Dr James Lush, MC, says:
Please make Dr Michael Kobor welcome.
[audience applauds]
Dr Michael Kobor, of the University of British Columbia, says:
Thank you, it's a real pleasure to be here.
I want to take the opportunity to thank the organisers for inviting me to spend two wonderful days here with you. I've been really, deeply impressed with the wonderful engagement that Australian FASD community has. It's really, very impressive.
I've been very touched by many sort of personal stories, that I've heard over these two days, and especially seeing that, as we'll see in a second, this is rather miraculous, very meaningful to me and I'm honoured to discuss today, our work with you, and, in particular, because it's such a diverse audience.
What I'd like to do today is just to spend a few minutes continuing from where Jeff started to discuss with you how epigenetics has become such a crucial player in the face between the genome and environment. Then, I've heard a lot of talk about environments, broadly speaking, the last couple of days. I want to share with you some of the work we have been doing over the years, to show that there are indeed a variety of environments, including, quality, and family environments get under the skin to affect our biology at the level of the DNA.
I'm going to move on to discuss the work we've been doing on fetal alcohol spectrum disorder, and collectively, what I love for you to take away from this, is that, indeed, epigenetics might be a key player in the face of nature and nurture. But also, that there are many different aspects, many different environments, that actually affect the epigenome.
Jeff already alluded briefly to what epigenetics is. You can see the most common definition here on the screen. I actually hate definitions, I'm German, and we're supposed to love definitions, I hate them. Which is probably why I live in Canada. The particular aspect of epigenetics that we are really focusing on is the social environment.
The idea that our experiences, environments and particularly those in early life, can get under our skin to affect our biology.
The kind of paradigm that we're working under is that, these experiences, they affect the biological pathways by getting in the skin, and they drill all the way onto the level, into the DNA, into the nucleus of a cell. Of course, this is not a one way road, because you can then easily imagine this to go the other way, as well.
When we think of epigenetics, we really just need to understand that little DNA piece, there. We need to understand that we have 25,000 genes.
With all due respect to my molecular colleagues, all you really need to know is that a gene is essentially a light bulb. Okay? A light bulb can be either totally on, or totally off. Or if you live in a fancy house, you have a dimmer, and you can put them somewhere in between. The epigenetics, in particular the kind of DNA methylation that Jeff alluded to, if it's in a gene regulatory region of genes, it's really just a dimmer. That's all you need to know about epigenetics, get your certificate as an epigeneticist.
How does that look in a particular situation? Let's say you have your favourite gene, in this particular situation, we have no methylation, and the light bulb is all the way on. Now we start to methylate. Methylation, it's just a tiny little chemical tear, consisting of some carbons and hydrogens. Now we start to methylate. You see how we methylate, we push that dimmer down. We push it further down and ultimately, we have fully methylated this gene, and we have turned the light bulb off. That's as simple as it is.
How does it look in real life? If you look at these five young mice, is there anyone in the room who thinks that they look exactly the same? Good. Otherwise, optometrists have some work to do.
[audience laughs]
So we all agree, I think that the coat colour is quite different. But what I'm going to tell you is that they are genetically, actually identical. It's not that they have different genes. The only reason that they have different coat colours, is that their mum was fed a different diet during pregnancy.
This particular coat colour has the tone by one gene, and the mice on the left, this particular gene, is totally on, the dimmer is all the way up. We have no methylation. But as we go through the mice on the right, we start to methylate and methylate this gene. We turn that light bulb off, and now, we have a different coat colour.
Again, this is solely dependent on the nutrition of the mum during pregnancy, there are a number of compounds that have been shown to affect this kind of circuitry.
Of course, most people really don't care, whether they have yellow or brown mice in their house, as long as they don't eat the cheese.
[audience laughs]
Let's just wait what happens to these mice when they grow up.
Keep in mind again, they're genetically identical, because those people in the basements, that grow the mice, it's what they do.
The only difference was that the mum was fed a different diet during pregnancy. You see the powerful impact of these epigenetics.
You see that there was a change in these dimmer switches that was dependent on the mum's diet during pregnancy. It got under the skin of these mice. It stuck with them for the rest of their life, to lead to these vastly different health outcomes.
This touches on all kinds of things, including our mental health, but also, as you can see, our metabolic health, and the like.
This is very powerful, and since, as Jeff alluded to, this has been actually shown to be true in humans, as well.
As we can see here in the Dutch famine, where we have these associations of the maternal diet, and a caloric intake, with the dimmer switches and the offspring. There's also prenatal smoking that has been shown to affect this circuitry, and a number of other compounds.
