Micheliolide and ACT001 as Covalent Modulators of Inflammation–Redox–Metabolism Networks in CNS Disorders: An Evidence-Stratified Review of Mechanisms and Translational Challenges Authors: Dongqing Jing, Hongshang Chu Published: 2026 ABSTRACT AbstractMicheliolide (MCL), a guaianolide sesquiterpene lactone bearing an α-methylene-γ-lactone Michael acceptor pharmacophore, and its orally bioavailable prodrug ACT001 have been studied in a growing range of central nervous system (CNS) disorders, including traumatic brain injury, Parkinson’s disease, neuropathic pain, neuromyelitis optica spectrum disorder, and glioblastoma. These compounds act beyond classical pathway inhibition (that is, blockade of any single linear signalling cascade in isolation) by modulating interconnected inflammatory, redox, and metabolic signalling modules. This review integrates preclinical evidence across CNS disease models and proposes a unified three-axis framework—encompassing innate immune-inflammatory regulation, redox homeostasis, and metabolism-associated signal integration—to organise mechanistic findings, stratify evidence by direct CNS validation status, and identify translational gaps. This review is based on a structured search of PubMed and Web of Science for publications on micheliolide, DMAMCL, or ACT001 in central nervous system disorders, and the diseases are discussed in the sequence in which the supporting evidence is presented below. The inflammatory axis (NF-κB/NLRP3/TLR4/MD2/STAT3) is the best-supported CNS-relevant mechanism, with convergent evidence across traumatic brain injury, Parkinson’s disease, neuropathic pain, and glioblastoma, including direct binding to the TLR4 co-receptor MD2 and IKKβ. Redox modulation (KEAP1/NRF2) remains plausible but rests primarily on peripheral models, awaiting CNS validation. The metabolism-associated axis (PKM2, GAPDH, MDK/c-Myc) is an innovative but incompletely validated frontier, with GAPDH Cys247 covalent engagement in glioblastoma as the strongest direct CNS evidence. We propose a disease-stratified model—neurodegeneration as inflammation-dominant, acute brain injury as inflammation–metabolism coupled, autoimmune and pain conditions as glial-lineage-expanded, and brain tumours as multi-axis dysregulated—requiring experimental validation. Although glioblastoma is the most translationally advanced indication for this chemotype, the shared neuroinflammatory, redox, and metabolic mechanisms engaged by MCL/ACT001 provide a strong rationale for their evaluation across non-malignant CNS disorders. MCL and ACT001 represent a promising covalent scaffold for modulating interconnected CNS networks, with glioblastoma as the most translationally advanced indication. Priorities include quantitative CNS pharmacokinetics, long-term safety of covalent pharmacology in post-mitotic neurons, and validation of metabolic–redox coupling in brain-resident cells via chemoproteomic and single-cell approaches.Keywords: micheliolide, ACT001, CNS disorders, neuroinflammation, covalent pharmacology, network modulation Introduction Micheliolide (MCL) is a guaianolide sesquiterpene lactone isolated from Magnolia grandiflora L.1 Its biological activity hinges on an α-methylene-γ-lactone pharmacophore that acts as a Michael acceptor, covalently modifying cysteine residues in redox-sensitive and signalling proteins.2–4 This electrophilic reactivity is the source of its pleiotropic pharmacology across inflammatory and oncogenic pathways.5,6 ACT001 (dimethylaminomicheliolide, DMAMCL) is an orally bioavailable prodrug developed to improve aqueous solubility and systemic exposure while retaining the reactivity of the parent scaffold.7 For much of their published history, that is, from the first pharmacological reports of this chemotype in the early 2010s until relatively recently, MCL and ACT001 have been studied in peripheral inflammatory conditions and cancer models — acute myeloid leukaemia4,8,9 and a range of solid tumours10–12 chief among them. Only more recently has their pharmacological scope extended into central nervous system (CNS) disorders: traumatic brain injury (TBI),13 Parkinson’s disease (PD),14,15 Alzheimer’s disease (AD),16 neuropathic pain,17 neuromyelitis optica spectrum disorder (NMOSD),18 and glioblastoma (GBM).19–22 Early evidence that MCL suppresses LPS-induced neuroinflammatory responses antedates much of the disease-model literature.23 This widening evidence base suggests that their activity may not be limited to conventional anti-inflammatory or anticancer effects. Instead, it appears to intersect with stress-response systems that are directly relevant to CNS pathology. What unites these disparate disease contexts is a set of recurring mechanistic themes — inflammatory signalling, redox regulation, and metabolism-linked transcriptional control, which are partially overlapping in practice. In the CNS, these processes are not abstract pathways but concrete physiological realities: microglial activation, astrocytic reactivity, mitochondrial dysfunction, and blood–brain barrier disruption collectively shape how disease unfolds. Yet most of the mechanistic insights we have still come from peripheral tissues or tumour systems. Translating them to CNS conditions is complicated by differences in pharmacokinetics, cellular composition, and immune microenvironment. Existing reviews have tended to treat MCL and ACT001 as broad-spectrum anti-inflammatory or antitumor agents, with little systematic distinction between CNS and non-CNS evidence.5,6,24 The result is a literature in which translational validity is highly uneven. Pathways such as nuclear factor-κB (NF-κB)/NLR family pyrin domain-containing 3 (NLRP3) inflammasome activation,25 Toll-like receptor 4 (TLR4)/myeloid differentiation factor 2 (MD2)-mediated innate immune signalling,17 signal transducer and activator of transcription 3 (STAT3)-dependent immunomodulation,26 and Kelch-like ECH-associated protein 1 (KEAP1)/nuclear factor erythroid 2-related factor 2 (NRF2)-regulated reactive oxygen species (ROS) control27,28 have been reported across diverse experimental systems, as have metabolic regulators including pyruvate kinase M2 (PKM2), glyceraldehyde-3-phosphate dehydrogenase (GAPDH), and the midkine (MDK)/c-Myc axis.25,29,30 But the evidential strength of these mechanisms varies substantially — from CNS disease models with functional readouts to peripheral or tumour-only systems that lack direct neurobiological validation. We have organised this review around a three-domain framework — inflammatory signalling, redox homeostasis, and metabolism-associated regulation — each representing a potential node of convergence for MCL and ACT001 pharmacology at the biochemical and molecular level. This is not presented as a fully mapped CNS interactome; rather, it is a pragmatic tool for sorting existing evidence according to biological function and translational relevance (Figure 1). Three-axis pharmacological network and evidence hierarchy of MCL/ACT001. (Left) Venn diagram of target distribution across the inflammation axis (blue: NF-κB/NLRP3, TLR4/MD2, STAT3), redox axis (green: KEAP1/NRF2, ROS, GSH), and metabolism axis (Orange: GAPDH, PKM2, MDK/c-Myc), converging on the shared α-methylene-γ-lactone Michael acceptor pharmacophore (structures inset; the dimethylamino modification of ACT001 improves oral bioavailability while retaining the reactive electrophile). (Right) Evidence pyramid ranking the three axes: inflammation (high evidence; direct CNS target engagement and multi-model pathway validation), redox (moderate; biologically plausible but limited direct CNS validation), and metabolism (emerging; supported mainly by peripheral or tumour-derived data). Dashed arrows denote the trajectory toward definitive CNS pharmacological status. Abbreviations: ACT001, dimethylaminomicheliolide fumarate; AD, Alzheimer’s disease; c-Myc, cellular myelocytomatosis oncogene; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GBM, glioblastoma; GSH, glutathione; KEAP1, Kelch-like ECH-associated protein 1; MCL, micheliolide; MD2, myeloid differentiation factor 2; MDK, midkine; NF-κB, nuclear factor-κB; NLRP3, NLR family pyrin domain-containing 3; NMOSD, neuromyelitis optica spectrum disorder; NRF2, nuclear factor erythroid 2-related factor 2; PD, Parkinson’s disease; PKM2, pyruvate kinase M2; ROS, reactive oxygen species; STAT3, signal transducer and activator of transcription 3; TBI, traumatic brain injury; TLR4, Toll-like receptor 4. Within this framework, we stratify available data into four categories: (i) CNS disease models, encompassing neurodegenerative, neurotraumatic, neuroimmune, and pain-related conditions with direct functional readouts; (ii) CNS tumour models, particularly glioblastoma, which offer insight into intracranial pharmacology but do not necessarily reflect non-malignant neural environments; (iii) peripheral disease systems, which support mechanistic plausibility but require CNS-specific validation; and (iv) translational gaps, including brain exposure, unbound drug concentration, target engagement, the long-term safety implications of covalent reactivity, and clinical development constraints. Our aim is to distinguish mechanisms that are robustly supported in CNS contexts from those that remain inferential, and to identify priority areas for future translational investigation of MCL and ACT001 as candidate neurotherapeutic scaffolds. Chemical Pharmacology and Prodrug Design of MCL/ACT001 MCL Structure and the α-Methylene-γ-Lactone Electrophilic Pharmacophore MCL’s defining structural feature is the α-methylene-γ-lactone (α-M-γ-L) moiety, an electrophilic Michael acceptor that reacts with accessible nucleophilic cysteine thiol groups in selected proteins.4,26,31 Crystallographic characterisation of C13-modified MCL derivatives confirms the preservation of the α-M-γ-L geometry upon exocyclic substitution.32 This cysteine-reactive motif gives MCL a chemical logic that is neither fully non-selective nor conventionally reversible. Instead, it sits in an intermediate pharmacological space: a natural product scaffold whose biological effects depend on the interplay between chemical reactivity, target accessibility, and cellular context. The α-M-γ-L pharmacophore can in principle influence multiple signalling modules involved in