This webinar explores how the epigenome is altered in embryonic development in mammalian and non-mammalian species. From a non-mammalian …
The Biochemical Society and Portland Press are pleased to welcome you to this webinar,
part of the biochemistry focus webinar series.
Topics in this series include different research
areas in the Molecular Biosciences as well
as practical sessions to support career development.
Each webinar will give you the opportunity
to ask questions via text, and we welcome
suggestions for future topics and speakers
to feature in our webinar series.
So, please see the website for more details.
My name is Marnie Blewitt, and I'm a research scientist in Australia, you can probably tell from my accent that I'm Australian.
I'm located in Melbourne. In Melbourne, I work at the Walter and Eliza Hall Institute of Medical Research,
where I run a lab and a division, focused on epigenetics in development.
Today's webinar is titled ‘Reprogramming
the epigenome through embryogenesis’. As
you know, epigenetic marks are by definition
‘mitotically heritable’.
However, in contrast to this really strong
heritability, epigenetic marks need to be
cleared between generations in order to remove
the marks acquired during the parents lifetime.
This is a critical process to ensure totipotency
and viability of the offspring. That has been
studied most extensively for DNA methylation,
particularly for DNA methylation at CG dinucleotides in mammals.
In this case we know there are two key rounds of reprogramming: one during germ cell development
and another shortly after fertilization during pre-implantation development. Interestingly,
the embryonic round of reprogramming occurs alongside activation of the zygotic genome,
which has been largely silent during the final stages of germ cell development. So, this
is also a process of key interest.
Now, some of these processes, as I said, have
been largely studied in mammals. But there
are several areas that we're really still
very unsure about. And these include the role of methylation at sites that aren't in CG
dinucleotides, so non CG methylation, for
example, but also what happens in other species,
and we all know that we can learn a lot by
comparison between different species.
So, today, we will explore how the epigenome is altered in embryonic development in mammalian
and non-mammalian species.
From a non-mammalian perspective, we will
hear about inheritance and developmental reprogramming of non CpG methylation in zebrafish.
And we’ll then hear about early mammalian
development from fertilization to gastrulation
with a focus on zygotic genome activation
and the transfer of the control from the mother,
or the oocyte, to the embryo.
So, this is perhaps, I think, one of the most
fascinating and important fields of biology,
although of course I’m biased, as it really
pertains to how much of life is created.
And we know that there's still a huge amount
that we don't understand based on the complexity of these processes.
Our first invited speaker today is doctor
Ozren Bogdanovic, a lab head and Senior Research
Fellow at the Garvan Institute of Medical
Research and the University of New South Wales in Sydney.
He leads a developmental epigenomics lab that
investigates the contribution of epigenetic
mechanisms to embryogenesis, germline formation and disease.
His lab combines zebrafish approaches, CRISPR-Cas9
genome editing, and genomics to study the
developmental dynamics of epigenetic marks
such as DNA methylation and histone modifications in vivo.
Our second invited speaker today is doctor
Melanie Eckersley-Maslin. She is a BBSRC Discovery
Fellow at the Babraham Institute in the UK.
She investigates how the epigenome sets up
and initiates transcriptional programs in
early mammalian development using embryonic
stem cells and mouse models.
She will soon be moving also to Australia,
to the Peter Maccallum Cancer Center in Melbourne across the road from where I am, where she's
gonna start a new group leading investigating into epigenetic parallels between early development and cancer.
So, we're gonna start with Oz, and then he's going to hand over to Melanie.
So welcome and we all look forward to your talk.
Thank you. First of all, I'd like to thank
the Biochemical Society and Marnie for organizing
this fantastic webinar.
Today I'm going to be talking about a very
peculiar DNA methylation signature that we
recently discovered then in zebrafish.
So as Marnie mentioned, DNA methylation has been extensively studied during embryonic
development. However, this was mostly studied within the context of CpG dinucleotides, which
is where most methylation is taking place
in vertebrates.
However, a number of recent studies have demonstrated that non-CpG methylation, or mCH methylation-
so where H is essentially any nucleotide besides the G- can also take place, particularly in
the neurons and in pluripotent stem cells,
as well as in oocytes.
So, while classical CpG methylation is pretty much anti-correlated with transcription, with
non-CpG or mCH methylation this is not so
clear. So some fantastic work from Ryan Lister
has described a few years ago that DNA methylation is actually anti-correlated with transcription
in vertebrate, well in mammalian brains, and it is positively correlated in pluripotent stem cells.
So, what happens during mammalian embryogenesis,
with this non-canonical form of methylation?
As I mentioned, it is present in the oocytes
it's not present in sperm, but upon fertilization it gets diluted. And it disappears only to
re-emerge again during fetal development as you can see by this purple line. And this
happens in both mice and humans in the neural system.
And recently, Ryan’s lab has shown, and
we participated in this study, it's currently
under review, that this form of non-CpG methylation in the brain is actually highly conserved
among vertebrates, particularly in this CAC
context, which is typical of neurons.
