Skip to Content

It Isn’t Ketamine

Ketamine is definitely an interesting compound, but not always in a good way. It’s been around for decades, and its main medical use is as an adjunct to anaesthesia. But it’s also long been a recreational drug/drug of abuse  (the line between those categories is famously thin, even for a single individual). It has strong and fast-acting CNS effects, including sedation, hallucinations, euphoria, and its exact opposite. A pretty fair number of recreational users over the years have gotten a lot less (or maybe a lot more) recreation than they signed up for; people get killed while on the stuff every year, often via inadvertent suicide while deranged. But for some years, the compound has been reported as a very unusual antidepressant, a property that’s brought a lot of attention, since the field is notoriously difficult to show efficacy in. The duration of the effect can be up to several days after a single dose, and the fact that there’s an effect at all is very much worth noting.

The problem is, even after all these years, no one’s completely sure how ketamine works, in any of its capacities. It definitely hits the NMDA receptor, but that’s only of limited use, since we don’t really understand the NMDA receptor well enough, either. Still, that almost certainly accounts for the anaesthetic action, but the compound also has effects on opioid receptors, dopamine reuptake, dopamine receptors, muscarinic receptors, sodium and calcium channels “and much, much more!” as the ads would say were it to be advertised on cheap cable stations. No wonder it’s hard to figure out what’s going on: this is one of those compounds – and there are many – that approaches the complicated multikeyboard instrument of the brain by giving it a Jerry Lee Lewis style bashing with both hands and a foot.

A new paper, though, may shed some light. A large team from the NIH and the University of Maryland has been working on evidence that the antidepressant action might be due to some sort of ketamine metabolite rather than the parent compound. This seems to be the case: the main human metabolite ((2S,6S;2R,6R)-hydroxynorketamine, or HNK, which results from loss of an N-methyl group and hydroxylation next to the ketone) has no anaesthetic activity at all, and no NMDA activity at all, but turns out to be quite active in a number of animal models that are broadly predictive of antidepressant activity. In a rather elegant experiment, the authors deuterated ketamine at the methylene next to its carbonyl group, and showed that this compound (as expected) displayed the same NMDA activity as the parent compound – but that it was far slower at producing the metabolite, and had no activity in the antidepressant screens.

What HNK does seem to hit is the AMPA receptor system, which had been already been suspected as a player in the ketamine story. But that evidence was being interpreted as “something that NMDA antagonists do that has an effect on the AMPA system”, which latest work clarifies quite a bit: NMDA seems to be out of the picture entirely now. HNK does not seem to cause disorientation in the animals or locomotor problems, and may well be free of the alarming ketamine effects (the headfirst-through-a-glass-door stuff) that have limited its use as an antidepressant. Interestingly, the compound’s effects are seen well after its concentrations in the brain go down below detection. This isn’t unheard of for CNS agents; things can disrupt signaling in ways that aren’t obvious. But all this makes a human trial of HNK seem like a very interesting project, and validates the interest in AMPA (here’s a recent report) as an antidepressant mechanism.

40 comments on “It Isn’t Ketamine”

  1. Barry says:

    Identifying the molecular target of HNK is suddenly the hottest topic in CNS in years. If it lights up an assay that’s already up and running, that’ll be fast. If it requires characterizing a new target (potentially more exciting) it could take a lot of work. Not that one can’t do med. chem. without knowing the molecular target, but that’s harder.

    1. anonao says:

      One can use phenotypic screening for this purpose.
      From this paper, if one can get something like the metabolite effect in the models with a different drugs they could go that way for optimization. (or use different model, even zebrafish are used for antidepressant model and can provide higher throughput)

      1. Barry says:

        yes, in the absence of a characterized molecular target, one can still do drug discovery. Plenty of drugs (Plavix, Omeprazole…) require metabolic activation to bind/modulate their actual target. But animal models of CNS disorders require a bit of faith. Just because they respond to known drugs doesn’t mean that they’re driven by the same (primary) target or that that’s the relevant one.

      2. Anon says:

        Indeed, one could do phenotypic screening and dispense with all this target-based hubris bollocks altogether, because time and time again we learn (or rather don’t learn) that what we think we knew is wrong.

      3. HTSguy says:

        Phenotypic screens work best when then the model is highly predictive of the desired activity in humans. I don’t think any current models of CNS (animal, iPSC, etc) fit that description. In the absence of such a model, “phenotypic screening” is just the buzzword du jour.