inflammation, redox regulation, and metabolism-associated signalling.25–28 In CNS disorders, this multi-node potential is particularly relevant because pathological progression typically involves coordinated dysregulation of inflammatory activation, oxidative stress, mitochondrial dysfunction, glial reactivity, and cell survival pathways. But the gap between chemical potential and biological validation is substantial. Only a handful of proteins have been supported by direct residue-level covalent binding evidence,3,26,29 and fewer still have been confirmed in CNS-relevant cellular or disease contexts. The reactivity of MCL provides a mechanistic rationale for multi-target pharmacology, but it does not by itself establish direct CNS target engagement. The coupling between redox and metabolic circuits has been proposed as a rational basis for MCL-based combination therapy in cancer cells.33 MCL is structurally related to parthenolide, another sesquiterpene lactone with a well-characterised electrophilic lactone motif.5,34,35 Comparative studies suggest that differences within the guaianolide scaffold may confer improved chemical stability and more favourable drug-like properties relative to parthenolide-derived compounds.5,34,36 These physicochemical features matter for CNS pharmacology because systemic stability, membrane permeability, tissue distribution, and intracellular accumulation all determine whether a cysteine-reactive compound reaches sufficient concentrations in neural, glial, or tumour compartments.7 The therapeutic application of MCL-derived compounds therefore hinges on balancing electrophilic reactivity for covalent target engagement against chemical stability required for systemic delivery and tolerability.3,34,37 ACT001 Prodrug Design and in vivo Conversion The translational development of MCL has been constrained by poor aqueous solubility and suboptimal oral bioavailability, problems common to sesquiterpene lactones.7,34 ACT001 was designed to circumvent these limitations,4,38 and its pharmacokinetic behaviour has been characterised in rats.7 The dimethylamino group enhances aqueous solubility and improves systemic exposure, while preserving the ability to regenerate the parent MCL electrophilic scaffold following in vivo biotransformation.4,7,38 This prodrug strategy effectively decouples systemic delivery properties from intracellular pharmacological reactivity. Subsequent derivatisation campaigns, including Heck-reaction-based modification of the guaianolide scaffold, expanded the series against drug-resistant acute leukaemic cells.39 Derivatives of both parthenolide and MCL have since been optimised as dual NF-κB/STAT3 inhibitors.40 After oral administration, ACT001 achieves measurable systemic exposure and undergoes bioconversion to MCL-related active species in vivo,7 confirming its feasibility as an orally deliverable platform. The pharmacokinetic parameters that determine CNS pharmacological sufficiency, namely the kinetics and tissue specificity of ACT001-to-MCL conversion, unbound brain exposure, and disease-dependent barrier heterogeneity — remain poorly defined and are analysed in CNS Pharmacokinetics and Blood–Brain Barrier Considerations. The comparative pharmaceutical, CNS exposure, and pharmacodynamic properties of MCL and ACT001 are summarised in Table 1. Table 1Comparative Pharmaceutical, CNS Exposure, and Pharmacodynamic Features of MCL and ACT001FeatureMCLACT001Ref.Chemical natureActive electrophilic scaffoldOrally bioavailable prodrug (DMAMCL)[7]BBB penetration evidenceNot establishedPreclinical brain distribution observed (no quantitative permeability metrics)[7]Systemic stabilityLowHigh (prodrug-enhanced)[7]Oral bioavailabilityPoorImproved[7]CNS exposure quantificationNoneNot available (no Kp uu or free brain concentration reported)[7]Active CNS speciesMCL (delivery-dependent)MCL (via in vivo conversion; spatial/temporal conversion dynamics undefined)[7]CNS target engagement (high confidence)GAPDH Cys247 (glioblastoma cells)Not established[29]CNS target engagement (high confidence, direct binding)NoneMD2 binding (neuropathic pain model; non-covalent interaction)[17]Pathway-level CNS effectsNF-κB/PI3K-Akt (AD); PKM2–p65 (stroke)NF-κB/STAT3/IKKβ (GBM); NF-κB/NLRP3 (TBI, PD)[13,15,16,21,22,25]Clinical translation stagePreclinicalEarly-phase clinical evaluation (GBM; trial registry ref. pending); PCNSL evidence preclinical onlyNote: Quantitative CNS pharmacokinetic parameters (eg, Kp, uu and unbound brain concentrations) have not been reported for either compound; pharmacokinetic-related entries are therefore qualitative descriptors, drawn primarily from.7. Comparative Pharmaceutical, CNS Exposure, and Pharmacodynamic Features of MCL and ACT001 Note: Quantitative CNS pharmacokinetic parameters (eg, Kp, uu and unbound brain concentrations) have not been reported for either compound; pharmacokinetic-related entries are therefore qualitative descriptors, drawn primarily from.7. Covalent Target Classification of Molecular Targets A persistent difficulty in interpreting MCL/ACT001 pharmacology is distinguishing direct covalent protein engagement from non-covalent molecular interactions, downstream pathway modulation, and mechanisms inferred from non-CNS systems. The presence of an α-methylene-γ-lactone electrophilic pharmacophore does not mean that every signalling node is covalently modified in a functionally relevant way. To avoid overinterpretation, we have organised reported molecular effects along an evidence continuum — from residue-resolved covalent engagement to indirect extrapolated mechanisms (Table 2 and Table 3). Table 2Evidence Hierarchy of MCL/ACT001 Mechanisms in CNS DisordersMechanismCNS-Relevant Evidence LevelDisease ModelCell/Tissue TypeMolecular Interaction TypeFunctional OutcomeConfidence (CNS Relevance)Ref.GAPDH Cys247Covalent engagement (residue-resolved)GBMGlioma cellsResidue-resolved covalent modificationTemozolomide sensitisationHigh (tumour-specific)[29]PKM2 Cys424Covalent engagement (residue-resolved; non-CNS model)LeukaemiaLeukaemia cellsResidue-resolved covalent modificationMetabolic modulationLow (CNS relevance)[3]TLR4/MD2Direct molecular interaction (non-covalent or unresolved covalent status)Neuropathic painSpinal glial systemDirect protein–ligand bindingAnalgesic effectHigh[17]IKKβ/NF-κB/ROSPathway-level pharmacodynamic effectsGBMGlioma cellsFunctional target interaction (pull-down evidenced; covalent site unresolved)Tumour cell deathMedium-high[21]STAT3/PD-L1Pathway-level pharmacodynamic effectsGBMGlioma, immune microenvironmentIndirect pathway modulationImmune suppression reversalMedium-high[22]Akt/NF-κB/NLRP3Pathway-level pharmacodynamic effectsTBIMicroglia, brain tissueIndirect pathway modulationBehavioural and histological improvementMedium-high[13]NF-κB/NLRP3Pathway-level pharmacodynamic effectsPDMicroglia, dopaminergic neuronsIndirect pathway modulationMotor and neurochemical improvement (with low-dose L-DOPA)Medium-high[14,15]NF-κB/PI3K-AktPathway-level pharmacodynamic effectsADMicroglia, brain tissueIndirect pathway modulationCognitive improvementLow (single unreplicated study)[16]AQP4-IgG–mediated astrocyte injuryPathway-level pharmacodynamic effectsNMOSDAstrocytesIndirect pathway modulationAttenuation of astrocyte injuryMedium[18]T cell-mediated anti-tumour immunityPathway-level pharmacodynamic effectsPCNSLT cells, tumour immune microenvironmentIndirect pathway modulationTumour growth inhibitionMedium[41]PKM2–p65Extrapolated mechanismsStroke-related macrophagesMacrophagesProtein–protein interaction disruptionInflammasome inhibitionMedium (extrapolated CNS relevance)[25]KEAP1/NRF2Extrapolated mechanismsAtherosclerosis, heart, pancreasMacrophages, peripheral tissuesRedox pathway modulationAntioxidant protectionMedium-low (CNS)[28,42,43]Note: Evidence levels follow the four-tier classification defined in Covalent Target Classification of Molecular Targets–2.4 (residue-resolved covalent engagement, direct molecular interaction, pathway-level pharmacodynamic effects, extrapolated mechanisms); rows are ordered by descending evidence tier. Evidence Hierarchy of MCL/ACT001 Mechanisms in CNS Disorders Note: Evidence levels follow the four-tier classification defined in Covalent Target Classification of Molecular Targets–2.4 (residue-resolved covalent engagement, direct molecular interaction, pathway-level pharmacodynamic effects, extrapolated mechanisms); rows are ordered by descending evidence tier. Table 3Representative Molecular Nodes Directly Modulated by MCL/ACT001: Evidence ClassificationNodeCompoundEvidence TypeCovalent SiteModel SystemCNS ValidationConfidenceCaveatRef.GAPDHMCLDirect covalent modificationCys247Glioma cellsYes (GBM, tumour CNS context)HighTumour-restricted CNS evidence[29]PKM2MCLDirect covalent modificationCys424Leukaemia cellsNoLow (CNS relevance)Haematological model[3]TLR4/MD2ACT001Direct protein–ligand interaction—Neuropathic painYesHighBinding mode not structurally resolved[17]IKKβACT001Direct binding (covalent site unresolved)—GliomaYes (GBM, tumour CNS context)Medium-highCovalent engagement not established[21]STAT3ACT001Functional pathway modulation—GliomaYes (GBM, tumour CNS context)Medium-highIndirect via phosphorylation[22]TrxR1MCLTrxR1-associated redox modulation (covalent engagement unresolved)—HeLa/cancer cellsNoLow (CNS relevance)Non-CNS model; extrapolated mechanism (Extrapolated Mechanisms)[44,45]PKM2–p65MCLProtein–protein interaction disruption—Stroke macrophagesPartialMediumPeripheral macrophages, not CNS-resident microglia[25]Notes: Nodes with direct molecular interaction evidence are listed together with two borderline cases retained for completeness (STAT3: functional pathway modulation; TrxR1: covalent engagement unresolved). Other pathway-level pharmacodynamic effects (eg, NF-κB, Akt, NLRP3) and the extrapolated KEAP1/NRF2 axis are classified by disease context in Table 2; MDK/c-Myc is discussed in Metabolism-Associated Signal Integration Axis: Emerging and Provisional and GBM/PCNSL: Multi-Axis Dysregulation. Representative Molecular Nodes Directly Modulated by MCL/ACT001: Evidence Classification Notes: Nodes with direct molecular interaction evidence are listed together with two borderline cases retained for completeness (STAT3: functional pathway modulation; TrxR1: covalent engagement unresolved). Other