A Change of Heart: Heart Development from Embryo to Adult

Part 1: A Change of Heart: Embryonic Heart Development

00:00:06.19 My name is Deepak Srivastava, and I direct the Gladstone
00:00:09.23 Institute of Cardiovascular Disease and the Roddenberry
00:00:12.12 Stem Cell Center at Gladstone, and I'm a professor at the
00:00:15.21 University of California in San Francisco. Today I'd like to
00:00:18.29 share with you a story that illustrates how we can begin
00:00:22.06 to use knowledge we gained from studying developmental biology,
00:00:26.11 so early embryonic development. To understand the
00:00:30.25 causes of human diseases and even find new therapeutics
00:00:35.04 for human diseases. And I'll largely focus on diseases of the heart,
00:00:40.23 but I'd like to highlight the fact that the principles I'll
00:00:44.20 share with you could be applied to many different organs, and
00:00:48.06 many different human diseases. So I have a background as a
00:00:53.20 pediatric cardiologist, and the disease that I take care of is
00:00:58.11 illustrated here. Which is a child, newborn, who was born
00:01:02.21 with a heart defect, where during embryonic development
00:01:05.23 this baby's heart just failed to form properly in the right
00:01:10.00 3-dimensional shape. And as a result, after birth the child
00:01:16.06 had a number of problems and required very early surgery,
00:01:19.11 as you can see here. And this type of problem is extremely common.
00:01:24.03 Heart defects occur in 1% of all live births worldwide.
00:01:29.11 And despite our best efforts at surgical intervention, which have now become
00:01:34.08 quite good, it remains the leading non-infectious cause of
00:01:37.17 death in the first year of life. So we have to understand
00:01:41.08 this disease better, and in the future, hopefully treat them in a better
00:01:45.29 way, both through preventative measures, but as well as
00:01:49.27 treatments that could change the natural history of this disease
00:01:54.28 over time. Now it's been our hope that if we understood how
00:01:58.22 nature normally forms a heart in an embryo, that it might
00:02:03.01 inform us of this disease. But it also might inform us about
00:02:07.00 a different kind of heart disease. And that's the type of
00:02:09.18 heart disease that occurs in adults, which is very different
00:02:12.18 from what I just mentioned to you. So in adults, heart disease
00:02:16.06 across the world now is the leading cause of death. Just within
00:02:20.25 the United States, there are over 1 million people every year who have
00:02:25.11 heart attacks, and because we're doing better and better with saving
00:02:29.06 those individuals, the problem has been that they are saved,
00:02:33.25 but they have damage to their heart. And as a result of that damage,
00:02:37.09 there are now 5-6 million people living in the United States
00:02:41.03 who suffer from what we call heart failure, which is when the pump
00:02:44.26 just because of the damage can't pump well enough to support the
00:02:48.26 circulation in the body. And if you look worldwide, that number is about
00:02:52.12 24 million. So it's an enormous health problem, and right now
00:02:56.26 we have very few approaches that can address this disease and none
00:03:02.12 that are actually curative. Because at the end of the day,
00:03:04.20 if you've lost heart muscle from damage, the only real way to
00:03:08.25 fix that is to create new heart muscle. So to regenerate the
00:03:12.29 heart, and right now we don't have any ways to do that.
00:03:15.08 And what I'm going to show you is that possibly, by using
00:03:18.18 the clues from nature of how it normally makes a heart,
00:03:22.17 that in this situation we might be able to create new heart
00:03:27.07 muscle for this disease. So the themes of what I'll share with
00:03:32.13 you today are: one, to understand how early cardiac cell fate
00:03:37.29 decisions are made in the embryo as it's first forming.
00:03:41.11 And that involves both understanding signaling pathways
00:03:46.05 that impinge of these decisions, transcriptional networks
00:03:50.01 that establish these very cell fates, and even translational
00:03:54.10 networks that are embedded in these pathways using microRNAs
00:03:58.22 that then titrate the right dose of these networks to exert
00:04:03.10 the proper outcome. And the second thing that we'll talk about,
00:04:07.28 is how we can utilize the knowledge of these networks and
00:04:12.22 overlay them on human genetic models to begin to understand
00:04:16.20 the basis of human disease. And not only understand the
00:04:19.21 genetics of those disease, but also understand the mechanisms
00:04:23.11 using induced pluripotent stem cells in a dish. And finally,
00:04:27.09 I'll share with you how we can use these developmental biology networks
00:04:31.01 to reprogram cells in situ in an organ for regenerative medicine.
00:04:36.11 So all of this work starts by going back to the basics of how
00:04:42.01 does an organ form in utero? And for what I'm showing you here, these are
00:04:47.03 the steps that we know of that are involved in a heart forming
00:04:51.06 early during embryogenesis. And so we know that as early
00:04:54.18 as two weeks after fertilization in the human embryo,
00:04:58.23 there are cells like you see here that begin to align themselves
00:05:02.28 in this crescent shape. And already at this stage, before there are any
00:05:07.08 organs in the body, these cells already know that they're going to become
00:05:12.17 heart muscle in the future. And even at this stage they're
00:05:17.00 already parsed in these pools of progenitors that I've highlighted
00:05:20.23 here in either red, purple, or yellow. And these cells go on
00:05:25.21 to form specific parts of the adult heart. And in particular,
00:05:31.02 you can see that these yellow cells align themselves in this
00:05:36.02 back of this tube that forms at about 3 weeks after fertilization.
00:05:41.07 This tube starts beating, and then these yellow cells come into
00:05:46.13 the top and form what's going to later become the right ventricle
00:05:51.08 of the heart. And the red cells form the left ventricle. So already,
00:05:56.05 we're beginning to see that there are different pools of progenitor
00:06:00.04 cells that will give rise to different chambers of the mature
00:06:03.20 heart. And this begins to explain an observation we've had
00:06:08.12 in pediatric cardiology for a long time, and that is that children
00:06:12.00 are often born where there are defects in just one part of the heart
00:06:15.28 but the rest of the heart is normal. So for instance, a child
00:06:19.14 might be born missing their whole left ventricle, or missing
00:06:23.25 their whole right ventricle and the rest of the heart is fine.
00:06:26.25 So now we can begin to understand why that might be.
00:06:29.18 Because you can imagine that there might be a mutation in
00:06:32.21 a gene that was critical for these yellow cells to form, but not the
00:06:37.08 red or vice versa, and if you have that mutation, you'd lose
00:06:41.03 only that part of the heart but not the rest of the heart.
00:06:44.28 And so we've learned a lot about this process at the
00:06:49.01 molecular level by using a variety of model organisms
00:06:53.02 including fruit flies, zebrafish, chick embryos, and mouse embryos.
00:06:59.18 And collectively in our field, we now know quite a bit about the
00:07:04.07 key genes that drive the events you see on this slide.
00:07:08.00 But animal models turned out to be very difficult to use
00:07:12.03 to get at the earliest cell fate decisions that occur in the
00:07:16.01 first two weeks of life, as I've mentioned earlier. And for
00:07:19.29 understanding that process, we as a field have turned to
00:07:24.14 the pluripotent stem cell system in a dish, where we can
