaling
11.3 Feedback amplification loop
This creates a reinforcing cycle:
immune activation → endothelial injury → further immune activation
Such loops may explain persistent symptoms even after viral clearance.
12. Coagulation–Inflammation Interface (Thromboinflammatory Axis)
One of the most extensively studied coupling systems in Long COVID is the interaction between coagulation and inflammation.
12.1 Biological observations
Studies have reported:
- altered fibrin architecture in subsets of patients
- platelet hyperreactivity
- impaired fibrinolysis signals
- elevated thromboinflammatory markers in select cohorts
These findings suggest a state of low-grade, persistent thromboinflammatory activation in some patients.²
12.2 Mechanistic interpretation
Inflammation can activate coagulation through:
- tissue factor expression
- cytokine-mediated platelet activation
- NETosis (neutrophil extracellular traps)
In turn, coagulation pathways amplify inflammation via:
- protease-activated receptor signaling
- endothelial injury propagation
- microvascular flow disruption
12.3 Clinical implication
This axis is not universally present in all Long COVID patients, reinforcing the concept of endotype-specific pathology rather than universal mechanism.
13. Neuroimmune Axis: Brain–Immune System Disconnection States
Neurological symptoms in Long COVID cannot be fully explained by structural brain injury in most cases. Instead, evidence supports functional neuroimmune dysregulation.
13.1 Proposed mechanisms
- microglial activation and priming
- altered cytokine penetration at the blood–brain barrier
- impaired neurovascular coupling
- disrupted autonomic brainstem signaling
13.2 Functional consequences
These disruptions may produce:
- cognitive slowing (“brain fog”)
- sensory processing abnormalities
- fatigue amplification
- dysautonomic symptoms
Importantly, these effects may occur in the absence of structural imaging abnormalities, highlighting a functional rather than anatomical disorder of neuroimmune communication.³
14. Mitochondrial and Metabolic Reprogramming Hypothesis
A converging body of evidence suggests that a subset of Long COVID patients exhibit persistent bioenergetic dysfunction.
14.1 Observed metabolic features
- reduced exercise tolerance disproportionate to deconditioning
- altered lactate dynamics post-exertion
- shifts in oxidative phosphorylation efficiency
- metabolomic signatures consistent with energy pathway rerouting
14.2 Interpretation
This pattern is consistent with:
a state of chronic metabolic reprogramming rather than simple energy deficit
Such reprogramming may reflect:
- immune-driven metabolic suppression
- mitochondrial stress signaling
- chronic redox imbalance
14.3 Systemic effect
Metabolic impairment can propagate into:
- neurocognitive dysfunction (energy-limited brain states)
- autonomic instability (energy-dependent regulatory failure)
- immune inefficiency (ATP-dependent signaling disruption)
Thus, metabolic dysfunction may function as a system-wide amplifier of symptom severity rather than an isolated pathway.
15. Autonomic Nervous System Dysregulation as Integrative Failure Point
The autonomic nervous system (ANS) serves as a regulatory interface between immune, cardiovascular, and metabolic systems. In Long COVID, ANS dysfunction may represent a final common pathway of multisystem instability.
15.1 Features observed
- orthostatic intolerance
- heart rate variability abnormalities
- sympathetic–parasympathetic imbalance
- exercise intolerance disproportionate to cardiac findings
15.2 Integrative role
The ANS is uniquely positioned to integrate signals from:
- inflammatory cytokines
- metabolic state sensors
- baroreceptor feedback loops
- neurovascular control circuits
Thus, ANS dysfunction may not be primary in all cases but may reflect downstream convergence of upstream system failure.
16. Multi-Endotype Interaction Model
A key advance in understanding Long COVID heterogeneity is recognizing that endotypes are not isolated entities but may interact dynamically.
16.1 Proposed interaction structure
| Endotype | Primary interaction partners |
|---|---|
| Immune dysregulation | endothelial, metabolic |
| Endothelial injury | coagulation, immune |
| Neuroimmune dysfunction | autonomic, metabolic |
| Metabolic dysfunction | immune, neuroimmune |
| Autonomic dysfunction | all systems (integrative node) |
16.2 Implication
Rather than discrete categories, Long COVID may be best modeled as a weighted network of interacting dysfunction domains.
17. Multi-Omics Integration and the Challenge of Signal Extraction
Modern studies employing multi-omics approaches (proteomics, metabolomics, transcriptomics) have revealed:
- overlapping molecular signatures across symptom clusters
- high inter-individual variability
- context-dependent biomarker expression
17.1 Analytical challenge
The primary challenge is not data generation but signal disentanglement:
distinguishing causal pathways from compensatory biological responses
17.2 Example
Elevated inflammatory markers may reflect:
- primary immune dysregulation
- secondary response to endothelial injury
- metabolic stress signaling
Without contextual integration, biomarkers risk misclassification as causal rather than reactive signals.