pathway-level pharmacodynamic effects (eg, NF-κB, Akt, NLRP3) and the extrapolated KEAP1/NRF2 axis are classified by disease context in Table 2; MDK/c-Myc is discussed in Metabolism-Associated Signal Integration Axis: Emerging and Provisional and GBM/PCNSL: Multi-Axis Dysregulation. Covalent engagement (residue-resolved). At the highest level of evidence are proteins with both compound–protein interaction data and site-specific identification of cysteine engagement. Two residue-resolved covalent targets have been reported for MCL: GAPDH at Cys247 in glioblastoma,29 the only case demonstrated in a CNS-relevant context — and PKM2 at Cys424 in leukaemia models,3 which supports the chemical plausibility of metabolism-associated cysteine reactivity. The mechanistic details and disease implications of both are developed in Metabolism-Associated Signal Integration Axis: Emerging and Provisional. Direct molecular interaction (non-covalent or unresolved covalent status). A second tier includes proteins for which direct binding and functional modulation have been demonstrated, but where residue-level covalent modification has not been conclusively established. The clearest example is the TLR4 co-receptor MD2: ACT001 binds MD2 directly and functionally disrupts TLR4/MD2 signalling in a neuropathic pain model (mechanistic details in Inflammatory Axis: the Best-Supported CNS-Relevant Axis).17 Classification as covalent engagement would require site-resolved biochemical evidence that is currently lacking. Pathway-level pharmacodynamic effects. A third tier comprises pathway-level effects where signalling outputs are consistently modulated after MCL/ACT001 exposure, but proximal molecular targets remain undefined. In traumatic brain injury, suppression of Akt phosphorylation, NF-κB activation, and NLRP3 inflammasome activity has been reported.13 In glioblastoma, reduced STAT3 phosphorylation and programmed death-ligand 1 (PD-L1) downregulation indicate immunomodulatory pathway activity.22 These are reproducible CNS pharmacodynamic effects, but whether Akt, NF-κB, STAT3, or NLRP3 represent direct molecular targets of ACT001 is unproven. They should be read as downstream pathway modulation rather than evidence of direct molecular engagement. Extrapolated Mechanisms Finally, extrapolated mechanisms are signalling relationships primarily supported in peripheral, haematological, or non-CNS tumour systems. KEAP1/NRF2 axis regulation,27,28,46 thioredoxin reductase 1 (TrxR1)-associated redox modulation,44,45 and PKM2-dependent metabolic–inflammatory coupling fall into this category when extended to non-neoplastic CNS disorders.25 These mechanisms are well established in peripheral immune or tumour contexts,27,28,47 but their transferability to brain-resident cell types (neurons, astrocytes, and microglia) remains incompletely validated. This evidence-based continuum (Table 3) reveals a graded pharmacological landscape: residue-resolved covalent engagement in CNS-relevant systems has been demonstrated for only one protein (GAPDH Cys247 in glioblastoma), whereas most CNS-relevant pharmacodynamic effects likely arise from indirect or network-level modulation. The stratification is intended as a practical tool for mechanistic prioritisation, not a definitive interactome; the methodological priorities needed to convert pathway-level observations into causal target assignments are discussed in Future Methodological Directions. CNS Pharmacokinetics and Blood–Brain Barrier Considerations Current Evidence for CNS Exposure Whether ACT001 and MCL achieve pharmacologically active concentrations in brain parenchyma is the threshold question for CNS therapeutic relevance. A pharmacokinetic study in rats detected ACT001 and MCL-related metabolites in brain tissue after systemic administration (that is, oral or intraperitoneal dosing, as opposed to local intracerebral or intrathecal delivery),7 and intracranial tumour models show functional pharmacodynamic effects.19–21,41 But detection in tissue homogenates does not establish sufficient unbound drug concentrations for consistent target engagement. Whole-tissue measurements conflate vascular, interstitial, cerebrospinal, and intracellular compartments, a critical limitation in tumour-bearing brains where blood–brain barrier (BBB) and blood–tumour barrier integrity is spatially heterogeneous, risking localised overestimation of CNS penetration. The quantitative parameters required for mechanism-based translation remain unavailable. No study has reported the unbound brain-to-plasma partition coefficient (Kp, uu), the metric that most directly reflects free drug distribution across the BBB. Quantitative target occupancy at disease-relevant nodes, such as GAPDH in glioblastoma29 or MD2/TLR4 in neuroinflammatory contexts,17 is also lacking. Whether ACT001 and MCL are substrates of BBB efflux transporters (P-glycoprotein, ABCG2), and how plasma protein binding and brain tissue binding influence the unbound fraction, remain unknown. The prodrug nature of ACT001 further complicates interpretation: CNS exposure may reflect parent compound penetration, peripheral conversion, or both.7,48 Until these pharmacokinetic (PK) determinants are resolved, preclinical efficacy remains mechanistically unanchored (Table 1). In vitro, micheliolide also inhibits several UDP-glucuronosyltransferase isoforms, indicating potential for metabolism-based drug–drug interactions that has not been explored in CNS contexts.36 Route of administration is a central determinant of CNS targeting. ACT001 is administered orally in clinical studies, and oral dosing therefore constitutes the reference route for translation; most preclinical CNS studies have likewise relied on systemic (oral or intraperitoneal) administration, which depends on blood–brain barrier penetration of the parent electrophile or its active metabolites. Alternative CNS-directed routes — such as intranasal delivery, convection-enhanced delivery, and nanoparticle-mediated transport across the blood–brain barrier — remain unexplored for this chemotype; nanoparticle-mediated delivery of micheliolide analogues has so far been demonstrated only in non-CNS settings, for example bone-targeted nanoparticles delivering MCL analogues to leukemic stem cells in the bone marrow,49 and whether similar carrier strategies can enhance CNS exposure remains to be tested. For non-malignant CNS indications, in which barrier integrity is comparatively preserved, optimising oral exposure and confirming unbound brain concentrations are likely to matter more than route innovation; for intracranial tumours with spatially heterogeneous barrier disruption, local or carrier-mediated delivery may warrant consideration as an adjunct to systemic dosing. Disease-Specific BBB and Blood–Tumour Barrier Considerations The implications of CNS exposure are not uniform across disease states. The BBB and blood–tumour barrier are dynamic, regionally heterogeneous interfaces whose integrity varies with pathological context. In glioblastoma, local barrier disruption may facilitate drug entry into contrast-enhancing tumour regions, but invasive cells infiltrate beyond leaky vasculature into non-enhancing parenchyma. Measurable accumulation in tumour tissue therefore does not imply uniform exposure. In acute brain injury (TBI, stroke), barrier permeability evolves temporally; early post-injury leakage gives way to partial restoration — meaning CNS exposure depends on treatment timing as well as systemic PK; notably, ACT001 treatment has been shown to preserve blood–brain barrier integrity in this setting.13 In chronic neurodegeneration (AD, PD), BBB integrity is relatively preserved outside inflamed regions, making CNS penetration a dominant efficacy constraint; accordingly, the cognitive benefit observed with MCL in AD models implies at least functionally relevant central exposure.16 Neuropathic pain and NMOSD — both responsive to ACT001 in preclinical models,17,18 span multiple anatomical compartments (spinal cord, dorsal root ganglia, peripheral nerves) whose barrier properties differ and remain uncharacterised for these compounds. These considerations underscore that binary “BBB penetrant” classification is insufficient. CNS exposure is a compartment- and disease-stage–dependent process in which regional barrier integrity, temporal pathology evolution, and cellular target localisation jointly determine pharmacologically relevant unbound drug availability. The primary limitation is not absence of detectable brain distribution, but the lack of a validated framework linking systemic exposure to target engagement across distinct CNS microenvironments. Until this quantitative bridge is established, translation into non-oncological CNS indications will remain empirical rather than mechanistically grounded. Mechanistic Axes of MCL/ACT001 Pharmacology in CNS Disorders Inflammatory Axis: The Best-Supported CNS-Relevant Axis Among the proposed mechanistic dimensions of MCL and ACT001 pharmacology, the inflammatory axis is currently the most robustly supported CNS-relevant component. This rests on convergent evidence across multiple disease models, direct or near-direct molecular interaction in selected contexts, and reproducible pathway-level pharmacodynamic effects linked to functional outcomes. Across CNS-related conditions (TBI, neuropathic pain, PD, AD, and glioblastoma), treatment with ACT001 or MCL has consistently been associated with attenuation of neuroinflammatory signalling and improvement in disease-relevant phenotypes.13–17,21–23 The most frequently implicated modules are Akt/NF-κB/NLRP3 signalling, TLR4/MD2/MyD88 pathway activation, and STAT3-associated immune regulation.13,17,22 But the level of mechanistic support varies. MD2 is supported by direct binding evidence in a neuropathic pain model,17 whereas Akt, NF-κB, NLRP3, and STAT3 are primarily supported by pathway-level pharmacodynamic data without confirmed direct target engagement.13–15,21,22 Unless otherwise stated, the pathway-level effects described below should be read as downstream pharmacodynamic modulation rather than direct molecular engagement, following the evidence continuum defined in Covalent Target Classification of Molecular Targets. It should be emphasised that, of the nodes discussed above, only MD2 and IKKβ are supported