So, knowing that in mammals both mCG and MCH is globally depleted, and this does not happen
in anamniotes such as fish and frogs, which is something that we've shown before, we wanted
to explore whether mCH could be present during zebrafish embryogenesis, and if so, what would
be the functional implications of this?
So first of all, I'd like to mention that
this project was and is being led by Sam Ross, an extremely talented student from my lab,
who did most of the work here.
So, the first thing that we did, we decided
to generate genotype corrected methylome maps, so whole-genome bisulfite sequencing maps
for four stages of zebrafish embryogenesis, covering blastula, pharyngula, tailbud, and the adult brain.
And as you can see on the left side,
this is essentially the lambda control, which is what we add to bisulfite sequencing experiments.
It is a negative control, so wherever you
see anything that's zero point two, zero point four,
this is likely just noise in the lambda,
so anything within that range should be taken
with a grain of salt.
And you can see essentially that there is
some enrichment, in CH and in particularly
in CA dinucleotides that go over this threshold
and particularly in the brain as shown before.
However, when we looked at the top 10000 most
methylated motifs in this non-CpG context,
what we discovered was this very peculiar
motif. So CATGCTAA motif, which was present in almost identical shape and format at all four stages.
And when we looked where these sequences were located,
where these motifs were located,
they were predominantly located in the MOSAT
DR repeats, which is a zebrafish or perhaps,
fish-specific repeat family mosaic satellite repeats.
And you can see an example of such a repeat
here with its corresponding DNA methylation pattern, as well as some single nucleotide
profiles where black circles denote methylated TGCT sequences and white circles denote unmethylated
TGCT sequences.
So, essentially embryonic mCH is enriched
at MOSAT repeats at this particular sequence, and one of the important thing is that we
see it almost exclusively on the TGCT strand, and not on the opposite AGCA strand, which
would be a possibility similarly to CG methylation, which happens on both strands.
So next what we wanted to do was, knowing that bisulfite sequencing is prone to all
kinds of sequence biases, we wanted to validate these results by an orthogonal approach, and
we applied enzymatic methylation sequencing, which is a technique that was developed by
New England Biolabs.
And it essentially entails TET oxidation,
and APOBEC conversion to unravel methylated and hydroxymethylated sites.
And what we saw is that the profile was pretty much identical with EM-seq and whole genome
bisulfite sequencing.
So, the MOSAT sites with this particular sequence
were highly methylated when located within MOSAT repeats, however, such sequences, within
the genomic context, outside of repeats were fully unmethylated.
So, we conclude that our mCH methylation that we observe by whole genome bisulfites sequencing
is not due to some conversion bias, but it
is bona fide a DNA methylation signal.
Next, what we wanted to look at is where these sequences are located in the genome.
Here you can see the breakdown of what we saw, so, most of them are found in either
intragenic regions or in introns. And you
can see two examples, one for each.
So, this would be the MOSAT sites in the intron and the MOSAT sites in the intergenic region
between these two genes. The blue signal is mCH and this greyer signal is mCG.
One other important thing to mention is that these MOSAT repeats are depleted of CpG dinucleotides.
So, it's likely that mCH within these sites
is not just the byproduct of DNMT activity
or methyltransferase activity that’s targeting CpG dimucleotides.
So, we can detect MOSAT by both whole genome bisulphite and EM-seq, we find these regions
and intergenic contexts as well as in introns, and we see that they are depleted of CpG dinucleotides.
And another very interesting observation,
that you can see here is that MOSAT mCH appears
to be increasing throughout development unlike CG, these regions. And that's something that
we wanted to study in more detail.
So, to do that, we decided to expand our range
of stages, and we re-analyzed a number of
publicly available datasets to span the whole
embryonic development of zebrafish, including gametes and also other adult tissues.
What Sam saw, amazingly, was the mCH, the MOSAT mCH, is reprogrammed or remodeled during
embryogenesis and you can see clearly here that it's present in sperm.
It's also present in the egg and in the zygote, it gets diluted during embryogenesis during
these Cleveland stages.
It reaches its lowest levels at zygotic genome
activation and then, coinciding with gastrulation it increases. And it is present in brain,
liver, and we also have some data showing
that it's present in other tissues, so likely
all tissues originating from all embryonic
So next, we wanted to look at histone modifications, and other epigenetic marks that might associate
with this motif, and we again utilized published datasets.
And what we saw was that the only mark that was enriched at most mCH sites was K9 trimethylation.
And none of the other histone modifications, the active ones, or K27 tri were enriched at MOSAT sites.
So, to explore this, in more detail, we remapped
these H3K9 trimethylation data that were generated by Mary Goll’s lab last year, published
in a beautiful study that explored the dynamics of K9 tri.
And essentially what we see is, as described in the paper, that this trimethylation signal
increases during embryogenesis.
However, on a small number of MOSAT sites,
it appears to be stable.
And you can see here clearly that MOSAT is
being reprogrammed, nevertheless H3K9 tri remains stable at these regions.
And this is a very beautiful example, so you
can see very nice correlation in one of these intergenic MOSAT sites.
You can see MOSAT slightly increasing and you can see the overlap is kind of almost
perfect with H3K9 tri.