00:07:28.04 take pluripotent stem cells and guide them along a path
00:07:32.02 in a stepwise fashion like you see here. Where they become
00:07:35.29 mesodermal progenitors, and those mesodermal progenitors
00:07:38.23 become these multipotent cardiac progenitors. And we can
00:07:44.08 even mimic the different pools of progenitors in cells in a dish here.
00:07:49.09 So those red and yellow cells we can isolate in a dish, and the
00:07:52.24 yellow cells are very similar to those in vivo, where they are
00:07:56.29 multipotent. They can become cardiomyocytes, endothelial
00:08:01.02 cells, or smooth muscle cells. And so we've utilized this system
00:08:05.29 to work out the intricacies of how these step-wise decisions are made
00:08:11.28 at the earliest time points, to tell an early cell that it's going to
00:08:15.26 become a cardiac progenitor and then a cardiac myocyte. Or
00:08:20.12 an endothelial cell or a smooth muscle cell. And together,
00:08:24.08 by using animal models and stem cells in a dish, we as a field
00:08:30.00 have developed a quite deep understanding of the gene
00:08:33.27 networks that are dictating cardiac cell fate and subsequent
00:08:38.14 morphogenesis. So we know a lot and maybe know most of the
00:08:42.22 players that are critical. And we're still trying to figure out
00:08:46.05 the exact mechanisms by which each of these critical
00:08:50.07 central players are actually exerting their effect. And currently we
00:08:54.25 now have the tools to do that in biology that we couldn't even
00:08:59.09 do actually just a few years ago. So, we've known enough
00:09:04.26 to know what the key players are, and so we've begun to ask,
00:09:08.15 are these gene networks that we've discovered the ones that are
00:09:12.19 in fact disrupted in the setting of human disease?
00:09:15.12 And the answer to that turns out to be as it's evolving,
00:09:19.20 is that yes, these are indeed the key genes that are disrupted.
00:09:23.22 So what I'm showing you here is a subset of the key gene
00:09:27.06 networks that are involved in controlling the yellow cells,
00:09:30.19 red cells, blue cells at the top, and most of the genes on this
00:09:35.14 slide are transcription factors that are playing key roles.
00:09:39.23 And we can begin to arrange them in a hierarchical network.
00:09:43.02 Like what you see here. And what's been satisfying is that each
00:09:47.22 of these gene networks is hit one or more times, with mutations
00:09:53.14 in humans. And I've indicated those mutations -- those genes that are mutated
00:09:57.07 by asterisks in this slide. And what I want to emphasize
00:10:03.25 here is that in every one of these cases, the gene mutations
00:10:07.11 are actually heterozygous mutations. So only one of the two
00:10:12.01 copies in the genome are mutated, and in many cases, the one
00:10:15.29 that's mutated is not even a complete loss of function mutation.
00:10:19.19 But maybe only a single amino acid has changed, so it's a hypomorph.
00:10:24.06 We've only reduced the dosage of that key gene in its network
00:10:28.27 a little bit. But that's enough to get over the disease threshold.
00:10:34.02 And so what that tells us that the dosage of these networks is
00:10:37.25 exquisitely controlled in a precise fashion. But it also
00:10:43.06 suggests that if we can just increase the dosages of these networks a little bit,
00:10:47.26 we might someday get over the disease threshold. And so
00:10:52.05 the hurdle might not be so high, to prevent some of these diseases.
00:10:57.00 I just hope that our understanding of this process might
00:10:59.27 actually lead to some preventative measures in the future.
00:11:03.20 Now our laboratory has, in specifically, identified years ago,
00:11:07.23 mutations in a very important developmental gene called
00:11:11.10 Notch1, that causes valve disorders. And we've also identified
00:11:15.22 mutations in another transcription factor called Gata4.
00:11:19.27 Which is very important for formation of the septum of the heart.
00:11:23.22 So the walls that divide the different chambers of the heart.
00:11:26.15 And I'm going to spend a little more time sharing with you
00:11:29.29 some of the things we've learned recently about the mechanism
00:11:32.27 by which this heterozygous mutation actually causes
00:11:36.29 human disease. Because at the end of the day, just identifying
00:11:40.20 the genes that are mutated doesn't get us that much further to a
00:11:44.16 therapy, we have to understand the mechanism by which
00:11:47.15 these gene mutations cause disease. Before I do that, I want
00:11:54.01 to also highlight the fact that we are now able to
00:11:58.28 generate exomes of thousands of individuals with disease.
00:12:04.17 And in an effort to a National Heart Lung and Blood sponsored
00:12:09.14 consortium, we and others have found that there are in fact
00:12:13.12 a number of genes that are involved in congenital heart disease.
00:12:18.02 And in most cases, those mutations are also heterozygous
00:12:22.00 mutations. And so, from the mutations I described to you before,
00:12:25.22 and the ones that we're beginning to find now through exome
00:12:29.09 sequencing. As I mentioned, if we really want to understand the
00:12:33.15 mechanism in a human cell, we've got to have human
00:12:36.17 cells in front of us to study. And for many years, we tried
00:12:40.07 to understand the mechanisms by making mouse models
00:12:42.22 of these mutations. And we learned some things, but at the end
00:12:46.11 of the day, most mice that were heterozygous for these
00:12:49.17 mutations were fine. The threshold in mice turned out to be very different
00:12:54.00 than in humans. And so the key advance that has allowed us
00:13:00.18 to now begin to understand mechanism has been the discovery
00:13:04.00 by our colleague here, Shinya Yamanaka, of induced
00:13:07.28 pluripotent stem cells. So Dr. Yamanaka found that we could
00:13:13.09 take fibroblasts or blood cells from an adult, and introduce
00:13:18.12 just four critical genes. And those genes were enough
00:13:22.26 to reprogram those cells into something that looked and behaved
00:13:26.18 just like a human embryonic stem cell, in that they were
00:13:30.03 pluripotent. He termed these iPSC cells, or induced pluripotent
00:13:34.12 stem cells. And this discovery opened up the possibility that we might
00:13:40.02 be able to take stem cells and make stem cells from a patient,
00:13:43.17 and use those cells for regenerative medicine by transplanting
00:13:47.19 them back into the patient because they'd be genetically identical.
00:13:50.19 And there are many efforts to do that, but that's still
00:13:54.02 years away. But what is happening now with this technology
00:13:57.22 and I'll share with you an example of this, is generating
00:14:01.05 these cells from an individual with a genetic mutation, and
00:14:05.13 then turning those pluripotent cells into the cell type
00:14:08.26 that you want to study the disease. And then you can now have
00:14:13.13 billions of cells in front of you that you can now study that has
00:14:17.07 the genetic mutation that's causing disease in the patient, and now
00:14:21.01 you can finally begin to interrogate those cells. And really
00:14:24.14 get at the nitty gritty of why did that mutation in that patient
00:14:28.00 cause a problem in the cells that are now suffering from
00:14:31.20 the disease. So this was a real breakthrough that allowed us to make
00:14:35.08 the progress that I'll share with you. So if you want to
00:14:38.20 do that for any cell type, the first thing you have to do is be able to
00:14:41.15 generate large numbers of the cells that are affected by disease.
00:14:45.12 So what you're seeing here is a sheet of cells that used to be
00:14:51.02 on somebody's skin. But now have been transformed into a
