Why Is Clozapine So Great?

[Epistemic status: Very speculative; I don’t fully understand a lot of the studies involved]

Clozapine is an antipsychotic drug sometimes used to treat schizophrenia. Like most antipsychotics, it works by blocking dopamine receptors in the brain. Conventional wisdom among psychiatrists goes that most antipsychotics are about equally good – except for clozapine, which is better.

It’s fun to listen to psychiatrists, usually pretty quick to admit how crappy a lot of the drugs they prescribe are, wax rhapsodic about clozapine. From Joober & Boksa:

Consensus of opinion is rare in psychiatry. Even in the field of clinical trials, where experimentation is tightly controlled and regulatory bodies scrutinize the proof, controversies are frequent and difficult to resolve. One issue for which there is a widespread consensus is the unique place that clozapine occupies in the treatment of severe mental illnesses, particularly refractory schizophrenia. This molecule is distinct because of its effectiveness, numerous and sometimes mysterious pharmacologic characteristics, serious side effects and under use…

Every clinician who has prescribed clozapine can recount a few experiences of seeing patients emerge from their chaotic psychotic experience. This is one of the most rewarding experiences that a psychiatrist can have in his or her professional life, and it is among the most important strikes we have made against one of the most devastating diseases affecting mankind.

Psychiatrists wish they could give clozapine to everyone who needs an antipsychotic. They can’t, because its increased efficacy goes side-by-side with greater side effects. The best known are agranulocytosis, metabolic syndrome, seizures, myocarditis, and eosinophilia, and the list just goes on from there.

This has led a lot of people to wonder: why is clozapine so uniquely great? And can we get a version which is just as good without the side effects?

Recently there’s been increased interest in the glutamate system (especially NMDA receptors) in schizophrenia, and in glutamatergic compounds as possible treatments. Various teams have taken schizophrenic patients already on antipsychotics and added NMDA modulators, especially d-serine, glycine, and sarcosine. Meta-analyses have been guardedly positive (Tiihonen & Walbeck) or explicitly positive (Singh & Singh, Tsai & Lin). There’s a widespread hope that the next generation of antipsychotics will be glutamatergic drugs which are able to attack more symptoms than the dopaminergics we have today.

But what if that next generation were already here?

Both of the recent meta-analyses of glutamatergic augmentation of antipsychotics noted the same exception. From Singh & Singh:

When added to clozapine, none of the drugs demonstrated therapeutic potential

And from Tsai & Lin:

Patients receiving risperidone or olanzapine, but not clozapine, improved.

Why should this be? A couple of recent studies have converged on an exciting possibility: clozapine is a combination antipsychotic + NMDAergic. That is, NMDA glycine site agonists don’t add anything to clozapine because clozapine is already agonizing the glycine site.

This is the suggestion of Schwieler et al, based on their electrophysiology studies in rats. They find that clozapine causes a characteristic change in the firing rates of certain rat neurons, a change which is reversed by the glycine site antagonist kynurenic acid. They conclude (I don’t know enough to confirm) that:

The enhanced response of [ventral tegmental area] [dopaminergic] neurons to clozapine seen following lowered [kynurenic acid] is what should be expected from a partial NMDA/glycine-site agonist.

Javitt et al also work with rats, and find that clozapine “inhibited transport of both glycine and MeAIB, but not other amino acids, at concentrations associated with preferential clinical response”. Other antipsychotics do nothing of the sort. A lot of their biochemistry is a little too in-the-weeds for me, but I think what they’re saying is that clozapine increases natural extracellular glycine levels, and so really is a direct analogue of medicinal glycine administration. They conclude:

This study suggests first that System A transporters, or a subset thereof, may play a critical role in regulation of synaptic glycine levels and by extension of NMDA receptor regulation, and second that System A antagonism may contribute to the differential clinical efficacy of clozapine compared with other typical or atypical antipsychotics.

