NSI-189

Neurogenic antidepressant

NSI-189 (Neurogenic antidepressant) is a cognitive & mood research compound. Stimulates hippocampal neurogenesis and increases hippocampal volume; studied for depression and cognitive recovery.

How NSI-189 works

Stimulates hippocampal neurogenesis and increases hippocampal volume; studied for depression and cognitive recovery.

Proposed benefits

Researched for focus, memory, neuroprotection, mood and stress resilience.

Where to get NSI-189

The evidence for NSI-189

Graded by what exists behind each claim.

✅ Clinically validated

• Reached phase 2 in humans for major depressive disorder, run by Neuralstem. The trial missed its primary endpoint on the MADRS depression scale.
• What did move is the part usually left out: secondary cognitive measures improved significantly, and patient-reported outcomes favored the drug. A later analysis argued the trial was underpowered for the effect size seen. That is a different situation from a compound that did nothing.

📊 Correlative data

• Community use for mood and cognition. Reported experience is of a slow onset over weeks, consistent with a neurogenesis mechanism rather than a monoamine one.

🧪 Theoretical / extrapolated

• A nicotinamide derivative that stimulates neurogenesis in the hippocampus — it increased hippocampal volume in rodent work, which is notable because hippocampal atrophy is a well-replicated finding in depression.
• Acting on neurogenesis rather than on serotonin is why the timescale is weeks and why the cognitive signal in the trial may be the more meaningful result.

How to read these tiers: they say how much human evidence exists, not how well something works — and ✗ flags harm, never a disappointing trial. How the evidence tiers work →

What NSI-189 actually does

NSI-189 is a small molecule with a published structure and no published target. The compound is NSI-189 phosphate, (4-benzylpiperazin-1-yl)-[2-(3-methyl-butylamino)pyridin-3-yl]methanone Liu 2019 — a benzylpiperazine joined by a ketone to an aminopyridine. Three recognizable fragments, none of them a clue to a receptor.

That is not an oversight. It is how the compound was found. It came out of a phenotypic screen for proliferation of human hippocampus-derived neural stem cells Liu 2019 — a screen that asks ‘does this make more cells?’ rather than ‘does this bind that?’. A molecule discovered that way carries no receptor with it, and NSI-189 reached a 220-person phase 2 trial without one ever being named. In a catalog where almost every entry is defined by its receptor, this one is defined by an effect.

What is known downstream, and it is real signaling rather than hand-waving. In acute hippocampal slices, NSI-189 increased the magnitude of long-term potentiation evoked by theta-burst stimulation, in a time- and dose-dependent way; the same work reports TrkB and Akt pathway activation and stimulation of neurogenesis both in human hippocampal neural stem cells in vitro and in mouse hippocampus in vivo Liu 2019. TrkB is the receptor for BDNF and Akt is its principal survival arm, so the compound is acting on the canonical neurotrophin axis. Activating a pathway is not the same as binding its receptor, and nothing published shows NSI-189 binding TrkB.

The second candidate mechanism is metabolic, and it is a completely different kind of claim. In Zucker diabetic fatty rats, oral NSI-189 elevated protein subunits of respiratory complexes III and V and increased the activities of complexes I and IV in brain cortex, while ameliorating AMPK phosphorylation in sensory ganglia Jolivalt 2022. That is mitochondrial respiratory-chain enhancement, measured as protein and as enzyme activity in tissue.

So there are two mechanistic stories and they have not been reconciled. One says neurotrophin signaling and new neurons; the other says more efficient mitochondria in existing ones. Both are supported by published measurements, neither identifies a binding site, and a molecule that improves mitochondrial respiration would be expected to increase stem-cell proliferation as a consequence rather than as a separate action. Nobody has run the experiment that would order them.

Cell, rodent, human — and where it stops

Step one, in cells and in slices. Proliferation of human hippocampus-derived neural stem cells; time- and dose-dependent enhancement of theta-burst long-term potentiation in acute hippocampal slices; TrkB and Akt activation Liu 2019.

Step two, in mice, in two unrelated disease models. In an Angelman syndrome model, daily injections reversed cognitive and motor impairments, with wild-type animals showing only slight enhancement Liu 2019. In type 1 and type 2 diabetes models, NSI-189 prevented small- and large-fiber peripheral neuropathy, increased hippocampal neurogenesis, synaptic markers and volume, protected long-term memory, and halted the progression of established peripheral and central neuropathy Jolivalt 2019.

