AF710B

The M1 muscarinic receptor has long been a target for Alzheimer's drugs, because the receptors in the cortex survive the disease better than the neurons that supply their acetylcholine [7]. Several muscarinic agonists reached phase 2 or 3 trials and were dropped for lack of efficacy or because they also stimulated muscarinic receptors outside the brain; xanomeline, which went furthest, caused nausea, vomiting, diarrhea, sweating and excess salivation [7]. AF710B was the next attempt from Abraham Fisher's group at the Israel Institute for Biological Research, which had earlier produced the orthosteric M1 agonists AF102B (cevimeline), AF267B and AF292 [1].

Instead of binding where acetylcholine binds, AF710B sits at a separate, allosteric site on the M1 receptor and makes the receptor respond more strongly to acetylcholine, while also activating the sigma-1 receptor, a chaperone protein in the endoplasmic reticulum [1]. It is the single active enantiomer of the racemic compound AF710; the other enantiomer, AF710A, was inactive in the rat memory test [1]. A 2019 study that tested it under the name ANAVEX3-71 gives its configuration as (S) [8].

The animal results are the reason anyone is interested. In rats, single oral doses of 1 to 30 µg/kg reversed the memory deficit caused by an M1-blocking drug, and the developer reported it as the most potent of the compounds it had tested in that model, including donepezil and its own earlier agonists [1]. In 3xTg-AD mice, which develop both amyloid and tau pathology, two months at 10 µg/kg a day improved water-maze performance and lowered amyloid-β, plaques, tau phosphorylation and glial activation [1]. In McGill-R-Thy1-APP rats, a transgenic amyloid model, 10 µg/kg orally prevented cognitive decline when started before plaques formed [4]. Started after plaques had formed, it restored novel object recognition and water-maze learning to wild-type levels: on the last training day treated transgenic rats found the hidden platform in 38 seconds, against 69 seconds untreated and 29 seconds for wild-type rats. It did not improve the probe trial, in which the platform is removed to test memory of where it was [3]. In both rat studies the effects were measured after the drug had been stopped for four to five weeks [3][4].

The human record is thin. A first-in-human single-ascending-dose study in healthy volunteers gave 5 to 200 mg and found linear pharmacokinetics, a half-life of about 3.6 hours, no food effect and no effect on the QT interval [2][5]. No trial in Alzheimer's disease or frontotemporal dementia followed [9]. Instead Anavex ran a phase 2 trial in 71 adults with schizophrenia on stable antipsychotics, and in October 2025 announced that it had met its primary endpoint, which was safety and tolerability. The release described "positive trends" in EEG and event-related-potential biomarkers and a reduction in the inflammation marker GFAP, with no numbers, and reported no results for the symptom, cognition and global-impression scales the trial had registered as secondary outcomes [6][10].

An allosteric M1 agonist that behaves like a modulator. In rat cortical membranes AF710B displaced the M1 radioligand [3H]-pirenzepine with a high-affinity constant of 0.046 nM [1]. Brucine, an M1 modulator that potentiates agonists acting at acetylcholine's own site, potentiated carbachol and AF267B but not AF710B, which places AF710B's binding site elsewhere on the receptor [1]. At 0.1 nM, AF710B lowered carbachol's high-affinity constant from 70 nM to 1 nM and raised carbachol's efficacy index (the ratio of its low- to high-affinity constants) from 110 to 1650 [1]. On its own it did not raise ERK1/2 phosphorylation in M1-expressing PC12 cells, but at nanomolar concentrations it increased the ERK1/2 and CREB phosphorylation produced by carbachol, the signaling route linked to memory consolidation [1]. Direct agonism needed far higher concentrations: in cells expressing each muscarinic subtype it was a partial agonist at M1 at 100 µM, inactive at 1 µM, and had no agonist activity at M2 to M5 [1].

One result separates it from other M1 modulators. Those potentiate acetylcholine's whole range of M1 responses, including the release of calcium inside the cell, but AF710B did not potentiate carbachol's calcium response; the authors report this, and several other cell results, as data not shown [1].

A sigma-1 agonist as well. AF710B displaced the sigma-1 ligand [3H]-pentazocine with a high-affinity constant of 1.3 nM and did not bind sigma-2 [1]. In the rat memory test the sigma-1 antagonist NE-100 cut its effect roughly in half, from a retention latency of 362 to 185 seconds, so the effect there depends partly on sigma-1; since the deficit was produced by an M1 blocker, the authors attribute the benefit to the two receptors acting together [1].

