KW-6356

Chemical identity

One structure, four registry identifiers, and two names that do not mean the same thing. PubChem holds a single record under the development code: CID 23144148, molecular formula C20H19N3O4S, molecular weight 397.4, InChIKey KMFLQPJJHQNKKF-UHFFFAOYSA-N, systematic name N-[4-(furan-2-yl)-5-(oxane-4-carbonyl)-1,3-thiazol-2-yl]-6-methylpyridine-3-carboxamide. The FDA substance registry carries the same structure as UNII FF2012ZV92, with CAS 858979-50-7 recorded as a primary code alongside INN number 12173, PubChem 23144148 and NCI Thesaurus C190445, and records the structure as achiral. ChEMBL5314441 returns the same formula, a molecular weight of 397.46, a maximum development phase of 2 and no first approval date. The four records carry the same formula and the same InChIKey. Reported molecular weights differ only by rounding convention: 397.4 (PubChem), 397.448 (PDB), 397.4494 (GSRS), 397.46 (ChEMBL).

KW-6356 is the Kyowa Kirin development code; sipagladenant is the International Nonproprietary Name. Both resolve to the same substance record. The Protein Data Bank holds the molecule as chemical component JQR, formula C20H19N3O4S, formula weight 397.448, which matches. One curated database does not carry it at all: a ligand query to the IUPHAR/BPS Guide to Pharmacology under either name returns no record, which is unusual for a compound whose complex with its target receptor has been solved and deposited. Structurally it is a 2-aminothiazole carrying a furan at C4 and a tetrahydropyran-4-yl ketone at C5, acylated with 6-methylnicotinic acid. It is not a xanthine, which separates it chemically from caffeine and from istradefylline.

A second chemical entity circulates under the same name. Human metabolism of the parent generates a metabolite the sponsor's papers call M6 and describe as an active adenosine A2A antagonist and inverse agonist in its own right. Vendor product listings sell material labelled as Metabolite 6 with the parent's development code in brackets, as though the two were one substance. They are not, and the difference cannot be resolved from any public identity source: a PubChem name query for the metabolite returns no compound, no CAS number or UNII for it was located in the FDA registry, and no published structure was found. Material sold under that label cannot be checked against anything.

What the receptor pharmacology measured

Ohno and colleagues published the in vitro profile in Molecular Pharmacology in 2023. Affinity at the human receptor was reported as a pKi of 9.93 plus or minus 0.01, and the dissociation rate constant as 0.016 plus or minus 0.006 per minute. Functional assays in that report separated the compound from istradefylline on two axes rather than one. It behaved as an inverse agonist, suppressing constitutive receptor activity rather than only blocking agonist binding, and its antagonism was insurmountable — raising agonist concentration did not restore the maximum response. Istradefylline in the same experiments was surmountable. The abstract does not state replicate counts, and the full text is not open access.

Cross-species affinities appear in the Chemical Probes Portal's expert review, dated 9 January 2024, which summarises that paper and gives pKi values of 9.49 in mouse, 10.0 in rat, 9.55 in dog, 9.95 in marmoset and 9.93 in human. The portal records a biochemical Kd of 0.13 nM and a cellular Ki of 0.12 nM, and sets a recommended cellular working concentration of up to 100 nM. It also records what the compound has not accumulated: one cellular-use review, none in model organisms, and no portal endorsement, which requires three stars or more.

Two crystal structures were deposited with the paper and released on 22 March 2023. Entry 8GNE is the human adenosine A2A receptor in complex with KW-6356 at 2.3 angstrom resolution; entry 8GNG is the same receptor with istradefylline at 3.2 angstroms. The authors attribute the inverse agonism to contacts with His250 and Trp246, and the insurmountable behaviour to interactions deep in the orthosteric pocket together with stabilisation of the extracellular loop lid. An independent group reached a compatible conclusion about the comparator a decade earlier: Bennett and colleagues in 2013 classified istradefylline as a neutral antagonist with equal affinity for active and inactive receptor conformations, grouping it with theophylline and caffeine.

Two marmoset studies from one laboratory

Anti-parkinsonian activity was assessed in MPTP-treated common marmosets in two 2023 reports. In the monotherapy study, published in the European Journal of Pharmacology, oral administration reversed motor disability in a dose-dependent manner up to 1 mg/kg, and the magnitude of that reversal was reported as significantly greater than istradefylline's in the same experiment. Repeated administration to animals previously primed with L-DOPA to exhibit dyskinesia induced little dyskinesia. Neither the abstract nor any other record retrievable here states the number of animals per group. The article is licensed open access — Unpaywall returns is_oa true, hybrid, CC BY-NC-ND, published version at the publisher — but the publisher host returned HTTP 403 to every retrieval attempt made on 19 August 2026.

