 Like some of the other speakers this morning, there was a small indel that occurred as I was coming down on the plane last night, and I'm not going to tell you about how de novo variants inform clinical phenotypes, but ask can they? And I'm going to take a basic science approach to this as is appropriate for this particular part of the day's meetings. So I'm going to tell you a little bit about critical congenital heart disease following up on the lovely talk we just heard. This is a very important problem that occurs in lots and lots of children around the United States and indeed the world. One percent of children are born with a heart malformation, and in fact, nine percent of those heart malformations are profoundly severe and require an intervention on the first year of life, what we deem as critical congenital heart disease. Now, until about 40 years ago, these children all died, and frankly, their healthcare long-term issues became a moot point because of their very shortened survival. But with increased improvement in interventions, we now know that this is the largest growing cardiovascular population, in particular among the young. As you can see, now one in .45 of 1,000 children has a repaired congenital malformation, a severely congenital malformation, and their mean age is 17 years. So these children are beginning to enter reproductive periods of life. When we think about the causes of these disorders, we've known about all sorts of acquired causes, rubella, toxins, et cetera, but there has been a recognition that in Western countries, especially in the U.S., these are unlikely etiologies for these devastating malformations. And moreover, as a challenge to the geneticists, as you can see, their pedigrees are really quite simple, perfectly normal mom and dad, and then whammo a child with some of these very profound malformations. Moreover, we know that about 11 percent of these children, after excluding anyone with a previously defined syndromic disorder, still has another congenital anomaly. And one of the most devastating issues is that even after we repair these babies' hearts, that the prognosis of these children is unfortunately very much compromised by lots of other associated lifelong issues, in particular neurodevelopmental delays. And the severity of the congenital heart disease correlates with the severity of the NDD, something that is obviously implies that these two features may be linked. The flip side of that is we've all recognized as during embryogenesis these children have less oxygenation, less potential supply of nutrients, and in the postnatal stages, they're enormously stressed with the interventions and surgical procedures that are performed to help them to survive. And so NHLBI took the lead on this devastating problem and put together a consortium to try and address severe forms of congenital heart disease. And as you would assume from the talk we've been talking, we've been discussing today, one of the hypothesis is that these were de novo mutations, that these were likely damaging mutations, and in fact that these could be identified by whole exome sequencing. And so we've put together now 10,000 children with a range of malformations. I'm going to tell you about the analysis of about 1,200 with severe congenital heart malformations. Again, we've excluded children who were recognized to have an overt syndrome. They are unaffected, their parents are unaffected, and they have no family history of disease. And the model was that we in fact would find a de novo mutation by whole exome sequencing of the entire TRIO. In addition to actually looking at whether a variant was discovered, we wanted to interpret whether in fact these variants might be meaningful using the most robust bioinformatic platforms available. And so what we did is we capitalized on some of the information that's been emerging in terms of how likely is one nucleotide to change in the context of how deeply the sequence was obtained, in the context of the flanking sequences around that particular variant, and in terms of whether that variant might be meaningful based on the divergence scores between related species and whether that was a conserved variant, sorry, a conserved residue and a conserved amino acid. And then as well, to further extend the work that was published by Mark Daley and his postdoctoral fellow, Caitlin Simochew, we've actually included some bioinformatic predictions of missense variants to identify those that might in fact be pathologic. Now these are expectation models, and so we wanted to have a control population in which we put through the same platform, and we capitalized on the Simon-Simplex collection of TRIOs. In fact they were quads in which a child with autism spectrum disorder and their unaffected sibling has been recruited. We therefore harnessed whole exome sequencing on the unaffected parents and the unaffected child as if you will attest to whether these bioinformatic platforms were useful. So this is what we found from these analyses of 1,200 congenital heart disease TRIOs and the 900 healthy, if you will, children of healthy parents. And what you can see is that as the model predicts that there's zenoval mutations in every one of us. I like to tell my friends that this is the best evidence I know that evolution does occur. It occurs in real time in each couple that has a child. And suffice it to say that when we take these variants and we then parse them to see if in fact