by direct compound–protein binding evidence; Akt, NF-κB, STAT3, and NLRP3 are currently pharmacodynamic readouts rather than confirmed direct targets of MCL or ACT001 (Covalent Target Classification of Molecular Targets; Table 3). These modules are nevertheless mechanistically related rather than independent: they constitute sequential and cross-regulated tiers of the innate immune signalling cascade, in which TLR4/MD2 engagement activates MyD88-dependent IKKβ/NF-κB signalling, NF-κB-driven priming licenses NLRP3 inflammasome activation, PI3K/Akt cross-regulates NF-κB activity, and STAT3 operates as a parallel cytokine-responsive node feeding back on inflammatory transcription. In TBI, ACT001 has been evaluated in a controlled cortical impact model, where treatment attenuated microglial activation, reduced inflammatory mediator production, and preserved blood–brain barrier integrity.13 These effects were associated with reduced Akt phosphorylation, inhibition of NF-κB p65 nuclear translocation, and suppression of NLRP3 inflammasome activity.13 In neuropathic pain, ACT001 provides the clearest example of direct molecular engagement in a CNS-relevant context. ACT001 binds the TLR4 co-receptor MD2, disrupts TLR4/MD2 complex formation, inhibits myeloid differentiation primary response protein 88 (MyD88)-dependent downstream signalling, and suppresses NF-κB/MAPK-associated spinal neuroinflammation.17 In PD models, ACT001 and DMAMCL suppressed NLRP3 inflammasome activation, decreased pro-inflammatory cytokine production, and preserved dopaminergic neurons in 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-induced neurodegeneration.14,15 But as in TBI, the evidence remains predominantly pathway-level and phenotypic, with no direct demonstration of target occupancy in nigrostriatal compartments. In AD models, MCL improved cognitive performance and attenuated hippocampal neuroinflammatory responses in 5×FAD transgenic mice.16 These effects were associated with suppression of NF-κB signalling and modulation of phosphoinositide 3-kinase (PI3K)/Akt pathways.16 While this provides CNS disease-model evidence in a neurodegenerative context, quantitative CNS exposure, cell-type-specific engagement, and long-term disease-modifying effects remain to be established. In glioblastoma, inflammatory signalling intersects with tumour cell survival, immune evasion, and redox regulation. ACT001 has been reported to bind IκB kinase-β (IKKβ) directly and suppress NF-κB signalling, thereby modulating the NF-κB/manganese superoxide dismutase (MnSOD)/ROS axis and promoting ROS-mediated tumour cell death;21 because the covalent site remains unresolved, this mechanism is classified as direct molecular interaction rather than residue-resolved covalent engagement (Covalent Target Classification of Molecular Targets), and its redox consequences are discussed further in Redox Axis: Plausible but Incompletely Validated in Non-Neoplastic CNS Disorders. Notably, residue-level covalent targeting of IKKβ has been demonstrated for the related analogue epoxymicheliolide (Leu281/Leu25) in renal cell carcinoma,50 supporting the chemical feasibility of IKKβ engagement by this chemotype. In parallel, suppression of STAT3 phosphorylation and downregulation of PD-L1 suggest immunoregulatory effects within the tumour microenvironment;22 these STAT3/PD-L1 effects rest on pathway-level pharmacodynamic data and should not be read as direct molecular targeting. Micheliolide has also been shown to suppress U251MG glioma cell viability, migration and invasion via the NF-κB pathway.51 The inflammatory axis exhibits the highest level of CNS-relevant support among the proposed mechanisms, integrating cross-disease reproducibility, partial direct molecular interaction (MD2), and consistent linkage between pathway modulation and functional outcomes. Most components remain supported primarily by pathway-level evidence rather than confirmed direct or covalent target engagement. Redox Axis: Plausible but Incompletely Validated in Non-Neoplastic CNS Disorders The redox axis, centred on KEAP1/NRF2 regulation and ROS homeostasis, is mechanistically plausible but less directly validated in CNS disorders than the inflammatory axis. Its evidence base is dominated by peripheral organ injury, vascular inflammation, and non-CNS tumour models,28,42,43,52,53 while direct demonstration of KEAP1/NRF2 engagement in neurons, astrocytes, microglia, or non-neoplastic brain tissue remains limited. Peripheral studies provide strong support for MCL-associated NRF2 activation. In atherosclerosis, MCL suppressed macrophage ferroptosis through the KEAP1/NRF2/HO-1/GPX4 axis, attenuating plaque formation.27,28 In myocardial ischaemia–reperfusion injury, MCL alleviated oxidative stress and inflammation through KEAP1/NRF2-mediated mechanisms.42 In severe acute pancreatitis, MCL potentiated NRF2-associated antioxidant and anti-inflammatory responses.43 These studies suggest that MCL can modulate NRF2-dependent cytoprotective programmes in peripheral tissues, although they do not establish whether equivalent mechanisms operate in CNS-resident cells. Direct CNS-related redox evidence is most evident in glioblastoma, but the functional direction differs from peripheral antioxidant protection. In GBM models, ACT001 modulated the NF-κB/MnSOD/ROS axis and promoted ROS accumulation, contributing to tumour cell death.21 This reflects pro-oxidant tumour suppression rather than NRF2-mediated neuroprotection. GBM redox evidence therefore cannot be extrapolated to support antioxidant effects in non-neoplastic CNS disorders. Instead, it highlights the context-dependent nature of MCL/ACT001 redox pharmacology: in non-malignant peripheral tissues, redox modulation may be cytoprotective, whereas in tumour cells with elevated basal oxidative stress, it may promote oxidative overload and cell death.54 Whether MCL/ACT001 redox modulation is net protective or cytotoxic in CNS-resident cells, where redox buffering operates under physiological rather than malignancy-associated oxidative stress — has not been systematically resolved. Several critical lines of evidence remain absent. Direct demonstration of KEAP1 engagement or covalent modification by MCL/ACT001 in brain tissue has not been established. Outside the CNS, direct redox-enzyme engagement has been reported, including thioredoxin reductase targeting in cervical cancer cells,45 but equivalent evidence in neural cells is lacking. NRF2 activation, including nuclear translocation or transcriptional upregulation of canonical target genes, has not been systematically validated in neurons, astrocytes, microglia, oligodendrocytes, or endothelial cells. Moreover, cell-type-specific responses to electrophilic modulation remain poorly characterised, despite likely heterogeneity driven by differences in glutathione availability, mitochondrial load, basal oxidative tone, and electrophile-handling capacity across CNS cell populations. The redox axis should therefore be regarded as a plausible but incompletely validated CNS mechanism. Peripheral KEAP1/NRF2 evidence supports biological plausibility, and GBM studies demonstrate CNS tumour-related ROS modulation; however, direct evidence for NRF2-mediated cytoprotection or redox reprogramming in non-neoplastic CNS disorders remains insufficient. In the absence of direct CNS target engagement and cell-type-resolved validation, current support for this axis remains inferential rather than definitive. Metabolism-Associated Signal Integration Axis: Emerging and Provisional The metabolism-associated signal integration axis is conceptually innovative but currently supported by the lowest level of direct CNS evidence. This axis includes mechanisms through which MCL or ACT001 may alter metabolic enzymes or metabolism-associated signalling nodes that also regulate inflammation, survival, stemness, and therapeutic resistance. Key mechanisms include PKM2-associated inflammatory transcription, GAPDH-associated non-metabolic signalling, and MDK/c-Myc-associated glioma stemness and temozolomide resistance.25,29,30 Most evidence, however, derives from peripheral macrophages, haematological malignancies, or glioma cell systems rather than from primary CNS-resident cells. The PKM2–p65 mechanism provides an important example of metabolism–inflammation coupling. MCL interfered with the interaction between PKM2 and NF-κB p65 in macrophage-related ischaemic stroke models, suppressing inflammasome activation and inflammatory cytokine production.25 This suggests that PKM2 can function beyond glycolysis as a transcriptional co-regulatory node linking metabolic state to inflammatory signalling. But because macrophages and microglia differ in ontogeny, metabolic state, and tissue-specific signalling architecture,55,56 direct validation in microglia and in vivo ischaemic brain tissue is required before this mechanism can be considered established for CNS pathophysiology. PKM2 Cys424 provides direct covalent evidence for the chemical capacity of MCL to engage metabolic enzymes, but not direct CNS evidence. MCL covalently modifies PKM2 at Cys424 in leukaemia cells, promoting PKM2 tetramer formation and altering malignant cell metabolism.3 This demonstrates residue-level engagement of a metabolism-associated target, but because the study was conducted in leukaemia models, it should be used to support the plausibility of MCL-mediated metabolic enzyme modulation rather than as direct evidence for CNS metabolic reprogramming. The strongest CNS tumour-related covalent evidence in this axis concerns GAPDH. MCL binds GAPDH at Cys247 in glioblastoma cells and promotes GAPDH nuclear translocation, contributing to glioblastoma cell death in combination with temozolomide.29 This provides residue-level evidence in a CNS tumour context. But its implications should be restricted to glioblastoma unless similar GAPDH engagement is demonstrated in non-neoplastic neural cells. GAPDH is known to have cell-type- and context-dependent non-glycolytic functions, and the consequences of modulating GAPDH nuclear translocation in neurons, astrocytes, or microglia may differ substantially from those observed in glioblastoma cells. MDK/c-Myc signalling has been implicated in glioma stemness and temozolomide resistance,30 although whether MCL or ACT001 functionally modulates