00:14:55.26 syncytium or a sheet of beating cells that are all beating
00:15:00.04 synchronously. And these are human iPSC derived
00:15:03.20 cardiomyocytes. We've become very good at making these, and we can make
00:15:07.28 billions of these cells with 90% pure cultures of these cells.
00:15:14.28 And so we've done this now for the human disease that I mentioned to you
00:15:20.03 where we've had this mutation in this critical gene, GATA4, that's
00:15:26.03 absolutely essential for normal development in the heart and
00:15:29.25 embryo. And so what you see here is the family that we
00:15:33.27 took care of and studied some years ago that spans 5 generations.
00:15:41.03 Everybody who has this single amino acid switch from a glycine
00:15:44.06 to serine at position 296 in GATA4, which is a transcription
00:15:49.00 factor, has the disease. And while we had identified this
00:15:54.09 gene mutation many years ago, this is an example of where
00:15:57.29 we got stuck by just studying the mice. Because the mice
00:16:02.04 that are heterozygous were pretty much okay and didn't
00:16:05.21 have a lot of the disease manifestations. But what we did
00:16:09.20 think a few years ago was that this point mutation might be
00:16:14.16 causing a problem in the ability of this protein to interact
00:16:18.23 with another protein called TBX5. And the reason we thought
00:16:23.18 that might be the case is because humans with mutations
00:16:25.25 in TBX5 have the same problem that humans with GATA4 have.
00:16:30.21 Which is that they get holes in their heart, or septal defects.
00:16:34.24 And the amino acid that's affected in this protein is shown here,
00:16:40.06 and it sits right next to a zinc finger in GATA4 that's very
00:16:45.27 critical for protein-protein interactions. And so when we made
00:16:51.04 this mutation of GATA4 and overexpressed this mutant
00:16:56.21 protein in a non-cardiomyocyte years ago, we could show that
00:17:00.24 GATA4 doesn't in fact interact with TBX5, and this mutation
00:17:05.19 breaks apart that interaction. But what we didn't know
00:17:08.22 is does that really happen on DNA in a human cardiomyocyte
00:17:13.14 when they're at their normal levels. And what I'll show you is
00:17:17.28 in fact they do, and we've been able to experimentally show that
00:17:20.20 and understand why that would cause trouble.
00:17:23.01 So, this is a heart from one of the family members,
00:17:28.13 that I showed you before, that does not have the GATA4
00:17:32.01 mutation, it's like a normal heart. It's an ultrasound where you
00:17:35.08 see the left ventricle is here, the right ventricle is here. And
00:17:40.04 these chambers are pumping normally. What we found is these patients
00:17:44.13 with the GATA4 mutation not only had holes in their heart,
00:17:48.01 but as some of the kids got older, they ended up having a different problem.
00:17:53.06 And here is an ultrasound of one of those kids who's now a teenager.
00:17:58.01 And what I'd like you to appreciate is here in the right ventricle,
00:18:02.15 you see a lot of white stuff here. That white stuff should not
00:18:06.08 be there. It should be black, which is meaning an empty
00:18:08.29 cavity full of blood. Instead, there's a lot of muscle fibers
00:18:12.28 that are invading into the cavity of this ventricular chamber.
00:18:17.13 And the heart isn't pumping as hard as it should.
00:18:20.23 And so, this is a problem that's often thought of as a problem
00:18:25.00 reflecting immaturity of the muscle cells. And so we thought
00:18:29.21 we might actually be able to see if that's actually happening
00:18:32.09 in their iPSC-derived cardiomyocytes. So, we've generated
00:18:37.29 iPSC cells from these family members that are shown here
00:18:42.07 in 8 individuals, four with the mutation and four without.
00:18:44.22 And we took skin biopsies from these individuals, transformed
00:18:49.10 them into stem cells, and then into beating cardiomyocytes
00:18:53.29 like I showed you earlier. And when we made these cells,
00:18:58.24 which were done by two talented trainees in the laboratory,
00:19:02.07 Yen-Sin Ang, a postdoc, and Renee Rivas, an MD PhD student.
00:19:06.05 They made these iPSC cells and looked at the RNA of
00:19:11.04 these cardiomyocytes. And it turns out, if you make iPSC
00:19:17.13 derived from all 8 of these individuals and try to compare
00:19:20.24 mutant ones with the mutation and ones without, there's a lot
00:19:24.10 of noise in the system. Because each one is a little bit different.
00:19:27.14 And so, the real breakthrough that allowed us to get to the next level
00:19:31.25 came with the advance in technology of gene editing.
00:19:36.21 So, by using CRISPR technology, we could go into these
00:19:41.14 iPSC cells and just in the patients who have the mutation,
00:19:45.23 correct that mutation so that everything else in the genome
00:19:48.27 is the same except for this one nucleotide change.
00:19:52.27 And so we called these isogenic controls, and now when
00:19:56.10 you compare the isogenic controls with the ones that
00:19:59.15 have a mutation, and everything else is the same, then
00:20:02.18 everything displayed itself because the noise went away.
00:20:05.26 And so, we were then able to figure things out that I'll show
00:20:10.08 you in the next few slides. So the first thing we wanted to know
00:20:13.00 was can we in fact recapitulate the problem of the muscle
00:20:17.13 being immature and having trouble contracting in these cells
00:20:22.13 even though they started on somebody else's skin?
00:20:24.06 And to do that, we used a system we developed in collaboration
00:20:28.24 with Beth Pruitt's lab, where we can grow these cardiomyocytes
00:20:36.26 not in the dish like the culture I showed you before, but
00:20:40.08 rather have each muscle cell in an individual well by itself
00:20:45.03 in these patterned microwells that are indicated here. So here,
00:20:50.17 each of these colored areas you see is a single cell in a well.
00:20:54.21 And these cells can beat and we can actually measure their
00:20:59.06 force of contraction as individual cells. And so we've done that
00:21:03.29 for a number of cells, so in this graph, each of these dots
00:21:08.02 represents a single cell and the force they generate.
00:21:11.09 And in the red cells are the ones with the mutation compared
00:21:14.22 to the ones without. And you can see that the force generation
00:21:18.03 by these mutant cells is less than the ones that don't have
00:21:22.20 the mutation. So in fact, we're able to recapitulate some aspect of this
00:21:26.29 cardiac dysfunction in this artificial system. So we then looked at
00:21:33.12 the transcriptome of these cells, since this is a transcription factor
00:21:37.17 mutation, we wanted to know how the transcriptional output was different.
00:21:41.24 So what you're seeing here is a heat map of the genes
00:21:45.21 that are downregulated with this mutation and those that are
00:21:48.27 upregulated. And what you'll notice is that there are a number
00:21:52.25 of genes, about a third of the genes were downregulated.
00:21:55.05 And if you look at the GO terms, of the gene ontology
00:21:58.07 terms, they characterize the genes that are downregulated.
00:22:02.12 It looks like this. It's enriched for genes involved in heart
00:22:06.09 development, cardiac chamber morphogenesis, contraction.
00:22:10.12 So all the genes that should normally be turned on so this
00:22:14.03 muscle cells knows it should beat vigorously and turn into
00:22:17.14 a heart are decreased. And in contrast, if you look at the
00:22:22.29 even larger number of genes that are apparently upregulated,
00:22:26.15 and look at the GO terms there. This was really quite interesting.
00:22:30.19 There are a whole host of genes that are involved in vascular
00:22:34.03 or endocardial development that are now up that should have
