NSI-189 (Amdiglurax)

Identity, and a name that changed mid-programme

The compound almost everyone calls NSI-189 is registered under a name hardly anyone uses. The FDA Global Substance Registration System files it as amdiglurax, INN 13622, UNII YVE9U408ZL. PubChem keeps a single record under the development code instead: CID 50922681, CAS 1270138-40-3, C22H30N4O, molecular weight 366.5 g/mol, InChIKey DYTOQURYRYYNOR-UHFFFAOYSA-N. The CAS and formula are identical in both. The structure is a benzylpiperazine amide of a 2-aminopyridine carboxamide, a small synthetic heterocycle rather than a peptide. Almost all published work used the phosphate salt: NSI-189 phosphate, CAS 1270138-41-4, PubChem CID 50922680, C22H33N4O5P, 464.5 g/mol. Animal doses in the literature are stated as free base.

The development code changed in 2021, and the trial record split along with it. Neuralstem, which discovered the molecule, became Seneca Biopharma and then, through a reverse merger, Palisade Bio. An Asset Transfer Agreement dated 18 October 2021, filed with the SEC as an exhibit to Alto Neuroscience's registration statement, transferred the defined "Seneca NSI-189 Know-How" and the associated patents to Alto. Alto's own annual report describes that same transaction as the acquisition of "all patent, know-how and other rights to ALTO-100". Reading the two filings together is what establishes that ALTO-100 and NSI-189 are one compound. Neither document says so in a single sentence, and the ClinicalTrials.gov records for the ALTO-100 studies never mention NSI-189.

Nothing bearing either name is approved anywhere. Queries to Drugs@FDA on amdiglurax and on NSI-189 as an active ingredient returned no matches, and the molecule appears in sponsor filings as an investigational product candidate. An International Nonproprietary Name records a naming decision, not a regulatory one. A PubMed search for amdiglurax returns zero records, so the published literature is indexed entirely under the development code.

The preclinical package that was never published

The Phase 1b paper is unusually candid about where the founding data sit. Fava and colleagues describe the molecule as stimulating neurogenesis of human hippocampus-derived neural stem cells in vitro and in mouse hippocampus in vivo, as showing behavioural effects across three doses in a novelty-suppressed feeding model after 28 days of oral administration, as increasing dentate gyrus neurogenesis and hippocampal volume in treated mice, and as acting in the hippocampus and subventricular zone and nowhere else in the central nervous system. Every one of those statements carries the same parenthetical: Data on file, Neuralstem, non-peer-reviewed. The in vitro metabolism characterisation and the Phase 1a single-dose safety result carry it too.

PubMed returns fourteen records for the search term. Three are incidental or false matches. None of the remaining eleven is the discovery paper. There is no published in vitro screen, no published dose-response for the neurogenesis assay, no published receptor or transporter selectivity panel, and no published toxicology. Alto's 2025 annual report states that the company believes the compound binds a receptor not targeted by other CNS therapeutics, and does not name it. Fifteen years after the first human dose, the molecular target is not in the public record.

The founding characterisation — the potency in the stem-cell assay, the anatomical selectivity, and the magnitude of the hippocampal volume effect in healthy mice — has never been published, and so has never been independently reproduced. Secondary pages describing those results as established are describing unpublished company material. Nor is the downstream animal record sponsor-independent. Every rodent paper in this ledger except Bauman 2018 carries Karl Johe, the compound's originating scientist, as a co-author; Tajiri 2017 and Jolivalt 2022 additionally list Neuralstem as an author affiliation, state that the sponsor supplied the test article, and disclose former-employee conflicts. The only rodent study with no sponsor authorship or affiliation is Bauman 2018, and it is the one reporting the narrowest effect.

Twenty-four inpatients: the Phase 1b

NCT01520649 randomised adults aged 18 to 60 with recurrent major depressive disorder to 40 mg once, twice or three times daily, or to placebo, for 28 days in an inpatient setting, with outpatient follow-up to day 84. Twenty-four patients were dosed across three cohorts of eight, six on drug and two on placebo in each. Mean terminal half-life was 17.4 plus or minus 3.06 hours, 20.5 plus or minus 3.51 hours and 18.6 plus or minus 4.14 hours across the three regimens, with steady state reached at 96 to 120 hours. Day-28 AUC over 24 hours was 1144 plus or minus 276, 2791 plus or minus 1443 and 4384 plus or minus 1217 hours times ng/mL.