this axis remains to be experimentally determined. This mechanism is best interpreted as tumour-associated signal integration rather than as a canonical metabolic mechanism. It is relevant to the broader theme of metabolism-associated survival and resistance networks in glioma, but it should not be overgeneralised to non-tumour CNS disorders. Overall, the metabolism-associated signal integration axis is promising but remains incompletely validated: residue-resolved engagement is confined to tumour contexts (GAPDH Cys247, PKM2 Cys424),3,29 PKM2–p65 disruption provides a plausible metabolism–inflammation link,25 and direct evidence in primary CNS-resident cells remains sparse. This axis should be framed as an emerging hypothesis requiring the chemoproteomic and cell-type-specific validation outlined in Future Methodological Directions. Cross-Axis Coupling: An Integrative but Provisional Framework The three axes described above are biologically interconnected, but the extent to which MCL and ACT001 coordinate these axes in CNS tissue remains to be established. Cross-axis coupling should therefore be presented as an integrative framework for organising existing evidence rather than as a confirmed systems-level mechanism (Figure 1). Inflammatory and redox signalling are linked through reciprocal regulation: NF-κB activation induces inflammatory and redox-associated mediators, while ROS modulates NF-κB through redox-sensitive signalling events. MCL/ACT001 engage this interface on both sides, suppressing NF-κB-associated inflammatory signalling (Inflammatory Axis: the Best-Supported CNS-Relevant Axis) and modulating ROS/NRF2 pathways in glioblastoma and peripheral tissues (Redox Axis: Plausible but Incompletely Validated in Non-Neoplastic CNS Disorders).21,28,42 Whether such bidirectional redox control operates in non-neoplastic CNS tissue remains unknown. Metabolism and inflammation may be connected through PKM2-associated signalling: the disruption of PKM2–p65 interaction described in Metabolism-Associated Signal Integration Axis: Emerging and Provisional provides a potential molecular bridge between glycolytic state and inflammatory transcription, pending confirmation in microglia or other CNS-resident cells. Metabolism and redox regulation may further interact through glycolytic flux, mitochondrial stress, and NADPH/ROS balance.57 Whether GAPDH Cys247 engagement and ROS-associated pathways in glioblastoma (Metabolism-Associated Signal Integration Axis: Emerging and Provisional) operate as a coordinated metabolism–redox programme remains uncertain; direct quantitative data in CNS cell types are lacking. The proposed three-axis model predicts that different CNS disease categories may emphasise different dominant mechanisms. This disease-stratified interpretation is developed in Disease-Stratified Actions in CNS Disorders and should be understood as a working model that requires further experimental validation. Converting this integrative framework into an established mechanism will require direct validation of KEAP1/NRF2 activation in CNS-resident cells, confirmation of PKM2–p65 disruption in microglia and ischaemic brain tissue, and assessment of GAPDH Cys247 engagement in non-neoplastic neural cells. The corresponding methodological platforms — chemoproteomic mapping, human iPSC-derived organoids, and PET-based occupancy imaging — are detailed in Future Methodological Directions. This hierarchy is offered not as a definitive ranking, but as a provisional map to guide where mechanistic investment is most likely to yield translational traction (Table 2). To make these interconnections explicit, Figure 2 integrates the disease-specific evidence into a single overview map: molecular nodes are grouped by axis and annotated with representative canonical pathway assignments (eg, KEGG Toll-like receptor, NF-kappa B, NOD-like receptor, JAK-STAT, PI3K-Akt, HIF-1, and glutathione metabolism pathways), and each node–disease pair is graded according to the evidence continuum defined in Covalent Target Classification of Molecular Targets. Reading the map column-wise highlights which targets or pathways should be prioritised in each disorder — TLR4/MD2 and NF-κB/NLRP3 in TBI, neuropathic pain, and NMOSD; KEAP1/NRF2 validation in AD and PD; GAPDH and STAT3/PD-L1 in GBM — whereas reading it row-wise exposes the cross-axis coupling (NF-κB–ROS feedback, PKM2–p65 metabolism–inflammation bridging, and GAPDH-mediated redox–metabolism integration) that may explain the context-dependent efficacy of a single covalent pharmacophore. Integrated disease–mechanism map of MCL/ACT001 pharmacology in CNS disorders. Molecular nodes (rows) are grouped by mechanistic axis — inflammation (blue), redox (green), and metabolism (Orange) — and annotated with representative canonical pathway assignments (eg, KEGG pathway maps). Columns represent CNS disease contexts. Cell shading indicates the highest level of supporting evidence according to the evidence continuum defined in Covalent Target Classification of Molecular Targets: residue-resolved covalent engagement (darkest), direct molecular interaction, pathway-level pharmacodynamic effect, and extrapolated or untested (lightest). The map is intended to guide mechanistic prioritisation: column-wise reading highlights priority targets for each disorder, whereas row-wise reading exposes cross-axis coupling (NF-κB–ROS feedback, PKM2–p65 metabolism–inflammation bridging, and GAPDH-mediated redox–metabolism integration). Abbreviations: ACT001, dimethylaminomicheliolide fumarate; AD, Alzheimer’s disease; CNS, central nervous system; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GBM, glioblastoma; KEGG, Kyoto Encyclopedia of Genes and Genomes; MCL, micheliolide; NF-κB, nuclear factor-κB; NMOSD, neuromyelitis optica spectrum disorder; p65, RELA proto-oncogene (NF-κB subunit); PD, Parkinson’s disease; PKM2, pyruvate kinase M2; ROS, reactive oxygen species; TBI, traumatic brain injury. Disease-Stratified Actions in CNS Disorders The pharmacological actions of MCL and ACT001 vary considerably across CNS disease categories, reflecting the distinct cellular substrates and molecular architectures of each pathological context. Rather than invoking independent disease-specific mechanisms, we interpret the available evidence through differential axis dominance: the inflammation, redox, and metabolism axes contribute to therapeutic response in variable proportion depending on the underlying pathology. This stratification represents differential readouts of the same underlying biochemical network architecture governed by covalent cysteine reactivity. This organisational structure is interpretive and provisional — it synthesises existing preclinical findings into an analytical model that awaits direct validation in comparative, standardised multi-disease studies. The following subsections apply this framework to four disease groupings, followed by an integrative synthesis. AD/PD: Inflammation-Axis Dominant Pattern Chronic neurodegenerative disorders such as AD and PD may be conceptualised as inflammation-axis-dominant pathologies: NLRP3 inflammasome suppression,14,15 NF-κB/PI3K pathway modulation,16 and attenuation of microglial hyperactivation have been reproducibly observed (Inflammatory Axis: the Best-Supported CNS-Relevant Axis). What distinguishes this context is the temporal dimension — inflammatory signalling operates as a sustained, self-amplifying circuit that progressively compromises neuronal viability over months to years.14–16,58 Redox and metabolic axes are interpreted as modulatory background influences: neither NRF2-dependent antioxidant modulation nor glycolytic enzyme engagement has been directly validated in AD or PD models (Redox Axis: Plausible but Incompletely Validated in Non-Neoplastic CNS Disorders–Metabolism-Associated Signal Integration Axis: Emerging and Provisional), and if operative, they likely act by attenuating the redox and bioenergetic signals that sustain NF-κB/NLRP3 activation rather than as axis-autonomous therapeutic determinants. TBI/Stroke: Inflammation–Metabolism Coupling Acute brain injury (TBI and ischaemic stroke) can be interpreted as an inflammation–metabolism coupling pattern, differing from chronic neurodegeneration in two dimensions. First, inflammatory cascades and metabolic derangements occur synchronously rather than sequentially — NF-κB/NLRP3 activation13 and PKM2-mediated metabolic reprogramming25,59 form a bidirectionally coupled network within minutes to hours of the insult (Inflammatory Axis: the Best-Supported CNS-Relevant Axis, Metabolism-Associated Signal Integration Axis: Emerging and Provisional). Second, the relative contribution of each axis likely shifts across disease phases: the acute phase (energy failure, excitotoxicity)55 may place greater weight on metabolic-axis intervention, whereas the subacute inflammatory phase may be more responsive to NF-κB/NLRP3-directed modulation. MCL/ACT001’s capacity to engage both axes may therefore offer phase-adaptive therapeutic coverage, although this sequential engagement hypothesis remains to be tested experimentally. NMOSD/Neuropathic Pain: Glial-Lineage-Expanded Regulation NMOSD and neuropathic pain may be interpreted as a glial-lineage-expanded pattern: the inflammation axis remains the primary pharmacological domain, but its cellular scope broadens beyond microglia. In neuropathic pain, ACT001 engages MD2 with direct binding evidence (Inflammatory Axis: the Best-Supported CNS-Relevant Axis),17 and efficacy requires modulation of spinal microglia, astrocytes, and the broader neuroinflammatory unit;17 in NMOSD, ACT001’s astrocyte-protective effect18 extends pharmacological action across CNS-resident glial populations. The distinction from the inflammation-dominant AD/PD pattern is therefore one of cellular scope rather than axis identity. GBM/PCNSL: Multi-Axis Dysregulation GBM may be interpreted as exhibiting multi-axis dysregulation: pathological transformation simultaneously remodels inflammatory, metabolic, and redox homeostasis, and micheliolide/ACT001 engage all three axes in this malignancy, inflammation (IKKβ/NF-κB/MnSOD/ROS,21 STAT3/PD-L1;22 Inflammatory Axis: the Best-Supported CNS-Relevant Axis), redox (ROS-mediated tumour cell death;21 Redox Axis: Plausible but Incompletely Validated in