00:22:37.15 been shut off. And so, somehow just because of this single
00:22:41.29 heterozygous point mutation, there's a whole host of genes
00:22:45.02 that should've been shut off that now aren't. So why is
00:22:48.27 that? So, to answer that, we began to look at where
00:22:54.27 do these genes TBX5 and GATA4 normally sit, and
00:23:00.21 how might that disruption actually cause the problem?
00:23:03.25 And so what I'm showing you here is the ChIP, chromatin
00:23:08.06 immunoprecipitation, or ChIP-seq sequencing data of GATA4,
00:23:13.11 where it sits on DNA, TBX5 and where it sits on DNA.
00:23:17.23 And a histone mark, Histone 3 K27 acetylation, which
00:23:21.29 marks active enhancers. And what you'll see is that in
00:23:27.13 these areas here that are blue, which indicate where GATA4normally sits on DNA, which each horizontal line representing
00:23:35.05 a single genomic locus where GATA4 is binding to the DNA.
00:23:39.27 If you look at the comparable areas where TBX5 sits,
00:23:43.24 about half of the places where GATA4 sits across the whole
00:23:46.24 genome, it's sitting there with GATA4. So these are really
00:23:51.15 obligate partners throughout the genome. And it's
00:23:56.06 essential for those to interact at those sites. And if we look
00:23:59.21 at this more broadly and look at the sites across the
00:24:03.04 genome where these two proteins are co-bound.
00:24:05.28 In the wild type setting, there are about 2500 spots.
00:24:11.11 And it turns out, that in the setting of this point mutation, about
00:24:17.21 half of those, indicated in blue here, are lost.
00:24:20.24 So with this mutation, at about half the sites where GATA4
00:24:24.14 normally sits on the DNA with TBX5, its partner, it no longer
00:24:28.28 can, because of this point mutation. And if you go a step
00:24:32.23 further and ask, what are these genes where this interaction
00:24:36.25 is lost? What types of genes are they? And look at the GO
00:24:40.20 terms of these, again, it's those genes that are most
00:24:44.22 important for contraction of the heart muscle, for the heart
00:24:48.14 to form, for general heart defects to occur. And so this begins to explain
00:24:53.11 why this point mutation in GATA4 is actually causing the
00:24:59.28 cardiomyopathy and the congenital heart defects at a very
00:25:03.03 fundamental level of its inability to now control transcription
00:25:08.09 in the proper way in the human genome. So this is an example
00:25:12.17 that the power that this approach can bring to understanding the
00:25:16.08 mechanism by which this disease occurs. Now, I've told you
00:25:22.19 so far about how we can use this to understand why a
00:25:25.17 muscle cell might not contract normally. But what about
00:25:28.11 the 3-dimensional aspect of this disease, where you have
00:25:31.26 holes in the heart? When we set out, we thought this might be
00:25:36.04 really difficult because it's a 2-dimensional system of these
00:25:40.04 iPSC-derived cardiomyocytes, yet we're trying to model a
00:25:42.27 3-dimensional -- to understand a 3-dimensional problem.
00:25:44.26 What I'm going to share with you in the next couple of slides
00:25:48.02 is how we believe that we've actually been able to
00:25:52.16 decipher the mechanism of this 3-dimensional defect, despite
00:25:56.06 having a 2-dimensional system, which we think bodes well for
00:25:59.29 other form and other types of disease modeling for genetic
00:26:03.27 defects. So to do that, I first need to give you a little bit of
00:26:07.00 background in one aspect of developmental biology.
00:26:10.17 And that is that early during development, there is a
00:26:14.27 signal that comes from the pulmonary endoderm, which when
00:26:20.01 the heart's forming early in an embryo, this pulmonary
00:26:23.00 endoderm sits right behind the developing heart, right next to it.
00:26:27.04 And it secretes a very important morphogen called
00:26:32.10 Sonic Hedgehog, which is involved in development of many
00:26:35.25 different organs. The neighboring cardiac muscle has to
00:26:40.29 receive that Sonic Hedgehog signal and in response to that
00:26:44.22 Sonic Hedgehog signal, actually forms this atrial septum that you see here.
00:26:50.06 And the Sonic Hedgehog signal is received by a receptor
00:26:56.08 on the muscle cells called Patched, and it leads to a series
00:27:00.09 of intracellular events that are mediated by a transcription
00:27:05.01 factor called GLI in the nucleus. And that then helps the muscle
00:27:10.29 cell respond to this Sonic Hedgehog signal, and then grow
00:27:14.18 this septum. And we know that if you delete Sonic Hedgehog
00:27:19.12 from the pulmonary endoderm, you end up with mouse hearts
00:27:23.23 that don't have a septum. And if you delete GLI from
00:27:28.13 the cardiac muscle cells, you also get holes in the heart
00:27:32.08 in a mouse. And the reason I'm telling you this is that
00:27:35.06 it turns out that in our iPSC-derived cardiomyocytes
00:27:40.04 with this GATA4 mutation, we found that Sonic Hedgehog
00:27:44.21 and the machinery in the cardiomyocyte to receive this
00:27:47.26 Sonic Hedgehog signal was downregulated broadly. And in particular,
00:27:52.03 the receptor PTCH1 and PTCH2 were severely
00:27:56.27 downregulated. And the transcription mediators, GLI2 and GLI3
00:28:00.12 were also downregulated. So what we think is happening
00:28:04.04 here is that in the setting of this mutation where GATA4
00:28:07.24 and TBX5 should normally sit at those regulatory regions
00:28:12.11 to turn on PTCH and GLI, they don't. And so now you get
00:28:16.21 downregulation. And so even though the Sonic Hedgehog
00:28:19.08 signal is coming from the developing lung, it can't be
00:28:22.28 received properly in the developing heart. Thus resulting in
00:28:26.27 this defect. And so, what we think is going on in this
00:28:33.10 setting based on what we've learned by going from the
00:28:36.05 developmental biology, understanding the genetic mutation, and
00:28:39.28 then modeling in iPSC-derived cells, is that normally GATA4
00:28:44.26 and TBX5 have to broadly across the genome sit at parts of the
00:28:48.26 DNA to robustly turn on cardiac genes. And at the same time,
00:28:54.25 they have to sit at other sites across the genome, like
00:28:59.17 endothelial or cardiac genes, and actively shut off those regions
00:29:04.03 by recruiting co-repressors. In data that I didn't show you, what we
00:29:07.22 found is that this is an active process and that has to happen
00:29:11.12 for normal development. And in the setting of a human
00:29:15.22 condition, where all you've done is made this protein
00:29:19.07 unable to recruit its cofactor, that now we see a situation
00:29:24.09 where there's aberrant downregulation broadly of cardiac
00:29:27.09 genes and aberrant upregulation of other genes that
00:29:32.18 should be shut off and now aren't shut off. Thus, affecting
00:29:35.14 the function and behavior of those muscle cells.
00:29:38.22 So, what I hope you can tell from this, and this example
00:29:42.07 could be applied to many, many diseases, is that by
00:29:45.07 deeply interrogating a human cell in a dish that has the mutation
00:29:50.19 that causes disease and using gene editing approaches,
00:29:54.01 we can really get at the nitty gritty of why this disease
00:29:57.24 occurs. And now we have cells in the dish that we can use
00:30:00.19 to screen for drugs that actually might reverse the problem
00:30:04.25 that we've identified. And we're actively doing that now.
00:30:07.25 So this is the end of the first part of my talk, related to
00:30:13.01 how developmental biology can be used to understand the
00:30:16.16 human disease. And in the second part, I will talk about
00:30:19.24 how we can use developmental biology gene networks
00:30:22.19 for regenerative medicine. Thank you very much.