they are predicted to be damaging, we see an enrichment in the children with congenital heart disease as compared to the controls and the same to with loss of function. If we combine those deleterious missense and loss of function into a junction that we call damaging mutations, we see an enrichment. Now this is genome-wide, but we know congenital heart disease affects the heart. And so if we do that same analysis and now only focus on genes that are highly expressed in the heart, the enrichment becomes very strikingly impressive, in my opinion. It doesn't occur in our controlled population. And if we did the same analysis, restricting the analysis to genes that are expressed in the lower quadrant of all transcripts expressed in the developing heart, we don't see this enrichment. So that tells us a couple of things. First, you look where the money is. If you have heart disease, you want something that's expressed in the heart, and you're probably going to do better when you concentrate on those genes that are very, very significantly expressed. So what is the evidence that these might in fact be pathogenic? So 21 of the genes had recurrent damaging mutations, and it was gratifying to see that some of these in fact are known CHD genes and disappointing that some of these in fact are syndromic genes. And that again reflects the fact that when you have a newborn infant, recognizing the child with noon and syndrome can be very, very difficult. That being said, we also identified some variants in genes that were not recognized to be implicated in congenital heart disease before. And one of the questions is, is this pathologic? Well, it's rather impressive that this occurred in three children and in none of our controls. Moreover, we had done exome sequencing in another code or horrid of children and looked at copy number variants using exome sequence data. And we had identified a substantial copy number variant that also deleted one copy of RBFOX2. And interestingly, all four of these children had exactly the same clinical phenotype. Something that I would suggest is highly unlikely to occur by chance. They all have hypoplastic left ventricular heart syndrome. And as you can see here, there's virtually no ventricular chamber on this schematic from the CDC. Moreover, RBFOX2, some of you may know, is an RNA binding protein that's important and has been hypothesized to be critical for alternative splicing, in particular in development. And so with that piece of information, we then asked, what are the targets of this molecule? And when we looked at those targets, we saw an enrichment of damaging mutations in the target of RBFOX genes that also were in the same cohort with congenital heart disease. So a lot of basic information that these are likely to be contributing to disease. But then we felt it was important to go back and really consider clinical phenotype. And I told you that often these children have not only a heart malformation but something else as well. And when we looked at who carried these damaging mutations, what we found is that there wasn't enrichment in the cohort overall, but it was strikingly increased if the child had an additional abnormality. And as you can see, parts in these right bars, if they in particular had an abnormality that involved just neurocognitive abnormalities, neurocognitive delays, rather. And I will remind you that because these were newborns when we recruited them that we could only restrict those analyses to children who were over 18 months of age and who had neurodevelopmental phenotypes. But we had about a third who could be classified, a third in whom we were certain did not have MDD and a third who were just too young to call. And when we only looked at those in who we could definitively look at their other abnormalities, be it in neurocognitive function or malformation, the enrichment was highest in those kids with both. So why is that? Well, if you look at the developing transcriptomes in the heart and the brain of a mouse, you can see, and it's hard to project, I apologize, but there's a smattering of gray dots across this graph which represents all of the genes of expression. In the top-core trial of expression in both the brain and heart, there's significant overlap of these genes. And the blue dots here represent those genes that are mutated in our children with congenital heart disease and as well are previously recognized to be mutated in children with a wide spectrum of neurodevelopmental abnormalities. The size of the dots reflects how often that gene was repetitively hit in the purple hue. Moreover, when we then took our children and said if we looked at genes known to be implicated in neurodevelopmental abnormalities and highly expressed in both the heart and the brain, we can see again that the children who have these neurodevelopmental abnormalities are increased in comparison to those without neurodevelopmental deficits, suggesting that there is a genetic link between these two clinical phenotypes. Okay, so what does this tell us? Well, if you then go back and do a bit more bioinformatics, not surprisingly, you take all of the mutated genes and congenital heart disease, you put them through David and the like, and you'll find that neurodevelopmental is one of the hits of the pathways that are activated. So two, cardiovascular development, again, not a very big surprise. But perhaps the most surprising was that we also had significant enrichment in pathways