And the correlation between MOSAT mCH, at least in stages where both marks are present-
so here and here at 1.75 hours and eight hours, so before and after the ZGA, the correlation,
the positive correlation between these two
marks is pretty strong.
So, we conclude that MOSAT coincides with K9 tri enrichments. Obviously K9 tri persists in the absence of
MOSAT mCH, so it's unlikely that MOSAT mCH would play a role in the recruitment of heterochromatin.
However, we cannot exclude the possibility
that H3K9 tri helps in recruiting MOSAT back
after the ZGA. So that is something that we
are currently looking into as well.
So, finally, we wanted to explore a bit further which DNA methyltransferase could be responsible
for this MOSAT mCH signal.
And as you can see in zebrafish, the situation
is pretty complicated when it comes to your DNMTs and in particular, de novo DNMTs.
So apart from the maintenance methyltransferase DNMT1, and the tRNA methyltransferase DNMT2,
zebrafish display six de novo methyltransferases.
So, these are I assume the old names, and
more recently, these even more confusing names have been adopted.
But what's really important here, what I'd
like to highlight is that DNMT3BB 2, as well
as DNMT3BA, have this weird CH, so calponin homology domain that is usually associated
with actin binding that we have no idea what it does.
And it's only specific to teleos-DNMTs.
So, to get a better idea of what to knock
out, we first explored the developmental expression profiles of different DNMTs and came up with
a couple of good candidates.
So, as you can see here on the left-hand panel,
DNMT3BB 2 coincides with, it peaks at the
ZGA and DNMT3BA basically tracks perfectly
how mCH is coming back after the ZGA.
But nevertheless, the first thing that we
wanted to do was to exclude the possibility
of DNMT1 or maintenance methyltransferase
activity playing the role in mCH as we knew
that in mammals, it's exclusively guided by DNMT3s.
So, we designed CRISPR knockouts.
So, these would be transient CRISPR knockouts or CRISP-ins, and we measured DNA methylation by a modified
reduced representation bisulfited sequencing approach that allows us to enrich for MOSAT sites.
Essentially what we saw- this would be the
wild type, and this is CpG, and this is MOSAT and that's CH methylation- that DNMT1 reduces the CG levels a bit.
However, we knew that the DNMT1 has this massive
maternal contribution, as you can see on this slide.
So, we also knocked out UHRF1, which is the cofactor that actually targets the DNMT1 and
here we achieved much better knockouts. But as you can see, this is exclusive to mCG methylation.
Nevertheless, for both the DNMT1, UHRF1 or DNMT2 the levels are pretty much stable on MOSAT sites.
So, next we knocked out the de novo DNMTs.
And what you can see here is that in the knockout where we combined guides targeting 4 DNMT3Bs,
as well as the DNMT3BA, here you can see a very significant reduction in MOSAT mCH levels.
And all of this has been done a number of
times in biological replicates.
So, we are pretty confident at the moment
that the DNMT3BA, so one of these weird methyltransferases
with a CH domain would be the main culprit for MOSAT mCH methylation.
Even though we're not completely excluding the possibility that the other ones might
somewhat participate in it.
So, we conclude the DNMT3 is the primary MOSAT
mCH methylase, we're currently exploring whether this weird CH domain would have a possible
functional in it, and of course there'll be
a lot of redundancy, with such a number of
de novo DNMTs. So we're currently very carefully designing these experiments.
So, since this is a reprogramming webinar,
I would like to end with the reprogramming figure.
And what you can see here is during zebrafish
embryogenesis- mCG levels, they're pretty
much stable somewhere around 24 hours when
thousands of enhancers are being demethylated, actively demethylated, you can see a slight
dip in methylation. And that increases as
the development proceeds.
But with MOSAT mCH you can see that it's reprogrammed very similarly as mammalian mCG and mCH are reprogrammed.
So, you have them in the gametes, it's practically erased,
and then it comes back again in a
developmental fashion.
So, in sum, we uncovered a novel mCH signature in this very peculiar motif, which hasn't
been described before. Unlike mCAC and mCAG described in mammals, we find that MOSAT mCH
is found throughout embryogenesis also in
a number of adult tissues.
So, it's not restricted to pluripotency or
the nervous system. It's reprogrammed during
embryogenesis coinciding with zygotic genome activation and its associated with heterochromatin.
And of course, the million-dollar question
is, how does MOSAT mCH contribute to embryogenesis?
How does it contribute to ZGA? What does it actually do? And that's something that we're
currently actively investigating.
And yes, of course it is deposited by DNMT3BA,
the peculiar DNA methyltransferase.
So, to conclude, I'd like to thank my lab,
and in particular, Sam, who did a fantastic
job driving this project forward, and Allegra
and Michael who also participated in this
Alex de Mendoza from Queen Mary University.
He helped a lot with genotype correction,
and motif calling, and some DNMT evolutionary comparisons.
And I'd like to thank our collaborators from
the Radboud University, Michiel Vermeulen
and Velin Sequeira, and Moreno from the Andalusian Center of Developmental Biology, they are
currently participating in ongoing work.