Part 2: A Change of Heart: In vivo Cellular Reprogramming

00:00:06.14 My name is Deepak Srivastava, and I direct the Gladstone
00:00:10.05 Institute of Cardiovascular Disease and the Roddenberry
00:00:12.28 Stem Cell Center at Gladstone, and I'm a professor
00:00:15.08 at the University of California in San Francisco.
00:00:17.04 This is the second part of my talk, and in this part, I'll share with you
00:00:23.19 how we can use the developmental biology gene networks
00:00:27.21 that we've discovered to impact not just our understanding
00:00:33.08 of birth defects, but also how we can use that for regenerative
00:00:37.24 medicine approaches through cellular reprogramming.
00:00:41.21 Can we use these developmental pathways we've discovered
00:00:47.01 to somehow take an adult heart and create new muscle in a
00:00:51.25 regenerative approach? And it turns out that the heart, like
00:00:56.28 most other organs, has not only muscle cells but a number of
00:01:01.29 support cells. In the heart, we call them cardiac fibroblasts,
00:01:05.18 but most other organs in your body have support cells around
00:01:10.02 the functioning cells that are there to form the architecture
00:01:13.19 and support the cells that are necessary for that organ to function.
00:01:17.07 The heart, it turns out, is quite special in this in that fully
00:01:21.29 half of the cells in the organ are of this type of cardiac fibroblast.
00:01:25.17 And those are the same ones that are activated in the
00:01:28.29 site of injury and make scars. So what you see here is a
00:01:33.01 heart, this is in the case of a mouse, but a human heart is very similar.
00:01:36.27 If it's been damaged in say a heart attack, where in red you see
00:01:40.29 normal muscle, and here in the left ventricle, you see this
00:01:44.12 broad purple area which is scar tissue that is present
00:01:47.17 after an injury to the heart. Now if we wanted to repair
00:01:53.20 this heart, you can tell that the fundamental problem is that we lost
00:01:57.20 muscle. We'd want to create new muscle there somehow.
00:02:01.06 And we and others are trying approaches to take the beating
00:02:03.23 heart cells that I've shown you before from stem cells and
00:02:06.22 somehow inject them into this heart to create new muscle.
00:02:11.01 It's turned out that we've had a lot of trouble there in terms of
00:02:14.15 getting those heart cells to stick and persist. And even if we
00:02:19.05 do that, to get them to electrically connect with their neighbors
00:02:22.09 so that they can beat in synchrony. So then years ago, we
00:02:25.15 asked could we take a different approach? Could we somehow
00:02:29.06 harness the fact that we have this abundant pool of
00:02:33.01 fibroblasts that are already in the organ. Could we somehow
00:02:36.26 trick those cells into not being fibroblasts, but somehow be
00:02:41.29 turning into new myocytes or reprogram them into new
00:02:46.21 myocytes. And this is really inspired by Shinya Yamanaka's
00:02:50.28 discovery that you could take a skin fibroblast and turn it
00:02:54.06 into a stem cell. So we thought maybe with the right cues we could
00:02:57.27 take this cardiac fibroblast and right there within the organ,
00:03:01.21 convert its fate into a new cardiomyocyte, if we knew the right
00:03:05.20 factors to induce that cell-fate switch. So, the question as illustrated
00:03:11.19 here. Can we take a cardiac fibroblast and convert it into
00:03:15.12 what we call an induced cardiomyocyte-like cell?
00:03:18.20 And to do this, we really relied upon all of these developmental
00:03:23.25 biology networks. Because we thought we'd learn from nature
00:03:25.29 -- how nature normally does it, maybe we could redeploy
00:03:29.08 the key factors in a dish. And that turned out to be the case.
00:03:33.14 When we screened a large number of transcription factors and
00:03:37.21 microRNAs that we knew to be critical for development of the heart,
00:03:41.18 and asked what was the minimal essential cocktail, it turned out
00:03:46.01 to be the three factors that I'm showing you here. Two of which we've
00:03:49.07 talked about extensively already, Gata4 and Tbx5. And maybe that
00:03:52.29 makes sense now, because of what I've showed you. That these
00:03:56.00 two proteins sit together to drive cardiac cell fate during
00:04:01.00 embryogenesis. And a third really critical factor, Mef2c.
00:04:04.22 That's also essential for normal cardiac formation in an embryo.
00:04:08.15 So it turned out that these three factors in a dish could push cells
00:04:13.07 towards this cardiac fate, somewhat inefficiently, but broadly
00:04:17.21 reprogram these cells. But what we really wanted to do
00:04:20.16 is know if these three factors could do the trick in the cells that are
00:04:24.26 already in the heart. And it turns out that they can.
00:04:28.05 So, if we take a mouse heart and tie off the coronary artery
00:04:33.07 and induce a heart attack, and do so in a mouse where we can
00:04:36.21 genetically label with a fluorescent protein the fibroblast
00:04:40.08 pool and its progeny, we could then track whether or not
00:04:43.27 we've really converted a resident fibroblast into a new myocyte.
00:04:48.13 And it turns out that we can not only do this, but we can do it in a way that
00:04:52.28 new myocytes that are formed electrically couple with their neighbors,
00:04:56.13 they're most similar to an adult ventricular cardiomyocyte.
00:05:00.21 And most importantly, we were able to show that these new
00:05:04.24 myocytes are sufficient to actually improve the cardiac output
00:05:09.06 in these mice after damage by MRI. And if you look at these
00:05:13.25 hearts, histologically, where we've done the experiment
00:05:16.08 that you see here and then wait about 3 months, and take these
00:05:19.21 hearts out and section them.. Normally, after injury, if we just
00:05:24.17 introduce with the virus these control genes a dsRed
00:05:29.00 marker, we see a ventricle like what you see here, with
00:05:32.01 a big large scar on the free wall of the ventricle, that's indicated
00:05:36.05 in blue here. In contrast, when we put these three genes
00:05:40.03 in, we still see a thin-walled ventricle, there's scar tissue.
00:05:43.19 But now you can begin to see these areas of red, threads of
00:05:48.15 myocardium in the scar area. And if you look by this yellow
00:05:52.03 fluorescent protein, these all used to be non-myocytes that have converted
00:05:56.24 into new myocytes. So this was encouraging, but we knew
00:06:01.22 that this process with just these three genes was relatively
