🧠 Nature 2026 · Neuroscience × Microbiome

Gut–Brain Memory Decline Pathway

How an aging gut bacterium hijacks vagal interoception to erase hippocampal memories — and the 10+ intervention points that can reverse it. Based on a landmark Nature 2026 study mapping the complete P. goldsteinii → MCFA → GPR84 → inflammation → vagal dysfunction → hippocampal engram failure cascade.

📄 Nature s41586-026-10191-6 🔬 Thaiss Lab, UPenn 🐁 C57BL/6J Mice 💊 10+ Druggable Targets
The Interoceptive Theory of Cognitive Aging
Memory loss is not solely a brain-intrinsic disease. This study reveals that the gut microbiome actively drives cognitive decline through a peripheral inflammatory pathway that silences vagal communication between the gut and hippocampus.
Key discovery: Transferring aged gut microbiota to young mice abolished their memory within one month — without affecting physical health, neurogenesis, or brain inflammation. The deficit was fully reversible by antibiotics, phage therapy, GPR84 inhibition, or vagal stimulation.
6
Cascade Layers
10+
Intervention Points
100%
Memory Restored
1,133
Species Changed w/ Age
955d
Mean Lifespan (Cohort)
2 wks
ABX → Recovery
🔬
The Thesis
Interoceptive dysfunction as a principle of brain aging

The traditional view of age-related memory loss focuses on brain-intrinsic damage: neuronal death, amyloid plaques, tau tangles, neuroinflammation. This study proposes a radically different model: the brain loses its ability to "listen" to the body. As we age, the gut microbiome shifts toward species like Parabacteroides goldsteinii that produce medium-chain fatty acids (MCFAs). These metabolites activate GPR84 on peripheral myeloid cells, triggering TNF and IL-1β production via the NLRP3 inflammasome. The resulting local inflammation in the gut and mesenteric adipose tissue silences vagal afferent neurons — the very neurons that relay interoceptive signals from the gut to the hippocampus via the nucleus tractus solitarii (NTS). Without this vagal input, hippocampal neurons fail to activate properly, engram formation collapses, and memory declines.

The beauty of this model is its complete reversibility. Every node in the cascade was validated as a therapeutic target: phage therapy to reduce P. goldsteinii, GPR84 inhibitors to block myeloid activation, anti-TNF/anti-IL-1β antibodies to block cytokines, capsaicin to stimulate vagal neurons, and even GLP-1 receptor agonists like liraglutide to restore hippocampal function. This is not a neurodegenerative dead end — it's a treatable peripheral inflammation problem.

📊
Key Experiments
4 orthogonal causal strategies

1. Co-housing: Young mice co-housed with 18-month-old mice for 1 month acquired an aged microbiome and lost NOR + Barnes maze performance.

2. FMT: Germ-free mice receiving aged donor microbiota replicated the cognitive deficit without social contact.

3. Germ-free aging: Germ-free mice showed delayed cognitive decline, retaining memory at 18 months.

4. Antibiotics: Broad-spectrum ABX before/during/after co-housing prevented or reversed memory loss within 2 weeks — even in naturally aged mice.

💡
Paradigm Shifts
What this changes

Brain aging is partly peripheral. You don't need brain inflammation to get memory loss — gut inflammation alone is sufficient.

Interoception ≠ just hunger signals. Vagal gut-brain communication directly feeds hippocampal engram formation.

GLP-1 drugs may protect cognition. Liraglutide restored memory in aged mice, explaining emerging epidemiological data on GLP-1 RA and dementia risk reduction.

Microbiome age ≠ host age. A young mouse with an old microbiome thinks like an old mouse. Microbiome rejuvenation may be sufficient to restore cognitive function.

📈
P. goldsteinii Abundance Over Lifespan
Longitudinal metagenomic tracking in C57BL/6 cohort (n=15)
The 6-Layer Cascade
Click any layer to see mechanistic details. Each node represents a validated step in the pathway from gut dysbiosis to hippocampal memory failure.
⊘ Antibiotics φPDS1 Phage ⊘ GPR84 inhibitor PBI-4050 ⊘ Anti-TNF / Anti-IL-1β NLRP3 KO ⊕ Capsaicin (TRPV1) CCK / GLP-1 / Liraglutide ⊕ Chemogenetic activation DREADD hM3Dq ⊕ CCK / GLP-1 Vagal stimulation LAYER 1 LAYER 2 LAYER 3 LAYER 4 LAYER 5 LAYER 6 🦠 Microbiome Dysbiosis P. goldsteinii overgrowth with aging ⚗️ MCFA Accumulation 3-HOA, decanoic acid, dodecanoic acid → gut lumen + adipose 🔥 GPR84 → Myeloid Inflammation NLRP3 inflammasome → TNF + IL-1β (gut + mesenteric adipose) ⚡ Vagal Interoceptive Dysfunction IL-1β → PHOX2B⁺/TRPV1⁺ vagal afferent silencing 📡 NTS Signal Collapse Nucleus tractus solitarii → medial septum relay failure 🧠 Hippocampal Engram Failure FOS⁺ neuron ↓ in CA1/CA3/DG → NOR + Barnes maze ↓ MEMORY LOSS
📊
Hippocampal FOS Response
Normalized novelty-evoked neuronal activation across conditions
Parabacteroides goldsteinii
The age-associated gut bacterium at the root of the cascade. Its abundance increases steadily with age and is transmissible to young mice through co-housing and FMT.
🦠
Taxonomy & Biology
KingdomBacteria
PhylumBacteroidota
ClassBacteroidia
OrderBacteroidales
FamilyTannerellaceae
GenusParabacteroides
SpeciesP. goldsteinii
Gram stainGram-negative, anaerobic
LPSContains MCFA in lipid A
📊
Evidence for Causality

