Australasian Science: Australia's authority on science since 1938

Sequencing seizures: Discovering new genetic mutations behind epilepsy

Neurologist Prof Sam Berkovic and molecular geneticist Prof David Goldstein describe their work uncovering chance mutations that cause childhood epilepsy.

DYANI LEWIS
I'm Dyani Lewis, thanks for joining us. The Human Genome Project, which published a completed sequence of our entire genetic code in 2003, introduced the world to large scale genomic sequencing efforts. Since then genome sequencing has become both faster and far more affordable. The result is that researchers and geneticists are now employing powerful sequencing strategies to investigate a great number of conditions, many for which a genetic cause has long been a mystery.

Epilepsy is one such condition and today on Up Close I am joined by two researchers who are using genomic sequencing technologies to identify the needle, or in this case the needles, in the haystack. They are looking for which genes out of the 20,500-odd genes in our genome are the faulty ones that cause epilepsy. My first guest today on Up Close is neurologist and epileptologist Professor Sam Berkovic. Sam is director of the Epilepsy Research Centre in Melbourne and Laureate Professor in the Department of Medicine at the University of Melbourne. Welcome to Up Close, Sam.

SAM BERKOVIC
Thank you, Dyani.

DYANI LEWIS
I'm also joined by Professor David Goldstein, Professor of Molecular Genetics and Microbiology, Professor of Biology and Director of the Centre for Human Genome Variation at Duke University. Welcome to Up Close, David.

DAVID GOLDSTEIN
Thank you, it's good to be here.

DYANI LEWIS
Sam, we refer to epilepsy as a single condition, but it's actually more correct to say epilepsies, isn't it?

SAM BERKOVIC
That's absolutely right. The epilepsies signify a group of diseases where the sufferers have epileptic seizures and we've learned that far from being a single condition it's very heterogeneous, both from what we see as clinicians and even more so now that we're digging into their molecular bases.

DYANI LEWIS
When you make a diagnosis of epilepsy, is it just on the basis of these seizures then?

SAM BERKOVIC
The diagnosis is based on what the patient suffers, which are the epileptic seizures, but together we traditionally put together investigations, such as the electro-encephalogram, the recording of brainwaves, and also and very importantly brain imaging, where we get a picture of the structure of the brain, which sometimes gives us the answer to what caused the epilepsy. But more often than not it does not and that's particularly where genetics comes in.

DYANI LEWIS
Can you explain what's going on with the brainwaves then when someone has a seizure?

SAM BERKOVIC
Yes, our brain is an electrochemical machine; we have many billions of neurons that talk to each other with little electrical signals, actually chemical signals that are then converted into electrical energy. That normally occurs in a very patterned and synchronised way. When that behaves abnormally, when there's an abnormal and excessive discharge, we have an epileptic seizure that's manifested by some change in behaviour, often loss of consciousness and it can be extremely fast and extremely dramatic.

DYANI LEWIS
David, you and Sam both are members of the Epi4K Consortium, which is headed by Duke. Can you tell us about what the goal of this consortium is?

DAVID GOLDSTEIN
The over-arching goal of the consortium is to find as many genetic explanations for the epilepsies as possible. The real motivation for the consortium is the recognition that now, because of so-called next generation sequencing, we really can perform genetic studies that we were never able to before. One way to really illustrate that is - you already mentioned the Human Genome Project, that effort essentially sequenced as its culmination, a complete human genome. It was actually a compositive around six individuals, but it completed determining most of the bases for most of the human genome, for a kind of representative human genome. The direct sequencing costs for that were $1 billion; the overall cost may be around $3 billion, but the direct sequencing costs $1 billion.

Today we can sequence an entire genome for about $4000 and we can do it in 10 days, running these so-called next generation machines normally. If we run them in a kind of a hot mode that makes them go fast and a bit more expensive, we can do it in a day. If we concentrate only on the part of the genome that actually makes protein, it's much cheaper and even faster. So what this means is that we can systematically find the variants that are present in any patient genome. What we set out to do in this consortium is take on a variety of different epilepsies and really find all the mutations that are present in those genomes and sift through them to find the ones that actually cause disease.

DYANI LEWIS
It's incredible change in technology that's led to that then?