The field really has exploded some 10 years ago, with the seminal work by Michael Meaney from McGill university, in Canada, who showed that the quality of the maternal care, also, can affect these dimmers. This work really has brought this entire field to the attention of the public.
I'm going to show you a little bit of data related to this. The closest we can come to these cute mice, are identical twins, that Jeff alluded to. So you see, those two cuties here, they're identical twins, but obviously, they don't quite look the same. Yet, if people look at their epigenome, it turns out the epigenome is very similar, early in life, and it becomes divergent later in life. In particular, if twins are separated early in life.
In turn, these kind of modifications are also associated with a host of health outcomes, including mental health, as shown primarily by Jonathan Mill, from the UK.
Taking all of this together, my team really focuses on the role of epigenetics in child health, broadly speaking. Really at the interface of nature, which is our DNA, and nurture, which is our environment.
Ultimately, what we're trying to understand with this integrative use, how can we provide the best and most healthy starts for our children, and how can we potentially prevent disease? We had the honour of working on a number of interesting projects, over the years, that I think are highly relevant for this aspect to really understand how experiences get under skin.
As you might know, there are very few experiences that are more powerful than socioeconomic status. We've been focusing on that, in a suite of work, to understand, does it matter whether you grew up in the dwelling on the left, or in the mansion on the right? We've also been focusing on the family unit. Does it matter, whether you have a family, with the nuclear family, with doggy woof woof, or the family where the parents constantly yell at each other.
Ultimately, how does that then turn around to affect childhood temperament, and behaviour and trajectories? This is really rooted in this idea that the developmental origins of health and disease, that can trace all the way back, as Jeff so beautifully showed to you, to the womb. But we're focusing on these early years of life, and trying to understand how these experiences get under the skin, and of course, the idea that happens by changing these dimmer switches, and the light bulbs.
In a suite of studies, we really focused on adults, trying to test whether we can see a vestige of early life quality. This was published some years ago, but what we did, we essentially had two groups of adults, that were all healthy adults, that were separated by whether they grew up, the first five years, in low versus high socioeconomic status, but were mixed in adulthood, so that whatever we measured was not confounded by the adult socioeconomic status.
Then we asked a number of questions. We first asked "are these light bulbs different?".
Indeed, as you can see, they are actually different.
Without going into too much detail, they're not different in a random way, but they're different in a way that suggests that the folks that grew up in low socioeconomic status, actually have an elevated immune response, even as adults, that can be traced back to their socioeconomic status in early life.
Lo and behold, when we then directly test their immune function, you can see that on the top right. The prediction came true, that these people had constantly a higher engagement of their innate immune pathways. For these folks, also had a consistently higher causal, so their arousal level, their stress level, if you wish, it was higher throughout the day. Again, all of this can be traced back to their socioeconomic status early in life, and not their current socioeconomic status.
We can also show, as seen by these highly skewed distributions of these parts, What you want to see is this going to the left.
There is an association of early life socioeconomic status in these dimmers, in these people.
Interestingly, at least in this study, and some of us that we've learned by now, there's actually no association of your current socioeconomic status with these dimmers. Suggesting again, that collectively, the early life quality gets under the skin, to set these dimmers in a persistent way that lasts all the way into adulthood, to change the light bulb setting, and to change the function of the immune system, and distress reactivity of people.
Socioeconomic status, of course, as you know, is a very highly complex construct.
We're constantly trying to tease it apart even more.
In the paper that we recently published together with Tom Boyce and Nicki Bush, in a cohort from the San Francisco area, we were actually able to take socioeconomic status apart into three components, all of which, actually, as you can see by the skewed distribution of these bars to the left, and these little dots that are either blue or red, all of those are associated with either income per dependent, parental education, or family adversity. Suggesting that each of those leaves a distinct signature, into the epigenome, in this case, of these children that are between 8 and 12 years old.
We also have been looking more closely, motivated by the fact that we wanted to see where we can potentially use this work to provide some recommendations as to what families can do. We have been interested whether it is indeed also a signature of the family unit that goes beyond just adversity.
In work that we published just last year, we've actually shown that these dimmers can be traced back to whether a neonate actually receives a lot of caretaker interaction versus very little caretaker interaction. This work of course, as I mentioned, caught quite a bit of public attention, sort of news services saying cuddling babies, or not, affects infants' gene expression. Scary stuff. But we'll come back to this later.