Combination effects were reported in the Journal of Pharmacological Sciences. Oral KW-6356 at 1 mg/kg enhanced the anti-parkinsonian activity of L-DOPA across doses of 2.5 to 10 mg/kg, and in animals primed to show dyskinesia it enhanced activity across 1.25 to 10 mg/kg without a corresponding increase in dyskinesia. Chronic co-administration for 21 days at 1 mg/kg with a low L-DOPA dose of 2.5 mg/kg did increase the degree of dyskinesia induced by that low dose, though the amplitude was lower than that induced by 10 mg/kg L-DOPA alone. This paper is gold open access in a DOAJ-indexed journal, but the publisher host returned HTTP 403 here too and the DOAJ record carries metadata only, so group sizes were again not read.

Both reports share an author group with each other and with the in vitro paper, and all three came from the originating company. No independent laboratory has published on this molecule. Binding constants are instrument readings; the MPTP marmoset endpoints are scored by observers, and all the observers worked for the sponsor. That screen is the standard preclinical test for anti-parkinsonian motor benefit and dyskinesia liability. Nothing in the published animal record addresses disease progression, and nothing addresses cognition, sleep architecture or arousal in any species.

Eight registered trials, none with posted results

ClinicalTrials.gov returns eight studies naming the compound, all sponsored by Kyowa Kirin entities — six by Kyowa Kirin Co., Ltd. in Japan, and the mass-balance and hepatic-impairment studies by the US subsidiary Kyowa Kirin, Inc. — all completed, and none carrying posted results. In start-date order: a phase 2a in early Parkinson's disease (NCT02939391, 175 enrolled, September 2016 to December 2017); the phase 2b adjunct study (NCT03703570, 502, September 2018 to April 2020); a phase 1 in Japanese and Caucasian healthy adults (NCT03830528, 48); an interaction study with midazolam, caffeine or rosuvastatin (NCT03970798, 50); an interaction study with clarithromycin or rifampicin (NCT04070495, 20); a radiolabelled mass-balance study (NCT04147910, 8); a hepatic impairment study (NCT04190654, 26); and a thorough QT study (NCT04342273, 128). Registered enrolment across the eight totals 957.

The registry labels both phase 2 arms only as low dose and high dose. The published paper and the conference abstract both identify these as 3 mg and 6 mg once daily, so the arm identities are recoverable from the literature but not from the registry record. One further human study appears in the literature with no registry record at all — the phase 1 that Tayama and colleagues label study 6356-001, which the paper itself states has no NCT number, and which tested single doses of 1, 3 and 10 mg and multiple doses of 6 mg once daily in healthy Japanese subjects.

Three peer-reviewed papers cover, between them, at most three of these studies. Six registered trials enrolling 734 participants have produced neither posted results nor a publication: the phase 2b, the thorough QT study, both drug-interaction studies, the mass-balance study and the hepatic impairment study. Cardiac repolarisation was assessed in 128 people against placebo and moxifloxacin, and what that assessment found is not in the public record. Absorption, metabolism and excretion were characterised with a radiolabel in eight people, and the mass balance is unpublished. Whether CYP3A inhibition or induction alters exposure was tested in twenty people, and the answer is not available.

The phase 2a that was published, and the phase 2b that was not

Maeda and colleagues published the monotherapy trial in Parkinsonism and Related Disorders in December 2023, six years after it completed. One hundred and sixty-eight patients with early, untreated Parkinson's disease were randomised and treated for twelve weeks — 55 to 3 mg daily, 58 to 6 mg daily and 55 to placebo — against a registry enrolment figure of 175. Least-squares mean changes from baseline in MDS-UPDRS Part III were -5.37 (95% CI -7.25 to -3.48) on 3 mg, -4.76 (-6.55 to -2.96) on 6 mg and -3.14 (-4.97 to -1.30) on placebo. Constipation was recorded in 4 and 6 patients on the two doses, nasopharyngitis in 4 and 5.

Those three point estimates and all three intervals had already appeared, unchanged, in a Kyowa Hakko Kirin press release dated 20 August 2018 describing a poster at the World Congress on Parkinson's Disease and Related Disorders. The phase 2b has had no such resolution. A release dated 21 October 2020 states that the trial studied 502 patients on levodopa-containing preparations, that the primary endpoint was change from baseline in MDS-UPDRS Part III over 26 weeks of administration, and that the difference from placebo reached statistical significance. It reports no number at all, and says results were planned for presentation at future conferences or in publications.