1. Correlative: Abundance increases with age across the full lifespan (longitudinal cohort, n=15 C57BL/6). Confirmed by both metagenomic sequencing and stool proteomics.

2. Transmissible: Transferred to young mice by co-housing and FMT. Young mice from facilities with naturally high P. goldsteinii show reduced memory.

3. Sufficient: Mono-colonization of germ-free or ABX-treated mice with P. goldsteinii alone induced cognitive impairment (NOR).

4. Specific: Alistipes, Lachnospiraceae, and Lactobacillus did NOT impair cognition despite similar age-related changes.

5. Mechanistic: Culture supernatants (<3 kDa fraction) are sufficient. The active molecule is 3-HOA (and other MCFAs).

⚗️
Medium-Chain Fatty Acid (MCFA) Metabolites
Small molecules that bridge microbial dysbiosis to immune activation
Metabolite Formula Chain Detection Cognitive Effect GPR84 Agonism
3-Hydroxyoctanoic acid (3-HOA) C₈H₁₆O₃ C8 Metabolomics Impairs NOR + FOS Strong
Decanoic acid (Capric) C₁₀H₂₀O₂ C10 Water-insoluble Impairs NOR + vagal Strong
Dodecanoic acid (Lauric) C₁₂H₂₄O₂ C12 Water-insoluble Impairs NOR Moderate

MCFAs increase in the gut lumen with age (conventional mice only, not GF or ABX-treated). They accumulate in mesenteric and inguinal adipose tissue where they trigger GPR84-dependent inflammation. Importantly, MCFAs do NOT cross the blood-brain barrier — their cognitive effects are entirely peripheral.

🧫
Bacteria Cognitive Impact Comparison
Only P. goldsteinii impairs memory among tested species
The GPR84 Myeloid Axis
GPR84 is a MCFA-sensing receptor expressed exclusively on myeloid cells. Its activation triggers the NLRP3 inflammasome and pro-inflammatory cytokine cascade that silences vagal neurons.
DBA mice are naturally GPR84-deficient due to a point mutation. They show delayed cognitive decline despite harboring P. goldsteinii, providing a natural loss-of-function model.
🎯
GPR84 Receptor Profile
GeneGpr84 (EBI2 family)
TypeG protein-coupled receptor (GPCR)
LigandsMCFAs (C9–C14), 3-HOA, embelin
ExpressionMacrophages, monocytes, neutrophils
NOT expressedT cells, B cells, ILCs, neurons, epithelium
DownstreamNLRP3 inflammasome, TNF, IL-1β
InhibitorPBI-4050 (clinical-stage)
AgonistEmbelin (experimental)
Natural KODBA/2 mice (point mutation)
🔥
Inflammatory Cascade

Step 1: MCFAs bind GPR84 on intestinal macrophages and adipose tissue myeloid cells.

Step 2: GPR84 activation triggers NLRP3 inflammasome assembly. Nlrp3−/− mice are fully resistant to 3-HOA-induced memory loss.

Step 3: NLRP3 cleaves pro-IL-1β into mature IL-1β. TNF is also upregulated.

Step 4: Cytokines accumulate locally in the gut and mesenteric adipose — NOT systemically in blood or brain. This is a peripheral inflammatory event.

Step 5: IL-1β acts on IL-1R1 receptors expressed on PHOX2B⁺ vagal afferent neurons. Conditional deletion of Il1r1 on Phox2b⁺ cells blocks the cognitive effect.