involved in chromatin modification. And this is where things began to all make sense. Clearly these are molecules that are critical for regulating transcription at the most primordial stage and are involved in morphogenesis of virtually every organ in our whole body. And yet this is exactly where the vast majority of these mutations were identified in the genes in children with both NDD and congenital heart disease. So what's the evidence when you find this in a cohort study that these are pathogenic and not just a variant of unknown significance? Well, from a pragmatic point of view, it's unlikely to have a chance occurrence of de novo recurrent mutations in genes that are genes and residues that are constrained across evolutionary principles and the co-occurrence of severe critical congenital heart disease. We also know that when we look at these, they're highly expressed in the tissues where the malformations arise. They might explain the unexpected phenotypes that we also observed in these children, namely extracardiac anomalies and as well as neurodevelopmental delays. And as well, when we see recurrent hits in the same genes and those children have like phenotypes, I think that adds to the supposition that these are probably pathogenic. Moreover, from a pathway analysis, when we take these proteins and we see that they are de novo mutations and ask do they function in the like pathway they do, indeed it's a very appealing faction observation to me that these chromatin modifying genes now link de novo mutations that cause these heart malformations with the very rare cases of familial congenital heart disease because these turn out to be mutations in cardiac transcription factor genes. I think that this also alludes to the fact that the ability to modify these phenotypes may reflect variations in many of the different molecules that participate in these broad translational and transcriptional pathways. What's the evidence that they're not pathogenic or that they don't fully account for all of these phenotypes? I think it's important to look in the flip side of that coin and I point out to you that of the 69 genes that carried 85 mutations in children with congenital heart disease or in children with neurodevelopmental deficits, the vast majority of these children in particular with NDD did not have severe congenital heart disease. That's not a phenotype you're going to miss when you're evaluating a child with autism spectrum disorder, for example. So too, not all of the children who carried these mutations and had congenital heart disease have any evidence of neurocognitive abnormalities. And so I think that we have to recognize that these mutations are not predictive of the full spectrum of phenotypes that can arise. And I would suggest to you, we've actually known that for a long time. This is a child with Holt-Orham syndrome and this is a Mendelian mutation in the TBX5 gene and we still to this day don't understand about the profound asymmetry of the limb malformations that most Holt-Orham patients exist. And yet we'd like to believe that this child's background genotype is all uniform across all organ systems. So the final thing I want to point out is it's important to recognize given these complexities, can we identify de novo mutation that is actually diagnostic in a single person? And I would suggest to you that that is to me a great challenge and something that we have to be very, very prudent in terms of putting on a patient's chart. Because if you give a parent the identification from clinical treatment that a child has a severe heart malformations and then also predict neurodevelopmental abnormalities that will ensue perhaps lifelong, the ramifications of that association are very, very profound and personal. So how do we get a better handle on it? And I'm going to transition a little bit in terms of the next session after lunch. One of the things that we've been interested in is whether we could build better models to understand the consequences of these mutations. And so we've been doing developmental transcriptional analysis at the single cell level in the mouse heart beginning at E9.5. And suffice it to say it's not sufficient to know that a gene is expressed in the heart. It's important to know at which time it is expressed. So in atrial genes, the blue represents genes that are off. Over time, these atrial genes turn to be, some turn to be turned on. Some are highly on early in development and are reduced later on. And some are increasing all throughout development. So too in ventricular cardiomyocytes, we see waves of development with gene transcriptions from E9.5 in deep purple to the lavender color here at 14.5 and again at P0. But there still is evolution into P21. 