Inflammation Source Interaction Network: Extended Causal DAG
Certainty: 0.25. The existing causal DAG for ME/CFS inflammation sources should be extended with the following nodes and edges based on emerging evidence:
Extended node set:
SPMS (SPM deficiency node): Reduced resolvin/protectin/maresin biosynthesis impairs active resolution. Edges: SPM deficiency β failed resolution β prolonged TLR4/RAGE signaling (positive feedback). SPM deficiency amplifies every other inflammation source by preventing normal termination. (Serhan, Libreros, and Nshimiyimana 2022) (Engert et al. 2026)
HMGB1 (DAMP amplification node): Nuclear HMGB1 translocates to extracellular space under stress, signaling via TLR4/TLR9/RAGE. Redox-dependent activity: disulfide-HMGB1 is pro-inflammatory, fully reduced is chemotactic. Exercise-induced oxidative stress shifts toward pro-inflammatory form. Edges: exertion β disulfide-HMGB1 β TLR4/RAGE β microglial activation β neuroinflammation. BBB breach allows peripheral HMGB1 to enter CNS. (Ibrahim, Wasim, and Rahman 2026) (Chen et al. 2026)
UPR (ER stress node): Unfolded protein response activation via PERK/IRE1alpha/ATF6 sensors. Chronic UPR drives IL-6/TNF-\(\alpha\) production (IRE1alpha-XBP1 axis) and NF-\(\kappa\)B activation (PERK-eIF2alpha). Edges: oxidative stress/mitochondrial dysfunction β UPR β inflammatory cytokine production β neuroinflammation. Bistable switch: once activated, persists after trigger removal. (Kawano et al. 2023)
S100A8/A9 (calprotectin node): Ca\([2+]\)-binding DAMPs released by activated neutrophils/monocytes/macrophages, signaling through TLR4 and RAGE. Directly documented in ME/CFS plasma proteomics (S100-A9 upregulation). Edges: S100A8/A9 β TLR4 on microglia β neuroinflammatory priming. Calprotectin correlates with NET burden (\(r >= 0.745\)). (Nunes et al. 2024)
Eicosanoid (COX-2/PGE2/TRPV1 amplification node): COX-2 upregulation produces PGE2 that sensitizes TRPV1, creating a feed-forward amplification loop. TRPV1 activation induces further COX-2 upregulation (\(\sim\) 30 min), perpetuating sensitization. Edges: PGE2 β EP receptors β TRPV1 sensitization β pain/neurogenic inflammation. TRPV1 activation β COX-2 upregulation β more PGE2 β sustained amplification. (Moriyama et al. 2005)
Iron (ferroptosis susceptibility node): Iron dysregulation catalyzes Fenton chemistry β hydroxyl radical β lipid peroxidation β ferroptosis. GPX4 normally reduces lipid peroxides; glutathione depletion compromises this defense. Edges: iron accumulation β lipid peroxidation β cell damage/DAMP release β inflammation. Inflammation β hepcidin β iron redistribution β tissue iron loading. Sublethal ferroptosis produces chronic oxidative stress without cell death. (Mantle et al. 2025)
Cross-edges (bidirectional amplifiers):
- HMGB1 βοΈ S100A8/A9: both signal through TLR4 and RAGE, creating convergent amplification
- UPR ββ HMGB1: ER stress promotes HMGB1 release; extracellular HMGB1 amplifies UPR via ROS
- SPM deficiency ββ all nodes: resolution failure prolongs every inflammatory signal
- Iron ββ eicosanoids: iron-dependent ROS activate COX-2, increasing PGE2; PGE2 modulates hepcidin
The extended DAG predicts that: (a) any node perturbation propagates to all others through the interaction structure; (b) SPM deficiency is the most destabilizing single node (it disables the systemβs only active resolution mechanism); (c) interventions targeting multiple nodes simultaneously (e.g., glycyrrhizin for HMGB1 + NAC for iron/redox + TUDCA for UPR) have synergistic potential; and (d) the DAG structure explains why single-node interventions (e.g., COX-2 inhibitors alone) produce only partial benefit β the network reroutes through alternative amplification paths.
Testable: Parameter estimation via multi-analyte profiling (HMGB1, S100A8/A9, GRP78, PGE2, 4-HNE, SPMs) in ME/CFS plasma at rest and post-CPET would allow construction of the first empirical interaction network. No such integrative profiling study exists.
Cross-reference: Chapter Formal Causal Hierarchy Analysis for the formal ODE models of selected sub-systems. Individual node mechanisms at Family 20: Inflammation Resolution and Lipid Mediators (HMGB1, S100A8/A9, SPM), Family 13: Protein Homeostasis and Degradation (UPR), Ferroptosis Susceptibility (iron).
Certainty: 0.35. The four trigger-capable root causes identified in Chapter Causal Hierarchy: Root Causes, Amplifiers, and Consequences (CNS energy crisis, metabolic safe mode, GPCR autoantibody cascade, TRPM3 channelopathy) each explain disease initiation but none alone fully explains the transition to chronicity. Resolution failure β impaired active termination of inflammation through deficient SPM biosynthesis, disrupted vagal-SPM coupling, and compromised DAMP clearance β may serve as a candidate chronicity mechanism contributing across multiple root causes. Caution: SPMs have never been measured in ME/CFS; the resolution failure model is a mechanistic hypothesis, not an established mechanism.
