| 1 | University of California Television (UCTV) | Cellular Reprogramming in Human Disease | 2740 | 113 | 4 | 67.8 | neutral | 58:26 | What I thought I would
do today is focus on a theme that I
think has evolved in our laboratory
where we can begin to both understand disease
and impact disease through the lens of
reprogramming cells either to do what
we want them to do or to recognize that in
certain disease states, the basis for disease is actually a cellular
reprogramming event, and once we understand that, that provides avenues for
potential intervention. I am trained as a
pediatric cardiologist, so my comments will be largely focused through the
lens of heart disease. But I do want to emphasize that the conceptual framework and the approaches could be applied, I think, to most human diseases. With regards to heart disease, of course, this remains the
number one killer worldwide, and we've gotten
better at keeping people actually alive
after acute events, but the end result
of that has been a growing population
that are left with damaged hearts and
what we call heart failure, clinically, and there are over 25 million people now worldwide who suffer
from heart failure. Ultimately, these
individuals would require a heart transplant, but even in the United States, there are only about 3,000 heart transplants done per year, and many countries don't
have any available at all. On the other end of the age spectrum is congenital
heart malformations, where children are born
with malformed hearts, and this is the most
common human birth defect. It occurs in 1% of all live
births across the world. It's a huge number. Here, too, we've
gotten better at palliating those defects
and keeping children alive, and as they get older, we're beginning to see a
number of sequelae that happen as they get older into adult life, including
heart failure, but also as I'll show you, we believe that some of the same genetic
abnormalities that result in malformations affect homeostasis
of the organ later on. My laboratory has over the years focused on trying to
deeply understand the gene networks
that are at play in cardiac sulfate
decisions early on and subsequent morphogenetic
events with the notion that we could utilize that knowledge
in two distinct ways. One, to regenerate
damaged hearts, and I'll show you a few
examples of approaches we've taken harnessing the cells
that are already in the organ, either the fibroblast population or the cardiomyocyte
population itself. Also that we could use this developmental
biology knowledge to understand the genetic
basis for heart disease, and not only understand that, but decipher
mechanism that would lead to new therapeutic
modalities, and I'll show you
an example of that. Now to start with the
regenerative area, we've published a body of
literature in this space, and I want to just in the
next few slides summarize for you some of what
we've learned from that, and I'll spend the
bulk of the time of this talk focusing
more on genetics. But in this realm, we've taken two
approaches to try to create new muscle in
the heart after damage. I should say that many
other groups are trying to transplant pluripotent stem cell derived cardiomyocytes
into the heart, and I'm very hopeful
that we'll, as a field, be able to overcome some of the current hurdles with that, and I'm eager to
see those develop. Our laboratory took a
different approach, and we decided to ask if we
could get the cells that are already in the organ to do something they
don't normally do. One thing that they
don't normally do, which is why the heart has little to no capacity
to regenerate, is they permanently exit the cell cycle very
soon after birth, and the result of that is
that if the muscle is lost, there's literally
no replenishment. Some years ago, we found a
way to quite efficiently get these adult
cardiomyocytes to re-enter the cell
cycle and divide, and that is one approach that we're pursuing to
regenerate the heart. The other is to harness the fibroblast population
that's in the heart and reprogram them directly into new
cardiomyocyte-like cells. Now, on this front, what we reported some years ago
was that a combination of four cell cycle regulators that are all highly
expressed during fetal development
when the heart is dividing and all get down
regulated soon after birth, if we reactivated those and stimulated both G1/S transition, as well as a G2/M transition, we could get 15-20% of adult myocytes in vivo to
re-enter the cell cycle, and that was enough to
increase cardiac output and improve function
in animal models. The important thing
here we found is that if you had
all four factors, you got stable cell division. What I mean by that is if
we remove any one of them, the cells actually could
divide in many cases, but then they'd undergo
mitotic catastrophe very quickly and
kill themselves. This is very encouraging
but for these experiments, we had delivered these
factors with the virus. That we presumed early on was a non-starter for
clinical translation because, of course, these could
also be oncogenic. We had tried years ago
to deliver this as mRNA, although one of the
problems with mRNA is that they only create protein for about three or
four days at the most. That was perfect for this
application where you want these cell cycle regulators
there for a few days. The cells divide and then
you want it to be gone. It turned out that
we could never get good delivery or expression, and it turned out that when the COVID
vaccine was developed, what many people don't realize
is the mRNA technology, of course, was quite old, but what made that possible
was new advances in lipid nanoparticle technology that allowed better delivery. With that in mind, we
visited this idea recently, and in collaboration
with Kevin Healy and Niren Murthy's laboratories at UC Berkeley in the
bioengineering department, we've screened for new LNPs
that might have tropism or propensity to deliver payloads specifically to the heart and specifically in areas of injury. What I'm showing you here
