Q&A With Josh Roffman. Questions by Hakon Heimer and Pat McCaffrey.
Q: Where does COMT stand, in terms of schizophrenia, now?
A: COMT, and in particular the Val/Met polymorphism, has been a really attractive target in schizophrenia research for a long time because we know it plays such an important role in prefrontal dopamine signaling, which many lines of evidence point to being abnormally low or otherwise impaired in schizophrenia. It’s fairly conclusive at this point that the Val allele by itself does not seem to be a risk factor for schizophrenia. Even in studies of working memory in patients and healthy controls—where it might be expected to have a stronger signal just because working memory is more of an endophenotype as opposed to a diagnosis, and might be closer to the level of the gene—regardless, the evidence there has also been inconsistent and suggests weak if any effects of this one polymorphism on working memory performance. Now, very important caveats: first of all, this does not mean that COMT doesn’t play a critically important role in working memory performance, and possibly for risk of schizophrenia, because this is just one polymorphism within the gene. And we now know that there are several polymorphisms in the gene that can be functional and that can affect dopamine signaling either by virtue of the enzyme activity or even the level of the COMT expression. And it remains to be seen whether some of these other variants in COMT, either by themselves or in aggregate, might end up being more of a robust risk factor for either schizophrenia or working memory impairment. However, even that being said, one of the most remarkable findings in imaging genetics, or you can say even neuroimaging in general, is that this one polymorphism has been associated with absolutely consistent effects on brain activation, where the Val allele under normal circumstances is associated with inefficient patterns of prefrontal activation. This has been seen in healthy subjects, in schizophrenia patients, and in siblings of schizophrenia patients. So it really suggests that on the level of brain biology this polymorphism is really doing something important with
prefrontal activation during working memory.
Q: Can you define "inefficient"?
A: That’s a good question, and it’s actually important because it speaks to the fact that most of these COMT imaging studies use the N-back test, which is a working memory test I’m not sure you’re familiar with. And what has been seen fairly consistently is for the same level of performance—meaning accuracy on the task—individuals who have the Val allele need to activate more of the prefrontal cortex. So that’s what has led to this inefficiency model of prefrontal activation.
Now this idea differs, and in some very important ways, from the task that we use in our study, which is another working memory test, but what we were looking at was the recruitment of the prefrontal cortex in response to increasing working memory loads. And the reason increased activity as opposed to decreased activity at the same level of performance is considered beneficial in our study is that previous studies by Dara Manoach and others showed that in schizophrenia patients, those who are able to activate more dorsolateral prefrontal cortex, are able to perform better on a test. Overall, this points to the importance of considering
prefrontal activation within the context of task that is being used in the study.
Q: You said that the Val polymorphism has effects on brain activation. How consistently has that been found in both schizophrenia and normal subjects?
A: Pretty much every study has shown the same thing. Exceptions have been under circumstances where other factors may be influencing prefrontal dopamine signaling. As we know from basic studies of primates and other human studies, basically the relationship between prefrontal dopamine signaling and how well the prefrontal cortex is functioning is not linear: it’s shaped like an inverted U.
So in general, schizophrenia patients are more on the left low dopamine side of the inverted U, whereas healthy controls are well to the apex, at that point of optimal function. However—and this is a very nice paper from Mattay and colleagues in Danny Weinberger's group a couple of years ago—if you give healthy subjects amphetamine, which increases availability of dopamine, what happens is that the Val allele carriers or Val/Val subjects improve, whereas Met/Met individuals get worse. And the idea behind that is that with the amphetamine, the Met/Met subjects get shifted to the right side of the inverted U curve, and therefore begin to show less optimal patterns of reactivation.
But that finding is not really inconsistent with the other COMT findings; it actually demonstrates the validity, at least the face validity of COMT findings with respect to this known inverted U model. So really every imaging paper that I can think of that’s used the N-back with COMT has demonstrated the same things, which are findings which are absolutely consistent with the inverted U. And this is in contrast to those behavioral studies and to these association studies that have been inconsistent. The best explanation that we have at this point is that with brain imaging, we're measuring something that is more closely downstream to the level of a gene, so its signal is essentially stronger. And by looking at either performance on a task or things like schizophrenia diagnosis, they may be so far downstream of COMT gene effects that we’re not really getting a consistent signal.
