Recent Advances in Therapies for Potassium Ion Channel Dysfunction in Autism

 

Ο ήλιος, η θάλασσα και το EpiphanyASD — οι καλύτερες συνθήκες για να ανοίξουν νέα κανάλια σκέψης!

A recent comment from Konstantinos prompted this update to my earlier posts on potassium channels. Progress is being made, particularly in Fragile X, but also potentially for a sub-group of idiopathic autism with a specific pattern in their EEG.

  

From the original channelASD hypothesis to drugs entering clinical trials

Back in 2015, I wrote about the possibility of classifying some forms of autism according to dysfunction in particular ion channels. I used the term channelASD, drawing attention to potassium (Kv), sodium (Nav) and calcium (Cav) channels.

At the time, this was largely a mechanism-driven hypothesis. We knew that ion channels regulate neuronal excitability, and genetic research was beginning to implicate ion-channel genes in neurodevelopmental disorders. The therapeutic question was whether correcting a specific channel abnormality might improve some of the symptoms associated with autism.

More than a decade later, there are some interesting developments.

Two different potassium-channel activators are now attracting attention. Azetukalner, a Kv7 potassium-channel opener developed by Xenon Pharmaceuticals, has reached late-stage clinical development for epilepsy, with a New Drug Application submitted to the FDA in September 2026. Meanwhile, Spinogenix has announced a new autism study of codabakalner (SPG601), which activates a different potassium-channel family, BK channels. A second BK-channel opener, KER-0193, is also heading into Phase 2 in Fragile X syndrome.

None of these drugs has been demonstrated yet to treat autism. Nevertheless, these developments suggest that the ion-channel hypothesis deserves another look.

 

1. Why potassium channels matter

Potassium channels help control the electrical activity of neurons. When a neuron becomes activated, the movement of potassium ions across its membrane helps regulate the duration of electrical signals and determines how readily the neuron can fire again.

Different potassium channels operate at different stages of neuronal activity.

•        Kv7 channels help stabilise the resting membrane potential and limit repetitive neuronal firing. Kv7 plays a role in sensory gating, which is very often disturbed in autism.

•        BK channels are activated by both intracellular calcium and membrane depolarisation, helping regulate action potentials and neurotransmitter release. BK channels may help regulate sensory filtering and habituation.

•        SK channels are activated by intracellular calcium and contribute to after-hyperpolarisation, influencing how frequently neurons fire. SK channels help regulate neuronal firing patterns and sensory-signal processing, making them plausible contributors to sensory-processing abnormalities in autism; however, their role is complex, and the evidence for a specific SK-channel mechanism in autism remains limited.

•        Kv3 channels help certain neurons fire at high frequencies with precise timing, which is important for synchronised neural activity. Kv3 channels are a promising target for investigating sensory hypersensitivity in autism

•        Slack (KNa) channels are sodium-activated potassium channels that are directly regulated by FMRP, the protein missing in Fragile X syndrome.

It is tempting to think of potassium-channel activation as a general way to calm an overactive brain. However, this is too simple. Different channels have different functions, and some forms of neurodevelopmental disorder involve channels that are overactive rather than underactive.

The therapeutic objective should therefore be to correct the relevant physiological abnormality, not simply to increase potassium-channel activity.

A useful review is Potassium channels and autism spectrum disorder: An overview (2021), which discusses several potassium-channel genes implicated in neurodevelopmental disorders.

Read the review

2. Azetukalner: Kv7 channel activation reaches the FDA

Our very active reader, Konstantinos, recently commented on my old channelASD article:

"Ancient post, but Xenon filed for FDA approval of azetukalner, a potassium channel opener. Epiphany is such a precious knowledge place :)"

Thank you, Konstantinos!

Azetukalner is a Kv7 channel opener developed by Xenon Pharmaceuticals. Positive Phase 3 results (X-TOLE2) in focal-onset seizures were announced in March 2026, and on 17 September 2026 the company announced submission of a New Drug Application to the FDA.

Kv7.2 and Kv7.3 channels, encoded by KCNQ2 and KCNQ3, contribute to the neuronal M-current. This current acts as an important brake on repetitive firing.

Enhancing Kv7 activity can reduce neuronal excitability, which is why this channel family has attracted considerable interest in epilepsy drug development.

The progress of azetukalner demonstrates that Kv7 modulation can be translated into a clinically advanced treatment strategy.