Non-Neoplastic CNS Disorders), and metabolism (GAPDH Cys247,29 PAI-1,60 AEBP1/PI3K/Akt;20 Metabolism-Associated Signal Integration Axis: Emerging and Provisional), with additional immune microenvironment modulation through CXCL10-driven CD8+ T cell recruitment.61 In primary CNS lymphoma (PCNSL), ACT001 enhances T cell-mediated anti-tumour responses,41 extending this immunomodulatory pattern across CNS malignancies. Crucially, the relative contribution of each axis has not been quantitatively deconvoluted (Cross-Axis Coupling: An Integrative but Provisional Framework), and whether axis-specific interventions would recapitulate multi-axis efficacy remains unknown. This multi-axis profile, the most complex within the stratification, underlies the tier positioning of GBM in Evidence Hierarchy and Translational Maturity. Cross-Disease Integration This cross-disease integration rests on the axis-dominance framework of Cross-Axis Coupling: An Integrative but Provisional Framework: Mechanistic Axes of MCL/ACT001 Pharmacology in CNS Disorders examines each axis in isolation, whereas this section weights those axes across disease architectures. The same covalent pharmacophore, the α-methylene-γ-lactone Michael acceptor, engages multiple nodes across the inflammation–redox–metabolism network, and each disease-specific therapeutic signature emerges from which nodes are most critically positioned within that disorder’s pathology. As emphasised in the preamble to this section, this stratification remains provisional pending comparative validation. Translational Readiness, Safety, and Future Directions Evidence Hierarchy and Translational Maturity Tier 1 — Glioblastoma. GBM represents the most translationally advanced indication for MCL/ACT001,62 supported by convergent evidence spanning tumour cell-intrinsic mechanisms, immune microenvironment modulation, and combination-therapy synergy63 (mechanistic details in Inflammatory Axis: The Best-Supported CNS-Relevant Axis–Metabolism-Associated Signal Integration Axis: Emerging and Provisional): residue-resolved covalent engagement of GAPDH Cys247 with temozolomide sensitisation,29 direct IKKβ binding with NF-κB suppression,21 STAT3/PD-L1 inhibition,22,64 and CXCL10-driven CD8+ T cell recruitment.61 Emerging data further highlight MDK/c-Myc signalling as a critical axis governing glioma stemness and temozolomide resistance;30,65 as noted in Metabolism-Associated Signal Integration Axis: Emerging and Provisional, functional modulation of this network by MCL or ACT001 has not been experimentally demonstrated. Collectively, these findings provide a multi-layered mechanistic rationale for therapeutic synergy with current standard-of-care regimens. However, despite this relatively mature evidentiary architecture, important translational gaps remain. Quantitative target engagement of key nodes such as GAPDH in orthotopic GBM models has not been established, and Kp, uu remains unreported. Thus, GBM currently represents the highest-confidence indication, although its translational maturity remains incomplete rather than definitive (Figure 3). Translational roadmap and future priorities for MCL/ACT001 CNS development. (Top left) Translational maturity pyramid (Tier 1 = highest readiness): Tier 1, GBM (most advanced indication; Kp, uu and target occupancy unquantified); Tier 2, TBI (BBB disruption; no PK/PD bridge and therapeutic window undefined); Tier 3, PD (L-DOPA adjuvant; no α-synuclein engagement and chronic-dosing data absent); Tier 4, pain/NMOSD/stroke (preclinical only; fragmented across models, without dose–response characterisation); Tier 5, AD (single study; no replication). (Top right) Three complementary methodological priorities: ABPP/chemoproteomics (purple) for target deconvolution — cysteine-reactive probe design, click-chemistry-enabled target mapping, orthotopic GBM validation, and off-target covalent engagement profiling; human iPSC-derived brain organoids (green) for human CNS validation — tumouroid–organoid fusion (GBM), LRRK2/SNCA-mutant PD lines, APP/PS1-edited AD lines, and humanised microglia co-culture; and PET imaging (teal) for quantitative CNS PK/PD — 11C/18F-labelled MCL analogues, brain penetration (Kp, uu) measurement, regional target occupancy mapping, and human-relevant exposure prediction. (Bottom) Chronic covalent safety in post-mitotic neurons. Key concerns: GSH competition and redox-buffering depletion; neuronal protein adduct accumulation; proteostatic stress; cell-type-specific vulnerability (neurons ≫ microglia); and delayed neurotoxicity risk. Required studies: 6-month chronic dosing in non-human primates; neuronal protein adduct histopathology; GSSG/GSH ratio monitoring; and CSF biomarker profiling (these specific study parameters are the authors’ proposed designs rather than established standards). The central challenge is achieving target engagement without cumulative off-target covalent modification in long-lived neurons. Abbreviations: ABPP, activity-based protein profiling; AD, Alzheimer’s disease; CSF, cerebrospinal fluid; GBM, glioblastoma; GSH, glutathione; GSSG, oxidised glutathione; iPSC, induced pluripotent stem cell; Kp, uu, unbound brain-to-plasma partition coefficient; L-DOPA, levodopa; NHP, non-human primate; NMOSD, neuromyelitis optica spectrum disorder; PCNSL, primary central nervous system lymphoma; PD, Parkinson’s disease; PET, positron emission tomography; PK/PD, pharmacokinetic/pharmacodynamic; TBI, traumatic brain injury. The red exclamation mark (!) denotes a cross-cutting safety alert highlighting the unresolved chronic covalent safety implications of the α-methylene-γ-lactone Michael acceptor in post-mitotic neurons (Safety of Chronic Covalent Pharmacology in the CNS). Tier 2 — Traumatic brain injury. TBI sits one tier below because the inflammation-axis mechanism is well-defined in a controlled cortical impact model — microglial attenuation, NF-κB/NLRP3 suppression, and BBB preservation13 (Inflammatory Axis: The Best-Supported CNS-Relevant Axis) — yet no study has demonstrated that neuroprotective plasma exposures correspond to therapeutically relevant brain concentrations, or that the observed anti-inflammatory effects confer direct BBB protection rather than reflecting reduced neuroinflammation. The therapeutic window post-injury also remains undefined. Tier 3 — Parkinson’s disease. In PD, ACT001 demonstrates anti-neuroinflammatory effects in the MPTP mouse model, attenuating microglial NLRP3 inflammasome activation and protecting dopaminergic neurons,15 and improves motor and neurochemical outcomes when combined with low-dose levodopa (L-DOPA).14 However, a conspicuous mechanistic gap limits its positioning: neither MCL nor ACT001 has been shown to directly engage α-synuclein pathology, and no direct evidence supports disease-modifying effects on protein aggregation or propagation. This positions ACT001 more appropriately as an L-DOPA adjuvant targeting neuroinflammatory symptom amplification rather than a disease-modifying therapy. The lack of chronic dosing studies in progressive PD models, and limited validation in other toxin-induced paradigms, further constrains the current evidentiary foundation. Tier 4 — NMOSD, neuropathic pain, and ischemic stroke. These indications share a broadly consistent profile in which mechanistically plausible but partially validated pathways have been reported across distinct acute and inflammatory CNS models. In NMOSD, ACT001 has been shown to attenuate AQP4-IgG–mediated astrocyte injury.18 In neuropathic pain, ACT001 disrupts TLR4 signalling through MD2 binding, thereby suppressing spinal neuroinflammatory activation.17 In ischemic stroke-related models, MCL has been reported to interfere with PKM2–NF-κB p65 signalling and reduce inflammatory responses in ischemia-associated injury contexts.25 However, the current evidence base remains fragmented across acute experimental paradigms, with limited dose–response characterization and insufficient integration into unified pharmacological models. Quantitative pharmacokinetic–pharmacodynamic relationships linking systemic exposure to CNS target engagement have not yet been established in any of these indications, limiting mechanistic translation. Tier 5 — Alzheimer’s disease. AD represents the least supported translational indication within the current evidence base. Although NF-κB-driven neuroinflammation and NLRP3 inflammasome activation are established contributors to AD pathogenesis, direct evidence for MCL in this context rests on a single study16 and remains limited. A single preclinical study has reported that MCL is associated with reduced microglial activation, attenuated neuroinflammatory responses, and decreased Aβ deposition in a 5×FAD mouse model.16 Whether the observed cognitive benefit reflects a direct effect on amyloid processing or plaque formation, or is instead secondary to suppression of NF-κB/PI3K/Akt-driven neuroinflammation, remains unresolved; effects on neuronal development have not been examined.16 However, this evidence has not been replicated across additional AD models or mechanistic layers. Moreover, no data are available regarding modulation of tau phosphorylation, synaptic degeneration, or neurofibrillary tangle propagation. Accordingly, the translational potential of MCL in AD remains preliminary and hypothesis-generating. Safety of Chronic Covalent Pharmacology in the CNS The α-methylene-γ-lactone Michael acceptor in MCL enables covalent modification of cysteine residues, as exemplified by covalent engagement of GAPDH at Cys247 in glioblastoma cells in a disease-relevant context.29 This mechanism introduces a dual safety profile that warrants careful dissection. On one hand, covalent binding reduces target dissociation and enables prolonged pharmacodynamic effects even after drug clearance — which may provide pharmacological advantages in CNS disorders where sustained pathway modulation is desirable. On the other hand, multi-node covalent reactivity introduces a fundamentally different risk calculus from reversible pharmacology: electrophilic natural products can engage multiple cysteine residues across the proteome, as demonstrated by probe-guided target profiling of MCL in leukaemia cells,2 and selectivity observed in reductionist systems may not fully predict off-target engagement in the complex