00:06:05.12 inefficient in vitro. And we knew that we had improvement even
00:06:09.18 in vivo. And so we embarked upon a chemical screen, where we
00:06:16.00 -- Tamer Mohamed, who has a pharmacist background, is a postdoc
00:06:21.22 in the lab who screened over 5,000 small molecule compounds
00:06:24.22 to ask could any of them make this process more efficient?
00:06:29.07 And it turns out that he identified about 20 different chemical
00:06:34.11 compounds that could. If we measured the number of cells that
00:06:38.24 were turning on a cardiac specific reporter, in this case alpha-myosin heavy chain.
00:06:43.15 At about two weeks after introducing these genes. And you can
00:06:47.08 see here that the number of hits range here from fold
00:06:50.07 increase from 2 to 7 fold. And what we're encouraged by is
00:06:55.16 about four of these 19 hits all affected the TgfB signaling
00:07:03.09 pathway. And about four of them affected the Wnt signaling pathway.
00:07:07.22 Both of them are very important signaling pathways in various
00:07:11.24 embryological pathways. So the fact that we're hitting the same
00:07:16.12 pathways multiple times encouraged us that these might be
00:07:19.25 true hits. And in fact, we showed that these small molecules
00:07:23.15 were working specifically through these pathways.
00:07:27.20 So it turns out that these small molecules not only increased
00:07:32.06 the number of cells that could convert, they also accelerated
00:07:36.09 the pace of reprogramming. So normally, in a dish,
00:07:39.28 it takes about 6 weeks for these cells to convert their fate all
00:07:43.29 the way to a beating cardiomyocyte, and with the TgfB
00:07:48.06 inhibitor, this process was accelerated so that it happened within
00:07:51.22 about 3 weeks. And most remarkably, when we added both
00:07:56.21 inhibitors, we could see beating cells in a dish as early as
00:07:59.28 just one week after introducing the genes. So this is really quite
00:08:04.26 remarkable, if you think about the wholesale epigenetic
00:08:07.22 shift that has to occur for a cell to go from one cell type
00:08:11.11 to another, just within a week and induce this process. And
00:08:16.11 we have interrogated that process very carefully. And we're beginning
00:08:20.01 to see how the epigenome changes, and it all happens
00:08:24.20 within the first few days in this setting, which is really quite remarkable.
00:08:28.00 So, we wanted to really know can these chemicals
00:08:33.08 improve the ability of these genes to reprogram in vivo. And
00:08:37.06 improve the function. And it turns out, if we repeat the experiment
00:08:40.24 that I showed you earlier, and now do it by injecting not only the
00:08:46.05 genes with the virus into the heart, but also treat the animals
00:08:50.28 with these chemical inhibitors of Wnt and TgfB
00:08:54.09 for a couple of weeks, now we see much more robust
00:08:58.08 formation of new muscle. So what you see here is the
00:09:01.21 control setting, where at the apex of the heart down at the bottom
00:09:05.13 of the heart, if you section you see a lot of scar tissue.
00:09:08.14 And that gets a little bit less as you get higher up in the ventricle.
00:09:11.07 In comparison, in this setting where we've put both genes and compounds,
00:09:16.10 even at the apex, all of this region right here that you see is all
00:09:20.13 muscle. IR reporter are all cells that used to be non-muscle, so we're getting
00:09:26.16 a lot of more muscle formation, about 5 times as much
00:09:29.22 as without the compounds. And if we look at the function
00:09:34.05 of the heart by measuring what we call the ejection fraction, which
00:09:37.29 is the fraction of blood squeezed out of the heart with every beat.
00:09:41.21 And we measure the decline in the ejection fraction over time,
00:09:46.01 by doing ultrasounds on a weekly basis, what you can appreciate
00:09:50.18 in the red and green lines is the decline that we saw in either just
00:09:55.07 a control virus or control virus plus the chemicals alone,
00:10:00.04 which didn't make a lot of difference. The purple line shows you the
00:10:03.02 change in ejection fraction with just the genes, which is better.
00:10:06.17 And then the blue line is the change with the genes plus these
00:10:11.17 chemical inhibitors, which you can see is significantly better
00:10:14.15 even with the genes alone. And it happens very quickly,
00:10:17.12 as early as a week, consistent with what we found
00:10:20.15 -- the rapidity of reprogramming. So we're hopeful that now
00:10:25.14 we've got a better combination that can actually improve the
00:10:29.07 heart function even more significantly in this setting.
00:10:31.23 But what I've showed you so far is just the mouse, and
00:10:35.01 obviously we'd like to translate this toward the human
00:10:38.21 condition. And so, it turned out that the three genes I mentioned to you in mouse
00:10:43.29 actually did nothing in a human cardiac fibroblast. So we had
00:10:48.07 to go back and ask, what are we missing? And it turns out,
00:10:50.26 that the addition of two more genes were enough, and again these
00:10:55.01 are transcription factors also, to reprogram a human cardiac
00:10:59.09 fibroblast to be more like what we saw in the mouse, but in a human
00:11:03.26 induced cardiomyocyte. But in mouse, we had see that the
00:11:07.27 efficiency in vivo was a lot better than in vitro, and so
00:11:12.00 we pushed forward with trying to test this combination
00:11:14.19 in a heart that's larger like the human, namely the pig.
00:11:19.24 And a pig's heart is physiologically more similar to the human.
00:11:23.14 And the three mouse genes alone didn't do the trick in
00:11:27.18 human cells either, so we thought this was a good system to
00:11:30.23 know if this might work in human hearts. And so, so far what
00:11:36.19 we've done is used this retrovirus, which only infects dividing cells, and
00:11:41.25 we haven't done it yet with the chemicals. But what we have
00:11:45.12 done is the following. We've taken pigs and blown up a balloon
00:11:49.17 in their left anterior descending artery, which is the one that's commonly
00:11:53.20 occluded in humans in an infarct. We've created a big
00:11:58.07 infarct and waited three days, put these pigs into an MRI machine,
00:12:02.14 to determine the degree of cardiac dysfunction that we
00:12:06.21 were left with. And then two days later, we've treated these
00:12:10.02 pigs with gene therapy using a retrovirus that infects dividing
00:12:16.02 cells, the fibroblasts namely and not the myocytes. And then waited two