Xenon is running Phase 3 studies of azetukalner in major depressive disorder (X-NOVA2 and X-NOVA3) and in bipolar depression, with topline data from X-NOVA2 expected in the first half of 2027. Those trials will tell us a good deal about how a Kv7 opener behaves in a psychiatric population, including its tolerability, although they say nothing directly about autism.

Xenon's September 2026 announcement

 

An important complication: some KCNQ mutations increase channel activity

Genetic studies have identified individuals with KCNQ2 or KCNQ3 gain-of-function variants who have developmental delay, language impairment and autistic features.

Sands et al. (2019) described 11 people with KCNQ3 gain-of-function variants. All had developmental delay and autistic features, while five received a formal autism diagnosis.

A separate study of 15 individuals with KCNQ2 gain-of-function variants found autistic features in ten.

These findings are particularly interesting because they challenge the idea that opening potassium channels should always be beneficial. In these particular genetic disorders, the channel is already excessively active. Further activation might be counterproductive.

The clinical implications are not straightforward, because the relationship between channel activity and neuronal network function is complex. Nevertheless, these studies illustrate why a genetic diagnosis and, where possible, functional characterisation may be important when considering targeted treatment.

3. BK channels: from Fragile X mouse studies to a human autism trial

Another strand of my earlier writing concerned calcium-activated potassium channels, particularly BK and SK channels.

The BK channel, encoded by KCNMA1, is activated by intracellular calcium and membrane depolarisation. It contributes to the regulation of neuronal firing and synaptic function.

Fragile X syndrome is particularly relevant to this line of research because loss of FMRP affects the regulation of numerous synaptic proteins and ion channels.

Human genetic evidence for BK channels

The case for BK channels does not rest only on mouse models. In 2006, Laumonnier and colleagues reported a person with autism and intellectual disability in whom the KCNMA1 gene was physically disrupted by a chromosomal translocation. The gene showed haploinsufficiency and the resulting BK channel activity was reduced, but could be enhanced in vitro by a BK-channel opener.

This is a single case, and a loss of BK function in one individual is not the same as a shared defect in Fragile X. It does, however, give the BK hypothesis a human genetic foothold.

Laumonnier et al., 2006, American Journal of Psychiatry

FRAXA published a helpful overview of the BK story in April 2026.

FRAXA: Opening the Gate — BK channel modulation in Fragile X

 

Chlorzoxazone

In 2024, Ferraguto and colleagues published a study of chlorzoxazone in a mouse model of Fragile X syndrome.

Chlorzoxazone is an existing muscle relaxant that can activate BK channels, although it is not a selective BK-channel drug.

In the Fmr1-knockout mouse model, chlorzoxazone improved several experimental outcomes, including hyperactivity, acoustic hyper-responsiveness and social-interaction measures. The investigators also reported restoration of impaired BK currents in hippocampal and cortical neurons.

The study provides preclinical evidence supporting BK channels as a potential therapeutic target in Fragile X syndrome.

This research is particularly relevant to the problem of sound sensitivity. In some individuals with autism, sound hypersensitivity can be severely disabling. Fragile X mouse models provide a way to investigate the underlying auditory physiology and test candidate treatments.

However, sound hypersensitivity can arise from several mechanisms, including abnormalities in peripheral auditory processing as well as central neural circuits. A drug that helps one mechanism may not help another.

 

Codabakalner (SPG601)

The most important recent development is the progression of BK-channel activation into human clinical research.

In April 2026, Pedapati and colleagues published a randomised, double-blind, placebo-controlled crossover study of SPG601 in ten adult men with genetically confirmed Fragile X syndrome.

Following a single 800 mg dose, the investigators reported changes in resting-state EEG gamma power and other EEG measures, alongside improvement on a measure of attention and inhibitory control.

Not every measure improved: auditory-evoked gamma outcomes did not show significant treatment effects.

The study was small and examined a single dose, so it cannot establish long-term efficacy. Nevertheless, it provides initial human evidence that BK-channel activation can produce measurable neurophysiological effects.

Read the original study in Scientific Reports

The new autism study

In October 2026, Spinogenix announced that it was adding a Phase IIb study of codabakalner to its CLARITY programme, specifically targeting idiopathic autism.