And Ryan Lister, for always being able to
discuss anything related to DNA methylation and non-canonical DNA methylation.
And finally, our funders, and thank you all
very much for listening.
And with that, I'll leave you to Melanie and her exciting research.
Thank you Oz for a wonderful talk. So just my presentation…
So, hello everyone, I’m Melanie, and I’d
like to start by thanking the Biochemical Society,
as well as Marnie for the invitation
to share my work.
So, I'm going to switch gears and talk about the mammalian system, in particular mouse
early development and a concept that we're
calling epigenetic priming, how this might
drive cell fate decisions that occur in the
So just a quick overview of the key steps
that take place in mammalian development
in the mouse, following fertilisation of the
oocyte by the sperm.
I can’t get my mouse to show up, I'm sorry.
You undergo a series of cleavage divisions,
which result in a blastocyst structure at
E3.5, which is the structure that implants
into the uterine wall.
So by this stage, you've had zygotic genome
activation that's taken place and you're getting the first cell fate decision has taken place
such that it’s just the cells of the inner
cell mass, the ICM, which will give rise to
the embryo itself.
So, after the blastocyst structure implants
you undergo gastrulation, so this is when
the three germ layers really take form, and
this is when cell fates really are starting
to be established.
So, during this time window you're getting
a decrease in development potency, but also
a lot of transcriptional and epigenetic changes.
So, at the transcriptional level, the oocyte
comes preloaded with maternal RNA stores, which you can see in the green line.
And these are rapidly degraded, as the zygotic genome is awakened, during zygotic genome activation.
Coinciding with this, there's changes in the DNA methylation landscapes,
so in mammalians DNA methylation occurs mostly in the CG context,
and most cells about 80% of their CGs are
However, in the pre-implantation embryo, you get this massive wave of global loss in DNA methylation.
And DNA methylation really only comes back up up again during gastrulation, as cell fates are being set up.
The same time, another epigenetic structure that I’ll talk to and introduce in a few
slides- bivalency- this is also lost, is absent
in the pre-implantation embryo and really
becomes established at the early stages of
So the question we have is how the genome really facilitates these early embryonic cell
fate transitions and can it in a way pre-empt these cell fate decisions set the cell up
for enabling it to be able to make these transitions?
So, this is a concept that we're calling epigenetic priming, which I'll just illustrate here with this cartoon.
So, the idea is that a priming factor shown
by the blue balls may be sitting at pluripotent
genes- lineage specific genes in pluripotent
cells, shown in the upper left quadrant.
And so by sitting at the promoters of these
genes, it can be preventing repressive DNA
methylation from coming in, keeping their
chromatin structure open by repelling repressive
epigenetic modifiers, shown in red, or recruiting active epigenetic modifiers, shown in green.
And in this way, it's keeping the chromatin
state, epigenetic state of these genes primed
and malleable. So that upon differentiation,
they can then be activated even in the absence
of the original epigenetic priming factor.
If you lack the priming factor, which you
can see in the bottom panels, then you can
have repressive DNA methylation coming in.
The repressing epigenetic modifiers may be able to be recruited to these genes and lock
the structure, such that upon differentiation, these genes are now no longer able to be effectively activated.
So there is a good example that we've known
about for several years now, of epigenetic
priming, and this is the context of bivalent chromatin.
So, this is the co-occurrence of both active
and inactive histone modifications on the
same nucleosomes, it occurs primarily in undifferentiated
cells, at key developmental gene promoters.
So, you have co-occurrence of the active H3K4
trimethylation mark, which is deposited by
the compass complex and the inactive H3K27
trimethylation, which is deposited by polycomb.
The idea is that upon differentiation these
poised structures are resolved into either
repressed H3K27 containing or an active H3K4
trimethylation containing structure as cell
fate decisions are made.
However we know little about the bivalent
chromatin is actually specifically targeted
to these key developmental gene promoters.
The proteins I want to talk to you today about,
that we've discovered as key epigenetic priming factors, are called developmental pluripotency
associated 2 and 4, or DPPA2 and 4. these
are quite small heterodymerising proteins,
they’re about 300 amino acids in length,
and they contain both a SAP domain that binds
DNA, as well as C terminal domain binds histone H3.
They have no known enzymatic activity that we have found.
What's been quite perplexing is knockout mice for these proteins survive embryogenesis but
die shortly after birth.
And this has been quite the conundrum in the
field because these proteins are exclusively
expressed in early embryos.
So, if you can see in the bottom right graph,
their expression is really limited to pre-implantation development.
By E6.5 when gastrulation starts, their expression is absent.
And they're no longer expressed throughout the
whole rest of embryogenesis into adulthood except for in the germline.
So, this actually makes them ideal priming
factors in that they are expressed at a time,
but their phenotype is seen at a later developmental time point.
We came across these proteins in a screen
we did looking for regulators of zygotic gene activation.
And in this paper that I don't have time to
share today, published last year, we found
that DPPA 2 and 4 act together by binding
a promoter of a transcription factor called
DUX, promoting its expression. DUX is a key
transcription factor for zygotic genome activation genes.