Jardemark et al find that clozapine facilitates NMDAergic neurotransmission through something called protein kinase C. My head is starting to hurt trying to keep track of all of these different chemicals, and in particular I’m not sure whether all of these people are positing different and incompatible mechanisms or if they can be unified into one big biochemical pathway. But it sure looks like a lot of people have found some kind of interesting NMDAergic effect.

This doesn’t really prove anything. For one thing, a bunch of drugs coincidentally have NMDAergic effects that have nothing to do with their mechanism of action – eg the antibiotic cycloserine treats tuberculosis and happens to be a pretty good NMDA agonist on the side. For another thing, what if all antipsychotics have NMDAergic effects? Blocking dopamine is probably going to do something or other upstream. Right now evidence seems pretty mixed on this, with one study suggesting they do (1) and a couple suggesting they don’t (1, 2, 3).

If this really was the source of clozapine’s special powers, it would be a really important big breakthrough. All through residency, I kept hearing “nobody really knows why clozapine is so great” or “it probably has something to do with its weak D2 binding affinity or something”. If we knew it was just because clozapine was an antipsychotic plus glycine, we could just give people an antipsychotic plus glycine, and avoid the agranulocytosis, metabolic syndrome, seizures, myocarditis, eosinophilia, etc.

This theory isn’t ready for prime time yet; it’s still not really proven that the NMDA agonists work for schizophrenia at all. And it probably never will be – it’s hard to patent these chemicals, meaning that it would take extreme creativity to jump the bureaucratic hurdles necessary to get them into the pharmacopoeia. Until that happens, using these chemicals will remain experimental, and replacing clozapine with them downright irresponsible.

Still, it’ll be hard to watch treatment-refractory schizophrenia patients get their agranulocytosis, metabolic syndrome, seizures, myocarditis, eosinophilia, etc without wondering whether another way is possible.

This entry was posted in Uncategorized and tagged . Bookmark the permalink.

27 Responses to Why Is Clozapine So Great?

  1. Scchm says:

    One of the explanations for a superior action of clozapine (at least against negative symptoms) is that it is a muscarinic M1 agonist unlike any other neuroleptic (most of gen1 antipsychotics are muscarinic antagonists). And as far as I remember, clozapine is also dopamine receptor D1 partial agonist, which should help with negative symptoms.
    As for glycine cite NMDA agonists, you can buy sunifiram on Amazon or any nootropics cites for about $30-40 per 5 g. It is a misnomer to call sunifiram a nootropic, but it stimulates “glycine-binding site of N-methyl-D-aspartate receptor” and “able to produce a NBQX sensitive reversal of the kynurenate-induced antagonism in the ‘‘kynurenate test’’” (see Gualteri F, Unifi nootropics from the lab to the web…)

  2. avturchin says:

    I have great experience with memantine, which also NDMA antagonist, but after it wears out, it increases the sensitivity of NDMA receptors, so works as agonist during hangover period (alcohol do it too).

    My protocol, after which I declared my depression controllable, is to take 2.5 mg of the memantine (it is small dose) every 3 days. Next day it starts to wear off – it has long half life of like 80 hours, so it is a slow process, and I feel better. I don’t do memantine constantly, as I now don’t feel depressed, but if some depressive feeling returns, I do it again.

    • Scchm says:

      A very perceptive comment. Note that taking memantine continuously does not work as shown in double blind controlled trials of memantine augmentation of antidepressants. I believe that ketamine may work the same way, during the rebound after strong NMDAR inhibition. That is why you need a once-a-week injections of ketamine.

  3. John D. Bell says:

    Scott – since sarcosine and glycine would be considered “dietary supplements” and not drugs, what would be the practical and/or medical ethical considerations for treating patients with conventional (non-clozapine) antipsychotics + sarcosine?

    • vV_Vv says:

      I presume that the brain level of anything that naturally occurs in food is tightly controlled by feedback loops, therefore you would need very high doses in order to perturb it.

  4. jobleonard says:

    a change which is reversed by the glycine site antagonist kynurenic acid.