CLARITY is a Phase 2b/3 basket trial (NCT07439510). The Fragile X part enrols men with full-mutation Fragile X syndrome, and the autism part is a Phase 2b study in adults. According to the ClinicalTrials.gov record, the trial was listed as not yet recruiting in October 2026, with an estimated start date of December 2026 and an estimated primary completion date of February 2030. Codabakalner has FDA Fast Track designation for Fragile X syndrome, and CLARITY has grant support from the FRAXA Research Foundation. Readers should therefore not expect results for several years.

The proposed recruitment strategy is particularly interesting. Participants will be screened using EEG, with enrolment focused on individuals who have elevated resting gamma-band power resembling a pattern observed in Fragile X syndrome.

In the reported EEG analysis, this signature was found in 35 of 159 autistic participants, or approximately 22%. Among adults, the proportion was 23 of 72, or approximately 32%.

The subgroup could not be reliably distinguished from other autistic participants using 27 conventional clinical assessments after statistical correction.

The investigators therefore propose that EEG may identify a biologically relevant subgroup that would otherwise be missed by behavioural assessment alone.

This is an important hypothesis, but it remains to be validated. An EEG pattern resembling that of Fragile X syndrome does not prove that the same molecular defect is present, nor does it establish that an individual will respond to a BK-channel activator.

The decisive question is whether participants selected by this EEG signature experience meaningful clinical improvement compared with placebo.

ClinicalTrials.gov record for CLARITY (NCT07439510)

Spinogenix announcement of the autism Phase 2b study

 

A second BK opener: KER-0193

Codabakalner is no longer the only BK-channel modulator in clinical development. In September 2025, Servier agreed to acquire KER-0193 from Kaerus Bioscience in a deal worth up to $450 million. KER-0193 is an oral small-molecule BK-channel modulator being developed for Fragile X syndrome.

A Phase 1 study in 56 healthy volunteers found it safe and well tolerated, and an EEG substudy was reported to show proof of mechanism. The FDA has granted it orphan drug and rare paediatric drug designations for Fragile X syndrome, and Servier planned to start a Phase 2 trial in Fragile X patients in 2026.

The significance here is that two independent companies are now testing BK-channel modulators in humans. That does not make either drug effective, but it makes the target harder to dismiss. As with codabakalner, the indication is Fragile X syndrome, not autism in general.

Servier's acquisition of KER-0193

 

4. My earlier interest in mefenamic acid (Ponstan)

Another old EpiphanyASD research thread concerned mefenamic acid, marketed as Ponstan, and its possible effects on potassium channels.

Ponstan is widely used as a pain medication in many countries. In Greece it is OTC and very cheap - there is even a liquid version for children. It is very common in India. In the US it is 10 to 100 times more expensive, so buyer beware.

In my 2016 post, I discussed research showing that related fenamate compounds, including meclofenamic acid and diclofenac, can activate Kv7.2/Kv7.3 currents in experimental systems.

This was an intriguing connection because the neuronal M-current is a well-established regulator of excitability.

In a later post, I explored mefenamic acid's multiple biological effects and its possible relevance to sound sensitivity in Monty. This was an individual observation and a mechanistic hypothesis, not evidence from a controlled autism trial.

My 2016 post on mefenamic acid and autism

My later post on fenamates

The lesson from this line of investigation is that drug repurposing can generate useful hypotheses, but a plausible molecular mechanism is only the beginning of the process.

Mefenamic acid has also been the subject of a patent family for autism, “Treatment and prevention of autism and autism spectrum disorders, associated with Knut Wittkowski and ASDERA. The filings cover mefenamic acid and related fenamates, alone or combined with other agents, and describe potassium, chloride and calcium channel modulators. Wittkowski’s published rationale, however, centres mainly on calcium-activated chloride channels (anoctamins), with mefenamic acid’s potassium-channel effects as only part of its pharmacology. ASDERA’s prodrug of mefenamic acid, ASD-002, was licensed to Q BioMed for development in autism with loss of language. A patent shows that an idea has been protected, not that the drug works. There is no mention in the patent that mefenamic acid or its prodrug benefits sound sensitivity.

In some people (Monty and others) a low dose of Ponstan alleviates sound sensitivity almost immediately.

 

5. What about SK, Kv3 and Slack channels?

BK channels are not the only potassium channels that might be relevant to sensory hypersensitivity.

SK channels are another calcium-sensitive potassium-channel family. They help regulate after-hyperpolarisation and the pattern of neuronal firing.

Chlorzoxazone can also affect SK currents, although its effects on these channels are less prominent than its BK-channel effects in the experimental literature. In the 2024 Fragile X mouse study, the investigators found that chlorzoxazone only partially restored impaired SK currents.