is an LNP we've found recently that has been injected systemically into
a reporter mouse where we're delivering mRNA of cre recombinase and if
the proteins expressed, you get a fluorescent
red marker activated. This is the heart here with the light sheet microscopy and I'm going to play a movie for you, and you can see at the
apex is where there's been damage after
coronary occlusion. You can see there's
accumulation of cre recombinase
protein depicted by the red here in this area
of damage specifically, and this is with the
whole body deliver. If you zoom in, you can see many of these cells
are actually quite large, and those are the
cardiomyocytes that we've documented in other ways, and you also see
these small cells which are the fibroblast cells. We believe we have a pretty
good delivery system now, and we're revisiting the ability to stimulate
proliferation and improve cardiac output with delivery of those cell cycle regulators
with this lipid nanoparticle. We're awaiting results for that, but we think at least we have a good delivery mechanism that's non-viral
now for the heart. Now, the other approach that
I mentioned is after injury, this heart that you
see here in purple is all the scar tissue that has replaced the viable myocytes. It turns out that the
cells, of course, that make this scar tissue
or the fibroblast pool, which turns out
to make up 50% of the heart in both
animals and in humans. These fibroblasts are also the ones that make scar tissue. Some years ago, we asked, could we utilize our
developmental biology knowledge when nature is making
heart cells and redeploy cues into these
adult fibroblasts that are already in the heart
and convert them to be more cardiac myocyte-like. In fact, we were able
to with the combination of essential developmental
transcription factors, particularly GATA4, MEF2C, and TBX5, and together
those three proteins, we found could bind to DNA in
a combinatorial code across the genome and wholesale
switch the epigenetics of an adult fibroblast to
be more cardiomyocyte-like. In mice, rats, and in pigs, we could do this with
enough efficiency, particularly in vivo, to improve cardiac output. Importantly, these cells electrically coupled
with their neighbors, which was essential for to
get a coordinated contraction in the heart so that
you can improve cardiac output and
not have arrhythmias, which has been a problem with transplanted cardiac
cells. That was great. We had the conceptual framework
for this reprogramming. But in fact, these three
factors did not fit into a single AAV that would be the best delivery
vehicle for this, and there was much work to be done to really translate this. Years ago, we put this
into a company we call Tenaya Therapeutics
in South San Francisco, and they've raised
a lot of money over the years to try to
move this forward and done a tremendous amount
of work to narrow the genetic material to
fit into a single AAV, develop a novel
capsid that could efficiently infect fibro blasts, not just not myocytes and a variety of other
delivery hurdles. They've done that, and I'm going to show
you just one side of data from Tenaya where they've done a coronary occlusion
in this pig heart, which is similar to human. You can see here at the bottom, this heart gets blanched. It's the area of damage. Then they waited a month to let the function decrease and
stabilize in the pigs. The ejection fraction of the
fraction of blood that is pumped out of the heart with every beat has gone down from, say, 60%, which is
normal to about 30%. Then they administer
the therapeutic after that one month post
myocardial infarction. What I'm showing you here is the data over a
two months period after delivering the dose compared to control pigs
who got a control vector. You can see here in the
control pigs in gray, the function stayed about flat, which is what we'd expect. They've already stabilized
their decrement. In contrast in the blue lines, you can see that
these pigs went from about 30% to low 40s,
which isn't normal. But it's enough to get out of the conversation of whether or not one should be considering
a heart transplant. Enough, people can walk
up a flight of stairs. They can walk several blocks, and so that essentially
is the goal here. Tonight has a number of other
hurdles still overcome, particularly with a
non-invasive delivery method that they're
continuing to pursue. But what I've tried
to share with you in this part is how in both cases, if we understand the gene
networks at play well enough, we can imagine ways to
re-program those cells to do something that they
don't normally do that could lead to
regenerative repair. Now, one of the things that I won't spend
much time today on, but I'll just briefly in one
slide in a cartoon form, summarize our recent findings that I'm also excited
about is the fact that even if you don't lose
any more myocytes and when somebody
has heart failure, what we do know is that
they reliably continue to decline in cardiac
function over time. The reason for that is this cardiac
fibroblast population gets inappropriately
activated and lays down more and more
fibrotic tissue throughout the heart globally and that
impairs cardiac function. In a body of work
in our lab led by two very talented postdoc who are both running their
own laboratories. Now, Michael Alexanian
and Arun Padmanabhan they started to investigate how is it that the heart senses this stress
that it's under, and how does it then signal to activate these fibroblasts
to become profibrotic? It's really quite
interesting what they found, which is that they
were able to map a very specific stress
dependent enhancer in the macrophage
population in the heart. These, we believe are the
sensing cells of stress. This enhancer activates
a number of cytokines, including maybe most
importantly interleukin-1 Beta. Interleukin-1 Beta is then
secreted and received by a specific receptor on
neighboring fibroblast, and that signaling pathway
then they mapped to another very specific
stress dependent enhancer that activates a transcription