Rationale. Each root cause produces inflammatory intermediates (cytokines, DAMPs, ROS) that normally trigger an SPM-mediated resolution response. If SPM biosynthesis is impaired (whether from omega-3 precursor deficiency, altered fatty acid metabolism, reduced vagal tone, or enzymatic dysfunction in the SPM synthesis pathway), every acute inflammatory trigger β whether from CNS energy crisis, safe mode activation, autoantibody binding, or TRPM3 dysfunction β converts from a self-limited episode to a sustained inflammatory event. Resolution failure thus acts as a permissive condition for chronicity rather than a root cause: it is necessary for the disease to become chronic but not sufficient to initiate it.
Mechanistic integration:
- CNS energy crisis β neuroinflammation β vagal tone reduction β impaired SPM synthesis β failed resolution of neuroinflammation β sustained CNS energy crisis (closed loop)
- Metabolic safe mode β IL-6/IFN-gamma elevation β altered lipid metabolism β reduced SPM precursor availability β failed resolution β prolonged safe mode signaling
- GPCR autoantibodies β beta2-AR activation on immune cells β altered eicosanoid balance β reduced pro-resolving lipid mediators β sustained autoantibody-driven inflammation
- TRPM3 channelopathy β impaired NK cell function β failed clearance of infected/stressed cells β sustained DAMP release β exhausted resolution capacity
Therapeutic implication: This integration predicts that resolution-enhancing interventions (SPM supplementation, tVNS for vagal-SPM coupling, aspirin for AT-SPM generation) should benefit all four root-cause subgroups, making them the first βpan-hierarchyβ treatment class. Conversely, single-root-cause interventions without resolution support may fail because resolution failure prevents durable recovery even when the root cause is addressed.
Testable predictions: 1. SPM deficiency severity correlates with illness duration more strongly than with initial trigger type 2. SPM deficiency is present across all four root-cause subgroups (i.e., is not subgroup-specific) 3. Resolution-enhancing interventions improve outcomes additively when combined with root-cause-targeted treatments
Limitations: SPMs have not been measured in ME/CFS in any cohort. The vagal-SPM coupling hypothesis in ME/CFS is inferred from the established cholinergic anti-inflammatory pathway. The causal direction (resolution failure as permissive vs. consequence of chronic inflammation) cannot be resolved without longitudinal SPM measurement from disease onset.
Cross-reference: SPM deficiency speculation at Family 20: Inflammation Resolution and Lipid Mediators. Causal hierarchy (Chapter Causal Hierarchy: Root Causes, Amplifiers, and Consequences) for the four root causes. tVNS protocol at Chapter Emerging and Investigational Therapies.
{{/* Structural vagal damage as distinct DAG node (V_str); brainstorm 7.1/7.2/7.3/7.5; vagal-gastric-denervation-longcovid stream */}}
Certainty: 0.20. (Origin: brainstorm; no ME/CFS data β a pre-data structural hypothesis.) The extended causal DAG (Inflammation Source Interaction Network: Extended Causal DAG) and the paperβs formal models currently represent vagal function as a rate (efferent tone modulating inflammation), implicitly assuming the substrate is intact and the deficit is functional. The structural denervation evidence (Structural Vagal Cholinergic Denervation as an Anatomical Substrate for Post-Viral Dysautonomia and Sustained Inflammation) motivates considering a separate node, \(V_{\text{str}}\) β a vagal structural capacity variable that would set a ceiling on achievable vagal signaling regardless of stimulation. In this formulation, functional vagal tone is bounded above by \(V_{\text{str}}\): if efferent cholinergic fibers are physically depleted, no amount of afferent activation (endogenous or via taVNS) could restore CAP output beyond the surviving-fiber ceiling. \(V_{\text{str}}\) is proposed as one contributor β among many β to three outcomes (CAP disengagement/inflammation, gastric motility, HF-HRV), each of which is multiply determined by numerous independent systems; it is emphatically not their single cause. The \(R=0.50\)β$ 0.61$ correlations Acanfora observed (\(r^2 \approx 0.25\)β$ 0.37$, most variance unexplained) are consistent with, but far from uniquely predicted by, such a shared-node structure β many alternative causal structures fit equally well. A distinct, slowly-varying structural variable is at least mechanistically natural because fiber loss and functional signaling have very different time constants (a bistable-damage variant is conceivable but not required).
Testable prediction: If \(V_{\text{str}}\) is a genuine separate node rather than a redundant relabeling of functional tone, then taVNS-induced TNF-\(\alpha\) suppression (a functional CAP assay, idea 9.3) should show a response ceiling that correlates with structural fiber density (gastric VIP+ IHC or a validated proxy) and is independent of baseline functional HRV. Concretely: in an adequately powered ME/CFS cohort (\(n \geq 60\)), the partial \(R^2\) of fiber density for the taVNS TNF-\(\alpha\)-suppression ceiling, after adjusting for baseline HRV, should exceed 0.05. Falsified if that partial \(R^2\) is below 0.05 (the ceiling is fully accounted for by functional tone), or if GI, HRV, and inflammatory readouts do not share a common latent factor.
Limitations: \(V_{\text{str}}\) is not yet parameterized β no ME/CFS structural vagal measurement exists to estimate it, and the taVNS-ceiling prediction has not been tested. It is included here as a placeholder pending ME/CFS gastric-biopsy data and should be removed from the ME/CFS causal model if ME/CFS gastric innervation proves normal. The node may also collapse into existing functional-tone variables if structural and functional deficits prove empirically inseparable (see Five Unresolved Threats to the Structural Vagal Denervation Model). This is a proposed model refinement motivated by Long COVID data, not an estimated model.
Consequence: Treating βhow many vagal nerve fibers surviveβ as a separate quantity from βhow well the surviving fibers are firingβ would tell researchers and trial designers whether vagus-stimulation therapies have a hard ceiling in a given patient β useful for predicting who can and cannot benefit, though it remains an untested modeling proposal.