This illustrates why it is important not to assume that a drug's behavioural effects arise exclusively from its most prominent molecular target.

Kv3 channels are also relevant to auditory processing. A 2019 study found abnormalities in high-threshold potassium currents in the auditory brainstem of Fragile X mice and investigated Kv3-channel modulators. These experiments suggest that different potassium-channel families may contribute to different aspects of sensory processing.

Read the Kv3 study

That work has continued. Autifony Therapeutics, the company that supplied the Kv3 modulators used in those experiments, lists Fragile X syndrome among the intended uses of its newer Kv3-modulator compounds. A 2026 abstract from a National Fragile X Foundation junior investigator award also reports effects of two Kv3 positive modulators on EEG biomarkers in Fmr1-knockout mice. This is still preclinical, but it is interesting that Kv3 research is now also using EEG measures, the same kind of biomarker approach used in the codabakalner studies.

FRAXA: Potassium channel modulators to treat Fragile X

NFXF 2026 junior investigator awardees

The Yale group behind the Kv3 work, led by Leonard Kaczmarek, also found that loss of FMRP has opposite effects on two different potassium channels: Kv3.1 currents are increased, while Slack currents are decreased. FMRP directly binds the Slack channel and activates it. This is a good illustration of the point made earlier in this post. Within one disorder, one channel may need to be dampened and another boosted, so "potassium-channel activation" is not a single strategy.

The broader point is that sensory hypersensitivity is not necessarily a single biological phenomenon. Different individuals may have different abnormalities in auditory processing, neuronal firing or network synchronisation.

 

6. Where does this leave the channelASD hypothesis?

I would now distinguish three levels of evidence.

First, genetic evidence. Pathogenic variants in several potassium-channel genes can cause neurodevelopmental disorders that include autistic features. This is strong evidence that ion-channel dysfunction can contribute to particular forms of autism. For BK channels, the human evidence is thinner, consisting of isolated cases such as the KCNMA1 disruption reported in 2006, plus the Fragile X link through FMRP.

Second, experimental therapeutic evidence. Studies of chlorzoxazone and other potassium-channel modulators in Fragile X models show that correcting aspects of channel function can improve selected experimental outcomes. Kv3 modulators have shown similar effects on auditory and sensory measures in mice. These findings support further research but do not establish clinical efficacy.

Third, clinical translation. Azetukalner has advanced into regulatory review for epilepsy, with Phase 3 depression studies reporting from 2027. Codabakalner has produced preliminary human EEG and cognitive findings in Fragile X syndrome and is about to be investigated in an EEG-selected autism subgroup, while KER-0193 is following it into Phase 2 in Fragile X.

These developments make the field more promising than it was when I first wrote about channelASD. However, none of these drugs has established efficacy in autism.

The most interesting question is whether we can identify a particular physiological abnormality, establish its mechanism, and then select a drug that corrects that abnormality.

Conclusion

The recent progress in potassium-channel drug development is encouraging. My original interest in channelASD was based on the idea that autism encompasses multiple biological conditions, some of which might be amenable to targeted pharmacological treatment.

Azetukalner shows that Kv7 modulation has reached an advanced stage of clinical development for epilepsy, and its depression trials will show how far the approach extends. Codabakalner is taking BK-channel modulation into a biomarker-enriched autism trial, and KER-0193 gives the BK target a second clinical programme. Earlier work with chlorzoxazone, Kv3 modulators and other compounds provides additional experimental support for investigating potassium channels in Fragile X syndrome and related neurodevelopmental conditions.

The challenge is to move beyond the broad observation that ion channels matter. We need to identify which channel is abnormal, in which individuals, in which direction, and whether correcting it produces a meaningful improvement.

That is the real promise of channelASD: not one treatment for all autism, but the possibility of matching a treatment to a demonstrable biological abnormality.

 

EpiphanyASD Research Copilot — an AI assistant with an EpiphanyASD research memory and perspective

  

 At last we can understand all of Peter's blog! and find those lost comments.

Some readers already make extensive use of AI. They will have realized that it very often needs some coaching to get the best results. AI needs context and when you give it context and your own operating instructions the results are much more usable.

A couple of months ago I decided to create a portable add-on to AI that gives it the starting knowledge from this blog and my book. It has been downloadable from the top right corner of the full-sized web page, but not visible on phones. 