factor they discovered Meox1 that turns out to
be the master regulator of a transcriptional switch
taking a fibroblast from a quiescent state into
an activated state. The reason I think
it's important they were able to trace
all these steps is each of these then serves as a potential
target to go after therapeutically either
an antibody against interleukin Beta
that they have shown is functional or deletion, or blocking of this enhancer, or this transcriptional switch. Each of these gives
you a level of specificity that we believe will reduce the potential
off target effects of, say, just blocking the
pathway completely. Especially if we can
intervene at the enhancers that are only active
during stress. That's all I'm going to say
about this part of the work, but we are very excited about
this potential to arrest, at least the cardiac function, so it doesn't continue
this decline. Now, for the remainder
of the talk, I'd like to focus on our efforts at understanding
genetics of heart disease. Here, also, our laboratory has published a body of
literature over time. I want to focus today
on two stories. One that represents a long
arc of investigation in my laboratory and the other that is more
recent unpublished work. The first is related to
this specific form of heart disease that
turns out to be the third most common form
of adult heart disease. It involves specifically the
aortic valve in the heart, which separates the left
ventricle from the body. What happens in a
large number of people is that this
aortic valve gets calcified over time
and doesn't open and close properly and ultimately, needs surgical replacement. It's an age dependent
phenomenon, so the incidence increases
as people get older. Interestingly, there is a link to a congenital
defect that occurs in a sub population of this
where some people are born. Instead of having three
nice valve leaflets that form this nice
Mercedes sign, as you see here, they're born with only two
aortic valve leaflets. Most of the time, they don't even
know they have it. But as they get older, about a third of these
individuals will develop early and more
aggressive calcification. It turns out that 1-2%
of the entire population worldwide is born with the
bicuspid aortic valve. But like I said,
most don't even know it until later decades in life. But sometimes, the valve is so malformed that even
in the newborn period. Children have difficulty
with blood getting out of the heart and they need an acute intervention
to stay alive. It's a whole spectrum of disease from fetal or childhood
to adult onset. The genetic cause
of this disease had not been known until
some years ago, when I took care of the little
boy here at the bottom of this family tree when I was
in UT Southwestern in Dallas. That boy was newborn and had severe erratic
valve stenosis from this thick erratic valve that we had to put a catheter across, blow up a balloon, rip it open, and then
he was able to survive. But what was interesting is
upon taking a family history, it turned out there
multiple generations of individuals indicated in
these black circles or squares that had already had open heart surgery because of calcified ertic
valve leaflets. This is clearly autosomal
dominant in transmission, meaning it's likely a
monogenic disorder, and we were able
to even back then map the single gene
that was causing this, and it turned out
that this disease in this family was caused
by a heterozygous, loss of function
mutation in NOTCH1. NOTCH1, I'm sure, is very
familiar to all of you. It's a famously studied gene. It's important in development
of almost every organ in our body and in some
maintenance in the adult. It was curious that reducing the dose of
the protein by 50%, which is what happens here, was caused disease just in the erratic valve
and nowhere else. Since then, our lab and many
other groups have identified a number of individuals and families with NOTCH mutations, ranging from erratic valve
calcification in the adult, all the way to even fetuses, that have such a severe
erratic valve obstruction in utero that their left ventricle doesn't even form as a fetus, and they are born with a very severe congenital heart lesion, missing a whole
chamber of the heart. That was great. This is our
first known genetic cause of this common disease. We were very excited,
and we thought, now we can begin
to understand it, but then we hit a roadblock. Mice that were heterozygous
for NOTCH, which, of course, had been made
much earlier were normal. Homozygous small mice died
at embryonic day 9.5 from vascular failure
because NOTCH is also on the end of the lining
of the whole vasculature. This became a nice to know, but we had very little
idea of mechanism. If you don't know mechanism, you can't really do much about it. I should say that this
smaller family here back at that time was
contributed by Paul Grossfeld, who's here at UC San Diego as a pediatric cardiologist and collaborated with us on
this work years ago. We had a breakthrough, though,
because Shinya Yamanaka, I had just recruited
after I moved to Gladstone to come to establish a laboratory
at Gladstone in 2007. He had described how to
make human iPS cells. We immediately flew the
family members from Dallas up to San Francisco
and did skin biopsies, made iPSL lines from four patients with the
mutation, four without. We turned those into
endothelial cells because we knew that those
were the culprit cells. I'd say for years,
we learned nothing. The reason any of you who try to model disease know
that unless you have isogenic controls and you've controlled the rest
of the genome, there's too much noise in the system to figure
these things out. It wasn't until we
were able to do gene editing
efficiently and make isogenic lines and then did full omic studies of
where NOTCH binds to DNA in the mutant setting, how it affects the epigenetics, and the transcriptomics
that finally the biology laid itself
out beautifully. This work was led by
Christina Theodoris, who was a very talented