DAVID GOLDSTEIN
It's really remarkable. For those of us that have been genetics for a while when we've been poking around tiny corners of the genome looking for risk factors, to suddenly be able to open up the whole genome and look through it systematically is really remarkable.

DYANI LEWIS
Sam, David mentioned that you're focusing in on a couple of epilepsies in particular. Which epilepsies are you looking at?

SAM BERKOVIC
Well, the whole project of Epi4K will sequence about 4000 genomes from patients and their families and that's in a series of projects. What's been reported just now is the project on what we call epileptic encephalopathies, which are about the most severe epilepsies there are. They affect predominantly children and infants and in these disorders there's not only epilepsy, but there's significant cognitive and behavioural deterioration. So these children, unfortunately, are very severely damaged and really finding the cause and then ultimately treatments is an urgent problem and that's what we're pleased to report at the moment.

DYANI LEWIS
As a neurologist, how common is it that you are able to tell a patient, or their parents, exactly what molecule in their brain, or what gene is faulty?

SAM BERKOVIC
That's where the game is changing. We've known that there's been a hereditary component to epilepsy really since time immemorial. Hypocrites thought that there was a genetic component. Over the years we've kind of known that there's been a higher rate of family history in most people with epilepsy, but what's becoming clear is that genetics is even more important than we'd realised from the straight clinical genetics. What we're reporting in this paper are de novo mutations. These are new mutations, so the epilepsy is not running through the family like a golden thread; it's appearing for the first time in the child. We believe those mutations usually occur either in the sperm or the egg. The parent is perfect healthy and there are no other affected members in the family and the child has a devastating illness.

These conditions were not particularly thought to be genetic and one of the real revelations that's happened in the last few years that's emphasised by this paper is that things that are genetic are not necessarily familial and we think that de novo mutation is going to have an even greater impact on epilepsies and probably other brain diseases as well.

DYANI LEWIS
David, you mentioned that you were sequencing only a portion of the genome and this is called exome sequencing. Can you explain for us what exome means and what kind of percentage of the genome is that?

DAVID GOLDSTEIN
We're looking at only just a bit under two per cent of the genome and that part of the genome that we're looking at is the part that directly encodes for proteins. So we have around 20,000 genes in the human genome that actually are turned into protein and this is where a lot of the mutations that influence human disease reside. Now by no means all, some mutations that influence disease don't directly change the protein that a gene encodes, but instead in some way influence how much of that protein is made, or where that protein is made, or when that protein comes on in the development of an individual. But some important proportion of the mutations that influence disease actually directly changed the protein sequences and that is therefore a very good place to start looking for disease-causing mutations. What the method does is concentrate that part of the genome that actually encodes for protein and then you sequence just that part, so it's much more economical.

Our approach in Epi4K is to start with that part of the genome and see how much we're able to explain concentrating on that accessible, easier to work with part of the genome. Then after we complete that phase of the work we have plans to expand out and try to take on the rest of the genome for those patients that remain genetically unexplained after our work concentrated on the exome.

DYANI LEWIS
With 20,000 genes it's still no mean feat, even though it is a small portion of the genome, it's still a huge job. But how does this approach differ from how geneticists would have previously looked for the cause of a disease?

DAVID GOLDSTEIN
In the past you really had to have some way of pointing towards part of the genome in order to be able to find disease-causing mutations. Probably the most common way of pointing at some of the part of the genome, certainly in terms of success, the most common way was to look at the coinheritance in families of disease status and also markers distributed throughout the genome. So these are just DNA differences that are not themselves causal of disease, but they just kind of flag different parts of the genome. Then what you do is you look for parts of the genome that are co-inherited along with disease status and then that tells you that that part of the genome includes the mutation-causing disease, so then you can zero in and focus on that part of the genome, so you get a pointer.

The difference now is that we don't need those pointers and that is why actually we can now take on these sporadic cases that really in many cases were absolutely refractory to genetic analysis before. What we can now do is look anywhere in the genome and we can, for example, take a child with an epileptic encephalopathy, sequence that child's exome or, indeed, the whole genome, sequence the child's parents and we can find just about all the new mutations in the child. There's some parts of the genome that are hard for us to sequence still, but the vast majority of the genome we can sequence and we can just identify the new mutations that are there anywhere in the genome. That really is why it's a different game now, because we can systematically look through genomes.

