HSP70 Pathway Enhancement for Neuroprotection
The Kang et al.(Kang et al. 2026) demonstration that exosome-delivered HSP70 mRNA reverses neuroinflammation raises the question of whether endogenous HSP70 enhancement — without engineered exosomes — could provide partial neuroprotection in ME/CFS. Several pharmacological and non-pharmacological approaches upregulate HSP70 expression.
Certainty: 0.45. HSP70 inducers are clinically available for other indications; neuroprotective effect in ME/CFS is inferred from Kang et al. and general HSP70 biology; BBB crossing mechanisms documented by Ramos-Zaldívar et al.; brain-targeted engineering platforms reviewed by Sanadgol et al.. (0.40→0.45: independent convergence with heat therapy as HSP70 induction strategy strengthens mechanistic rationale.) No ME/CFS-specific data.
Candidate agents with existing human safety data:
- Arimoclomol: Co-inducer of the heat shock response; prolongs HSF1 activation; crosses BBB; Phase III completed for ALS (orphan drug status in EU). No ME/CFS data.
- Geranylgeranylacetone (GGA): Gastric mucosal protectant used in Japan for decades; induces HSP70 in CNS at standard doses. Low cost, favourable safety profile, but CNS penetration modest.
- BGP-15: Hydroxylamine derivative; insulin sensitiser + HSP70 co-inducer; Phase II completed for diabetes. CNS penetration data limited.
Mechanistic rationale. Kang et al.(Kang et al. 2026) showed that HSP70 in CNS cells reduces TNF-α, IL-6, and IL-1β while increasing IL-10, BDNF, and pCREB. Pharmacological HSP70 induction would not replicate the spatial targeting of RVG-exosomes but would provide systemic HSP70 elevation, potentially crossing the BBB in regions where permeability is already increased (Section Blood-Brain Barrier Dysfunction). The HSP70 pathway is already relevant to ME/CFS via the HSF1-HSAT2 axis — heat therapy recommendations exist for ME/CFS (Section Chapter Lifestyle and Non-Pharmacological Interventions) — but the HSAT2 activation concern (pericentromeric satellite repeat transcription under HSF1) must be weighed against HSP70’s anti-inflammatory benefits.
Synergy with exosome delivery. Idea 1.5 from brainstorm: HSP70-inducing drugs could provide sustained chaperone capacity while exosome-delivered HSP70 mRNA provides immediate peak levels. The combination would address the delay between drug administration and transcriptional HSP70 induction (hours to days).
Safety. Arimoclomol: generally well-tolerated (GI disturbance most common AE). BGP-15: limited long-term data. Heat therapy: already in the paper with explicit HSAT2 cautions. All pharmacological HSP70 inducers are contraindicated in active infection (HSP70 supports viral replication in some contexts).
Testable prediction. ME/CFS patients receiving HSP70-inducing agents (arimoclomol or GGA, 8 weeks) will show increased peripheral blood mononuclear cell HSP70 levels by ≥30% from baseline and corresponding reduction in serum inflammatory biomarkers (IL-6, TNF-α) compared to placebo. Falsified if HSP70 induction occurs without neuroinflammatory biomarker improvement, or if HSAT2 retroelement RNA increases in plasma EVs proportionally with HSP70 elevation (indicating HSF1-mediated retroelement activation outweighs anti-inflammatory benefit).
Limitations. Certainty 0.35. No ME/CFS data for any HSP70 inducer. The optimal agent, dose, and duration for CNS HSP70 elevation are unknown. The HSAT2 transcription risk under sustained HSF1 activation has not been quantified in ME/CFS patients — this could be a net-harm scenario if retroelement activation outweighs anti-inflammatory benefit. This is a research direction, not a treatment recommendation.
Certainty: 0.20. Macropinocytosis enhancement documented for pioglitazone in cancer cell lines; no CNS exosome uptake data; no ME/CFS data.
Ramos-Zaldívar et al. identified macropinocytosis as one of several mechanisms by which EVs cross the BBB. PPARγ agonists (pioglitazone, rosiglitazone) upregulate macropinocytosis pathway components and have been shown to enhance nanoparticle uptake in some experimental systems. If PPARγ agonism increases exosome internalisation by CNS target cells, it could lower the required therapeutic exosome dose — addressing the very low (less than 1 percent) CNS delivery fraction limitation of systemic administration.
ME/CFS-specific considerations. Pioglitazone has anti-inflammatory effects independent of exosome uptake enhancement (PPARγ activation suppresses NF-κB, reduces microglial activation) — these may provide independent benefit in ME/CFS neuroinflammation. However, pioglitazone carries FDA black-box warnings for congestive heart failure and is associated with weight gain, fluid retention, and bone fracture risk — all problematic in a bedbound/severe population with pre-existing orthostatic intolerance and potential cardiac dysfunction.
Testable prediction. Pioglitazone pretreatment will increase exosome uptake in primary hippocampal neurons by ≥1.5-fold in vitro, measured by flow cytometry or fluorescence microscopy; this effect will be abolished by the macropinocytosis inhibitor EIPA.
Limitations. Certainty 0.20. No CNS exosome uptake data exist for any PPARγ agonist. The macropinocytosis-to-exosome-uptake pathway is documented for some cell types but not neurons or microglia. Pioglitazone safety profile is unfavourable for a non-life-threatening condition; even if effective, the risk-benefit ratio would be challenging. This is a mechanistic probe, not a treatment proposal.
Vagus nerve stimulation transiently increases BBB permeability via cholinergic anti-inflammatory pathway activation. If tVNS applied 30 minutes before exosome administration increases CNS exosome accumulation — as hypothesised from the inflammation-enhanced EV permeability documented by Ramos-Zaldívar et al. — it could provide a non-pharmacological method to boost therapeutic delivery. This is completely untested. The risk is that tVNS-mediated BBB opening is non-selective, potentially allowing entry of peripheral inflammatory mediators alongside therapeutic exosomes. Pilot biodistribution studies in animal models would be required before any human application.