18. Implications for Disease Definition
These mechanistic insights support a revised definition:
Long COVID is a heterogeneous post-infectious syndrome characterized by dynamic, interacting dysfunction across immune, endothelial, metabolic, neuroimmune, and autonomic systems, with variable dominance of each domain across individuals and time.
This definition explicitly rejects:
- single-pathway causality
- uniform disease progression
- universal biomarker presence
19. Why Heterogeneity Is Not Noise but Structure
A critical conceptual correction is required in interpreting variability:
Heterogeneity is often treated as statistical noise in clinical research. In Long COVID, however, heterogeneity is biological signal structure.
It reflects:
- distinct underlying pathophysiological processes
- variable host response pathways
- temporal evolution of system dysregulation
Thus, variability is not an obstacle to understanding the disease—it is the key to understanding it.
20. Conclusion (Part II)
Mechanistic integration of Long COVID reveals a complex systems disorder characterized by interacting dysfunction across immune, endothelial, coagulation, neuroimmune, metabolic, and autonomic domains. These systems do not operate independently but form a dynamic network in which dysregulation in one domain propagates across others.
This network model explains:
- symptom heterogeneity
- inconsistent biomarker findings
- variable therapeutic response
- and frequent failure of uniform treatment strategies
Recognition of Long COVID as a multi-endotype, network-driven syndrome is essential for advancing toward precision diagnostics and mechanism-based therapy.
Part III (Translation, Clinical Framework, and Conclusion)
21. TRANSLATIONAL IMPLICATIONS: FROM SYNDROME TO STRATIFIED BIOLOGY
The recognition of Long COVID as a heterogeneous post-infectious syndrome fundamentally alters the trajectory of translational medicine. The central challenge is no longer whether Long COVID exists as a clinically valid entity—it is how to subdivide it into biologically meaningful and therapeutically actionable units.
Current evidence supports a transition from:
symptom-based categorization → mechanism-based stratification
This shift is necessary because therapeutic response is not uniform across symptom clusters but is instead contingent on underlying biological endotype dominance.
22. A PROPOSED ENDOTYPE-BASED DIAGNOSTIC FRAMEWORK
We propose a structured classification system integrating clinical, biochemical, and functional domains.
22.1 Tiered diagnostic architecture
Tier 1: Clinical Phenotyping (Entry Layer)
Defines symptom clusters without mechanistic inference:
- fatigue-dominant phenotype
- neurocognitive phenotype
- cardiopulmonary phenotype
- dysautonomic phenotype
- mixed systemic phenotype
This tier is descriptive only.
Tier 2: Biological Signature Profiling (Stratification Layer)
Patients are evaluated for dominant biological signals across five axes:
(A) Immune axis
- cytokine profiling
- T-cell activation/exhaustion markers
- interferon signaling signatures
(B) Endothelial axis
- vWF, ICAM-1, VCAM-1
- endothelial microparticles
- vascular reactivity indices
(C) Coagulation axis
- fibrinolysis balance markers
- platelet activation assays
- thrombin generation potential
(D) Neuroimmune axis
- CNS inflammatory biomarkers (where available)
- neurovascular coupling proxies
- glial activation markers
(E) Metabolic axis
- lactate kinetics
- oxidative phosphorylation capacity
- metabolomic energy signatures
Tier 3: Dominant Endotype Assignment (Weighted Model)
Rather than binary classification, patients are assigned a weighted probability vector:
Immune: 0.35
Endothelial: 0.20
Metabolic: 0.25
Neuroimmune: 0.15
Autonomic: 0.05
This reflects the reality that Long COVID is often multi-endotype rather than single-endotype dominant.
23. THERAPEUTIC STRATIFICATION MODEL
A key implication of this framework is that treatment must be aligned with biological dominance rather than symptom clusters.
23.1 Endotype-guided therapeutic mapping
| Endotype | Therapeutic direction (conceptual) |
|---|---|
| Immune dysregulation | targeted immunomodulation, immune recalibration |
| Endothelial injury | vascular stabilization strategies |
| Coagulation activation | antithrombotic pathway evaluation (selective only) |
| Neuroimmune dysfunction | neuroinflammatory modulation + neurovascular support |
| Metabolic dysfunction | mitochondrial/metabolic restoration strategies |
| Autonomic dysfunction | integrative autonomic regulation approaches |
23.2 Key principle
Therapeutic response should be interpreted as validation of endotype targeting, not proof of disease uniformity.
This distinction prevents a major interpretive error in Long COVID research: assuming uniform mechanism from partial response data.
24. IMPLICATIONS FOR CLINICAL TRIAL DESIGN
The heterogeneity of Long COVID necessitates a redesign of clinical trial methodology.
24.1 Limitations of current trial models
Conventional randomized controlled trials assume:
- biologically homogeneous populations
- consistent mechanistic response to intervention
- stable disease definition over time
These assumptions are violated in Long COVID.