MD PhD student in our laboratory at the time and went on to train in pediatric genetics at
Boston Children's, and we've now recruited to start her own laboratory at Gladstone. What Christina found back then, is that NOTCHs normal job in
the heart turns out to be to block the valve cells from turning into
osteoblast like cells. That's its job is to put
a brake on this system. You may ask, why would nature set it up that there
would have to be a gene to block this
sulfate transition, which is essentially a
cellular reprogramming event. It turns out that
cardiac valves are not that different in their
tissue compared to cartilage. Nature has used many of the same gene networks
that it uses to make cartilage to make cardiac valves during
embryogenesis. But it doesn't want
it to go all the way to become bone like, and so NOTCH is there to
put the brakes on that. It seems like the crux of this very common disease
is quite simple. It's a cellular reprogramming
defect where the cells are changing their fate and losing their identity and
becoming osteoblast like. Not only did Christina
figure that out, but since she started
off by mapping broadly the gene networks
that were shifting, what she found is that these gene networks
were largely being driven by upregulation of just three key transcription
factors, SOX7, TCF4, which mediates Wnt
signaling and SMAD1, which mediates BMP signaling. If there are about
1,000 genes that were dysregulated in
the sulfate transition, and about 80% of those
are 800 were directly, indirectly or indirectly
related to these three factors, and that if she
knocked those down, these three factors down, you could largely shift
the network back. That was great. Now we understood
mechanism a bit and it looked like it funneled
down to a discrete pathway. Christina decided to undertake a very interesting drug screen, not looking for molecules
that upregulated NOTCH, but rather molecules that
would shift the whole network. The whole gene network
closer to normal, and she set up a machine
learning algorithm back then to call normal cells or
mutant cells and ask if thousands of molecules, if any of one of them would
shift the whole network back. I won't go through the details
of that as it's published, but suffice it to
say she did find a very interesting
small molecule that seemed to correct the whole gene network
quite effectively. This molecule turned out to be an estrogen receptor
related Alpha inhibitor. We validated that as a target subsequently with
SINA experiments and testing a host of
other small molecules that also hit this target, and we're convinced
now that this is in fact the true target. That was great.
We've now finally, after these years of effort, had a small molecule that
might actually be able to shift this
aberrant gene network back closer to normal. But you'll recall that we didn't have a mouse model
to test this in, and we weren't going to go
from a human iPS model, to a clinical trial, of course. Christina had the clever idea that putting together
a few observations, and one is that, as I mentioned, this is an age
dependent disease. We know that telomeres get
shorter as people age, and we also knew that mice have much longer telomeres
than humans. She asked whether mice
might be protected from the disease simply because
of their longer telomeres. That's a testable hypothesis. She tested that by
crossing NOTCH1 one heterozygous mice that were normal with mice lacking TERC, which is the RNA
component of telomerase. When you do this in each
generation of breeding, the telomeres get
shorter and shorter. It turns out that just in the second generation where the telomeres are just
a little bit shorter, that resulted in nearly
complete recapitulation of the human phenotype. It's really quite striking. I'm showing you that here in a cross-section through the
aortic valve of these mice, just after a month of age. It's really quite rapid. On the top panels, what you see are the three
nice aortic valve leaflets and a normal mouse
that's turk null, but not wild type. In contrast, you can
see on the bottom that these valves
are very thick, and they're heavily calcified, as indicated by
Alizarin Red here. This is what we see in humans and by ultrasound in these mice, we can detect the
level of stenosis, which is exactly the way
we detected in humans, where if the valve
aperture gets narrow, blood has to go across
it faster in order to get the same
amount of blood out to support the circulation. You can measure the speed
of blood across there, and that tells you how
much stenosis there is. But what this mouse model also allowed us to do
is to go back and ask if that observation in vitro was really
happening in Vivo, namely, are these
cells undergoing a sulfate switch to become
more osteoblast like in vivo. For that, Christina
stained these with valves with a marker of
osteoblast called Runx2, which is a master
transcriptional regulator that drives the fate even
and you can, I think, easily appreciate here
that these cells are all positive for Runx2
suggesting that, in fact, this is likely a disease of cellular
reprogramming. This mouse model then
gave us a way to test this drug in vivo in the setting of erratic
valve stenosis. We recognize that most
in the human population, these individuals would come to the attention of a physician you can hear the level of stenosis
with the stethoscope. Then normally what we
do right now is that we detect that and see it by
ultrasound, document it, and then we just watch
people over years, and as it gets worse and worse, then we do surgical
intervention. The reason for that is
that right now there's no medical therapy whatsoever
for this condition. What we did in mice
is we took the mice, I should say this was
incompletely penetrant, so not all the mice had a degree of stenosis that we
could detect by ultrasound. We did ultrasounds at a month of age and
isolated those with documented aortic valve stenosis and then randomized those in a preclinical trial to
either drug or placebo. Then we treated
them for a month, and then came back and repeated the ultrasounds after