        1. Mark Thorson says:

          I looked at that Chinese article — interesting that there’s an antidepressant in cordecyps that modulates AMPA. I guess those ancient Chinese herbalists really knew a thing or two. It mentioned a rodent model of depression I was unfamiliar with, the Tail Suspension Test. Looking around for some information on it, I found this good paper:

          Interesting. Even within a strain of rodent, their performance on the TST and FST will vary by rodent vendor. Whoa, deep voodoo!

        2. Heisenberg says:

          Respectfully disagree, HTSguy. It is exactly because we look to animal models to be predictive of human behavior (which is an impossible standard in most areas of CNS) – and because we get so hung up on putative target (which is generally proven wrong, or to be only part of the story) – that we’ve wasted so much time and $ in this area. Phenotypic screening is rather most useful when done in vivo and when neural circuitry is largely conserved, and when drugs with similar MOAs read out with similar results. Like zebrafish, as pointed out in an earlier reply.

          1. M says:

            I’m thoroughly confused by this. If we don’t think animal models are predictive what’s the point of screening in them, unless you’re actually trying to help out depressed zebra fish?

            Isn’t the argument that “neural circuitry is conserved” basically an argument that it is predictive, and what does “similar MOAs” mean if not that it may break down if I’m totally agnostic about “target” (even i I don’t know the exact target or its polypharmacology or whatever.)

          2. M reply says:

            M – Predictive of drug response, yes. Predictive of behavior, no. Whether flies or worms or fish or rodents, if similar drugs (say antidepressants) act similarly in a species, and we can quantify the similarity, then we can use the model to screen for compounds with similar activity, without regard to MOA. But not to predict that an animal behaves like a human.

          3. Dr CNS says:


            It is not that simple… I wish it were.
            Since we don’t know exactly what many CNS diseases are, it is not possible to create a model that replicates it entirely in a preclinical species. For novel mechanisms of action, it is impossible to predict with any useful degree of certainty whether the drug will do anything good to patients. So, instead, we try to study different mechanistic aspects to make sure the compound is having the functional effects we THINK it should have.

            In fact, we know that what we call a CNS disease is actually the manifestation of more than one dysregulation in the body… In other words, there are many independent ways to generate the same CNS disease.

            Even humans are not great models to study CNS drugs… Example: for depression, the typical response to a clinically effective drug is around 30%… So, it does not work on more patients than it does…

      4. Mark Thorson says:

        Do you know what the zebrafish model is called? I know the one with rodents is the Forced Swimming Test, but presumably that one doesn’t work with fish.

        1. HTSguy says:

          I believe that both the FST and the TST are examples of “learned helplessness”:

          I suspect that the zebrafish model is just another version of this.

          1. Mark Thorson says:

            Maybe that explains the intrastrain differences between rodent vendors — some rodents are raised in more comfortable situations and have more time with Mom, while other rodents are raised in colder, more austere environments. So, some feel more abused by being hung by their tail or waterboarded.

  2. Peter Barfuss says:

    Amusingly, the first thing I did was search for interactions between AMPA receptor activation, BDNF transcription, and signalling activity at TrkB. Turns out, yep: – “Positive AMPA receptor modulation rapidly stimulates BDNF release and increases dendritic mRNA translation.” – “Positive modulation of AMPA receptors increases neurotrophin expression by hippocampal and cortical neurons.”
    etc etc etc. There’s a bunch. There’s also interesting results showing *chronic* elevations in BDNF after short but consistent acute treatment with strong AMPA PAMs (see for instance).

    And an effect that is dependent on increasing BDNF mRNA transcription would be consistent with the delay seen in effects (with the effects actually starting only after the compound is actually undetectable), and also (to the best of my knowledge) seems to be consistent with SSRI effects when they are actually effacious for someone, and also exercise-induced improvements in mood.

    Which makes me again wonder if there have been any clinical trials *in humans* of small molecule TrkB agonists. I’d not be too surprised if they’re less effacious than ketamine or even not effacious at all, but I really want to know *if* they are, since to me more & more it seems the common player in depression seems to be a lack of neurotrophin signalling, an I’m curious if there’s any merit at all to that hypothesis in vivo.

  3. That guy with that thing says:

    Nice metaphor with Killer on the keyboards!