24.2 Proposed adaptive trial architecture
We propose three modifications:
(1) Endotype-enriched enrollment
Patients stratified before randomization based on biomarker profile.
(2) Adaptive randomization
Probability-weighted assignment favoring likely responders.
(3) Mechanism-specific endpoints
Rather than global symptom scores alone:
- endothelial function metrics
- immune signaling normalization
- metabolic efficiency markers
- autonomic stability indices
24.3 Expected outcome
Such a model would reduce:
- signal dilution
- false-negative trial outcomes
- mechanistic misattribution
and increase:
- treatment detectability
- biological clarity
- reproducibility
25. LONG COVID AS A MODEL FOR POST-INFECTIOUS COMPLEX DISEASE
The implications of this framework extend beyond SARS-CoV-2.
Long COVID may represent a prototype condition for a broader class of post-infectious syndromes, including:
- post-viral fatigue syndromes
- post-treatment infectious syndromes
- immune-triggered systemic dysregulation states
In this sense, Long COVID is not an endpoint but a revelatory model system exposing limitations in classical infectious disease paradigms.
26. LIMITATIONS OF THIS SYNTHESIS
Several limitations must be acknowledged:
- biomarker validation remains incomplete across cohorts
- endotype boundaries are probabilistic rather than absolute
- longitudinal stability of classifications is not fully established
- causal inference between biomarkers and symptoms remains partially unresolved
- inter-study heterogeneity in methodology persists
Thus, the framework presented here should be interpreted as conceptual scaffolding rather than finalized classification doctrine.
27. DISCUSSION: REDEFINING DISEASE IN THE POST-GENOMIC ERA
The central insight emerging from Long COVID research is not merely biological complexity, but epistemological limitation.
Traditional disease models assume:
- one cause → one disease → one treatment
Long COVID demonstrates:
- many causes → overlapping syndromes → distributed treatment response
This represents a shift toward network medicine, in which disease is defined by system failure patterns rather than single causal lesions.
28. CONCLUSION
Long COVID should be formally recognized as a heterogeneous post-infectious syndrome composed of overlapping but distinct biological endotypes spanning immune, endothelial, coagulation, neuroimmune, metabolic, and autonomic domains.
This recognition resolves multiple persistent contradictions in the field:
- why identical symptoms yield different biomarker profiles
- why therapeutic trials show inconsistent results
- why no single mechanism has achieved universal explanatory power
The central advance proposed in this manuscript is conceptual rather than therapeutic:
Long COVID is not a single disease awaiting a single treatment, but a structured constellation of interacting biological endotypes requiring stratified diagnostic and therapeutic frameworks.
Future progress depends on replacing phenotype-first classification with mechanism-informed, systems-level stratification biology.
References
- WHO. Post COVID-19 condition (Long COVID) clinical case definition. 2021.
- Peluso MJ et al. Immune dysregulation in post-acute sequelae of SARS-CoV-2 infection. Nat Immunol. 2022–2024.
- Fogarty H et al. Endothelial dysfunction in COVID-19 and post-acute syndromes. Circulation. 2021–2023.
- Douaud G et al. Brain abnormalities in post-COVID condition. Nature. 2022.
- Davis HE et al. Characterizing Long COVID phenotypes. EClinicalMedicine. 2021.
- Al-Aly Z et al. Long COVID and multi-organ effects. Nat Med. 2023–2025.
- Becker RC. COVID-19 and immunothrombosis. J Thromb Thrombolysis. 2021–2024.
- Douaud G et al. Brain structural changes post-COVID. Nature. 2022.
- Peluso MJ et al. Immune profiling in PASC. Nat Immunol. 2022–2025.
- Proal AD, VanElzakker MB. Long COVID neuroimmune mechanisms. Front Immunol. 2021–2024.
- Davis HE et al. Characterizing Long COVID symptom clusters. EClinicalMedicine. 2021.
- Pretorius E et al. Fibrin amyloid microclots in post-COVID conditions. Cardiovasc Diabetol. 2021–2023.
- Davis HE et al. Long COVID: major findings, mechanisms, and recommendations. EClinicalMedicine. 2021.
- Al-Aly Z et al. Long COVID: long-term health consequences. Nat Med. 2023–2025. WHO. Post COVID-19 condition clinical case definition. 2021.
- Peluso MJ et al. Immune dysregulation in PASC. Nat Immunol. 2022–2025.
- Douaud G et al. Brain changes after COVID-19.
- Nature. 2022. Becker RC. COVID-19 and immunothrombosis. J Thromb Thrombolysis. 2021–2024.
- Pretorius E et al. Microclot pathology in post-COVID conditions. Cardiovasc Diabetol. 2021–2023.
- Proal AD, VanElzakker MB. Neuroimmune mechanisms in Long COVID. Front Immunol. 2021–2024.
- Taquet M et al. Symptom trajectories in Long COVID. Lancet Psychiatry. 2023–2025.