one month and asked, did this drug slow
down progression? Because in humans, I think
all we would have to do is slow down progression,
not even arrest it. Because it's a disease of aging, most people will
ultimately succumb to other conditions rather than need an intervention for this. What I'm showing you
here the results from the placebo group,
the control group, and on the y-axis is the amount percent change over that one-month period of the degree of aortic
valve stenosis. You can see it's a
rather wide range, but most of them are
progressing as we'd expect, just in a month with a mean of somewhere around 80%
further stenosis. Now, in stark contrast, when we looked at the cohort of mice that had been treated, it looked something like this. While there are a couple
of mice that did progress, the vast majority had
little to no progression. But most strikingly, half of the mice actually
had regression. It looks like with this drug, we're actually also able
to not just stop but maybe even reverse in part some of the
degree of stenosis. This is really exciting. We
think we might actually have a drug that could be effective
for this common disease. But everything I've shown
you so far is related to animals with notch mutations. It turns out that I
think I didn't mention this probably only 4-5% of the population with this
disease will present with mutations and not
your pathway members. Those are going to be
more rare conditions. Most people will not have an identifiable
genetic condition, even though it's
likely a combination of factors leading
to this disease. But we do believe
that for most people, even if they don't
have notch mutations, the cause of the disease
is going to be similar, which is a shift
of cell identity into an osteoblast-like fate and then laying down of calcium. We decided to test this
idea by collecting primary aortic valve
endothelial cells from patients who had had their valve explanted
at time of surgery, and normally the surgeons
would throw in the trash. We could get those and grow these actual aortic human cells. We did that for a bunch
of normal valves, a number of calcified
three-leaflet valves, and calcified
two-leaflet valves. The first thing we observed
by sequencing their RNA is that the same gene
network was shifted, and in particular, the same
three transcriptional drivers were all upregulated. That suggested that, yes, the same network
is being altered even in the absence
of notch mutations. Most importantly, the
same ERR-Alpha inhibitor restored the gene network
broadly and in particular, these three
transcription factors. With this information, what
we do believe we have now is a viable drug that can be taken forward
in clinical trials. I should say we've advanced the chemistry of this molecule, so it's got high potency
and low toxicity, features, and good pKa values. Our vision now is that we finally may have an oral
once-a-day drug that could be used to at
least slow down and arrest the progression of
this very common disease. That story, I should say, as a physician scientists, particularly, probably
represents one of the most satisfying arcs
of discovery of my career, going from taking
care of a patient to identifying the genetic cause,
understanding mechanism, and actually having a
potential therapeutic that could alter the way we
treat this disease. The last story I want
to share with you is, again, related to genetics, but it is an unpublished
story related to understanding why
congenital lesions occur. I should first say
that as a result of a broad nationwide consortium
funded by the NHLBI, where we've sequenced over 5,000 children and their parents
with congenital heart disease. We can now understand with a genetic basis about half
of congenital heart defects, which is a big
advancement for us. That still leaves
about half that are in this pie chart that
are unknown causes, and we're still working on. But even amongst the half
that wet "understand', I want to draw your attention
to this light blue pie. That is the pie where we understand the cause
because of aneuploidies. Now, that gives us some satisfaction that we
understand the genetic basis, but aneuploidy is often involve large portions
of the chromosome. Actually, we generally
don't understand the specific genes that
might be driving this. Again, if you don't
understand mechanism, there's not a lot we can do. Of course, the most
common of all of these aneuploidies
is trisomy-21, which results in down syndrome. It turns out that children
with Down syndrome, about half of them are
born with cardiac defects, mostly of the septal
defects where the walls between the atria and ventricles have holes in them, and there's communication of
blood between the chambers. Now, there's one very
specific type of septal defect that I've
illustrated here in this cartoon, which is at the level of where the separation is between the atrial chambers at the top and the ventricular
chambers here, involves the valves
that separate these. Here, there's free communication between all four
chambers in what we call an atrioventricular
septal defect because of a defect at this
specific level of the heart. This happens in children
without Down syndrome also, but in Down syndrome, this occurs with a thousand-fold
increase incidence, thousand-fold. There's clearly something
very specific going on here in trisomy-21 that is resulting in this specific area of the
heart being affected. We haven't known the genetic
cause of this in the past. This is the problem
that Sanjeev Ranade, a postdoc fellow
in the laboratory, undertook several years ago, and Sanjeev is now here at the Sanford Burnham Institute
running his own laboratory. I want to share with you the
beautiful work that he's done while he was in the lab and is continuing
here in his own laboratory. The question obviously is,
what are the gene or genes on chromosome 21 that are
driving this type of defect? Before I share with
you the story, there's one piece of developmental biology
background you need to know, which is that if you look at