The Copilot combines your AI's general reasoning with the knowledge accumulated from the EpiphanyASD blog and book Game Changer, reader experiences and research hypotheses. The more extensively a subject has been explored in EpiphanyASD, the more additional context and perspective the Copilot can potentially bring to its answer, while current scientific literature can be used to test and challenge those ideas.

Areas such as summertime raging in autism or bumetanide use can therefore produce a much richer response than a standard AI. For simple technical questions, however, the Copilot will give a very similar answer to ordinary AI.

Public-facing AI is deliberately conservative when it comes to health topics — for good reason. But that same caution means it is not well placed to weigh the cost of doing nothing for someone with severe, treatment-resistant symptoms, which is a judgment call only a parent or clinician close to the situation can really make.

The Copilot does not remove any of the AI's built-in safety behavior — it still will not, for example, tell you to start, stop, or adjust a medication. What it adds is the accumulated EpiphanyASD context, so the AI's caution is better informed rather than reduced. There is not going to be an AI autism doctor app.

 The Copilot is portable and reproducible. It is not a proprietary black box—you have the knowledge base and the instructions, and can recreate it in a compatible AI of your choice.

The EpiphanyASD knowledge stays separate from the AI model, so as AI models improve, the same knowledge base can potentially be used with them.

In simple terms: EpiphanyASD provides the memory and perspective; your chosen AI provides the reasoning.

 

How to use the portable Copilot

The Copilot has been tested successfully with ChatGPT and Claude. Other AI platforms may have different file-handling and knowledge-retrieval capabilities; Gemini was not fully compatible.

You do not need a paid AI account to use the Copilot, although free accounts may have usage limits. You can ask questions in any language.

 

Step 1: Download the two files

Click each link below to open it in Google Drive, then click the Download icon (the small downward arrow near the top right of the screen) to save the file.

1. The Knowledge-Base File:
EpiphanyASD_Knowledge_Base_v0.4.2_COMPLETE.zip

Download Knowledge File

2. The Instruction File:
EpiphanyASD_Research_Copilot_v5.4_Instructions.txt

Download Instruction File

 

If you are using a PC, the files will normally appear in your Downloads folder. You do not need to open or read them yourself.

On a phone, the files will normally appear in your device's Files or Downloads area. If you are not sure where they went, search your Files app for the filenames.

 

Step 2: Upload both files to your AI

1.     Open ChatGPT or Claude.

2.     Look next to the text box where you type your prompt for the + (plus) icon or Paperclip / Add file button.

3.     Upload both downloaded files into the chat.

 

Check that the knowledge base was actually loaded

The Copilot is designed to check automatically whether it can access the knowledge base.

If you are unsure whether it has worked, ask:

“Initialize the knowledge base”

The AI should tell you what material it has actually been able to read.

Depending on your own settings, ChatGPT or Claude might not open the ZIP file, but should do when you ask directly. Otherwise extract the ZIP on your computer or phone and upload the individual files inside it instead, together with the instruction file.

 

Step 3: Start asking your questions!

Once the knowledge base has been successfully accessed, simply ask your questions as normal.  Once it is loaded, you can ask as many questions as you like.

If your question relates to topic areas, reader experiences or hypotheses covered extensively by EpiphanyASD, the AI can draw on that research base while providing broader scientific context and, where appropriate, using current scientific literature to test and challenge those ideas.

For simple technical questions, it will behave much like an ordinary AI.

Important: The Copilot is specific to the chat in which you uploaded the files. When you start a new chat, simply attach the two files again to recreate it.

What can it do?

For longtime users of the blog, it should be amusing because the Copilot can reference long-forgotten comments.

For example:

“What did Tyler write about BCAAs?”

Clever guy, that Tyler.

Not only can you recover Tyler's reported treatment for self-injury, the AI can explain the proposed mechanism and identify other reader experiences associated with the idea. What more could you ask for? All without searching throughout the blog.

Or ask:

“What did Seth Bittker write about in the blog and later publish?”

He certainly has been busy.

You can get a summary of Seth's online autism activities and his publications.

That is the point of the EpiphanyASD Research Copilot: instead of searching through years of blog posts and comments yourself, you can ask your AI a question and let it search the accumulated EpiphanyASD knowledge for you. And as a bonus, it can add context from the wider scientific literature, helping to test and challenge the ideas found in the EpiphanyASD material.

 

Does paying for AI make a difference?

