Leveraging Congenital Heart Disease Mouse Model Findings to Improve Clinical Outcome  – Cecilia Lo

Leveraging Congenital Heart Disease Mouse Model Findings to Improve Clinical Outcome  – Cecilia Lo

Cecilia Lo:
I want to thank Teri and the organizer for the invitation to participate this very exciting
meeting. My interest has been in congenital heart disease for some time, and Cricket gave
a beautiful presentation earlier and she really introduced a lot of the background so I’m
going to be very brief. Congenital heart disease, as you know, is one of the most common birth
defects. It’s characterized by abnormalities in the cardiovascular structures. In particular,
human and air breathing animals in general have four chamber hearts with separate pulmonary
system and circulation. And so you have a right atrium, right ventricle pumping the
oxygenated blood through the pulmonary artery to the lung where it gets oxygenated. It returns
to the left side of the heart into the right — into the left atrium, left ventricle, out
the aorta, and it’s systemically bringing oxygenated blood to the rest of the body.
So there’s this very distinct right-left asymmetry and four chamber anatomy that’s
really critical for survival of air breathing animals, and it’s disruption of this four
chamber anatomy that actually underlies congenital heart disease. As Cricket mentioned, advances in surgical
palliation now allows most congenital heart disease patients to survive their structural
heart defects, and in fact many of these patients are living into adulthood with more adults
with congenital heart disease now then there are children or infants born with congenital
heart disease. What has been found over time, and again Cricket mentioned that, that patients
with some of the same structural heart defects can have very different outcome. Of particular
note for example are patients with hypoplastic left heart syndrome. They have a very stenotic
aorta with hypoplastic left ventricle and atretic or stenotic mitral valve so the whole
left side of the heart is very hypoplastic. Patients with HLHS can have variable outcomes.
Some after surgical palliation could live a relatively normal life and others undergo
heart failure. Studies by the PHN Network has shown that
in fact it’s — what’s most important for the long term outcome of these patients
are patient intrinsic factors and not surgical parameters, and so suggests that in fact something
intrinsic to the patient is actually determining outcome. And so our interest is in the possibility
of exploring whether genetic factors might be a contributing factor to possibly the differential
outcome in patients with congenital heart disease. And of course the first step is the
deciphering what are the genes that may contribute to congenital heart disease, and so that’s
what I’m going to focus on. So we have undertaken a large scale mutagenesis screen to look at
genes that can cause congenital heart disease using the mouse as our model system. And the
reason for that is that mice have the same four chamber cardiac anatomy as humans so
that’s obviously important because it’s the substrates of congenital heart disease.
And also with inbred mice, we can avoid the problem of genetic heterogeneity that you
see in human studies so it’s really an ideal model system for studying the genetics of
congenital heart disease. So we undertook a systems genetic approach,
as I said, with a large scale four genetic screen with chemical mutagenesis with ethylnitrosourea.
This introduces random mutations in the genome and allows us a way to interrogate in a non-gene
bias way for genes that can contribute to congenital heart disease. So we’re doing
a phenotype driven screen and the goal is to identify genes and pathways that may drive
congenital heart disease pathogenesis. By driving this screen with phenotype, we’re
hoping the collection and the totality of genes may give us some insight. What are the
main pathways that are contributing to the pathogenesis congenital heart disease? And
of course we’re also hoping that the totality of data we collect may also give us some insights
into the genomic context for disease pathogenesis. So this is just some ultrasound images to
show you the way we do the phenotyping is by in utero fetal ultrasound imaging. This
is a whole conceptus that you can see and that’s the frontal view. You can rotate
the transducer and you can see in the sagittal view the same animal, vertebral column, the
heart beating away, and here’s the head of the animal. And so this very relatively
low resolution ultrasound will allow you to locate where the fetuses are, and with the
in utero fetal screen we don’t have to worry about fetal demise because we can actually
track them horizontally over time. And so using more high resolution imaging with Vevo
2100, we can actually look and get a lot of details and structures of the heart. So using
four different imaging modality, we can actually collect very specific information about heart
structure and function, pretty much as what you would do for human fetuses or even human
adult cardiac phenotyping. And so with fetal ultrasound, we are able to not only be very
high throughput because it’s noninvasive, but also have very high detection sensitivity
and specificity because this imaging modality is specifically developed to phenotype for
cardiac anomalies. So using this noninvasive phenotyping strategy
over a five year period of screening, we’ve scanned 100,000 fetuses. This is a recessive
screen so it’s a G2 by G1 backcross so they come from over 3,000 pedigree, and out of
the 100,000 fetuses, about 3,000 show evidence of cardiovascular defects. Obviously only
a subset of these will have structural heart disease, but by doing the fetal ultrasound