a developing mouse embryo, like you see here from the side, here's the left ventricle,
the left atrium, and this is the
region that's between the atrium ventricular
chambers that we call the atrioventricular canal. If you do a cross-section
through this, you can easily see that the
muscle has a cuff around this specific area is different from the
neighboring muscle in the ventricle and atria. You can see that by
gene expression of this transcription factor, Tbx2, where it's only expressed in this cuff of muscle
and not here, in contrast to other
genes that are only expressed in the
chamber myocardium, but not this
atrioventricular myocardium. This is a very
specialized muscle right around where the valves will form and the
septa will form, separating the chambers, and is exactly where the
defect occurs in this disease. The approach Sanjeev
took was to use iPS cells that were made
from people walking around who are mosaic
for trisomy 21. These people don't
have any disease. They have a higher
risk of having an offspring with trisomy
21, but they're fine. But when you make iPS
cells from these mosaics, in the same dish, you'll get disomic iPS lines and trisomic. Again, the key is this gives
you isogenic controls. He did that and
differentiated them to cardiomyocytes and did
single-cell RNA sequencing. Here we can annotate cells shown here that are
actually of the type, specifically of the
atrioventricular myocardium, and others that are
ventricular myocardium. We can distinguish
by single cell RNA seek this specialized
population. Here's what I thought was a really key observation that made the rest of
the project possible. That is when you look at just the transcriptome
in these specific cells, there was a very
clear difference between disomic and
trisomic cells. That's illustrated
here in the dot plot of a number of genes, where you can clearly
see the pattern is different between
disomic and trisomic. Importantly, what
we observed is that the genes that
were normally high in the atrioventricular
canal myocardium, the specialized myocardium,
were all downregulated. In contrast, the genes
that were normally low, ventricular myocardial
genes, were now upregulated. What it seemed like
was going on here is maybe this disease
is one of, again, reprogramming in part of the specialized
myocardium to now be more like ventricular
myocardium, and that could be the
crux of the disease. Now this gives us an assay to
go after where some gene or genes on chromosome 21 must be causing
this pattern shift. Now the problem
becomes quite simple, although it was a lot
of work to find it, but conceptually it's simple. We were aiding in the fact that chromosome 21 is one of
the smaller chromosomes. There are 241 genes, and then there had been
a mouse model already made of trisomy 21 with
a syntenic region, and that model recapitulates the cardiac defects and only had duplication of less
than 150 genes. It must be one or more of those genes because we are
seeing the cardiac defect. I should say, from the
mouse data genetics, it was clear that they're likely at least two loci that both need to be duplicated to be
sufficient for the defect, but each being necessary. Then our laboratory had
over the years made a mouse atlas of all the genes expressed during the heart
during development, and of these 148 genes, only 66 were ever expressed in any cell
type in the heart, and so we figured
it had to be those since the mouse also
has the defect. That got it down to a
more narrow number. What Sanjeev decided to
do is take advantage of the CRISPR-Activation
approach. Now, any of you have done CRISPR-A know that
the problem with CRISPR-A is that you don't ever get really high
levels of activation. You get maybe 50%, maybe 100%, but you
don't get t10-fold. But for us, that was
perfect because in trisomy we were only
getting 50% activation. Sanjeev inserted the
Cas9 VPR cassette into the disomic line so he could use guide RNAs to each of
the 66 genes and ask, do any of the guide RNAs then recapitulate that pattern
shift that I showed you? He introduced, using
a CROP-seq approach, a lentiviral library
of the guide RNAs, differentiated them
to cardio myocytes, and then did single
cell RNA seq. Just in that atrioventricular canal myocardial
population, asked, do any of the guide RNAs make that pattern shift occur
to be more trisomy-like? Working with Sean Whalen
and Katie Pollard's lab, they set up a machine
learning algorithm that would call hits. Here, what I'm
showing you in red in this UMAP plot are
the disomic cells, blue are the trisomic, so they clearly
separate in space. The gray dots are each
of the guide RNAs. You can see most guide
RNAs don't do anything, which is what we'd expect. But some cause a shift closer
to the trisomic state, and we then took those and did secondary tasks
with them individually, not in a pool screen. To make a long story short, one of the guide RNAs virtually recapitulated
the pattern shift, and that turned out
to be a guide RNA activating a gene called HMGN1, which many of you will know is a nucleosome binding protein that functions often
as a co-activator, sometimes as a corepressor, at target sites, particularly of cell type specific
transcription factors. Hat I'm showing you here
again is that dot plot with the disomic control here, the trisomic all the
way on your right, and then in the
middle is activation just of HMGN1 with
the guide RNA, and you can see it largely
recapitulates the pattern. Particularly, these genes like TBX2 that I showed you
in R-spondin 3 that are normally high in the atrioventricular
canal myocardium are similarly down
regulated by HMGN1, like in the trisomic condition, and the chamber myocardial genes which are up-regulated in trisomy are also up-regulated
just with HMGN1. This suggests that
this one gene on chromosome 21 at least is
shifting the transcriptome, similar to what we
see in trisomy, with this identity shift, if you will, for the specialized myocardium
to be less specialized. That was encouraging.