    I’m have no background in this field so maybe this is irrelevant, but I attended a lecture last year about red muscle and its effects producing proteins that convert kynurenine (produced by stress) into kynurenic acid, protecting the brain from stress-related illnesses or triggers for illnesses. Let’s see if I can find an online source (on mobile right now)… OK, here is a press release from Karolinska from a few years ago:

    The researchers discovered that mice with higher levels of PGC-1a1 in muscle also had higher levels of enzymes called KAT. KATs convert a substance formed during stress (kynurenine) into kynurenic acid, a substance that is not able to pass from the blood to the brain. The exact function of kynurenine is not known, but high levels of kynurenine can be measured in patients with mental illness.

    In the current study, the researchers demonstrated that when normal mice were given kynurenine, they displayed depressive behaviour, while mice with increased levels of PGC-1a1 in muscle were not affected. In fact, these animals never show elevated kynurenine levels in their blood since the KAT enzymes in their well-trained muscles quickly convert it to kynurenic acid, resulting in a protective mechanism.

    “It’s possible that this work opens up a new pharmacological principle in the treatment of depression, where attempts could be made to influence skeletal muscle function instead of targeting the brain directly. Skeletal muscle appears to have a detoxification effect that, when activated, can protect the brain from insults and related mental illness,” says Jorge Ruas.

    http://ki.se/en/news/how-physical-exercise-protects-the-brain-from-stress-induced-depression

    They also have a follow-up paper where they attempt to figure out what red muscle uses kynurenic acid for, which is what the lecture I attended was about. Something to do with converting white muscle into “pink” muscle if I recall correctly.

    Anyway, thought this might be interesting. Now if you’ll excuse me I am off to do some mental health protecting aerobic exercise 😉

  5. Dog says:

    Sarcosine could be the most promising of the 3 NMDA modulators, in that it may work similarly to glycine but at a much more reasonable dosage (2g per day vs. 60g per day or more), and maybe more reliably then d-serine. There are also a lot of anecdotal reports of glycine etc. being effective for depersonalization disorder. It would be amazing if this line of research pans out in both cases, since DPD, like the negative symptoms of schizophrenia, has no great treatment options.

  6. userfriendlyyy says:

    The problem with rodent studies on antipsychotics is that you can’t make a genetically schizophrenic rat and you can’t test for efficacy of treatment even if you could make one. All those studies rely on the quaint notion that Ketamine and PCP make people and rats similar to schizophrenic people and then work backwards trying to prevent damage from massively overdosing on one of those dissociatives. Anyone who has ever done drugs can tell you that tripping is a lot closer to schizophrenia than a k-hole. And now they are developing k into the world’s first instant antidepressant. They just figured out that it is actually a metabolite of k working on the AMPA receptor that does the amazing job as an antidepressant and that ketamine’s NMDA bonding is probably what creates the dissociative effect. But if you are really interested in how Clozapine works I’d check out this paper. http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0177036

  7. Sam Reuben says:

    Scott, I always love hearing about your profession and the forays into the human mind. It is, if you’ll pardon me, excellent food for thought.

    Something I do wonder about, though, is along the same lines as rahien.din’s conclusion: aren’t a lot of these biochemical attempts to explain certain brain functions running a serious risk of oversimplifying extraordinarily complex interactions? This is more-or-less the same problem you’ve noted time and time again with the high presence of NMDA-receptors in key studies and functions. NMDA-receptor-yes and NMDA-receptor-no can’t possibly explain the wide range of human personalities all on their own, meaning that a naive account of NMDA-receptor-importance is going to fail to get at what’s really going on in some of these diseases, but at the same time the evidence shows that those gosh-darn receptors have something serious to do with it all.

    My natural inclination is to wonder about the kinds of “striping” patterns we tend to see in (especially embryo-fetus) growth. I’m sure you’re familiar with the rough outline of how it works: certain signalling chemicals start to be built at one part of the li’l cell-ball, they diffuse through the cell-ball, and different functions manifest at different concentrations of the signalling chemicals. For each different point where some significantly different function manifests, a new “stripe” of the organism takes shape: for instance, the arm-stripe and the leg-stripe. There’s more than just one signalling chemical, and the real process is more complicated than that, but I think this might be applicable to the NMDA-receptor problems that you’re working with.