DYANI LEWIS
You're listening to Up Close. My guests today are neurologist Sam Berkovic and geneticist David Goldstein and we're discussing the genetics of epilepsies. I'm Dyani Lewis.

David, how many children with epilepsy did you sequence for this paper published August 2013 in Nature?

DAVID GOLDSTEIN
This paper reports on 265 children with epileptic encephalopathies, we sequenced them and their parents. The primary focus was looking for new mutation in the patients that contribute to disease. One point to emphasise is that that's actually a relatively modest, or even small, genetic study by contemporary standards. Even though it's relatively small, we have very, very clear evidence of many, many mutations that influence risk in this data set. From that we really draw a lot of encouragement that as sample sizes grow, we'll identify a lot of additional genes and mutations that influence risk.

DYANI LEWIS
De novo mutations are the essential ingredients for natural selection, so without this genetic variation between people and between parents and their children, human beings and every other organism just wouldn't be able to evolve. How did you go about working out whether the mutations you were finding were the cause of epilepsy or just some variation?

DAVID GOLDSTEIN
It's a great question, how you identify out of all the variants that you see, the specific ones that are pathogenic. The truth is it's very much a work in progress. There are lots of different approaches that can be used. There's some controversy in the field about the best way to identify pathogenic mutations, but one point to emphasise is that when you're working with brand new or de novo mutations it's much easier, in fact, than when you're working with inherited variants that are polymorphisms in the general population. The reason for that is that we all carry many, many variants that are relatively rare in the human population and that look kind of nasty. So for example, if you look at an individual without any known diagnosis, so they don't have any disease that they know about and you sequence their exome, just as we did for the children in our study, and you identify all the variants there that are rare in the general population, so therefore they might be bad, and that are predicted by available software to do something bad to proteins, any person will have a couple hundred of those.

So we sitting here in this room, we each have a couple hundred of those. So sifting through those can be challenging. With respect to de novo mutations however, there are many fewer that you have to consider. So if you look at only the exome on average a person will have just one de novo mutation. If you look at their genome as a whole, that number might be anywhere between 50 and 250, or 300. In fact, the number depends very, very heavily on dad's age. We know that the number of these brand new mutations that an individual carries is strongly influenced by the age of the father. If you have a younger father you have fewer of these, maybe closer to 50 throughout the genome. You have an older father and you have many more, maybe closer to 250 or 300. But looking at only the exome you have just about one and so what we can do is look across all the individuals we sequenced and ask whether there are particular genes that have more de novo mutations than they ought to have, given the mutation rate of that gene. If there's a sufficiently significant excess of mutations then we can declare that that gene must be influencing risk.

DYANI LEWIS
Sam, from a molecular perspective a lot of people would be familiar with recessive mutations, where they need to have one copy from their mother, one copy from their father and it's only when the child has two copies that something goes wrong. But in these de novo mutations you've got a dominant effect essentially, so what's happening with the protein that these only need one copy for something to go terribly wrong?

SAM BERKOVIC
That's a very good question and one that we're still struggling with. As you know we have two copies of our genes and in these disorders that are dominant there are probably at least three mechanisms by which a dominant disorder can cause disease. The first is that one may need a certain amount of the protein to have the normal function. So if one has just got half then there will be failure of function and disease. The second is there can be a so-called dominant negative effect where when you've got one bad copy of the protein it interacts with the good protein and there is, in fact, no useful function. The third is what we call gain ofr function and we're actually working on one such mutant now and that is where the mutation does not cause the protein to have loss of its function, but there's actually more function and that can, in fact, be if you like poisonous to the brain and there are some good examples of that as well. So that's one of the big challenges now to really understand the mechanism of all these mutations.

DYANI LEWIS
David, all of the children that you looked at had severe epilepsy. How many different genes were involved? Was it just one gene for all of the children?