Our initial testing suggests that it can. We tested the Copilot on paid and unpaid versions of ChatGPT and Claude using natural questions about autism, self-injury and prognosis. The paid versions generally showed stronger evidence synthesis and calibration, with paid Claude performing best in our initial comparison and paid ChatGPT close behind. The free versions were still useful, particularly for retrieving relevant corpus material, but were somewhat more likely to make confident interpretations without equally strong evidence qualification. This is an early, informal comparison rather than a controlled benchmark, so we regard the result as provisional rather than proof that a paid AI will always perform better.

 

What makes the Copilot different from simply asking an AI about autism?

A general AI starts with its general training and whatever current sources it can access. The Copilot adds a separate, structured research layer containing the EpiphanyASD blog, Game Changer book, reader experiences, source/provenance information and selected reference material.

This matters because much of the EpiphanyASD material consists of observations and hypotheses that are scattered across many years of writing and comments. The Copilot can retrieve that material, identify who reported what, distinguish a personal experience from published evidence, and then use current scientific literature to test or challenge the interpretation.

 

For our very computer literate readers

Some readers have been doing clever things like reanalysing genetic testing results.

My Copilot project started out as a custom GPT that could be shared with ChatGPT users, including those with free accounts. That was quite quick to set up, but sharing was automatically blocked because the GPT is health-related.

So I decided to make the Copilot portable instead. If you have a paid ChatGPT or Claude account, you can very easily create your own custom version using my knowledge ZIP file and instruction file. Then you have it permanently available without uploading the files for every session.

You can also adjust the instructions to suit your own taste, or add your own knowledge files.

 

One practical tip: if you build up a huge thread of questions in the portable version, the AI may gradually lose track of the knowledge base. If that happens, refresh it by re-uploading the files or starting a new chat.

 

A second Korean NMDA therapy tackles the same problem as AST-001 — But from a different angle

 

SLC6A20-ASO restores NMDA receptor function in SHANK2 and SHANK3 autism models, including adult mice and human brain organoids

 

Lay Summary

A second Korean research programme is now targeting the same NMDA receptor problem as the experimental drug AST-001, but from a different direction.
NMDA receptors are important for learning, communication and social behaviour, and they need two keys to work properly.
One key is glutamate; the second is either glycine or D-serine.
AST-001 provides L-serine, which can potentially increase the availability of D-serine and glycine—the molecules that provide the second key.
The new treatment, SLC6A20-ASO, takes a different approach by reducing the removal of glycine from the brain. This leaves more glycine available to activate NMDA receptors.

In mice with SHANK2 or SHANK3 mutations, SLC6A20-ASO restored NMDA receptor function and improved several autism-related behaviours. Some of these effects occurred in adult mice, suggesting that the signalling problem may remain reversible after development.
The researchers also reproduced the NMDA receptor rescue in human brain organoids carrying SHANK2 or SHANK3 mutations. SHANK3 is particularly important because mutations in this gene cause Phelan-McDermid syndrome.
The two treatments are therefore different, but they may be tackling the same underlying bottleneck: insufficient NMDA receptor co-agonism.
AST-001 has already produced a positive signal in a Phase II autism trial, and a follow-up study found that many previous responders continued to respond after a year.
One Korean reader of this blog has also reported that L-serine appeared to help their child, although this is only an anecdotal observation.

 ---

A few years ago, I became interested in the possibility that NMDA receptor dysfunction might be an important part of autism in at least some people.

That led me to write about AST-001, a Korean drug based on the amino acid L-serine. The results from its Phase II clinical trial were encouraging, and a subsequent long-term study has provided additional evidence that some children who respond may maintain that benefit.

Epiphany: High dose L-Serine to treat children under 7 with severe autism + ID ? It works in Korea

 

Now another Korean research group has reported a fascinating new approach to the same biological system.

The treatment is called SLC6A20-ASO, and unlike AST-001, it does not supply L-serine. Instead, it targets a protein involved in transporting glycine, another important molecule involved in NMDA receptor activation. 

Glycine-modulating Slc6a20a-ASO restores NMDA receptor function in SHANK2 and SHANK3-mutant mice and cortical organoids | Nature Communications

 

The two approaches are therefore different, but they may be trying to solve the same underlying problem: insufficient NMDA receptor co-agonism.

That makes the new study particularly interesting.

 

The NMDA receptor has two keys

To understand why these two therapies might be related, it helps to think of an NMDA receptor as a door with two keys.