imaging, we can go from the 100,000 to the 3,000 relatively efficiently. And we’ve
been able to recover more than 300 mutant mouse lines with a wide range of congenital
heart defect phenotypes. They’re just grossly summarized here. I’m not going to go through
that with you, but suffice it to say most of the congenital heart disease observed clinically
have been recovered in our screen. One of the surprising things was that we found that
25 percent of the congenital heart disease mutant lines actually exhibit left right patterning
defects. If you remember to the early part of my introduction, I showed you the left
— the cardiovascular system is very left right asymmetric and that asymmetry is actually
really important for establishing systemic pulmonary circulation. So maybe you shouldn’t
be surprised that mutations that affect laterality defects actually give you some of the worst
and most complex congenital heart disease. One of the — also one of the other interesting
observations from our screen is that when you have a mutation that cause left right
patterning defects and congenital heart disease that most of them will present with three
phenotypes. So they can have situs solitus meaning this — the visceral organ side is
completely normal. Situs inversus where it’s reversed, a mirror symmetric reverse, so for
example the heart’s pointing to the right, but the stomach is on the left side — sorry,
on the right side as well and — which is the mirror symmetric of normal. Or you can
have heterotaxy where it’s randomized. So this animal has the heart pointing to the
right, but actually, the stomach’s on the left. And so when you have heterotaxy that’s
actually when you have congenital heart disease, but animals that have situs solitus or situs
inversus are completely normal. I wanted to bring this up because this suggests that — well
we know that a single mutation can give you three distinct phenotypes. Only the heterotaxy
mutants will have congenital heart disease so when we’re doing exome sequencing or
whole genome sequencing, we’re using sibling controls, I just want to point out that you
may actually be filtering out mutations that could cause congenital heart disease. So for
example in the ExAC database, I think that that’s something we should really keep in
mind as a potential complication. So all of the mutant lines were recovered.
We crowd preserve the sperm so that they’re available from Jackson Laboratory, but we
also actually provide very detailed curations of their phenotype. There’s also deposit
of the mouse genome informatics database. So this is an example of some of the images
that you might see from one of these pages of a mutant. We actually have added several
hundred mouse phenotype ontology terms for congenital heart disease and we also used
the Boston file codes so that we can correlate the mouse congenital heart disease phenotype
with the human clinical terms so that they are cross referenced. And we also provide
— I forgot to — yes. We also provide a lot of images and [unintelligible] clips so if
you go to MGI, you can get a lot of detailed information about the phenotype of these mutant
lines. So of course it’s all about mutation recovery,
not just getting interesting phenotypes and we’re able to do this by whole exome sequencing
because we do our screen in the context of completely inbred C57 black six mice. So we
can simply do exome sequencing then compare any changes relative to the reference genome.
And so at the end of our screen, we’ve generated over 12,000 mutations in over 7,000 genes.
If you assume 24,000 genes, it would suggest that we’re at 30 percent saturation and
of the 1,500 homozygote mutations, we see that 147 in fact are pathogenic or we were
able to show that they’re pathogenic. Two thirds of them are missense mutation and the
remainder are split between nonsense mutation and splicing defect mutations, and these reside
in 98 genes. All of the mutations recovered, the more than 12,000 mutations, are searchable
in a mouse model organism search page in the Bench to Bassinet website. And remembering
that we’ve crowd preserved all the sperm, all of these alleles are actually available
because you can regenerate them from reanimating the sperm. Of the 98 genes we recovered, 47 are novel
genes, novel meaning not previously known to cause congenital heart disease, and we
also have 23 genes with multiple alleles. And using that screen metrics and a statistical
modeling approach known as unseen species method, we can get an estimate of the total
number of congenital heart disease genes in the mouse genome. So this includes looking
at the total numbers of congenital heart disease genes recovered, the number of genes with
one mutation, and the total number of mutations recovered altogether. And so this estimate
is 272 genes, and remembering this in the context of a recessive screen. So this would
suggest that our screen is at about 35 percent saturation, remembering that we’ve recovered
98 genes. So this number, 35 percent, is remarkably close to the 30 percent we came up with by
just looking at the exome sequencing metrics. So taking a close look at the homozygote known
mutations in the mutant lines where we recovered the pathogenic mutation, we see that there
are 151 of these homozygote known mutations. We would expect that 30 percent of these would
be embryonic lethal based on the comp studies. Of these 151, 108 have been curated with known
Knockout mouse phenotype. And we see that of the 108, 104 or 96 percent are viable to
weaning. That’s not surprising because again we’re screening the mice in the gestation
so this is selecting for genes that are not going to cause early embryonic lethality.