00:12:20.00 months, put them back in the machine, measured if there's any
00:12:24.04 improvement in their cardiac function. So this we did a bit to mimic
00:12:28.23 the human condition, and what you're seeing here are the results
00:12:33.19 of that experiment in a group of about 5 pigs treated with
00:12:37.23 the factors, and the four pigs with the controls. And we haven't
00:12:43.09 done it yet with the most robust effort with the chemicals,
00:12:47.20 or with the really proper gene therapy vector. But despite
00:12:51.14 that, what we're seeing is if you compare the change in
00:12:56.08 ejection fraction from the point where there's been damage,
00:12:59.15 and compare every pig to itself every time. What we can see
00:13:03.11 is that there's a significant difference between the treated pigs
00:13:06.24 and the control pigs over that two month period.
00:13:10.20 So this type of difference of about 7% in the absolute
00:13:16.04 ejection fraction is significant, it's about a 25% relative
00:13:19.26 improvement from where they started. And we think that we can actually
00:13:23.19 even do much better than this by testing with these chemical inhibitors
00:13:27.29 that I mentioned. And with the proper vector that can get into
00:13:31.02 more of the cardiac fibroblasts. So, we've tested whether or not
00:13:37.14 these chemical inhibitors I mentioned actually help the human
00:13:41.08 cardiac fibroblasts reprogramming, and it turns out they
00:13:43.18 do. And not only do they help the reprogramming like they
00:13:48.01 did in mouse, it turns out it's simplifies the gene cocktail
00:13:52.07 that's required to only the three genes you see here
00:13:55.11 that can give us pretty good reprogramming with these
00:13:59.09 human cardiac fibroblasts that you see have these nice
00:14:02.00 patterns that we call sarcomeres in these cells.
00:14:05.12 So, the reason this is important is that with this combination
00:14:09.28 we're getting down to a small enough number of genes
00:14:13.17 that you could imagine packaging those genes in a gene
00:14:18.06 therapy vector that's non-integrating in the genome, which
00:14:21.26 would be the most clinically viable approach. And that approach would involve
00:14:25.21 an adeno-associated vector. Now, the problem with adeno
00:14:31.23 associated vectors turned out to be that if you wanted to infect a
00:14:35.16 cardiac myocyte, we actually have great AAVs to do that.
00:14:40.11 But if you want to infect a cardiac fibroblast, which would be
00:14:44.05 the target in this therapeutic approach, none of the currently
00:14:47.19 available AAVs do that very well. And so to overcome this
00:14:52.20 obstacle, we took advantage of the fact that our colleague, David Schaffer at
00:14:57.05 Berkeley, had developed this error prone AAV library, where
00:15:02.06 there are millions of random mutations that are going on
00:15:05.02 in these AAVs. So you can choose from those millions
00:15:10.17 and try to identify one that have tropism for a human, and in
00:15:16.04 our case pig, fibroblast over a human cardiac myocyte. So the way
00:15:20.28 we did that is we did rounds of selection, asking are there any
00:15:25.24 AAVs in this pool of millions that can efficiently infect human
00:15:30.29 cardiac fibroblasts? And then we take those and we ask, to remove
00:15:35.29 those that could efficiently infect a human embryonic stem cell
00:15:39.19 derived cardiac myocyte. And we did this round of selection
00:15:42.26 about seven times. And finally came up with an AAV that
00:15:47.14 had in fact greater tropism for a human cardiac fibroblast.
00:15:51.17 So then the question is, can this vector actually express
00:15:55.20 high enough levels of these reprogramming factors to
00:15:58.21 actually convert the cell. And so we cloned in the genetic
00:16:04.22 material into this vector, put them into a human cardiac fibroblast.
00:16:08.11 And what you see here is a cell that used to be a fibroblast, and I think
00:16:13.20 you can appreciate is that it's vigorously beating just one week
00:16:16.18 after introducing these genes. So we feel like we've now
00:16:21.04 got a good cocktail that can be used to induce cell fate conversion.
00:16:27.04 We've got a good vector to deliver these, and we're ready
00:16:30.27 now to do the final trials in large animals before moving
00:16:35.14 this technology towards a human clinical trial. And there are many hurdles that
00:16:39.17 have to be overcome still, in terms of safety issues, and delivery.
00:16:42.28 But we can begin to see a line of sight towards the clinical
00:16:47.17 application of this reprogramming technology. And this notion
00:16:51.21 of harnessing cells in your body to reprogram them to the type
00:16:56.17 that you want is really not unique to the heart, but in fact,
00:17:00.19 could be utilized for many different organs in the body, because all
00:17:04.13 we have to do is be able to crack the code, if you will, for
00:17:08.07 what are the minimal essential genes that will dictate a
00:17:11.28 cell fate. And then introduce that in the proper organ, in
00:17:16.18 that setting. And others have now begun to show this for a
00:17:21.24 variety of different organs, including cells in the brain, the liver,
00:17:25.22 pancreas, spinal cord, etc ... And so we believe this paradigm
00:17:30.23 of in vivo cellular reprogramming could be used for many
00:17:35.10 different organs that might benefit from a regenerative approach.
00:17:39.18 And I think we'll see that in the coming years. So, what I've shown
00:17:43.11 you today is how we can utilize the knowledge of the key
00:17:48.22 developmental biology networks that are involved in cell fate
00:17:52.18 determination and organogenesis. I've shown it to you in the heart, but the same
00:17:57.19 is going on for many different organs in the body, and how we can
00:18:00.24 use that information to both understand the basis on a deep mechanistic
00:18:05.27 level of human birth defects, but also how we can apply that
00:18:10.11 knowledge to regenerative medicine in the adult.
00:18:13.13 So, with that I'd just like to thank the members of the
00:18:17.29 laboratory that did the work that I shared with you.
00:18:20.15 Those individuals are shown here and range from very
00:18:23.02 talented graduate students and postdocs to very dedicated
00:18:26.11 research assistants. And we've had a number of collaborators
00:18:31.15 throughout our scientific community that have greatly
00:18:34.24 contributed to the work. And thank you for your attention,
00:18:37.28 and I hope this has been informative.