DAVID GOLDSTEIN
No, there are a number of different genes that have clear evidence of influencing risk in this cohort that we looked at. Two of those genes were newly implicated in this study in influencing epileptic encephalopathies and several of the genes were already known to influence epileptic encephalopathies. In fact, across all of the genes that we can say are securely implicated, even newly implicated here or already known to influence epileptic encephalopathies, there are 25 distinct mutations that we have confidence are disease-causing spread across, as I said, several genes. Collectively they account for something like 10 per cent of the cases. Moreover these are just the genes that we have high confidence of. There are actually a number of other genes that really do look suspicious, that they probably are influencing risk and so they're kind of on a watch list until we see more data come in. We, in fact, already know from additional work that's going on in Epi4K that hasn't yet been published, we already know that several of these genes that are on the watch list will, in fact, over time be shown with statistical confidence to also influence epileptic encephalopathies.

DYANI LEWIS
For that statistical confidence you're just waiting to sequence more children and hopefully the same gene will crop up multiple times, is that right?

DAVID GOLDSTEIN
That's exactly right. So the way that it works is that for any given condition that you look at there is this technical term that we use called locus heterogeneity. What that refers to is the number of different genes that can carry a mutation to influence the condition. The higher the locus heterogeneity, so that is the more different genes that can influence the condition, the larger the sample size needs to be to track down the responsible genes. The reason for that is pretty obvious, if there are lots and lots of different genes, then the proportion of the patients that you look at that are due to any given gene will be relatively small. So to have multiple observations of that given gene, the overall cohort has to be relatively large. So what's happening is that so far we've been able to pick up the numerically more important of the genes and we're going to have to just keep going in sample size to pick up the remaining ones. We estimate from this study, in fact, one of the really nice things about this study is that because it was systematically performed it gives us a statistical insight into what we call the genetic architecture.
We can make inferences about things like how many genes can influence the disease, as I talked about earlier, and also how big an impact on risk any individual mutation makes. When we perform those analyses we estimated that the number of genes out of a particular subset of genes in the human genome, the number of genes that can influence risk, is 90. So what that tells us is that we've got a ways to go and it's going to take a larger sample size.

DYANI LEWIS
Now, Sam, as a neurologist when you're looking at all of this data coming out from the genetics and seeing some of the genes that are mutated, did they make sense to you as a neurologist?

SAM BERKOVIC
I'd really like to say yes, but the answer is no. I guess the big perspective as a clinician is that we know that epilepsy is complicated from a clinical point of view and we diagnose what we call specific syndromes, which mean collections of particular seizure types, e.g. abnormalities, ages of onset, et cetera and intellectual capacity, et cetera. I guess about 20 years ago we had a fairly simple view that as the genetics tumbled out, we'd have one syndrome that we recognise as clinicians and that there'd be one gene for that. How wrong we were. What has turned out is that each syndrome that we recognise as relatively homogenous turns out to have a number of genes, and apparently an increasing number, and more than that one gene particularly if its mutated in slightly different ways can cause different syndromes.

So it's a very complex jigsaw puzzle. As a clinician it makes you want to throw up your hands and say, well, nothing will be sorted out here. As a researcher it's even more exciting, because it then gives us the stimulus to try and figure it all out and make nature simple again.

DYANI LEWIS
But were there any particular proteins or cellular mechanisms that were always affected, or affected quite a lot?

SAM BERKOVIC
Yeah, so the first large group of genes that were discovered for epilepsy, which are still thought to be important are ion channels. Our nerve cells have simple chemicals in them like sodium, potassium and chloride and these are maintained at particular gradients inside and outside the cell. In fact, most of the energy that we consume in the brain is actually working those pumps, it's sort of pumping iron to keep those gradients at the correct level for normal functioning. Disorders of those ion channels have turned out to be an important cause of epilepsy and that kind of makes sense for a disorder that is due to abnormal excitability, the brain appears to be functioning normally most of the time and then suddenly there's abnormal excitability, so ion channels make a lot of sense.

However, there is far more to epilepsy than ion channels and we're learning that genes that are important for brain development also cause epilepsy. Genes that are important for the formation and function of synapses, the structures where nerve cells communicate with one another are also important. The landscape of genes involved in epilepsy is now increasing.

DYANI LEWIS
The two genes that were identified in this particular study as having very strong statistical and genetic support and were new to epilepsy research, could you describe those two genes and what they do?

SAM BERKOVIC
Well, one is a GABA receptor and GABA is the major inhibitory transmitter in the brain and the GABA receptors have been longstanding target for epilepsy genetics. But this one had not been securely identified in epileptic encephalopathies before. So that one, perhaps, was a little less of a surprise, but the other one we really did not expect at all and we're learning more and more about new pathways in epilepsy by this sort of study.