Surprisingly we also recovered mutations for known mutations in four genes that are not
known to cause early embryonic lethality, but we actually get them to survive to near
term. So this suggests the notion of genetic resiliency which is recently presented in
an interesting paper, Nature of Biotechnology. So what did we recover in the totality of
congenital heart disease genes? We find that there is an enrichment for cilia-related mutations.
Fifty of the 98 are cilia related with another 21 involved in cilia transduced cell signaling.
We also see a number of endocytic vesicular trafficking genes. Overall this would suggest
that the disturbance of cilia and cilia related function may play an important role in the
pathogenesis of congenital heart disease. This is just a diagram to show you some of
the cilia genes that were recovered. I’m not going to go through this, but I just want
to point at that we obviously recovered some of the cilia genes that are involved in human
diseases including genes involved in multi cilia functions shown here. They can cause
primary cilia dyskinesia, but also genes involved in ciliopathies that are thought largely to
involve primary cilia including genes that cause Joubert syndrome, Meckel-Gruber syndrome,
nephronophthisis. And many of these genes you can see actually are found in the cilia
transition zone which is a very important region that gates trafficking of proteins
in and out of the cilia. We also found many cilia transduce cell signaling
genes, 21 in total including those involved in Wnt signaling, in hedgehog signaling, PGF
signaling, and PGF beta signaling. And so the question is, do these pathways and mutations
really have any relevance to human congenital heart disease? So we borrowed some data from
the Pediatric Cardiac Genomics Consortium that Cricket mentioned earlier. This is data
actually from the earlier study that was published in 2013. Out of 27 de novo variants that were
recovered, we found that 11 of them were in pathways identified by a mouse congenital
heart disease screen including genes involved in TGF beta signaling, Wnt signaling, hedgehog
signaling, ciliome related, and endocytic trafficking. Again to reverberate something that Cricket
mentioned, we found that actually many of these genes are involved in — also have a
role to play in exome guidance and neurogenesis and synaptic transmission. Here is just a
subset of the genes recovered and actually now with a 98 genes; if we put it through
pathway analysis, we see that one of the top pathways recovered indeed is involved in exome
guidance and neurogenesis, so again suggesting the sharing of developmental pathways in heart
development and development of the brain and the nervous system providing an explanation
for the poor neural developmental outcome often seen with congenital heart disease.
Another interesting finding that we made was that many of the pathogenic mutations recovered
actually code for interacting proteins, that is proteins that are direct interacting partners.
So here are just some examples. So we have a mutation in one animal that’s Ank6 that
causes a disease, a different animal in Nek8. And we know both of these proteins are direct
interacting partners, but these were recovered in separate animal completely independent
of each other. And this is repeated many, many times. I don’t — I’m — and for
the transition zone complex there are too many proteins to list. And so the question is whether in fact genes
that are involved in congenital heart disease pathogenesis might actually be part of an
interactome network. And so to look at that question we took the congenital heart disease
gene and created the HPRD BioGRID database that has information protein-protein interactions.
A generated interactome network is shown here, and we see that the shortest distance between
nodes, which are the congenital heart disease genes in red, has 4.7 edges. If you do the
same kind of interactome assembly using random gene sets and you simulate that 10,000 times,
we get 14.9 edges, suggesting that this congenital heart disease gene network actually is of
functional significance. And so this led to this notion that maybe the interactome network
itself may provide the genomic context for congenital heart disease pathogenesis and
that this may explain the complex genetics of congenital heart disease which is really
more relevant to human disease then recessive mutations. And so can we provide some experimental evidence
to support that? So with that in mind, we looked at Ank6 and Nek8. In collaboration
with Jaga Shaw [spelled phonetically] we previously showed that Ank6 actually is required to activate
Nek8. Nek8 is a kinase. It’s actually the only kinase that’s been identified in the
cilia and so in the presence of Ank6, Nek8 kinase is activated. And so the question is
we have this mutation in Ank6 and a mutation in Nek8, can these two genes interact in an
epistatic fashion to cause disease? To look at that question, we intercross heterozygous
Ank6, heterozygous Nek8. Each one by themselves in heterozygosity has no disease phenotype.