Presentation came in 2022, as abstract 743 at the International Congress of Parkinson's and Movement Disorders. It gives 503 patients randomised — 168 placebo, 168 on 3 mg, 167 on 6 mg — treated for 24 weeks, with p = 0.006 for 3 mg and p = 0.049 for 6 mg against placebo on the primary endpoint, plus a post-hoc finding on the Parkinson's Disease Sleep Scale-2 at 6 mg (p = 0.013 in patients with baseline sleep problems, p = 0.018 in the overall population). No effect size, confidence interval or least-squares mean difference for this trial has ever been published. The abstract and the press release disagree on the number randomised, 503 against 502, and on treatment duration, 24 weeks against 26; the registry records 502 enrolled and a primary outcome measured up to 26 weeks. Nothing retrievable reconciles them.

A pattern runs across both phase 2 studies that no published analysis addresses. In the monotherapy trial the lower dose produced the larger point estimate, -5.37 against -4.76. In the adjunct trial the lower dose produced the stronger p-value, 0.006 against 0.049, the higher dose clearing the conventional threshold by one thousandth. The sleep finding the abstract highlights, and that secondary coverage repeats, belongs to the 6 mg arm and is post-hoc.

Discontinued, and what remains in circulation

Rights outside Japan and the Asian region went to H. Lundbeck under a licensing agreement announced on 5 October 2010. That agreement was terminated by mutual agreement, and the August 2018 press release records that all rights had reverted to Kyowa Hakko Kirin, which said it would proceed independently. Development stopped four years later. A release dated 15 July 2022 announces the discontinuation of the clinical development programme and states that the phase 2 studies conducted in Japan found the compound potentially effective in relieving motor and non-motor symptoms, both as monotherapy and in combination with levodopa-containing therapy.

No clinical reason is given. The same release attributes the decision to an evaluation of the global regulatory landscape, development hurdles, and timelines for potential market entry. Nothing about a safety finding, nothing about a failed endpoint; the sentence describing the compound's efficacy sits immediately before the sentence announcing its abandonment. Four years on, the 502-patient trial that prompted the 2020 announcement remains unpublished and the eight registry records remain empty. A discontinued programme is under no obligation to publish, and that is the mechanism by which most of this dataset became unavailable.

Approval exists in no jurisdiction. ChEMBL records no first approval date, the FDA substance record carries no approved product, and no ATC code has been assigned. The compound is sold as powder and as solution by vendor sites that classify it as research material and disclaim human use; that classification is a commercial description rather than a regulatory status, and it corresponds to no category recognised by a medicines regulator. Istradefylline is approved in Japan and the United States. KW-6356 stopped at phase 2. The 2022 release attributes the decision to the global regulatory landscape, development hurdles and timelines for market entry, and cites no safety finding or missed endpoint.

What is not known

Most of what was measured in people is not in the public record. Eight registered trials enrolled 957 participants and none posted results; six of them, covering 734 participants, produced no publication either. That includes the thorough QT study, so nothing about cardiac repolarisation is retrievable; the radiolabelled mass-balance study, so the routes and extent of elimination are not established outside a population model; both drug-interaction studies, so whether CYP3A inhibition or induction alters exposure is unanswered in public, as is the interaction with caffeine that one of them was designed to test; and the hepatic impairment study. Renal impairment was never studied. The phase 2b, the largest dataset by a wide margin, exists as a press release that gives no number and a conference abstract that gives p-values and no effect sizes, and the two disagree on how many patients were randomised and for how long they were treated. Human pharmacokinetics are partly retrievable and partly not: the phase 1 abstract gives a terminal half-life range of 18.4 to 43.1 hours across single doses of 1 to 60 mg, while the single-point figures in wide circulation are cited to a paper containing no human pharmacokinetic data. Two of the animal papers are licensed open access but could not be retrieved here — the publisher host returned HTTP 403 — so the animals-per-group figures behind the marmoset results remain unread rather than absent. Every phase 2 patient was Japanese, and the only ancestry comparison in the record is one arm of a 48-participant phase 1. No exposure beyond 26 weeks has been studied in anyone, and nothing has been published on what happens after discontinuation. Outside Parkinson's disease the record is empty in every species: no cognition endpoint, no sleep-architecture measurement, no arousal or wakefulness study, and no measurement in an undamaged nervous system beyond safety pharmacology. All published pharmacology originates with the company that discovered the molecule, and no independent laboratory has replicated any of it.

Questions