Step three, in rats, orally, on the longest schedule in the file. Zucker diabetic fatty rats left diabetic and untreated for 16 weeks, then given oral NSI-189 for a further 16 weeks, with nerve conduction velocity, tactile and thermal sensitivity and cognitive behavior assessed monthly. Peripheral neuropathy indices reversed, memory and plasticity improved, and the mitochondrial changes above were measured in cortex Jolivalt 2022. Note the design: treatment started after the disease was established, which is the harder and more relevant test.

Step four, humans, phase 1B, and this is where the enthusiasm came from. 24 patients with major depressive disorder in three cohorts — 40 mg once daily (6 on drug, 2 on placebo), 40 mg twice daily (6 and 2), and 40 mg three times daily (6 and 2) — treated for 28 days as inpatients with follow-up to day 84. The primary endpoint was safety and tolerability, and no serious adverse effects were reported. Secondary efficacy measures — the SDQ, MADRS, CGI-I and CPFQ — all improved significantly, with medium to large effect sizes. The measured human half-life was 17.4 to 20.5 hours Fava 2016.

Step five, humans, phase 2, and it missed. 220 subjects randomized to 40 mg daily, 80 mg daily or placebo for 12 weeks. The primary endpoint was reduction in MADRS score. It was not met at either dose: pooled mean difference −1.8, P = 0.22 at 40 mg, and P = 0.34 at 80 mg. Secondary measures at 40 mg did separate — SDQ P = 0.04, CPFQ P = 0.03, QIDS-SR Stage 2 P = 0.04 Papakostas 2020.

Read the two doses against each other, because that is the strangest thing in this compound’s record. 40 mg outperformed 80 mg on the primary endpoint and on every secondary that separated Papakostas 2020. Doubling the dose made the drug look worse. A monotonic dose-response is the most basic evidence that a drug effect is a drug effect; this one is inverted, and no published analysis explains it.

Step six, the post-hoc analysis, labeled as such. A re-analysis of the same 220 patients found that 80 mg produced a MADRS-6 benefit in the moderately depressed subgroup, defined as baseline MADRS under 30 (P = 0.046), and that 31% of CogScreen variables improved in the moderate subgroup against 14% in the severe one Johe 2020. A subgroup, chosen after the fact, at P = 0.046. It is a hypothesis for the next trial, and the next trial has not happened.

The obstacles, one at a time. (1) The registered phase 2 missed its primary endpoint at both doses Papakostas 2020. (2) The phase 1B’s large effect sizes came from 18 people on drug against 6 on placebo, in an inpatient setting, with efficacy as a secondary measure Fava 2016. (3) The dose-response is inverted and unexplained. (4) The strongest animal efficacy is in diabetic neuropathy Jolivalt 2019 Jolivalt 2022, an indication no human trial has tested. (5) No molecular target has been identified, so there is no way to check whether a given batch engages anything.

NSI-189 pharmacokinetics — how much of it actually gets in

The card says ‘~Hours’. The published human figure is much more specific than that: the plasma half-life measured in the phase 1B was 17.4 to 20.5 hours Fava 2016. That is a real human number for a research-market compound, and it is worth more than the vague word on the card.

What a 19-hour half-life implies about the two human regimens. Steady state arrives after four to five half-lives, so about 3 to 4 days. On a once-daily schedule the accumulation factor is about 1.7× a single dose’s exposure; on the eight-hourly schedule used in the top phase 1B cohort it is about 3.9×. So the 40 mg three-times-daily arm was not simply ‘three times the dose’ — it was three times the daily dose sitting on top of a much flatter, higher trough, and that cohort still reported no serious adverse effects across 28 inpatient days Fava 2016.

What degrades it, and the honest answer is that nobody has published it. The structure carries a 4-benzylpiperazine and a secondary aminopyridine Liu 2019. Both are canonical cytochrome P450 substrates: piperazines undergo N-dealkylation and aromatic rings undergo hydroxylation. No metabolism study, no metabolite identification and no CYP phenotyping appears in the published record for this molecule, which means the interaction profile of a twice-daily oral compound is entirely unmapped.

