Remesal 2025 — Neuronal FTL1 as a driver of hippocampal cognitive aging

TL;DR

Remesal et al. identify neuronal ferritin light chain 1 (FTL1) as a pro-aging factor in the hippocampus whose age-associated rise drives cognitive decline. In young C57BL/6 mice, hippocampal AAV-mediated Ftl1 overexpression (synapsin-1 promoter) shifts iron redox state toward oxidized Fe³⁺, reduces dendritic complexity, decreases excitatory and inhibitory synaptic markers, impairs LTP, and produces deficits in novel-object recognition (NOR) and Y-maze cognition — recapitulating aging-like phenotypes in young animals. Conversely, in aged (18–24 mo) mice, two independent neuronal-FTL1-loss-of-function approaches — AAV-shRNA knockdown and synapsin-1-Cre-driven CRISPR conditional knockout — restored synaptic markers and rescued NOR + Y-maze performance. Bulk + single-nuclei RNA-seq converge on ATP synthesis / oxidative phosphorylation as the most consistently regulated process, and Seahorse confirms that FTL1 OE in primary neurons cuts mitochondrial ATP production rate (P<0.0001). Exogenous NADH (100 mg/kg IP daily × 9 days, with the final dose 2 h before testing) rescues both dendritic morphology and cognition in FTL1-OE young mice. Senior author Saul Villeda (UCSF); funded by NIA AG081038/AG067740/AG062357 and the Simons Foundation. Published as a Letter (short format) in Nature Aging.

This paper establishes:

A neuronal cell-autonomous driver of hippocampal cognitive aging that is targetable in old animals (rare in the aging-brain field — most rejuvenation work targets systemic or non-neuronal compartments).
A mechanistic link between brain iron dyshomeostasis and mitochondrial-dysfunction at the level of ATP synthesis — distinct from (though potentially upstream of) ferroptotic cell death.
A pharmacological rescue paradigm — NADH supplementation — that extends prior NAD-precursor cognitive-aging work (nmn, nr) using the reduced cofactor directly.

Design

Mouse lines

C57BL/6J young (2–3 mo) and aged (18–22 or 18–24 mo per experiment) — Jackson Lab 000664; aged mice from NIA colony
B6;129-Gt(ROSA)26Sor^tm1(CAG-cas9°,-EGFP)Fezh/J — inducible Cas9 transgenic (Jackson Lab 024857), used for neuron-specific conditional Ftl1 knockout

Viral manipulation paradigms

Direction	Vector	Promoter	Cargo	Target population
Overexpression	Lentivirus (in vitro) / lentivirus (in vivo)	synapsin-1	Syn1-Ftl1-IRES-eGFP vs Syn1-IRES-eGFP control	Hippocampal neurons of young mice (CA1 + dentate gyrus, bilateral stereotaxic)
Knockdown	Lentivirus, pGreenPuro shRNA backbone	U6	shRNA targeting Ftl1 (SH1–SH4) vs luciferase shRNA control	Hippocampal neurons of aged mice
Conditional KO	Lentivirus + Cas9 transgenic mouse	synapsin-1 (Cre) + U6 (gRNA)	Syn1-Cre + gRNA targeting Ftl1 exons 2–4 (vs scrambled gRNA)	Hippocampal neurons of aged Rosa26-loxP-STOP-loxP-Cas9-eGFP mice

All in vivo stereotaxic injections: 2 μl per site at 0.2 μl/min, bilateral CA1 + dentate gyrus, viral titer 1.0 × 10⁸ vp/ml.

Cognitive assays

Novel Object Recognition (NOR) — White 2020 protocol; discrimination index = (T_novel − T_familiar) / (T_novel + T_familiar)
Y-maze forced alternation — discrimination index for novel arm vs trained arm after 45-min delay
Radial Arm Water Maze (RAWM) — Alamed 2006 protocol; used in cognitive composite z-score
Open field — control for general locomotor / anxiety phenotypes (no differences detected in any cohort)