Efficacy was exploratory, and the three active cohorts were pooled to give eighteen on drug against six on placebo. Effect sizes ranged from 0.57 on the clinician-rated CGI-I to 0.95 on MADRS, with statistical significance reached only on the two self-rated instruments, the SDQ and the CPFQ. Reported adverse events were dominated by headache, at 50% in every group including placebo; dizziness reached four of six on the highest regimen. No serious adverse events occurred. A 2016 erratum corrected a duplicated image across two panels of one figure and a typo in a table; PubMed types the record as a published erratum and attaches no other flag.

Structural MRI was the closest thing to a mechanistic readout in that trial, and it did not deliver one. Analysis of covariance found no significant difference in hippocampal volume between pooled active and placebo patients at day 28 or day 84. A post-hoc repeated-measures analysis suggested a modest increase on the left side at p = 0.12. In that same post-hoc analysis the amygdala, included as a control site, reached significance on the right at p = 0.049, where the target region did not, and the authors described these as trends of similar magnitude and wrote that this underscored the possibility of a spurious finding. This remains the only MRI dataset in the published human record.

Phase 2, and the analyses that followed it

NCT02695472 randomised 220 outpatients with major depressive disorder to 40 mg daily, 80 mg daily or placebo for twelve weeks, using a sequential parallel comparison design of two six-week stages. The primary outcome, MADRS reduction against placebo, missed at both doses: pooled mean difference minus 1.8 at 40 mg with p = 0.22, and minus 1.4 at 80 mg with p = 0.34. The 40 mg arm showed advantages on three self-rated measures, SDQ at minus 8.2 and p = 0.04, CPFQ at minus 1.9 and p = 0.03, and QIDS-SR in stage 2, and on some CogScreen subtests. On the Cogstate battery every combined p value was at or above 0.05.

Tolerability ran in an unusual direction. Discontinuation for intolerance during the first six weeks was significantly higher on placebo, at 25 patients or 18.9%, than on 40 mg at 9.1% or 80 mg at 2.3%, with p = 0.013. No participant randomised to the compound experienced a serious adverse event. Neither the Phase 1b nor the Phase 2 has results posted on ClinicalTrials.gov; the Phase 2 registry entry still carries a status of UNKNOWN against an estimated completion date of December 2017.

A separate post-hoc analysis of that same 220-patient dataset, published in 2020 with the compound's originating scientist as first author, split the sample at a baseline MADRS of 30. In the moderately depressed subgroup the 80 mg dose separated from placebo on the six-item MADRS at p = 0.046, and 11 of 36 CogScreen variables improved, against 5 of 36 in the severe subgroup and 7 of 36 across the whole sample. Those analyses were not the registered primary outcome. They are the origin of most circulating claims that the compound worked in moderate depression.

The second Phase 2b, under the new code

Alto Neuroscience has registered four Phase 2 studies of ALTO-100. NCT05117632 was an open-label biomarker study in 245 adults with major depressive disorder or post-traumatic stress disorder. NCT05419869 was a 21-participant decentralised pilot. NCT05712187 was the randomised, quadruple-masked Phase 2b: 301 adults with major depressive disorder across 34 United States sites, enrolled against a pre-randomisation memory-based cognitive biomarker, with a six-week MADRS primary endpoint. It completed in October 2024. The fourth, NCT06656416 in bipolar depression, is described below. None of the four has results posted on ClinicalTrials.gov, and a PubMed search for ALTO-100 returns no publication of any of them.

Figures for that Phase 2b come from the sponsor's annual report rather than from a journal. In the biomarker-positive modified intent-to-treat population, 97 on drug against 99 on placebo, week-6 least-squares mean MADRS change was minus 10.3 against minus 9.8, Cohen's d 0.05, p greater than 0.1. The monotherapy subgroup ran the other way at d = minus 0.13. The adjunctive subgroup, 30 against 31 and not powered for significance, gave d = 0.47 at p = 0.09. The sponsor attributed the split to non-compliance, reporting that blood sampling found detectable drug in 56% of the monotherapy group against 100% of the adjunctive group.

Bipolar depression is where the programme went next. NCT06656416 is recruiting toward roughly 200 adults with bipolar I or II disorder in a current major depressive episode, taking ALTO-100 or placebo adjunctively to a stable mood stabiliser or atypical antipsychotic, with topline data expected in the second half of 2026. As of August 2026 it has not reported. Two randomised trials of this molecule in depression have now missed their primary endpoint.