But of course, what we want to know really is, in vivo, is this critical? We do have a way to test that, of course, because I told you, we have a mouse model,
and we have mice that are heterozygous for HMGN1. The simple experiment
is to cross the trisomic mouse with mice that are
heterozygous for HMGN1, and now you reduce the dose
of HMGN1 to two alleles, where everything else
has stayed the same. If this is the key gene
or one of the key genes, then we should rescue
the phenotype. We did that, and this
is the result we found, particularly for the more
serious cardiac defects, septal and
atrioventricular defects. You can see here
the trisomic mice, we get a very significant increase compared to wild type, and just by removing
one allele of HMGN1, we rescue this phenotype of
the cardiac defects in vivo. We believe this is
convincing evidence that HMGN1 is likely at least
one of the major genes. There has been very good
literature suggesting that a kinase DYRK1A is
another critical gene, and we're now trying to
test if both of these together could be sufficient. But as I mentioned, obviously, the key
question now is, where does HMGN bind
in these cells, and how does it affect
gene transcription? We're eagerly pursuing
those experiments in collaboration with
Vijay Ramani's laboratory, who has expertise in this area. But what we think is going on, even as we learn mechanism, is that the crux of this disease is likely a reprogramming
event, again, where this specialized
myocardium gets reprogrammed to be more like the ventricular myocardium
that it shouldn't be, and then that affects its
signals that it's sending to the endocardium to then affect the development
of the valves and the septa in this region. What I've tried to show
you today is how we can go from understanding the
developmental biology and molecular biology of
the system to being able to leverage that both for regenerative
medicine purposes and mechanistic understanding
of the genetics of human disease in ways that
will allow us to intervene. I tried to mention the brilliant trainees that have come through
the lab and worked on these projects
over the years. Many of them are listed here, including Sanjeev, and
great collaborators. This is my current laboratory, a fairly recent picture. With that, I'll close
and be happy to take any questions.
Thank you very much. Incredible work. It's really mind blowing. Anyway, I had two questions. One was about the last story. You showed RBFOX1 changes, and that is a regulator
of embryonic splicing. Have you looked
into alterations in splicing as an epitranscryptomic
instability that could actually lead to what's going on in the specification
of what should be atrioventricular tissue as opposed to myocardial
ventricular wall? Yeah, it's a great question, and we have worked on
splicing in other projects, and we're eager to
understand that it's hard to do with the
10x platform with single cell because
you don't get full transcriptomes to
understand splicing. There are some
computational approaches that try to do that, and we've been trying to collaborate with Lars
Steinmetz at Stanford, who's developed some of those, but we haven't succeeded. You can do the five prime
version, just saying. We can. Then the other question
is about anakinra. You're showing
IL-1 up-regulation and IL-1 Beta specifically. You can give anakinra, which is the IL-1
receptor that's exogenous that could
mop up that IL-1 Beta. Have you thought about that
in terms of preventing that myocardial reprogramming
towards a fibrotic state? We have. In fact, we have used
that and it was effective. There had been a
clinical trial of the IL-1 antibody
that had been done. It didn't get approval because of a number of side effects of blocking the immune
system more broadly. It led to greater
risk of infection. So that's why I think getting
with a greater degree of specificity that wouldn't
affect more broadly is key. But conceptually, that
I think showed us that this is the right
pathway to go after. Absolutely fantastic. I have three questions
about the Notch-1 work, but I promise they're short. They're all mostly
yes/no questions. For the Notch-1 reprogramming to a more osteoblastic
like state, that makes total sense. I'm curious whether there's
any explanation for perhaps the valve formation of a bicuspid format as
opposed to tricuspid. I'm not sure what you
thought about that. In the valve ex-plants
that you got from people who got the removal
or the swap surgery, was there any genomic
evidence for, say, somatic variance in
the Notch-1 pathway? Then, finally, I was
wondering when you tested the ERR-Alpha
blocking drug, whether or not there
were any adverse effects on the female mice. I'll try to be brief
in my answers as well. Sorry. For the first step, we
haven't been able to gain a lot of insight on the developmental
part of the anomaly, which is disappointing
to me because that's actually what I'm probably
most interested in. But our mice don't get bicuspid aortic valve
with this condition, so that's been a
rate limiting step. We don't fully
understand that part. For the second question
related to the? Somatic variance. Somatic variance. We
haven't looked at that, so I don't know if
that's the case, but it's a possibility
that there could be specific variation there. The third, we don't see sex
differences in those mice. I should say, which
I forgot to mention, that the vascular calcification
that occurs commonly in humans pathologically looks very similar to what we
see in the valve. When we do the clinical trial, we'll be looking at vascular
calcification also. We just don't know if it would work for that also
because there's not a good mouse model at all
for vascular calcification, but we have every
reason to think that it'll affect there also. Yes. Amazing talk. Really inspiring. I love the idea of
transforming fate in vivo, and that's happening
in the heart. We see it a lot in
epithelial tissue, so seeing it in the heart,
it's absolutely amazing. My question relates more on
lots of those changes that you've shown in those heat maps are actually niche factors, TGF Beta, WNT, VGFs. Maybe what are your
thoughts around this fate switch and how the
cells are actually undergoing this fate switch change
in their environment? You're absolutely right. What I didn't say is myocardial cells, we know a lot of the genes that are involved in valve formation, and they're dependent on
secreted factors like those that get sent and received by the
neighboring endocardium. Then the endocardium forms valves in these
acceptal regions. We do think that those are critical downstream
targets that are directly affected by the abnormality of that specialized myocardium. There are specialized
TGF Betas that are only secreted from
that valve myocardium, and there's special
TGF Beta receptors. They're only expressed in the endocardium
next to the valve, not in the endocardium
in the chambers. There's decades of
research on that, so we know the
pathways quite well. Exactly, you were spot on. Yes. Trisomy 21 has these
high incidence of these endocardio
cushion defect things. But I'm wondering,
since you seem like you maybe have one of
the main factors there, half the patients
with trisomy 21 don't have congenital
heart defects, but this would suggest they
have some predisposition. I'm wondering if
any work has been done in those that don't, just look at maybe gene
environment interactions or modifier genes that
might predispose to that more severe phenotype. It's a great point, and why is it that half
don't have disease? If you could figure that out, maybe there's a
prevention modality. There's actually been a lot
of effort, not by our lab, but by others in the field
to sequence large cohorts of people with trisomy 21
that have heart disease and those that don't to
look for modifier genes. It's been frustrating
for the field. So far, it hasn't been
obvious, there must be, but we haven't had
the power, I think, to identify that as yet.