One key is glutamate.

The second key is either glycine or D-serine.

Both keys are needed for the receptor to open properly.

So you can have plenty of glutamate, but if there isn't enough of the second key, NMDA receptor signalling can still be reduced.

This is called NMDA receptor hypofunction.

The important point is that AST-001 and SLC6A20-ASO approach this second key from different directions.

 

AST-001: helping provide the second key

AST-001 is a pharmaceutical formulation of L-serine.

L-serine is not itself the second key. Instead, it is a precursor that can contribute to the production of D-serine and glycine, both of which can act at the NMDA receptor co-agonist site.

So the simplified idea is:

L-serine

↓

D-serine / glycine pathways

↓

more potential availability of the second NMDA key

↓

greater NMDA receptor activation

 

This was one of the possible mechanisms I discussed in my earlier article about AST-001.

The important thing is that this is a potential mechanism, not something the clinical trial definitively proved.

 

What happened with AST-001?

The original Phase II trial was conducted in 151 children aged 2–11 years with autism.

The high-dose AST-001 group showed a statistically significant improvement in the Vineland Adaptive Behavior Composite, a measure of adaptive functioning, compared with the placebo-to-high-dose control group.

There was also an improvement in the Clinical Global Impression-Severity measure.

One particularly interesting observation was that the younger children appeared to benefit more strongly, especially those aged 2–6 years. This was a subgroup finding, however, and should not be interpreted as proof that AST-001 only works in younger children.

Since then, there has been a 52-week follow-up study.

The follow-up included 61 children who had previously responded to treatment. Among the 48 children who had been classified as CGI-I responders at week 24, 45 continued to meet the responder criteria at week 52.

The treatment was also generally well tolerated.

This is encouraging, although the long-term study did not include a concurrent placebo group and therefore cannot prove that the improvement over the year was entirely due to AST-001.

So AST-001 is still an investigational treatment, but the evidence is becoming increasingly interesting.

 

And one reader of this blog actually tried L-serine

There is also an interesting real-world observation from this blog.

A Korean reader who had followed the AST-001 research told me that they tried L-serine and found that it appeared to help their child.

Obviously, this is only an anecdote.

It cannot establish that L-serine caused the improvement, and there was no placebo control or blinded assessment.

But it is interesting to record because it is consistent with the direction of the clinical research.

We now have a positive Phase II signal, encouraging long-term follow-up among previous responders, and at least one independent experience reported to this blog.

That certainly doesn't prove that L-serine works for autism generally.

But it makes the AST-001 story worth following.

 

Now comes the second Korean NMDA approach

The new study comes from the Institute for Basic Science (IBS) Center for Synaptic Brain Dysfunctions in South Korea.

The researchers were interested in glycine, the other molecule that can provide the second key for the NMDA receptor.

They focused on a protein called SLC6A20, which transports glycine.

Their idea was remarkably simple:

What if you prevent some of the glycine from being transported away?

More glycine would then remain available around NMDA receptors.

The researchers used an antisense oligonucleotide, or ASO, to reduce production of SLC6A20.

This is very different from giving a glycine supplement.

The ASO is a highly targeted molecular intervention designed to reduce expression of a particular protein.

 

SLC6A20-ASO: protecting the second key

The simplified mechanism is therefore:

 

SLC6A20 inhibition

↓

less glycine transported away

↓

more local glycine available

↓

more of the second NMDA key

↓

restored NMDA receptor function

 

So we now have two approaches:

AST-001

L-serine → D-serine/glycine pathways → potentially more NMDA co-agonism

SLC6A20-ASO

less glycine transport → more local glycine → more NMDA co-agonism

Different approaches for the same biological bottleneck.

That is what makes the new study so interesting in the context of AST-001.

 

Testing the idea in SHANK2 and SHANK3 autism models

The researchers tested SLC6A20-ASO in mice carrying mutations in SHANK2 and SHANK3.

These are important autism-associated genes involved in the organisation and function of excitatory synapses.

SHANK3 is particularly relevant to Phelan-McDermid syndrome, where loss or disruption of one copy of SHANK3 can produce a neurodevelopmental disorder with autism, intellectual disability and other features.

The researchers found that the SHANK2 and SHANK3 mutant mice had reduced NMDA receptor function.

After SLC6A20-ASO treatment, NMDA receptor function was restored.

And the effects were not limited to electrical measurements.

Several autism-related behavioural abnormalities also improved.