The oral question, which for this compound is not a barrier but an unknown. Every human dose and the 16-week rat course were given orally Fava 2016 Papakostas 2020 Jolivalt 2022, so the compound is clearly absorbed. What is not published is the absolute bioavailability, the fraction lost to first-pass hepatic extraction, or any food effect. A once-daily oral drug with an unpublished food effect is a drug whose users are adding an unmeasured variable every morning.

The brain, which is where it has to get to. Its effects are on hippocampal neurogenesis and cortical mitochondria Liu 2019 Jolivalt 2022, so it must cross the blood-brain barrier — and no brain-to-plasma partition ratio has been published for it in any species. The measured tissue effects are the only evidence that it gets there.

Route, stated plainly. Oral. There is no injectable, intranasal or sublingual human data for this molecule at any dose.

What would have to be true, and how you would know it was not

Four predictions. The first two both say this compound should do less than its reputation, and one of them is directly contradicted by the way people dose it.

1. Mood will not separate from placebo, and the registered trial already said so. The 220-person phase 2 missed MADRS at both doses Papakostas 2020. If you are going to run this on yourself, use a structured weekly self-rating from day −14 to day 84 so there is a pre-treatment slope to compare against, because depression scores fall on their own over that window and a 12-week uncontrolled improvement is exactly what placebo looks like.

2. 80 mg will not beat 40 mg, and this is the falsifiable prediction that matters most to a user. In the phase 2, the higher dose was worse on the primary endpoint and on every secondary that separated Papakostas 2020. The community instinct on a compound that underwhelms is to escalate; the only registered dose-comparison ever run says escalation is the wrong move. Run 4 weeks at 40 mg, 4 weeks at 80 mg, 4 weeks back at 40 mg, with a fixed daily rating, and this prediction is testable by one person.

3. Objective cognition should not move in a healthy person. The cognitive benefit in the post-hoc analysis was confined to the moderately depressed subgroup, at 31% of CogScreen variables against 14% in the severe group Johe 2020, and in wild-type mice the effect was described as slight where in the disease model it was a reversal Liu 2019. Take a MoCA at baseline and 12 weeks and a Trail Making test at the same two points; expect the practice effect and nothing else. A compound whose animal data shows large effects in disease and small ones in healthy animals is a compound whose healthy-user market is buying the wrong half of the evidence.

4. Watch the liver, because nobody has mapped the metabolism. Draw a CMP at baseline, 6 and 12 weeks. There is no published metabolite identification for this molecule and no CYP phenotyping, and it is taken orally every day. The prediction is that nothing moves — 220 people took it for 12 weeks in a monitored trial Papakostas 2020 — but this is the one blood test that is checking something the published record cannot.

What nobody has tested yet

Five experiments. The first has been outstanding since the compound was discovered.

1. Nobody has identified the target. NSI-189 came out of a phenotypic proliferation screen Liu 2019 and went to a 220-person phase 2 Papakostas 2020 without a named binding site. A standard broad receptor and kinase panel — a routine commercial service — would either find the target or exclude several hundred candidates. Its absence from the literature after fifteen years is the single strangest thing about this molecule.

2. Nobody has explained the inverted dose-response. 40 mg beat 80 mg on everything Papakostas 2020. The candidate explanations — an off-target effect appearing at the higher concentration, a U-shaped neurotrophin response, or simply noise in a trial that missed — make different predictions and are distinguishable with a three-dose study. None has been published.

3. Nobody has taken the diabetic neuropathy result to a human. Two rodent studies, two species, both types of diabetes, prevention and reversal of established peripheral neuropathy, with a mechanistic mitochondrial readout to go with it Jolivalt 2019 Jolivalt 2022. That is a stronger and more coherent preclinical package than the depression indication ever had, and the compound has never been trialed in it.

4. Nobody has published its metabolism. No metabolite identification, no CYP phenotyping, no food effect, no absolute bioavailability. For a daily oral small molecule taken alongside other daily oral small molecules, that is the most practically important missing table and the cheapest to produce.