During this study, we also noticed that DPPA2 and 4 also bind non-ZGA genes and it’s this
that we decided to follow up on further. So,
this is them binding non-ZGA genes.
In particular, because when we made double knockout embryonic stem cells for DPPA 2/4
and subjected them to pluripotent- to differentiation essays, we saw quite a striking phenotype.
So, if you differentiate wild type embryonic
stem cells over nine days of embryonic, of
embryoid body culture, usually we see downregulation of pluripotency genes by about day 4 of differentiation.
And coincides with the upregulation of mesoderm, endoderm and ectoderm genes at about day 4,
as you can see on the left panels of the heat map.
However, if you take the double knockout embryonic stem cells and put them through the same differentiation
assay, you can see that the pluripotency genes take until day seven to be downregulated,
whereas the mesoderm, endoderm and ectoderm genes are now also being upregulated at a later time at day 7.
You can also see this in a principal components analysis,
where we take the whole transcriptome
and cluster the cells based on how similar
they are in transcription.
When you do this, you see that the double
knockout cells at day seven and day nine,
which is shown by the pink and purple diamonds,
more closely resemble the wild type cells
at day four which is shown by the red circles
than the wild type cells day seven and day
nine which are at the end of the arrow.
And so this really points to these cells,
although they have no problems in their pluripotent
state are having problems differentiating.
So, when we looked at where these cells are
binding throughout the genome, we noticed
that they are highly enriched at the promoters
of active and bivalent genes.
So, on the left I’m showing you an example
where you can see some active gene promoters highlighted by the green bar and some bivalent
gene promoters highlighted by the yellow bar.
The tracks you can see are DPPA2 and DPPA4
binding in the blue and the purple and the
K3 and the K27 trimethylation in the green and the red.
And you can see nice enrichment at these promoters.
When you actually do genome wide analysis and look for the enrichment, you can see that
these genes are really enriched at active
and bivalent promoters.
So, this prompted us to look at the epigenetic state of the embryonic stem cells when you remove DPPA2/4.
So, on the top, I'm showing H3K4 trimethylation,
and the bottom H3K27 trimethylation, this
is whole genome wide ChIP-seq.
I'm showing here the peaks. And you can see that in the H3K4 trimethylation in the knockout
cells, you're losing a large number of peaks
which are overlapping with DPPA4 peaks in
purple, as well as promoters shown in orange.
You are also losing H3K27 peaks, although
to a lesser extent.
And indeed, these peaks are overlapping each
other and are seen co-occurring at a lot of
bivalent genes.
So, if we zoom in particularly at the bivalent
gene promoters, we can categorize them into
four different categories- three different
categories, sorry.
We have those in orange, shown in the top
row, which have lost both H3K4 trimethylation,
H3K27 trimethylation as well as the epigenetic machinery responsible for depositing these
modifications, and we’re terming these genes DPPA dependent bivalent gene promoters.
We also have a second category of promoters which are sensitive to loss of DPPA2/4,
losing only H3K4 trimethylation.
But the H3K27 trimethylation remains unchanged.
And lastly, we have a group of promoters which are independentof DPPA2/4, so these are
remaining largely unchanged.
What you might appreciate from this graph,
and we’ve done some machine learning analysis as well, to support this, is that the main
feature that distinguishes these three categories of bivalent genes is the original amount of
H3K4 trimethylation at the genes.
So DPPA2 dependent bivalent genes have fairly
low levels of H3K4 to start with, whereas
the sensitive and the independent ones have
higher levels of H3K4 trimethylation.
When we look specifically at these different
bivalent gene categories of their behavior
during differentiation, we can see that all
three categories in wild type cells usually
become upregulated upon differentiation.
However, the DPPA2 dependent gene promoters shown in the orange, in the knockout cells
they are now no longer able to be activated
upon differentiation.
So, we next ask, well, what is keeping them
We turn to DNA methylation, as this is a well-known mechanism for repressing gene expression in mammalian cells.
We did whole genome bisulfite sequencing,
and you can see in the left plot that most
of the genome remains largely unaltered.
However, we have several hundred regions of the genome which are gaining DNA methylation,
and these are overlapping gene bodies shown in pink and promoters shown in orange.
And you can see the quantification of these
in the pie chart. And indeed, when we look
specifically at the dependent gene promoters shown in orange in the right plot you can
see that these are the regions that are gaining DNA methylation in the knockout cells.
And we think this is a subsequent event to
the loss of bivalency is what is keeping them
silent during differentiation.
Lastly, we wanted to develop a system to look
into the dynamics in the recovery, whether
or not we can rescue this phenotype.
So to do this we turned to an inducible shRNA system, whereby addition of dox enables the
expression of shRNAs targeting DPPA2 or DPPA4, and you can see in the Western blot by seven
days of treating with dox you have lost DPPA2 and DPPA4.
What is quite nice about this system is you
can then wash dox out and allow the cells
to recover- DPPA2 and 4 levels return.