Other key assays

Neuronal nuclei RNA-seq (snRNA-seq) — FACS-sorted NeuN⁺ Hoechst⁺ nuclei; Smart-seq2 protocol; ~13,100–15,950 nuclei per sample; ~27,000 reads/nucleus; STAR + RSEM + DESeq2; PCA + Louvain + UMAP; Seurat
Synaptosome proteomics — sub-dissected hippocampus + cortex; sucrose-gradient synaptosome fractionation; TMT-6plex isobaric labeling; Orbitrap Fusion Lumos LC-MS/MS
Iron redox imaging — DNAzyme-based fluorescent sensors selective for Fe²⁺ (Fe(II)-H5) and Fe³⁺ (Fe(III)-B12), adapted from Wu 2023; confocal imaging of CA1/CA2/CA3/DG
Seahorse XFe24 — primary neuron mitochondrial ATP production rate (oligomycin + rotenone/antimycin protocol)
Field EPSP LTP — acute hippocampal slices from 16–18-week-old male mice ~1 month after stereotaxic OE injection; Schaffer collateral stimulation; 100 Hz × 2 trains; 50-min recording
Western blot, qPCR, confocal synaptic-puncta quantification (synapsin / PSD95 / gephyrin) — standard

Key findings

1. FTL1 rises in aged hippocampal neurons and tracks cognitive decline

Neuronal nuclei bulk RNA-seq (NeuN⁺ FACS-sorted nuclei from hippocampi): 28 upregulated and 81 downregulated genes between young (N=3, 3 mo) and aged (N=2, 18 mo) mice; GO enrichment on differentially expressed genes was dominated by synapse-related processes (carbohydrate metabolic process, regulation of synapse structure/activity, regulation of synapse organization, axonogenesis, synapse assembly) ¹
Synaptosome proteomics (sub-dissected hippocampus + cortex, sucrose-gradient synaptosomes, TMT-6plex MS/MS): 27 upregulated and 19 downregulated proteins with age (N=3 young, N=3 aged 18–20 mo, 20 mice pooled per replicate); GO enriched for immune response + complement activation ¹
Cross-referencing the transcriptomic and proteomic datasets yielded a single shared upregulated factor: FTL1 ¹
Validation: hippocampal Ftl1 mRNA increased by qPCR (N=6 young vs N=6 aged 18–22 mo, P=0.0219); hippocampal FTL1 protein increased by western blot (N=12 young vs N=9 aged 24 mo, P<0.0001) ¹
In an independent cohort, hippocampal FTL1 protein levels (western blot) correlated negatively with a behavioral z-score composite of Y-maze + NOR + RAWM: Pearson r = −0.434, P = 0.0268 (N=15 young + N=11 aged 20 mo) ¹

2. Mimicking the age-rise in young mice produces aging-like deficits

In vitro (primary hippocampal neurons; DIV8 lentiviral OE → DIV18 analysis):

Total dendrite length: control vs Ftl1-OE P<0.0001 (N=5 vs 5 replicates, 5–9 neurons each)
Total Sholl intersections: P<0.0001
Intersections-vs-distance from soma: P<0.01 (control > Ftl1-OE) ¹

In vivo (young C57BL/6, bilateral CA1+DG Syn1-Ftl1-IRES-eGFP):

Hippocampal FTL1 protein ↑ by ~30% vs Syn1-eGFP control (N=6 vs 6, P=0.0434) ¹
Iron redox shift (DNAzyme sensors): Fe²⁺ unchanged (P=0.2567); Fe³⁺ ↑ (N=4 per group, P=0.0382); Fe³⁺/Fe²⁺ ratio ↑ (P=0.0279) ¹
Excitatory synaptic markers ↓ (N=4 control vs 7 Ftl1-OE): synapsin P=0.0265; PSD95 P=0.0006; co-localized excitatory synapses P=0.0089 ¹
Inhibitory synaptic markers ↓: synapsin P=0.0129; gephyrin P=0.0053; inhibitory synapse co-localization P=0.2001 (NS) ¹
NR2A, AMPA receptor, synapsin ↓ on western (Extended Data Fig 3c)
LTP impaired: fEPSP slope (% baseline, average of final 5 min) reduced (N=16 control vs 16 Ftl1-OE slices, P=0.029) ¹
NOR: control mice showed novel-object preference (N=23, P=0.0023); Ftl1-OE young mice showed no preference (N=20, P=0.4315) ¹
Y-maze: same pattern — control N=18 P=0.0336; Ftl1-OE young N=22 P=0.4315 ¹
No locomotor/anxiety differences in open field (Extended Data Fig 3d–f) ¹

3. Targeting FTL1 in aged hippocampus rescues cognition

Aged shRNA knockdown (18 mo, bilateral CA1+DG sh-Ftl1-eGFP vs sh-Luc-eGFP):