🧪
Genetic & Pharmacological Validation
Convergent evidence from 8+ experimental approaches
ApproachTargetEffect on MemoryType
Gpr84−/− miceGPR84ProtectedGenetic KO
DBA/2 mice (natural KO)GPR84ProtectedNatural variant
PBI-4050GPR84RestoredPharmacological
Embelin (agonist)GPR84ImpairedPharmacological
Nlrp3−/−NLRP3ProtectedGenetic KO
Anti-TNF antibodyTNFRestoredBiologic
Anti-IL-1β antibodyIL-1βRestoredBiologic
PLX-3397 (CSF1Ri)All myeloidRestoredPharmacological
Clodronate liposomesPeripheral phagocytesRestoredPharmacological
Ccr2−/−Peripheral myeloid recruitmentProtectedGenetic KO
GPR84−/− → WT chimeraHematopoietic GPR84ProtectedBM chimera
Phox2bCre-Il1r1fl/flIL-1R on vagal neuronsProtectedConditional KO
📊
Cytokine Expression by Tissue
Inflammation is LOCAL (gut + adipose), not systemic
Vagal Interoceptive Circuit
The vagus nerve is the information highway between gut and brain. Aging silences this highway through inflammatory damage to specific sensory neuron subtypes.
Vagal Neuron Subtypes Involved
Intersectional genetic targeting identifies the critical population
MarkerPopulationManipulationEffect on Memory
TRPV1⁺ Vagal + spinal afferents Trpv1DTA ablation / hM4Di silencing Phenocopies aging
PHOX2B⁺ Vagal only (not spinal) Chemogenetic silencing Phenocopies aging
PHOX2B⁺ TRPV1⁺ Vagal TRPV1 (most specific) Intersectional targeting Phenocopies aging
CCKAR⁺ Gut-innervating vagal CCK-SAP ablation Impairs memory
Spinal TRPV1⁺ Spinal afferents Resiniferatoxin / ganglionectomy No effect

The critical population is PHOX2B⁺ TRPV1⁺ vagal afferents innervating the gut. Spinal afferents are not involved. These neurons relay through the nodose ganglion to the nucleus tractus solitarii (NTS), then through the medial septum to the hippocampus.

🧠
Hippocampal FOS⁺ Response by Region
Novelty-evoked neuronal activation (% of young control)
CA1
100%
Young control
CA3
100%
Young control
Dentate Gyrus
100%
Young control
NTS
100%
Young control
📊
Vagal Calcium Imaging
Nodose ganglia neuron responses to duodenal Ensure infusion
Therapeutic Arsenal
Every node in the cascade is a druggable target. Several interventions have existing clinical-stage compounds. Filter by intervention layer.
Microbiome
φPDS1 Bacteriophage
Phage therapy · Preclinical

Targets Parabacteroides. Reduces luminal MCFA levels by altering P. goldsteinii LPS gene expression. Consistent cognitive improvement in aged mice.

75% — Strong preclinical
Microbiome
Broad-Spectrum Antibiotics
Microbiome ablation · Clinical

2 weeks of ABX fully restored memory in both co-housed young mice AND naturally aged mice. Prevents and reverses the phenotype.

90% — Causally validated
Microbiome
Young FMT
Microbiome rejuvenation · Clinical

Implied by reverse experiments: young microbiome → young cognition. GF mice colonized with young donor microbiota show full cognitive capacity.

65% — Indirect evidence
Receptor
PBI-4050 (GPR84 inhibitor)
Small molecule · Phase 2 (IPF)

Negated 3-HOA's cognitive effect. Counteracted P. goldsteinii. Restored youthful NOR in aged mice. Already in clinical trials for idiopathic pulmonary fibrosis.

85% — Clinical-stage drug
Inflammation
Anti-TNF Antibody
Biologic · Approved (RA/IBD)

Rescued memory in aged mice, co-housed young mice, and 3-HOA-treated mice. Existing approved drugs: infliximab, adalimumab, etanercept.

80% — Approved class
Inflammation
Anti-IL-1β Antibody
Biologic · Approved (canakinumab)

Protected aged and co-housed young mice. IL-1β signals directly on vagal IL-1R1. Canakinumab already approved for inflammatory diseases.

80% — Approved class
Inflammation
PLX-3397 (CSF1R inhibitor)
Small molecule · Phase 3 (TGCT)

Depletes myeloid cells. Restored cognition in co-housing AND 3-HOA treatment. Approved as pexidartinib for giant cell tumors.

70% — Broad target
Inflammation
Clodronate Liposomes
Experimental · Preclinical

Depletes peripheral phagocytes without crossing BBB. Rescued 3-HOA effect. Proves brain microglia are NOT involved — the inflammation is entirely peripheral.

65% — Mechanistic tool
Vagal
Capsaicin (TRPV1 agonist)
Natural compound · OTC

Low-dose capsaicin rescued NTS activation, hippocampal FOS, and memory in BOTH aged and co-housed mice. Overcame P. goldsteinii effect. Requires TRPV1⁺ neurons.

85% — Multiple validations
Vagal
DREADD Chemogenetic Activation
hM3Dq · Preclinical

Chemogenetic activation of TRPV1⁺ neurons fully restored hippocampal FOS and NOR in co-housed young mice. Even overcame IL-1β administration.