Q: Okay. Where does your current work tie into this?
A: We became interested in MTHFR initially not directly because of its potential role in dopamine but because we know that the MTHFR enzyme plays a role in folate metabolism, which has also been implicated as being abnormal in schizophrenia. The one particular finding that led to this whole line of research in our group was a paper by Don Goff, a couple of years ago in the American Journal of Psychiatry, where he found, without looking at any genotypes, that there was a relationship in schizophrenia patients between serum folate levels and negative symptoms. Patients who had the lower serum folate levels had worse negative symptoms. And given the central role of the MTHFR gene in folate metabolism, the first question that we asked was whether this C677T polymorphism might be playing a role in this overall relationship between folate and negative symptoms. The hypothesis was that patients who had the T version of MTHFR, which is a functional variant resulting in reduced activity in MTHFR, would show more pronounced negative symptoms, and that’s actually what we found. In addition, we had serum folate levels available on a subset of those patients, and what we found was that among T/T patients, those who had low serum folate levels actually had the worst negative symptoms, whereas those who had higher folate levels had negative symptoms that weren’t any worse than any of the other genotype groups. Which makes sense because folate basically provides the substrate for the MTHFR reaction. So it’s almost like you have a dysfunctional enzyme, but if you’re able to overwhelm it by providing more nutritional substrate, then you can ameliorate the downstream consequences of the gene. So our initial findings were around the T version of MTHFR and negative symptoms. We also found very similar effects looking at executive dysfunction, where, again, the T allele was disadvantageous. And all that being said, we were really unclear as to what the mechanism might be on a cellular or biochemical level. How do you go from folate metabolism and this particular gene to something as far downstream as symptom clusters in schizophrenia? So given what we know relating these kinds of symptoms to low prefrontal dopamine signaling, we wondered whether the MTHFR gene might ultimately be influencing prefrontal dopamine. The first probe of that was to see whether there was an interaction between the MTHFR genotype and the COMT Val/Met genotype, just on the level of behavior. And what we found in the study published in the American Journal of Medical Genetics earlier this year was that, indeed, patients who had the Val version of COMT, meaning that they had likely lower prefrontal lobe dopamine signaling, and who also had the T version of MTHFR, had substantially worse performance on the Wisconsin Card Sort Test compared to all of the other compound genotype groups. And in fact, the interaction was more than additive, suggesting that there might be some epistasis.
Our imaging findings from this PNAS paper back that idea up completely, in that patients who had the Val version of COMT and the T version of MTHFR showed the least advantageous pattern of prefrontal recruitment. And at the same time, looking at healthy subjects—and this is really the first time we had looked at MTHFR effects in healthy subjects either with mirror imaging or with behavioral testing—we found basically exactly the opposite, which is that there was very little effect of MTHFR genotype in healthy subjects who carried the Val version of COMT. But in Met/Met healthy subjects, who had higher prefrontal dopamine signaling by virtue of COMT genotype, there was an effect and actually it was the C version, the generally good version of MTHFR, that was disadvantageous. That finding reminded us very much of Mattay's finding with amphetamine and COMT, and again, suggests to us that ultimately these two genes may be jointly influencing prefrontal dopamine signaling, in a manner consistent with that inverted U curve.
Q: So would you view this as more evidence that the inverted U-shaped curve is a real thing, or do you think the evidence for that was already so strong it’s firmly established?
A: I think that’s pretty strongly established at this point. Now that being said, you know one significant limitation of using imaging studies to validate the inverted U curve is that we’re not measuring dopamine signaling—we’re measuring a BOLD signal. What we really need to do more of is have more direct measurements of prefrontal dopamine signaling, and now we can really begin to do this with PET. There was a paper in Molecular Psychiatry earlier this year by Anissa Abi-Dargham, and she was looking at just the effect of COMT genotype on D1 receptor binding profiles in the prefrontal cortex, and found, exactly as predicted, individuals who had the Val version of COMT—the supposedly low dopamine version—had increased D1 receptor binding, consistent with a compensatory upregulation. So those PET findings I think for the first time in humans really back up the COMT prefrontal cortex relationship, with a good, more direct marker on dopamine signaling.