With the intercrosses, you get four genotype outcome. Of course what’s of an interest
is the double heterozygote. You would expect that at 25 percent ratio. We can see there’s
a depression. We’re harvesting these in utero so they’re late gestation embryos
and we can see that actually we’re losing some of them much earlier. And actually out
of the 27 embryos we did recover, 17 of them show mutant phenotypes that are seen only
in the homozygote Ank6 or our homozygote Nek8 mutants. So it shows that digenic interaction
double heterozygosity can actually give you disease phenotype similar to what you would
see in the homozygote adult animals suggesting this concept of multigenic interactions in
an interactome network may really be relevant for thinking about human congenital heart
disease. Finally, I just want to show very quickly
that we also have been very interested in hypoplastic left heart syndrome which Cricket
also mentioned earlier. It was surprising. This is one of the rarest phenotype recovered
in our screen. We recovered eight mutant lines with HLHS. Surprisingly exome sequencing analysis
showed that these eight mutants have no genes in common suggesting this notion that they
have a multigenic etiology. We’ve been able to validate that for one particular line where
we can show that there are two genes required to generate HLHS. So we believe that this
is a disease phenotype that is intrinsically multigenic in etiology and we believe that
there is a couple of other congenital heart disease phenotype in that same category. So
to close I just want to leave you with this thought that using system genetics with mutagenesis
may provide a segue for really interrogating the complex genetics of human congenital heart
disease. So not only can we use forward genetics to recover and gain insight into recessive
mutations that are in a Mendelian genetic model of disease, but that these may lead
us to insights on the complex genetics of disease because I believe the same genes that
are in these Mendelian model can also contribute to more complex genetics of disease. And moreover
we think that using this system genetic approach may give us some insight on the genomic context
of disease pathogenesis so that looking at the totality of mutations recovered, we may
actually gain some insight on genetic resiliency. And one thing I didn’t even talk about is
whether there is evidence for protective versus pathogenic alleles which I believe our screen
may have some evidence for, and also this concept of penetrance which we touched upon
a number of times. And finally, I want to suggest that maybe
there’s real value possibly provided by generating a mutagenesis database. There are
many, many labs that are doing screens. The collection of that data may perhaps allow
another means to query sequenced variants for possible functional significance, and
again this is something that obviously will require some resource to collate. And so in
closing then regarding the question of animal modeling of human disease, I would like to
propose that in selecting animal models, you obviously need to make sure that there is
similar anatomy and physiology in your animal model to the human disease under study, and
so for congenital heart disease, you need a four chamber heart. The availability of
inbred strains is also important in the context of genetic analysis and mice is uniquely suited
given that we have completely inbred strains. Phenotype ontology is also very important.
It’s really critical that the phenotype ontology used to describe phenotypes in animal
models parallel the human phenotype ontology, and I think this is something that needs to
expand with different animal model system under use. And finally that there needs to be a way to
quickly disseminate the phenotype genotype data in public databases so clinicians and
other scientists can readily access that. For us, Jackson Lab has been wonderful in
providing MGI as a context in which we can do that. And then finally I think Haoyi Wang
may have mentioned the use of CRISPR-Cas. Ultimately with animal model, can we then
use CRISPR-Cas as a quick way to do validation of human sequence variants? And so with that,
I’d like to thank a lot of people that contributed to the work and support from heart, lung,
and blood. I did have a lot of clinical collaborations. I didn’t really have time to show the data,
but we provided a lot of mouse models for a number of papers that have been published
on different human diseases. So with that I thank you for your attention. Male Speaker:
Thank you. I have a quick question for you and then I think we’ll move to a discussion.
So your heterozygotes are fascinating. So you didn’t see complete penetrance with
your two — with the two heterozygotes. So have you looked at these animals in more detail
and let them live longer to see what happens? Cecilia Lo:
So the double heterozygote animals, if they are situs solitus, they’re perfectly fine.
If they’re situs inversus, they’re perfectly fine, but it’s the one with heterotaxy.
They always have complex congenital heart disease and those expire, but the ones with
situs inversus are perfectly fine just like the homozygote animals. So this is a line
where we see laterality phenotypes, and we actually have another line that does not have
left-right patterning defects so we also have seen this — digenic interactions. So I think
— I think that this is — this is, you know, something that I think will be important that
we can explore experimentally. We’re trying to figure out a way to do this more systematically
based on the protein interactome. So — but, yeah. I don’t know if that answered your
question. Male Speaker:
And have you crossed any of the other combinations? So you just looked at that one set of genes.