FTL1 protein ↓ (N=6 vs 6, P=0.0323) ¹
Excitatory: synapsin P=0.0132, PSD95 P=0.1324, excitatory synapses P=0.0777 (trend)
Inhibitory: synapsin P=0.0257, gephyrin P=0.0031, inhibitory synapse co-localization P=0.2964
NOR rescued: control aged N=16 (P=0.4977 — fails); Ftl1-KD aged N=17 (P=0.0010 — succeeds)
Y-maze rescued: control aged N=17 (P=0.1895 — fails); Ftl1-KD aged N=18 (P=0.0072 — succeeds)
Open field unchanged ¹

Aged neuron-specific CRISPR cKO (22 mo, Rosa26-LSL-Cas9-eGFP transgenics + Syn1-Cre + gRNA-Ftl1):

FTL1 protein ↓ (N=5 vs 6, P=0.0076)
Excitatory: synapsin P=0.0028, PSD95 P=0.0140, excitatory synapses P=0.0773
Inhibitory: synapsin P=0.1091, gephyrin P=0.1091, inhibitory co-localization P=0.0472
NOR rescued (control N=21 P=0.5796 fails; Ftl1-cKO N=14 P=0.0087 succeeds — but note discrepancy: in Fig 3l, control N=15 P=0.4977 fails and Ftl1-cKO P=0.0010 succeeds; the cKO and KD cohorts converge on the same effect direction)
Y-maze: control N=21 P=0.5796; Ftl1-cKO N=14 (P=0.0883 — strong trend, not significant); see Fig 3m ¹

Two independent loss-of-function approaches (shRNA + CRISPR cKO) converging in aged animals materially derisks an off-target / sequence-specific artifact.

4. Mechanism: FTL1 disrupts mitochondrial ATP synthesis

Bulk neuronal nuclei RNA-seq:

Young Ftl1-OE: 100 differentially expressed genes; GO enriched for translation, oxidative phosphorylation, ATP metabolic process, aerobic respiration, proton transmembrane transport ¹
Aged Ftl1-KD: 309 differentially expressed genes; GO enriched for regulation of trans-synaptic signaling, transport, synapse organization, synapse plasticity, neurogenesis ¹
32 genes were bidirectionally changed (down in Ftl1-OE young, up in Ftl1-KD aged): GO enriched for oxidative phosphorylation, proton-motive-force-driven mitochondrial ATP synthesis, translation, aerobic respiration, energy derivation by oxidation of organic compounds ¹
Specific ETC genes upregulated in aged Ftl1-cKO: Sdhb (Complex II succinate dehydrogenase), Atp5o, Atp5c1, Atp5j (Complex V ATP synthase), Ndufa10 (Complex I) ¹

Single-nuclei RNA-seq of aged Ftl1-cKO vs control (N=2 pooled mice per group, ~10,000 nuclei/sample; 10x Genomics 3’ kit):

Cell-type-resolved GO across CA1 pyramidal, CA2/CA3 pyramidal, DG granule cell clusters all converged on metabolic processes: proton-motive-force-driven mitochondrial ATP synthesis, oxidative phosphorylation, mitochondrial electron transport, NADH dehydrogenase complex assembly ¹

Seahorse XFe24 mitochondrial ATP production rate (primary hippocampal neurons, DIV8):

Ftl1-OE vs control: P<0.0001 (decreased; N=12 control vs N=18 OE replicates)
Ftl1-KD vs control: P=0.0784 (trend toward increased; N=17 control vs N=18 KD replicates) ¹

5. NADH supplementation rescues FTL1-driven cognitive deficits

Rationale: NADH is a coenzyme that drives ATP synthesis at Complex I (NADH:ubiquinone oxidoreductase). Boosting NADH should counter FTL1’s effects on ATP synthesis if that mechanistic pathway is causal.

In vitro rescue (primary neurons, 200 μM NADH from DIV14–DIV18):

Total dendrite length: control+saline vs OE+saline P<0.0001 (deficit confirmed); OE+saline vs OE+NADH P=0.0002 (deficit mitigated)
Total Sholl intersections: control vs OE+saline P=0.0075; OE+saline vs OE+NADH P=0.0039
Intersections vs distance: control vs OE+saline P<0.0001; OE+saline vs OE+NADH P<0.0001
N=5 control+saline, 5 OE+saline, 6 control+NADH, 6 OE+NADH replicates ¹

In vivo rescue (young adult mice, bilateral Syn1-Ftl1-IRES-eGFP injection ~42–51 days prior; NADH 100 mg/kg/day IP × 9 days, final dose 2 h before testing):

