Can prepubertal NMN improve cognitive deficits in a schizophrenia mouse model?

A prepubertal window tested in a two-hit model

The paper asks whether boosting NAD+ precursor supply before puberty can soften cognitive deficits in a mouse model built to mimic schizophrenia risk. The rationale is timing: cognitive problems in schizophrenia are described as a core driver of long-term disability, existing treatments mainly control positive symptoms, and the prepubertal period is framed as a window when metabolic and plasticity abnormalities may still be reversible.

The model combines two insults. Male offspring received maternal immune activation on gestational day 9, which establishes neurodevelopmental susceptibility, and then chronic unpredictable stress during adolescence (postnatal day 30 to 40), which triggers the phenotype. NMN, a direct precursor of NAD+, was given at 2 g/L in drinking water from postnatal day 21 to 30, so the intervention sits before the stress exposure rather than after the full phenotype appears. Only male offspring were used, and the authors state that sex differences remain to be evaluated.

Learning and recognition improve toward control levels

Two behavioral tests carried different weight. In the Morris water maze, the two-hit group had longer escape latencies than controls on all training days and did not shorten them across training; the NMN group had significantly shorter latencies (all P < 0.001) that approached control levels. After platform removal, the two-hit group spent less time in the original platform quadrant and crossed the former platform location less often than controls, and NMN increased both measures (all P < 0.001). Swimming speed did not differ among groups (P = 0.60), which argues against a motor explanation for the maze result.

The novel object recognition test is weaker evidence, and the paper says so. On the test day, the two-hit group showed no preference for the novel object (P > 0.99), while the NMN group had a significantly greater discrimination index (P < 0.001). But the two-hit group also explored less overall, and the authors note that motivation, anxiety, arousal, or novelty-seeking could produce that pattern without a recognition-memory defect. They therefore treat the NOR result as supporting evidence for altered cognitive or exploratory performance, not definitive proof of recognition-memory rescue, and rest the overall cognitive conclusion on both tests together.

More proliferating cells and mature neurons in the dentate gyrus

Hippocampal neurogenesis here means neural stem cells in the subgranular zone of the dentate gyrus proliferating, differentiating, and integrating into circuits. The study measured markers for stages of that lineage rather than the process itself. Relative to controls, two-hit mice had fewer Ki67+/DCX+ proliferative cells, fewer EdU+/SOX2+/GFAP+ radial glia-like cells, fewer total EdU+ cells, fewer DCX+ immature neurons, and fewer NeuN+ mature neurons (all P < 0.001), while S100β+/GFAP+ reactive astrocytes were increased. NMN was associated with significantly greater numbers of the neurogenesis-related populations (all P < 0.001) and fewer S100β+/GFAP+ cells (P = 0.005).

Western blots matched: SOX2, Tuj-1, and DCX were lower in two-hit mice and higher after NMN, while GFAP was higher in two-hit mice and lower after NMN. One index did not move: the NeuN+/DCX+ ratio, a maturation index, did not differ significantly between control and two-hit mice (P = 0.27) or between two-hit and NMN groups (P = 0.12). The authors also caution that the GFAP/S100β pattern could reflect a stem-cell lineage bias toward glia, reactive astrogliosis, or both, and the population-level data cannot separate these.

Marker densities track behavior only associatively

Densities of EdU+/SOX2+/GFAP+, Ki67+/DCX+, and NeuN+ cells were positively correlated with water-maze target-quadrant dwell time, platform crossing frequency, and the NOR discrimination index (r = 0.68–0.81; all P < 0.001). This is the paper's strongest bridge between histology and behavior, and it is explicitly correlational. The authors state that the data support an associative, not necessarily causal, link between neurogenesis-related changes and behavioral improvement.

A labeling detail limits interpretation further. EdU labeling at postnatal day 48 to 50 reflects cumulative proliferative output after both the NMN exposure and the stress period, so the counts cannot be read as a snapshot of proliferation during the treatment window alone. Whether marker changes are causally required for cognitive improvement is listed among the questions left for future studies.