DYANI LEWIS
The GABA receptor, this is working at the synapses you were describing before and the other one asparagine-linked glycosylation 13, what on earth does that do?

SAM BERKOVIC
What a mouthful? We know that glycosylation of proteins, that is putting on sugars on proteins is important for many functions. Although there are a number of other glycosylation disorders that include epilepsies, they haven't really been implicated in epileptic encephalopathies to this date, so that gene was a big surprise. If you were to have shown me a list of genes and asked me to pick what might be likely, I certainly would have thought about the GABA receptor, but I wouldn't have thought about this one. That's the beauty of this whole exome approach where we're taking an agnostic view and not going in with any preconceived ideas.

DYANI LEWIS
But David, I guess also coming up with genes that have previously been implicated must be a great endorsement for the approach that you are taking?

DAVID GOLDSTEIN
So I think that's right. We were very reassured, in fact, when we saw very, very clear evidence for genes that have, in fact, for many years been known to influence the epileptic encephalopathies, so that's a really nice proof of concept that we can track down the causal genes. Of course, one expects that we would get those known genes first; because of course the reason that they were found is that they're the relatively more important ones. That tell us that the technique is working and that we can continue, funding permitting, this work until we track down most of them. I think one of the very most encouraging things, at least to me in the work that we're doing, is that the genes that we're implicating are not - very clearly - not just drawn at random from the human genome. They very, very clearly are drawn from a relatively small number of different biological pathways.
This really is all important, because if it's true that there are 90 or 100 or a bit more genes that can influence epileptic encephalopathies and from patient to patient, the causal mutations in those genes are different, there really is a concern that every patient is different from every other patient and in that case how would the genetics really help you, because you're still dealing with each patient being their own kind of example? But on the other hand, if the different genes and the mutations in them influence a small number of distinct biological pathways, then what we can do is we can really concentrate on organising the patients into what pathway is dysregulated and try to figure out the right treatment to use when that particular pathway is dysregulated. We can also think about developing drug discovery programs around the particular dysregulated pathways and lots and lots of different kinds of analyses strongly indicate to us that we really are seeing particular pathways implicated.

Just a comment about the number of genes that we think is involved in influencing the epileptic encephalopathies. When we started this endeavour, now a couple of years or so back, we had a kick-off party and a dinner where Sam came to Duke and somewhere into the evening I asked him, well, how many genes do you think that might influence the epileptic encephalopathies? I won't try to do the accent, I probably could practise that, but he said, look, I've had three glasses of wine, but if you ask me it's going to be somewhere between 90 and 100. After we did a whole bunch of super complicated statistical genetics, as you heard that's exactly the inference that we came to, so I hope you will ask him how he did that calculation?

DYANI LEWIS
I want to know what the wager was.
[Laughter]

DAVID GOLDSTEIN
Unfortunately I didn't put anything on it.

SAM BERKOVIC
No, no money changed hands.

DYANI LEWIS
You're listening to Up Close. My guests today are neurologist Sam Berkovic and geneticist David Goldstein and we're discussing the genetics of epilepsies. I'm Dyani Lewis.

David, you mentioned some studies that have shown that older parents, or when they're older at the time of conception, that in the children there are more de novo mutations. Is this what your study found?

DAVID GOLDSTEIN
We didn't look at that directly, that wasn't the focus of our study. I think the best study on that question came out of Iceland from a company called deCODE genetics and what they did is generate whole genome sequence data on families and that way they can identify most of the de novo mutations, whereas we only get a small fraction of them. They were able to ask how much the father's age influences the number of de novo mutations. They did find that it really is the principal determinant of the number of de novo mutations. You can really think of these de novo mutations as kind of taking a sledge hammer to the genome, that really is the way to think of them. It's just a change that falls down somewhere in the genome and if it falls in an important part of the genome, just like taking a sledge hammer to any complicated instrument, if you bang on an engine you're unlikely to improve it, just exactly the same way if a de novo mutation falls into an important part of the genome, it is very, very, very unlikely to do something helpful and it has a pretty decent chance of doing something harmful. Now, of course, a very, very small fraction of them in an evolutionary sense will actually do something useful, but the vast majority of them do something harmful. So these are things that you really don't want in your genome.