How can we apply all of this to really get the bigger picture? We can do that by adding more and more studies, and that's exactly what we have done.
We have published studies that show kids growing up in orphanages in Eastern Europe, being adopted to the United States. We see these signatures of dimmers. We see the signatures of the immune system.
We have done studies in the Philippines, where we can show that in young adults, again, there is a signature of both social, but also physical environment, in their lifestyles.
Collectively, really a picture is emerging by which these epigenetic modifications, early in life, are being sculpted by a number of different environments.
Together, with my mentor and friend, Tom Boyce, we have then put this recently into a more comprehensive review, trying to understand how this can serve as the synapse of the gene environment, to help our children have a rocket start to life.
How can we apply all of this to FASD? As you know, much better than I do, FASD is a multifaceted syndrome.
In Canada, as we heard yesterday, the incidence is between 2 and 5%.
We have then put this epigenetic biology to the test, in the context of FASD. We've done this under the umbrella of the kids brain health work, network, formally NeuroDevNet, a Canadian organisation, that is dedicated to understanding neurodevelopmental disorders, from the basic biology, all the way to the advocacy.
What we try to do with this work is twofold. We try to ask the question whether epigenetics can tell us something about the biology of the disease, of the disorder. Perhaps more tangibly, whether epigenetics can serve as a new biomarker for better, earlier, more sophisticated diagnosis, or just another tool in the toolkit.
I'm happy to say that I've seen the early results, which suggest that both of these could be the case.
I've been building this in two Canadian cohorts, the NeuroDevNet cohort, FASD cohort, and the bigger one, a Pan-Canadian cohort.
You can see 110 cases, and around 96 controls, and 110 cases, distributed across the range of FASD, versus the second cohort, the smaller one, it's the Manitoba cohort, and it's small, but it's only children with FASD.
So using the first cohort, we asked again the question are there differences between the cases and controls, in these dimmer switches, and lo and behold, as you can see here, there are lots of blue dots. You just want to focus on the blue dots, and you want to focus on the blue dots that go far, to both, away from the middle, and they go high up. This is called the volcano plot. The furthermost up, and far, is really the most meaningful associations.
You can see there are roughly 650 or so CpGs, and more than 400 different genes, that are associated with fetal alcohol spectrum disorder in this first cohort.
When we look more closely at those, we can see the direction, a number of them that make biological sense, even though this is cheek swabs Including components, such as DLD4, that are part of the dopaminergic system, and we can come back to DLD4, in I believe, three slides, so hold your questions.
We can also look beyond these individual thoughts, and look for broader reasons. When we do that, we have to buy informatic tools that are available. We can identify epigenetic neighbourhoods that differ between kids with FASD and the controls.
Again, we see changes in genes that make a lot of biological sense. We see changes in genes that are related to the immune response.
As you know, individuals with FASD tend to have compromised immune functioning, higher incidence of arthritis, and the like.
We also see genes that are associated with autism spectrum disorder - perhaps not entirely surprisingly.
When we then do even more sophisticated analysis, and I'm gonna spare you, because I barely understand it myself. To look at sort of network analysis, we can see that there is indeed an overlap, potentially with other neurodevelopmental processes in this order, again, including autism, anxiety, and epilepsy.
Of course, in our world, you're only worth as much as you can replicate, so this is why we turned, after this initial promising work to NeuroDevNet, to ask to what extent can we actually replicate this? We worked with Alex Lussier, a very talented former PhD student, published just earlier this year, and see that we can replicate roughly 180 of those 650 CpGs.
Those are the ones, if you look at the left, that are red, you can see that many others also go in the same direction, but do not pass our fairly rigorous statistical threshold. As Jeff pointed out, this is all done by very sophisticated technology, called Array.
Often, what we try to do is to use a different technology, and we use what's called pyrosequencing, to look at whether this actually holds up, and it's not just an artefact of the technology.
And as you can see here for this gene, that's involved in brain development, and substance-related disorders, and epilepsy. You can see that both technologies give a very similar pattern of association between FASD and controls.
The reason I will put this up is that, you can also see, if you take a single individual on this plot, you take a single individual, just based on this one gene, you actually cannot tell that this person would be part of the control group, or that will be part of the FASD cohort.
How do we get around this? We get around this by using the black box, and a magic word that everybody loves these days, machine learning. What Alex did to use the first cohort, the neurodefinite cohort one, and had this DNA, this signature of the FASD, that I showed to you.