Mitochondrial and NAD+ indices move in the same direction

Transmission electron microscopy showed disorganized cristae and vacuolization in hippocampal neuronal mitochondria of two-hit mice, which the paper describes as ameliorated after NMN. Two-hit mice had lower ATP and NAD+ levels and higher ROS than controls (all P < 0.001), and NMN was associated with recovery of all three (all P < 0.001 versus two-hit). Mitochondrial biogenesis indicators followed: mtDNA copy number and citrate synthase activity were lower in two-hit mice (both P < 0.01) and restored after NMN (both P < 0.01), and coenzyme Q10 was lower (P < 0.001) and restored after NMN.

Controls given NMN alone also showed higher hippocampal ATP and NAD+ than untreated controls (both P < 0.01), so NMN shifted these indices beyond the normal baseline as well as in the model. These are biochemical and structural indices of mitochondrial content and homeostasis, not direct measures of cognitive function, and the paper presents them as concurrent with, not upstream proof of, the behavioral changes.

SIRT1–p53–PI3K/AKT/mTOR as a candidate route

Network pharmacology screening pointed to SIRT1–p53–PI3K/AKT/mTOR, and the paper tested it in vivo and in neural stem cells under neuroinflammatory conditions. Two-hit mice and the in vitro inflammatory model had lower SIRT1 and higher p53 acetylation and pan-acetyllysine than controls (all P < 0.001). NMN was associated with elevated SIRT1, p-PI3K, p-AKT, p-mTOR, and PGC-1α and reduced p53 acetylation (all P < 0.001 in most comparisons; SIRT1 in vivo P = 0.002, p-PI3K in vivo P = 0.004). The p-mTOR/mTOR ratio did not differ between control mice given NMN and untreated controls, which the authors read as no excessive mTOR activation under physiological conditions.

The docking result argues against direct binding: the predicted binding free energy of NMN to SIRT1 was −3.4 kcal/mol, above the favorable threshold of −5.0 kcal/mol, which the authors interpret as NMN acting through conversion to NAD+ rather than by binding SIRT1 itself. The functional test used EX527, a SIRT1 inhibitor, in the in vitro model. Adding EX527 to NMN reduced neurosphere diameters and EdU+ proliferative cell proportions back to inflammatory-model levels (P = 0.09 and P = 0.51 versus that model), raised apoptosis above the NMN group (P < 0.001) though still below the inflammatory model (P = 0.02), and abolished the NMN-induced drop in p53 acetylation (P = 0.65 versus the inflammatory model). This supports SIRT1 involvement in vitro; it does not establish the pathway as the in vivo mechanism of cognitive rescue.

Where the two-hit design limits the claim

The largest interpretive gap is the model itself. There were no maternal-immune-activation-only or stress-only control groups, so the study cannot say whether NMN acts on the immune-activation component, the adolescent stress component, or their interaction. Because NMN was given before the stress exposure, it may instead have changed vulnerability to later stress. The findings establish efficacy in the combined two-hit model without dissecting either hit.

Other boundaries sit with their results: only male offspring were studied; the NOR data are qualified by unexplored motivation and anxiety confounds; the neurogenesis evidence is marker density plus correlation, with EdU labeling spanning the post-treatment period; and the pathway evidence is strongest in vitro, where the inhibitor experiment was done. The paper's own conclusion is correspondingly framed as association, with causality and hit specificity left open.

What matters most is the timing, not the molecule alone

The paper's distinctive contribution is not that NMN changed a marker in a schizophrenia model, but that a short prepubertal exposure, ten days from postnatal day 21 to 30 and before the adolescent stress hit, coincided with behavioral performance approaching control levels alongside neurogenesis-related and mitochondrial indices. That combination of timing and convergence across behavioral, histological, biochemical, and in vitro readouts is what makes the result worth following.

The priority for readers is therefore the window and the convergence, held together with the design's blind spot: because the two hits were never separated and NMN preceded the second hit, the most defensible reading is that early metabolic intervention in a combined-risk model is associated with better outcomes, with the causal chain from SIRT1 signaling to new neurons to cognition still unproven. A study that isolates the hits, includes females, and tests whether blocking neurogenesis removes the behavioral benefit would be the decisive next comparison.

Sources used in this answer