21 days after birth, the transcriptome is still continuing to evolve. And so when we think about new models using this information, we'd like to believe that we could begin to incorporate things like IPS cells, which are human. They're not models of a different system. They are derived from patients or are isogenic, which is what we tend to use into the lab. And as all of you know, we can now introduce these mutations. It's easy to say that we have to do a lot to make these begin to look like adult cardiomyocytes. But we now have the capacity to not only have them look morphologically the same, but we can measure force and contraction in a dish. That being said, when we look at the genetic consequences, the transcriptional changes in response to these mutations, we see vast changes in each and every mutation. I won't spend any time about this except to point out one feature. These model systems are a new way to interrogate different mutations in the same gene. I pull your attention to a KDM5 loss of function mutation in a heterozygous state. We've produced that in an IPS cell and turned it into a cardiomyocyte and in a missense mutation in the same gene. And the congruence of transcriptional changes are exactly the same. So while this is all well and good, it's not necessarily going to be the answer for the best model system. This is a linear representation of the transcriptional changes that occur from E9.5 over time. And when you look at mouse cardiomyocytes differentiated from IPS cells or human IPS cell derived cardiomyocytes, we see that these only make it to the very middle stages of embryogenesis. And so they are not going to, in my opinion at this time at least, be a good model for adult onset diseases. Moreover, when we do the same kind of experiment in a mouse that carries a definitive mutation in NKX2.5, which we know is a cause of congenital heart disease, we can see that the mouse cells have a delay in the maturation of their transcriptional profile at E9.5, but slowly over time it catches up. And so cells, alas, are not going to provide, I think, a definitive answer of what is the pathogenic mutation or whether, in fact, over time there can be resolution of the phenotype that may contribute to the human diversity that we, that I've tried to tell you about. So I've tried to tell you a lot about how we use de novo interpretation to inform our insights about new disease genes, and also as well how we can begin to use new strategies to interrogate these. But I would, again, end by telling you it's a long way to take that information in terms of saying that this functional annotation will tell you the outcome of this particular patient. And I'll stop there. Okay. Thank you. Questions? Yes. Sorry. That was great. I wanted to ask you for the de novo mutations that you identified in the CHD cohort. How much overlap is there between a particular gene and the phenotype? So you know, you have all different phenotypes in that collection of patients, and you identified these high-impact variants. So I'm just wondering how much overlap is there in terms of phenotype overall between all the different variants that were recovered? So the one gene that I showed you, the Arbifox mutation that all causes hypoplastic love heart syndrome in four children, that is the only close correlation that we've seen in all of the other mutations, de novo mutations that we've identified. The range of malformations range from conotrunkle to outflow tract obstruction to some of the hetero-taxis. This hetero-taxis are probably with more enrichment of that cohort, possibly defining a genotype-phenotype correlation, largely I suspect because those are recessive mutations. But for the most part, we are seeing no precise correlation. And that speaks to the complexity of these pathways. It isn't just a mutation in one gene, it's the interaction of that mutant protein with the surrounding molecules that are going to modify transcription and translation, I think. So just a comment. So I think, so I really appreciated that you went back and forth and tried to caution, yes, you won't always see these things fully associated, not fully penetrant. But we ran a study for the sequencing centers to look at, across bio-view, just a polygenic predictor of autism, so what other phenotypes would be associated. So people can think about what kinds of phenotypes you wouldn't want to use in controls. And congestive heart failure is the first medical conditions that are associated with autism across bio-view. So I think the possibility that there's more overlap in these genes is a really interesting one. Because these, as the kids age, they'll be at higher risk for cardiac phenotypes potentially because of some of the overlap to the congenital heart defects. I would agree with you, Nancy. I would also say that the highest pickup rate of mutation was not just with NDD, but NDD with any of their morphologic phenotype, an ear, an arm, a spleen. So what that says to me is that there are going to be broad developmental pathways that can be perturbed, but the nuance of which one is perturbed is yet really to be understood at a single-person level. So, Cricket, I was curious about the really elegant work that you did in cardiomyocytes, IPS-derived. Is similar work going on in neurologic cells, which obviously, there isn't as nice a phenotype that you can characterize, but is that work going on, too? Very, very much so. And in fact, I think this is something where scientific groups need to collaborate because we would love to have these same cells differentiated into neuronal cells and understand what are the consequences there and what we can learn from that. I would also say that one of the remarkable progresses of IPS-derived, any differentiated cell type is that it's human. But the barrier of not getting it into an adult state is problematic. Part one. And part two, it is a pure culture. The ability of cells to catch up, in my opinion, may as much reflect the cells that are surrounding that cell as compared to the intrinsic cell machinery and mutations that that particular cell burdens. So, more to do. Okay. Yes. So, we can just open it up for discussion. I think I could go, yeah.