In mitochondria isolated from mouse forebrain, a Montpellier laboratory, testing it alongside other sigma-1 ligands in a study partly funded by Anavex, found that it raised reactive oxygen species by about 12% at 0.3–1 µM and increased complex I activity at 1 and 10 µM — concentrations far above those at which it binds its receptors. The experiments showing that the complex I effect depends on calcium and is reversed by a sigma-1 antagonist were done with the reference agonist PRE-084, not with this compound [8]. Against amyloid-β, it did less than the other agonists: its reduction of the amyloid-induced rise in reactive oxygen species did not reach significance, it did not protect respiration or complex I, and it only partly offset the loss of complex IV activity [8].

Both receptors were needed to protect spines. Mushroom-shaped dendritic spines, the stable kind associated with memory, are lost in hippocampal cultures from PS1 and APP knock-in mice. A 16-hour exposure to 30 nM AF710B restored them nearly to normal in both lines, while 1 µM of the orthosteric agonist AF267B gave a partial rescue in one line and none in the other [1]. The M1 antagonist pirenzepine abolished the rescue and NE-100 did not; but knocking down sigma-1 itself reduced mushroom spines from 38.4% to 22.1% in normal neurons, and AF710B could not reverse that [1]. The authors proposed that AF710B acts on an M1–sigma-1 receptor complex in the spine, and said further experiments would be needed to test it [1].

Downstream it touches each Alzheimer's hallmark measured. In 3xTg-AD mice, two months at 10 µg/kg a day lowered the β-secretase BACE1 and its APP product C99 while leaving APP and the α-secretases unchanged, reduced soluble and insoluble Aβ40 and Aβ42 and thioflavin-positive plaques, raised inhibitory phosphorylation of GSK3β, reduced the CDK5 activator p25, lowered tau phosphorylated at five epitopes including PHF-1, and reduced reactive astrocytes and microglia near plaques [1].

In McGill-R-Thy1-APP rats treated from 13 months of age, after plaques had formed, 4.5 months at 10 µg/kg followed by a 5-week washout cut mature plaques in the subiculum by 84% and insoluble and soluble cortical Aβ42 by 61% and 66%. Aβ42 in the cerebrospinal fluid rose, which the authors read as amyloid being cleared from the brain rather than trapped in plaques. Microglial clustering around CA1 neurons, hippocampal Iba1 protein and cortical IL-10 expression, all raised in the transgenic rats, returned to wild-type levels, as did the reduced synaptophysin mRNA and protein; the drug changed none of these in wild-type rats [3]. Started at 7 months, before plaques, and given for 7 months, it reduced plaque numbers and cortical Aβ42 and Aβ40, reduced glial recruitment toward amyloid-laden neurons and restored the conversion of the BDNF precursor to mature BDNF [4].

The evidence for how it works in people is limited to biomarkers. In the part A dose-escalation stage of the schizophrenia trial, a review reports dose-dependent increases in 40 Hz auditory steady-state response coherence and resting-state alpha power, measures linked to cortical function in schizophrenia [11]. The topline release for the whole trial described only "positive trends" in EEG and event-related-potential measures and lower GFAP than placebo, without values or statistics [6].

In animals, toxicity was far from the effective dose. In a GLP study in rats, single oral doses up to 50 mg/kg produced no toxic effects and no gross pathology, and the developer calculated a margin of 50,000 between the toxic and effective doses, against 400 for AF267B and 115 for cevimeline in the same rat model; the toxicology data themselves were not shown [1]. Two months of daily dosing in mice did not change body weight [1].

In healthy volunteers, single doses of 5 to 200 mg did not affect the heart's electrical cycle. There was no clinically relevant effect on heart rate or the PR and QRS intervals, the concentration–QTcF slope was slightly negative, and an increase above 10 ms was excluded at every plasma concentration observed [2]. The full texts of the two phase 1 papers were not available for this entry, so their adverse-event listings are not summarized here; the registered study enrolled 36 people [12].

In schizophrenia, adverse events were no more common than on placebo. In the 28-day part B, 11 of 28 people on 60 mg (39.3%) had a treatment-emergent adverse event against 13 of 27 on placebo (48%), and 5 (17.9%) against 4 (15%) had one judged related to treatment. None was serious or severe, and no one stopped the drug because of an adverse event, against two on placebo. In part A, one of six people on 30 mg had a treatment-related adverse event and none on 60 mg did [6]. The release did not say what the adverse events were, and these figures are company data that have not been peer reviewed.

The class risk is cholinergic. Earlier muscarinic agonists were abandoned largely because stimulating M1, M2 and M3 receptors outside the brain caused nausea, vomiting, diarrhea, sweating and excess salivation [7]. AF710B had no agonist activity at M2 to M5 in cell assays [1], but whether that spares people those effects cannot be judged until the adverse events from the human trials are itemized.