NOR: control+saline P=0.0046 (preference); Ftl1-OE+saline P=0.6516 (deficit confirmed); control+NADH P=0.0016; Ftl1-OE+NADH P=0.0017 (rescued)
Y-maze: control+saline P=0.0250; Ftl1-OE+saline P=0.0551 (trend deficit); control+NADH P=0.0238; Ftl1-OE+NADH P=0.0020 (rescued)
N=12 control+saline, 12 Ftl1-OE+saline, 14 control+NADH, 13 Ftl1-OE+NADH per cohort ¹
Open field unchanged across all four groups ¹

NADH rescue suggests the FTL1 cognitive phenotype is downstream of impaired ATP synthesis — not solely a direct iron-toxicity / ferroptotic effect — and lies upstream of dendritic-arbor maintenance and synaptic plasticity.

Wiki impact

Hallmark connections

mitochondrial-dysfunction — primary mechanism. FTL1 acts as an iron-routing perturbant that disrupts proton-motive-force-driven ATP synthesis at the inner membrane. Atp5o/c1/j, Sdhb, and Ndufa10 are bidirectionally regulated.
loss-of-proteostasis — secondary. Aged-rise in neuronal ferritin protein is a proteostatic shift (whether driven by iron-loading or independent protein-handling change is not resolved in this paper).
Iron redox shift (↑Fe³⁺/Fe²⁺) overlaps with ferroptosis machinery ²³, but Remesal explicitly notes that the direction of the Fe³⁺/Fe²⁺ shift in FTL1-OE young mice (increase) is opposite to what was previously reported in Alzheimer’s mouse brain (decrease) ⁴. Author interpretation: ferroptotic cell death and the FTL1-driven sub-lethal dysfunction described here are mechanistically distinct.

Intervention modalities

AAV-shRNA / synapsin-1-Cre CRISPR cKO of neuronal Ftl1 — neuronal-FTL1-knockdown joins a small set of validated late-life cognitive-rescue gene-therapy targets in mouse. Companion to Wheatley 2019 (O-GlcNAcylation) and Pratt 2022 (TET2) from the same Villeda lab lineage.
NADH supplementation — extends the NAD-cognitive-aging story (cf. nmn, nr). Distinguishing feature: Remesal used NADH directly (the reduced cofactor), not a precursor — bypassing the de novo / salvage biosynthetic steps that NMN/NR rely on. This may have translational implications if NADH bioavailability differs from precursors. The IP-100-mg/kg × 9 days, 2-h pretrial paradigm is short-duration and tested only against FTL1-OE-induced deficit, not aging itself.

Translational implications

Authors note three convergent lines for human relevance:

Neuroferritinopathy — autosomal-dominant disorder caused by mutations in FTL (the human ortholog of Ftl1); presents with movement disorders, dystonia, and cognitive impairment ⁵⁶⁷ — demonstrating that pathological ferritin variation is cognitively impactful in humans.
CSF ferritin in AD — Ayton et al. 2015 reported that CSF ferritin levels in AD cohorts predicted cognitive performance over 7 years and predicted conversion from MCI to AD ⁸. This is the most direct human correlative evidence that ferritin biology tracks cognitive trajectory.
General iron-dyshomeostasis aging biology — iron accumulation in the aging brain is a long-recognized phenomenon; this paper supplies a discrete mechanistic node (neuronal FTL1 → iron redox → ATP synthesis → cognition) that is potentially druggable.

Gaps and limitations

Single laboratory replication — all in vivo behavioral and synaptic data are from the Villeda lab. Independent replication of the cognitive rescue is not yet published. needs-replication
Single species — mouse only. Direct demonstration that the same mechanism operates in human hippocampal aging awaits postmortem proteomic confirmation + human-genetic association (CSF ferritin / GWAS hits exist but are correlative). needs-human-replication
Mechanism of FTL1 upregulation with age is not identified — what drives the age-rise of neuronal Ftl1 in the first place? Transcriptional regulator(s), iron-IRP-IRE post-transcriptional regulation, microglia/astrocyte iron-handoff biology — all are plausible upstream causes but none are resolved here. no-mechanism (upstream)
Sex — paper does not break out by sex; electrophysiology used 16–18-week-old males only; behavioral cohorts are not sex-segregated in the figure legends. needs-sex-stratification
NADH dose-response and duration are minimal — single dose (100 mg/kg IP), short window (9 days). Effect on baseline aged cognition (without FTL1 OE) is not tested. dose-response-unclear; long-term-unknown
Bidirectional gene-set is small (32 genes) — high statistical power to call the OXPHOS signature, but mechanistic specificity to ATP synthesis (vs collateral effects on Ca²⁺ handling, redox, peroxisomal metabolism etc.) is limited.
Cognitive composite z-score correlation (r=−0.434, P=0.0268) is moderate. The relationship is consistent but not deterministic; many FTL1-low aged mice still show impairment, and vice versa.