Q: What are the goals of this line of research?
A: There are a couple of different directions that we’re going to take it, and it’s really "bi"-translational research at this point, because we’re looking not only at whether we can clarify the basic mechanism for MTHFR and this interaction with COMT, but also whether there are any direct implications on treatment. One thing that we’re looking at now, from a basic science standpoint, is methylation at the COMT promoter, which is something that we would expect to be influenced by MTHFR genotype, because it’s so important in regulating the availability of these methyl moieties for intracellular methylation reactions, including at the COMT promoter, and other promoter methylation. The implications there are that differential degrees of promoter methylation can change the amount of protein expressed by the gene. So that would be one potential mechanism for the interaction. In terms of the clinical relevance, obviously the goal of this whole line of research that all of us who are doing imaging genetics work in schizophrenia would like to get to is somehow putting together multiples of these functional variants into a model that can account for as much of the variance as possible, in prefrontal activation or in working memory. There was, in fact, a poster at ACNP, by Weinberger’s group, where they have now put three SNPs together in three different genes, to account for activation during working memory. Although if I remember right, what they were looking at there was more hippocampal activation than prefrontal activation. But in any event, ultimately, what we’d like to be able to do is put as many of the SNPs together as is feasible in order to count for as much variance as we can. Now, that being said, the statistical methods for doing that really have not been marked out yet. You get into multiple comparison issues very quickly, and we have to be very careful in taking this approach that we’re not finding something just because we’re doing enough comparisons to see something.
So the methods there are really being worked out. But I think more directly clinically relevant is we’ve been looking at two biochemical systems that are very amenable to pharmacologic intervention. So with folate genes, you just give folate. It was suggested in our initial negative system finding, even among patients who have this dysfunctional T allele of MTHFR, high enough serum folate levels kind of compensate for the fact that they have this enzyme deficiency. Don Goff has a rather large RO-1 going right now where we are actually giving supplemental folate to schizophrenia patients—a double-blind placebo-controlled study. Not only are we looking to see whether folate will have a beneficial effect on things like negative symptoms or cognitive impairment, but also whether there are interactive effects with MTHFR genotype. We’re also imaging the patients before and after their course of treatment, so we can also see interactive effects with folate and genotype on prefrontal function during working memory. So that’s the folate side of things.
On the dopamine side of things, there are all sorts of potential agents that can be given to augment prefrontal dopamine signaling. Of course, there you’ve got some risks, because of the possibility of exacerbating psychosis. So that actually speaks to one of the potentially important ways in which the imaging genetics approach could potentially be clinically useful—you've got these potential genes and mechanisms that we think could account for improved symptoms, but they also come with a certain amount of risk.
It would be very nice if we had ways of predicting ahead of time who was likely to develop treatment-induced side effects, because you wouldn’t want to give a risky intervention to those people if you thought that they were often very likely to do poorly with side effects. There was a nice proof of concept of that idea in a paper by Alejandro Bertolino, in the American Journal of Psychiatry a number of years ago, where he did a prospective trial of olanzapine in schizophrenia and found not only that the COMT Met heterozygotes showed more improvement in terms of their performance and clinical symptoms with olanzapine, but also had more pronounced beneficial changes in prefrontal activation over the course of treatment. This obviously needs to be studied in larger cohorts, but I think the important take- home point is that you’ve got this drug olanzapine, and we’re not entirely sure by which mechanism it is leading to treatment improvement in schizophrenia, but for sure in many people it is going to cause obesity and hypertension, all sorts of lipid abnormalities. Wouldn’t it be nice to know before starting a medication like that that someone is likely to derive some benefit, based on these biological markers? A very similar idea could be applied using agents that we know to be directly involved in prefrontal dopamine signaling.