And you can ask what, now, what is the epigenetic
state in the recovery cells two weeks after
So when we do this, you can see on the left
hand panels are PCAs, so we start with the
untreated cells as shown in the gray circles
and squares, addition of dox, which leads
to down regulation of DPPA2 and 4 in the pink cells, you can see these are very different
in the epigenetic profiles for both K4 and
However now when we allow DPPA2 and 4 to return to the cells, in the recovery cells in blue
you can see that they now closely resemble
the original untreated cells, so we are restoring
this altered epigenetic state.
If we look at the regions of the genome that
are changing in H3K4 in the top and H3K27
in the bottom.
You can see that between untreated and dox treated or DPPA depleted cells, you can see
that we're losing a large number of H3K4 trimethylation peaks, H3K27 doesn't change as much.
And indeed, if you look specifically at the
DPPA dependent genes, in the right panel that
are highlighted by orange, you can see that
these are the promoters which are losing H3K4
trimethylation in the dox treated or DPPA
depleted cells, and they are now being restored
in the recovery cells.
This suggests that DPPAs are sufficient for
actually targeting this bivalent structure
to these genes.
And, lastly, we also looked to see well, what's happening at the DNA methylation at these
sites in this inducible recovery system.
And we did this using amplicon bisulfite sequencing,
and if you look here this is a list of DPPA
dependent genes, the top two rows are control regions.
You can see in the untreated dox cells they
have relatively low levels of DNA methylation.
This increases in the cells treated with dox
which have depleted DPPA2 and 4 and is now lost again in the recovery cells.
This again points at DPPA2 and 4 targeting
bivalent structure and really, the loss of
them is DNA methylation coming in subsequently.
So just to summarize this study with an example gene.
This is a DPPA dependent gene, you can see
I've highlighted the promoter region with
this pale orange box.
In the top tracks you see binding of DPPA2
and 4 at the promoter of this gene.
And in wild type cells, there's a nice peak
of H3K4 and H3K27 trimethylation. And this
is lost in the knockout cells, completely
Coincidentally, we also have a loss in accessibility, and we have a gain in DNA methylation shown
by the long lines from wild type cells shown
in grey to knockout cells shown in green.
And subsequently upon differentiation, you
can see these are transcription tracks from
days 4 of differentiation in wild type cells,
in the top row you can see that there is transcription,
and this is lost in the knockout cells from
the bottom row.
This led us to our model, whereby there are
several classes of bivalent genes- the DPPA
independent genes and DPPA dependent genes.
The DPPA genes shown in the bottom row, DPPA2/4
are present, and really required to recruit
polycomb and trithorax and maintain this active
and inactive poised bivalent structure.
In the absence of DPPA2/4 you fail to recruit
polycomb and trithorax, you lose this bivalent structure, and subsequently DNA methylation
comes in and represses the genes such that upon differentiation these genes are now no
longer able to be expressed.
And so, we think that DPPA2 and 4 act in this
way as epigenetic factors and are really important for multi-lineage commitment.
And they do this by targeting bivalency to
a subset of bivalent promoters preventing
repressive DNA methylation, so that these
genes are activated efficiently during differentiation.
And we think that this temporal uncoupling
of when the proteins are expressed and the
phenotype of the genes that they regulate
may explain the preweaning lethality of the
knockout mice.
And we think that there may be actually a
more general role for DPPA2 and 4, in promoting epigenetic plasticity in general and facilitating
cell fate transitions.
So, with that, I'd like to thank the people
in the in the lab who have helped with this
project in particular Aled, Christel, Elisa
and Marloes, as well as collaborators Clive
D’Santos and Masashi Narita at CRUK.
And just to let you all know that I’ll soon
be starting my lab at Peter Mac in Australia,
looking at epigenetic priming in and development and cancer so if you're interested there'll
be future opportunities here.
Thank you for your attention and I guess we'll
pass back to Marnie for the question-answer session.
Thank you so much Melanie and also thank you Oz for your fantastic talks.
So, everybody please remember to send in your questions. I'll start asking the ones that I can see already.
So, the first one for Oz says: does the ablation of DNMT3BA and thus mCH result in an increase in mCG at the same regions?
And this comes from Eytan Zlotorynski.
Yeah, we haven't seen that. So, it appears
to be specific to mCH not to mCG.
Excellent. OK. So, for Melanie, Frank Schubert says:
very clear data- does DPPA2/4 dependence identify genes that are critical for cell
fate decisions or other genes involved in
subsequent differentiation processes.
It's a great question. So, they tend to be
genes that are involved in actually the cell
fate decisions themselves, along with a lot
of bivalent genes.
When we do gene ontology, we don’t really
pick much up.
Although if we look at the phenotype of the
knockout mice that have been generated for
these dependent genes, a lot of them have
very clear developmental phenotypes but at
later stages- so, later stages of embryonic
development or preweaning or just after birth.
There's another question from Melanie, from Samuel Ross. Just a clarification, I think.
Are DPPA2/4, are they similar to Stella or
is it just a generic name?