It's a great point. Well, while we wait for
more questions in the room, we'll take one off of the Zoom. Given that you found
an association between telomere shortening
and aortic stenosis, does that give you any insight into aging related
diseases of the heart? Yeah, or period, aging related. When we had that finding, I immediately, of course, went to Liz Blackburn's office, which is on campus before
she'd come to San Diego briefly and went to share the data with her
and ask her for insights. I was surprised to learn that the field didn't have a good handle on why telomere shortening was
associated with aging, even though it's known to be. What I am actually very
excited about that we're pursuing right now is the idea that we actually
might have one of the best models that exist to figure out mechanism for telomere shortening
relationship to aging. One theory in the field is
that the telomeres loop back to contact promoters throughout the genome and
regulate transcription. The way to test that
experimentally, would be able to do Hi-C or three dimensional
conformation experiments in the setting of these
various mutations. The problem is, it's a
very small subset of cells in the valve
that are affected. You can't really figure
that out unless you can do single Hi-C and only recently has the technology
developed to do that. Longzhi Tan at Stanford is one who has developed some
of that technology. We're working with
his lab right now and have that single
cell Hi-C working. It looks like that's
exactly what we're seeing, and I think we might
be able to crack that, but we don't have the
answer right now. Dyskeratosis congenita, they
have mutations in TERC. Could you make an
IPS line out of that and do your
Hi-C experiment? Yes, that could be useful. What we have done just
in the last few weeks is a Hi-C experiment with the
mice that are null for TERC, and then mice that are null for TERC and
heterozygous for notch. I'll just share with you
the preliminary results, which I'm still
trying to understand, it turns out that in the
notch heterozygous setting, there's a decrease
in contacts and transcription of many of the genes that are
osteoblast-like. Then in the TERC knockout, that's true of TERC
knockout by itself. Then together, it's exaggerated. They seem to be hitting the same genes
even by themselves, and it's synergistic. Now we're still
trying to figure out. Is there a window of
opportunity after an infarct where you cannot do reprogramming in the heart? In other words, the
conversion from noncardiomyocytes
into a cardiomyocyte where nothing you
do will change. Yes, I think for a
variety of reasons, it wouldn't be done in the
acute setting immediately. Part of that is, if we were able to do it very efficiently, you want your fibroblast there to make scar and not
have a cardiac rupture. I actually don't
think we're at danger for that because we're
not so good at it. The efficiency is not
where we want it to be. The bigger reason
that won't be done is that for a clinical
trial purpose, there's so much variability
patient to patient in the acute setting of
where they'll end up. That clinical trials
become nearly impossible to see a
signal because there's so much variability from the time an injury occurs
to where that patient will land at their baseline after revascularization and
other medical therapies. You're just going to
get a ton of noise. I don't think you'd
ever see a signal. I think in clinical trials, it'll have to be once a
patient has stabilized, like we did in the pig. They've stabilized in whatever their decrement will be in
their cardiac function, so you have a reliable
change you can track. You define when it's too early, is there a point
where it's too late? We don't know that answer. It's hard to do really
chronic studies in animals to wait a long time and do it,
particularly in pigs. Honestly, we don't
know the answer. We'll have you be
the last question. I have one last question about the stability
of the reprogramming. You're showing one of
your first studies that you can reprogram fibroblasts to cardiomyocytes
in a scar tissue. How stable is this change? Whether they stay forever cardiomyocytes or
they can change fate. As Prince would say, forever is a mighty long time. But, what we do know, having studied the
epigenetics of this switch is that it's a wholesale epigenetic shift that appears to be very stable. In vivo and mice, when we've gone
several months out, it appears to be stable. We have every reason to
believe that after that, you've overcome that
epigenetic barrier and landed in a
different valley, let's say, using the Waddington
landscape as an example, that it would be an equally
difficult hurdle to get back. I think that's how
we think about it, but time will tell. We want to thank Deepac for just a really inspiring example of what we've often
talked about here, bedside to bench and
then back to the bedside and for just truly
illuminating. Thanks so much. Thank you. | ↗ |