DYANI LEWIS
But these kids have just been incredibly unlucky as well.

DAVID GOLDSTEIN
That's exactly right. So these de novo mutations we all carry them and most of them, fortunately, do tend to fall in parts of the genome that aren't so important, maybe they do a little bit. Maybe they make one of us a bit near-sighted or something, maybe they do a little bit, but maybe they don't do anything too bad. Maybe in some cases, certainly, they do nothing at all, but when they fall in critical parts of the genome then they can do something very serious.

DYANI LEWIS
Sam, will this information change the way that children are diagnosed with epilepsy, or the information that they receive when they get diagnosed?

SAM BERKOVIC
Absolutely. One thing I'd like to emphasise before I answer that specifically is that although patients and their families come to us wanting treatment, or wanting a cure, they also really want an answer as to what caused it. That's a really important part of the therapeutic interaction that I see as a clinician. So in the past many families carried around the belief that the fall from the change table, or the fact that they didn't get to the hospital in time and birth took a little longer, or was a bit traumatic, they carry these false beliefs that had that not happened their child would be healthy and that's actually an incredibly heavy burden. Or alternatively, they go around with absolutely no idea and they go to all sorts of alternative practices, they go to the 50th doctor and they spend their lives in a fruitless search. Getting that closure is really important and when they get that it really allows them to focus and then to go on.

They're often interested to then meet up with families with children with the same or similar problems. That gives them a sense of community, a sense of strength, a sense of empowerment and that's incredibly important. So when you achieve that it really is a milestone in the treatment of that particular child and that particular family. In some cases then we can say, well look, this child has this particular syndrome due to this particular gene and we know that this particular group of drugs is helpful, or not helpful, and it allows us to tailor the treatment. It also stops us doing things that we don't need, in other words very invasive testing. In the past many of these children had investigations for epilepsy surgery, which although applicable to a very small proportion is not applicable to the majority. So this was quite traumatic for many of the children going through some of this testing.

Of course, the end point will be when we can come up with directed treatment for the particular mutations. That's a much longer road, but one that we can now actually see the landscape for and we can see the path to move down. That's happening here at this campus and in many other places around the world. We've now got the substrate to ask the question, how can we possibly fix this?

DYANI LEWIS
David, where to from here for the Epi4K?

DAVID GOLDSTEIN
In a way we have started with the genetically more attractable problem, the epileptic encephalopathies and we want to move on from there to what we anticipate will be a harder, but very important problem of looking at risk factors for epilepsies that occur later in life, or that also occur in families with multiple affected individuals. So we really are moving on to trying to take on now really the full spectrum of what we refer to as non-acquired epilepsy, so those epilepsies where we don't have a clear identified cause for the epilepsies. We're going to systematically now look through those types of epilepsies for genetic risk factors looking at many, many, many more genomes than we have so far, so that's the next step genetically. After that what we hope to do is take the mutations that we're identifying and put them into appropriate biological assay systems that will allow us to develop what we refer to as screening programs, where we can take a mutation, we can put it into neuronal cultures and we can ask, how do those mutations change the properties of the neuron and, in fact, of networks of neurons that are interacting?

If we can see how the mutations we are identifying change, say, the network properties of groups of neurons that have the mutation, then what we can do is we can actually give those cultures drugs and we can ask, can we actually push the network behaviour back towards normal, looking at the behaviour of neurons sitting in a dish on the bench in a laboratory? If we can do that we can actually start to work out how to develop new treatments and how to pick the best treatment for the particular mutations we've identified. That's really what's got us really excited in Epi4K right now.

DYANI LEWIS
Sam Berkovic, David Goldstein, thank you for being my guests on Up Close today and discussing your work on epilepsy.

SAM BERKOVIC
Thank you very much.

DAVID GOLDSTEIN
Thank you very much.

DYANI LEWIS
Professor Sam Berkovic is the Director of the Epilepsy Research Centre based in Melbourne and Laureate Professor in the Department of Medicine at the University of Melbourne. Professor David Goldstein is Professor of Molecular Genetics in Microbiology, Professor of Biology and Director of the Centre for Human Genome Variation at Duke University.