90% — Strongest rescue
Hormonal
CCK (Cholecystokinin)
Gut peptide · Preclinical

Peripheral CCK administration made aged/co-housed mice cognitively indistinguishable from young controls. Restores hippocampal FOS. Requires TRPV1⁺ neurons. Also overcomes TNF effect.

80% — Strong rescue
Hormonal
Liraglutide (GLP-1 RA)
Approved drug · Obesity/T2D

GLP-1 receptor agonist enhanced memory in aged mice. GLP-1 itself restored cognition via vagal activation. Major implications for semaglutide/liraglutide cognitive benefits.

85% — Approved, massive market
📊
Intervention Efficacy Comparison
NOR discrimination index restoration (% of young control)
Cognitive Risk Estimator
Interactive model based on the pathway nodes identified in the study. Adjust parameters to estimate relative cognitive risk from the gut-brain axis.
🎚️
Risk Parameters
Age (years) 50
Gut Microbiome Diversity Moderate
Systemic Inflammation (CRP) Normal
Spicy Food / Capsaicin Intake Moderate
GLP-1 RA Use None
Dietary Fiber Intake Moderate
Presets
📊
Risk Profile
Moderate Risk
42
COMPOSITE RISK SCORE (0–100)
References
Key publications supporting the gut-brain memory decline pathway.
  1. Thaiss, C.A. et al. Intestinal interoceptive dysfunction drives age-associated cognitive decline. Nature (2026). doi:10.1038/s41586-026-10191-6
  2. Cryan, J.F. & Dinan, T.G. Mind-altering microorganisms: the impact of the gut microbiota on brain and behaviour. Nat. Rev. Neurosci. 13, 701–712 (2012).
  3. de Araujo, I.E. et al. Food reward in the absence of taste receptor signaling. Neuron 57, 930–941 (2008).
  4. Boehme, M. et al. Microbiota from young mice counteracts selective age-associated behavioral deficits. Nat. Aging 1, 666–676 (2021).
  5. D'Amato, A. et al. Faecal microbiota transplant from aged donor mice affects spatial learning and memory via modulating hippocampal synaptic plasticity. Gut 69, 2164–2172 (2020).
  6. Suarez, A.N. et al. Gut vagal sensory signaling regulates hippocampus function through multi-order pathways. Nat. Commun. 9, 2181 (2018).
  7. Josselyn, S.A. & Tonegawa, S. Memory engrams: Recalling the past and imagining the future. Science 367, eaaw4325 (2020).
  8. Wood, S.J. et al. GPR84 mediates inflammation-induced myeloid cell activation. J. Immunol. 195, 1092–1102 (2015).
  9. Reczynska, D. et al. Medium-chain fatty acids modulate innate immune responses. Immunology 168, 542–555 (2023).
  10. Han, W. et al. A neural circuit for gut-induced reward. Cell 175, 665–678 (2018).
  11. Kaelberer, M.M. et al. A gut-brain neural circuit for nutrient sensory transduction. Science 361, eaat5236 (2018).
  12. Matsumoto, M. et al. Impact of intestinal microbiota on intestinal luminal metabolome. Sci. Rep. 2, 233 (2012).
  13. Franceschi, C. et al. Inflamm-aging: an evolutionary perspective on immunosenescence. Ann. N.Y. Acad. Sci. 908, 244–254 (2000).
  14. Batista, C.R.A. et al. Lipopolysaccharide-induced neuroinflammation as a bridge to understand neurodegeneration. Int. J. Mol. Sci. 20, 2293 (2019).
  15. Drucker, D.J. GLP-1 receptor agonists and the brain: does neuroprotection extend to neurodegenerative diseases? Nat. Rev. Drug Discov. 23, 1–16 (2024).
  16. Swanson, K.V. et al. The NLRP3 inflammasome: molecular activation and regulation to therapeutics. Nat. Rev. Immunol. 19, 477–489 (2019).
  17. Bäckhed, F. et al. The gut microbiota as an environmental factor that regulates fat storage. Proc. Natl Acad. Sci. USA 101, 15718–15723 (2004).
  18. Clemente, J.C. et al. The impact of the gut microbiota on human health: an integrative view. Cell 148, 1258–1270 (2012).
  19. Nøhr, M.K. et al. GPR41/FFAR3 and GPR43/FFAR2 as cosensors for short-chain fatty acids in enteroendocrine cells vs FFAR3 in enteric neurons and FFAR2 in enteric leukocytes. Endocrinology 154, 3552–3564 (2013).
  20. Villumsen, M. et al. GLP-1 based therapies and disease course of neurodegenerative disorders: protocol for a network meta-analysis. BMJ Open 14, e078489 (2024).