  4. PF says:

    Give it a few years. Carlos said two years ago that HNK was responsible for the antidepressant effects of ketamine ( but attributed it to nicotine alpha7. So what now makes it Nature worthy this time around? I can’t access the full paper in Nature: does he give alpha7 a mention?

    1. TX raven says:

      Reference #12 in the paper is the only mention to such alpha-7 effects.

      The story looks robust, but that does not mean it is complete.
      Indeed, the effects in vivo and in electrophys also outlast compound washout… so, there is something else here. Any thoughts?

  5. Rapid antidep man says:

    Also from the same group: No significant correlation of this metabolite with treatment response in MDD patients (17 responders from 45 MDD patients tested) (
    Wouldn’t some sort of correlation be expected if this metabolite was the origin of ketamines AD action?

    1. TX raven says:

      @ Rapid antidep man,
      The relationship between efficacy and metabolite concentrations would depend on the mechanism of action…
      Anything is possible in CNS drug discovery…

      1. Diver dude says:

        Also, you will be measuring peripheral concentrations of the drug. Who knows what’s happening at the effect site (where and whatever that is). I spent years working on AMPA receptors when I was working in epilepsy. Nasty tendency to hallucinations and convulsions if you hit them wrong. I suspect HNK is doing something else as well.

        1. tangent says:

          > I suspect HNK is doing something else as well.

          I hope the paper showed non-activity at the kainate receptor? Who knows, maybe kainic acid is an undiscovered antidepressant in low doses,
          but I wouldn’t be excited about administering a close relative and crossing fingers that nobody develops amnesic shellfish poisoning…

  6. Barry says:

    a case in point of drug discovery without a known molecular target is Plavix (clopidigrel). The operative assay was in rodents (in vivo). That’s the expensive end of what we mean when we say “phenotypic screen”; who would launch such an effort today? And still, one would have to trust a rodent model of depression.

    “Clopidogrel…could not be discovered today through an in vitro high throughput screening because they are prodrugs, which must be transformed in the body into an active metabolite. The active metabolite is very unstable and cannot be obtained by chemical synthesis or stored. Moreover, its structure cannot be predicted by rational drug design”

    1. Dr CNS says:

      Barry… (or anyone else)
      Can you clarify something for me, please?

      If you don’t know the target or the mechanism of efficacy of your phenotypic drug, how do you know it will translate to human? That is, unless you look for the target and link the preclinical biology to human disease… or have a human phenotypic screen that you can do preclinically? (e.g., coagulation time in human blood)

      Thank you.

      1. HTSguy says:

        Patient-derived iPSC neurons?

      2. Barry says:

        there are various levels of confidence. With e.g. clopidigrel/Plavix we’re satisfied that the clotting cascade in rodents is like the clotting cascades in humans, so what works in a rodent is very very likely to translate to humans, even if we don’t know the molecular site of action. So a blockbuster drug was discovered/developed without need to know the target. On the other end of the spectrum I’d put all the CNS indications. Correlation of the animal models to human disease ranges from weak to terrible (but you’ve got to have something to show the FDA before you go into humans)

      3. matt says:

        In this case, the serendipitously discovered response has already been seen in humans. The only question is whether the metabolite is responsible for the response, and in that case it seems reasonable to argue that the metabolite’s effects on the mouse model of depression might correspond to human depression effects already seen.

        Just to play devil’s advocate (hops up and straps in to the deep-sea-fishing chair), how much of this hand-wringing about the target is synonymous with pharma med chems whining that academia hasn’t provided the free early stage work to which they are accustomed? Can’t get good help these days, can you? Especially when you aren’t paying for it. Haha.

        Also, is it strictly necessary to treat HNK as just a lead compound and get all fancy fiddlin’ around with it? (Other than for patent purposes, which Martin Shkrelli would tell you is necessary for moneymaking only due to a lack of imagination.) Couldn’t you go straight to the clinic with HNK, since you’ve already been in the clinic dosing with what seems to be the pro-drug?

        My sister tells me veterinarians commonly use ketamine as an anesthetic; I imagine there is a vast amount of information on ketamine dosing, safety, and toxicology, both in people and in an enormous range of animal species. Shouldn’t that give you a reasonable starting point to work out a NAEL for the metabolite, and an upper limit for the dose required? (Of course you’d need to add the “O” back in with confirmatory animal tests on actual HNK.)

        I would think this is a good deal safer than the LSD dude testing his own product out on himself for the first time.