Videos in this Talk

Part 1: A Change of Heart: Embryonic Heart Development
Audience:
- Researcher
Part 2: A Change of Heart: In vivo Cellular Reprogramming
Audience:
- Researcher

Speaker: Deepak Srivastava

Total Duration: 0:49:55

Recorded: February 2017

All Talks in Development and Stem Cells

Talk Overview

During embryogenesis, the heart needs to form a specific three-dimensional shape or a child will be born with a defective heart. Srivastava and his colleagues hope that by better understanding the molecular pathways involved in normal heart development, it will possible to improve treatments for both congenital and adult onset heart disease. In his first talk, Srivastava describes studies from his lab and others which use animal models and induced pluripotent stem cells to elucidate many of the gene networks that determine cardiac cell fate. iPS cells have been particularly important for identifying a mutation in the human transcription factor GATA4. By understanding the importance of GATA4 during heart development, it has been possible to develop a model that explains how cardiac specific genes can be activated while genes for other cell types are repressed.

About half of the cells in an adult heart are cardiac myocytes, or muscle cells, and about half are cardiac fibroblasts or support cells. Following a heart attack, muscle is lost and fibroblasts form scar tissue. In his second talk, Srivastava asks whether our understanding of embryonic heart development can be used to reprogram fibroblasts to myocytes to repair damaged adult hearts. His lab showed that introducing the genes for 3 transcription factors important for embryonic cardiac development resulted in an increase in the number of myocytes in a mouse heart after an induced heart attack. Similar results were obtained in vivo in pigs and in vitro in human cells suggesting that in vivo cellular reprogramming by gene therapy has broad implications for organ regeneration.

Speaker Bio

Deepak Srivastava

Deepak Srivastava is the President of Gladstone Institutes and a Senior Investigator and Director of the Gladstone Institute of Cardiovascular Disease and the Roddenberry Center for Stem Cell Biology and Medicine at Gladstone. He is also a Professor of Pediatrics and of Biochemistry and Biophysics at the University of California, San Francisco. Srivastava’s lab investigates… Continue Reading