Then he used "machine learning", which you can talk about later, to essentially come up with a predictor for FASD, that's not just based on one single of these dimmer switches, but uses some fancy thing, called neural nets, to identify patterns in this data.
When he does that, he then tests this, any stuff, the FASD cohort, the second one, and then an active control, which is an autism spectrum disorder cohort, publicly available.
When you look at this curve, this receiver operating characteristic curve, I think we’re getting somewhere. We definitely, if there was no predictive value, the blue and the grey things would just go diagonally up the middle. You can see that there is this nice shape that we're looking for. It's not perfect.
When we then start mixing this, and putting it right to the test in the NeuroDevNet two cohort, you can see that the predictor actually predicts 18 of the 24 FASD cases. It misses six, and also mislabels two controls as FASD. It's not perfect yet, but I think it's a very promising beginning that we can use these kind of technologies, and these kinds of approaches to understand, to come up with biomarkers, and a predictor, for fetal alcohol spectrum disorder, that hopefully can be added to the toolbox.
We also then wanted to understand the biology just a wee bit more, and there's very little you can do in the human biology, other than measuring these associations that I talked about.
We turned, together with my dear friend, Joanne Weinberg, we turned to the rat. Alex has just published a paper, it's not actually published, it's just accepted this week, where he uses a paradigm of three different feeding pallets prenatally with the rat control, nutritional control, called Pair-Fed, and the prenatal alcohol exposure.
There's a long story to this, but Alex was essentially then wondering how are these dimmers switches, looking in the brain, in the hypothalamus, which is a key brain area involved in FASD in the stress response in these rats.
He identified 30 different regions across the rat development that were differentially methylated and would be associated with the exposure to prenatal alcohol.
Then, of course, the key question becomes, can we actually draw something about this biology, and compare it to the human data? You see it from the title, it's a little bit like comparing apples and oranges.
The good message is, we humans aren't rats.
[audience laughing]
Of course our brain and our genes are quite different, but nonetheless, there was actually one particular gene, among all of different dimmer switches, that Alex identified that was different, and that was common between those two paradigms.
Any wild guesses? I told you this three slides ago - DLD4.
The DLD4 gene, which is involved in the dopamine pathway, as I said earlier, is actually, you can see that here again, in these pyro sequencing panels, it's very differently methylated.
These dimmers switches are very different between the exposed rats, which are in red at the top, and the blue rats, which are the control rats. This is true across the entire span of development.
If you look at the human again, it's also differentially methylated, It's not exactly the same region, it goes in different directions, but nonetheless, I think it's very telling, very exciting.
It's even more exciting when we look at a paper that these lovely Australian folks, some of which are in this room, published. They also looked at prenatal, and maternal alcohol consumption, and DLD4 methylation. Maybe it's a sign that we should work closely together, to explore that.
With that, I want to close, and really share with you some of my thoughts on all of this.
First of all, and I sort of divided up in different pieces. First of all, I want to make sure that you understand this. This is all correlation, and it does not imply causality. That's a very important point.
Secondly, I also want to highlight that by and large, with some exceptions, these DNA methylation changes aren't necessarily bad or good. They can mean all kinds of different things, or they can mean nothing, it can be just a biomarker. I also want to follow up and highlight again that, the gap, we have a gap between these group averages, and individual prediction sets. We're trying to close it with these machine learning approaches, but the gap still exists, very much. Again, this is an important aspect.
On a more technical level, of course, our cohorts were such, that we can't really distinguish a signature between PAE versus FASD. As you probably can tell from all the different environments that are shown to be associated DNA methylation, many, many different confounding factors contribute to shaping the epigenome, some of which might be alarming and some of which might not.
There is also this issue of tissue-specific versus tissue-agnostic measures.
And then of course, what we have not yet touched upon is that there might well be a genetic underpinning to all of this.
What I'm really most excited about is this knowledge translation, the opportunities, that all of this offers.
As you can probably tell, it's a fairly complex biological concept. The key point that I think I'd really love to discuss further, is this idea that there could be an epigenetic risk messaging, and that the idea that there could be a biological method, determinism. I want to strongly argue that this work is actually not the case.
I'm very curious to learn when you think about this work, whether it actually increases or decreases FASD stigmatisation. We're actually starting to have focus groups to address these questions, that include a number of different stakeholders.
Collectively, I hope that I've shared a bit of my excitements about early life experiences, going all the way back to prenatal, and their way of getting under the skin, modifications of these dimmers, the epigenome, the interaction of the light bulbs, to affect phenotypes, broadly speaking.
Thank you so very much.
[audience applauds]