        1. Anonymous says:

          It’s not just vets who use it. Ketamine is commonly used as a sedative for children in the emergency room when undergoing painful procedures like setting broken bones, stitching up wounds and so on, since it does not cause respiratory depression, and patients are able to protect their own airway. On the other hand, with young children you’re not worried about getting them “hooked” either.

          For more information, see:

  7. Vince says:

    From my perspective this leaves GLYX-13 in an interesting position. It’s certainly NMDAR active and specific, with efficacy in early depression models.

    Perhaps Derek’s comments cutting the NMDAR out of the Depression story entirely is a little premature; especially for these highly interconnected systems?!

    1. Chris says:

      Ageed, it would be premature to through out the NMDAR from the story at this point in time. In addition to exhibiting pre-clinical efficacy for depression, GLYX-13 has also demonstrated robust POC in Phase II trials with similar outcomes reported for ketamine. Doesn’t partial agonist/antagonist activity at the NMDAR (whether at the GluNR1 or GluNR2 subunit) impact AMPAR recruitment and activity? GLYX-13 also stimulates BDNF production. More to the story and this recent paper is quite fascinating.

      1. Chris says:

        throw out

  8. Paul D. says:

    OT: avasimibe returns as a possible treatment for pancreatic cancer?

  9. steve says:

    Two questions: First, how many good drugs have been abandoned because pharma didn’t know the target? (Not that it’s entirely a bad thing, I started a company based on one such case. They had a great drug, never pursued it because the target hadn’t been discovered; it was simple enough to invent around their old patents now that the target is known). Second, how many drugs did pharma move ahead with only to find out what they thought was the target wasn’t? I believe in the old adage in vivo veritas. Forget trying to pin down the target when you have a drug that works. You’re likely to be wrong anyone, biology isn’t as simple as ye old lock and key.

  10. steve says:

    I’d like to add one more comment. Anyone who’s had a loved one who has suffered from depression knows it’s a horrible disease. Don’t get lost in all the minutiae of ligand binding; if HNK works and is safe, get it to patients and let a generation of post-docs figure out the mechanism. Same with pain. Chronic pain is a horrible unmet need. If there’s a ketamine derivative that can be used safely and effectively, don’t get caught up in co-crystallizing it with a receptor. Just do the damn clinical studies and figure it out later. This is where pharma fails and biotech succeeds – as long as you have evidence that it’s safe and effective, get it to the patient as quickly as possible. There are millions of people who are suffering.

    1. Parmuas says:

      Totally agree, this was the way it was done in the past. Olanzapine was discovered by phenotypic screening, although it is off patent it still makes significant revenue for Lilly compared to their supposedly targeted approaches. The need to have a mechanistic-link has kill innovative discovery over the last decade! Drugs are rarely discovered on excel spreadsheets or in PowerPoint presentations.

  11. Me says:

    ‘Pharma’ deserves a good kicking from time to time no doubt. But might I remind all that any discovery program does have an in vivo translational model component at the end. In effect they are all phenotypic ultimately, because whatever it does at XYZ receptor, it will live and die by it’s effect and PD in the animal models of disease.

  12. Morten G says:

    Am I an idiot? That doesn’t look very stable to me…

    1. Me says:

      Maybe….secondary alcohol and tertiary amine ain’t as reactive as you’d think – plenty of them knocking around drug-space.

    2. Barry says:

      Even if it’s ultimately demonstrated that HNK is the bioactive species, that doesn’t imply that it is or could be a drug. Consider the active species generated on metabolism of e.g. Plavix. The gamma-mercapto acrylate is necessary for platelet deactivation, but it’s not drug-like.
      It’s already reported that anti-depressant activity persists after HNK is no longer detectable in the periphery, so its stability need not be a show-stopper. But it seems likely to me that HNK will be a lead from which med. chem. starts rather than a drug in its own right.

      1. Morten G says:


        PS “Me”: secondary carbon, tertiary carbon. I think what threw me was the reducing sugar-looking bit.

  13. James says:

    Whenever I see ketamine mentioned I’m reminded of that drugs dealer I met in Rehab who was hiding out in there on the presumption that it would be the last place that his former colleagues would look for him. He was hiding out because he’d ripped them off for about £20,000 worth of said drug, and understandably they were a little annoyed with him. The thing is, it was an up market rehab, and it must have cost him a fairly large percentage of his profit margin. Doesn’t make much sense, but maybe he’d been sampling a bit to much of his own products …

Comments are closed.