5. Nobody has tested it in healthy people. Every human participant in both trials had major depressive disorder Fava 2016 Papakostas 2020, and the market for this compound is overwhelmingly people who do not. Whether a neurogenic compound does anything measurable in a healthy hippocampus is unstudied in humans and the mouse data hint that the answer is ‘much less’ Liu 2019.

NSI-189 — its own safety story, not its class's

The class block on this page is written for nootropics generally. Five things below are this compound’s own.

1. The human safety database is about 244 people and 12 weeks. 24 in the phase 1B, 28 days, inpatient, up to 120 mg a day, no serious adverse effects Fava 2016; 220 in the phase 2, 12 weeks, up to 80 mg a day Papakostas 2020. That is a genuinely reasonable early-phase record and it is also the whole of it: no long-term exposure, no elderly population, no pregnancy data, and nothing beyond three months.

2. No identified target means no predictable off-target profile. For most compounds in this catalog you can reason from the receptor to the likely side effects. Here you cannot, because the receptor has never been named Liu 2019. Everything known about what this molecule does to a person comes from having given it to 244 of them and watching.

3. The scaffold deserves a question nobody has answered. The molecule contains a 4-benzylpiperazine Liu 2019. Benzylpiperazine as a standalone chemotype has its own stimulant pharmacology at monoamine transporters. That does not make NSI-189 a stimulant — a fragment is not a molecule, and nothing in the trial record describes stimulant-like effects. It does mean the obvious screening question has an obvious answer available from a monoamine transporter panel, and no such panel has been published.

4. Promoting proliferation is the mechanism, so it is also the long-term question. The compound was selected for making neural stem cells divide Liu 2019. Nothing published follows animals or people for long enough to say what sustained stimulation of a proliferative compartment does over years, and the longest exposure anywhere in this file is 16 weeks in a rat Jolivalt 2022.

5. The development status is the honest framing. This compound missed the primary endpoint of its registered phase 2 Papakostas 2020 and no confirmatory trial followed. It is not an approved medicine anywhere and it is not an abandoned toxic one either — it is a molecule with a clean short-term safety record, a negative pivotal result, an unexplained inverted dose-response and an unidentified target, sold as a nootropic on the strength of an 18-person open secondary endpoint from 2016.

Sources read for this page

• Fava M, et al. A Phase 1B, randomized, double blind, placebo controlled, multiple-dose escalation study of NSI-189 phosphate, a neurogenic compound, in depressed patients. Molecular Psychiatry 2016 · PMID 26643541
• Papakostas GI, et al. A phase 2, double-blind, placebo-controlled study of NSI-189 phosphate, a neurogenic compound, among outpatients with major depressive disorder. Molecular Psychiatry 2020 · PMID 30626911
• Johe KK, et al. NSI-189 phosphate, a novel neurogenic compound, selectively benefits moderately depressed patients: A post-hoc analysis of a phase 2 study of major depressive disorder. Annals of Clinical Psychiatry 2020 · PMID 32722729
• Jolivalt CG, et al. Amelioration of Both Central and Peripheral Neuropathy in Mouse Models of Type 1 and Type 2 Diabetes by the Neurogenic Molecule NSI-189. Diabetes 2019 · PMID 31492662
• Jolivalt CG, et al. Enhancement of Mitochondrial Function by the Neurogenic Molecule NSI-189 Accompanies Reversal of Peripheral Neuropathy and Memory Impairment in a Rat Model of Type 2 Diabetes. Journal of Diabetes Research 2022 · PMID 35967127
• Liu Y, et al. Enhancement of synaptic plasticity and reversal of impairments in motor and cognitive functions in a mouse model of Angelman Syndrome by a small neurogenic molecule, NSI-189. Neuropharmacology 2019 · PMID 30408487

NSI-189 — safety, predicted from mechanism

NSI-189 — interference & stacking

Bloodwork to run alongside NSI-189

Baseline first, then again at 8–12 weeks.

The Brain Fog & Cognition panel covers these in one order — 12 markers, $233.06 with the discount applied.

Check results you already have → · All 103 markers A–Z

NSI-189 — frequently asked questions

NSI-189 inside a finished plan

One arm of 1 Protocol Blueprint, free to read in full.

What NSI-189 is used for

NSI-189 appears under 2 goals in the goal router.

Related Cognitive & Mood compounds

← Explore the full Protocol Vault

↑ Back to on this page