milieu of brain tissue. The CNS amplifies these concerns because of neuronal vulnerability. Post-mitotic neurons do not divide and cannot dilute covalent protein adducts through cell turnover. Cumulative modifications to proteins involved in axonal transport, mitochondrial dynamics, or synaptic function could, in principle, produce delayed neurotoxicity that reversible inhibitors would not cause. This concern is not merely theoretical: chronic exposure to α,β-unsaturated carbonyl-containing electrophiles is known to impose proteostatic stress in long-lived cells, although direct evidence for MCL/ACT001 in post-mitotic neurons is lacking. Furthermore, the brain’s proteostatic buffering capacity may be relatively constrained compared to peripheral organs, raising questions about long-term proteostatic resilience in CNS tissue. A critical, yet underappreciated, aspect of this pharmacology is the intrinsic electrophilicity of these compounds and their competition with abundant endogenous nucleophiles, particularly glutathione (GSH), in biological systems. In the CNS, where GSH is present at millimolar concentrations but its distribution is tightly regulated and regionally heterogeneous, electrophilic Michael acceptors may undergo rapid conjugation with thiol-containing biomolecules prior to selective target engagement. This introduces a fundamental pharmacological tension: moderate electrophilicity is required for covalent modification of targets such as GAPDH Cys247,29 yet the same reactivity can drive preferential scavenging by GSH and other low-molecular-weight thiols. Such non-specific thiol quenching reduces the fraction of compound available for productive target engagement and may perturb intracellular redox homeostasis in a manner that reductionist target-engagement assays do not capture. Neurons are particularly vulnerable to this redox burden. Unlike peripheral tissues, neurons possess limited de novo GSH synthesis capacity and depend substantially on astrocytic GSH supply via the xc− exchange system. Chronic MCL/ACT001 administration could therefore deplete the reduced GSH pool — MCL has been shown to downregulate glutathione in tumour cells66 — either through direct Michael addition or through GSH–drug conjugate formation — and shift the cellular redox balance toward oxidative stress. This is not immediately cytotoxic in acute dosing paradigms, but it may progressively sensitise neurons to endogenous oxidative insults, mitochondrial dysfunction, and excitotoxicity. The extent to which GSH buffering protects neurons versus compromising their long-term redox resilience during chronic electrophile exposure remains unquantified. This competition between target proteins and cellular thiol pools varies significantly across CNS cell types. Microglia and glioma cells, which are generally considered to operate at elevated basal oxidative tone and maintain larger GSH reserves, may tolerate electrophilic challenge differently from neurons and astrocytes. Such heterogeneity complicates predictions of on-target versus off-target engagement in complex CNS microenvironments and implies that safety margins derived from peripheral or tumour models may not translate directly to non-malignant neural tissue. Therefore, the therapeutic window of MCL/ACT001 in CNS disorders is likely governed not only by pharmacokinetic exposure but by the balance between productive target engagement and non-specific thiol quenching within redox-buffered cellular compartments. Quantitative monitoring of GSH redox status — including oxidised glutathione (GSSG)/GSH ratios and GSH–drug conjugate formation — in brain tissue after repeated dosing would address a critical gap in the current safety profile. Current preclinical data indicate that MCL is well tolerated at pharmacologically active doses in a murine model of rheumatoid arthritis,67 and that ACT001 is tolerated in mouse models of traumatic brain injury13 and Parkinson’s disease,14 without dose-limiting toxicity reported in these disease-specific studies. ACT001’s fumarate salt formulation enhances aqueous solubility and oral exposure (ACT001 Prodrug Design and in Vivo Conversion),7 which may indirectly contribute to improved safety margins of the parent scaffold; medicinal-chemistry efforts toward sulfide prodrugs with improved safety profiles have likewise been reported,48 and nanoparticle-mediated delivery of MCL analogues has been explored as an alternative formulation strategy.49 Yet these observations, while reassuring, do not resolve the core safety question. Systematic chronic CNS safety evaluation, including histopathological assessment of neuronal protein adduct burden and biomarkers of cumulative electrophilic stress, has not been reported for MCL or ACT001. It is also important to distinguish the exposure ranges associated with different effect classes. Cytotoxic and tumour-suppressive effects in glioma models generally require sustained micromolar concentrations in vitro,19–22 whereas anti-inflammatory and neuroprotective effects in non-malignant CNS models have been observed at well-tolerated systemic doses without reported dose-limiting toxicity;13–18,23 formal dose–response comparisons across malignant and non-malignant CNS cell types are nonetheless lacking and should be prioritised. Beyond glioblastoma, the principal toxicity concern associated with the anti-inflammatory properties of these compounds is not acute cytotoxicity but the cumulative electrophilic burden and glutathione depletion in post-mitotic neural cells discussed above, together with the theoretical risk of excessive immunosuppression during chronic NF-κB/NLRP3 suppression. The central safety question for MCL/ACT001 in CNS indications is therefore not whether the compounds are pharmacologically effective — the preclinical evidence on this point is substantial — but whether a long-term tissue-selective safety window exists. This window must satisfy two conditions: sufficient target engagement in inflammatory cells (microglia, infiltrating macrophages) and tumour cells to produce therapeutic benefit, without cumulative off-target covalent modification in neurons or glial subpopulations that compromises long-term CNS integrity. Demonstrating such a window will require direct measurement of target versus off-target engagement in brain tissue, rather than reliance on plasma exposure data and peripheral safety readouts. Future Methodological Directions Activity-based protein profiling (ABPP) and cysteine chemoproteomics. The most immediate need is residue-resolved deconvolution of Michael acceptor–cysteine engagement at the biochemical level, moving beyond pathway-level inference to direct, site-specific characterization of covalent adduct formation in CNS-relevant proteins. MCL-derived chemoproteomic probes equipped via late-stage C–H functionalization have already enabled site-directed mapping of cysteine-reactive hotspots in leukaemia cells,2 and target-fishing studies of related sesquiterpene lactones have delineated anti-glioma target profiles;68 adapting such probe platforms to CNS-relevant systems (including cellular systems and ex vivo brain tissue preparations) would elucidate the full landscape of residue-specific cysteine engagement in neural tissue, distinguishing productive on-target modification from non-specific thiol quenching. Human induced pluripotent stem cell (iPSC)-derived brain organoids. Rodent models, while informative, cannot fully predict human CNS pharmacology due to species differences in drug metabolism, BBB transporter expression, and innate immune cell biology. Human iPSC-derived brain organoids — particularly those incorporating humanized microglia through either co-culture systems or genetic engineering (eg, CSF1R-driven microglial differentiation) — offer a physiologically relevant platform for assessing MCL/ACT001 effects on human neuroinflammatory networks. Organoid models of GBM (tumoroid–organoid fusion), PD (LRRK2 or SNCA mutant lines), and AD (APP/PS1-edited lines), albeit with varying levels of maturity, would enable indication-specific evaluation in human neural tissue prior to clinical translation. PET imaging and quantitative target occupancy. The pharmacokinetic/pharmacodynamic (PK/PD) disconnect that plagues MCL/ACT001 CNS pharmacology — no published Kp, uu values, no target occupancy data — can be resolved through radiochemistry. Development of carbon-11 or fluorine-18 labeled MCL/ACT001 analogues would enable positron emission tomography (PET) imaging to measure brain penetration, regional distribution, and — with appropriate tracer kinetic modeling — quantitative engagement of key protein targets such as GAPDH. Radiolabeled probes could also be used to visualize off-target binding hotspots, providing a direct safety readout. The measurement of Kp, uu via PET would finally anchor preclinical PK predictions to human-relevant CNS exposure data. Limitations This review identifies five specific limitations that constrain the strength of conclusions and should guide future primary research. First, the peripheral-to-CNS evidence gap. A substantial fraction of the mechanistic literature cited derives from non-CNS models — cardiovascular,28,37,42 hepatic,44,69 pancreatic,10,43 and immune cell systems70–72 — in which MCL/ACT001 effects on NF-κB, NLRP3, and metabolic reprogramming pathways are well characterized. The implicit assumption that these mechanisms translate directly to CNS cell types (microglia, astrocytes, neurons) is biologically plausible but experimentally unverified for many pathways. Direct validation in CNS-relevant cellular systems remains incomplete. Second, the PK/PD disconnect. Despite extensive pharmacodynamic characterization in cellular and animal models, Kp, uu remains unreported for either compound (Current Evidence for CNS Exposure), nor has target engagement of covalently modified proteins (eg, GAPDH) been quantified in brain tissue. Without exposure-response relationships linking plasma pharmacokinetics to CNS pharmacodynamics, preclinical efficacy cannot be reliably bridged to human dosing predictions. Third, species translation uncertainty. The preclinical evidence base is overwhelmingly derived from rodent models (mouse TBI,13 MPTP PD,14 MCAO stroke25). Differences between rodent and human BBB transporter expression, microglial