Footnotes

Cross-references

Subject protein: ftl1 (drafted alongside this study page; verified-by-claude 2026-05-20)
Senior author lab work (Villeda) — cognitive-rejuvenation-via-blood-factors paradigm: young-blood context (planned); gdf11 (verified); related papers: Villeda 2011 Nature, Villeda 2014 Nat Med, Castellano 2017 Nature (umbilical cord plasma), Horowitz 2020 Science (exercise factors), Schroer 2023 Nature (platelet factors), Pratt 2022 Cell Rep (TET2)
Cognitive-aging gene-therapy precedents: tet2 (verified) — Pratt 2022 same-lab study
Hallmark coverage: mitochondrial-dysfunction (drafted) — to be propagated with the FTL1 iron-ATP-synthesis axis; loss-of-proteostasis (drafted)
Tissue coverage: brain (verified) — to be propagated with a brain-iron section
Intervention coverage: nmn (drafted), nr (drafted) — NAD-precursor cognitive-aging story; this paper uses NADH directly rather than a precursor
Model organism: mus-musculus (drafted) — C57BL/6J young + aged

remesal-2025-ftl1-brain-cognitive-aging · this study · n’s and p-values per the individual results sections above · randomized + blinded (per Methods statement: “All experiments were randomized and blinded by an independent researcher”) · model: C57BL/6J mice young (2–3 mo) + aged (18–24 mo); primary hippocampal neuron cultures from E17 C57Bl/6J embryos · published as Letter, Nature Aging 5(10):1957–1969 · DOI 10.1038/s43587-025-00940-z · PMID 40830655 · PMC12532579 · OA via CC-BY 4.0 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³ ↩¹⁴ ↩¹⁵ ↩¹⁶ ↩¹⁷ ↩¹⁸ ↩¹⁹ ↩²⁰ ↩²¹ ↩²² ↩²³ ↩²⁴ ↩²⁵ ↩²⁶
doi:10.1016/j.cell.2012.03.042 · Dixon SJ et al. · Cell 2012;149(5):1060-1072 · ferroptosis-defining paper · in-vitro · model: human cancer cell lines · not independently re-read for this study page ↩
doi:10.1016/j.cell.2017.09.021 · Stockwell BR et al. · Cell 2017;171(2):273-285 · ferroptosis review · not independently re-read for this study page ↩
doi:10.1126/sciadv.eade7622 · Wu Y et al. · Sci Adv 2023 · DNAzyme Fe²⁺/Fe³⁺ sensor methodology paper · model: Alzheimer’s disease mouse brain · Fe³⁺ enriched over Fe²⁺ in AD brain (decreased Fe³⁺/Fe²⁺ ratio reported in original — opposite sign vs Remesal 2025 FTL1-OE in non-AD young brain; author interpretation: distinguishes ferroptotic from sub-lethal FTL1 dysfunction) · not independently re-read ↩
doi:10.1016/j.beha.2004.08.022 · Levi S et al. · Best Pract Res Clin Haematol 2005;18(2):265-276 · neuroferritinopathy review · not_oa likely · not independently re-read ↩
doi:10.7916/D8KS6PR1 · Kumar N et al. · Tremor Other Hyperkinet Mov 2016;6:355 · neuroferritinopathy clinical review · not independently re-read ↩
doi:10.1002/mds.29162 · Marchand A et al. · Mov Disord 2022;37(9):1948-1952 · neuroferritinopathy iron-chelation case report · not independently re-read ↩
doi:10.1038/ncomms7760 · Ayton S et al. · Nat Commun 2015;6:6760 · observational · CSF ferritin predicted AD outcomes over 7 yr; regulated by APOE · model: human AD cohort · not independently re-read ↩

🧬 Aging Wiki

Explorer

Targeting iron-associated protein Ftl1 in the brain of old mice improves age-related cognitive impairment