 

The adult-mouse finding

Perhaps the most striking part of the study was that the treatment worked in adult mice.

The researchers were not simply preventing a developmental abnormality from occurring.

They were intervening after the animals had already developed.

The treatment improved several social, communication and repetitive behaviours, although it did not correct every abnormal behaviour in every model.

That distinction is important.

This is not evidence that autism can simply be "reversed" in adulthood.

But it does suggest that at least some of the consequences of these genetic mutations remain functionally modifiable after development.

The underlying mutation remains.

The researchers are instead correcting a downstream signalling problem.

 

Then they tested human brain organoids

The researchers went one step further.

Using CRISPR gene editing, they created human cortical organoids carrying SHANK2 or SHANK3 mutations.

These human brain models showed reduced NMDA receptor function.

They then used an ASO designed to target human SLC6A20.

Again, NMDA receptor function was restored towards normal levels.

This is important because the same basic mechanism was seen in genetically modified mice and human cortical organoids.

It is still very early-stage research, but it gives the finding greater human relevance than a mouse study alone.

 

Could this identify a subgroup of autism?

This may ultimately be the most important question.

The new study does not suggest that every autistic person has insufficient NMDA receptor activity.

Autism is biologically heterogeneous.

Some people may have NMDA hypofunction.

Others may have different forms of synaptic dysfunction.

And the same individual could potentially have different abnormalities in different brain regions.

This means that simply increasing NMDA receptor activity is unlikely to be a universal treatment for autism.

The more interesting possibility is that we might eventually identify people whose autism involves NMDAR hypofunction or inadequate co-agonism.

For those people, treatments such as L-serine or more targeted approaches affecting glycine availability could potentially be more relevant.

 

Why the story is becoming more interesting

When I first wrote about AST-001, the idea that L-serine might improve autism through NMDA receptor signalling was still speculative.

The Phase II trial provided an encouraging clinical signal.

The longer-term follow-up has added evidence that previous responders can continue to show benefit.

And now a completely different intervention—SLC6A20-ASO—has restored NMDA receptor function in genetic models involving SHANK2 and SHANK3, including adult mice and human cortical organoids.

This does not prove that AST-001 works because of glycine.

It does not prove that SLC6A20-ASO will work in humans.

And it certainly does not mean that NMDA receptor modulation is the answer for everyone with autism.

But the two lines of research are beginning to tell a coherent story.

Different interventions may be approaching the same biological bottleneck from different directions.

For AST-001, the starting point is L-serine.

For SLC6A20-ASO, the starting point is glycine transport.

But the destination may be the same:

restoring NMDA receptor co-agonism when that particular signalling system is not functioning properly.

What about using D-serine?

D-serine is particularly interesting because it is one of the actual "second keys" that can activate an NMDA receptor. Unlike AST-001, which supplies L-serine, D-serine can act directly at the NMDA receptor co-agonist site. In theory, this makes D-serine a more direct way of increasing NMDA receptor activation.

However, there has been a safety concern with D-serine, particularly dose-related kidney toxicity seen in animal studies. Human studies so far have been more reassuring, but D-serine still requires appropriate safety monitoring. L-serine therefore has an important potential advantage: it can provide a more indirect and potentially safer way of influencing the same system.

This may be one reason why AST-001 uses L-serine rather than D-serine.

Why might glycine itself not work?

If glycine is the second key for the NMDA receptor, it is natural to ask: why not simply take glycine?

The problem is that the amount of glycine reaching an individual NMDA receptor is tightly regulated by transporters that remove glycine from the extracellular space.

So taking more glycine does not necessarily mean that more glycine will reach the particular synapses where NMDA receptors are under-functioning.

There is also an important complication: glycine has two very different roles in the nervous system. It can enhance NMDA receptor activity, but it also acts as an inhibitory neurotransmitter, particularly in the brainstem and spinal cord.

This means that simply increasing glycine throughout the nervous system may not be the ideal strategy.

The SLC6A20 researchers took a more targeted approach: instead of supplying large amounts of glycine, they reduced a transporter involved in glycine handling, potentially allowing more glycine to remain available where it is needed.

This does not prove that oral glycine cannot work. It simply means that the new study has not tested that question.

The important distinction is therefore:

Glycine supplementation = increase the supply

SLC6A20 inhibition = change how the brain handles the glycine already present

The latter may ultimately prove a more precise way of providing the second key to the NMDA receptor.