    Suppose that during brain growth, different parts of the brain are conditioned for different things through striping (I believe this is how it works, but I don’t want to pretend to be definitive on anything I’m no expert with). In particular, the conditioning makes different clusters of neurons in different parts of the brain respond to NMDA-receptor-functioning in different ways. The various molecules which bind to NMDA-receptors are released in the brain in different locations according to different stimuli, causing different reactions in sites at varying distances from the source, which in turn accounts for different behavior. This could account for a lot of how the brain functions, in different amounts.

    So this would explain why it’s so hard to pin down what NMDA-receptors really do: they do different things in different parts of the brain. It also explains why certain agonists have massive side-effects, I think. What’s wrong in the person to begin with is that a specific area of the brain is either consistently getting the wrong concentrations or else isn’t reacting properly to the correct concentrations. When a powerful drug of whatever sort comes in, it’s flowing in through the blood and mixing through the brain evenly, rather than appearing at a particular location. This means that while the one thing which was out of whack is now getting the right concentrations, everything else is getting weird and bad concentrations, which leads to major side-effects.

    I’m quite interested to hear what you have to say about this interpretation. What I’m expecting is that it’s either a popular idea or that it’s been completely debunked, or perhaps that it’s something that’s hard to prove or disprove at present time. Either way, it’s far better for learning to ask an expert than speculate on one’s own. Always eager to gain more insight!

    • Scott Alexander says:

      I agree all of this is fiendishly complex, but we’ve previously been able to identify some broad chemical-symptom-area links; for example, conventional antipsychotics work by antagonizing mesolimbic dopamine.

  8. Dr Martina Feyzrakhmanova says:

    I am aware of some unpublished research about clozapine hindering platelet aggregation having a positive effect on CVS risk even despite the associated metabolic syndrome. Wonder what will come of it.

    Have you ever used glycine medically?

  9. Sniffnoy says:

    Question: How do we know antipsychotic + glycine wouldn’t just have the same side effects as clozapine?

    • Mediocrates says:

      Clozapine is one of the classical examples of a “dirty” drug, that is, one that promiscuously binds a bunch of molecular targets instead of specifically zapping exactly one. The latter case is generally considered the ideal for modern drug development, and biopharmas spend a ton of resources beating laserlike specificity into their molecules, but many of the dirtier drugs from the old days are probably effective precisely because they engage multiple targets; the body has a lot of redundancy built in, and it tends to be hard to get a major effect by tweaking a single knob.

      Wikipedia lists 17 different receptors or isoforms that clozapine binds pretty tightly (<100 nM Ki), and even more weaker targets that you'd hit at higher doses. Based on that, I wouldn't be surprised if it's hitting a bunch of other stuff – kinases, transporters – that we don't know about. Maybe some of those targets are key to the efficacy, but if it was just due to the combination of dopamine antagonism plus increased glycine, all those extra targets would just be buying you toxic side effects. In that case, a combo of cleaner molecules would probably be superior.

      Also, it looks like people think that the eosinophilia and agranulocytosis are due to hypersensitivity (basically, an allergic reaction to the clozapine molecule). If so that's probably just a random quirk of the structure, and another drug could hit the same target(s) without provoking the reaction.

      • Sniffnoy says:

        I see, thanks!

      • Speaker To Animals says:

        I prefer the term ‘promiscuous’ to dirty.

      • vV_Vv says:

        Clozapine is one of the classical examples of a “dirty” drug, that is, one that promiscuously binds a bunch of molecular targets instead of specifically zapping exactly one. The latter case is generally considered the ideal for modern drug development, and biopharmas spend a ton of resources beating laserlike specificity into their molecules, but many of the dirtier drugs from the old days are probably effective precisely because they engage multiple targets; the body has a lot of redundancy built in, and it tends to be hard to get a major effect by tweaking a single knob.

        Interesting. According to the “omnigenic” hypothesis, almost every gene affects almost everything, which if I understand correctly implies that almost every enzyme promiscuously binds with almost everything. Of course, since everything evolves together, negative effects from these interactions are usually avoided.