The animal record, and one head-to-head that went the other way

Rodent work has been more consistent than the human work, but it is not sponsor-independent. Every rodent paper in this record except Bauman 2018 carries Karl Johe, the compound's originating scientist, as a co-author; Tajiri 2017 and Jolivalt 2022 additionally list Neuralstem as an author affiliation, state that the sponsor supplied the test article, and disclose former-employee conflicts. Tajiri and colleagues gave 48 male Sprague-Dawley rats 30 mg/kg by oral gavage starting six hours after middle cerebral artery occlusion and continuing daily for twelve weeks, 24 animals per arm drawn from 60 operated, and reported motor and neurological improvement maintained to 24 weeks after stroke.

Allen and colleagues irradiated Long-Evans rats with 27 Gy in a fractionated cranial protocol, then gave 30 mg/kg by oral gavage for four weeks, with 15 to 16 rats per group; the paper does not state the sex of the animals. Performance improved significantly on all four spontaneous exploration tasks — novel place recognition, novel object recognition, object in place and temporal order — while on the separate contextual fear-conditioning task only a non-significant trend was reported. Neurogenesis increased and fewer activated microglia were counted, the difference persisted a month after dosing stopped, and hippocampal volume showed no major change.

Diabetes models produced the largest coherent body of data. Jolivalt and colleagues dosed female Swiss Webster mice modelling type 1 diabetes at 10 mg/kg and male db/db mice at 30 mg/kg by daily oral gavage, 8 to 10 per group, and reported effects on nerve conduction, tactile and thermal sensitivity, hippocampal neurogenesis, synaptic markers and long-term memory. A 2022 follow-up in male Zucker diabetic fatty rats used 30 mg/kg orally for sixteen weeks after sixteen weeks of untreated diabetes and reported raised activity of respiratory complexes I and IV in brain cortex. Liu and colleagues reported enhanced long-term potentiation and reversal of motor and cognitive deficits in an Angelman syndrome mouse model; that paper is paywalled, and the corresponding patent application specifies 30 mg/kg intraperitoneally in four groups of four to five mice.

One direct comparison deserves separating out. Bauman and colleagues, whose primary subject was the unrelated compound P7C3-A20 in rhesus macaques, ran a side-by-side rodent assay in male C57BL/6J mice at 10 mg/kg per day intraperitoneally for both molecules. In that mouse assay, dentate gyrus neural precursor proliferation one hour after a BrdU pulse was approximately 70% higher in NSI-189-treated animals than in vehicle controls (p < 0.001; male C57BL/6J mice, 10 mg/kg/day ip, PMID 30258178). In the 15-day survival assay in the same paper, the NSI-189 group showed only a minor, non-significant difference, while the P7C3-A20 comparator increased net neurogenesis. The authors concluded that the pro-neurogenic effect is likely confined to proliferation, which is a narrower claim than the description usually attached to the compound. This is also the only rodent study in this record with no sponsor authorship or affiliation.

What is not known

The molecular target is unknown and has never been published, fifteen years after the first human dose; the current sponsor states only that it believes the compound binds a receptor no other CNS therapeutic targets. No in vitro screen, dose-response, selectivity panel or toxicology package appears anywhere in the peer-reviewed literature, which means the founding claims about potency and anatomical selectivity cannot be checked by anyone outside the sponsor. Human exposure data are thin: the only published pharmacokinetics come from 24 inpatients over 28 days, there is no published data in healthy volunteers beyond a registry entry for a 35-participant single-dose study that was never published, and no exposure beyond twelve weeks has been reported in any population. Neither of the two published trials, nor any of the four ALTO-100 studies, has results posted on ClinicalTrials.gov, and the Phase 2 registry record still carries a status of UNKNOWN more than eight years after its estimated completion date of December 2017. The 301-patient Phase 2b exists in the public record only as a table in an annual report; its adverse-event data, its full analysis populations and its biomarker definition have not been published. Nothing has been reported on reproductive or developmental effects, carcinogenicity, drug interactions, or use in anyone under 18 or over 70. Nothing has been studied in people without a psychiatric diagnosis. Two randomised trials in depression have now missed their primary endpoint, and no trial has ever been run for a cognitive-enhancement indication in a non-clinical population.

Questions