Have you looked at any of the other combinations to see what happens as heterozygotes? Cecilia Lo:
So we — yes. We have another combination that we’ve done that we also see interactions,
and the phenotype in that case is actually milder than in either of the parental homozygotes,
but we still see the congenital heart disease penetrance. Some of the extra cardiac phenotypes
seem to be milder or not present, but the cardiac phenotype, it still persists. Male Speaker:
Okay, any specific questions for Cecilia, or I will open it for — yeah, Cricket? Female Speaker:
Cecilia that was great. I was just curious about your genetic resiliency. Did those mice
not exhibit the phenotype or did they have a phenotype and survive? Cecilia Lo:
They have congenital heart disease that’s related to another gene that we were tracking,
but the severe embryonic lethality phenotype type known for the knockout is obviously not
observed because we’re getting these mice to term. So in other words — and one example
is like LRP1B. It’s one of the homozygote genes we found in homozygosity. The knockout
mouse actually dies around nine and half, 10 and half days’ gestation so our mutants
are being examined near term so they’re coming to term. So basically that knockout
— that homozygote knockout really has been rescued if you will by other mutations in
that line. And I’m not saying that it’s necessarily the pathogenic gene that we recovered,
but suffice it to say that there — the sort of the genomic context of that animal has
rescued that null phenotype. Female Speaker:
So have you sequenced the resilient strains and defined variants that are [unintelligible]
in that network or something like — Cecilia Lo:
Yeah so we have the exome data just from the one animal and so yes, there are some suggestion
of other genes in endocytic trafficking that actually may provide some recovery of function,
but obviously, you know, we don’t have experimental data. It’s just argument based on the exome
sequencing data. Male Speaker:
Calum? Calum MacRae:
Cecilia, I was going ask you if you’d looked at cellular phenotypes across clutches without
— or sorry, litters — across litters without — and correlated them with cardiac phenotypes? Cecilia Lo:
So — Calum MacRae:
Except — for example ancillary. Cecilia Lo:
Right, right. So yes. So we’ve looked at ciliation [spelled phonetically] efficiency,
percent ciliation and, you know, it — you’re asking is it consistent between animals? Yes,
they are consistent between animals and right now what we’re trying to do is actually
look at kinase activity. So in the double heterozygote, what happens to the kinase activity
and the interaction between Ank6 and Nek8? So we’re trying to do some biochemistry
to really address that question, and we also of course have animals with tissues that we
can analyze as well. So we’re looking at that and this — these two mutations cause
nephronophthisis and so we actually can also see that they have a kidney — this is the
kidney phenotype, the double hats, but I think they are milder although we haven’t looked
so carefully that we’ve quantitated. And I can say it’s truly milder, but they definitely
have nephronophthisis just like the homozygote, but it seems like they’re kind of — you
know, it takes longer for them to get out to that more severe state. Male Speaker:
Other questions? Comments from up here? Male Speaker:
I wonder if the terms reduced penetrance and genetic resiliency shouldn’t just be dropped
because they’re both euphemisms for oligogenic inheritance and actually imply things about
the situation which might not be true. I mean if you say reduced penetrance, your focus
is on one mutation and you’re saying well this mutation has the characteristic that
it doesn’t always cause disease. Well actually the situation might just be that there’s
one, two, three, four other genes that determine together whether disease comes up, and your
focus on this one gene is just a historical accident. That’s what I mean. It’s not
really a good term. Teri Manolio:
Sure, no, I understand, but what would be the alternative? Just that these are polygenic
traits that are complex? I mean that seems like a long way of saying reduced penetrance. Male Speaker:
Well I think it’s a more honest way of saying it. Male Speaker:
[inaudible] [laughter] Male Speaker:
Cricket? Female Speaker:
But it’s not necessarily genotype. It could be dietary. It could be other things so I
think we can’t just assume everything is genomic. Male Speaker:
Well especially with left-right. There’s a — Female Speaker:
Right. Male Speaker:
— clear stochastic component. Male Speaker:
Right. Female Speaker:
Right. Male Speaker:
So — Female Speaker:
Exposures matter. We know viruses cause disease. [end of transcript]

Be the first to comment

Leave a Reply

Your email address will not be published.