        I can see why biopharmas can’t just rely on natural selection to get rid of genotypes that react badly to their drugs, but is it possible that their focus on specifically-binding drugs runs the risk of missing out more powerful molecules?

        • Mediocrates says:

          I suspect that some (maybe weaker) form of the omnigenic hypothesis is probably true; basically every interesting trait is highly heritable, but 9.5 times out of 10 gene association studies on that trait find a huge list of alleles of small effect, which added together only explain a fraction of the variation. That seems consistent with most of the variance coming from random little dings and dents in genes that aren’t directly related to the trait in question but just make the cell/tissue work a little worse or better in general.

          Note though that this wouldn’t necessarily imply that enzymes directly bind to a bunch of stuff. Most of the interactions are probably indirect; like, say, if you tweaked an enzyme that regulated a calcium pump, which slightly changes the calcium concentration in the cell, which slows the rate at which you can dump out some waste product that’s co-transported with calcium ions, which backs up a metabolic pathway, and on and ever on.

          You’re right to worry that the focus on molecules that specifically hit a single known target might bias you away from more powerful drugs. This is both because (i) blocking one node in a huge, redundant hairball of a network with a bunch of compensatory mechanisms may not do much, and (ii) we generally have a pretty poor idea of what targets to go after in the first place. Back in the “golden age” of pharma (roughly the middle 20th century) it was much more common to do phenotypic screening: take some animal model of your disease and just throw a bunch of compounds at it semi-randomly until you find something that works, then develop it into a drug and worry about figuring out the mechanism later (maybe decades later). That’s how a lot of the classical cancer chemo drugs came about, except in that case the “animal model” was actually child leukemia patients at St. Jude.

          As we learned more about human biology, we became enamored of the idea that we could cure diseases more quickly and safely by first figuring them out at a molecular level and then rationally designing drugs based on that information. This paradigm has had plenty of home runs, but isn’t without its discontents. And plenty of diseases lack either predictive animal models or any real mechanistic understanding, which is why Alzheimer’s clinical trials have a 99.9% failure rate.

    • Scott Alexander says:

      Glycine is an amino acid. You take at least a little glycine every time you eat a steak. I think side effects are usually minimal even at high doses.

  10. rahien.din says:

    Just goes to show that we know next-to-nothing about how the brain actually works, or how to elegantly interface with it.

    With respect to its substrate, neuropsychiatric pharmacology is about as sophisticated as slamming an open palm onto your computer keyboard. Or changing every instance of C# in a symphony to B. Even when it works, we don’t really have a good idea why.

    • Alethenous says:

      We know a huge amount about how the brain works. It’s just that there’s an even more huge amount left to learn.

      • Speaker To Animals says:

        Yes, I’d say ‘Have a lot more to learn’ is a better statement of our knowledge than ‘know next to nothing’.

        • rahien.din says:

          I’m not being flippant. We can help a lot of people. We can predict a fair number of things. But most of our knowledge is along the lines of “If we create a big lesion in [X region], then [Y thing] happens” or “If we turn every dial in the control room 30 degrees clockwise, [Z thing] happens.”

          That’s not knowing much more than nothing. We’ve simply operationalized a system of interlocking lesions.

          Case in point : we’ve only just figured out that interneurons are important for seizures, but we don’t know how. We don’t really know how a seizure starts, or stops. It’s only in the past few years that we’ve recognized that seizures are network phenomena, but we’re not sure of much beyond that. We can’t explain why nearly every seizure treatment strategy – regardless of mechanism – is about 60% effective for the appropriately-selected patient. We can’t explain why two drugs that disable the same target in different ways have the same spectrum of action, but a third drug acting on the same target has a totally different spectrum of action.

          And besides, “It’s not that we know so little, it’s that there is so much left to learn!” rings hollow to me. It strikes me as a way of congratulating ourselves for having expended a great deal of effort, without sufficiently acknowledging our current deficient state of knowledge.