It's a good question, actually. So, there
are five DPPA's and they were named in a screen
looking for proteins which have similar expression pattern to Oct4.
So, they’re all based on their expression
patterns, not necessarily in their structure.
Coincidentally DPPA2/4 are very similar DPPA3, which is also known as Stella, is fairly similar.
But, as far as we know, it doesn't act in
this sort of manner.
OK, I'm just trying to work out who this one’s for, it says many thanks for these great talks
it's from Susanne Dietrich: How come if differentiation is globally inhibited the phenotypes appear
so late? So, cardiomyocytes you need very
early, for example.
So, this is obviously for Melanie. Sorry,
I worked it out after reading.
Yeah so, that's something that we're still
trying to understand, actually.
So, one idea is that, if you think about cell
fate transitions and activation barriers.
So, if you might still be able to get through,
eventually. So, if you saw in the heat maps,
eventually, the genes get activated upon differentiation, but they're just delayed.
So, it's just not as efficient at doing it.
So, you might be able to start making the
tissues just not as efficiently and they're
not working as well.
So, that's why you might be able to get through all of development to birth, lay out the rough
body plans but not perfect body plans.
And it's really at birth when the mice need
to start breathing, and eating, and so on,
and that's when you get the lethality.
So, actually, they died just after birth from
lung defects, which obviously makes sense,
because that's the first time they need to
use their lungs.
So, we’re following this up at the moment,
we’ve made knockout mouse models, and we're
going in and profiling them at earlier stages
to see how their epigenetic state looks and
later stages to see how they look phenotypically.
But that's the, that's the idea that we have
at the moment,
Maybe a related question to that from me,
so there are many different epigenetic regulators.
When you knock them out on mutate them, you
get quite a wide variance.
And maybe you haven't changed overall the
average, but you have this huge variance.
And it sort of looks like, well, something
goes wrong and it just depends on what goes wrong first.
Do you see that in these mice or is it they’re
all dying of a lung defect or do some die
in late gestation and you know they’re kind
of dropping off.
Um, we don't really see much of a dropout
We haven't got enough numbers now to really have proper statistics, but most of them seem
to be making it through. The litter sizes
seem fairly similar.
Occasionally, you get some nice that survive, and while they’re lighter, they’re smaller
than their littermates they catch up eventually.
So, this does, there is heterogeneity in the
letters themselves, and amongst the different litters.
So, it's not an all or nothing, black and
white, sort of thing.
So does point us to inefficiency of these
proteins, really, improving the efficiency
of making these decisions, rather than being the key gateholder if that makes sense.
I'm gonna ask one to Oz now, if that's OK.
So, I was wondering, Oz, do you know, or can
you postulate about any of the particular
binding factors for the mCH methylation in MOSAT?
So, I guess in mammals, MeCP2, which classically
I have thought of as a CG binding protein
has now been shown to bind CA.
So, do you think that could be factors like
That we might have previously thought had
maybe a more important role at another site,
might be binding these?
Yeah, so we're currently looking into that,
but it appears, and it's also based on the
Some people probably picked it up and it very strongly resembles the Oct4 motif.
And actually it appears also to coincide with sites of Oct4 binding during embryogenesis.
And we have some preliminary data that MOSAT mCH when methylated has high affinity for Oct4.
So, how it all plays into the context of embryonic
development, it's something that we're currently looking into.
We think that what might be happening is that some of these sites might actually serve as
sinks or reservoirs for transcription factors,
which then get redistributed upon the ZGA.
And recently, Julia Hirshfield showed something similar with cohesin as well. So, it is both.
It participates in a similar mechanism.
Yes, it's a more efficient way of doing things early in development.
OK, there are a few more talks, a few more questions.
So, next one is from Grant Mitchell says:
two nice talks to translate to humans is the
DPPA phenotype detectable in fetuses and would
it appear in humans by prenatal ultrasound.
So intrauterine growth restriction for example?
Yes, that's a great question. I have looked into it. And there's actually no cases of
humans that have mutations in DPPA2/4.
Well, we don't know all the mutations but
loss of the DPP2 and 4 locus.
So, whether that's just because they haven't,
they don't have a phenotype or whether that's because these humans don't exist.
Of course, we don't know. We do have some ongoing work in the lab, looking at human
embryonic stem cells, just to get at some
of these types of questions because it is
quite interesting whether or not the same
thing is also happening in the human system.
What about, as a separate way to think about relating it to humans would be to say, loss of bivalency and again,
in DNA methylation sounds distinctly like what happens in tumors at CG islands anyway.
So, do you see a change in them in the context of cancer?
So, they're not normally expressed then.
Do you see them being aberrantly expressed, or something weird going on?
Yes, exactly and that's something I plan
to follow up on in my new lab is that there’s
quite a lot of cancers that upregulate DPPA2/4 so these genes are regulated themselves by DNA methylation.
And so, we think their upregulation
it's not through copy number amplifications, but it’s actually through changes in the methylation
at the promoters that lead to the upregulation of these genes in a lot of different cancers.