ontogeny, and drug metabolism (particularly for sesquiterpene lactones undergoing hepatic metabolic processing36) raise substantial uncertainty about human dose projection. The absence of non-human primate PK/PD data or human CNS penetration measurements compounds this limitation. Fourth, chronic covalent safety unknowns. As discussed in Safety of Chronic Covalent Pharmacology in the CNS, the safety implications of chronic irreversible modification in post-mitotic neurons are largely unexplored. No published study has evaluated neuronal protein adduct accumulation, proteostatic stress, or delayed neurotoxicity after extended MCL/ACT001 administration. Prospective evaluation of this risk will require extended chronic dosing studies — on the order of six months or longer, so that cumulative adduct accrual can be detected over a meaningful fraction of the neuronal lifespan — and should include non-human primates, whose glutathione turnover, redox-buffering capacity, and lifespan approximate human physiology far more closely than those of rodents. Such studies should incorporate histopathological assessment of neuronal protein adducts and biomarker monitoring, as proposed in Figure 3. This chronic safety risk is visually flagged by the red exclamation mark (!) in Figure 3 as a cross-cutting barrier spanning all translational tiers. Fifth, evidence heterogeneity across indications. The strength of evidence varies by approximately an order of magnitude from GBM (Tier 1; mechanistic closure and advancement toward clinical translation) to AD (Tier 5; single study, speculative mechanism). This heterogeneity limits the ability to draw unified conclusions about MCL/ACT001 as a therapeutic platform for CNS disorders. Readers should weigh conclusions according to the tier assigned in Evidence Hierarchy and Translational Maturity, recognizing that generalization from the strongest to the weakest indication is not warranted without additional primary evidence. Conclusion The α-methylene-γ-lactone Michael acceptor pharmacophore endows micheliolide with a distinctive capacity for covalent, multi-node protein modification across inflammation–redox–metabolism networks, and its orally bioavailable prodrug ACT001 improves systemic pharmacokinetic properties, with MCL-related species detectable in brain tissue after systemic administration.7 However, the translational parameters required for dose optimisation remain uncharacterised: Kp, uu, target occupancy at specific CNS protein nodes, and exposure–response relationships linking plasma pharmacokinetics to brain pharmacodynamics have all yet to be quantified (CNS Pharmacokinetics and Blood–Brain Barrier Considerations). We propose an evidence-supported integrative hypothesis organising existing data along three coupled axes. The inflammation axis represents the best-supported CNS-relevant dimension, with convergent pathway-level evidence — and, for MD2 and IKKβ, direct binding evidence — demonstrating NF-κB/NLRP3 suppression and microglial functional reprogramming in traumatic brain injury,13 Parkinson’s disease,14,15 neuropathic pain via MD2-targeted TLR4 antagonism,17 and glioblastoma through IKKβ-associated NF-κB suppression and STAT3 inhibition.21,22 The redox axis, centred on KEAP1-NRF2 and broader redox regulatory networks, is mechanistically plausible but remains predominantly supported by peripheral model evidence.27,28,42 The metabolism-signal axis represents an innovative frontier, with site-specific covalent modification of GAPDH Cys247 in GBM29 and emerging evidence for PKM2-mediated inflammation-metabolism coupling, yet direct CNS validation in non-malignant contexts remains fragmentary. These axes form a coupled network in which NF-κB–ROS bidirectional feedback and metabolic enzyme signal transduction create interdependent modulation points that may explain context-dependent efficacy across divergent disease states. This framework supports a disease-stratified interpretation in which distinct patterns of axis dominance emerge across CNS disorders. Neurodegenerative conditions may represent an inflammation-dominant pattern in which sustained NF-κB/NLRP3-driven glial activation constitutes the principal therapeutic target.14,16 Acute brain injury exhibits inflammation–metabolism coupling in which rapid metabolic reprogramming synchronises with inflammatory cascades to amplify secondary damage.13 Autoimmune and pain conditions display glial-lineage-expanded regulation extending to astrocytic and spinal neuroinflammatory compartments.17,18 Brain tumours exhibit multi-axis dysregulation — the simultaneous pathological remodelling of the inflammatory, redox, and metabolic axes described in GBM/PCNSL: Multi-Axis Dysregulation, rather than dysregulation of anatomical axes of the brain — requiring coordinated engagement of tumour cell-intrinsic metabolism, immune microenvironment modulation, and redox perturbation for therapeutic efficacy.20–22,29,60,61,68 This stratification remains provisional pending direct comparative validation (see Disease-Stratified Actions in CNS Disorders preamble). From a translational perspective, glioblastoma represents the most mature indication, with combination therapy potential alongside temozolomide supported by residue-resolved covalent target engagement and synergy studies.29 For non-neoplastic CNS disorders, clinical progression depends on resolving two fundamental gaps: quantitative PK validation confirming therapeutically relevant brain exposure at safe systemic doses, and chronic safety assessment of irreversible covalent modification in post-mitotic neural tissue. Activity-based protein profiling, single-cell transcriptomics, human iPSC-derived organoids, and PET imaging may collectively address the “brain exposure threshold” question that currently constrains translation across non-malignant indications. MCL and ACT001 stand as a promising scaffold for CNS network pharmacology, offering a compelling prototype for how covalent natural product-derived agents may modulate integrated pathological networks; yet their ultimate therapeutic utility will be decided not by the breadth of preclinical mechanism catalogued, but by the rigour with which CNS target engagement, exposure-response relationships, and long-term neural safety are validated in the path forward to human trials. Funding Statement This work was supported by the Scientific Research Foundation for doctoral of Affiliated Hospital of Shandong Second Medical University (2024BSQD08), Taishan Scholars Project of Shandong Province (tsqnz20240857), and The Natural Science Foundation of Shandong Province (ZR2025QC2026Z to H.C.; ZR2024QH154 to D.J.). AI Disclosure During the preparation of this manuscript, we used Kimi K3 (Moonshot AI) to improve language expression, grammatical accuracy, and manuscript structure. This tool was employed solely for editorial assistance, including refining readability, standardising scientific terminology, and enhancing the logical flow of the text. Kimi K3 was not used to generate core research content, scientific hypotheses, mechanism analysis, data interpretation, or conclusions. All key scientific content, logical reasoning, and final manuscript decisions were completed independently by us. We confirm the originality and accuracy of all manuscript content, including the integrity of the entire manuscript and the accuracy of all references. We take full responsibility for the integrity of the whole content, including the accuracy of references. We have reviewed and confirmed that the terms of use of Kimi K3 are appropriate and suitable for publication, and therefore confirm that the use of this AI tool complies with the journal’s requirements. Abbreviations ABCG2, ATP-binding cassette subfamily G member 2; ABPP, activity-based protein profiling; ACT001, dimethylaminomicheliolide fumarate (orally bioavailable prodrug of micheliolide); AD, Alzheimer’s disease; AKT, protein kinase B; AQP4-IgG, aquaporin-4 immunoglobulin G; BBB, blood–brain barrier; CD8, cluster of differentiation 8; c-Myc, cellular myelocytomatosis oncogene; CNS, central nervous system; CXCL10, C-X-C motif chemokine ligand 10; DMAMCL, dimethylaminomicheliolide; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GBM: glioblastoma; GPX4, glutathione peroxidase 4; GSH, glutathione; GSSG, oxidised glutathione; HIF-1, hypoxia-inducible factor 1; HO-1, haem oxygenase-1; IKKβ, IκB kinase-β; iPSC, induced pluripotent stem cell; KEAP1: Kelch-like ECH-associated protein 1; KEGG, Kyoto Encyclopedia of Genes and Genomes; Kp, uu, unbound brain-to-plasma partition coefficient; L-DOPA, levodopa; MAPK, mitogen-activated protein kinase; MCAO, middle cerebral artery occlusion; MCL, micheliolide; MD2, myeloid differentiation factor 2; MDK, midkine; MnSOD, manganese superoxide dismutase; MPTP, 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine; MyD88, myeloid differentiation primary response protein 88; NADPH, reduced nicotinamide adenine dinucleotide phosphate; NF-κB, nuclear factor-κB; NLRP3, NLR family pyrin domain-containing 3; NMOSD, neuromyelitis optica spectrum disorder; NRF2, nuclear factor erythroid 2-related factor 2; PAI-1, plasminogen activator inhibitor-1; PCNSL, primary central nervous system lymphoma; PD, Parkinson’s disease; PD-L1, programmed death-ligand 1; PET, positron emission tomography; PI3K, phosphoinositide 3-kinase; PK, pharmacokinetics; PK/PD, pharmacokinetic/pharmacodynamic; PKM2, pyruvate kinase M2; ROS, reactive oxygen species; STAT3, signal transducer and activator of transcription 3; TBI, traumatic brain injury; TLR4, Toll-like receptor 4; TrxR1, thioredoxin reductase 1. Ethics Statement No new animal or human subjects were involved in this study. No additional ethical approval was required. Author Contributions All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work. Disclosure Dr Dongqing Jing reports Support for the manuscript from The Natural Science Foundation of Shandong Province, during the conduct of the study. Dr Hongshang Chu reports Support for the manuscript from Scientific Research Foundation for doctoral of Affiliated Hospital of Shandong Second Medical University, Taishan Scholars Project of Shandong Province, The Natural Science Foundation of Shandong Province, during the conduct of the study. The authors declare no conflict of interest. References