We don't know the significance of this yet,
but my hypothesis is that their presence and
upregulation promotes this epigenetic plasticity in the cancer cells, which you can see in
the altered epigenetic profile of cancer cells.
And this might actually be contributing to
tumorigenesis or metastasis when the cells actually have to change what they're doing.
And become more malignant for example, but it's something that- stay tuned, hopefully
we'll have an answer sometime soon.
Yeah, there are two questions from Angel Liang
so they kind of follow on, I'll ask them both
at the same time.
The first says how do DPPA2 and 4 know where to be recruited to, and do you think that
they could be used to enhance pluripotency
in some way, efficiently?
Yeah. So, they bind CG rich regions of the
genome so they like promoters, CG rich promoters,
they’re actually found that two thirds of
So, we don't know exactly why only some of the bivalent promoters are affected.
Whereas other bivalent promoters that have DPPA2/4 binding are unaltered.
I think it's to do with levels of K4 and transcription of these genes, which are able to maintain
them in a state that, you lose the initiator
and they’re self-sustaining. We still need more data on that.
Could it beacetylation? The amount of acetylation,
they have there rather than K4 tri or K27
Nothing that really stood out. I think we've
gone through comprehensively.
We’ve done motif analysis, but you just
get GC rich sequences because they like GC rich regions.
And for enhancing pluripotency
there's a really nice study that was done
in the US looking at DPPA2 and 4 in IPS cell
And actually, by over expressing DPPA2/4 you really greatly enhance IPS reprogramming,
up to 80% of cells can make it.
So, there is certainly something about cell
fate transitions in general, not just the
ones at bivalent genes and gastrulation but also ZGA, reprogramming and so on.
I have one more for Oz: can you in your expression analysis,
can you do expression analysis at
the same time as your CH methylation analysis?
Perhaps not during the early stage where K9 methylation seems to take over or be retained
while the CH is being reprogrammed, but later can you correlate the expression, the amount
of CH methylation and the context?
So, when they are in the gene, and whether
that gene is being expressed, or whether they're in the surrounding genes and intergenic region?
So, we looked into that. Yeah, unfortunately, because of the restrictions on the time I
didn't put it in. But we don't think that
it correlates at all.
Actually, interestingly and similar to mammals, most of these, most mCH sites, they’re found
in neuronal genes.
So you've got, even though the developmental
increase has nothing to do with these genes because they get switched later on, there
must be something that attracts MOSAT mCH to the whole neural story, I don't know if
it is because neural genes are longer.
So, they have a higher probability of having
such a massive repeat cluster.
There's something else that's something related
to the 3D structure, but there's definitely
no correlation in terms of MOSAT mCH and gene
expression, or at least at that very gene
where it sits in.
The other thing that I thought was interesting was that you had the CH, the CH is where you are CG poor, and at CT.
So, you know, is there any way, can you go
back in evolutionary time and ask when that repeat landed in those places? And what it
was like beforehand, to try and see if there's any relationship?
Yes. So, we looked a bit into it, it appears
that it's specific to zebrafish and carps,
and there's some data, published data that
found similar repeats in the Asian sea bass.
But, yes, that's something that we would like to look into it.
But it is, it has to be done in a comprehensive manner. And it's not so trivial actually.
I'm sure it's very hard.
But it would be interesting to look at.
So, there's one final question, and this is
for Oz and then we’ll wrap it up.
So, Susanne Dietrich says: is this MOSAT enrichment because of primary neurogenesis in anmania.
You might understand that better than I do.
Yeah, I'm not sure, really. But I don't think
that it correlates really with neurogenesis
because this enrichment we're seeing happens even,
starts happening even before that.
So, we think it has something to do with general
regulation at the ZGA and the fact that these sites reside in neuronal genes.
It's likely because of their length, or because of some other, I guess, structural concept
that it would fit nicely in that environment.
OK, so it just remains to thank both of our speakers.
I'm so very grateful that you're
able to join us tonight.
And everybody who came along and asked questions and listened.
So, we're going to be running an Embryogenesis collection in the Biochemical Society Transactions
and that can be visited at the journal website for more information.
If any of you that's watching out there would like to suggest key topics that we should
commission around, so commissioning short reviews, or if you'd like to submit a mini
review to the collection for consideration,
please contact editorial at
And we're going to also welcome suggestions for future topics and speakers to feature
in this Biochemistry Focus webinar series.
So, if you've got a good idea for a webinar,
then please do submit a proposal for an upcoming webinar.
Next one is going to be held next Friday,
the 11th of September at 3PM, UK time, I think, or 15:00,
and it's on 'How to deliver biomolecular science honours research projects with limited lab bench time',
which is relevant to many
of us that are supervisors.
And this is part two in that webinar series.
Where Dr Aysha Divan an associate professor
in the School of Molecular and Cellular Biology, and Dr Vas Ponnambalam a reader in the Human
Disease Biology at University of Leeds will
talk about their experiences, running non-traditional
research projects under the constraints of
limited access to the lab.
So you can register for that one and find
out more information about the upcoming webinars
and of course, propose your own webinar at
Thank you very much.

Leave A Comment

Your email address will not be published